The 10 Rules of Castings - ایران مواد

به نام خدا

Castings Practice

The 10 Rules of Castings

www.iran-mavad.com ايران مواد - مرجع مهندسى مواد و متالورژى

Dedication

To Merton C. Flemings of MIT for inspirational teaching and research


Castings PracticeThe 10 Rules of Castings

John Campbell


Elsevier Butterworth-Heinemann

Linacre House, Jordan Hill, Oxford OX2 8DP

200 Wheeler Road, Burlington, MA 01803

First published 2004

Copyright Ó 2004, John Campbell. All rights reserved

The right of John Campbell to be identified as the author of this work

has been asserted in accordance with the Copyright, Designs and

Patents Act 1988

No part of this publication may be reproduced in any material

form (including photocopying or storing in any medium by

electronic means and whether or not transiently or incidentally

to some other use of this publication) without the written

permission of the copyright holder except in accordance with the

provisions of the Copyright, Designs and Patents Act 1988 or under

the terms of a licence issued by the Copyright Licensing Agency Ltd,

90TottenhamCourtRoad,London,EnglandW1T4LP.Applications

for the copyright holder's written permission to reproduce any part

of this publication should be addressed to the publisher

Permissions may be sought directly from Elsevier's Science &

Technology Rights Department in Oxford, UK: phone: (�44) 1865 843830,

fax: (�44) 1865 853333, e-mail: [email protected]. You may

also complete your request on-line via the Elsevier homepage

(http://www.elsevier.com), by selecting `Customer Support' and

then `Obtaining Permissions'

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library

Library of Congress Cataloguing in Publication Data

A catalogue record for this book is available from the Library of Congress

ISBN 07506 4791 4

For information on all Elsevier Butterworth-Heinemann

publications visit our website at http://books.elsevier.com

Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India

Printed and bound in Great Britain by Biddles Ltd., King's Lynn.


Contents

Dedication iiPreface viiSummary xi

Rule 1. Achieve a good quality melt 11.1 Background 11.2 Melting 31.3 Holding 31.4 Pouring 41.5 Melt treatments 5

1.5.1 Degassing 51.5.2 Additions 6

1.6 Filtration 71.6.1 Packed beds 71.6.2 Alternative varieties of

filters 81.6.3 Practical aspects 8

Rule 2. Avoid turbulent entrainment (thecritical velocity requirement) 92.1 Maximum velocity

requirement 102.2 The `no fall' requirement 132.3 Filling system design 15

2.3.1 Gravity pouring ofopen-top moulds 15

2.3.2 Gravity pouring of closedmoulds 16

2.3.3 Horizontal transfercasting 68

2.3.4 Counter-gravity 722.3.5 Surface tension controlled

filling 752.3.6 Inclusion control: filters

and traps 782.3.7 Practical calculation of the

filling system 93

Rule 3. Avoid laminar entrainment of thesurface film (the non-stopping,non-reversing condition) 1023.1 Continuous expansion of the

meniscus 1023.2 Arrest of vertical progress 103

3.3 Waterfall flow 1043.4 Horizontal stream flow 1043.5 Hesitation and reversal 106

Rule 4. Avoid bubble damage 1084.1 Gravity-filled running

systems 1114.2 Pumped and low-pressure filling

systems 112

Rule 5. Avoid core blows 1145.1 Background 1145.2 Prevention 117

Rule 6. Avoid shrinkage damage 1206.1 Feeding systems design

background 1206.1.1 Gravity feeding 1236.1.2 Computer modelling of

feeding 1246.1.3 Random perturbations to

feeding patterns 1246.1.4 Dangers of solid

feeding 1256.1.5 The non-feeding roles of

feeders 1256.2 The seven feeding rules 126

Rule 1: Do not feed 126Rule 2: Heat-transfer

requirement 127Rule 3: Mass-transfer

requirement 128Rule 4: Junction

requirement 132Rule 5: Feed path

requirement 133Rule 6: Pressure gradient

requirement 138Rule 7: Pressure

requirement 1406.3 The new feeding logic 142

6.3.1 Background 1426.3.2 The new approach 143

6.4 Active feeding 1456.5 Freezing systems design 146


6.5.1 External chills 1476.5.2 Internal chills 1496.5.3 Fins 150

Rule 7. Avoid convection damage 1577.1 Convection: the academic

background 1577.2 Convection: the engineering

imperatives 1577.3 Convection damage and casting

section thickness 1607.4 Countering convection 162

Rule 8. Reduce segregation damage 163

Rule 9. Reduce residual stress (the `no waterquench' requirement) 1669.1 Introduction 1669.2 Residual stress from casting 1669.3 Residual stress from

quenching 1679.4 Distortion 1729.5 Heat treatment developments 1739.6 Epilogue 174

Rule 10. Provide location points 17510.1 Datums 17510.2 Location points 176

10.2.1 Rectilinear systems 17710.2.2 Cylindrical systems 17810.2.3 Trigonal systems 17910.2.4 Thin-walled boxes 179

10.3 Location jigs 18010.4 Clamping points 18010.5 Mould design: the practical

issues 18110.6 Casting accuracy 18210.7 Tooling accuracy 18310.8 Mould accuracy 18310.9 Summary of factors affecting

accuracy 18610.10 Metrology 186

Appendix 188The 1.5 factor 188The Bernoulli equation 189Rate of pour of steel castings from a

bottom-pour ladle 191Running system calculation record 191Design methodology for investment

castings 194

References 195

Index 199

vi Contents


Preface

Castings can be difficult to get right. Creatingthings never is easy. But sense the excitement ofthis new arrival:

The first moments of creation of the newcasting are an explosion of interacting events;the release of quantities of thermal and chemicalenergy trigger a sequence of cataclysms.

The liquid metal attacks and is attacked by itsenvironment, exchanging alloys, impurities, andgas. The surging and tumbling flow of the meltthrough the running system can introduceclouds of bubbles and Sargasso seas of oxidefilm. The mould shocks with the vicious blast ofheat, buckling and distending, fizzing with thevolcanic release of vapours that flood throughthe liquid metal by diffusion, or reach pressuresto burst the liquid surface as bubbles.

During freezing, liquid surges through thedendrite forest to feed the volume contractionon solidification, washing off branches, cuttingflow paths, and polluting regions with excesssolute, forming segregates. In those regions cutoff from the flow, continuing contraction causesthe pressure in the residual liquid to fall, possi-bly becoming negative (as a tensile stress in theliquid) and sucking in the solid surface of thecasting. This will continue until the casting issolid, or unless the increasing stress is suddenlydispelled by an explosive expansion of a gas orvapour giving birth to a shrinkage cavity.

The surface sinks are halted, but the internaldefects now start.

The subsequent cooling to room temperatureis no less dramatic. The solidified casting strivesto contract while being resisted by the mould.The mould suffers, and may crush and crack.The casting also suffers, being stretched as on arack. Silent, creeping strain and stress changeand distort the casting, and may intensify to the

point of catastrophic failure, tearing it apart,or causing insidious thin cracks. Most trea-cherous of all, the strain may not quite crackthe casting, leaving it apparently perfect, butloaded to the brink of failure by internal resi-dual stress.

These events are rapidly changing dynamicinteractions. It is this rapidity, this dynamism,that characterizes the first seconds and minutesof the casting's life. An understanding of them iscrucial to success.

This new work is an attempt to provide aframework of guidelines together with the back-ground knowledge to ensure understanding; toavoid the all too frequent disasters; to cultivatethe targeting of success; to encourage a profes-sional approach to the design and manufactureof castings.

The reader who learns to guide the produc-tion methods through this minefield will find therare reward of a truly creative profession. Thestudent who has designed the casting method,and who is present when the mould is openedfor the first time will experience the excitementand anxiety, and find himself asking the ques-tion asked by all foundry workers on suchoccasions: `Is it all there?' The casting designrules in this text are intended to provide, so faras present knowledge will allow, enough pre-dictive capability to know that the casting willbe not only all there, but all right!

The clean lines of the finished engineeringcasting, sound, accurate, and strong, are a plea-sure to behold. The knowledge that the castingcontains neither defects nor residual stress is anadditional powerful reassurance. It represents amiraculous transformation from the originaltwo-dimensional form on paper or the screen toa three-dimensional shape, from a mobile liquid


to a permanently shaped, strong solid. It is anachievement worthy of pride.

The reader will need some backgroundknowledge. The book is intended for final yearstudents in metallurgy or engineering, for thoseresearching in castings, and for casting engi-neers and all associated with foundries that haveto make a living creating castings.

Good luck!This new book is the second of three books

dealing with castings. The three books are(i) Principles (the new metallurgy of cast metals;the metallurgist's book) (ii) Practice (the prac-tical founder's book) and (iii) Processes (anappraisal of the various methods of makingcastings; perhaps a casting buyer's book). Thethree are intended as a sequence, dealing withthe theory and practice of the casting of metals.At the rate at which new understanding isemerging, an additional text may also berequired; (iv) Properties (a book for everyone).

The second in the series is devoted to the TenRules. These are my own checklist to ensure thatno key aspect of the design of the manufacturingroute for the casting is forgotten.

The Ten Rules listed here are proposed asnecessary, but not, of course, sufficient, for themanufacture of reliable castings. It is proposedthat they are used in addition to existingnecessary technical specifications such as alloytype, strength, and traceability via internationalstandard quality systems, and other well-knownand well-understood foundry controls such ascasting temperature etc.

Although not yet tested on all cast materials,there are fundamental reasons for believing thatthe Rules have general validity. They have beenapplied to many different alloy systems includ-ing aluminium, zinc, magnesium, cast irons,steels, air- and vacuum-cast nickel and cobalt,and even those based on the highly reactivemetals titanium and zirconium. Nevertheless, ofcourse, although all materials will probablybenefit from the application of the Rules, somewill benefit almost out of recognition, whereasothers will be less affected.

The Ten Rules are first listed in summaryform. They are then addressed in more detailin the following ten chapters with one chapterper Rule.

The Rules originated when emerging froma foundry on a memorable sunny day. Theauthor was discussing with indefatigable Boeingenthusiasts for castings, Fred Feiertag and DaleMcLellan, that the casting industry had speci-fications for alloys, casting properties, andcasting quality checking systems, but what didnot exist but was most needed was a processspecification. Dale threw out a challenge: `Write

one!'. The Rules and this book are the outcome.It was not perhaps the outcome that either Daleor I originally imagined. A Process Specificationhas proved elusive, proving so difficult thatI have concluded that it will need a moreaccomplished author.

The Rules as they stand therefore constitute afirst draft of a Process Specification; more like achecklist of casting guidelines. A buyer of cast-ings would demand that the list were fulfilled ifhe wished to be assured that he was buying thebest possible casting quality. If he were to spe-cify the adherence to these Rules by the castingproducer, he would ensure that the quality andreliability of the castings was higher than couldbe achieved by any amount of expensivechecking of the quality of the finished product.

Conversely, of course, the Rules are intendedto assist the casting manufacturer. It will speedup the process of producing the casting rightfirst time, and should contribute in a major wayto the reduction of scrap when the casting goesinto production. In this way the caster will beable to raise standards, without any significantincrease in costs. Quality will be raised to thepoint at which castings of quality equal to thatof forgings can be offered with confidence. Onlyin this way will castings be accepted by theengineering profession as reliable, engineeredproducts, and assure the future prosperity ofboth the casting industry and its customers.

It is recognized that many users of this bookwill be students of casting technology. Forcompleteness therefore, the strict description ofthe Rules as intended as the caster's checklisthas been relaxed a little. A small addition hasbeen made to paragraph 10, extending the sec-tion describing the requirement for locationpoints. This extension includes related aspectsnot included elsewhere, such as the accuracy ofthe whole mould assembly, and the many-sidedproblems of mould design.

A further feature of the work that emerged asthe book was being written was the dominanceof Chapter 2, the design of the filling systems ofcastings. It posed the obvious question `why notdevote the book completely to filling systems?'. Idecided against this option on the grounds thatboth caster and customer require products thatare good in every respect. The failure of any oneaspect may endanger the casting. Therefore,despite the enormous disparity in length ofchapters, none could be eliminated; they were allneeded.

Finally, it is worth making some generalpoints about the whole philosophy of makingcastings.

For a successful casting operation, one of therevered commercial goals is the attainment of

viii Preface


product sales being at least equal to manufactur-ing costs. There are numerous other requirementsfor the successful business, like management,plant and equipment, maintenance, accounting,marketing, negotiating etc. All have to be ade-quate, otherwise the business can suffer, andeven fail.

This text deals only with the technical issuesof the quest for good castings. Without goodcastings it is not easy to see what future a castingoperation can have. The production of goodcastings can be highly economical and reward-ing. The production of bad castings is usuallyexpensive and damaging.

The `good casting' in this text is defined as onethat meets or exceeds the customer's specification.

It is also worth noting at this early stage, thatwe hope that meeting the customer's specifica-tion will be equivalent to meeting or exceedingservice requirements. However, occasionally it isnecessary to live with the irony that the aims ofthe customer and the requirements for serviceare sometimes not in the harmony one wouldlike to see.

These problems illustrate that there are easierways of earning a living than in the castingindustry. But few are as exciting.

J.C.West Malvern

3 September 2003

Preface ix


The 10 Rules: Summary

1. Start with a good quality melt

Immediately prior to casting, the melt shall beprepared, checked, and treated, if necessary, tobring it into conformance with an acceptableminimum standard. However, preferably, pre-pare and use only near-defect-free melt.

2. Avoid turbulent entrainment of thesurface film on the liquid

This is the requirement that the liquid metalfront (the meniscus) should not go too fast.Maximum meniscus velocity is approximately0.5 msÿ1 for most liquid metals. This maximumvelocity may be raised in constrained runningsystems or thin section castings. This require-ment also implies that the liquid metal must notbe allowed to fall more than the critical heightcorresponding to the height of a sessile drop ofthe liquid metal.

3. Avoid laminar entrainment of thesurface film on the liquid

This is the requirement that no part of the liquidmetal front should come to a stop prior to thecomplete filling of the mould cavity. Theadvancing liquid metal meniscus must be kept`alive' (i.e. moving) and therefore free fromthickened surface film that may be incorporatedinto the casting. This is achieved by the liquidfront being designed to expand continuously. Inpractice this means progress only uphill in acontinuous uninterrupted upward advance; i.e.(in the case of gravity poured casting processes,

from the base of the sprue onwards). Thisimplies

� Only bottom gating is permissible.� No falling or sliding downhill of liquid metal

is allowed.� No horizontal flow of significant extent.� No stopping of the advance of the front due to

arrest of pouring or waterfall effects etc.

4. Avoid bubble entrainment

No bubbles of air entrained by the filling systemshould pass through the liquid metal in themould cavity. This may be achieved by:

� Properly designed offset step pouring basin;fast back-fill of properly designed sprue;preferred use of stopper; avoidance of the useof wells or other volume-increasing featuresof filling systems; small volume runner and/oruse of ceramic filter close to sprue/runnerjunction; possible use of bubble traps.

� No interruptions to pouring.

5. Avoid core blows

� No bubbles from the outgassing of cores ormoulds should pass through the liquid metalin the mould cavity. Cores to be demonstratedto be of sufficiently low gas content and/oradequately vented to prevent bubbles fromcore blows.

� No use of clay-based core or mould repairpaste unless demonstrated to be fully driedout.


6. Avoid shrinkage

� No feeding uphill in larger section thicknesscastings because of (i) unreliable pressuregradient and (ii) complications introducedby convection.

� Demonstrate good feeding design by follow-ing all Feeding Rules, by an approvedcomputer solidification model, and by testcastings.

� Control (i) the level of flash at mould and corejoints; (ii) mould coat thickness (if any); and(iii) temperatures of metal and mould.

7. Avoid convection

Assess the freezing time in relation to the timefor convection to cause damage. Thin and thicksection castings automatically avoid convectionproblems. For intermediate sections either(i) reduce the problem by avoiding convectiveloops in the geometry of the casting and rigging,

(ii) avoid feeding uphill, or (iii) eliminate con-vection by roll-over after filling.

8. Reduce segregation

Predict segregation to be within limits of thespecification, or agree out-of-specificationcompositional regions with customer. Avoidchannel segregation formation if possible.

9. Reduce residual stress

No quenching into water (cold or hot) followingsolution treatment of light alloys. (Polymerquenchant or forced air quench may be accept-able if casting stress is shown to be negligible.)

10. Provide location points

All castings to be provided with agreed locationpoints for pickup for dimensional checking andmachining.

xii The 10 Rules: Summary


Rule 1

Achieve a good quality melt

1.1 Background

It is a requirement that either the process forthe production and treatment of the melt shallhave been shown to produce good quality liquid,or the melt should be demonstrated to be of goodquality,or,preferably,both.Agoodqualityliquidis one that is defined as

(i) Substantially free from suspensions of non-metallic inclusions in general, and bifilms inparticular.

(ii) Relative freedom from bifilm-openingagents. These include gas in solution andcertain alloy impurities (such as Fe in Alalloys) in solution.

It should be noted that such melts are not to beassumed, and, without proper treatment, areprobably rare. (Additional requirements, notpart of this specification, may also be placed onthe melt. For instance, low values of particularsolute impurities that have no effect on bifilms.)

Unfortunately, many melts start life withpoor, sometimes grossly poor, quality in terms ofcontent of suspended bifilms. Figure 1.1 givesseveral examples of different poor qualities ofliquid aluminium alloy. The figures show resultsfrom reduced pressure test (RPT) samplesobserved by X-ray radiography. Since the sam-ples are solidified under only one tenth of anatmosphere (76 mm compared to 760 mm offull atmospheric pressure) any gas-containingdefects, such as bubbles, or bifilms with airoccluded in the centres of their sandwich struc-tures, will be expanded by ten times. Thus rathersmall defects can become visible for the first time.

We can assume [following the conclusions ofCastings (2003)] that bifilms always initiate

pores, and that the formation of rounded poressimply occurs as a result of the bifilm beingopened by excess precipitation of gas, finallyachieving a diameter greater than its originallength. Thus the RPT is an admirably simpledevice for assessing (i) the number of bifilms,but (ii) gas content is assessed by the degree ofopening of the bifilms from thin crack-likeforms to fairly spherical pores.

If the melt contained no gas-containingdefects the radiographs would be clear.

However, as we can see immediately, andwithout any benefit of complex or expensiveequipment, the melts recorded in Figure 1.1 arefar from this desirable condition. Figure (a)shows a melt with small rounded pores indi-cating that the bifilms that initiated these defectswere particularly small, of the order of 0.1 mmor less. The density of these defects, however,was high, between 10 and 100 defects per cm3.Sample (b) has a similar defect distribution, butwith slightly higher hydrogen content. Sample (c)illustrates a melt that displayed a deep shrinkagepipe, normally interpreted to mean good qual-ity, but showing that it contained a scatteringof larger pores, probably as a result of fewerbifilms, so that the available gas was con-centrated on the fewer available sites. Melt (d)has considerably larger bifilms, of size in theregion of 5 mm in length, and in a concentrationof approximately 1 per cm3. Samples (e) and (f )show similar samples but with increasing gascontents that have inflated these larger bifilmsto reasonably equiaxed pores.

Naturally, it would be of little use for thecasting engineer to go to great lengths to adoptthe best designs of filling and feeding systems ifthe original melt was so poor that a good castingcould not be made from it.


Thus this section deals with some of theaspects of obtaining a good quality melt. Itshould be noted that many of these aspects havebeen already been touched upon in Castings(2003).

In some circumstances it may not be neces-sary to reduce both bifilms and bifilm-openingagents. An interesting possibility for futurespecifications for aluminium alloy castings(where residual gas in supersaturated solutiondoes not appear to be harmful) is that a doublerequirement may be made for the content ofdissolved gas in the melt to be high, but thepercentage of gas porosity to be low. Themeeting of this double requirement will ensureto the customer that bifilms are not present.Thus these damaging but undetectable defectswill, if present, be effectively labelled and madevisible on X-ray radiographs and polished

sections by the precipitation of dissolved gas. Itis appreciated that such a stringent specificationmight be viewed with dismay by presentsuppliers. However, at the present time we havemainly only rather poor technology, makingsuch quality levels out of reach.

(For steels, the content of hydrogen is a moreserious matter, especially if the section thicknessof the casting is large. In some steel castings ofsection thickness above about 100 mm or so, thehydrogen cannot escape by diffusion during thetime available for cooling or during the time ofany subsequent heat treatments. Thus the highhydrogen content retained in these heavy sec-tions can lead to hydrogen embrittlement, andcatastrophic failure of the section by cracking.)

The possible future production of Al alloysfor aerospace, with high hydrogen content butlow porosity, is a fascinating challenge. As ourtechnology improves such castings may befound not only to be manufacturable, but offerguaranteed reliability of fatigue life, and there-fore command a premium price.

The prospect of producing ultra-cleanAl alloys that can be demonstrated in this wayto be actually extremely clean, raises the issueof contamination of the liquid alloy from thenormal metallurgical additions such as the var-ious master alloys, and grain refiners, modifiers,etc. It may be that for clean material, normalmetallurgical additions to achieve refinement ofvarious kinds will be found unnecessary, andpossibly even counter-productive.

For ductile iron production the massiveamounts of turbulence that accompany theaddition of magnesium in some form, such asmagnesium ferro-silicon, are almost certainlyhighly damaging to the liquid metal. It isexpected that immediately after such nodular-ization treatment that the melt will be massivelydirty. It will be useful therefore to ensure thatthe melt can dwell for sufficient time for theentrained magnesium-oxide-rich films to floatout. The situation is analogous to the treatmentof cast iron with CaSi to effect inoculation (i.e.to achieve a uniform distribution of graphite ofdesirable form). In this case the volume ofcalcium-oxide-rich films is well known, so thatthe CaSi treatment is known as a `dirty' treat-ment compared to FeSi inoculation. The authoris unsure about in-mould treatments thereforewith such oxidizable elements as Ca and Mg. Dothey give results as good as external treatments?If so, how is this possible? Some work to clarifythis situation would be valuable.

For nickel-based superalloys melted and castin vacuum, it is with regret that the material is,despite its apparently clean melting environ-ment, found to be sometimes as crammed with

Figure 1.1 Radiographs of RPT samples of Al±7Si,ÿ 0.4Mg alloy illustrating different bifilm populations(courtesy S. Fox)

2 Castings Practice: The 10 Rules of Castings

(a) (b)

(e) (f)

(c) (d)


oxides (and/or nitrides) as an aluminium alloy(Rashid and Campbell 2004). This is because themain alloying element in such alloys isaluminium, and the high temperature favoursrapid formation and thickening of the surfacefilm on the liquid. This occurs even in vacuum,because the vacuum is only, of course, diluteair. Although thermodynamics indicates thataluminium oxide (and perhaps also the nitride)is unstable in vacuum, there is no doubt that afilm forms rapidly, and can become entrained,thus damaging the liquid and any subsequentcasting. The melting and pouring processes ofthese alloys also leaves much to be desired. Thealloy is melted by induction, and is pouredunder gravity via a series of sloping launders,falling several times, and finally falling one ortwo metres or more into steel tubes that act asmoulds. This awfully turbulent primary pro-duction process for the alloy impairs all thedownstream products.

A recent move towards the production of Ni-base alloy bar by horizontal continuous castingis to be welcomed as the first step towardsa more appropriate production technique forthese key ingredients of our modern aircraftturbines. Production of improved material iscurrently limited, but should be the subject ofdemand from customers. Poor casting technol-ogy has been the accepted norm within the air-craft industry for too long. (However, as we allknow, the aircraft industry is not alone.)

1.2 Melting

The quality of melts can be significantly affectedby the type of melting furnace.

For instance the melting of the charge in asingle pot is usual. Furnaces of this type auto-matically ensure that all the oxides on the surfaceof the charge material will be incorporated intothe melt. Thus crucible furnaces, whether bale-out or tilting, or whether heated by gas or elec-trical resistance, are all of this type. Also includedare induction furnaces and reverbatory (meaningreflected heat from a roof ) melting units.

The remelting of aluminium alloy sand cast-ings probably represents one of the worst cases,since the oxide skin on a sand casting is particu-larly thick, having cooled from high tem-perature in a reactive environment. When theskin is submerged in the melt it can floatabout, substantially complete. When remeltingscrapped cylinder heads by adding them in toone end of a combined melting and holdingfurnace, the author has fished out the skin of acomplete cylinder head from the opposite end.

The remelting of aluminium alloy gravity die(permanent moulded) castings have an oxideskin that is much thinner, and seems to give lessproblems. Whether this is a real or imaginedadvantage in view of the damage that can becaused by any entrained oxide skin, irrespectiveof its thickness, is not clear at this time.

Induction furnaces enjoy the great advantageof extremely rapid melting. However, theyhave long been regarded with some reserveby aluminium melters because of the electro-magnetic stirring, with the suspicion that oxidesmay therefore be entrained. With normal induc-tive coil geometry there is a high-pressure regionnear the centre of the wall that drives a doubletorroid (a torroid is a ring shape like a doughnut)in directions away from this point. However,there is no evidence known to the author that thestirring is sufficiently rapid to entrain oxide,although such a problem cannot be ruled out.What is certain is that any oxide will have noopportunity to settle out, but this is also true ofmost of the above crucible furnaces because ofthe presence of natural convection. Because of theheat input via the walls of the crucible, and heatloss from the top surface, the convective stirringwill be expected to take the form of a simpletorroid, the flow direction being upwards at thewall, and downwards in the centre. The onlysignificant difference, if any, between these twostirring modes is the rate of stirring. It is possiblethat the higher energy in the induction furnacemay shred films, whereas the natural convectiveregimes in other furnaces would be expected toconserve the original film size distribution.

The dry hearth type of furnace is quite dif-ferent. The charge to be melted is heaped onto adry, sloping refractory floor, called a hearth. Asthe charge melts, the liquid alloy flows out ofits oxide skin and down the sloping hearth intothe main melt. The oxide skins present on thesurface of the charge materials remain behind,accumulating on the hearth. The pile of dross israked off the hearth at intervals via a side door.Such melting units are useful in aluminium andcopper alloy production.

1.3 Holding

Holding furnaces can also have a significanteffect on melt quality.

Holding furnaces were originally selected fortheir utility in smoothing the supply of moltenmetal between batch melters and a fluctuatingdemand from the casting requirements. Anadditional advantage was the smoothing oftemperature and chemical analysis that wasunavoidably variable from batch to batch.

Rule 1. Achieve a good quality melt 3


The Cosworth Process was perhaps the firstto acknowledge that for liquid aluminium alloysthe oxide inclusions in a holding furnace couldbe encouraged to separate simply by a sink andfloat principle; the metal for casting being takenby a pump from a point at about midway depthwhere the best quality metal was to be expected.

In contrast, the holding of melts in closedvessels for the low-pressure die casting ofaluminium, or the dosing of the liquid metal, areusually impaired by the initial turbulent pour ofthe melt to fill the furnace. The total pour heightis often of the order of a metre. Not only are newoxides folded into the melt in this pouring action,but those oxides that have settled to the floor ofthe furnace since the last filling operation arestirred into the melt once again. Finally, theseenclosed units suffer from the inaccessibility ofthe melt. This usually restricts any thoroughaction to improve the melt by any kind ofdegassing technique.

The Alotech approach to the design of aholding furnace is patented and not available forpublication at the time of writing. It is hoped torectify this in future editions of this work asdetails of the process are published. Why evenmention it at this stage? The purpose of men-tioning it here is to illustrate that even apparentlysimple equipment such as a holding furnace iscapable of considerable sophistication, leadingto the production of greatly improved processingand products. It takes the concept of meltcleaning and degassing to an ultimate level thatprobably represents a limit to what can beachieved. In addition, the technique is simple,low capital cost, low running cost, has no movingparts and is operator-free. At this time thetechnique is being applied only to aluminiumalloys.

1.4 Pouring

Most foundries handle their metal from onepoint to another by ladle. The metal is, ofcourse, transferred out of the ladle by pouring.In most foundries multiple pours are needed totransfer the liquid metal from the melting unit tothe mould.

At every pouring operation, it is likely thatlarge areas of oxide film will be entrained in themelt because pour heights are usually not con-trolled. It is known that pour heights less thanthe height of the sessile drop cannot entrain thesurface oxide. However, such heights are verylow; 16 mm for Mg, 13 mm for Al, and only8 mm for dense metals such as copper-base, andiron and steel alloys.

However, this theoretical limit, while abso-lutely safe, may be exceeded for some metals withminimum risk. As long ago as 1928 Beck descri-bed how liquid magnesium could be transferredfrom a ladle into a mould by arranging thepouring lip of the ladle to be as close as possibleto the pouring cup of the mould, and relativelyfixed in position. In this way the semi-rigid oxidetube that formed automatically around the jetremained unbroken, and so protected the fallingstream.

Experiments by Din and Campbell (2002) onAl±7Si±0.4Mg alloy have demonstrated thatin practice the damage caused by falls up to100 mm appears controlled and reproducible. Thisis in close agreement with early observations byTurner (1965) who noted that air was taken intothe melt, reappearing as bubbles on the surfacewhen the pouring height exceeded about 90 mm.

Above 200 mm, Din and Campbell (2003)found that random damage was certain. Atthese high energies of the plunging jet, bubblesare entrained, with the consequence that bubbletrails add to the total damage in terms of area ofbifilms.

In general, it seems that the lower the pourheight the less damage is suffered by the melt. Inaddition, of course, less metal is oxidized, thusdirectly saving the costs of unnecessary meltlosses. Ultimately, however, it is, of course, bestto avoid pouring altogether. In this way lossesare reduced to a minimum and the melt ismaintained free from damage.

Until recent years, such concepts have beenregarded as pipe dreams. However, the devel-opment of the Cosworth Process has demon-strated that it is possible for aluminium alloycastings to be made without the melt sufferingany pouring action at any point of the process.Once melted, the liquid metal travels alonghorizontal heated channels, retaining its con-stant level through the holding furnace, andfinally to the pump, where it is pressurized to fillthe mould in a counter-gravity mode. Suchtechnology would also appear to be relativelyeasily applied to magnesium alloys.

The potential for extension of this technologyto other alloy base systems such as copper-basedor iron-based alloys is less clear. This is becausemany of these other alloy systems either do notsuffer the same problems from bifilms, or do nothave the production requirements of some ofthe high volume aluminium foundries. Thus innormal circumstances, many irons and steels arerelatively free from bifilms because of the largedensity difference between the inclusions andthe parent melt, encouraging rapid flotation.Alternatively, many copper-based and steelfoundries are more like jobbing shops, where



the volume requirement does not justify acounter-gravity system, and high technologypumps, if they were available, would haveproblems surviving the oxidation and thermalshock of a stop-go production requirement.

Despite these reservations, the counter-gravityGriffin Process has been impressively successfulfor the volume production of steel wheels forrail rolling stock. The process produces wheelsthat require no machining (apart from thecentre hub). The outer cast rim runs directly onthe steel rail. The products outperform forgedsteel wheels in terms of reliability in service,earning the process 80 per cent of the market inthe USA. What a demonstration of the sound-ness of the counter-gravity concept, contrastingdramatically with steel castings producedworld-wide by gravity pouring, in which defectsand expensive upgrading of the casting arethe norm.

1.5 Melt treatments

1.5.1 Degassing

Gases dissolved in melts are disadvantageousbecause they precipitate in bifilms and causethem to unfurl, and even open further as cracksor voids, thereby progressively reducing themechanical properties of the cast product.

Treatments to reduce the gas content of meltsinclude vacuum degassing. Such a technique hasonly been widely adopted by the steel industry.If a melt were simply to be placed under vacuumthe rate of degassing would be low because thedissolved gas would have to diffuse to the freesurface to escape. The slow convection of themelt will gradually bring most of the volumenear to the top surface, given time. The processis greatly speeded by the introduction of mil-lions of small bubbles of inert gas via a porousplug or other technique. In this way the processis accomplished in a fraction of the time.

Traditionally the steel industry has used thetechnique of a carbon boil, in which the creationand floating out of carbon monoxide bubblesfrom the melt carries away unwanted gases suchas oxygen (actively by chemical reaction withcarbon) and hydrogen (passively by simpleflushing action). The nitrogen in the melt maygo up or down depending on its starting valueand the nitrogen content of the environmentabove the melt, since it will tend to equilibratewith the environment. The more recent oxygensteelmaking processes in which oxygen is injectedinto the melt certainly reduce hydrogen andnitrogen, but require the oxygen to be reducedeither chemically by reaction with C or other

deoxidizers such as Si, Mn or Al etc., or useeven more modern techniques such as AOD(argon-oxygen-decarburization). We shall notdwell further on these sophistications, since thesespecialized techniques, so well understood andwell developed for steel, are almost unusedelsewhere in the casting industry.

By comparison, the approaches to thedegassing of aluminium alloys, containing onlyhydrogen, has for many years been primitive.Only recently have effective techniques beenintroduced.

For instance until approximately 1980,aluminium was commonly degassed usingimmersed tablets of hexachlorethane that therm-ally degraded in the melt to release largebubbles of chlorine and carbon (the latter assmoke). Alternatively, a primitive tube lancewas used to introduce a gas such as nitrogen orargon. Again large bubbles of the gas wereformed. These techniques involving the genera-tion of large bubbles were so inefficient thatlittle dissolved gas could be removed, butthe creation of large areas of fresh melt to theatmosphere each time a bubble burst at thesurface of the melt provided an excellentopportunity for the melt to equilibrate with theenvironment. Thus on a dry day the degassingeffect might be acceptable. On a damp day, orwhen the flue gases from the gas-fired furnacewere suffering poor extraction, the melt couldgain hydrogen faster than it could lose it. Thispoor rate of degassing, combined with the highrate of regassing from interaction with theenvironment, led to variable and unsatisfactoryresults.

Rotary degassing came to the rescue. The useof a rapidly rotating rotor to chop bubbles ofinert gas into fine clouds raised the rate ofdegassing and lowered the rate of regassing.Thus effective degassing could be reliablyachieved in times that were acceptable in aproduction environment.

It is to be hoped that this is not the end of thestory for the degassing of aluminium alloys. Forinstance the use of an `inert' gas is a convenientuntruth. Even sources of high purity inert gascontain sufficient oxygen as an impurity tocreate an oxide film on the inside of the surfaceof the bubbles. Thus millions of very thin bifilmsmust be created during this degassing process.For a new rotor, or after a weekend, the rotorand its shaft will have absorbed considerablequantities of water vapour (most refractoriescan commonly absorb water up to 10 per cent oftheir weight). Furthermore, the gas lines areusually not clean, or contain long rubber orplastic tubing. These materials absorb andtherefore leak large quantities of volatiles into



the degassing line. Thus by the time the gasarrives in the melt it is usually not particularlyinert. At this time no-one knows whether theoxides created as by-products of degassing arenegligible or whether they seriously reduceproperties. What is known is that aluminiumalloys are usually greatly improved by rotarydegassing. Whether further improvements can besecured is not clear.

Part of the considerable benefit of rotarydegassing is not merely the degassing action. Itseems that the millions of tiny bubbles attach tooxides in the melt and float them to the top,where they can be skimmed off. After treatment,after the rotor has been raised out of the meltand the surface skimmed, a test of the efficiencyof the cleaning action is simply to look at thesurface of the melt. If the cleaning action is notcomplete, during the next few minutes smallparticles will be seen to arrive at the surface,under the surface oxide film, which is seen to bepushed upwards by the particle. The treatmentcan be repeated until no further arrival of debriscan be seen. The melt surface then retains itspristine mirror smoothness. The melt can thenbe pronounced as clean as the treatment canachieve.

The action of rotary degassing on oxide filmsmerits further examination. Much has beenwritten on the benefits of small bubbles on therate of degassing (although the rate of regassingfrom a damp rotor, or from the environmenthas been neglected). The action of the bubbles toeliminate films has in general been overlooked.However, the size of bubbles in relation to theefficiency of removal of films is probably crit-ical. For instance, large bubbles will displacelarge volumes of melt during their rise to thesurface, thus displacing films sideways, so thatthe film and bubble never make contact. On theother hand, small bubbles will displace relativelylittle liquid, and so be able to impact on rela-tively large films in their path. Thus such con-tacted films will be buoyed up to the surface.The mutual contact would be expected tobecome important when bubbles and filmswere approximately similar in size. Thus smallbubbles will take out correspondingly smalleroxide films. This predicted effect deservesto be demonstrated experimentally at somefuture date.

An experience by the author illustrates someof the misconceptions surrounding the dual roleof hydrogen and oxide bifilms in aluminiummelts. An operator used his rotary degasser for5 minutes to degas 200 kg of Al alloy. The meltwas tested with a reduced pressure test (RPT;see below) sample that was found to contain nobubbles. The melt was therefore deemed to be

degassed. The melt was immediately pouredinto a transfer ladle on a fork lift truck andconveyed to a low-pressure die casting furnace,into which it was poured. The melt in the low-pressure furnace was then tested again by RPT,and the sample found to contain many bubbles.The operator was baffled. He could not under-stand how so much gas could have re-enteredthe melt in only the few minutes required for thetransfer.

The truth is, of course, that the melt wasinsufficiently degassed with only 5 minutes oftreatment. In fact with a damp rotor the gaslevel is likely to rise initially (getting worse beforeit gets better!). The RPT showed no bubblesnot because the hydrogen was low, but becausethe short treatment had clearly been sufficientlysuccessful to remove a large proportion ofthe bifilms that were the nuclei for the bubbles.This high hydrogen metal was then pouredtwice, each time from a considerable height,re-introducing copious quantities of oxidebifilms that act as excellent nuclei, so thatthe RPT could now reveal its high hydrogencontent.

1.5.2 Additions

Additions to melts are made for a variety ofreasons. These can include additives for chemi-cal degassing (as the addition of Al to steel to fixoxygen and nitrogen) or grain refinement (as inthe addition of titanium and/or boron or carbonto Al alloys). Sometimes it seems certain that thepoor quality of such materials (perhaps meltedpoorly and cast turbulently, and so containing ahigh level of oxides) can contaminate the meltdirectly.

Indirectly, however, the person in charge ofmaking the addition will normally be underinstructions to stir the melt to ensure the dis-solution of the addition and its distributionthroughout the melt. Such stirring actions candisturb the sediment at the bottom of melts,efficiently re-introducing and re-distributingthose inclusions that had spent much time insettling out. The author has vivid memories ofwrecking the quality of early Cosworth melts inthis way: the addition of grain refiners gavewonderfully grain refined castings, and shouldhave improved feeding, and therefore thesoundness of the castings. In fact, all the cast-ings were scrapped because of a rash of severemicroporosity, initiated almost certainly on thestirred-up oxides that constituted the sedimentin the holding furnace.

Additions of Sr to aluminium melts haveoften been accused of also adding hydrogenbecause of the porosity that has often been



noted to follow such additions. Here again, itseems unlikely that sufficient hydrogen could beintroduced by such a small addition. Perhapsthe problem is with stirred sediment, or othermore complex reasons as are discussed at somelength in Castings 2003.

1.6 Filtration

Filtration is perhaps the most obvious way toremove suspended solids from liquid metals.However, it is not without its problems, as weshall see.

The action of a filter in the supply of liquidmetal in the launder of a melt distribution sys-tem in a foundry or cast house (i.e. a foundry forcontinuous casting of long products) is ratherdifferent from its action in the running system ofa shaped casting.

The filters used in castings act for only thefew seconds or minutes that the casting is beingpoured. The velocities through them are high,usually several metres per second, compared tothe more usual 0.1 m/s rates in launders. Thefiltration effect is further not helped by theconcentration of flow through an area often afactor of up to 100 smaller than that used inlaunder systems. The total volume per secondrates at which casting filters are used are there-fore higher by a factor of around 1000. It ishardly surprising therefore that the filtrationaction is reduced nearly to zero. However, theeffect on the flow is profound. The use of filtersin running systems is dealt with in detail later inthe book. We concentrate in this section on fil-ters in launder systems.

The treatment of tonnage quantities of metalhas been developed only for the aluminiumcasting industry. The central problem for mostworkers in this field is to understand how thefilters work, since the filters commonly have poresizes of around 1 mm, whereas, puzzlingly, theyseem to be effective in removing a high percent-age of inclusions of only 0.1 mm diameter. Mostresearchers expand at length, listing themechanisms that might be successful to explainthe trapping of such small solid particles.Unfortunately, these conjectures are probablynot helpful, and are not repeated here.

The fact is that the important solids beingfiltered in aluminium alloys are not particlesresembling small solid spheres, as has generallybeen assumed. The important particles are films(actually always double films that we havecalled bifilms). Once this is appreciated thefiltration mechanism becomes much easierto understand. The films are often of size 1 to10 mm, and so are, in principle, easily trapped

by pores of 1 mm diameter. Such bifilms are noteasily seen in their entirety in an optical micro-scope, the visible portions appearing to be muchsmaller, explaining why filters appear to arrestparticles smaller than their pore diameters.

We need to take care. This explanation, whileprobably having some truth, may oversimplifythe real situation. In Castings (2003), the life ofthe bifilm was described as starting as the fold-ing in of a planar crack-like defect as a result ofsurface turbulence. However, internal turbu-lence wrapped the defect into a compactform, reducing its size by a factor of 10 or so. Inthis form it could pass through a filter, andfinally open once again in the casting as theliquid metal finally came to rest, and bifilm-opening (i.e. unfurling) processes started tocome into action.

In more detail now, the trapping of compactforms of films is explained by their irregularand changing form. During the compactedstage of their life, they will be constantly in a stateof flux, ravelling and unravelling as they travelalong in the severely turbulent flow. Two-dimensional images of such defects seen onpolished sections always show loose trailingfragments. Thus in their progress through thefilter, one such end could become attached, pos-sibly wrapping over a web or wall of filter mate-rial. The rest of the defect would then roll out,unravelling in the flow, and be flattened againstthe internal surfaces of the filter, where it wouldremain fixed in the tranquil boundary layer.

1.6.1 Packed beds

In practice, in DC (direct chill) continuouscasting plants, filters have been used for manyyears, sited in the launder system between themelting furnaces and casting units. Commonly,the filtration system is a large and expensiveinstallation, comprising a crucible furnace thatcontains a divided crucible. One half is filledwith refractory material such as alumina balls,or tabular alumina. The flow of the melt downone half of the crucible, through a connectingport, and up through the deep packed bed ofsuch systems has been shown to be effective ingreatly reducing the inclusion count (the num-ber of inclusions per unit area). However, it isknown that the accidental disturbance of suchfilters releases large quantities of inclusions intothe melt stream. This has also been reportedwhen enthusiastic operators see the filterbecoming blocked by the increasing upstreamlevel of the melt, and stir the bed with ironrods to ease the flow of metal. It is noteasy to imagine actions that could be morecounter-productive.



Some work has been carried out on filteringliquid aluminium through packed beds of tab-ular or ball alumina (Mutharasan et al. 1981)and through bauxite, alumino-silicates, magnesia,chrome-magnesite, limestone, silicon carbide,carbon, and steel wool (Hedjazi et al. 1975).This latter piece of work demonstrated thatall of the materials were effective in reducingmacro-inclusions. This is perhaps to be expectedas a simple sieve effect. However, only thealumino-silicates were really effective in remov-ing any micro-inclusions and films, whereasthe carbon and chrome-magnesite removedonly a small percentage of films and appearedto actually increase the number of micro-inclusions. The authors suggest that the wett-ability of the inclusions and the filter materialis essential for effective filtration.

An interesting application of a chemicallyactive packed bed is that by Geskin et al. (1986),in which liquid copper is passed through char-coal to provide oxygen-free copper castings. It iscertain, however, that the charcoal will havebeen difficult to dry thoroughly, so that the finalcasting may be somewhat high in hydrogen.Because of the low oxygen content it is likelythat hydrogen pores will not be nucleated (asdiscussed in Castings (2003)). The hydrogen isexpected therefore to stay in solution andremain harmless.

1.6.2 Alternative varieties of filters

Other large and expensive filtration systemsinclude the use of a pack of porous tubes, sealedin a large heated box, through which thealuminium is forced. The pores in this case areof the order of 0.25 mm, with the result that thefilter takes a high head of metal to prime it.However, the technique is not subject to failurebecause of disturbance, and guarantees highquality of liquid metal.

Other smaller and somewhat cheaper systemsthat have been used include a ceramic foamfilter, usually designed to be housed in a box,sited permanently below the surface of the melt.The velocities through the filter are usually low,encouraged by the large area of the filter,usually at least 300� 300 mm. The filter is onlybrought out into the air to be changed when the

metal level either side of the filter box shows alarge difference, indicating that the filter isbecoming blocked.

The efficiency of many filtration devices canbe understood when it is assumed that theimportant filtration action is the removal offilms (not particles). Thus glass cloth is widelyused to good effect in many different forms inthe Al casting industry.

1.6.3 Practical aspects

It is typical of most filtration systems that thehigh quality of metal that they produce (often atconsiderable expense) is destroyed by thought-less handling of the melt downstream.

Even the filter itself can give difficulties inthis way. For instance if the melt exits thefilter downwards, or even horizontally, causingfine jets of metal to form in the air, and plungeinto the melt, additional oxide defects arenecessarily created downstream. The avoidanceof this problem is an important aspect of thedesigning of effective filters into the runningsystems of shaped castings, as will be discussedlater.

Many have wondered whether the filter itselfcauses oxides because the flow necessarilyemerges in a divided state, and therefore mustcreate double films in the hundreds of con-fluence events. Video observations by theauthor on a stream of aluminium alloy emergingfrom a ceramic foam filter with a pore diameterclose to 1 mm have helped to clarify the situa-tion. It seems true that the flow emerges dividedas separate jets. However, within a few milli-metres (apparently depending on the flow rateof the metal) the separate jets merge. Thus theoxide tubes formed around the jets appear to beup to 10 mm long when the melt travelled ataround 500 mm sÿ1 but remained attached tothe filter. The oxide tubes did not extend furtherbecause after the streams merged oxygen wasnecessarily excluded. The forest of tubes wasseen to wave about in the flow like underwatergrass. It is possible that more rapid flows mightcause the grass to detach as a result of its greaterlength and the higher speed of the metal. Thisseems much more likely in the conditions of therunning system of a casting.



Rule 2

Avoid turbulent entrainment (the criticalvelocity requirement)

The avoidance of surface turbulence is probablythe most complex and difficult Rule to fulfilwhen dealing with gravity pouring systems.

The requirement is all the more difficult toappreciate by many in the industry, sinceeveryone working in this field has alwaysemphasized the importance of working with`turbulence-free' filling systems for castings.Unfortunately, despite all the worthy intentions,all the textbooks, all the systems, and all thetalk, so far as the author can discover, it seemsthat no-one appears to have achieved this targetso far. In fact, in travelling around the castingindustry, it is quite clear that the majority (atleast 80 per cent) of all defects are directlycaused by turbulence. Thus the problem ismassive; far more serious than suspected bymost of us in the industry.

To understand the fundamental root of theproblem, it will become clear in Section 2.1 thatany fall greater than the height of the sessiledrop (of the order of 10 mm) causes the metal toexceed its critical velocity, and so introduces thedanger of defects in the casting. As most falls arein fact in the range 10 to 100 times greater thanthis, and as the damage is likely to be propor-tional to the energy involved (i.e. proportionalto the square of the velocity) the damage socreated will usually be expected to be in therange 100 to 10 000 times greater. Thus in thegreat majority of castings that are poured sim-ply under the influence of gravity, there is amajor problem to ensure its integrity. In fact,the situation is so bad that the best outcome ofmany of the solutions proposed in this book isdamage limitation. Effectively, it has to be

admitted that at this time it seem impossible toguarantee the avoidance of some damage whenpouring liquid metals.

This somewhat depressing conclusion needsto be tempered by a number of factors.

First; the world has come to accept castingsas they are. Thus any improvement will bewelcome. This book described techniques thatwill create very encouraging improvements.

Second; this book is merely a summary ofwhat has been discovered so far in the devel-opment of filling system design. Better designsare to be expected now that the design para-meters (such as critical velocity, critical fallheight, etc.) are defined.

Third; there are filling systems that can yield,in principle, perfect results.

Of necessity, such perfection is achieved byfulfilling Rule 2 by avoiding the transfer of themelt into the mould by pouring downhill undergravity. Thus, considering the three directionsof filling a mould:

(i) downhill pouring under gravity;(ii) horizontal transfer into the mould

(achieved by tilt casting in which the tiltconditions are accurately controlled);

(iii) uphill (counter-gravity) casting in which themelt is caused to fill the mould in only anuphill mode;

only the last two processes have the potential todeliver castings of near perfect quality. In myexperience, I have found that in practice it isoften difficult to make a good casting by grav-ity, whereas by a good counter-gravity process


(i.e. a process observing all the 10 Rules) it hasbeen difficult to make a bad casting. The jury isstill out on horizontal transfer by tilting. Thisapproach has great potential, but requires adedicated effort to achieve the correct conditions.

Thus in summary, filling of moulds can becarried out down, along, or up. Only the `along'and `up' modes totally fulfil the non-surfaceturbulence condition.

However, despite all its problems, it seemsmore than likely that the downward mode,gravity casting, will continue to be with us forthe foreseeable future. Thus we shall devotesome considerable time to the damage limitationexercises that can offer considerably improvedproducts, even if, unfortunately, those productscannot be ultimately claimed as perfect. Mostwill shed no tears over this conclusion.Although potential perfection in the along andup modes is attractive, the casting business is allabout making adequate products; products thatmeet a specification and at a price a buyer canafford.

The question of cost is interesting; perhapsthe most interesting. Of course the costs have tobe right, and often gravity casting is acceptableand sufficiently economical. However, moreoften than might be expected, high quality andlow cost can go together. An improved gravitysystem, or even one of the better counter-gravitysystems, can be surprisingly economical andeffective. Such opportunities are often over-looked. It is useful to watch out for such benefits.

2.1 Maximum velocity requirement

One day, I was seated in the X-ray radiographicroom using an illuminated screen to study aseries of radiographic films of cylinder headcastings made by our recently developed castingsystem at Cosworth. Each radiograph in turnwas beautiful, having a clear, `wine glass' per-fection that every founder dreams of. I was atpeace with the world. However, suddenly, aradiograph appeared on the screen that was atotal disaster. It had gas bubbles, shrinkageporosity, hot tears, and sand inclusions. I wasshocked, but sensed immediately what hadhappened. I shot out to query Trevor, our manon the casting station. `What happened to thiscasting?' He admitted instantly `Sorry. I put themetal in too fast.'

This was a lesson that remained with me foryears. This chance experiment by counter-gravity, using an electromagnetic pump allow-ing independent control of the ingate velocity,had kept constant all the other casting variables

(temperature, metal quality, alloy content,mould geometry, aggregate type, binder type,etc.), showing them to be of negligible impor-tance. Clearly, the ingate velocity was domi-nant. By only changing the speed of entry ofmetal we could move from complete success tocomplete failure.

Thinking further about this, common sensetells us all that there is an optimum velocity atwhich a liquid metal should enter a mould. Theconcept is outlined in Figure 2.1. At a velocity ofzero the melt is particularly safe (Figure 2.1a),being free from any danger of damage. Regret-tably, this condition is not helpful for the fillingof moulds. Conversely, at extremely high velo-cities the melt will enter like a jet of water from afire-fighter's hosepipe (Figure 2.1c), and isclearly damaging to both metal and mould. At acertain intermediate velocity the melt rises tojust that height that can be supported by surfacetension around the periphery of the spreadingdrop (Figure 2.1b). The theoretical backgroundto these concepts is dealt with at length in thefirst book in this series Castings (Principles)(2003). For nearly all liquid metals this criticalvelocity is close to 0.5 m sÿ1. This value is ofcentral importance in the casting of liquidmetals, and will be referred to repeatedly in thissection.

The liquid drop, emerging close to its criticalvelocity, and spreading slowly from the ingate isclosely in equilibrium, its surface tension hold-ing the drop in its compact shape, just balancingthe head of pressure tending to spread it becauseof its density. This slowly expanding drop isclosely similar to a sessile (Latin `sitting') drop.(The word contrasts with glissile drop, meaninga gliding or sliding drop.) A sessile drop of Alsitting on a non-wetted substrate is approxi-mately 12.5 mm high. Corresponding values forother liquids are Fe 10 mm, Cu 8 mm, Zn 7 mm,Pb 4, water about 5 mm.

Figure 2.1 The extremes of velocity entering the mouldfrom the gate; zero, critical and high.


(a) (b) (c)


Recent research has demonstrated that if theliquid velocity exceeds the critical velocity thereis a danger that the surface of the liquid metalmay be folded over by surface turbulence. If aperturbation to the surface exceeds the height ofthe sessile drop the liquid can no longer besupported by its surface tension. It will thereforefall back under gravity and may thus entrain itsown surface in an enfolding action. At risk ofoverly repeating this important phenomenon,this entrainment of the surface can occur if thereis sufficient energy in the form of velocity in thebulk liquid to perturb the surface against thesmoothing action of surface tension. In addi-tion, notice that damage is not necessarily cre-ated by the falling back of the metal. The fallingis likely to be chaotic, so that any folding actionmay or may not occur. The significance of thecritical velocity is clear therefore: above thecritical velocity there is the danger of surfaceentrainment leading to defect creation. Belowthe critical velocity the melt is safe fromentrainment problems.

(It is perhaps useful to remind the reader thatentrainment may still occur as a result of surfacecontraction. Thus the steady forward advanceof the meniscus, Rule 3, remains another centralaxiom of good filling systems.)

Japanese workers optimizing the filling oftheir design of vertical stroke pressure die cast-ing machine using both experiment and com-puter simulation (Itamura 1995) confirm thecritical velocity of 0.5 m sÿ1, finding that airbubbles have a chance to become entrainedabove this value. (These workers go further todefine the amount of liquid that needs to be inthe mould cavity to suppress entrainment athigher ingate velocities.)

Normally, the surface is covered with anoxide film, although many other types of filmsare possible in different circumstances. A com-mon alternative is a graphitic film. There is achance therefore that if the speed of the liquidexceeds this critical velocity its surface film maybe folded into the bulk of the liquid. This fold-ing action is an entrainment event. It leads to avariety of problems in the liquid that we cancollectively call entrainment defects. The majorentrainment defects are air bubbles and dou-bled-over oxide films. The author has namedthese folded-in films `bifilms' to emphasize theirdouble, folded-over nature. Because the filmsare necessarily folded dry side to dry side, thereis little or no bonding between these dry inter-faces, so that the double films act as cracks. Thecracks (alias bifilms) become frozen into thecasting, lowering the strength and fatigue resist-ance. Bifilms may also create leak paths, causingleakage failures.

The folding-in of the oxide is a randomprocess, leading to scatter and unreliability inthe properties and performance of the producton a casting to casting, day to day, and monthto month basis during a production run.

The different qualities of metals arriving inthe foundry from batch to batch will also beexpected to contain different quantities anddifferent forms of bifilms. Thus the performanceof the foundry will suffer further variation. Thisis the reason for Rule 1. The foundry needs tohave procedures in place to smooth variationsof its incoming raw material.

Looking a little more closely at the detail ofcritical velocities for different liquids, it is closeto 0.4 m sÿ1 for dense alloys such as irons, steelsand bronzes and about 0.5 m sÿ1 for liquid alu-minium alloys. The value is 0.55 to 0.6 m sÿ1 forMg and its alloys. Taking an average of about0.5 m sÿ1 for all liquid metals is usually goodenough for most purposes related to the designof filling systems for castings, and will begenerally used in this book.

The maximum velocity condition effectivelyforbids top gating of castings (i.e. the plantingof a gate in the top of the mould cavity,causing the metal to fall freely inside the mouldcavity). This is because liquid aluminiumreaches its critical velocity of about 0.5 m sÿ1

after falling only 12.5 mm under gravity. Thecritical velocity of liquid iron or steel is ex-ceeded after a fall of only about 10 mm (theseare, of course, the heights of the sessile drops).Naturally, such short fall distances are alwaysexceeded in practice in top gated castings,leading to the danger of the incorporation ofthe surface films, and consequent leakage andcrack defects.

Castings that are made in which velocitieseverywhere in the mould never exceed the cri-tical velocity are consistently strong, with highfatigue resistance, and are leak-tight (if properlyfed, of course, so as to be free from shrinkageporosity).

Experiments on the casting of aluminiumhave demonstrated that the strength of castingsmay be reduced by as much as 90 per cent ormore if the critical velocity is exceeded. The cor-responding defects in the castings are not alwaysdetected by conventional non-destructive testingsuch as X-ray radiography or dye penetrant,since, despite their large area, the folded oxidefilms are thin, and do not necessarily give rise toany significant surface indications.

The speed requirement automatically excludesconventional pressure die-castings as havingsignificant potential for reliability, since the fillingspeeds are usually 10 to 100 times greater thanthe critical velocity.

Rule 2. Avoid turbulent entrainment 11


Over recent years there have been welcomemoves, introducing some special developmentsof high-pressure technology that are capable ofmeeting this requirement. These include thevertical injection squeeze casting machine, andthe shot control techniques. Such techniquescan, in principle, be operated to fill the cavitythrough large gates at low speeds, and withoutingress of air into the liquid metal. Such castingsrequire to be sawn, rather than broken, fromtheir filling systems of course. Unfortunately,the castings remain somewhat impaired by theaction of pouring into the shot sleeve. Evenhere, these problems are now being addressed bysome manufacturers, with consequent benefitsto the integrity of the castings.

Other uphill filling techniques such as low-pressure filling systems are capable of meetingRule 2. Even so, it is regrettable that the criticalvelocity is practically always exceeded duringthe filling of the low-pressure furnace itselfbecause of the severe fall of the metal as it istransferred into the pressure vessel, so that themetal is damaged even prior to casting. Inaddition, many low-pressure die castingmachines are in fact so poorly controlled onflow rate, that the speed of entry into the diegreatly exceeds the critical velocity, thus negat-ing one of the most important potential benefitsof the low-pressure system. Processes such as theCosworth Process avoid these problems bynever allowing the melt to fall at any stage ofprocessing, and control its upward speed intothe mould by electromagnetic pump.

Metals that can also suffer from entrainedsurface films are suggested to be the ZA (zinc±aluminium) alloys and ductile irons. Carbonand stainless steels are thought to be similar,although in some of these systems the entrainedbifilms agglomerate as a result of being partiallymolten and therefore somewhat sticky. Theytherefore remain more compact, and float outmore easily to form surface imperfections in theform of slag macroinclusions on the surface.For a few materials, particularly alloys based onthe Cu±10Al types (aluminium- and manganese-bronzes) the critical velocities were originallythought to be much lower, in the region of only0.075 m sÿ1. However, from recent work atBirmingham this low velocity seems to havebeen a mistake, probably resulting from theconfusion caused by bubbles entrained in theearly part of the filling system. With well-designed filling systems, the aluminium-bronzesaccurately fulfil the theoretically predicted0.4 m sÿ1 value for a critical ingate velocity(Halvaee and Campbell 1997).

Because of the central importance of theconcept of critical velocity, the reader will

forgive a re-statement of some aspects in thissummary.

(i) Even if the melt does jump higher than theheight of a sessile drop, when it falls backinto the surface there is no certainty that itwill enfold its surface film. These tumblingmotions in the liquid can be chaotic,random events. Sometimes the surface willfold badly, and sometimes not at all. This isthe character of surface turbulence; it is notpredictable in detail. The key aspect of thecritical velocity is that at velocities less thanthe critical velocity the surface is safe.Above the critical velocity there is thedanger of entrainment damage. The criter-ion is a necessary but not sufficient condi-tion for entrainment damage.

(ii) If the whole, extensive surface of a liquidwere moving upwards at a uniform speed,but exceeding the critical velocity, clearlyno entrainment would occur. Thus thesurface disturbance that can lead to en-trainment is more accurately described notmerely as a velocity but in reality a velocitydifference. It might therefore be definedmore accurately as a critical velocity gra-dient measured across the liquid surface.For those of a theoretical bent, the criticalgradient might be defined as the velocitydifference achieving the critical velocityalong a distance in the surface of the orderof the sessile drop radius (approximatelyhalf its height) in the liquid surface. Toachieve reasonable accuracy, this approachrequires one to allow for the reduction indrop height with velocity. Hirt (2003) solvesthis problem with a delightful and novelapproach, modelling the surface distur-bances as arrays of turbulent eddies, andachieves convincing solutions for the simu-lation of entrainment at hydraulic jumpsand plunging jets. Such niceties are neg-lected here. The problem does not arisewhen considering the velocity of the meltwhen emerging from a vertical ingate into amould cavity. In that situation, the ingatevelocity and its relation to the criticalvelocity is clear.

(iii) If the melt is travelling at a high speed, butis constrained between narrowly enclosingwalls, it does not have the room to fold-overits advancing meniscus. Thus no damage issuffered by the liquid despite its high speed,and despite the high risk involved. This isone of the basic reasons underlying thedesign of extremely narrow channels forfilling systems that are proposed in thisbook.



2.2 The `no fall' requirement

It is quickly shown that if liquid aluminium isallowed to fall more than 12.5 mm then itexceeds the critical 0.5 m sÿ1. The critical fallheight can be seen to be a kind of re-statementof the critical velocity condition. Similar criticalvelocities and critical fall heights can be definedfor other liquid metals. The critical fall heightsfor all liquid metals are in the range 3 to15 mm.

It follows immediately that top gating ofcastings almost without exception will lead to aviolation of the critical velocity requirement.Many forms of gating that enter the mouldcavity at the mould joint, if any significant partof the cavity is below the joint, will also violatethis requirement.

In fact, for conventional sand and gravity diecasting, it has to be accepted that some fall ofthe metal is necessary. Thus it has been acceptedthat the best option is for a single fall, con-centrating the loss of height of the liquid at thevery beginning of the filling system. The falltakes place down a conduit known as a sprue, ordown runner. This conduit brings the melt tothe lowest point of the mould. The distributionsystem from that point, consisting of runnersand gates, should progress only uphill.

Considering the mould cavity itself, therequirement effectively rules that all gates intothe mould cavity enter at the bottom level,known as bottom gating. The siting of gates intothe mould cavity at the top (top gating) or at thejoint (gating at the joint line) are not options ifsafety from surface turbulence is required.

Also excluded are any filling methods thatcause waterfall effects in the mould cavity. Thisrequirement dictates the siting of a separateingate at every isolated low point on the casting.

Even so, the concept of the critical fall dis-tance does require some qualification. If thecritical limit is exceeded it does not mean thatdefects will necessarily occur. It simply meansthat there is a risk that they may occur. This isbecause the energy of the liquid is now suffi-ciently high that the melt is potentially able toenfold in its own surface. Whether a defectoccurs or not is now a matter of chance. (Thiscontrasts, of course, with falls of less than thecritical height. In this case there is no chancethat a defect can occur, the regime being com-pletely safe.)

There is, however, further qualification thatneeds to be applied to the critical fall distance.This is because the critical value quoted abovehas been worked out for a liquid, neglecting thepresence of any oxide film. In practice, it seemsthat for some liquid alloys, the surface oxide has

a certain amount of strength and rigidity, sothat the falling stream is contained in its oxidetube and so is enabled to better resist the con-ditions that might enfold its surface. Thisbehaviour has been investigated for aluminiumalloys (Din, Kendrick and Campbell 2003). Itseems that although the original fall distancelimit of 12.5 mm continues to be the safestoption, fall heights of up to about 100 mm mightbe allowable in some instances, possiblydepending somewhat on the precise alloy com-position. However, falls greater than 200 mmdefinitely entrain defects; the velocity of the meltin this case is about 2 m sÿ1 so that entrainmentseems unavoidable. Also, of course, other alloysmay not enjoy the benefits of the support of atube of oxide around the falling jet. This benefitrequires to be investigated in other alloy systemsto test what values beyond the theoretical limitsmay be used in practice.

The initial fall down the sprue in gravity-filled systems does necessarily introduce someoxide damage into the metal. For this reason itseems reasonable to conclude that gravity-poured castings will never attain the degree ofreliability that can be provided by counter-gravity and other systems that can avoid surfaceturbulence.

Of necessity therefore, it has to be acceptedthat the no-fall requirement applies to thedesign of the filling system downstream of thebase of the sprue. The damage encountered inthe fall down the sprue has to be accepted;although with a good sprue and pouring basindesign this initial fall damage can be reduced toa minimum as we shall see.

It is a matter of good luck that it seems thatfor some alloys much of the oxide introduced inthis way does not appear to find its way throughand into the mould cavity. It seems that much ofit remains attached to the walls of the sprue.This surprising effect is clearly seen in many top-gated castings, where most of the oxide damage(and particularly any random leakage problem)is confined to the area of the casting under thepoint of pouring, where the metal is falling.Extensive damage does not seem to extend intothose regions of the casting where the speed ofthe metal front decreases, and where the fronttravels uphill, but there does appear to be somecarry-over of defects. Thus the provision of afilter immediately after the completion of thefall is valuable. It is to be noted, however, thatsignificant damage will still be expected to passthrough the filter.

The requirement that the filling systemshould cause the melt to progress only uphillafter the base of the sprue forces the decisionthat the runner must be in the drag and the gates



must be in the cope for a horizontal singlejointed mould (if the runner is in the cope, thenthe gates fill prematurely, before the runneritself is filled, thus air bubbles are likely to enterthe gates). The `no fall' requirement may alsoexclude some of those filling methods in whichthe metal slides down a face inside the mouldcavity, such as some tilt casting type operations.This undesirable effect is discussed in moredetail in the section devoted to tilt casting.

It is noteworthy that these precautions toavoid the entrainment of oxide films also applyto casting in inert gas or even in vacuum. This isbecause the oxides of Al and Mg (as in Al alloys,ductile irons, or high temperature Ni-base alloysfor instance) form so readily that they effectively`getter' the residual oxygen in any conventionalindustrial vacuum, and form strong films on thesurface of the liquid.

Rule 2 applies to `normal' castings with wallsof thickness over 3 or 4 mm.

For channels that are sufficiently narrow,having dimensions of only a few millimetres, thecurvature of the meniscus at the liquid front cankeep the liquid front from disintegration. Thusnarrow filling system geometries are valuable intheir action to conserve the liquid as a coherentmass, and so acting to push the air out of thesystem ahead of the liquid. The filling systemstherefore fill in one pass.

A good filling action, pushing the air aheadof the liquid front as a piston in a cylinder, is acritically valuable action. Such systems deservea special name such as perhaps `one pass filling(OPF) designs'. Although I do not usually carefor such jargon, the special name emphasizes thespecial action. It contrasts with the turbulentand scattered filling often observed in systemsthat are over-generously designed, in which themelt can be travelling in two directions at oncealong a single channel. A fast jet travels underthe return wave that rolls over its top, rolling inair and oxides.

For a wide, narrow, horizontal channel, anyeffect of surface tension is clearly limited tochannels that have dimensions smaller than thesessile drop height for that alloy. Thus for Alalloys the maximum channel height would be12.5 mm, although even this height would exertlittle influence on the melt, since the roof wouldjust touch the liquid, exerting no pressure on it.Similarly, taking account that the effect of sur-face tension is doubled if the curvature of theliquid front is doubled by a second componentof the curvature at right angles, a channel ofsquare section could be 25 mm square, and becontained just by surface tension. In practice,however, for any useful restraint from the wallsof channels, these dimensions require to be at

least halved, effectively compressing the liquidinto the channel.

For very thin walled castings, of sectionthickness less than 2 mm, the effect of surfacetension in controlling filling becomes pre-dominant. The walls are so much closer than thenatural curvature of a sessile drop that themeniscus is effectively compressed, and requiresthe application of pressure to force it into suchnarrow gaps. The liquid surface is now so con-strained that it is not easy to break the surface,i.e. once again there is no room for splashing ordroplet formation. Thus the critical velocity ishigher, and metal speeds can be raised byapproximately a factor of 2 without danger.

In very thin walled castings, with walls lessthan 2 mm thickness, the tight curvature of themeniscus becomes so important that fillingcan sometimes be without regard to gravity (i.e.can be uphill or downhill) since the effect ofgravity is swamped by the effect of surface ten-sion. This makes even the uphill filling of suchthin sections problematical, because the effec-tive surface tension exceeds the effect of gravity.Instabilities therefore occur, whereby the mov-ing parts of the meniscus continue to moveahead in spite of gravity because of the reducedthickness of the oxide skin at that point. Con-versely, other parts of the meniscus that dragback are further suppressed in their advanceby the thickening oxide, so that a run-awayinstability condition occurs. This dendriticadvance of the liquid front is no longer con-trolled by gravity in very thin castings, makingthe filling of extensive sections, whether hori-zontal or vertical, a major filling problem.

The problem of the filling of thin walls occursbecause the flow happens, by chance, to avoidfilling some areas because of random mean-dering. Such chance avoidance, if prolonged,leads to the development of strong oxide films,or even freezing of the liquid front. Thus thefinal advance of the liquid to fill such regions ishindered or prevented altogether.

The dangers of a random filling patternproblem are relieved by the presence of regularlyspaced ribs or other geometrical features thatassist organizing the distribution of liquid.Random meandering is thereby discouragedand replaced by regular and frequent penetra-tion of the area, so that the liquid front has abetter chance to remain `live', i.e. it keepsmoving so that a thick restraining oxide is givenless chance to form.

The further complicating effect of themicroscopic break-up of the front known asmicro-jetting (Castings 2003) observed in sec-tions of 2 mm and less in sand and plastermoulds is not yet understood. The effect has not



yet been investigated, and may not occur at allin the dry mould conditions such as are found ingravity die casting.

2.3 Filling system design

Getting the liquid metal out of the crucible ormelting furnace and into the mould is a criticalstep when making a casting: it is likely that mostcasting scrap arises during this few seconds ofpouring of the casting.

Recent work observing the liquid metal as ittravels through the filling system indicates thatmost of the damage is done to castings by poorfilling system design. It is also worth reflectingon the fact that every gram of metal in thecasting has, of necessity, travelled through thefilling system. Leaving its design to chance, oreven to the patternmaker (with all due respect toall our invaluable and highly skilful pattern-makers), is a risk not to be recommended.

The early part of this section presents thedesign background, outlining the generalthinking and some of the detailed logic behindthe design of filling systems. The detailed cal-culations that are required to determine theprecise dimensions of the various parts of thesystem are presented later (Section 2.3.7).

2.3.1 Gravity pouring of open-top moulds

Most castings require a mould to be formed intwo parts: the bottom part (the drag) forms thebase of the casting, and the top half (the cope)forms the top of the casting. However, somecastings require no shaping of the top surface.In this case only a drag is required. The absenceof a cope means that the mould cavity is open,so that metal can be poured directly in. Thefoundryman can therefore direct the flow ofmetal around the mould using his skill duringpouring (Figure 2.2). Such open-top mouldsrepresent a successful and economical technique

Figure 2.2 (a) An open and(b) closed mould partially sectioned.



for the production of aluminium or bronze wallplaques and plates in cast iron, which do notrequire a well-formed back surface. The firstgreat engineering structure, the Iron Bridgebuilt across the River Severn by the greatEnglish ironmaster Abraham Darby in 1779, hadall its main spars cast in this way. This spectac-ular feat, with its main structural members over23 m long cast in open-top sand moulds, her-alded the dawn of the modern concept of astructural engineering casting.

Other viscous and poorly fluid materials arecast similarly, such as hydraulic cements, con-cretes, and organic resins and resin/aggregatemixtures that constitute resin concretes. Moltenceramics such as liquid basalt are poured in thesame way, as witnessed by the cast basalt curbstones outside the house where I once lived, thathave lined the edge of the road for the lasthundred years or so and whose maker's name isstill as sharply defined as the day it was cast.

The remainder of this section concentrates onthe complex problem of designing filling systemsfor castings in which all the surfaces are moul-ded, i.e. the mould is closed. In all such cir-cumstances, a bottom-gated system is adopted(i.e. the melt enters the mould cavity from oneor more gates located at the lowest point, or ifmore than one low point, at each lowest point).

2.3.2 Gravity pouring of closed moulds

The series of funnels, pipes and channels toguide the metal from the ladle into the mouldconstitutes our liquid metal plumbing, and isknown as the filling system, or running system.Its design is crucial; so crucial, that this iswithout doubt the most important chapter inthe book.

However, the reader needs to keep in mindthat the elimination of a running system bysimply pouring into the top of the mould (downan open feeder, for instance) may be a reason-able solution in certain cases. Although appar-ently counter to much of the teaching in thisbook, there is no doubt that a top-pouredoption has often been demonstrated to be pre-ferable to some poorly designed running sys-tems, especially poorly designed bottom-gatedsystems. There are fundamental reasons for thisthat are worth examining right away.

In top gating the plunge of a jet into a liquidis accompanied by relatively low shear forces inthe liquid, since the liquid surrounding the jetwill move with the jet, reducing the shearingaction. Thus although some damage is alwaysdone by top pouring, in some circumstances itmay not be too bad, and may be preferable to acostly, difficult, or poor bottom-gated system.

In poor filling system designs, velocities inthe channels can be significantly higher than thefree-fall velocities. What is worse, the walls ofthe channels are stationary, and so maximize theshearing action, encouraging surface turbulenceand the consequential damage from the shred-ding and entraining of bubbles and bifilms.

Ultimately, however, a bottom-gated system,if designed well, has the greatest potential forsuccess.

Most castings are made by pouring the liquidmetal into the opening of the running system,using the action of gravity to effect the fillingaction of the mould. This is a simple and quickway to make a casting. Thus gravity sand castingand gravity die-casting (permanent mould cast-ing in the USA) are important casting processesat the present time. Gravity castings have,however, gained a poor reputation for reliabilityand quality, simply because their running systemshave in general been badly designed. Surfaceturbulence has led to porosity and cracks, andunreliability in leak-tightness and mechanicalproperties.

Nevertheless, there are rules for the design ofgravity-running systems that, although admit-tedly far from perfect, are much better thannothing. Such rules were originally empirical,based on transparent-model work and someconfirmatory tests on real castings. We are nowa little better informed by access to real-timevideo radiography of moulds during filling, andsophisticated computer simulation, so thatliquid aluminium or liquid steel can be observedas it tumbles through the mould. Despite this,many uncertainties still remain. The rules for thedesign of filling systems are still not the maturescience that we all might wish for. Even so,some rules are now evident, and their intelligentuse allows castings of the highest quality tobe made. They are therefore described in thissection, and constitute essential reading!

It is hoped to answer the questions `Why isthe running system so complicated?' and `Whyare there so many different features?' It is asalutary fact that the apparent complexity hasled to much confused thinking.

An invaluable general rule that I recommendto all those studying running and gating systemsis `If in doubt, visualize water'. Most of us haveclear perceptions about the mobility and generalflow behaviour of water in the gentle pouring ofa cup of tea, the splat as it is spilled on the floor,the flow of a river over a weir, or the spray froma high-pressure hose pipe. A general feeling forthis behaviour can sometimes allow us to cutthrough the mystique, and sometimes even thecalculations! In addition, the application of thissimple criterion can often result in the instant



dismissal of many existing filling systemsintended for the production of a reliable qualityof casting as being quite clearly useless!

Closed moulds represent the greatest chal-lenge to the casting engineer. There are numer-ous ways to get the metal into the mould, somedisastrously bad, some tolerable, some good. Toappreciate the good we shall have to devotesome space to the bad (Figure 2.3). If this readslike a sermon, then so be it. The Good RunningSystem is a Good Cause that deserves the pas-sionate concern of the casting engineer. Toomany castings with hastily rigged running sys-tems have appeared to be satisfactory in limitedprototype trials, but have proved to have dis-astrous levels of scrap when put into produc-tion. This is normally the result of surfaceturbulence during filling that produces non-reproducible castings, some apparently good,some definitely bad. This result confirms thenature of turbulence. Turbulence implies chaos;and chaos implies unpredictability. When usinga running system that generates surface turbu-lence a typical scrap rate for a commercialvehicle casting might be 15 per cent, whereasa turbine blade subjected to much more strin-gent inspection can easily reach 75 per centrejections.

In general, however, experience shows thatfoundries that use exclusively turbulent fillingmethods such as most investment foundries,experience on average about 20±25 per centscrap, of which 5±10 per cent is the total ofmiscellaneous minor processing problems suchas broken moulds, castings damaged during cut-off, etc. The remaining 15 per cent is composedof random inclusions, random porosity, andmisrunsÐthe standard legacy of turbulent run-ning systems: the inclusions are created by thefolding of the surface, as are the random pock-ets of porosity; and the misruns by the unpre-dictable ebb and flow in different parts of thecasting during filling. In sand casting foundries,most of the so-called mould problems leading tosand inclusion are actually the result of the poorfilling system designs. With good filling systemssand problems such as mould erosion and sandinclusions usually disappear.

In a foundry making a variety of castings,the 15 per cent running system scrap is madeup of difficult castings which might run at85±95 per cent scrap (almost never 100 per cent!)and easy castings which run at 5 per cent scrap(almost never zero!). The non-repeatable resultscontinuously raise the characteristic false hopethat the problems are solved, only to have thehopes dashed again by the next few castings.The variability is baffling, because the foundryengineer will often go to extreme lengths to

ensure that all the variables believed to be undercontrol are held constant.

Only a carefully worked out running systemwill give filling that is characterized by lowsurface turbulence, and which is thereforereproducible every time. Interestingly, this canmean 100 per cent scrap. However, this is notsuch a bad result in practice because the defectwill be reproducibly repeated in every casting. Itis therefore easy to identify and correct, andwhen corrected, stays corrected. After the firsttrials, the good running system should yieldreliable, repeatable castings, and be character-ized by a scrap rate close to zero.

A good running system, perhaps somethinglike that shown in Figure 2.3b, will also betolerant of wide variations in foundry practice,in contrast with the normal experienceaccompanying turbulent filling, in whichpouring conditions are critical. Many foundrieswill know the problem that certain castings canonly be poured successfully by certain opera-tors. The good running system will ensure thatpouring speed will now be under the control ofthe running system, not the pourer, and castingtemperature will no longer be dictated by theavoidance of misruns, but can be set indepen-dently to control grain size without the addi-tion of grain refiners. It is clear, therefore, thata good running system is a good ally inthe creation of economical products of highquality.

The elements of a good system are:

1. Economy of size. A lightweight system willincrease yield (the ratio of finished castingweight to total cast weight), allowing thefoundry to make more castings from theexisting melt supply. It may also help to getmore castings into a given mould size.This has a big effect on productivity andeconomy.

2. The filling of the mould at the required speed.In the method proposed in this book, thewhole running system is designed so thatthe velocity of the metal in the gates is belowthe critical value. This value varies from onealloy to another, but is generally close to0.5 m s

ÿ1. There is now much experimentaland theoretical data to support this value(Runyoro 1992). Data on the density ofcastings produced by gating uphill haveshown that air entrapment can occur aboveapproximately 0.5 m sÿ1 (Suzuki 1989). Incomputer simulations of flow, Lin andHwang (1988) show that when liquid alumi-nium enters the mould horizontally at1.1 m sÿ1 it hits the far wall with suchforce that the reflected wave breaks, causing



Figure 2.3 (a) Poor top gates and side-fed running system, compared with (b) a more satisfactory bottom-gated andtop-fed system (c) poor system gated at joint and (d) recommended economical and effective system.


FillingSystemConical

basin

(c) (d)

Reversetapersprue

Correctlytaperedsprue

Mouldcavity

in cope

Feeder

Stopper

Offset stepbasin

Runner exitsdirectly in cavity

Fall of metalinside mould cavity

Runnerin drag

Gatesin cope

Well

Drag

Cope

FeedingSystem

Conicalcup

Side feeder

Gatesin drag

Castingin drag

Castingin cope

Feeder athighest point

Offsetstepbasin

Correcttapersprue

Nowell

Taperrunner

Gates in cope

Reversetapersprue

Runnerin cope

(a)

(b)

�� '�� ; �� ,( *��


surface turbulence. These figures confirmthe safety of 0.5 m sÿ1, and the danger ofexceeding 1 m sÿ1.

3. The delivery of only liquid metal into themould cavity, i.e. not other phases such asslag, oxide, and sand. However, in most casesthe overwhelmingly common and unwelcomephase is air (probably contaminated withother mould gases of course). The design offilling systems to achieve the exclusion ofair will constitute a major preoccupation inthis book.

4. The elimination of surface turbulence, pre-ferably at an early stage in the runner system,but certainly by the time that the metalarrives in the mould cavity. The problemhere is that by the time the metal has fallenthe length of the sprue to reach the lowestlevel of the casting, its velocity is well abovethe critical velocity for surface turbulence.Despite this danger, the running systemshould, so far as possible, prevent the result-ing fragmentation of the stream. Any frag-mentation will result in permanent damage tothe casting in most alloys. However, iffragmentation occurs, the best that can nowhappen is that it should be followed by anaction to gather the stream together again. Inthis way the melt enters the mould as acoherent, compact spreading front, prefer-ably at a velocity sufficiently low that thedanger of any further break-up of the front iseliminated.

5. Ease of removal. Preferably the systemshould break off. As a next best option, itshould be removable with a single stroke of aclipping press, or a straight cut. Curved cutstake more time and are more difficult to dressto finished size by grinding or linishing.Internal or shielded gates may need to bemachined off, in which case the expense ofsetting up the casting for machining might be

avoidable by carrying out this task later,during the general machining of the casting.

(Note that in general practice it is usually best toassume that there is no requirement for thefilling system to act as a feeder, i.e. to compen-sate for the contraction on solidification. Weshould ensure that the feeding function ifnecessary at all, is carried out by a separatefeeder placed elsewhere, preferably high up, onthe casting (Figure 2.3b). In some cases it ispossible to use a running system that can alsoact as a feeder. These special systems should beused whenever possible. They are considered inChapter 6. It is worth noting that in investmentcasting the almost universal confusion betweenfilling and feeding systems is deeply regrettable.In this book the two functions are treated totallyseparately.)

Because the above list of criteria have been sodifficult to meet in practice, there has been amove away from gravity casting as a result ofwhat have been believed to be insoluble barriersto the attainment of high quality and reliability.Uphill filling, against gravity, known as coun-ter-gravity casting (and, more colloquially andless helpfully, as low-pressure casting), hasprovided a solution to the elimination of surfaceturbulence. It has seemed to be the ultimatedevelopment of bottom gating (Figure 2.4). Thisdevelopment has therefore provided the impetusfor the growth of low-pressure die casting, low-pressure sand casting, and various forms ofcounter-gravity filling of investment castings. Aform of high-pressure die-casting has also beendeveloped to take advantage of the qualitybenefits associated with counter-gravity fillingfollowed by high-pressure consolidation. Thesedifferent techniques of getting the metal into themould will all be discussed later.

However, although counter-gravity fillingfulfils all the above requirements, our main aim

Figure 2.4 Various direct gatingsystems applied to a box shapedcasting. Possible filter locations areshown as dashed outlines. Note thatall of the gravity systems shown hereare poor: the sprue base connectsdirectly with the ingate into thecasting. All need mechanisms (notshown for clarity) to reduce thevelocity of the melt.



in this section is to evaluate gravity filling, to seehow far it can meet this difficult set of criteria.

Requirement 3 for good gating is important:only liquid metal should enter the casting. Thusall bubbles entrained by the surface turbulencecharacterizing the early part of the runningsystem should have been eliminated by thisstage. If the running system is poor, and bubblesare still present, their rise and bursting at theliquid surface in the mould violates Rule 4. Thisviolation results in a number of problems,including bubble trails, splash defects, and theretention of the scattering of smaller bubblesthat remain trapped under the oxide skin of therising metal. These cause concentrations ofmedium-sized pores (0.5±5 mm diameter) atspecific locations in the casting, usually at uppersurfaces of the casting above the ingates.

The other point in Requirement 3, that drossor slag does not enter the mould cavity isinteresting. In the production of iron castings itis normal for the runner to be placed in the copeand the gates in the drag, as is illustrated inFigure 2.3a. The thinking behind this design ofsystem is that slag will float to the top of therunner, and thus will not enter the gates. Suchthinking is at fault because it is clear that at leastsome of the first metal to enter the runner willfall down the first gate that it meets, taking withit not only the first slag but also air. This pre-mature delivery of metal into the mould beforethe runner is full is clearly unsatisfactory. Themetal has had insufficient time to settle down, toorganize itself free from dross, oxide and bub-bles. The fact that such systems are widely used,and are found in practice to reduce bubbledefects in the casting actually reveals how poorthe front end of the running system is. Clearly,bubbles are being generated throughout thepour, so the off-take of gates at the base of therunner is valuable in this case.

A more satisfactory system is illustrated inFigure 2.3b. Here the runner is in the drag andthe gates in the cope. In this system the runnerhas to fill first before the gates are reached. Thusthe metal has a short but valuable time to riditself of bubbles and dross, most of which can betrapped in the dross trap or against the uppersurface of the runner. Only a limited amount ofslag or dross will be unfortunately placed toenter the gate. Provided the velocity of the metalin the gate is not too high, even this slag still hasa good chance of being held against the ceilingof the gate, and thus not entering the casting.Figure 2.3d illustrates an optimum system(contrasting with 2.3c), designed to resist theentrainment of air at all stages of the system.

Statement 4 is deceptively simple. However,the requirement of no surface turbulence is so

important, and so central to the quest for goodcastings, that we have to consider it at length.

Texts elsewhere often refer to turbulence-freefilling as laminar filling. The implication here isthat turbulence as defined by Reynold's numberis involved, and that the desirable criterion isthat of laminar flow of the bulk. As discussed inCastings 2003, it is not bulk turbulence that isrelevant since turbulent flow in the bulk liquidcan still be accompanied by the desirablesmooth flow of the surface. Our attentionrequires to be concentrated on the behaviour ofthe liquid surface. Thus provided we ensure thatby `laminar fill' we mean `surface laminar fill',then we shall have our concepts correct, and ourthinking accurate.

Requirement 4 above is clearly violatedby splashing during filling. It can be seenimmediately that top gating will probablytherefore always introduce some defects (theexception is very thin wall castings where surfacetension takes over control of surface turbulence).Figure 2.4 illustrates a poor running systemwhere the metal enters from top or side gatesthat allow the metal to suffer a free fall intothe mould cavity. Bottom-gated systems arealways required if surface turbulence is to beeliminated.

However, although bottom gating is neces-sary, it is not a sufficient criterion. It is easy todesign a bad bottom-gating system! In fact, it ispossible to state the case more forcefully: a badbottom-gated system is usually worse than mosttop-gated systems.

For instance, it is common to see bottom-gated systems proudly displayed with the baseof the runner turned so that metal directlyenters the mould (Figure 2.5). Such systems are

Figure 2.5 A poor filling system because of direct entryof high velocity metal into the mould cavity.



compact, and appear economical until the per-centage scrap figures are inspected. The sequenceof events is clear if we consider the fall of thefirst liquid down the length of the sprue. The highvelocity of the metal on its impact at its base isnot contained. The resulting splash may belikened to an explosion of high-velocity drops orjets fired like projectiles directly into the mould.The bulk of the metal follows in an untidyfashion, mixed with air and mould gases, andricochets from the far wall, causing more surfaceturbulence as the rebounding wave breaks,rolling over and entraining yet more surface andmore gas. The elimination of the entrainedbubbles by bursting as they rise to the surface ofthe melt causes additional droplets to be createdby splashing. It is important, therefore, todesign the down-runner with care so that it willfill quickly, excluding air as quickly as possible,and to design the runner and gate to constrainthe metal, avoiding any provision of room forsplashing (Figure 2.6a). Further improvementsmight be allowable as in Figures 2.6b and 2.6c inwhich the fall heights down the sprue are pro-gressively reduced, reducing velocities in themould, by simply re-orienting the casting.

The base of the sprue should be the lowestpoint in the whole system: having reached here,all subsequent flow of the liquid should beuphill, displacing the air ahead in a controlledand progressive advance. So far as possible, theliquid should be slowed as it goes, experiencingas much opportunity as possible to becomequiescent before entering the mould. It shouldfinally enter the mould at a velocity less than itscritical velocity for the entrainment of defects.In this way a good and reproducible casting isfavoured.

2.3.2.1 Pressurized versus unpressurized

In the book Castings 1991 the author recom-mended the achievement of velocity reductionby the progressive enlargement of the area of theflow channels at each stage, with the aim ofprogressively reducing the rate of flow. This isknown as an unpressurized running system. Theaim was to ensure that the gate was of a suffi-cient area to make a final reduction to the speedof the melt, so that it entered the mould at aspeed no greater than its critical velocity. Morerecent research, however, has demonstrated thatthe enlargement of the system, by, for instance,a factor of two as the flow emerges from the exitof the sprue and enters the runner, usually failsto fill the runner. Thus the unpressurized sys-tems unfortunately behaved poorly, entrainingbubbles and oxides, because much of thesystem runs only partly full. The other standard

criticism (but incidentally of much less impor-tance) was that unpressurized systems are heavy,thus reducing metallic yield, and thus costly.

In fact, video radiography reveals that at theabrupt increase in cross-section at the base ofthe sprue on the entry to the runner, theentrainment of air occurs with dramatic effec-tiveness. This is because the melt jets along thebase of the runner (not filling the additionalarea provided) and hits the end of the runner.

Figure 2.6 (a) An improved bottom-gated system;(b) and (c) further improved by height reductions.


(a)

(b)

(c)


The reflected back wave rolls over the under-lying fast jet, rolling in oxides and bubbles at theinterface between the two (Figure 2.7a). Theeffect can be long-lived, developing into a stablehydraulic jump. The bubbles travel along theinterface between the two opposing streams(probably because of the presence of two non-wetting oxide films separating the two flowingstreams) and progress to the ingate, usually

collecting in a low pressure zone on one side ofthe ingate, before proceeding to swim upthrough the metal in the mould cavity (2.7e).Naturally, these bubbles and oxides bequeathserious permanent damage to the casting (2.7f).

The cast iron foundryman had some justifi-cation therefore to champion his own favouritepressurized systems. For the benefit of thereader, the so-called pressurized running system

(a) (b)

(c) (d)

(e) (f)

.�� # + "�� /� �� 1�� . �� /� �� 1�� /� �� 7�� 5&&�� ,&&�� ,&�� /��. �� '

�� '�� ; �� ,( *��


is one in which the metal flow is choked (i.e.limited by constriction) at the gate; i.e. its rate offlow into the mould is controlled by the area ofthe gate, the last point in the running system(Figure 2.7b). This causes the running system toback-fill from this point, and become pressur-ized with liquid, forcing the system to fill andexclude air. Thus the system entrains fewerbubbles and oxides. However, it also forced themetal into the mould as a jet. Clearly this systemviolates one of our principal rules, since themetal is now entering the mould above its crit-ical speed. The resulting splashing and otherforms of surface turbulence inside the mouldintroduces its own spectrum of problems, dif-ferent from those of the unpressurized system,but usually harming both the quality of themould and the casting.

Thus neither the unpressurized nor the pres-surized traditional systems are seen to worksatisfactorily. This is a regrettable appraisal ofpresent casting technology.

Because for many years the pressurized sys-tems were mainly used for cast iron, there werespecial reasons why the systems appeared to beadequate:

1. In the days of pouring grey iron into green-sand moulds the problems of surface turbu-lence were minimized by the tolerance of themetal±mould system. The oxidizing environ-ment in the greensand mould produced aliquid silicate film on the surface of the liquidiron. Thus when this was turbulently en-trained it did not lead to a permanent defect(Castings (2003)). In fact, many good cast-ings were produced by tipping the metal intothe top of the mould, using no running systemat all! Nowadays, with the use of certain corebinders and mould additives that cause solidgraphitic surface films on the metal, andconsequently reduce its tolerance to surfaceturbulence, the pressurized systems are pro-ducing defects where once they were workingsatisfactorily. This problem has become moreacute as it has become increasingly commonfor irons to have alloy additions such asmagnesium (to make ductile iron) and chro-mium (for many alloyed irons).

2. Over recent years the standards required ofcastings have risen to an extent that thetraditional foundryman is shocked anddazed. Whereas the pressurized system wasat one time satisfactory, it now needs to bereviewed. The achievement of quality is nowbeing seen to be not by inspection, but byprocess control. Turbulence during fillingintroduces a factor that will never be pre-dictable or controllable. This ultimately will

be seen as unacceptable. Reproducibility ofthe casting process will be guaranteed only bysystems that fill the mould cavity withlaminar surface flow. At one time this wasachievable only with counter-gravity fillingsystems. Nowadays, as we shall see, we canachieve some success with gravity systems,provided they are designed correctly.

The conclusion given by the author in Castings(1991) was `Unpressurized systems are recom-mended therefore. Pressurized are not.' Thisbold statement now requires revision in the lightof recent research since we now find that neithersystem is really satisfactory.

In summary, the unpressurized system hadthe praiseworthy aim to reduce the gate velocityto below the critical velocity. Unfortunatelysuch systems usually run only part-full, causingdamage to the castings because of entrained airbubbles and oxides. The pressurized systemprobably benefited greatly from its ability to fillquickly and to run full, greatly reducing thedamage from bubbles. However, the highvelocity of the melt as it jetted into the mouldcreated its own contribution to havoc.

Turning now to another sacred cow of run-ning system design that requires to be addres-sed. This is the concept of a choke. The choke isa local constriction designed to limit flow. In thenon-pressurized system the choke was generallyat the base of the sprue, whereas the pressurizedsystem was choked at the ingates into themould. Unfortunately, a choke is an undesirablefeature. Flow rates are usually sufficiently highthat the melt will be speeded up through aconstriction and emerge as a jet, entraining aironce again downstream, with much con-sequential damage.

All these systems were devised before thebenefits of computer simulation and videoX-ray radiography. They also pre-dated thedevelopment of the concepts of surface turbu-lence, critical velocity, critical fall height andbifilms. It is not surprising therefore that allthese traditional approaches to the design offilling systems gave less than satisfactory results.

In the history of the development of fillingsystems most of the early work was of limitedvalue because the emphasis was on steady stateflow through fully filled pipework, following theprinciples of hydraulics. This does, of course,sometimes occur late during the filling process.However, the real problems of filling are asso-ciated with the priming of the filling system, i.e.its behaviour before the filling system is filled.Thus these early studies give us relatively littleuseful background on which to base effectivedesigns for real castings.



A completely new approach is described inthis book that attempts to address these issues.We shall abandon the concept of a localizedchoke. The whole of the length of the fillingsystem should experience its walls in permanentcontact and gently pressurized by the liquidmetal. Thus, effectively, the whole length of therunning system should be designed to act like achoke; a kind of continuous choke principle. Inall probability, it seems that we really needuniformly pressurized systems. An alternativedescription might be `naturally pressurized' sys-tems, because the new design concept is basedon designing the flow channels in the mould soas to follow the natural form that the flowingmetal wishes to take.

For instance, at the base of the sprue we candefine its area as unity. After the right anglebend into the runner, if the stream loses energyso that its velocity falls by 20 per cent, we canexpand the channel by this amount. The run-ner can remain at this area of 1.2 along itslength. After turning through a further rightangle bend into the gate this gives a series ofpermissible area ratios of 1 : 1.2 : 1.4, althoughit will be noticed that the ingate velocity hasonly fallen by approximately 40 per cent fromthat at the sprue exit.

If the 20 per cent expansion of area after eachbend is not entirely allowed (for instance if only10 per cent expansion were provided) the streamwill experience a gentle pressurization. Thismodest pressure against the walls of the runningsystem will be valuable to counter any effect ofbubble formation and will act to support thewalls of the running system against collapse (aspecial problem in large running systems forlarge castings). Thus to be more sure of main-taining the system completely full, and slightlypressurized, a ratio of 1 : 1.1 : 1.2 or even 1 : 1 : 1might be used.

Examples of area ratios are shown inTable 2.1.

From the ratios it is clear that the naturally (orslightly) pressurized system is part-way betweenthe pressurized and unpressurized systems.

However, there is a major problem with theuse of these new systems that the reader mayalready have noticed. The naturally pressurizedsystem has no built-in mechanism for any sig-nificant reduction in velocity of the stream.Thus the high velocity at the base of the sprue ismaintained (with only minor reduction) into themould. Thus the benefits of complete priming ofthe filling system to exclude air are lost onceagain on entering the mould cavity.

This fundamental problem alerts us to thefact that the naturally pressurized approachrequires completely separate mechanisms toreduce the velocity of the melt through theingates. The options include

(i) the use of filters;(ii) the provision of specially designed runner

extension systems such as flow-offs;(iii) a surge control system;(iv) the use of a vertical fan gate at the end of

the runner. Additional mechanisms mightbe possible in the future, when properlyresearched, such as

(v) the use of vortices to absorb energy whileavoiding significant surface turbulence.

We shall consider all these options in detail indue course, but the reader needs to be awarethat, unfortunately, at this time the use ofnaturally pressurized systems is in its infancy. Inparticular, the rules for such designs are not yetknown for some features such as joining roundor square section sprues to rectangular runners.Filters are not easily incorporated, nor arevortex systems fully understood.

This creates a familiar problem for thefoundry person: in the real world, the castingengineer has to take decisions on how to makethings, whether or not the information is avail-able at the time to help make the best choice.Thus insofar as the rules are presently under-stood for the majority of castings, they are setout below, for good or for bad. I hope theyassist the caster to achieve a good result. Oneday I hope we in the industry will all have thebetter answers that we need.

In the meantime, computers are starting tosimulate successfully the flow of metal in fillingsystems. At the present time such simulationsare highly computationally intensive, andtherefore slow and/or not particularly accurate.It is necessary to be aware that some simulationpackages are still highly inaccurate. However,time will improve this situation, to the greatbenefit of casting quality.

Table 2.1 Examples of area ratios (sprue exit area :runner area : gate area)

Examples ofarea ratios

Pressurized 1 : 0.8 : 0.61 : 1 : 0.8

Unpressurized 1 : 2 : 41 : 4 : 4

Natural 1 : 1.2 : 1.4Slightly pressurized 1 : 1 : 1

1 : 1.1 : 1.2With foam filter in gate 1 : 1 : 4With speed reduction or by-pass designs 1 : 1 : 10



2.3.2.2 Design of pouring basin

The conical basin

The in-line conical basin (Figure 2.8a), usedalmost everywhere in the casting industry,appears to be about as bad as could be envi-saged for most casting operations. It is probablyresponsible for the production of more castingscrap than any other single feature of the fillingsystem. It is not recommended.

The problems with the use of the conicalbasin arise as a result of a number of factors:

1. The metal enters at an unknown velocity,making the estimation of the design ofthe remainder of the running systemproblematical.

2. The metal enters at high, unchecked velocity.Since the main problem with running systemsis to reduce the velocity, this adds to thedifficulty of reducing surface turbulence.

3. Any contaminants such as dross or slag thatenter with the melt are necessarily takendirectly down the sprue.

4. The device works as an air pump, concen-trating air into the flow (the action isanalogous to other funnel-shaped pumps inwhich a fast stream of fluid directed down thecentre of the funnel is designed to entrain asecond surrounding fluid. Good examplesare steam ejectors and the vacuum suctiondevice that can be driven from a compressedair supply). Because air is probably the singlemost important contaminant in runningsystems, this is probably the most severedisadvantage, yet is not widely appreciated tobe a problem.

An example that the author has witnessedmany times can be quoted. Bottom-teemed

steel ingots were produced by a conventionalarrangement that consisted of pouring thesteel into a central conical cup, affixed to thetop of a spider distribution system of ceramictubes connected to the centre of the base of agroup of four or six surrounding ingotmoulds. Because the top of the ingot mouldremained open during filling, the upwellingcascade of air bubbles in the centre of therising metal was clear for all to see. (Thebottom fill technique was designed to deliveran improved surface condition of the ingot asa result of the gentle rolling action of the liquidmeniscus against the wall of the ingot mouldas the metal ascended. However, the overallcleanness of the ingot would have beensignificantly impaired by the passage of somuch air. It would have been useful to retainthe benefits of the bottom-teemed ladles andyet achieve improved castings by reducingthe entrainment of air into the system.)

5. The small volume of the basin makes itdifficult for the pourer to keep full (itsresponse time is too short, as explained later),so that air is automatically entrained as thebasin becomes partially empty from time totime during pouring. The pourer is usuallyunaware of this, since the aspiration of airusually takes place under the surface at thebasin/sprue junction.

6. The mould cavity fills differently depending onprecisely where in the basin the pourer directsthe pouring stream, whether at the far side ofthe cone, the centre, or the near side. Thus thecastings are intrinsically not reproducible.

7. This type of basin is most susceptible tothe formation of a vortex, because any slightoff-axis direction will tend to start a rotationof the pool. There has been much written

Figure 2.8 A rogues gallery of non-recommended scrap generating systems. Conical basin and sprue combinationsshowing (a) perhaps least damaging; (b) basin too large; (c) cup form; (d) basin too small; (e) enlarged sprue toact as a combined basin and sprue.


(a) (b) (c) (d) (e)


about the dire dangers of a vortex, and somebasins are provided with a flat side todiscourage its formation. In fact, however,this so-called disadvantage would only havesubstance if the vortex continued down thelength of the sprue, along the runner and intothe mould cavity. This is unlikely. Usually, avortex will `bottom out,' giving an air-freeflow into the remaining runner system as willbe discussed later. This imagined problem isalmost certainly the least of the difficultiesintroduced by the conical basin.

If this long list of faults was not already damningenough, it is made even worse for a variety ofreasons. A basin that is too large for the sprueentrance (Figure 2.8b) jets metal horizontally offthe exposed ledge formed by the top of themould, creating much turbulence and preventingthe filling of the sprue. The problem is unseen bythe caster, who, because he is keeping the basinfull, imagines he is doing a good job. Thecup shape of the basin (Figure 2.8c) is bad forthe same reason. The basin that is too small(Figure 2.8d) has painful memories for the writer:a casting with an otherwise excellent runningsystem was repeatedly wrecked by such a simpleoversight! Again, the caster thought he was doinga good job. However, the aspirated air causeda staggering amount of bubble damage in analuminium sump casting.

The expansion of the sprue entrance to act asa basin (Figure 2.8e) may hold the record for airentrainment (however the author has no plansto expend effort investigating this black claim).Worse still, the top of this awful device is usuallynot sufficiently wide that the pourer can fill itbecause it is too small to hit with the stream ofmetal without the danger of much metal spla-shed all over the top of the mould and sur-roundings. Thus this combined `basin/sprue'necessarily runs partially empty for most of thetime. Furthermore, the velocity of the melt isincreased as the jet is compressed into the nar-row exit from the sprue (this point is discussedin detail later). The elongated tapered basinsystem has been misguidedly chosen for its easeof moulding. There could hardly be a worse wayto introduce metal to the mould.

For very small castings weighing only a fewgrams, and where the sprue is only a few milli-metres diameter, there is a strong element ofcontrol of the filling of the sprue by surfacetension. For such small castings the conicalpouring cup probably works tolerably well. It issimple and economical, and, probably fills wellenough. This is as much good as can be saidabout the conical basin. Probably even this ispraising too highly.

Where the conical cup is filled with a handladle held just above the cone, the fall distanceof about 50 mm above the entrance to the sprueresults in a speed of entry into the sprue ofapproximately 1 m sÿ1. At such speeds the basinis probably least harmful. On the other hand,where the conical cup is used to funnel metalinto the running system when poured directlyfrom a furnace, or from many automatic pour-ing systems, the distance of fall is usually muchgreater, often 200 to 500 mm. In such situationsthe rate of entry of the metal into the system isprobably several metres per second. From thebottom-poured ladles in steel foundries themetal head is usually over 1 m giving an entryvelocity of 5 m sÿ1. This situation highlights oneof the drawbacks of the conical pouring basin; itcontains no mechanism to control the speed ofentry of liquid.

The pouring cup needs to be kept full ofmetal during the whole duration of the pour. Ifit is allowed to empty at any stage then air anddross will enter the system. Many castings havebeen spoiled by a slow pour, where the pouringis carried out too slowly, allowing the stream todribble down the sprue, or simply poured downthe centre without touching the sides of thesprue, and without filling the basin at all (whichis the trouble with the expanded sprue type).Alternatively, harm can be done by inattention,so that the pour is interrupted, allowing thebush to empty and air to enter the down-runnerbefore pouring is restarted. Even so, because ofthe small volume of the basin, it is not easilykept full so that these dangers are a constantthreat to the quality of the casting.

Unfortunately, even keeping the pouring cupfull during the pour is no guarantee of goodcastings if the cup exit and the sprue entranceare not well matched, as we have seen above.This is the most important reason for mouldingthe cup and the filling system integral with themould if possible.

Finally, even if the pour is carried out as wellas possible, any witness of the filling of a conicalbasin will need no convincing that the highvelocity of filling, aimed straight into the top ofthe sprue, will cause oxides and air to be carrieddirectly into the running system, and so into thecasting. For castings where quality is at a pre-mium, or where castings are simply required tobe adequate but repeatable, the conical basin isdefinitely not recommended.

Inert gas shroud

A shroud is the cloth draped as a traditionalcovering over a coffin. This sober meaning does



convey the sense in which the word is used in thefoundry.

The inert gas shroud has been adopted insome steel foundries. The device is a protectiveshield around the metal stream issuing from abottom-poured ladle, rather like a collar, pro-viding an inert gas environment, usually argon.Its purpose is to reduce re-oxidation of the steelduring casting.

It is difficult to believe that a user wouldthink that the short distance between the ladleand the conical basin was influential in anysubstantial reduction of the oxidation of themelt. Usually, the time involved in this shortjourney will be probably only a few hundredmilliseconds. It is not easy therefore to escapethe conclusion that users were in fact tacitlyacknowledging the air pump action of the con-ical basin. The shroud therefore encouragesargon to be sucked into the cone instead of air,assuming that the rate of delivery of argon issufficient (since such pumps usually transferroughly equal volumes of pumping fluid andentrained fluid).

The beneficial action of an argon shroud isthat the reactive gas is simply replaced by aninactive gas. Thus although volumes of bubbleswill continue to be entrained with the flow, theyat least do not react to produce oxides ornitrides.

In fact of course, the shroud will never becompletely protective for various reasons: thegas itself will be contaminated with oxygen,water vapour and other gases and volatiles inthe plumbing system that delivers the gas. Moreimportant still, the seal of the shroud around thestream cannot be made proof against leakage ofair; and finally the outgassing from the mould,especially in the case of an aggregate (sand)mould, will be massive.

Even so, when used appropriately, theshroud is useful. It greatly reduces re-oxidationproblems of steels during casting as demon-strated by research carried out by the SteelFounders Society of America (2000). The resultemphasizes the damage done by the emulsion ofsteel and air bubbles that characterizes theaverage poorly designed casting system.

The shroud has been taken to an extremeform as a long silica tube mounted directly tothe underside of a bottom-pour ladle (HarrisonSteel, USA, 1999). The tube acts as a re-usablesprue, and is inserted through the top of themould and lowered carefully, so that its exitreaches the lowest point of the filling system.The stopper is then opened. If the seal betweenthe ladle and tube is good, the filling rate of themould is high. If leakage of air occurs at the sealthe rate of mould filling is significantly reduced,

implying the strong pumping action of the fall-ing stream to create a vacuum in the upper partof the tube, drawing in air if it can, and thusdiluting the falling stream with air. Severalcastings in succession can be poured from onetube. However, after the tube cools the silicafragments, and requires to be replaced.Although this solution to the protection of themetal stream from oxidation is to be admiredfor its ingenuity, it does appear to the author tobe awkward in use. The leakage problem isalways an attendant danger.

In general, the author has not opted for theshroud solution, but has preferred to put inplace systems that avoid the ingestion of gasesinto the filling system. These various systems aredescribed below.

Contact pouring

The attempt to exclude air during the pouring ofcastings is carried to its ultimate logical solutionin the concept of contact pouring. In this systemthe metal delivery system and the mould arebrought into contact so that air is effectivelysealed out.

The direct contact system is of course neces-sary, and taken for granted in the case ofcounter-gravity systems, in which the mould isplaced directly over a source of metal. The metalis then displaced upwards by pump or differ-ential pressure.

In the case of gravity pouring, however, theauthor is only aware of one use in a foundry(VAW, now Hydro Aluminium Limited, Dilligem,Germany) casting aluminium alloy. The meltis brought to the casting station by launder(a horizontal channel). The mould is also broughtup to the underside of the launder in the base ofwhich is a nozzle closed by a stopper. When themould is presented to and pressurized againstthe nozzle the stopper is opened. After themould is filled the stopper is closed and themould can be removed, in this particular case tobe rolled immediately through 180 degrees toavoid convection and aid feeding. This systemworks reliably and well.

The thought of transferring the concept tosteel castings, using the stopper in the base of thebottom-poured ladle to deliver directly intothe mouth of a sprue is quite another matter.The engineering problems for steels are daunt-ing at this time, but may be solved one day.

The offset basin

Another design of basin (sometimes called abush) that has been recommended from time totime, is the offset basin (Figure 2.9a).



The floor of this basin is usually arranged tobe horizontal (but sometimes sloping). Theintention is that the falling stream is brought torest prior to entering the sprue. This, unfortu-nately, is not true. The vertical component offlow is of course zero, but the horizontal com-ponent is practically unchecked. This sidewaysjet across the entrance to the sprue preventsapproximately half of the sprue from fillingproperly, so that air is entrained once again. Thehorizontal component of velocity continuesbeneath the surface of the liquid throughout thepour, even though the basin may be filled.

There has been research using this type ofbasin over the years, in which the dischargecoefficients from sprues have been measuredand found to be in the region of 50 per cent orless. These low figures confirm that the sprue isonly 50 per cent or less filled, so that the majorfluid being discharged is air. The quality of anycastings produced from such devices must havebeen lamentable.

This type of basin is definitely not recom-mended.

The offset step (weir) basin

The provision of a vertical step, or weir, in thebasin (Figure 2.9b and c) brings the horizontaljet across the top of the sprue to a stop. It is anessential feature of a well-designed basin.

Interestingly, this basin has a long history.Sexton and Primrose described a closely similar

design (but without a well-formed step) in theirtextbook on ironfounding published in 1911. Ifthis basin is really valuable (as is recommendedhere) the reader will be curious as to why it hasbeen known for so long, but has been extremelyunpopular in foundries, whose experience of ithas been discouraging. There are several reasonsfor this bad experience. Sometimes the basin hasbeen made incorrectly, neglecting the importantdesign features listed below. However, moreserious than this, it has been usual to place thisexcellent design of basin on a filling system thatcompletely undoes all the benefits provided bythe basin. Thus the benefits of the basin arenever realized, and the basin is unjustly blamed.

Despite the revered age of this basin design,the precise function and importance of eachfeature of the design had not been investigateduntil recent computer studies by Yang andCampbell (1998). These studies make it clearthat

(i) The offset blind end of the basin isimportant in bringing the vertical down-ward velocity to a stop. The offset alsoavoids the direct inline type of basin, such asthe conical basin, where the incoming liquidgoes straight down the sprue, its velocityunchecked, and taking with it unwantedcomponents such as air and dross, etc.

In older designs of this device the blindend of the basin was often moulded as ahemispherical cup. This was not helpful,

Figure 2.9 Offset pouring basins (a) without step (definitely not recommended); (b) sharp step(not recommended); (c) radiused step (recommended).


Offset

(a) (b) (c)


since metal could easily be returned out ofthe basin by the sloping sides. The flat floorand near-vertical sides of the basin weretherefore significant advantages. In fact theuse of sharp corners to the offset side of thebasin is positively helpful to avoid metalbeing ejected by the basin as discussed later.

(ii) The step (or weir) is essential to eliminatethe fast horizontal component of flowover the top of the sprue, preventing itfrom filling properly. Basins without thisfeature commonly only approximately halffill the sprue, giving an effective so-calleddischarge coefficient of only approximately0.5 (how could it be higher if the sprue isonly half full?). The provision of the stepyields a further bonus since it reverses thedownward velocity to make an upwardflow, giving some opportunity for lighterphases such as slag and bubbles to separateprior to entering the sprue. Floating debristhat has separated in this way is shownschematically in Figure 2.9b, c). Again, earlydesigns were less than ideal because the stepwas not vertical (Swift 1949) so that itseffect was compromised. The step needs avertical height at least equal to the height ofthe stream at that point to ensure that itbrings the horizontal component of flow toa complete stop. Commonly, this height willbe at least a few millimetres for a smallcasting, and might be 10 to 20 mm for acasting weighing several tonnes.

(iii) Finally, the provision of a generous radiusover the top of the step (Figure 2.9c),smoothing the entrance into the sprue,further aids the smooth, laminar flow ofmetal. Swift and co-workers (1949) illus-trated this effect clearly in their watermodels of various basins. The effect is alsoconfirmed by the computer study by Yangand the author (1998).

The practice of placing a boom, or dam acrossthe top of the basin (Figure 2.10) to hold backfloating debris is probably counter-productive.It is seen to interfere with the natural circulationin the basin that will automatically favour theseparation of buoyant phases. A dam is notrecommended.

In practice, compared to the conical type, theoffset step design of basin is so easy to keep fullit becomes immediately popular with both casterand quality technologist alike. And, naturally,when teamed up with a well-designed filling sys-tem, the basin can demonstrate its full potentialfor quality improvement of the casting.

An understandable criticism is that the basinsare so voluminous that they reduce yield and are

thus costly. The usual design is shown inFigure 2.11a. Clearly the yield criticism can becompletely met by ensuring that the basin drainsas completely as possible by arranging it to besufficiently higher than the casting. However, ofthose cases where the basin has to be placedlower and will not drain, the problem is to someextent addressed by the design variant shown inFigure 2.11b. In addition to saving money, thisbasin works even better because it constrains themelt more effectively. It encourages the funnel-ling of the melt into the sprue with excellentlaminar directional guidance.

These offset step basins can be made asseparate cores, stored, and planted on moulds,matching up with the sprue entrance whenrequired. However, because they will berequired for many different castings, and so willneed to mate up with different sprue entrancediameters, there is concern about any mis-matchof the basin exit and the sprue entrance. How-ever, the problem is much less acute than mis-match of conical basins, because the speed of thefalling stream at this point is considerably lower,in fact only at about its critical velocity. In thesecircumstances surface tension is able to bridgemodest outstanding ledges without significantentrainment of the liquid surface. An over-hanging ledge is probably more serious and tobe avoided. Thus a selection of stored basinswith excess exit diameter is to be preferred. Infact it may be preferable to arrange the bush tohave its base completely removed on the sprueside. The bush will then fit practically anymould. Provided the entrance to the sprue onthe top surface of the cope is nicely radiused, themetal will probably be adequately funnelled intothe sprue (see Figure 2.23).

Figure 2.10 Basin with dam (probably not helpful).



Ultimately, however, the author prefers tomould the basin integral with the sprue, and soavoiding the link-up and alignment problems.This is easily achieved with a vertical mouldjoint, but less easy, but still possible, with ahorizontally jointed mould.

The basin is easier to use, and works moreeffectively, if its response time is approximately1 second. To the author's knowledge there is nodefinition of response time. I therefore adopt aconvenient measure as the time for the basin toempty completely if the pourer stops pouring. Inpractice, of course, the pourer does not usuallystop pouring, so that the actual rate of change oflevel of the basin is usually at least double theresponse time as defined above. Such times arerelatively leisurely, allowing the pourer tomaintain a consistent level of melt in the basin.Different pourers or pouring systems mayrequire times shorter or faster than this.

The volume of the basin Vb (m3) to give aresponse time tr (in seconds) at a pouring rateQ (m3 sÿ1) is given simply by

Vb � Q=tr

Clearly, when tr� 1 second, Vb�Q when usingthe recommended SI units.

Offset stepped basin with a bottom-pour ladle

Ladles equipped with a nozzle in the base arecommon for the production of large steel

castings. The benefits are generally describedto be:

(i) the metal is delivered from beneath thesurface of the melt, so avoiding the transferof slag;

(ii) for large castings the tipping of a ladle toeffect a lip pour becomes impractical;

(iii) the accuracy of the placing and the direc-tion of the pour is valuable. Even so it iswidely known in the trade that foundriesusing bottom pour ladles suffer dirtiercastings than those steel foundries that uselip pour ladles. This follows as a naturalconsequence of the great difference inpouring speeds into the conical basin, withthe consequent great difference in the rateof entrainment of air. (The use of bottom-pour ladles with an offset stepped basin atthe entry to the mould has the potential toavoid this central problem. However, it isnot without its own set of requirements thatneed to be studied carefully, as we shall seebelow.)

The common problem when using an offsetstepped basin is that although a pourer using alip pour ladle can continue to adjust the rate ofpour to maintain the level of liquid at therequired height in the basin, this is easier saidthan done if the melt is being supplied from abottom-pour ladle whose rate of delivery oftencannot be controlled, the stopper is either open

Figure 2.11 Side and plan viewsof offset basins (a) conventionalrectangular; (b) slimmed shapeto streamline flow and improvemetal yield.


(a) (b)


or closed. Any attempt to adjust the rate ofdelivery results in sprays of steel in all directions.

In addition to this problem, as the bottom-teemed ladle gradually empties it reduces its rateof delivery. In the case of pouring a singlecasting from a ladle, it is fortunate that thefilling system for the casting actually requires afalling rate of delivery as the net head (the levelin the basin minus the level of metal in themould) of metal driving the flow around thefilling system gradually falls to zero. Even so, itis clear that the two rates are independentlychanging, and may be poorly matched at times.The match of speeds might be so bad that thebasin runs empty, but even well before thismoment, filling conditions are expected to bebad. At a filling level beneath the designed filllevel in the basin the top of the liquid will appearto be covering the entrance to the sprue, butunderneath, the sprue will not be completelyfilled, and so will be taking down air. It isessential therefore to ensure, somehow, that thelevel in the basin remains at least up to itsdesigned level. At this time the problem ofsatisfactorily matching speeds can only besolved in detail by computer. Most softwaredesigned to simulate the filling of castingsshould be able to tackle this problem. However,it is perhaps more easily solved by simply havinga basin with greatly increased depth, forinstance perhaps up to four times the designdepth. The ladle nozzle size is then chosen todeliver at a higher rate, causing the basin tooverfill its design level, and so effectively run-ning the casting at an increased speed. Thisincreased speed is far preferable to the danger ofunderfilling the basin with the consequentialingestion of air into the melt.

In general therefore, a greatly increaseddepth to the basin is very much to be recom-mended. The problem of overfilling andincreased speed of running may not be as seriousas it might first appear. The reason is quicklyappreciated. If the rate of delivery from the ladleis 40 per cent higher (a factor of 21/2) than thedesigned rate of filling of the casting, the heightof metal in the pouring basin will rise to a leveltwice as high (provided the basin has beenprovided with sufficient depth of course). Abasin four times the minimum height willaccommodate delivery from the ladle at up totwice as fast as the running system was designedfor. The increase in pressure that this provideswill drive the filling system to meet the higherrate. (Notice that the narrow sprue exit is notacting as a so-called choke, illustrating howwrong this concept is.) Thus the system is,within limits, automatically self-compensating ifthe basin has been provided with sufficient

freeboard. It is important therefore to makesure that offset stepped basins in collaborationwith a bottom poured ladle do have sufficientadditional height.

The preferred option to overfill the basin interms of height is valuable in the other commonexperience of using a large bottom-pour ladle tofill a succession of castings. Let us take as anexample a 20 000 kg ladle that is required topour nine castings each of 2000 kg. (The final2000 kg in the ladle will probably be discardedbecause it will pour too slowly, contain toomuch slag and be too low in temperature; thereare sometimes real problems when pouringsuccessive castings from one ladle.) The firstcastings will be poured extremely rapidlybecause the head of metal in the ladle will behigh. However, the most serious problem is thatthe final castings in the sequence will be pouredslowly, perhaps too slowly, and so might suffersevere damage from air entrainment.

The important precaution therefore is toensure that the final casting is still poured suf-ficiently quickly that the minimum height in thepouring basin is still met. This is a key require-ment, and will ensure that the final casting isgood. Thus all of the filling design should bebased on the filling conditions for the last cast-ing. Clearly, all the preceding castings will all beoverpressurized by increased heights of metal intheir pouring basins, and so will fill corre-spondingly faster, with correspondingly highervelocities entering the mould. This should bechecked to ensure that the velocities are not sovery high as to cause unacceptable damage.Usually, this approach can be made to workout well.

In some cases the first castings may have theirpouring basins filled high, but the metal not yetarrived in the feeders to give a signal to theoperator to stop pouring. In this case the onlyoption is to monitor the progress of the pour bysome other factor, such as precise timing, orbetter still, a direct read-out load cell on theoverhead hoist carrying the ladle.

The matching of the speed of delivery fromthe ladle with the speed of flow out of thepouring basin is greatly assisted if the rate ofdelivery from the ladle is known. This is acomplex problem dependent on the height ofmetal in the ladle, its diameter, and the diameterof the nozzle. The interaction of all these factorscan be assessed using the nomogram provided inthe Appendix.

The sharp-edged or undercut offset weir basin

In addition to the matching of the rate of flowbetween the ladle and the casting, there are



additional problems with the application ofoffset weir basins for use with bottom-pouredladles.

As we have discussed above, the velocity ofthe melt exiting the base of the bottom-pouredladle when the stopper is first opened is sober-ingly high. This is because the melt at the base ofa full ladle is highly pressurized. Effectively ithas fallen from the upper surface of the melt inthe ladle; often as much as a metre or more.Thus the exit speed is often in the region of 4 or5 m sÿ1. This is so high that if this powerful jet isdirected into the blind end of a step basin, theliquid metal flashes outwards over the base, hitsthe radii in the corners of the vertical sides,where it is turned upwards to spray all over thefoundry (Figure 2.12a). Such spectacular pyro-technic displays are not recommended; littlemetal enters the mould.

The small radii around the four sides of theoff-axis well of the basin are extremely effectivein redirecting the flow upwards and out of thebasin. One solution to this problem is thereforesimply the removal of the radii. The provisionof sharp corners to all four sides reduces thesplashing tendency to a minimum (the top of theweir step leading over to the sprue entranceshould still be nicely radiused of course).

The sharp cornered basin is a useful design.However, an ultimate solution to the splashingproblem is provided by a simple re-entrantundercut at the base of the basin (Figure 2.12b).(The author demonstrated such a basin in a steelfoundry while foundry personnel hid behindpillars and doors. On the opening of the ladlestopper the stream gushed into the basin, butnot a drop emerged. The pouring process wasquiet; its intense energy tamed for the first time.

The foundry personnel emerged from theirhiding places to gaze in wonder.)

The undercut is, of course, a problem formany greensand moulding operations makinghorizontally parted moulds. This is why thesharp-edged basin is so useful. Even so, whereextreme incoming velocities are involved, anundercut edge to all four sides of the filling wellof the basin may be the only solution.

The undercut may be difficult to mould, butit can be machined. The upgrading of a spruecutter to 3-D machining unit equipped with aball-ended high speed cutter would make shortwork of the basin, complete with its undercutand sprue entrance, and providing all this withinthe moulding cycle time. Such a unit would bean expensive sprue cutter, but would be a goodinvestment.

The undercut is not a problem for verticallyjointed moulds. Its use on machines such asDisamatics is popular and welcomed by thefoundry operators. Its quiet filling is easilycontrolled, and there is complete absence ofsplashed metal (commonly seen as pools, some-times nearly lakes, swimming around on the topsof moulds). The reduction of pouring overspill isa significant contribution to the raising of metalyield in the foundry.

The moulding of the sprue cover (Figure 2.12b)ensures that metal is never poured in errordirectly down the sprue, and saves a little metal,making a further small contribution to yield.(In some iron foundries, however, the designmay be less good at holding back slag sincethere is now less volume provided for slag toaccumulate.)

If the offset stepped basin is successfullymaintained full, the head of metal provided by

Figure 2.12 Offset basins for highvelocity input. (a) No undercutempties spectacularly upwards (notrecommended). (b) The provision ofan undercut gives a basin that does notsplash a drop. The shaded area can bemoulded in a vertical jointed mould tofurther improve flow and metal yield,and prevents any risk of pouring directlydown the sprue.


(a) (b)


the height of the down-runner will be steady,and the rate of flow will be controlled bythe sprue. The filling rate will be no longer at themercy of the human operator on that day. Therunning system will the have the best chanceto work in accord with the casting engineer'scalculations.

Stopper

As a further sophistication of the use of theoffset step basin, some foundries place a smallsand core in the entrance to the sprue. The corefloats only after the bush is full, and thereforeensures that only clean metal is allowed to enterthe sprue. Alternatively, a wire attached to thecore, or a long stopper rod lifted by handaccomplishes the same task. For a large castingthe raising of the stopper will require a moreruggedly engineered solution, involving thebenefit of the action of a long lever to add tothe mechanical advantage and keep the opera-tor well away from sparks and splashes. How-ever it is achieved, the delayed opening of thedown-runner is valuable in many foundrysituations.

The early work on the development of fillingsystems at Birmingham concentrated on the useof the offset step basin. A stopper was not usedbecause it was considered to be too muchtrouble. However, after about the first 12months, as a gesture to scientific diligence, itwas felt that the action of a stopper should bechecked, if only once, by observing the filling ofa sprue using the video X-ray radiographic unit,comparing filling conditions with and withouta stopper. A stopper was placed in the sprueentrance, sealing the sprue. The metal waspoured into the basin. When the basin was filledto the correct level the stopper was raised. Thepouring action to keep the basin full was thencontinued until the mould was filled. The resultswere unequivocal. The use of a stopper greatlyimproved the filling of the sprue. It was withsome resignation that the author affirmed thisresult. For all castings after that day, a stopperwas always used.

Latimer and Read (1976) demonstrated thatthe use of a stopper reduced the fill time by60 per cent. This is further proof that the systemruns much fuller.

There seems little doubt therefore that,despite the inconvenience, when the best qualitycastings are required, a stopper is advisable.Thus the author always recommends its use foraerospace products.

In addition, the use of stoppers is particularlyuseful for very large castings where differentlevels of the filling system are activated by the

progressive opening of stoppers as the melt levelrises in the mould, so bringing into action newsources of metal to raise the filling speed.

2.3.2.3 Sprue (down-runner)

The sprue has the difficult job of getting the meltdown to the lowest level of the mould whileintroducing a minimum of defects despite thehigh velocity of the stream.

The fundamental problem with the design ofsprues is that the length of fall down the spruegreatly exceeds the critical fall height. Theheight at which the critical velocity is reachedcorresponds to the height of the sessile drop forthat liquid metal. Thus for aluminium this isabout 13 mm, whereas for iron and steel it isonly about 8 mm. Since sprues are typically 100to 1000 mm long, the critical velocity is greatlyexceeded. How then is it possible to preventdamage to the liquid? This question is not easilyanswered and illustrates the central problem tothe design of filling systems that work usinggravity. (Conversely, of course, counter-gravitysystems can solve the problem at a stroke, whichis their massive technical advantage.)

For the sprue at least, the problem is soluble.It seems that the secret of designing a goodsprue is to make it as narrow as possible, so thatthe metal has minimal opportunity to break andentrain its surface during the fall. The concepton protecting the liquid from damage is either(i) to prevent it from going over its criticalvelocity, or (ii) if the critical velocity has to beexceeded, to protect it by constraining its flow inchannels as narrow as possible so that it is notable to jump and splash.

Theoretically a design of the sprue can beseen to be achieved by tailoring a funnel in themould of exactly the right size to fit around afreely falling stream of metal, carrying just theright quantity of metal per second (Figure 2.13).We call the funnel the down-runner, or sprue forshort. Many old hands call it the spue, or spew(which, incidentally, does not appear to be ajoke).

Most sprues are oversized. This is bad formetallic yield, and thus bad for economy.However, it is much worse for the metal quality,which is damaged in two important ways:

(i) The sprue takes more time to fill. Air istherefore taken down with the metal, caus-ing severe surface turbulence in the sprue.This, of course, leads to a build-up of oxidein the sprue itself, and much consequentialdamage downstream from oxide and en-trained air. The amount of damage to themetal caused by a poor basin and sprue can



be quickly appreciated from the commonobservation of the blockage of filters. Evenwith good quality liquid metal, a poor basinand sprue will create so much oxide that afilter is simply overloaded. Such poor frontends to filling systems are so common thatfilter manufacturers give standard recom-mendations of how much metal a filter can beexpected to take before becoming choked.However, in contrast to what the manufac-turers say, with a good basin and sprue (andproviding, of course, the quality of the meltis not too bad) a filter seems capable ofpassing indefinite quantities of liquid metalwithout problem.

(ii) The free fall of the melt in an oversizedsprue, together with air to oxidize away thebinder in the sand, is a potent combinedassault that is highly successful in destroy-ing moulds. The hot liquid ricochets andsloshes about, its high speed and agitationpunishing the mould surface with a ham-mering and scouring action. At the sametime the pockets of air in this unsteady flowwill be displaced through the sand likeblasts from a blacksmith's bellows, causingthe organic matter in the binder to glow,and, literally, to disappear in a puff ofsmoke! When the binder is burned away,reclaiming the sand back to clean, un-bonded grains, the result is, of course,severe sand erosion. Figure 2.14 shows atypical result for an aluminium alloy castingin a urethane resin-bound mould. An over-size sprue is a liability.

Conversely, if the sprue is correctly sized themetal fills quickly, excluding air before anysubstantial oxidation of the binder has a chanceto occur. The small amount of oxygen in thesurface region of the mould is used up quicklyby the burning of a small percentage of binder,but further oxidation has to proceed at the rateat which new supplies of air can arrive by dif-fusion or convection through the body of themould. This is, of course, slow, and is thereforenot important for those parts of the mould suchas the sprue, that are required to survive for onlythe relatively short duration of the pour. Fur-thermore, since the liquid metal now fills thevolume of the down-runner, the oxide filmforming the metal±mould interface is stationary,protecting the mould material in contact withthe sprue, and transmitting the gentle pressureof the steady head of metal to keep it intact. Theresult is a perfectly cast sprue (Figure 2.14), freefrom sand erosion and oxide laps. A correct-sized sprue for an aluminium alloy casting willshine like a new pin. (But beware, an undersizedsprue will too!) Figure 2.3 illustrates someexamples of good and bad systems. A test of agood filling system design in any metal is howwell the running system has cast. It should beperfectly formed.

How then is it possible to be sure that thesprue is exactly the right size? The practicalmethod of calculating the dimensions of thesprue is explained in Section 2.3.7 `Practicalcalculation of the filling system'. Basically, thesprue is designed to mimic the taper that thefalling stream adopts naturally as a result of itsacceleration due to gravity (Figure 2.15). Theshape is a hyperbola (interestingly, not a para-bola as widely stated). Because most sprues

Figure 2.14 An oversize sprue that has suffered severeerosion damage because of air entrainment during thepour. A correctly sized sprue shows a bright surface freefrom damage.

Figure 2.13 The geometry of the stream falling freelyfrom a basin.



approximate the shape to a straight taper, thecurved sides of the stream encourage the metalto become detached from the walls at abouthalf-way down as shown in this figure. Formodest-sized castings this (together with othererrors, mainly due to the geometry and frictionof the flow in the basin) is simply corrected bymaking the sprue entrance about 20 per centlarger in area (corresponding of course to about10 per cent increase in diameter). Thus straighttapered sprues are commonly used, and appearto be satisfactory.

For very tall castings the straight taperedapproximation to the sprue shape is definitelynot satisfactory. In this case it is necessary tocalculate the true diameter of the sprue at closeintervals along its length. The correct form ofthe falling stream can then be followed withsufficient accuracy, and air entrainment duringthe fall can be avoided.

Using this detailed approach the author hassuccessfully used sand sprues for very largecastings (including a steel casting of about50 000 kg and 7 m high. The sprue was assem-bled from a stack of tubular sand cores, accu-rately located by an annular stepped joint. Only

one core box was required, but the centralhole required a pile of separately turned loosepieces). The conventional use of ceramic tubesfor the building of filling systems for steel cast-ings was thereby avoided, with advantage to thequality of the casting. As an interesting aside,the appearance of this sprue after being brokenfrom the mould was at first sight disappointing.It seemed that considerable sand erosion hadoccurred, causing the sprue to increase in dia-meter by over 10 mm (about 10 per cent). Oncloser examination however, it became clearthat no erosion had occurred, but the chromitesand had softened and been compressed, losingits air spaces between the grains to become asolid mass. It had partially softened probably asa result of the use of a silicate binder system; thesilicate had probably reacted with the chromiteto form a lower melting point phase. Since sucha growth in diameter would necessarily haveoccurred by a kind of creep process, in whichpressure, temperature and time would beinvolved, it follows that much of this expansionwould have happened after the casting had fil-led, since pressure was then highest, the sandfully up to temperature, and more time would be

Figure 2.15 The theoretical hyperbola shape of the falling stream, illustrating the complicating effects of the basinand sprue entrance.


Basin

Theoreticaloutline offreely fallingstream

Gap often seen inX-ray video radiography

Sprue entrance

Straighttaper sprue

Sprue exit


available because the time for the solidificationof the sprue would be as much as ten timeslonger than the pouring time. Thus during thepour the sand-moulded sprue would almostcertainly have retained a satisfactory shape, ascorroborated by the predicted fill time beingfulfilled, and the cleanness of the metal risingin the mould cavity was clearly seen to besatisfactory.

Sand-moulded filling systems for steel cast-ings are, of course, prone to erosion if the sys-tem design is bad, and particularly if, as is usual,the system is oversized. In this case however, thesand-moulded sprue worked considerably betterthan the conventional ceramic tube system.However, there would be no doubt that theceramic tubes would be excellent if they could bespecifically designed and produced for thesprues for each individual steel casting. Nat-urally, at the present time this is not easilyarranged. Even so, it may be found to be aneconomic option in view of the expensive sandsand mould coatings required if sand is usedalone. In addition, the ceramic tubes are extre-mely easy and quick to incorporate into a sandmould, often avoiding the problem of creating anew joint line in the mould.

The cross-section of the sprue can be roundor square. Some authorities have stronglyrecommended square in the interests of reducingthe tendency of the metal to rotate, forming avortex, and so aspirating air. This probably wasimportant in castings using conical pouringbasins because any out-of-line pouring wouldinduce rotation of the melt. However, theauthor has never seen any vortex formation withan offset step basin. The problem seems not toexist with good basin design.

In addition, of course, the vortex appears tobe unjustifiably maligned. The central cone of

air will only act to introduce air to the casting ifthe central cone extends into the mould cavity.This is unlikely, and in its use with the vortexsprue and other benign use of vortices, thedesign is specifically arranged to suppress thispossibility. The vortex can be a powerful friend,as we shall see.

The attempt to provide gating or feeding offvarious parts of the sprue at various heights isalmost always a mistake, and is to be avoided.Examples are shown in Figure 2.16. Overflowfrom such channels can introduce metal into themould prematurely, where it can fall, splashing,and damaging the casting and mould before thegeneral arrival of the melt via the intendedbottom gate. Even if the channels are carefullyangled backwards to avoid premature filling,they then act to aspirate air into the metalstream. Thus divided sprues usually either act tolet out metal or let in air. They are not easilydesigned. Extreme caution is recommended.Perhaps one day we shall be able to design suchfeatures with complete safety as a result of highquality computer simulation. Those days areawaited patiently.

To summarize: for ease and safety of designat this time, the sprue should be a single,smooth, nearly vertical, tapering channel, con-taining no connections or interruptions of anykind. The rate of filling of the mould cavityshould be under the absolute control of its cross-section area. If, therefore, the casting is found inpractice to be filling a little too fast or too slow,then the rate can be modified without difficultyby slight adjustment of the size of the sprue.

Significantly, it is not simply the sprue exitthat requires modification in this case. If cor-rectly designed, the whole length of the sprueacts to control the rate of flow. This is what ismeant by a naturally pressurized system. We

Figure 2.16 An illustration of various kinds of common junctions or misalignments of the sprue. None are recommended.


(a) (b) (c) (d)


can get the design absolutely correct for thesprue along its complete length. Althoughmethoding engineers have been carrying outsuch calculations correctly for many years,somehow only the sprue exit has been con-sidered to act as the choke. We need to takecareful note of this widespread error, and per-haps take time to re-think our filling systemconcepts.

Turning now to a common problem withmany automatic moulding units for the manu-facture of horizontally parted moulds. It isregrettable that a reverse-taper sprue is usuallythe only practical option, flagging up a majorproblem with the design of nearly all of ourmodern automatic moulding machines. (What isworse, these units also cannot usually providefor a properly moulded basin. Despite such abasin being possible to be machined as men-tioned above, production by cutting is usuallynever actioned.) The sprue pattern needs to bepermanently fixed to the pattern plate, andtherefore has to be mouldable (i.e. the mouldhas to be able to be withdrawn off the spruewhen stripping the mould off the pattern) asseen in Figure 2.3a. In this case all is not yet lost.The top of the sprue should be designed tomaintain its correct size, and the taper (now thewrong sign, remember) down the length of thesprue should be kept to a minimum. (A polishedstainless steel sprue pattern can often workperfectly well with zero taper providing thestripping action is accurately square.)

Even though all precautions are taken in thisway to reduce the surface turbulence to aminimum, the consequential damage to the meltby a reverse or zero tapered sprue is preferablyreduced by the provision of a filter as soon aspossible after the base of the sprue. The frictionprovided by the filter acts to hold back the flow,and thus assist the poorly shaped sprue to back-fill as completely and as quickly as possible, andso reduce the rate of damage. The filter will alsoact to filter out some of the damage, although ithas to be realized that this filtering action is notparticularly efficient. The use of filters is dealtwith in detail later in Section 2.3.6.3.

We need to dwell a little longer on theimportance of the use of the correct taper, so faras possible, for sprues.

The effect of too little, or even negative taperhas been seen above to be detrimental to castingquality. Surely, one might expect that the oppo-site condition of too much taper would not be aproblem, since it seems reasonable to assumethat the velocity of the metal depends only onthe distance of fall. However, this is not true.The head of metal in the pouring basin is thedriving force experienced by the melt entering

the sprue. If the sprue tapers to match thenatural taper of the falling stream the onlyacceleration experienced by the melt is theacceleration due to gravity. If, however, thetaper of the sprue is greater than this, the melt iscorrespondingly speeded up as the sprue con-stricts its area. This extra speed is unwelcome,since the task of the filling system designer is toreduce the speed. The effect of varying taper hasbeen studied by video X-ray techniques. Inexperiments in which the sprue exit area wasmaintained constant, a doubling of the sprueentrance area was seen to nearly double the exitspeed, with the generation of additional turbu-lence in the runner. Three times greater entrancearea led to such increased velocities in the run-ner that severe bubble entrainment was created(Sirrell and Campbell 1997). This is one ofthe reasons why the elongated basin/sprue(Figure 2.8e) is so bad.

This effect is illustrated in Figure 2.17. Forthe negative tapers (a) and (b) the velocity at thesprue exit is merely that due to the fall of metal.The rate of arrival (kg sÿ1) is of course con-trolled by the area of the sprue top. For thecorrectly sized sprue (c) the velocity and rate ofdelivery are substantially unchanged, althoughit will be noticed that the whole of the length ofthe sprue is now contacting and controlling thestream, to the benefit of the melt quality. Thosesprues with too much taper (d) and (e) continueto deliver metal at nearly the same rate (in kg sÿ1

for instance), but at much higher speed (in msÿ1

for instance) in proportion to the reduction inarea of the exit. Far from acting as an effectiverestraint, the narrow sprue exit merely increasesproblems.

These effects were studied using real-timeX-ray radiography (Sirrell et al. 1995) to opti-mize the taper, measuring the time for the sprueto back-fill, and the speed of the exiting melt(Figure 2.18). This work confirmed that thelong-used 20 per cent increase of the area ofthe sprue entrance was a valuable correction.The consequential 20 per cent increase in velo-city into the runner was an acceptable penalty toensure that the sprue primed faster and morecompletely despite its straight-wall approximateshape.

Thus to summarize the effect of sprue taper;the taper has to be correct (within the 20 per centoutlined above). Too little or too much taper bothlead to damage of the melt.

Multiple sprues

In magnesium alloy casting the widespread useof a parallel pair of rectangular slots to act asthe sprue seems to be due to the desire for the



reduction in vortex formation (especially, as wehave noted, if poor designs of pouring basin areemployed). Swift et al. (1949) have used threeparallel slots in their studies of the gating ofaluminium alloys. However, the really usefulbenefit of a slot shape is probably associatedwith the reduction in stream velocity by theeffect of friction on the increased surface area.The slots can be tapered to tailor their shape tothat of the falling stream. However, to be strictlyaccurate, their area should be modified to makeallowance for the additional frictional losses.

Such sprues would probably benefit the widercasting industry. A study to confirm the extentof this expected benefit would be valuable.

A really important benefit from the use of aslot sprue appears to have been widely over-looked. This is the accuracy with which it can beattached to a slot runner to give an excellentfilling pattern. This benefit is described in detailin the section below concerning the design pro-blems of the sprue/runner junction. It seems thatwe should perhaps be making much more reg-ular use of slot sprues.

When pouring a large casting whose volumeis greater than can be provided from the ladle, itis common to use more than one ladle. Thesequential pouring of one ladle after the otherinto a single basin has to be carried outsmoothly because any interruption to the pouris almost certain to create defects in the casting.Simultaneous pouring is often carried out.Occasionally this can be accomplished with asingle sprue, but using an enlarged pouringbasin, often with a double end, either side of thesprue, allowing the ladles access from eitherside. Often, however, two or more sprues areused, sited at opposite ends of the mould, so asto give plenty of accessibility for ladles andcranes, and reduce the travel distance for themelt in the filling system. The correspondinglysmaller area used when using more than onesprue is an advantage because they fillmore easily and quickly, excluding their airmore rapidly. Multiple sprues for larger castingsare to be recommended and should beconsidered more often.

Figure 2.18 Experimental data from video radiographicobservations of sprue filling time and velocity of discharge.A taper of 1.2 is shown to be close to an optimum choice(Sirrell 1995).

Figure 2.17 A variety of straight tapered sprues. Too little or too much taper is bad. Only the centre taper to matchthe falling stream is recommended. Even this could be improved by 20 per cent additional entrance area, or better still,shaped to follow the shape of the stream.


(a)

V V V 2 V 4 V

(b) (c) (d) (e)

+

++ +++++

4

3

2

1

0

2

1

00 1 2

Area actual sprue entrance/Area of optimum sprue

Spr

ue fi

ll tim

e (s

)

Mel

t vel

ocity

(m

s–1)

3 4


For very large castings, an interesting tech-nique can be adopted. Several sprues can con-nect to runners that are arranged around themould cavity at different heights. In the pour ofa 3 m high iron casting weighing 37 000 kgdescribed by Bromfield (1991), four sprues werearranged to exit from two pouring basins. Thesprues were initially closed with graphite stop-pers. The trough was first filled. The stoppers tothe lowest level runner were then opened. Theprogress of the filling was signalled by themaking of an electrical contact at a criticalheight of metal in the mould. In other instanceswitnessed by the author, the progress of fillingcould be observed by looking down risers orsighting holes placed on the runners. When thenext level of runner was reached, announced bythe bright glow of metal at the base of thesighting hole, the next level of sprues wasbrought into action to deliver the metal to thislevel of runner. In the case of the casting thatwas witnessed, three levels of runners wereprovisioned by six sprues. The technique hadthe great advantage that the rate of pouring didnot start too fast, and then slowly decrease tozero during the course of the pour. The ratecould be maintained at a more consistent levelby the action of bringing in additional sprues asrequired. In addition, the temperature of theadvancing front of the melt could also bemaintained by the fresh supplies of hot metalarriving at the different levels, thus reducing theneed for excessive casting temperatures to avoidmisruns. Again, the significant advantages ofmultiple sprues are clear.

2.3.2.4 Sprue base

The point at which the falling liquid emergesfrom the exit of the sprue and executes a right-

angle turn along the runner requires specialattention. The design of this part of the liquidmetal plumbing system has received muchattention by researchers over the years, but withmixed results that the reader should note withcaution.

The well

One of the widely used designs for a sprue baseis a well. This is shown in Figure 2.19a. Itsgeneral size and shape has been researched inan effort to provide optimum efficiency inthe reduction of air entrainment in the runner.The final optimization was a well of double thediameter of the sprue exit and double the depthof the runner. This optimization was confirmedin an elegant study by Isawa (1993) who foundthat the elimination of the hundreds and thou-sands of bubbles that were generated initiallyreduced exponentially with time. The exponen-tial relationship gave a problem to define a finitetime for the elimination of bubbles because thedata could not be extrapolated to zero bubbles;clearly the extrapolation predicted an infinitetime! He therefore cleverly extrapolated back tothe time required to arrive at the last bubble,and used `the time to the last bubble' to comparedifferent well designs.

However, it should be noticed that both thisand all the research into wells had been carriedout on water models, and all had used runnersof large cross-section that were not easy to fill.The result was a well design that, at best, clearedthe liquid of bubbles after about 2 seconds.

For small castings that fill in only a few sec-onds we have to conclude that such well designsare counter-productive. In these cases it is clearthat much of the filling time will be taken upconveying highly damaged metal into the mould

Figure 2.19 A variety of sprue/runner junctions in side and plan views from poorest (a) to best (d). The offsetjunction at (e) forms a vortex flow along the cylindrical runner.


(a) (b) (c) (d) (e)


cavity. Thus the comforting and widely heldimage of the well as being a `cushion' to softenthe fall of the melt is seen to be an illusion. Inreality, the well was an opportunity for the meltto churn, entraining quantities of oxide andbubble defects.

Systematic X-ray radiographic studies star-ted in 1992 have been revealing. They haveshown that in a sufficiently narrow fillingchannel with a good radius at the sprue/runnerjunction, the high surface tension of the liquidmetal assists in retaining the integrity of acompact liquid front, constraining the melt.These investigative studies on dramaticallynarrow channels in real moulds with real metalsquickly confirmed that the sprue/runner junc-tion was best designed as a simple turn (Figure2.19c and d), provided that the channels were ofminimum area.

The studies showed that if a well of any kindwas provided, the additional volume created inthis way was an opportunity for additionalsurface turbulence, so damaging the melt. Fur-thermore, after the well was filled, the rotationof the liquid in the well was seen to act as a kindof ball bearing, reducing the friction on thestream at the turn. In this way the velocity in therunner was increased. These higher speedsobserved out of right-angle turns provided by awell were unhelpful. For a narrow turn withouta well the velocity of the metal in the runner hadthe benefit of additional friction from the wall,giving a small (approximately 20 per cent) butuseful reduction in metal speed. Thus the con-clusion that the filling systems perform betterwithout a well seems conclusive.

On a note of caution, it is perhaps necessaryto bear in mind that all this research has beenconducted on rather small castings. Even so,there seems no a priori reason why the principlesshould not also apply to large products.

It is unlikely that wells will disappear fromthe casting scene without strong defence fromtheir supporters. It should be borne in mind thatwells may once have been appropriate wherelarge section runners were used.

In summary, despite what was recommendedby the author in Castings 1991, more recentresearch confirms that wells are no longerrecommended, particularly for narrow sectionfilling systems.

The radius of the turn

It has been shown that for small castings, gen-erally up to a few kilograms in weight, the meltcan be turned through the right angle at thebase of the sprue simply by putting a right-anglebend into the channel. However, if no radius is

provided, the melt cannot follow the bend, sothat a vena contracta is created (Figures 2.7a and2.20). The trailing edge of this cavitated regionis unstable, so that its fluttering and flappingaction sheds bubbles into the stream.

The vena contracta is a widely observedphenomenon in flowing liquids. It occurswherever a rapid flow is caused to turn througha sharp change of direction. An importantexample has already been met in the offsetpouring basin if no step is provided (Figure2.9a). This creates a vena contracta that showersbubbles down the sprue. However, the base ofthe down-runner is probably an even moreimportant example if, as is usually the case,speeds are much higher here. The loss of contactof the stream from the top of the runnerimmediately after the turn has been shown to bethe source of much air in the metal. Experimentswith water have modelled the low-pressureeffect here, demonstrating the sucking ofcopious volumes of air into the liquid as streamsand clouds of bubbles (Webster 1967). This isexpected to be particularly severe for sandmoulds, where the permeability will allow agood supply of air to the region of reducedpressure.

In fact, when pouring castings late at night,when the foundry is quiet, the sucking of airthrough into the liquid metal can be clearlyheard, like bath water down the plug-hole! Suchcastings always reveal oxides, sand inclusionsand porosity above the gates, which are thetell-tale signs of air bubbles aspirated into therunning system.

In contrast, provided that the internal cornerof the bend is given a sufficiently large radius,the melt will turn the corner without cavitationor turbulence (Figure 2.19c). In fact, the actionof the advancing metal is like a piston in acylinder: the air is simply pushed ahead of the

Figure 2.20 The vena contracta problem at aright angle with inadequate radius.



advancing front, never becoming mixed. To beeffective, the radius needs to be at least equal tothe diameter of the sprue exit, and possibly twicethis amount. The precise radius requires furtherresearch. The action of the internal radius isimproved further if the outside of the bend isalso provided with a radius (Figure 2.19d).

For larger casting where surface tensionbecomes progressively less important, thechannels are filled only by the available volumeof flow. Initially, during the first critical periodas the filling system is priming, there is con-siderable danger of significant damage to themetal.

To limit such damage it is helpful to take allsteps to prime the front end of the filling systemquickly. This is assisted by the use of a stopper.However, in Castings (1991) the author con-sidered the use of various kinds of choke at theentrance to the runner as a possible solution tothese problems. Again, recent research has notupheld these recommendations. It seems thatany such constriction merely results in the jet-ting of the flow into the more distant expandedpart of the runner.

This finding emphasizes the value of theconcept of the naturally pressurized system. It isclearly of no use to expand the running systemto fulfil some arbitrary formula of ratios, in thehope that the additional area will persuade theflow velocity to reduce. The flow will obey itsown rules, and we need to design our system tofollow these rules.

The use of a vortex sprue, or even simply avortex base or vortex runner (Figure 2.19e) tothe conventional sprue represent exciting andpotentially important new developments inrunning system design. These concepts aredescribed more fully in Section 2.3.2.12.

2.3.2.5 Runner

The runner is that part of the filling system thatacts to distribute the melt horizontally aroundthe mould, reaching distant parts of the mouldcavity quickly to reduce heat loss problems.

The runner is usually necessarily horizontalbecause it simply follows the normal mouldjoint in conventional horizontally partedmoulds. In other types of moulds, particularlyvertically jointed moulds, or investment mouldswhere there is little geometrical constraint, therunner would often benefit from being inclineduphill.

It is especially useful if the runner can bearranged under the casting, so that the runner isconnected to the mould cavity by vertical gates.All the lowest parts of the mould cavity can thenbe reached easily this way. The technique is

normally achieved only in a three-part mould inwhich the joint between the cope and the dragcontains the mould cavity, and the joint betweenthe lower mould parts (the base and the drag)contains the running channels (Figure 2.21a).The three-part mould is often an expensiveoption. Sometimes the three-level requirementcan be achieved by use of a large core (Figure2.21b), or the distribution system can beassembled from ceramic or sand sections, andbuilt into the mould as the moulding box is filledwith sand (Figure 2.21c). These options areoften worth considering, and might prove aneconomic investment.

More usually, however, a two-part mouldrequires both casting and running system to bemoulded in the same joint between cope anddrag. To avoid any falls in the filling system therunner has to be moulded in the drag, and thegates and casting in the cope (Figure 2.3d).

The usual practice, especially in iron and steelfoundries, of moulding the casting in the drag

Figure 2.21 Bottom-gated systems achieved by(a) a three-part mould with accurately moulded runningsystem; (b) making use of a core; and (c) a two-partmould with preformed channel sections.


Cope

Drag

Base

Core

Drag

Cope

Cope

Drag

Preformed sandor ceramic tubes

(a)

(b)

(c)


(Figure 2.3a) is understandable from the pointof view of minimizing the danger of run-outs.A leak at the joint, or a burst mould is a possibledanger and a definite economic loss. This wasan important consideration for hand-mouldedgreensand, where the moulds were rather weak(and was of course the reason for the use of thesteel moulding box or flask). However, theplacing of the mould cavity below the runnercauses an uncontrolled fall into the mould cav-ity, creating the risk of imperfect castings. It isno longer such a danger for the dense, stronggreensand moulds produced from modernautomatic moulding machines, nor for theextremely rigid moulds created in chemicallybonded sands. For products whose reliabilityneeds to be guaranteed, the arrangement of therunner at the lowest level of the mould cavity,causing the metal to spread through the runningsystem and the mould cavity only in an uphill

direction is a challenge that needs to be met(Figure 2.22). Techniques to achieve this includethe clever use of a core (Figure 2.21b) or forsome hollow castings the use of central gating(Figures 2.23 and 2.24b).

Figure 2.22 An external running system arrangedaround an automotive sump (oil pan).

Figure 2.23 Cross-section of aninternal running system for thecasting of a cylinder.

Figure 2.24 Ring casting produced using (a) anexternal and (b) an internal filling system.


Offset step basin(with open delivery side)

Radiused sprue entrance

Annular feeder(if necessary)

Sprue contained incentre core

Cylinder casting

Slot gates formed in mould

‘Spider’ of radial runners

(a)

(b)


Webster (1964) carried out some earlyexploratory experiments to determine optimumrunner sizes. We can summarize his results interms of the comparative areas of the runner/sprue exit. He found that a runner that has onlythe same area as the sprue exit (ratio 1) will havea metal velocity that is high. A ratio of 2 heclaims is close to optimum since the runner fillsrapidly and excludes air bubbles reasonablyefficiently. A ratio of 3 starts to be difficult tofill; and a ratio of 4 is usually simply wasteful formost castings. Webster's work was a prophesy,foretelling the dangers of large runners thatfoundries have, despite all this good advice,continued to use.

For the best results, however, recent carefulstudies have made clear that even the expansionof the area of flow by a factor of 2 is not easy toachieve without a serious amount of surfaceturbulence. This is now known from videoX-ray radiographic studies, and from detailedexamination of the scatter of mechanical prop-erties of castings using highly sensitive Weibullanalysis.

The best that can easily be achieved withoutdamage is merely the reduction of about20 per cent in velocity by the friction of thesprue/runner bend, necessitating a 20 per centincrease in area of the runner as has been dis-cussed above. Any greater expansion of therunner will cause the runner to be incompletelyfilled, and so permit conditions for damage.

Greater speed reductions, and thus greateropportunities for expansion of the runner occurif the number of right-angle bends is increased,since the factor of 0.8 reduction in speed iscumulative from one bend to the next. Afterthree such bends the speed is reduced by half(0.8� 0.8� 0.8� 0.5). Right-angle bends wereanathema in filling system designs when largecross-sections were the norm. However, withvery narrow systems, there is less room forsurface turbulence. Even so, great care has to betaken. For instance video X-ray studies haveconfirmed that the bends operate best if theirinternal and external radii provide a parallelchannel. The lack of an external radius cancause a reflected wave in larger channels.

One of the most effective devices to reducethe speed of flow in the runner is the use of afilter. The close spacing of the walls of itscapillaries ensures a high degree of viscous drag.Flow rate can often be reduced by a factor of4 or 5. This is a really valuable feature, andactually explains nearly all of the beneficialaction of the filter (i.e. when using good qualitymetal in a well-designed filling system the filterdoes very little filtering. Its really importantaction in improving the quality of castings is its

reduction of velocity). The use of filters isconsidered later (Section 2.3.6).

There has over the years been a considerableinterest in the concept of the separation of sec-ond phases in the runner. Jeancolas et al. (1969)carried out experiments on ferrous metals toshow that at Reynold's numbers below therange 7000±12 000, suspended particles of alu-mina could be deposited in the runner but atvalues in excess of 15 000, they could not pre-cipitate. Although these findings underline theimportance of working with the minimum flowvelocities wherever possible, it is quicklyshown that for a steel casting of height 1 m,giving a velocity of flow of 4.5 m sÿ1, forZ� 5.5� 103 N s mÿ2, and for a runner of80 mm square, Re is over 100 000. Thus it seemsthat conditions for the deposition of solidmaterials such as sand and refractory particlesin runners will not be easily met. Even so, everycast iron foundry worker knows that slag willaccumulate on the tops of runners, where it ismuch to be preferred than in the casting.Separation in this case happens because of thegreat difference in density between the slag andthe metal, and because of the large size of theslag droplets. Thus there are some conditions inwhich a slow runner speed is valuable to assistcleaning the metal.

If there is a choice, the runner should bemoulded in the lower half of the mould (thedrag). As emphasized previously, this willencourage the runner to fill completely prior torising through the gates (moulded for preferencein the cope) and into the mould cavity.

The basic plan of the filling design starts tobecome clear: the metal arrives in some chaos atthe bottom of the sprue. Here, after this initialtrauma, it is gathered together once again by theintegrating action of a feature such as a filter toprovide some delay and back-pressure, afterwhich it is allowed to rise steadily againstgravity, filling section after section of the run-ning system, and finally arriving in the mould ingood order at a speed below the critical velocity.

It should be noted that such a logical systemand its consequential orderly fill is not to betaken for granted. For instance, a usual mistakeis to mould the runner in the cope. This ismainly because the gates, which are in either thedrag or the cope, will inevitably start to fill andallow metal into the mould cavity before therunner is full, as is clear from Figure 2.3a. Thetraditional running of cast iron in this way failsto achieve its potential in its intended separationof metal and slag. This is because the first metaland its load of slag enters the gates immediately,prior to the filling of the runner, and thus priorto the chance that the slag can be trapped



against the upper surface of the runner. In short,the runner in the cope results in the violation ofthe fundamental `no fall' criterion. The runnerin the cope is not recommended for any type ofcastingÐnot even grey iron!

In gravity die castings the placing of therunner in the cope, and taking off gates on thedie joint (Figure 2.3a), is especially bad. This isbecause the impermeable nature of the die pre-vents the escape of air and mould gases from thetop of the runner. Thus the runner never prop-erly fills. The entrapped gas floating on thesurface of the metal will occasionally dislodge,as waves race backwards and forwards along therunner, and as the gases heat up and expand.Large bubbles will therefore continue to migratethrough the gates from time to time throughoutthe pour, and possibly even afterwards. Becauseof their late arrival, it is likely that not only willbubble trails and splash problems occur, butalso the advancing solidification front will trapwhole bubbles.

This scenario is tempered if a die joint isprovided along the top of the runner to allowthe escape of air. Alternatively, a sand core sitedabove the runner can help to allow bubbles todiffuse away.

Even so, the complexity of behaviour of somefilling system designs is illustrated by a runner ina gravity die, positioned in the cope, that actedto reduce the bubble damage in the casting(Figure 2.25). This result, apparently in com-plete contradiction to the behaviour describedabove, arose because of the exceptionally tallaspect ratio of the runner, which was shaped likea vertical slot. This shape retained bubbles highabove the exits to the gates moulded below. Infact it seems that the reduction in bubbles intothe casting by placing the gates low in this wayonly really resulted because of the extremelypoor front end of the filling system. This was abubble-producing design, so that almost anyremedy had a chance to produce a better result.

However, there is a real benefit to be noted(running systems are perversely complicated)because the gates would prime slowly as a headof metal in the runner was built up, thusavoiding any early jetting through into themould cavity. This is a benefit not to beunderestimated, and highlights the problem ofgeneralizing for complex geometries of castingsand their filling systems that can sometimescontain not just liquid metal but sometimesemulsions of slag and/or air.

The tapered runner

It is salutary to consider the case where therunner has two or more gates, and where thestepping or tapering of the runner has beenunfortunately overlooked. The situation isshown in Figure 2.26a. Clearly, the momentumof the flowing liquid causes the furthest gate,number 3, to be favoured. The rapid flow pastthe opening of gate 1 will create a reduced-pressure region in the adjacent gate at this point,

Figure 2.25 Tall slot runner with bottom gates.

Figure 2.26 (a) An unbalanced delivery of melt intothe mould as a result of an incorrect runner design;(b) a tolerably balanced system.


Runner

Gates intomould cavity

1 2 3

1 2 3

(a)

(b)


drawing liquid out of the casting! The flow maybe either in or out of gate 2, but at such areduced amount as to probably be negligible. Inthe case of a non-tapered runner it would havebeen best to have only gate 3.

Where more than one gate is attached to therunner, the runner needs to be reduced in cross-section as each gate is passed, as illustrated inFigure 2.26b. In the past such reductions haveusually been carried out as a series of steps,producing the well-known stepped runnerdesigns. For three ingates the runner would bereduced in section area by a step of one third theheight of the runner as each gate was passed.However, real-time X-ray studies have notedhow during the priming of such systems,because of the high velocity of the stream, thesteps cause the flow to be deflected, leaping intothe air, and ricocheting off the roof of the run-ner. Needless to say, the resulting flow washighly disturbed, and did not achieve its inten-ded even distribution. It has been found thatsimply reducing the cross-section of the runnergradually, usually linearly, cures the deflectionproblem. A smooth, straight taper geometrydoes a reasonable job of distributing the flowevenly (Figure 2.26b).

Kotschi and Kleist (1979) allow a reductionin the runner area of just 10 per cent more thanthe area of the gate to give a slight pressuriza-tion bias to help to balance the filling of thegates. However, they used a highly turbulentnon-pressurized system that will not haveencouraged results of general applicability. Incontrast, computer simulation of the narrowrunners recommended in this work has shownthat the last gate suffers some starvation as aresult of the accumulation of friction along thelength of the runner. Thus for slim systems thefinal gates require some additional area, notless. The author usually provides for this in anad hoc way by simply extending the runner pastthe final gate, and providing a linear taper tothis more distant point (Figure 2.26b). The tapercan, of course, be provided horizontally orvertically (an important freedom of choice oftenforgotten).

Finally, avoid tapering the runner to zero.The thinning section adds no advantage but toprovide points on which people keep stabbingthemselves in the foundry. It aids safety in theworkplace to stop the taper at about 5 mm sec-tion thickness.

The expanding runner

In an effort to slow the metal in its early pro-gress in the runner, a number of methods ofexpanding the area of the runner have been

tried. The simple expansion of the runner at anarbitrary location along the runner is of no useat all (Figure 2.27a). The melt progresses with-out noticing the expansion. Even expanding therunner directly from the near side of the sprue(shown as having a square section for clarity) isnot helpful (Figure 2.27b). However, expandingthe runner from the far side of the sprue (Figure2.27c) does seem to work considerably better.Even here, however, the front tends to progressin two main streams on either side of the central

Figure 2.27 Plan views of a square section sprueconnected to a shallow rectangular runner showingattempts to expand the runner (a and b) that fail com-pletely. Attempt (c) is better, but flow ricochets off thewalls generates a central starved, low pressure region;(d) a slot sprue and slot runner produce a uniform flowdistribution in the runner shown in (e) (recommended)and (f) (probably acceptable).


(a)

(b)

(c)

(d)

(e)

(f)


axis of the runner, leaving the centre empty, orrelatively empty, forming a low-pressure regionsome distance down-stream in the runner. Thisdevelopment of this double jet flow seems to bethe result of the attempted radial expansion ofthe flow as it impacts on the runner, but findsitself constrained by and reflected from thewalls of the runner. This situation for high-temperature liquids such as irons and steelsleads to the downward collapse of the centre ofthe runner in sand moulds, since this becomesheated by radiation, and so expands, but isunsupported by the pressure of metal. Theclosing down of the runner in this way can beavoided by a central moulded support, effec-tively separating the runner into two separate,parallel runners. In practice I find that a slotrunner about 100 mm wide for irons and steels isclose to the maximum that can resist collapse.

A further pitfall for the unwary is the possibleconstricting effect that sometimes occurs as aresult of attempting to connect a round orsquare section sprue on to a thin flat runner(Figure 2.28). For instance, if the runner werepaper thin the constriction at the exit of thesprue would be nearly total; only a fraction ofthe flow would be able to squeeze into the nar-row runner. To eliminate a constriction at thispoint the runner may need to be thickened, or,preferably, the fillet radius at the bend mayrequire to be increased.

Even so, ultimately, it may come as somesurprise to the reader to learn that the linking ofa round or square section sprue to a slot runner,especially when attempting an expansion of therunner to reduce the velocity of the liquid, is notyet developed. To the author's knowledge,techniques for the satisfactory design of thisjunction do not yet exist. Some limited expan-sion in a horizontal plane might be achievable asindicated in Figure 2.28, but should probably beaccompanied by at least a partial correspondingreduction in the vertical plane (not shown inFigure 2.28). The reduction in velocity wouldbenefit from the friction provided by the extrasurface area, but would probably not be suc-cessful to fill an expansion of a factor of 2. Thusthe effect is of limited value. More research isrequired to evaluate what can be achieved bycareful runner design.

What seems more certain, is that the dis-tribution of flow would be simpler if a narrowslot sprue were simply to turn to link onto ahorizontal slot type of runner (Figure 2.27dand e). The more uniform action of friction mightassist better to achieve a modest expansion andcorresponding speed reduction. This has yet to betested. Even so, the use of slot sprues linked toslot runners promises to be a complete solution

to the problem of the sprue/runner junction anddeserves wider exploration.

2.3.2.6 Gates

Siting

When setting out the requirements for the site ofa good gate, it is usual to start with the questions

Figure 2.28 Potential constriction to flow at the sprueto runner junction.


Runner

Sprue

Fillet radius

Min area

Min length � min area


`Where can we get the gate on?'

and

`Where can we get the gate off?'

Other practical considerations include

`Gate on a straight side if possible'

and

`Locate at the shortest flow distance to thekey parts of the casting'.

This is a good start, but, of course, just the start.There are many other aspects to the design of agood gate.

Direct and indirect

In general, it is important that the liquid metalflows through the gates at a speed lower thanthe critical velocity so as to enter the mouldcavity smoothly. If the rate of entry is too high,causing the metal to fountain or splash, then thebattle for quality is probably lost. The turbu-lence inside the mould cavity is the most seriousturbulence of all. Turbulence occurring early inthe running system may or may not producedefects that find their way into the castingbecause many bifilms remain attached to thewalls of the runners and many bubbles escape.However, any creation of defects in the mouldcavity causes unavoidable damage to thecasting.

One important rule therefore follows verysimply:

Do not place the gate at the base of the down-runner so that the high velocity of the fallingstream is redirected straight into the mould, asshown in Figures 2.3c, 2.5 and 2.7b. In effect,this direct gating is too direct. An improved,somewhat indirect system is shown in Figures2.3b and 2.6, illustrating the provision of aseparate runner and gate, and thus incorporat-ing a number of right-angle changes of directionof the stream before it enters the mould. Theseprovisions are all used to good effect in reor-ganizing the metal from a chaotic mix of liquidand gases into a coherent moving mass of liquid.Thus although we may not be reducing theentrainment of bifilms, we may at least be pre-venting bubble damage in the mould cavity.

As we have mentioned above, all of the oxi-des created in the early turbulence of the prim-ing of the running system do not necessarily findtheir way into the mould cavity. Many appear to`hang up' in the running system itself. Thisseems especially true when the oxide is strong asis known to be the case for Al alloys containingBe. In this case the film attached to the wall of

the running system resists being torn away, sothat such castings enjoy greater freedom fromfilling defects. The wisdom of lengthening therunning system, increasing friction, especially bythe use of right-angle bends, adds back-pressurefor improved back-filling and reduces velocity.It also provides more surface to contain andhold the oxides generated during priming.

Total area of gate(s)

A second important rule concerns the sizing ofthe gates. They should be provided with suffi-cient area to reduce the velocity of the melt tobelow the critical velocity of about 0.5 m sÿ1. Theconcept is illustrated in Figure 2.1. Occasionally,the author has permitted himself the risk of avelocity up to 1 m sÿ1 and has usually achievedsuccess. However, velocities above 1.2 m sÿ1 forAl alloys always seem to give problems. Velo-cities of 2 m sÿ1 in film-forming alloys, unlessonto a core as explained below, would beexpected to have consequences sufficiently ser-ious that they could not be overlooked. Witheven higher velocities the problems simplyincrease.

Occasionally, there is a problem obtaining asufficient size of gate to reduce the melt speed tosafe levels before it enters the mould cavity. Insuch cases it is valuable if the gate opens at rightangles onto a thin (thickness a few millimetres)wall. This is because the melt is now forced tospread sideways from the gate, and suffers nosplashing problems because the section thick-ness of the casting is too small. As it spreadsaway from the gate it increases the area of theadvancing front, thereby reducing its velocity.Thus by the time the melt arrives in a thickersection of the casting it is likely to be moving ata speed below critical. In a way, the techniqueuses the casting as an extension of the fillingsystem.

This is a good reason for gating direct onto acore. This, once again, is contrary to conven-tional wisdom. In the past, gating onto a corewas definitely bad because of the amount of airentrained in the flow. The air-assisted hammeraction and oxidation of the binder thus led tosand erosion. With a good design of filling sys-tem, however, in which air is largely excluded,the action of the hot metal is safe. Little or nodamage is done despite the high velocity of thestream, because the melt merely heats the corewhile exerting a steady pressure that holds thecore material in place. Thus with a good fillingsystem design, gating directly onto a core isrecommended.

Returning to the usual gating problemwhereby the gate opens into a large-section



casting. If the area of the gate is too small thenthe metal will be accelerated through, jettinginto the cavity as though from a hosepipe.Figure 2.1c shows the effect. In many castingsthe jet speed can be so high that the metaleffectively blasts its way around the mouldcavity. Historically, many castings have beengated in this way. At the present time most steelsand grey cast iron appear to be cast with thistechnique. The approach has enjoyed tolerablesuccess while greensand moulding has beenemployed, but it seems certain that better cast-ings and lower scrap rates would have beenachieved with less turbulent filling. In the case ofcores and moulds made with resin binders thatcause graphitic films on the liquid iron, thepressurized system is usually unacceptable. Thesame conclusion is true for ductile irons in alltypes of moulds.

We may define some useful quick rules fordetermining the total gate area that is needed.For an Al alloy cast at 1 kg sÿ1 assuming a densityof approximately 2500 kg mÿ3 and assuming thatwe wish the metal to enter the gate at its criticalspeed of approximately 0.5 m sÿ1 it means weneed approximately 1000 mm2 of gate area. Theelegant way to describe this interesting ingateparameter is in the form of the units of areaper mass per second; thus for instance`1000 mm2 kgÿ1 s'.

Clearly we may pro-rata this figure in dif-ferent ways. If we wished to fill the casting attwice this rate (i.e. in half the time) we wouldrequire 2000 mm2 and so on. It can also be seenthat the area is quickly adjusted if it is decidedthat the metal can be allowed to enter at twice thespeed, thus the 1 kg sÿ1 would require only500 mm2, or if directly onto a core in a thinsection casting, perhaps twice the rate onceagain, giving only 250 mm2.

Allowing for the fact that denser alloys suchas irons, steels and copper-based alloys, have adensity approximately three times that of alumi-nium, but the critical velocity is slightly smaller at0.4 m sÿ1, the ingate parameter becomes, withsufficient precision, 500 mm2 kgÿ1 s.

The values of approximately 1000 mm2 kgÿ1 sfor light alloys and 500 mm2 kgÿ1 s for densealloys are useful parameters to commit tomemory.

Gating ratio

In its progress through the running systemthe metal is at its highest velocity as it exits thesprue. If possible, we aim to reduce this in therunner, and further reduce as it is caused toexpand once again into the gates. The aim is toreduce the velocity to below the entrainment

threshold (the 0.4 or 0.5 m sÿ1) at the point ofentry into the mould cavity.

It is worth spending some time belowdescribing an alternative method of definingrunning systems which is widely used, buterroneous. It is to be noted that it is notrecommended!

It has been common to describe runningsystems in terms of ratios based on the area ofthe exit of the sprue. For instance, a widelyused area ratio of the sprue/runner/gates hasbeen 1 : 2 : 4. Note that in this abbreviatednotation the ratios given for both the runnerand the gates refer back to the sprue, so that fora sprue exit of 1, the runner area is 2 and thetotal area of the gates is 4. It is clear that suchratios cannot always be appropriate, and thatthe real parameter that requires control isthe velocity of metal entering the mould. Thuson occasions this will result in ratios of 1 : 5 : 10and other unexpected values. The design ofrunning systems based on ratios is therefore amistake.

Having said this, I do allow myself to use theratio of the area of sprue exit to the (total) areaof the gates. Thus if the sprue is 200 mm tall(measured of course from the top of the metallevel in the pouring basin) the velocity at its basewill be close to 2 m sÿ1. Thus a gate of fourtimes this area will be required to get to below0.5 m sÿ1. (Note therefore that the old 1 : 2 : 4and 1 : 4 : 4 ratios can be seen to be applicableonly up to 200 mm sprue height. Beyond thissprue height the ratios are insufficient to reducethe speed below 0.5 m sÿ1.)

I am often asked what about the problemthat occurs when the mould cross-sectional areareduces abruptly at some higher level in themould cavity. The rate of rise of the metal willalso therefore be increased suddenly, perhapsbecoming temporarily too fast, causing jettingor fountaining as the flow squeezes through theconstriction. Fortunately, and perhaps surpris-ingly, this is extremely rare in casting geome-tries. In forty years dealing with thousands ofcastings I have difficulty recalling whether thishas ever happened. The most narrow area isusually the gate, so the casting engineer candevote attention to ensuring that the criticalvelocity is not exceeded at this critical location,and at the location just inside the mould becauseof the sideways spreading flow (see below). Ifthe velocity in these two situations is satisfac-tory it usually follows that the velocity is satis-factory at all other levels in the casting.

Even in a rare situation where a narrowing ofthe mould is severe, it would still be surprising ifthe critical velocity were exceeded, because thevelocity of filling is at its highest at the ingate,



and usually decreases as the metal level rises,finally becoming zero when the net head is zero,as the metal reaches the top of the mould.

Once again, of course, counter-gravity fillingwins outright. In principle, and usually withsufficient accuracy in practice, the velocities canbe controlled at every level of filling.

Multiple gates

Premature filling problem via early gates Sutton(2002) applied Bernoulli's theorem to drawattention to the possibility that a melt travellingalong a horizontal runner will partly enter ver-tical gates placed along the length of the runner,despite the fact that the runner may not yet havecompletely filled and pressurized (Figure 2.29).This arises as a result of the pressure gradientalong the flow, and is proportional to the velo-city of flow. In real casting conditions, the meltmay rise sufficiently high in such gates thatcavities attached to the gates might be partiallyfilled with a slow dribble of upwelling metalprior to the filling of the runner, and thereforeprior to the main flow up the vertical gates.These dribbles of metal in the cavity are poorlyassimilated by the arrival of the main metalsupply, and so usually constitute a lap defectresembling a misrun or part-filled casting.

This same effect would be expected to be evenmore noticeable in horizontal gates moulded inthe cope, sited above a runner moulded in thedrag (Figure 2.3b). The head pressure requiredto simply cross the parting line and start anunwanted early filling of part of the mouldcavity would be relatively small, and easilyexceeded.

Horizontal velocity in the mould When calcu-lating the entry velocity of the metal through thegates, it is easy to overlook what happens tothe melt once it starts to spread sideways intothe mould cavity. The horizontal sidewaysvelocity away from the gate can sometimes behigh. In many castings where the ingate enters a

vertical wall the transverse spreading speedinside the mould is higher than the speedthrough the gate, and causes a damaging splashas the liquid hits the far walls (Figure 2.30). Wecan make an estimate of this lateral velocity VL

in the following way.The lateral travel of the melt will normally be

at about the height h of a sessile drop. (In a thinwall the height of the flow might reach 2h,reducing the problem considered below. We shallneglect this complication, and consider onlythe worst case.) We shall assume the sectionthickness t, for a symmetrical ingate, area Ai.The melt enters the ingate at the critical velocityVC, and spreads in both directions away fromthe gate. Equating the volume flow ratesthrough the gate and along the base of thecasting gives

VC � Ai � 2 VL � h � tIf we limit the gate velocity and the transversevelocity to the same critical velocity VC (forinstance 0.5 m sÿ1) and adopt a gate thickness tthe same as that of the casting wall, the relationsimplifies to the fairly self-evident geometricalrelation in terms of the length of the ingate Li

Li � 2h

The message from this simple formula is thatif the length of the gate exceeds twice the heightof the sessile drop, even if the gate velocity isbelow the critical velocity, the transverse velo-city may still be too high, and surface turbulencewill result from the impact of the transverse flowon the end walls of the mould cavity.

To be sure of meeting this condition, there-fore, for aluminium alloys where h� 13 mm,gates must always be less than 26 mm wide(remembering that this applies only to gates thathave the same thickness as the wall of the cast-ing). For irons and steels gates should notexceed 16 mm wide. Since we often require areas

Figure 2.29 The partial filling of vertical ingates alongthe length of a runner. Figure 2.30 Sideways flow inside mould cavity.


V1V2

L1

h

Area A1

t

Hydraulic gradient


considerably greater than can be provided bysuch a short gate, it follows that multiple gatesare required to achieve the total ingate area tobring the transverse velocity below critical.

Clearly, it is a concern that in practice, gatelengths are often longer than these limits andmay be causing quality problems from thisunsuspected source.

For those conditions where more length (orarea) is needed than the above formula willallow, the solution is the provision of moregates. Two equally spaced gates of half thelength will halve the problem, and so on. In thisway the individual gates lengths can be reduced,reducing the problem correspondingly. Ourrelation becomes simply for N ingates of totalingate length Li

Li=N � 2h

Or directly giving the number of ingates N thatwill be required

N � Li=2h

These considerations based on velocity throughthe gate or in the casting take no account ofother factors that may be important in somecircumstances. For instance, the number ofingates might require to be increased to (i) dis-tribute heat more evenly throughout the mould;(ii) avoid localized hot spots as a result ofjunction problems (see below); and (iii) provideliquid at all the lowest points in the mould cavityto avoid waterfall effects.

Junction effect When the gates are planted onthe casting they create a junction. This self-evident statement requires explanation.

Some geometries of junction create the dan-ger of a hot spot. The result is that a shrinkagedefect forms in the pocket of liquid that remainstrapped here at a late stage of freezing. Thuswhen the gate is cut off a shrinkage cavity isrevealed underneath. This defect is widely seenin foundries. In fact, it is almost certainly thereason why most traditional moulders cut suchnarrow gates, causing the metal to jet into themould cavity with consequent poor results tocasting quality.

The magnitude of the problem dependsstrongly on what kind of junction is created.Figure 2.31 shows the different kinds of junc-tions. An in-line junction (c) is hardly more thanan extension of the wall of the casting. Very littlethermal problem is to be expected here. TheT-junction (a) is the most serious problem. It isdiscussed below. The L-junction (b) is anintermediate case and is not further discussed.

The reader can make his or her own allowancesassuming conditions intermediate between thezero (in-line junction) and T-junction cases.

To help to solve this problem it is instructive toexamine the freezing patterns of T-sections. Inthe 1970s Kotschi and Loper carried out someadmirable theoretical studies of T-junctionsas shown in Figure 2.32 using only simple

Figure 2.31 Maximum allowable gate thickness toavoid a hot spot at the junction with the casting.

Figure 2.32 Geometry of a T-shaped junction.



calculations based on modulus. These studiespointed the way for experimental work byHodjat and Mobley in 1984 that broadly con-firmed the predictions. The data are interpretedin Figure 2.33 simply as a set of straight linesof slopes 2/1, 1/1, and 1/2. (A study of the scatterin the data shows that the predictions are notinfallibly correct in the transitional areas, sothat some caution is required.)

Figure 2.34 presents a simplified summary ofthese findings. It is clear that a gate (the uprightleg of the T) of 1 : 1 geometry, i.e. a section equalto the casting section (the horizontal arm of theT), has a hot spot in the junction, and so isundesirable. In fact, Figure 2.33 makes it clearthat any medium-sized gate less than twice asthick, or more than half as thick, will give a

troublesome hot spot. It is only when the gate isreduced to half or less of the casting thicknessthat the hot spot problem is removed. (Otherlessons can be learned from the T-junctionresults: (1) an appendage of less than one half ofthe section thickness will act as a cooling fin,locally enhancing the rate of cooling in themanner of a metal chill in a sand mould; and(2) an appendage of section double that of thecasting will freeze later without a hot-spot pro-blem. This is the requirement for a feeder whenplanted on a plate-like casting, as will be dis-cussed in Chapter 6.

In the case of gates forming T-junctions withthe casting (Figure 2.34), the requirement tomake the gate only half of the casting thicknessensures that under most circumstances no

Figure 2.33 Solidification sequence forT-shaped castings (A� arm, J� junction,L� leg). Experimental data from Hodjat andMobley (1984).

Figure 2.34 Array of different T-junctions.


t = 2 t = 1 t = ½ t → 0Lastmetalto solidify

Neutral junction (usefulas a

feeder)

Neutral junction Cold junction (Cooling fin)

Hot spot (Avoid)

Pore

Good options for ingates


localized shrinkage defect will occur, and almostno feeding of the casting will take place throughthe gate.

For slot gates much less than half the sectionof the casting that act as cooling fins the effectcan be put to good use in setting up a favourabletemperature gradient in the casting, encoura-ging solidification from the gate towards the topfeeder. Such cooling fin gates have been used togood effect in the production of aluminium andcopper-based alloys because of their high ther-mal conductivity (Wen et al. 1997). (The effect ismuch less useful in irons and steels.) Also, thecommon doubt that the cooling effect would becountered by the preheating because of the flowof metal into the casting is easily demonstratedto be, in most cases, a negligible problem. Thepreheating occurs only for a relatively shorttime compared to the time of freezing of thecasting, and the thin gate itself has little thermalcapacity. Thus following the completion of itsrole as a gate it quickly cools and converts toacting as a cooling fin. Where a gate cannotconveniently be made to act as a cooling fin, theauthor has planted a cooling fin on the sides ofthe gate. (This is simple if the slot gate is on ajoint line.) By this means the gate is stronglycooled, and in turn, cools the local part of thecasting.

The current junction rules have been statedonly in terms of thickness of section. For gatesand casting sections of more complex geometryit is more convenient to extend the rules, repla-cing section thickness by equivalent modulus.The more general rule that is inferred is `the gatemodulus should be half or less than the localcasting modulus'.

It is worth drawing attention to the fact thatnot all gates form T-junctions with the casting.For instance, those that are effectively merelyextensions of a casting wall may clearly becontinued on at the full wall thickness withoutany hot-spot effect (Figure 2.31c).

Gates which form an L-junction with the wallof the casting are an intermediate case (Sciama1974), where a gate thickness of 0.75 times thethickness of the wall is the maximum allowablebefore a hot spot is created at the junction(Figure 2.31b).

It is possible that these simple rules may bemodified to some degree if much metal flowsthrough the gate, locally preheating this region.In the absence of quantitative guidelines on thispoint it is wise to provide a number of gates,well distributed over the casting to reduce suchlocal overheating of the mould. The Cosworthsystem devised by the author for the ingating ofcylinder heads used ten ingates, one for everybolt boss. It contrasted with the two or three

gates that had been used previously, and at leastpartly accounts for the immediate success of thegating design (although it did not help cut-offcosts, of course!).

If the casting contains heavy sections thatwill require feeding then this feed metal willhave to be provided from elsewhere. It isnecessary to emphasize the separate roles of (i)the filling system and (ii) the feeding system.The two have quite different functions. In theauthor's experience attempts to feed the castingthrough the gate are to be welcomed if reallypossible; however, there are in practice manyreasons why the two systems often work betterwhen completely separate. They can then beseparately optimized for their individual roles.

It is necessary to make mention of someapproaches to gating that attempt to evaluatethe action of running systems with gates thatoperate only partly filled (Davis 1977). Thereader will confirm that such logic only appliesif the gates empty downhill into the mould, likewater spilling over a weir. This is a violation ofone of our most important filling rules. Thusapproaches designed for partially filled gates arenot relevant to the technique recommended inthis book. The placing of gates at the lowestpoint of the casting, and the runner below that,ensures that the runner fills completely, thenthe gates completely, and only then can thecasting start to fill. The complete prior filling ofthe running system is essential; it avoids thecarrying through of pockets of air as wavesslop about in unfilled systems. Complete fillingeliminates waves.

As in most foundrywork, curious prejudicescreep into even the most logical approaches.In their otherwise praiseworthy attempt to for-malize gating theory, Kotschi and Kleist (1979)omit to limit the thickness of their gates toreduce the junction hot spot, but curiouslyequalize the areas of the gates so as to equalizethe flow into the casting. In practice, making thegates the same is rarely desirable because mostcastings are not uniform. For instance, a doubleflow rate might be required into part of thecasting that is locally twice as heavy.

The design of gates may be summarized inthese concluding paragraphs.

The requirement for gates to be limited to amaximum thickness naturally dictates that thegates may have to become elongated into a slot-type shape if the gate area is also required to belarge. Limitations to the length of the slots tolimit the lateral velocities in the mould may berequired of course, dictating more than one gateas explained above. The limitation of lateralvelocities in addition to ingate velocities is avital feature.



The slot form of the gates is sometimesexasperating when designing the gating systembecause it frequently happens that there is notsufficient length of casting for the requiredlength of slot! In such situations the castingengineer has to settle for the best compromisepossible. In practice, the author has found thatif the gate area is within a factor of 2 of the arearequired to give 0.5 m sÿ1, then an aluminiumalloy casting is usually satisfactory. Any furtherdeviation would be cause for concern. Greyirons and carbon steels are somewhat moretolerant of higher ingate velocities.

As a final part of this section on gating, itis worth examining some traditional gatingdesigns.

The touch gate The touch gate, or kiss gate, isshown in Figure 2.35a. As its name suggests, itonly just makes contact between the source ofmetal and the mould cavity. In fact there is nogate as such at all. The casting is simply placedso as to overlap the runner. The overlap istypically 0.8 mm for brass and bronze castings(Schmidt and Jacobson 1970; Ward and Jacobs1962) although up to 1.2 mm is used. Over2.5 mm overlap causes the castings to be diffi-cult to break off, negating the most importantadvantage. The elimination of a gate in the casedescribed by the authors was claimed to allowbetween 20 and 50 per cent more castings in amould. Furthermore, the castings are simplybroken off the runner, speeding production andavoiding cut-off costs and metal losses fromsawing. The broken edge is so small that formost purposes dressing by grinding is notnecessary; if anything, only shot blasting isrequired.

A further benefit of the touch gate is that acertain amount of feeding can be carried outthrough the gate. This happens because (1) thegate is preheated by the flow of metal through it,and (2) the gate is so close to the runner andcasting that it effectively has no separate exist-ence of its own; its modulus is not that of atiny slot, but some average between that of therunner and that of the local part of the castingto which it connects. Investigations of touchgate geometry have overlooked this point, withconfusing results. More work is needed to assesshow much feeding can actually be carried out.The result is likely to be highly sensitive to alloytype so that any study would benefit from theinclusion of short and long freezing rangealloys, and high and low conductivity metals.

Ward and Jacobs report a reduced incidenceof mis-run castings when using touch gating.This observation is almost certainly the result of

the beneficial effect of surface tension control inpreventing the penetration of the gates beforethe runner is fully filled and at least partlypressurized. Only when the critical pressure toforce the metal surface into a single curvature of0.4 mm is reached (in the case of the 0.8 mmoverlap) will the metal enter the mould cavity.This pressure corresponds to a head of 30 to40 mm for copper-based alloys.

With such a thin gate, variations of only0.1 mm in thickness have been found to changeperformance drastically. With the runner in onehalf of the mould and the casting in the other,this is clearly seen to be a problem from smallvariations in mismatch between the mouldhalves. The problem can be countered in prac-tice by providing a small gate attached to thecasting, i.e. in the same mould half as the castingcavity (Figure 2.35b), so that the gate geometry

Figure 2.35 (a) Touch gate, (b) knife gate,(c) pencil gate, (d) normal and reversed horn gates.



is fixed regardless of mismatch. This is some-times called a knife gate.

Although it is perhaps self-evident, touch andknife gates are not viable as knock-off gates onthe modern designs of accurate, thin-walled,aluminium alloy castings. This is simply becausethe gate has a thickness similar to the casting, sothat on trying to break it off, the casting itselfbends! The breaking off technique works onlyfor strong, chunky castings, or for relativelybrittle alloys.

The system was said to be unsuitable foraluminium-bronze and manganese-bronze, bothof which are strong film-forming alloys (Schmidtand Jacobson 1970), although this discouragingconclusion was probably the result of the runnerbeing usually moulded in the cope and thecastings in the drag and a consequence of theirpoor filling system, generating quantities ofoxide films that would threaten to choke gates.The unfortunate fall into the mould cavity wouldfurther damage quality, as was confirmed byWard and Jacobs (1962). They found that uphillfilling of the mould was essential to providinga casting quality that would produce a perfectcosmetic polish.

The system has been studied for a number ofaluminium alloys (Askeland and Holt 1975),although the poor gating and downhill fillingused in this work appears to have clouded theresults. Even so, the study implies that a betterquality of filling system with runner in the dragand casting impressions in the cope could beimportant and rewarding.

The fundamental fear that the liquid may jetthrough the narrow gate may be unfounded. Infact, there may actually be no jetting problem atall. This appears to be a result of the high sur-face tension of liquid metals. Whereas watermight be expected to jet through such a narrowconstriction, liquid aluminium is effectivelycompressed when forced in to any section lessthan its natural sessile drop height of 12.5 mm.The action of a melt progressing through a thingate, equipped with an even thinner sectionformed by a sharp notch was observed for alu-minium alloys in the author's laboratory byCunliffe (1994). The gate was 4 mm thick andthe thickness under the various notches wasonly 1 to 2 mm. The progress of the melt alongthe section was observed via a glass windowfrom above. The metal was seen to approach,cross the notch constriction, and continue on itsway without hindrance, as though the notchconstriction did not exist! This can only beexplained if the melt immediately re-expands tofill the channel after passing the notch. It seemsthe liquid meniscus, acting like a compressed,doubled-over leaf spring, immediately expands

back to fill the channel when the point of highestcompression is passed.

If the surface turbulence through touchgates is tolerable, or minimal, then they deserveto be much more widely used. It would be sowelcome to be able to end the drudgery ofsawing castings off running systems, togetherwith the noise and the waste. With goodquality metal provided by a good front end tothe filling system, and uphill filling of themould cavity after the gate, it seems likely thatthis device could work well. It would probablynot require much work to establish a properdesign code for such a practice.

The pencil gate Many large rolls for a varietyof industries are made from grey cast iron ingreensand moulds. They often contain amassive proportion of grey iron chills aroundthe roll barrel to develop the white iron wearsurface of the roll. It is less common nowadaysto cast rolls in loam moulds produced bystrickling. (Loam is a sand mixture containinghigh percentages of clay and water, like a mud,which allow it to be formed by sleeking intoplace. It needs to be thoroughly dried prior tocasting.) Steel rolls are similarly cast.

Where the roll is solid, it is often bottom-gated tangentially into its base. Where the rollor cylinder is hollow, it may be centrifugallycast, or it may be produced by a special kind oftop gating technique using pencil gates.

Figure 2.35c represents a cross-sectionthrough a mould for a roll casting. Such acasting might weigh over 60 000 kg, and havedimensions up to 5 m diameter by 5 m facelength, with a wall thickness 80 mm (Turner andOwen 1964). It is cast by pouring into an opencircular runner, and the metal is metered intothe mould by a series of pencil gates. The metalfalls freely through the complete height of themould cavity, gradually building up the casting.The metal±mould combination of grey iron ingreensand is reasonably tolerant of surface tur-bulence. In addition, the heavy-section thick-ness gives a solidification time in excess of30 minutes, allowing a useful time for the floatingout and separation of much of the oxideentrained by splashing. The splashing is limitedby the slimness of the falling streams from thenarrow pencil gates.

The solidification geometry is akin to con-tinuous casting. The slow, controlled build-upof the casting ensures that the temperaturegradient is high, and thus favouring good feed-ing. The feeder head on top of the casting istherefore only minimal, since much of the cast-ing will have solidified by the time the feeder is



filled. This beneficial temperature gradient isencouraged by the use of pencil gates: the nar-row falling streams have limited energy and sodo not disturb the pool of liquid to any greatdepth (a single massive stream would be adisaster for this reason).

Top gating in this fashion using pencil gatesis expected to be useful only for the particularconditions of: (i) grey iron; (ii) heavy sections;and (iii) greensand or inert moulds. It is notexpected to be appropriate for any metal±mouldcombinations in which the metal is sensitive tothe entrainment of oxide films, especially in thinsections where entrained material has limitedopportunity to escape.

Even so, this top pouring, although occurringin the most favourable way possible as discussedabove, still results in occasional surface defect inproducts that are required to be nearly defect-free. The use of bottom gating via an excellentfilling system, entering the mould at a tangent tocentrifuge defects away from the outer surfaceof the roll would be expected to yield a superiorproduct. Even vertical-axis centrifugal castingwould benefit from better filling design, apply-ing the liquid metal to the rotating mould in aless turbulent fashion. No matter what thecasting method, there is no substitute for a goodfilling system.

The horn gate The horn gate is a device usedby a traditional greensand moulder to make aquick and easy connection from the sprue intothe base of the mould cavity without the need tomake and fit a core or provide an additionaljoint line (Figure 2.35d). The horn pattern couldbe withdrawn by carefully easing it out of themould, following its curved shape. Although theingenuity of the device can be admired, inpractice it cannot be recommended. It breaksone of our fundamental rules for filling systemdesign by allowing the metal to fall downhill. Inaddition, there are other problems. When usedwith its narrow end at the mould cavity it causesjetting of the metal into the mould. This effecthas been photographed using an open-topmould, revealing liquid iron emerging from theexit of the gate, and executing a graceful arcthrough the air, before splashing into a messy,turbulent pool at the far side of the cavity(Subcommittee T535 1960). It has occasionallybeen used in reverse in an attempt to reduce thisproblem (Figure 2.35b). However the irregularfilling of the first half of the gate by the metalrunning downhill in an uncontrolled fashionand slopping about in the valley of the gate issimilarly unsatisfactory. Furthermore, the largeend junction with the casting now poses the

additional problem of a large hot spot thatrequires to be fed to avoid shrinkage porosity.

The horn gate might be tolerable for greyiron in greensand. Otherwise it is definitely to beavoided.

Vertical gate Sometimes it is convenient toplace a vertical gate at the end of a runner.Whereas the slowing of the flow by expandingthe channel was largely unsuccessful for thehorizontal runner, an upward-oriented expand-ing fan-shaped gate can be extremely beneficialbecause of the aid of gravity. As always, theapplication is not completely straightforward.Figure 2.36a shows that if the fan gate is siteddirectly on top of a rectangular runner, the flowis constrained by the vertical sides of the runner,so that the liquid jets vertically, falling back tofill the fan gate from above. Figure 2.36b showsthat if the expansion of the fan is started from thebottom of the runner, the flow expands nicely,filling the expanding volume and so reducing inspeed before it enters the mould cavity. Thisresult is valuable because it is one of the very fewsuccessful ways in which the speed of the metalentering the mould cavity can be reduced.

The work in the author's laboratory (Rezvaniet al. 1999) illustrates that this form of gateproduces castings of excellent reliability. Com-pared to conventional slot gates, the Weibullmodulus of tensile test bars filled with the nicelydiverging fan gate was raised nearly four times,indicating the production of castings of fourtimes greater reliability.

Itamura and co-workers (2002) have shownby computer simulation that the limiting0.5 m sÿ1 velocity is safe for simple vertical gates,but can be raised to 1.0 m sÿ1 if the gate isexpanded as a fan. However, expansion doesnot continue to work at velocities of 2 m sÿ1

where the flow becomes a fountain. Similarresults have been confirmed in the author'slaboratory by Lai and Griffiths (2003) who usedcomputer simulation to study the expansion ofthe vertical gate by the provision of a generousradius at the junction with the casting. All thesedesirable features involve additional cutting offand dressing costs of course.

Surge control systems The flowing of metalpast the gates and into some kind of dump hasbeen widely used to eliminate the first coldmetal, diverting it away, together with anyinitial contamination by sand or oxide. Whenthe dump is filled the gates can start to fill.If there is any raising of back-pressure as, forinstance, the accumulation of friction along thelength of the runner extension, particularly if



(a)

t = 0.40

t = 0.50 t = 0.60

t = 0.65 t = 0.75

.�� # &) � �� 7' D�� * ��-?�'


t = 0.10 t = 0.25

t = 0.40 t = 0.50

t = 0.75 t = 1.13

(b)

.�� # &) !��


the runner is narrow, then the gates may start tofill earlier, before the dump is fully filled. Thisprinciple is nicely addressed by the use of theBernoulli equation as used by Sutton (2002).The concept is developed further below.

The by-pass principle can be used to generatemore important benefits. It can assist gainingcontrol over the initial velocity of the meltthrough the gates. Usually, the first metalthrough the gate is a transient jet, the metalspurting through when the runner is suddenlyfilled. This is not a problem for castings of smallheight where the jet effect can be negligible, butbecomes increasingly severe for those with ahigh head height. For a tall casting the velocityat the end of the runner is high so the momen-tum of the melt, shocked to an instant stop,causes the metal to explode through the far gate,and enter the cavity like a javelin. This dama-ging initial transient can occur despite the cor-rect tapering of the runner, since the taper is

designed to distribute the flow evenly into themould only after the achievement of steady stateconditions.

The problem can be greatly reduced bydiverting the initial flow away from the casting.The provision of an additional gate at the end ofthe runner, beyond the casting, and not con-nected with the mould cavity, is a valuabletechnique for the reduction of the shock of thesudden filling of the runner and the impact ofmetal through the gates. The design of this flow-off device is capable of some sophistication, andpromises to be a key ingredient, particularly forlarge, expensive one-off castings. This intro-duces the concept of surge control systems.

A gate that channels the initial metal into adump below the level of the runner is probablythe least valuable form of this technique (Figure2.37a). The downward facing gate will continueto fill without generating significant back-pressure, the metal merely falling into the trap,

Figure 2.37 By-pass designs showing (a) and (b) dross trap type (better than nothing, but not especially recommended);(c) non-return trap; (d) vertical runner extension for gravity deceleration; (e) and (f) surge control systems using aterminal vortex; (g) surge control system with in-line vortex with axial (central) outlet.


H

(a) (b)

(c)

(d) (e)

(f) (g)

Net headto fillgate

h

Casting

Vortexsurgecylinder

hh


until the instant the trap is completely filled. Atthat instant, the shock of filling is then likely tocreate, albeit at a short time later, the spurtingaction into the mould that it was designed toavoid. However, by this time the gates andperhaps even the mould are likely to containsome liquid, so there is a chance that any dele-terious jetting action will to some extent besuppressed. This flow-off dump has the benefitof working as a classical dross trap, of course.A taper prior to the trap can prevent the back-wave reversing debris out of the trap, sincethere is only room for the inflow of metal(Figure 2.37c).

An improved form of the device is easilyenvisaged. A gate into a dump moulded abovefrom the runner has a more positive action(Figure 2.37d). It provides a gradual reductionin flow rate along the runner because it gen-erates a gradually increasing back-pressure as itfills, building up its head height. When placed atthe end of the runner, the gate acts to reducetemporarily the speed into the (real) gates byproviding additional gate area, and is valuablefor reducing the unwanted final filling shock bysome contribution to reducing speed. A simplyupturned end to the runner as a runner exten-sion (Figure 2.37b) will help in this way, but itslimited area will mean that it generates its back-pressure too quickly, so any benefit of a slowincrease in speed through the gates is limited.Additional volume of the dump is an advantageto delay the build-up of pressure to fill the gate.

The economically minded casting engineermight find that some castings could be made as`free riders' in the mould at the end of suchgates. The quality may not be high, especiallybecause of the impregnation of the aggregatemould by the momentum of the metal. Even so,the part may be good enough for some pur-poses, and may help to boost earnings permould.

A more sophisticated design incorporates allthe desirable features of a fully developed surgecontrol system. It consists of extending therunner into the base of an upright circularcylinder, entering tangentially (Figure 2.37e andf). The height and diameter of the cylinder arecalculated to raise the back-pressure into thegates at a steady rate (avoiding the applicationof the full head from the filling system) for asufficient time to ensure that the gates and thelower part of the casting are filled. When thecylinder (a kind of vortex dump) is completelyfilled only then does the full pressure of thefilling system come into operation to acceleratethe filling of the mould cavity. The final fillingof the dump may still occur with a `bang'Ðthewater hammer effectÐannounced by the shock

wave of the impact as it flashes back along therunner at the speed of sound. However, thisfinal filling shock will be considerably reducedfrom that produced by the metal impacting theend of a simple closed runner.

Although the device actually controls thespeed of metal through the ingate, it is not calleda speed control since its role is over within thefirst few seconds of the pour. The name `surgecontrol' emphasizes its temporary nature.

An even more sophisticated variant that canbe suggested is the incorporation of the surgecontrol dump in line with the flow from thesprue (Figure 2.37g). The design of the dump asa vortex as before brings additional advan-tages: on arrival at the base of the sprue andturning into the runner at high speed, the speedcreates a centrifugal action. This action isstrongly organizing to the melt, retaining theintegrity of the front rather than the chaoticsplashing that would have occurred in an impactinto a rectangular volume, for instance. Therotary action also centrifuges the entrained air,slag (and possibly some oxides) into the centrewhere they have opportunity to float if thecylinder is given sufficient height. The goodquality melt is taken off from the centre of thebase. The small fall down the exit of the surgecylinder is not especially harmful in this casebecause the rotational action assists the flow toprogress with maximum friction down the wallsof the exit channel. The system acts to take thefirst blast of high-velocity metal, graduallyincreasing the height in the surge cylinder. Inthis way a gradually increased head of metal isapplied to the gates. Furthermore, of course, themetal reaching the gates should be free of airand other low density contaminants.

These surge control concepts promise torevolutionize the production of large steel cast-ings, for which other good filling solutions are,in general, either not easy or not practical. Theby-pass and surge control devices representvaluable additions to the techniques of con-trolling not only the initial surge through thegates, but if their action is extended, as seemspossible, they can also make a valuable con-tribution to slowing velocities during the com-plete vulnerable early phase of filling.

The action of a by-pass to double as aclassical dross trap is described further inSection 2.3.6.

2.3.2.7 Direct gating (from gates)

If the casting engineer has successfully designedthe running system to provide bottom gatingwith minimal surface turbulence, then the cast-ing will fill smoothly without the formation of



film defects. However, the battle for a qualitycasting may not yet be won. Other defects can liein wait for the unwary!

For the majority of castings the gate connectsdirectly into the mould cavity. I call this simply`direct gating'. In most cases it is allowable, ortolerable, but it sometimes causes other prob-lems because of the effect it has on the solidifi-cation pattern of the casting.

Flow channel structure

Consider the direct-gated vertical plate shown inFigure 2.38a. Imagine this casting being filled

slowly to reduce the potential for surface tur-bulence. If the filling rate was reduced to thepoint that the metal just reached the top of themould by the time the metal had just cooled toits freezing point, then it might be expected thatthe top of the casting would be at its coldest, andfreezing would then progress steadily down theplate, from the top to the gate. (At that timethe gate would be assumed to be hot because ofthe preheating effect of the hot metal that wouldhave passed through.) Nothing could be furtherfrom the truth.

In reality, the slow filling of the plate causesmetal to flow sideways from the gate into thesides of the plate, cooling as it goes, and freezingnear the walls. Layers of fresh hot metal wouldcontinue to arrive through the gate. The suc-cessive positions of the freezing front are shownin Figure 2.38. The final effect is a flow pathkept open by the hot metal through a castingthat by now has mainly solidified. Rabinovich(1969) describes these patterns of flow in thinvertical plates, calling them jet streams. Flowchannel is suggested as a good name, if some-what less dramatic. The final freezing of the flowchannel is slow because of the preheated mouldaround the path, and so its structure is coarseand porous. The porosity will be encouraged bythe enhanced gas precipitation under the con-ditions of slow cooling, and shrinkage maycontribute if local feeding is poor because eitherthe flow path is long or it happens to be distantfrom a source of feed metal.

Reducing the subsequent feeding problem ina flow channel by feeding down the channelfrom above, or by limited feeding uphill frombelow, is facilitated in thicker sections where thefeeding distance is greater (see Chapter 6). Thusbottom gating into bosses can take advantage ofthe boss as a useful feed path (Figure 2.38b).However, this action increases the problemswith slower cooling, leading to enhanced gasporosity and coarse structure.

The flow channel structure is a standardfeature of castings that are filled slowly fromtheir base. This serious limitation to structurecontrol seems to have been largely overlooked.

Moreover, the defect is not easily recogniz-able. It can occasionally be seen as a region ofcoarse grain and fine porosity in radiographs oflarge plate-like parts of castings. The structurecontrasts with the extensive areas of clear,defect-free regions of the plate on either side. Itis possible that many so-called shrinkage prob-lems (for which more or less fruitless attemptsare made to provide a solution by extra feedersor other means) are actually residual flowchannels that might be cured by changing ingateposition or size, or raising fill rate. No research

Figure 2.38 (a) Direct bottom-gated vertical plate, and(b) the use of a boss to assist feeding after casting is filled.



appears to have been carried out to guide us outof this difficulty.

Nevertheless, in general, the problem isreduced by filling faster (if that is possiblewithout introducing other problems).

However, even fast filling does not cure theother major problem of bottom gating, which isthe adverse temperature gradient, with thecoldest metal being at the top and the hottest atthe bottom of the casting. Where feeders areplaced at the top of the casting this thermalregime is clearly unfavourable for effectivefeeding. In addition, particularly when solidifi-cation is slow, the problem of convection maybecome important. This serious problem isconsidered later under Rule 7.

2.3.2.8 Indirect gating (from gates)

There is an interesting gating system that solvesthe major features of the flow channel problem.The problem arises because the hot metal that isrequired to fill the casting is gated directly intothe casting and has to travel through the castingto reach all parts.

The solution is not to gate into the casting: amain flow path is created outside the casting. Itis called a riser or up-runner. Metal is thereforediverted initially away from the casting, throughthe riser, only entering subsequently by dis-placement sideways from the riser as fresh sup-plies of hot metal arrive. The fresh suppliesflood up into the top of the riser, ensuring thatthe riser remains hot, and that the hottest metalis delivered to the top of the mould cavity. Thesystem is illustrated in Figure 2.39. The systemhas the special property that the riser and slotgate combination acts not only to fill but also tofeed. (The reader will notice that the use of theterm `riser' in this book is limited to this specialform of feeder which also acts as an `up-runner',in which the metal rises up the height of thecasting. It is common in the USA to refer toconventional feeders placed on the tops ofcastings as risers. However, this terminology isavoided here; such reservoirs of metal are calledfeeders, not risers, following the simple logic ofusing a name that describes their action per-fectly, and does not get confused with other bitsof plumbing such as whistlers, up which metalalso rises!)

The final parts of the casting to fill inFigure 2.39 will probably also require somefeeding. This is easily achieved by planting afeeder on the top of the riser, as a kind of riserextension. This retains all the benefits of thesystem, since its metal is hot, and hotter metalbelow in the riser will convect into the feeder.The disadvantages of the riser and slot gate

system are as follows:

1. The considerable cut-off and finishing pro-blem, since the gate often has to be sited onan exterior surface of the casting, and sorequires much subsequent dressing to achievean acceptable cosmetic finish.

2. There appears to be no method of predictingthe width and thickness of the gate at thepresent time. Further research is requiredhere. In the normal gate where it is requiredto freeze before the casting section to avoidthe hot-spot problem in the junction, the

Figure 2.39 Riser and slot gate to both gate and feeda vertical plate from (a) its side, or (b) its centre.


Hf

(a)

(b)

(c)

L L

L

X


thickness of the gate is held to half of thecasting thickness or less. However, in thiscase the gate is really equivalent to a feederneck, through which feed metal is required toflow until the casting has solidified. Whereasa thickness of double the casting thicknesswould then be predicted to avoid the junctionhot spot under conditions of uniform startingtemperature, the preheating effect of the gatedue to the flow of metal through it mightmean that a gate as narrow as half of thecasting section may be good enough tocontinue feeding effectively. There are, un-fortunately, no confirmatory data on this atthe present time.

It is important to caution against the use of agate which is too narrow for a completely dif-ferent reason; if filling is reasonably fast then theresistance to flow provided by a narrow slot gatewill cause the riser to fill up to a high level beforemuch metal has had a chance to fill the mouldcavity. The dynamics of filling and surface ten-sion, compounded by the presence of a strongoxide film, will together conspire to retain theliquid in the riser for as long as possible. Themetal will therefore spill through the slot intothe casting from an elevated level (the height offall Hf shown in Figure 2.39c). Again, our no-fall rule is broken. It is desirable, therefore, tofill slowly, and/or to have a gate sufficientlywide to present minimal resistance to metalflow. In this way the system can work properly,with the liquid metal in the riser and castingrising substantially together. Metal will thenenter the mould gently.

It is also important for the gate not to bethinner than the casting when the casting wallthickness reduces to approximately 4 mm. Agate 2 mm thick would hold back the metalbecause of the effect of surface tension and thesurface film allowing the metal head to build upin the riser. When the metal eventually breaksthrough, the liquid will emerge as a jet, and falland splash into the mould cavity. For castingsections of 4 mm or less the gate should prob-ably be at least as thick as the wall.

In general, it seems reasonable to assume thatconditions should be arranged so that the falldistance h in Figure 2.39c should be less than theheight of the sessile drop. The fall will then berelatively harmless.

For thinner-section castings (for instance,less than 2 mm thickness) made under normalfilling pressure, the feeding of thin-sectioncastings can probably be neglected (as will bediscussed in Chapter 6). Thus any hot-spotproblems can also be disregarded, with theresult that the ingate can be equal to the casting

thickness. Surface tension controls the entrythrough the gate and the further progress of themetal through the mould cavity, reducing theproblems of surface turbulence. Fill speed cantherefore be increased.

A further important point of detail in Figure2.39 should be noted. The runner turns upwardon entry to the riser, directing the flow upwards.A substantial upward step is required to ensurethis upward direction to the flow. This is asimilar feature to that shown in Figure 2.6 andcontrasts with the poor system shown in Figure2.5. If the provision of this step is neglectedcausing the base of the runner to be level withthe gate and the base of the casting, metalrushing along the runner travels uncheckeddirectly into the mould cavity. A flow pathwould then be set up so that the riser wouldreceive no metal directly, only indirectly after ithad circulated through the casting. The base ofthe casting would receive all the heated metal,and the riser would be cold. Such a flow regimeclearly negates the reason for the provision ofthe system! Many such systems have failedthrough omitting this small but vital detail.

What rates are necessary to make the systemwork best? Again we find ourselves without firmdata to give any guide. We can obtain someindication from the following considerations.

The first liquid metal to flow through the gateand along the base of the cavity travels as astream. Being the first metal travelling over thecold surface of the mould, it is most at risk fromfreezing prematurely. Subsequent flow occursover the top of this hot layer of metal, andtherefore does not lose so much heat from itsundersurface. Thus if we can ensure that con-ditions are right for the first metal to flow suc-cessfully, then all subsequent flow should be safefrom early freezing. In the limiting conditionwhere the tip of the first stream just solidifies onreaching the end of the plate, it will clearly haveestablished the best possible temperature gra-dient for subsequent feeding by directionalsolidification back towards the riser. Sub-sequent layers overlying this initial metal will, ofcourse, have slightly less beneficial temperaturegradients, since they will have cooled less duringtheir journey. Nevertheless, this will be the bestthat we can do with a simple filling method;further improvements will have to await theapplication of programmed filling by pumpedsystems.

Focusing our attention, therefore, on the firstmetal into the mould, it is clear that the problemis simply a fluidity phenomenon. We shallassume that the height of the stream corre-sponds to the height h of the meniscus which canbe supported by surface tension (Figures 2.30).



If the distance to be run from the gate is L, andthe solidification time of the metal is tf in thatsection thickness x, then in the limiting condi-tion where the metal just freezes at the limit offlow (thus generating the maximum temperaturegradient for subsequent feeding):

tf � L=V (2.1)

where V is the velocity of flow (m sÿ1) of themetal stream. The corresponding rate of flow Q(kg sÿ1) for metal of density r (kg mÿ3) is easilyshown to be:

Q � Vhxr

� Lhxr=tf

(2.2)

At constant filling rate the time t to fill a castingof height H is given by:

t � tf �H=h�A 10 mm thick bar (considering the first lengthof melt to travel along the base of the mould) inAl±7Si alloy would be expected to freeze com-pletely in about 40 seconds giving a flow life forthe solidifying alloy of perhaps 20 seconds. Themeniscus height h is approximately 12.5 mm,and so for a casting H� 100 mm high thepouring time would be 8� 20� 160 seconds,or nearly 3 minutes. This is a surprisingly longtime.

This conclusion is not likely to be particu-larly accurate, but does emphasize the impor-tant point that relatively thin cast sections donot necessarily require fast filling rates to avoidpremature solidification. What is important isthe steady, continuous advance of the meniscus.Naturally, however, it is important not to pressthis conclusion too far, and the above first-order approximation to the fill time probablyrepresents a time that might be achievable inideal circumstances: in fact, if the rate of fillingis too slow, then the rate of advance of theliquid front will become unstable for otherreasons:

1. Surface film problems may cause instabilityin the flow of some materials, as is explainedin Castings (2003). Film-free systems will notsuffer this problem, and vacuum casting mayalso assist.

2. Another instability that has been little re-searched is the flow of the metal in a pastymode. Flow channels revealed in the radio-graph in Figure 2.40 (Runyoro and Campbell1990) show the curious behaviour in whichchannels take a line of least resistancethrough the casting, abandoning the riser.

They adopt the form of magma vents in theearth's crust, and form volcano-like struc-tures at the top surface. (Additionally in theseradiographs a metal±mould reaction betweenA356 alloy and the furan resin binder hasproduced many minute bubbles that havefloated to decorate the upper surfaces of flowchannels, revealing the outline of the lastregions to remain liquid.)

2.3.2.9 Central versus external systems

Most castings have to be run via an externalrunning system as shown in Figure 2.22. Whilethis is satisfactory for the requirements of therunning system, it is costly from the point ofview of the space it occupies in the mould. Thisis especially noticeable in chemically bondedmoulds, whose relatively high cost is, of course,directly related to their volume, and whosevolume can be modified easily since the mouldsare not contained in moulding boxes, i.e. theyare boxless. Naturally, in this situation it wouldbe far more desirable if the running systemcould somehow be incorporated inside thecasting, so as to use no more sand than neces-sary. This ideal might be achieved in somecastings by the use of direct gating in conjunc-tion with a filter as discussed later (Section 2.3.6,Direct pour).

Figure 2.40 Radiograph of an Al±7Si±0.4Mg alloyvertical plate filled via a riser and slot gate from theleft-hand side, revealing unexpected filling behaviourwhen cast particularly cool. Remains of thermocouplescan be seen (Runyoro 1992).


100 mm


For castings that have an open base, how-ever, such as open frames, cylinders or rings, anexcellent compact and effective solution is pos-sible. It is illustrated for the case of cylinder andring castings in Figures 2.23 and 2.24. Therunners radiate outwards from the sprue exit,and connect with vertical slot gates arranged asarcs around the base of the casting. Ruddle andCibula (1957) describe a similar arrangement,but do not show how it can be moulded (with alldue respect to our elder statesmen of the foun-dry world, their suggested arrangement looksunmouldable!), and omit the upward gates. Thevertical gates are an important feature for suc-cess, introducing useful friction into the system,and making for easy cut-off.

Feeders can be sited on the top of the cylinderif required. Alternatively, if the casting is to berolled through 180 degrees after pouring, thefeeding of the casting can take the form of a ringfeeder at the base (later to become the top, ofcourse).

Experience with internal running has found itto be an effective and economical way to pro-duce hollow shapes. It is also effective for theproduction of other common shapes such asgearboxes and clutch covers, where the spruecan be arranged to pass down through a rathersmall opening in one half of the casting and thenbe distributed via a spider of runners and gateson the open side.

However, it has been noted that aluminiumalloy castings of 300 mm or more internal dia-meter exhibit a patternmaker's contractionconsiderably less than that which would havebeen expected for an external system. This seemsalmost certainly to be the result of the expansionof the internal core as a result of the extraheating from the internal running system. For asilica sand core this expansion can be between1 and 1.5 per cent, effectively negating the pat-ternmaker's shrinkage allowance, which isnormally between 1 and 1.3 per cent.

2.3.2.10 Sequential filling

When there are multiple impressions on a hor-izontal pattern plate, it is usually unwise toattempt to fill all the cavities at the same time.(This is contrary to the situation with a verti-cally parted mould, in which many filling sys-tems specifically target the filling of all thecavities at once to reduce pouring time. How-ever, such vertically parted moulds have notbeen subjected to the same degree of study interms of the defects probably introduced by thissystem. In the absence of data therefore, theyare not described further here. We look forward

to good data becoming available at some futuredate.)

The reasoning in the case of the horizontalmould is simple. The individual cavities arefilling at a comparatively slow rate, and notnecessarily in a smooth and progressive way. Infact, despite an otherwise good running systemdesign, it is likely that filling will be severelyirregular, with slopping and surging, becauseof the lack of constraint on the liquid, andbecause of the additional tendency for the flowto be unstable at low flow rates in film-formingalloys. The result will be the non-filling of anumber of the impressions and doubtful qualityof the others.

Loper (1981) provided a solution to thisproblem for multiple impressions on one plateas shown in Figure 2.41. He uses runner dams toretard the metal, allowing it time to build up ahead of metal sufficient to fill the first set ofimpressions before overcoming the dam andproceeding to the next set of impressions, andso on.

The system has only been reported to havebeen used for grey iron castings in greensandmoulds. It may give less satisfactory results forother metal±mould systems that are more sus-ceptible to surface turbulence. However, thedesign of the overflow (the runner down the farside of each dam) could be designed as a min-iature tapered down-runner to control the fall,and so reduce surface turbulence as far aspossible. Probably, this has yet to be tested.

Another sequential-filling technique, `hor-izontal' stack moulding, has also only so farbeen used with cast iron. This was invented inthe 1970s by one of our great foundry char-acters from the UK, Fred Hoult, after hisretirement at the age of 60. It is known in hishonour as the `H Process'. Figure 2.41 outlineshis method. The progress of the metal acrossthe top of those castings already filled keepsthe feeders hot, and thus efficient. The lengthof the stack seems unlimited because the coldmetal is repeatedly being taken from the frontof the stream and diverted into castings. (Thereader will note an interesting analogy with theup-runner and slot gate principle; one is hor-izontal and the other vertical, but both aredesigned to divert their metal into the mouldprogressively. The same effect is also used inthe promotion of fluidity as described byHiratsuka et al. 1966.) Stacks of 20 or moremoulds can easily be poured at one time.Pouring is continued until all the metal is usedup, only the last casting being scrappedbecause of the short pour, and the remainingunfilled moulds are usable as the first mouldsin the next stack to be assembled.



The size of castings produced by the H Pro-cess is limited to parts weighing from a fewgrams to a few kilograms. Larger parts becomeunsuitable partly because of handling problems,since the moulds are usually stacked verticallyduring assembly, then clamped with longthreaded steel rods, and finally lowered to thefloor to make a horizontal line. Larger parts arealso unsuitable because of the fundamentallimitation imposed by the increase of defects asa necessary consequence of the increased dis-tance of fall of the liquid metal inside the mould,and possibly greater opportunity to splash inthicker sections.

2.3.2.11 Two-stage filling

There have been a number of attempts over theyears to introduce a two-stage filling process.The first stage consists of filling the sprue, afterwhich a second stage of filling is started in whichthe runner and gates, etc. are allowed to fill.

The stopping of the filling process after thefilling of the sprue brings the melt in the sprue toa stop, ensuring the exclusion of air. The melt isthen allowed to start flowing once again. Thissecond phase of filling has the full head H of

metal in the sprue and pouring basin to drive it,but the column has to start to move from zerovelocity. It reaches its `equilibrium' velocity(2gH)1/2 only after a period of acceleration.Thus the early phase of filling of the runner andgates starts from a zero rate, and has a graduallyincreasing velocity. The action is similar to our`surge control' techniques described earlier.

The benefits of the exclusion of air from thesprue, and the reduced velocity during the earlypart of stage 2, are benefits that have beenrecorded experimentally for semi-solid (actuallypartly solid) alloys. These materials are other-wise extremely difficult to cast without defects,almost certainly because their entrainmentdefects cannot float out but are trapped in sus-pension because of the high viscosity of themixture.

Workers from Alcan (Cox et al. 1994)developed a system in which the advance of themelt was arrested at the base of the sprue by alayer of ceramic paper supported on a ceramicfoam filter (Figure 2.42a). When the sprue wasfilled the paper was lifted from one corner by arod, allowing the melt to flow through the filterand into the running system. These authors calltheir system `interrupted pouring'. However, the

Figure 2.41 (a) Sequential filling fora number of impressions on a patternplate (after Loper 1981), and(b) sequential filling for horizontalstack moulded castings (H Process).


(a)

(b)


name `two-stage pour' is recommended as beingmore positive, and less likely to be interpreted asa faulty pour as a result of an accident.

The two-stage pour was convincinglydemonstrated as beneficial by Taghiabadi andcolleagues (2003) for both partly solid andconventional aluminium casting alloys. Theseauthors used Weibull statistics to confirm thereality of the benefits. They used a steel sheet toform a barrier, as a slide valve, in the runner(Figure 2.42b). After the filling of the spruethe sheet was withdrawn, allowing the mouldto fill.

A second completely different incarnation ofthe two-stage filling concept is the snorkel ladle,sometimes known as the eye-dropper ladle. It isillustrated in Figure 2.42c. The device is usedmainly in the aluminium casting industry, butwould with benefit extend to other castingindustries. Instead of transferring metal from afurnace via a ladle or spoon of some kind, andpouring into a pouring basin connected to asprue, the snorkel dips into the melt, and canbe filled uphill simply by dipping sufficientlydeeply, or by a shallow dip and the melt suckedup by a reduced pressure applied in the body ofthe snorkel. The ladle is then transferred to themould where it can deliver its contents into aconventional basin and sprue system, or, inthe mode recommended here, lowered downthrough the mould to reach and engage with therunner. Only then is its stopper raised and themelt delivered to the start of the running systemwith minimal surface turbulence. The approachis capable of producing excellent products.

Two-stage filling in its various forms seems tooffer real promise for many castings.

2.3.2.12 Vortex systems

The vortex has usually been regarded in foun-dries as a flow feature to be avoided at all costs.If the vortex truly swallowed air, and the airfound its way into the casting, the vortex wouldcertainly have to be avoided. However, in gen-eral, this seems to be not true.

The great value of the vortex is that it is apowerful organizer of the flow. Designers ofwater intakes for hydroelectric power stationsare well aware of this benefit. Instead of thewater being allowed to tumble haphazardlydown the water intake from the reservoir, it iscaused to spiral down the walls. At the base ofthe intake duct the loss of rotational energyallows the duct to back-fill to some extent. Thecentral core of air terminates at the level surfaceof a comparatively tranquil pool, only gentlycirculating, near the base of the duct. (Thespiralling central core of air does not extendindefinitely through the system.)

Several proposals to harness the benefits ofvortices to running systems have originated inrecent years from Birmingham, following thelead by Isawa (1994). They are potentiallyexciting departures from conventional approa-ches. Only initial results can be presented here.The systems merit much further investigation.

Vortex sprue

The benefits of the vortex for the action of asprue were first explored by Campbell and

Figure 2.42 Two-stage filling techniques: (a) paper seal ontop of ceramic foam filter, lifted by wire; (b) steel slidegate at entrance to runner; (c) stopper in the base of asnorkel ladle.


(a)

(b)

(c)


Isawa (1994) as illustrated in Figure 2.43. Analuminium alloy was poured off-axis, beingdiverted tangentially into a circular pouringbasin. The melt spun around the outside of thebasin, gradually filling up and so progressingtowards the central sprue entrance. As it pro-gressed inwards, gradually reducing its radius,its rotation speeded up, conserving its angularmomentum like the spinning ice skater whocloses in outstretched arms. Finally, the meltreached the lip of the sprue, where, at maximumspinning speed, it started its downward fall.

The rationale behind this thinking is that theinitial fall during the filling of the sprue is con-trolled by the friction of a spiral descent downthe wall of the sprue. Once the sprue starts tofill, the core of the vortex terminates near thebase of the sprue; it does not in fact funnel airinto the mould cavity. The hydrostatic pressuregenerated by this system, driving the flow intothe runner and gates, arises only from the smalldepth of liquid at the base of the vortex, so thatthe net pressure head h driving the filling of themould is small. Because the level of the pool atthe base of the vortex and the level of metal inthe mould cavity both rise together, the net

head to drive the filling of the mould remainsremarkably constant during the entire mouldfilling process. Thus the filling of the mould isnecessarily gentle at all times.

Despite some early success with this system,it seems that the technique is probably notsuitable for sprues of height greater than per-haps 200 or 300 mm, because the benefits of thespiral flow are lost progressively with increasingfall distance. More research may be needed toconfirm the benefits and limitations of thisdesign. For instance, the early work has beenconducted on parallel cylindrical sprues, sincethe taper has been thought to be not necessaryas a result of the melt adopting its own `taper' asit accelerates down the walls, becoming a thin-ner stream as it progresses. However, a tapermay in any case be useful to favour the speedingup of the rate of spin, and so assist maintainingthe spin despite losses from friction against thewalls. Also of course, the provision of a taperwill assist the sprue to fill faster, and increaseyield. Much work remains to be done to definean optimum system.

Vortex well or gate

The provision of a cylindrical channel at thebase of the sprue, entered tangentially by themelt, is a novel idea with considerable potential(Hsu et al. 2003). It gives a technique for dealingwith the central issue of the high liquid velocityat the base of the sprue, and the problem ofturning the right-angle corner and successfullyfilling the runner. What is even better, it pro-mises to solve all of these problems withoutsignificant surface turbulence.

The vortex well can probably be orientedeither horizontally or vertically as seen in Figure2.44. The horizontal orientation may be usefulfor delivery to a single vertical gate. Alter-natively, the vertical orientation is often con-venient because the central outlet from thevortex can form the entrance to the runner,allowing the connection to many gates.

Notice that the device works exactly oppositeto the supposed action of a spinner designed tocentrifuge buoyant inclusions from a melt. Inthe vortex well the outlet to the rest of the fillingsystem is the outlet that would normally be usedto concentrate inclusions. Thus the device cer-tainly does not operate to reduce the inclusioncontent. However, it should be highly effectivein reducing the generation of inclusions bysurface turbulence at the sprue base of poorlydesigned systems.

Once again, these are early days for thisinvention. Early trials on a steel casting of about4 m height have suggested that the vortex is

Figure 2.43 Vortex sprue (after Isawa and Campbell1994).


Flow spins downwall of sprueVortex core‘bottoms out’

Net head todrive filling h

opposed rotation exitto halt rotation anddelay entry to runner


extremely effective in absorbing the energy ofthe flow. In this respect its action resembles thatof a ceramic foam filter. To enable the device tobe used in routine casting production the energyabsorbing behaviour would require to bequantified. There is no shortage of future tasks.

Vortex runner (the offset sprue)

The simple provision of an offset sprue causesthe runner to fill tangentially, the melt spinningat high speed (Figures 2.19e and 2.45). Thetechnique is especially suited to a verticallyparted mould, where a rectangular cross-sectionsprue, moulded on only one side of the partingline, opens into a cylindrical runner moulded onthe joint. The consequential highly organizedfilling of the runner is a definite improvementover many poor runner designs, as has beendemonstrated by the Weibull statistics ofstrengths of castings produced by conventionalin-line and offset (vortex) runners (Yang et al.1998, 2000 and 2003). The technique producesconvincingly more reliable castings than con-ventional in-line sprues and runner systems.

However, at this time it is not certain that theaction of the vortex runner is better than otheruseful solutions such as the slot sprue/slot run-ner, or the vortex well, but it has the greatbenefit of simplicity. It promises to be valuablefor vertically parted moulds; it deals effectivelywith the problem of high flow velocities in suchmoulds because of the great fall heights oftenencountered.

2.3.3 Horizontal transfer casting

The quest to avoid the gravity pouring of liquidmetals has led to systems employing horizontaltransfer and counter-gravity transfer. Thesesolutions to avoid pouring are clearly seen to bekey developments; both seem capable of givingcompetitive casting processes that offer productsof unexcelled quality. The two major approachesto the first approach, horizontal transfer, aredescribed below.

2.3.3.1 Level pour (side pour)

The `level pour' technique was invented by ErikLaid (1978). At that time this clever techniquedelivered castings of unexcelled quality. It seemsa pity that the process is not more generallyused. This has partly occurred as the result ofthe process remaining commercially confidentialfor much of its history, so that relatively littlehas been published concerning the operationaldetails that might assist a new user to achievesuccess. Also, the technique is limited to the typeof castings, being applied easily only to plate,box or cylinder type castings where a long slotingate can be provided up the complete height ofthe casting. In addition, of course, a fairlycomplex casting station is required.

The arrangement to achieve the so-calledlevel filling of the mould is shown in Figure 2.46.An insulated pouring basin connects to a hori-zontal insulated trough that surrounds three ofthe four sides of the mould (a distribution sys-tem reminiscent of a Roman aqueduct). Themelt enters the mould cavity via slot gates thatextend vertically from the drag to the cope.Either side of each slot gate are guide plates thatcontain the melt between sliding seals as it flowsout of the (stationary) trough and into the(descending) mould.

Casting starts with the mould sitting on thefully raised mould platform, so that the troughprovides its first metal at the lowest level of thedrag. The mould platform is then slowly low-ered while pouring continues. The rate of with-drawal of the mould is such that the metal inthe slot gate has time to solidify prior to its

Figure 2.44 (a) Vortex well (with horizontal axis)and (b) vortex gate (with vertical axis).


(a)

(b)


appearance as it slides out below the level of thetrough.

In one of the rare descriptions of the use of theprocess by Bossing (1982) the large area of meltcontained in the pouring basin and distributiontrough would materially help to smooth therate of flow from the point of pour to deliveryinto the mould. Also in this description isan additional complicated distribution systeminside the mould, in which tiers of runners areprovided to minimize feeding distances andmaximize temperature gradients. In general, suchsophistication would not be expected to benecessary for most products.

2.3.3.2 Controlled tilt casting

It seems that foundrymen have been fascinatedby the intuition that tilt casting might be asolution to the obvious problems of gravitypouring. The result has been that the patentliterature is littered with re-inventions of theprocess decade after decade.

Even so, the deceptive simplicity of the pro-cess conceals some fundamental pitfalls for theunwary. The piles of scrap seen from time totime in tilt-pour foundries are silent testimonyto these hidden dangers. Generally, however,the dangers can be avoided, as will be discussedin this section.

Tilt casting is a process with the unique fea-ture that, in principle, liquid metal can be

transferred into a mould by simple mechanicalmeans under the action of gravity, but withoutsurface turbulence. It therefore has the potentialto produce very high quality castings. This wasunderstood by Durville in France in the 1800sand applied by him for the casting of aluminium-bronze in an effort to reduce surface defects inFrench coinage.

The various stages of liquid metal transferin the Durville Process are schematicallyillustrated in Figure 2.47a. In the process as

Figure 2.45 (a) A conventional runner, and (b) a vortex runner, useful on a vertically parted mould (courtesy of X. Yangand Flow 3-D).

Figure 2.46 Level pour technique.


(a) (b)

Insulatedguideplates

Insulateddistributiontrough

Frozen ingate

Mould platform (incorporating mouldclamping + lifting features)lowered by electric motor drive

Mould

Liquidmetalpoured from insulatedladle

Core


originally conceived by Durville, the metal ismelted in the same crucible as is used for the tiltmachine. No pouring under gravity takes place atall. Also, since he was casting large, open-ended

ingots for subsequent working, he was able tolook into the crucible and into the mould,observing the transfer of the melt as the rotationof the mould progressed. In this way he couldensure that the rate of rotation was correct toavoid any disturbance of the surface of theliquid. During the whole process of the transfer,careful control ensured that the melt progressedby `rolling' in its skin of oxide, like inside arubber sack, avoiding any folding of its skin bydisturbances such as waves. The most sensitivepart of the transfer was at the tilt angle close tothe horizontal. In this condition the melt frontprogresses by expanding its skin of oxide, whileits top surface at all times remains horizontaland tranquil.

In the USA, Stahl (1961) popularized theconcept of `tilt pouring' for aluminium alloysinto shaped permanent mould castings. Thegating designs and the advantages of tilt pouringover gravity top pouring have been reviewedand summarized in several papers from thissource (Stahl 1963, 1986, 1989).

A useful `bottom-gated' tilt arrangement isshown in 2.47c, d. Here the sprue is in the drag,and the remainder of the running and gatingsystem, and the mould cavity, is in the cope.Care needs to be taken with a tilt die to ensurethat the remaining pockets of air in the die canvent freely to atmosphere. Also, the die side thatretains the casting has to contain the ejectors ifthey are needed. The layout in Figure 2.47c illus-trates a unique benefit enjoyed by tilt casting:a single operator can fill both pouring cups froma large ladle prior to starting the tilt. Staticgravity casting would require two pourers to filltwo pouring basins.

In an effort to understand the process insome depth, Nguyen and Carrig (1986) simu-lated tilt casting using a water model of liquidmetal flow, and Kim and Hong (1995) carriedout some of the first computer simulations ofthe tilt casting process. They found that acombination of gravity, centrifugal and Coriolisforces govern tilt-driven flow. However, for theslow rates of rotation such as are used in mosttilt casting operations, centrifugal and Corioliseffects contribute less than 10 per cent of theeffects due to gravitational forces, and couldtherefore normally be neglected. The angularvelocity of the rotating mould also made somecontribution to the linear velocity of the liquidfront, but this again was usually negligiblebecause the axis of rotation was often not farfrom the centre of the mould.

However, despite these studies, and despiteits evident potential, the process has continuedto be perfectly capable of producing copiousvolumes of scrap castings.

Figure 2.47 Tilt casting process (a) Durville;(b) Semi-Durville; (c) twin-poured tilting die (adaptedfrom Nyamekye et al. 1994); and (d) outline of tiltrunning system design at the critical moment that metalreaches the far end of the `sprue'.


Mould Melting crucible

(a)

(b)

Metal leveltoo low

Metal onbrink of pour

Metal startspour at steepangle of tilt

Persistentoxideflowtube

Initialsurfaceturbulence

(c)

h1

h2

(d)

Casting

Ingate

Crossrunner

Down-runneror sprue


The first detailed study of tilt casting usingthe recently introduced concepts of criticalvelocity and surface turbulence was carriedout in the author's laboratory by Mi (2002).In addition to the benefits of working withinthe new conceptual framework, he had avail-able powerful experimental techniques. Heused a computer controlled, programmablecasting wheel onto which sand moulds could befixed to produce castings in an Al±4.5%Cualloy. The flow of the metal during the fillingof the mould was recorded using video X-rayradiography, and the consequential reliabilityof the castings was checked by Weibullstatistics.

Armed with these techniques, Mi found thatat the slow rotation speeds used in his work themechanical effect of surface tension and/orsurface films on the liquid meniscus could not beneglected. For all starting conditions, the flowat low tilt speeds is significantly affected bysurface tension (most probably aided by theeffect of a strong oxide film). Thus below aspeed of rotation of approximately 7 degrees persecond the speed of the melt arriving at the endof the runner is held back. Gravity only takescontrol after tilting through a sufficiently largeangle.

As with most casting processes, if carried outtoo slowly, premature freezing will lead to mis-run castings. One interesting case was foundin which the melt was transferred so slowly intothe runner that frozen metal in the mouthof the runner acted as an obstructing ski jumpto the remaining flow, significantly impairingthe casting. At higher speeds, however, althoughski jumps could be avoided, the considerabledanger of surface turbulence increased.

The radiographic recordings revealed thatthe molten metal could exhibit tranquil orchaotic flow into the mould during tilt casting,depending on (i) the angle of tilt of the mould atthe start of casting, and (ii) the tilting speed. Thequality of the castings (assessed by the scatter inmechanical properties) could be linked directlyto the quality of the flow into the mould.

We can follow the progress of the melt duringthe tilt casting process. Initially, the pouringbasin at the mouth of the runner is filled. Onlythen is the tilting of the mould activated. Threestarting positions were investigated:

(i) If the mould starts from some position inwhich it is already tilted downward, oncethe metal enters the sprue it is immediatelyunstable, and runs downhill. The meltaccelerates under gravity, hitting the farend of the runner at a speed sufficient tocause splashing. The splash action entrains

the melt surface. Castings of poor reliabilityare the result.

(ii) If the mould starts from a horizontal posi-tion, the metal in the basin is not usuallyfilled to the brim, and therefore does notstart to overflow the brim of the basin andenter the runner until a finite tilt angle hasbeen reached. At this stage the vertical falldistance between the start and the far end ofthe runner is likely to be greater than thecritical fall distance. Thus although slightlybetter castings can be made, the dangerof poor reliability remains. This unsatisfac-tory mode of transfer typifies many tiltcasting arrangements, particularly the so-called Semi-Durville type process shown inFigure 2.47b.

(iii) If, however, the mould is initially tiltedslightly uphill during the filling of the basin,there is a chance that by the time the changeof angle becomes sufficient to start theoverflow of melt from the basin, the angleof the runner is still somewhat above thehorizontal. The nature of the liquid metaltransfer is now quite different. At the startof the filling of the runner the meniscus iseffectively climbing a slight upward slope.Thus its progress is totally stable, itsforward motion being controlled by addi-tional tilt. If the mould is not tilted furtherthe melt will not advance. By extremelycareful control of the rate of tilt it is possiblein principle to cause the melt to arrive at thebase of the runner at zero velocity ifrequired. (Such drastic reductions in speedwould, of course, more than likely becounter-productive, involving too great aloss of heat, and are therefore not recom-mended.) Even at quite high tilting speedsof 30 degrees per second as used by Mi in hisexperimental mould, the velocity of the meltat the end of the runner did not exceed thecritical value 0.5 m sÿ1, and thus producedsound and repeatable castings.

The unique feature of the transfer when startedabove the horizontal in this way (mode iiiabove) is that the surface of the liquid metal isclose to horizontal at all times during thetransfer process. Thus in contrast to all othertypes of gravity pouring, this condition of tiltcasting does not involve pouring (i.e. a freevertical fall) at all. It is a horizontal transferprocess. It will be seen that in the critical regionof tilt near to the horizontal, the nature of thetransfer is the same as that employed originallyby Durville.

Thus the optimum operational mode for tiltcasting is the condition of horizontal transfer.



Horizontal transfer requires the correct choiceof starting angle above the horizontal, and thecorrect tilting speed.

An operational map was constructed(Figure 2.48), revealing for the first time awindow for the production of reliable castings.It is recognized that the conditions defined bythe window are to some extent dependent on thegeometry of the mould that is chosen. However,the mould in Mi's experiments was designed tobe close to the size and shape of many industrialcastings, particularly those for automotiveapplications. Thus although the numerical con-clusions would require some adaptation forother geometries, the principles are of generalsignificance and are clear: there are conditions,possibly narrowly restricted, but in which hor-izontal transfer of the melt is possible, and givesexcellent castings.

The problem of horizontal transfer is that it isslow, sometimes resulting in the freezing of the`ski jump' at the entrance to the runner, or eventhe non-filling of the mould. This can usually besolved by increasing the rate of tilt after therunner is primed. This is the reason for theextended threshold, denoted a marginal fillingcondition, on the left of the window shown onthe process map (Figure 2.48). A constant tiltrate (as is common for most tilt machines at thistime) cannot achieve this useful extension of thefilling conditions to achieve good castings.Programmable tilt rates are required to achievethis solution.

A final danger should be mentioned. At cer-tain critical rates of rise of the melt against an

inclined surface, the development of the trans-verse travelling waves seems to occur to give lapproblems on the cope surface of castings (illu-strated later in Figure 2.60). In principle, suchproblems could be included as an additionalthreshold to be avoided on the operationalwindow map (Figure 2.48). Fortunately, thisdoes not seem to be a common fault. Thus in themeantime, the laps can probably be avoided byincreasing the rate of tilt during this part of thefilling of the mould. Once again, the benefits ofa programmable tilt rate are clear.

Insummary, theconclusionsfortiltcastingare:

1. If tilt casting is initiated from a tilt orienta-tion at, or (even more especially) below thehorizontal, during the priming of the runnerthe liquid metal runs downhill at a rate out ofthe control of the operator. The acceleratingstream runs as a narrow jet, forming apersistent oxide flow tube. In addition, thevelocity of the liquid at the far end of therunner is almost certain to exceed the criticalcondition for surface turbulence. Once themould is initially inclined by more than10 degrees below the horizontal at theinitiation of flow, Mi found that it was nolonger possible to produce reliable castingsby the tilt casting process.

2. Tilt casting operations benefit from using asufficiently positive starting angle that themelt advances into an upward sloping run-ner. In this way its advance is stable andcontrolled. This mode of filling is character-ized by horizontal liquid metal transfer,promoting a mould filling condition freefrom surface turbulence.

3. Tilt filling is preferably slow at the earlystages of filling to avoid the high velocities atthe far end of the running system. However,after the running system is primed, speedingup the rate of rotation of the mould greatlyhelps to prevent any consequential non-filling of the castings.

2.3.4 Counter-gravity

There are some advantages to the use of gravityto action the filling of moulds. It is simple, lowcost and completely reliable, since gravity hasnever been known to suffer a power failure. It iswith regret, however, that the advantages finishhere, and the disadvantages start. Furthermore,the disadvantages are serious.

Nearly all the problems of gravity pouringarise as a result of the velocity of the fall. After atrivial fall distance corresponding to the criticalfall height, gravity has accelerated the melt to itscritical velocity. Beyond this point there is the

Figure 2.48 Map of variables for tilt pouring, showingthe operational window for good castings (Mi et al. 2002).


–200 0.5

Maximum velocity ofmetal front in runner (m s–1)

10

–10

0

10

20

30

Reliable, filledcastings

Incompletefilling

Initi

al a

ngle

of m

ould

(de

gree

s ab

ove

horiz

onta

l)

UnreliablecastingsMarginal

filling


danger of entrainment defects. Because thecritical fall distance is so small, being only theheight of a sessile drop of the liquid, nearly allactual falls exceed this limit. In other words theenergy content of the melt, when allowed to falleven only relatively small distances under grav-ity, is nearly always sufficiently high to lead to thebreak-up of the liquid surface. (It is of littlecomfort at this time to know that foundries onthe moon would fare better.)

A second fundamental drawback of gravityfilling is the fact that at the start of pouring, atthe time the melt is first entering the ingates, thenarrowest part of the mould cross-section wherevolume flow rate should be slowest, the speed offlow by gravity is highest. Conversely, at a latestage of filling, when the melt is at its coldestand approaching the top of the mould cavity,and the melt needs to be fastest, the speed offilling is slowest. Thus filling by gravity givescompletely the wrong filling profile.

Thus to some extent, there are always prob-lems to be expected with castings poured bygravity. The long section on filling systemdesign in this book is all about reducing thisdamage as far as possible. It is a tribute to thedogged determination of the casting fraternitythat gravity pouring, despite its severe short-comings, has achieved the level of success that itcurrently enjoys.

Even so, over the last 100 years and more, thefundamental problems of gravity filling haveprompted casting engineers to dream up anddevelop counter-gravity systems.

Numerous systems have arisen. The mostcommon is low-pressure casting, in which air oran inert gas is used to pressurize an enclosedfurnace, forcing the melt up a riser tube and intothe casting (Figure 2.49). Other systems use apartial vacuum to draw up the metal. Yet othersuse various forms of pumps, including directdisplacement by a piston, by gas pressure(pneumatic pumps), and by various types ofelectromagnetic action.

Clearly, with a good counter-gravity system,one can envisage the filling of the mould atvelocities that never exceed the critical velocity,so that the air in the mould is pushed ahead ofthe metal, and no surface entrainment occurs.The filling can start gently through the ingates,speed up during the filling of the main part ofthe mould cavity, and finally slow down andstop as the mould is filled. The final decelerationis useful to avoid any final impact at the instantthe mould is filled. If not controlled in this way,the transient pressure pulse resulting from thesudden loss of momentum of the melt can causethe liquid to penetrate any sand cores, openmould joints to produce flash, and generallyimpair surface finish.

When using a good counter-gravity system,good filling conditions are not difficult toachieve. In fact, in comparison with gravitypouring where it is sometimes difficult to achievea good casting, counter-gravity is such a robusttechnique that it is often difficult to make a badcasting. This fundamental difference betweengravity and counter-gravity filling is not widely

Figure 2.49 (a) Low-pressure casting process, and (b) the usual poor filling technique.


(a) (b)Air cylinderto open die

Gulde bars

Ejector

Gas pressure

Riser tube

Melt withaccumulatedoxides

Dieset

Top dieBottom die


appreciated. In general, only those who havesuffered gravity filling and finally acceptedcounter-gravity appreciate and are amazed bythe powerful benefits.

That is not to say that the technique is notsometimes used badly. The usual failure is tokeep the metal velocity under control. However,in principle it can be controlled, in contrast togravity pouring where, in principle, control isoften difficult or impossible.

A concern often expressed about counter-gravity is that the adoption of filling speedsbelow the critical speed of approximately0.5 m sÿ1 will slow the production rate. Suchfears are groundless. For instance if the castingis 0.5 m tall (a tall casting) it can, in principle, befilled in one second. This would be a challenge!

In fact the unfounded fear of the use of lowvelocities of the melt leading to a sacrifice ofproduction rate follows from the confusion of(i) flow velocity (usually measured in m sÿ1) and(ii) melt volume flow rate (usually measuredin m3 sÿ1). For instance the filling time can bekept short by retaining a slow filling velocitybut increasing the volume flow rate simply byincreasing the areas of the flow channels.Worked examples to emphasize and clarify thispoint further will be given in Section 2.3.7dealing with the calculation of the filling system.

2.3.4.1 Programmable control

The varying cross-sectional areas of the metal asit rises in the mould pose a problem if the fillrate through the bottom gate is fixed (as isapproximately true for many counter-gravityfilling systems that lack any sophistication ofprogrammable control). Naturally, the meltmay become too slow if the area of the mouldincreases greatly, leading to a danger of coldlaps or oxide laps. Alternatively, if the localvelocity is increased above the critical velocitythrough a narrow part of the mould, the metalmay jet, causing entrainment defects.

Counter-gravity filling is unique in havingthe potential to address this difficulty. In prin-ciple, the melt can be speeded up or sloweddown as required at each stage of filling. Evenso, such programming of the fill rate is not easilyachieved. In most moulds there is no way todetermine where the melt level is at any timeduring filling. Thus if the pre-programmed fill-ing sequence (called here the filling profile) getsout of step, its phases occurring either early orlate, the filling can become worse than thatoffered by a constant rate system. The mis-timing problems can easily arise from splashesthat happen to start timers early, or from

blockage in the melt delivery system causing thetime of arrival of the melt to be late.

2.3.4.2 Feedback control

The only sure way to avoid such difficulties is toprovide feedback control. This involves a sys-tem to monitor the height of metal in the mould,and to feed this signal back to the delivery sys-tem, to force the system to adhere to a pre-programmed fill pattern.

One system for the monitoring of height isthe sensing of the pressure of the melt in the meltdelivery system. This has been attempted by theprovision of a pocket of inert gas above the meltcontained in the permanent plumbing of theliquid metal delivery system, and connected to apressure transducer via a capillary.

A non-contact system used by the Cosworthsand casting process senses the change in capa-citance between the melt and the mould clampplate when the two are connected as a parallelplate condenser. Good feedback control solvesmany of the filling problems associated withcasting production.

However, elsewhere, feedback control is littleused at this time. The lack of proper control incounter-gravity leads to unsatisfactory modes offilling that explain some of the problems withthe technique. The other problems relate to theremainder of the melting and melt handlingsystems in the foundry, that are often poor,involving multiple pouring operations from meltfurnaces to ladles and then into the counter-gravity holding vessel. A widespread re-chargingtechnique for a low-pressure casting unit isillustrated in Figure 2.49b; much of theentrainment damage suffered in such processesusually cannot be blamed on the counter-gravitysystem itself. The problem arises earlier becauseof inappropriate metal handling in the foundrybefore any casting takes place.

The lesson is that only limited success can beexpected from a foundry that has added a coun-ter-gravity system on to the end of a badlydesigned melting and melt-handling system.There is no substitute for an integrated approachto the whole production system. Some of the veryfew systems to achieve this so far have been theprocesses that the author has helped to develop;the Cosworth Process (see description later underRule 7) and Alotech Processes. In these processes,when properly implemented, the liquid metal isnever poured, never flows downhill, and is finallytransferred uphill into the mould.

Finally, the concept of an integrated approachnecessarily involves dealing with convectionduring the solidification of the casting. This ser-ious problem is usually completely overlooked.



It has been the death of many otherwise goodcounter-gravity systems, but is specificallyaddressed in the Cosworth and Alotech systems.The problem is highlighted by the author asRule 7.

The numerous forms of counter-gravitytechniques will be discussed in detail in Volume 3`Casting Processes'.

2.3.5 Surface tension controlled filling

This section starts with the interesting situationthat the liquid may not be able to enter themould at all. This is to be expected if the pres-sure is too low to force melt into a narrowsection. It is an effect due to surface tension. Ifthe liquid surface is forced to take up a sharpcurvature to enter a non-wetted mould then itwill be subject to a repulsive force that will resistthe entry of the metal. Even if the metal enters, itwill still be subject to the continuing resistanceof surface tension, which will tend to reverse theflow of metal, causing it to empty out of themould if there is any reduction in the fillingpressure. These are important effects in narrow-section moulds (i.e. thin-section castings) andhave to be taken into account.

We may usefully quantify our formulation ofthis problem with the well-known equation

Pi ÿ Pe � g�1=r1 � 1=r2� (2.3)

where Pi, is the pressure inside the metal, and Pethe external pressure (i.e. referring to the localenvironment in the mould). The two radii r1 andr2 define the curvature of the meniscus in twoplanes at right angles. The equation applies tothe condition when the pressure differenceacross the interface is exactly in balance with theeffective pressure due to surface tension. Todescribe the situation for a circular-section tubeof radius r (where both radii are now identical),the relation becomes:

Pi ÿ Pe � 2g=r (2.4)

For the case of filling a narrow plate of thick-ness 2r, one radius is, of course, r, but the radiusat right angles becomes infinite, so the recipro-cal of the infinite radius equates to zero (i.e. ifthere is no curvature there is no pressure dif-ference). The relation then reduces to the effectof only the one component of the curvature, r:

Pi ÿ Pe � g=r (2.5)

We have so far assumed that the liquid metaldoes not wet the mould, leading to the effect ofcapillary repulsion. If the mould is wetted thenthe curvature term g/r becomes negative, soallowing surface tension to assist the metal to

enter the mould. This is, of course, the familiarphenomenon of capillary attraction. The poresin blotting paper attract the ink into them; thecapillary channels in the wick of a candle suckup the molten wax; and the water is drawn upthe walls of a glass capillary. In general, how-ever, the casting technologist attempts to avoidthe wetting of the mould by the liquid metal.Despite all efforts to prevent it, wetting some-times occurs, leading to the penetration of themelt into sand cores and moulds.

Continuing now in our assumption thatthe metal±mould combination is non-wetting,we shall estimate what head of metal will benecessary to force it into a mould to make a wallsection of thickness 2r for a gravity castingmade under normal atmospheric pressure. If thehead of liquid is h, the hydrostatic pressure atthis depth is rgh, where r is the density ofthe liquid, and g the acceleration due to gravity.The total pressure inside the metal is thereforethe sum of the head pressure and the atmo-spheric pressure, Pa. The external pressure issimply the pressure in the mould due to theatmosphere Pa plus the pressure contributed bymould gases Pm. The equation now is

�Pa� rgh� ÿ �Pa� Pm� > g=r (2.6)

giving immediately

rghÿ Pm > glr (2.7)

The back-pressure due to outgassing in themould lowers the effective head driving thefilling of the mould. It is good practice, there-fore, to vent narrow sections, reducing thisresistance to practically zero if possible.

It is also clear from the above result that,provided the mould is permeable and/or wellvented, atmospheric pressure plays no part inhelping or resisting the filling of thin sections inair, since it acts equally on both sides of theliquid front, cancelling any effect. Interestingly,the same equation and reasoning applies tocasting in vacuum, which, of course, can beregarded as casting under a reduced atmo-spheric pressure. Clearly, a vacuum casting istherefore not helpful in overcoming the resist-ance to filling provided by surface tension(although, to be fair, it may help by reducing Pmby outgassing the mould to some extent prior tocasting, and it will help where the permeabilityof the mould is low, where residual gases may becompressed ahead of the advancing stream.Vacuum casting may also help to fill the mouldby reducingÐbut not eliminatingÐthe effect ofthe surface film of oxide or nitride).

The case of vacuum-assisted filling (notvacuum casting) is quite different, since the



vacuum is not now applied to both the front andback of the liquid meniscus, thus cancelling anybenefit as above, but applied only to theadvancing front as illustrated in Figure 2.50.This application of a reduced pressure to oneside of the meniscus creates a differentialpressure that drives the flow. The differential

pressure acts by atmospheric pressure continu-ing to apply to the liquid metal via the runningsystem, but the atmospheric pressure in themould is reduced by applying a (partial) vacuumin the mould cavity. This is achieved by drawingthe air out either through the permeable mould,or through fine channels cut through to thesection required to be filled (as is commonlyapplied to the trailing edge of an aerofoil bladesection). In this way Pm is guaranteed to be zeroor negligible, and Pa remains a powerful pres-sure to assist in overcoming surface tension asthe equation indicates:

Pa� rgh > g=r (2.8)

It is useful to evaluate the terms of this equationto gain a feel for the size of the effects involved.Taking, roughly, g as 10 m sÿ2, and the liquidaluminium density r as 2500 kg mÿ3 and g as1.0 N mÿ1 (for steels and high-temperaturealloys the corresponding values are approxi-mately 7000 kg mÿ3 and 2.0 N mÿ1), the resist-ance term g/r works out to be 2 kPa for a 1 mmsection (0.5 mm radius) and 10 kPa for a 0.1 mmradius trailing edge on a turbine blade.

For a head of metal h� 100 mm the headpressure rgh is 2.5 kPa, showing that the 1 mmsection might just fill. However, the 0.1 mmtrailing edge has no chance; the head pressurebeing insufficient to overcome the repulsion ofsurface tension. However, if vacuum assistancewere applied (NB not vacuum casting, remem-ber) then the additional 100 kPa of atmosphericpressure normally ensures filling. In practice itshould be noted that the full value of atmo-spheric pressure is not easily obtained invacuum-assisted casting; in most cases a valuenearer half an atmosphere is more usual. Evenso, the effect is still important: one atmospherepressure corresponds to 4 m head of liquid alu-minium, and approximately 1.5 m head of densermetals such as irons, steels and high-temperaturealloys. In modest-sized castings of overall heightaround 100 mm or so, these valuable pressures toassist filling are not easily obtainable by othermeans. The pressure delivered by a feeder placedon top of the casting may only apply the addi-tional head corresponding to its height of per-haps 0.1 to 0.4 m; only one tenth of the pressuresthat can be applied by the atmosphere.

For those castings that have sections of only1 or 2 mm or less, the surface tension wieldsstrong control over the tight radius of the front.Filling is only possible by the operation ofadditional pressure, such as that provided by thejeweller's centrifuge, or the application ofvacuum assistance. Filling can occur upwards ordownwards without problems, being always

Figure 2.50 (a) A plaster mould encased in a steel boxusing vacuum-assisted filling through the base of themould. No formal running system is required for suchsmall thin-walled castings. (b) Sand mould to make fourcover castings, using narrow slot filling system to max-imize benefits from surface tension and wall friction.


Gates 50 × 4 tapered to 2

Runner 25 × 5

(Steel container)

(Permeableplaster mould)

(a)

(b)

Vacuum


under the control of the surface tension, which iseffectively so strong in such thin sections that itkeeps the surface intact. Surface turbulence isthereby suppressed. The liquid has insufficientroom to break up into drops, or to jet or splash.The integrity of the front is under the control ofsurface tension at all times. This special featureof the filling of very thin-walled castings meansthat they do not require formal running systems.In fact, such thin-walled investment castings aremade successfully by simply attaching waxpatterns in any orientation directly to a sprue(Figure 2.50). The metal flows similarly witheither gravity or counter to gravity, and no`runner' or `gate' is necessary.

To gain an idea of the head of metal requiredto force the liquid metal into small sections,from Equation 2.8 we have:

rgh � g=r

h � g=rrg (2.8a)

Using the values for aluminium and steel givenabove, we can now quickly show that to pene-trate a 1 mm section we require heads of at least80 and 60 mm respectively for these two liquids.

If the section is halved, the required head forpenetration is, of course, doubled. Similarly, ifthe mould shape is not a flat section thatimposes only one curvature on the meniscus, butis a circular hole of diameter 1 mm, the surfacethen has an additional curvature at right anglesto the first curvature. Equation 2.4 shows thehead is doubled again.

In general, because of the difficulty of pre-dicting the shape of the liquid surface in com-plex and delicate castings, the author has foundthat a safety factor of 2 is not excessive whencalculating the head height required to fill thinsections. This safety factor is quickly used upwhen allowances for errors in the wall thickness,and the likely presence of surface films is takeninto account.

The resistance to flow provided by surfacetension can be put to good effect in the use ofslot-shaped filling systems. In this case the slotsare required to be a maximum thickness of only1 or 2 (perhaps 3 at the most) mm for engi-neering castings (although, clearly, jewelleryand other widget type products might requireeven thinner filling systems). Figure 2.50b showsa good example of such a system. A similarfilling system for a test casting designed by theauthor, but using a conical basin (not part of theauthor's original design!), was found to performtolerably well, filling without the creation ofsignificant defects (Groteke 2002). It is quiteevident, however, when filling is complete such

narrow filling channels offer no possibility ofsignificant feeding. This is an important issuethat should not be forgotten. In fact, in thesetrials, this casting never received the properattention to feeding, and as a consequence suf-fered surface sinks and internal microporosity(the liquid alloy was clearly full of bifilms thatwere subsequently opened by the action ofsolidification shrinkage).

Finally, however, in some circumstancesthere may be fundamental limitations to theintegrity of the liquid front in very thin sections.

(i) There is a little-researched effect that theauthor has termed microjetting (Castings2003). This phenomenon has been observedduring the filling of liquid Al±7Si±0.4Mgalloy into plaster moulds of sections be-tween 1 and 3 mm thickness (Evans et al.1997). It seems that the oxide on such smallliquid areas temporarily restrains the flow,but repeatedly splits open, allowing jets ofliquid to be propelled ahead of the front. Theresult resembles advancing spaghetti. Themechanical properties are impaired bythe oxide films around the jets that becomeentrained in the maelstrom of progress ofthe front. Whether this unwelcome effectis common in thin-walled castings is un-known, and the conditions for its formationand control are also unknown. Very thinwalled castings remain to be researched.

(ii) In pressure die-castings a high velocity v ofthe metal through the gate is necessary tofill the mould before too much heat is lost tothe die. Speeds of between 25 and 50 m sÿ1

are common, greatly exceeding the criticalvelocity of approximately 0.5 m sÿ1 thatrepresents the watershed between surfacetension control and inertial control of theliquid surface. The result is that entrain-ment of the surface necessarily occurs on ahuge scale. The character of the flow is nowdictated by inertial pressure, proportionalto v2, that vastly exceeds the restraininginfluences of gravity or surface tension.This behaviour is the underlying reasonfor the use of PQ2 diagrams as an attempt tounderstand the filling of pressure die cast-ings. In this approach a diagram is con-structed with vertical axis denoting pressureP, and horizontal axis denoting flow rate Q.The parabolic curves are linearized bysquaring the scale of the Q values on thehorizontal axis. The approach is describedin detail in much of the pressure die-castingliterature (see, for instance, Wall and Cocks1980). In practice, it is not certain how valu-able this technique is, now that computer



simulation is beginning to be accepted as anaccurate tool for the understanding of theprocess.

2.3.6 Inclusion control: filters and traps

The term `inclusion' is a shorthand generallyused for `non-metallic inclusion'. However, it isto be noted that such defects as tungsten drop-lets from a poor welding technique can appearin some recycled metals; these, of course, con-stitute `metallic inclusions'. Furthermore, one ofthe most common defects in many castings is thebubble, entrained during pouring. This con-stitutes an `air inclusion' or `gas inclusion.'

The fact that bubbles are trapped in thecasting from the filling stage is remarkable initself. Why did the bubble not simply rise to thesurface, burst and disappear? This is a simplebut important question. In most cases the bub-ble will not have been retained by the growth ofsolid, because solid will, in general, not havetime to form. The answer in practically all casesis that oxide films will also be present. In fact thebubbles themselves are simply sections of theoxide films that have not perfectly folded backtogether. The bubbles decorate the double films,as inflated islands in the folds. Thus manybubbles, entangled in a jumble of films, neversucceed to reach the surface to escape. Eventhose that are sufficiently buoyant to powertheir way through the tangle may still not burstat the surface because of the layers of oxide thatbar its final escape.

This close association of bubbles and films(since they are both formed by the same turbu-lent entrainment process; they are both entrain-ment defects) is called by me bubble damage. Weneed to keep in mind that the bubble is thevisible part of the total defect. The surroundingregion of bifilms to which it is connected act ascracks, and can be much more extensive andoften invisible. However, the presence of suchfilms is the reason that cracks will often appearto start from porosity, despite the porosityhaving a nicely rounded shape that would not initself appear to be a significant stress raiser.

Whereas inclusions are generally assumed tobe particles having a compact shape, it is essen-tial to keep in mind that the most damaginginclusions are the films (actually always double,unbonded films, remember, so that they act ascracks), and are common in many of our com-mon casting alloys. Curiously, the majority ofworkers in this field have largely overlooked thissimple fact. It is clear that techniques to removeparticles will often not be effective for films, andvice versa. The various methods to clean metalsprior to casting have been reviewed in Chapter 1

as a fulfilment of Rule 1. The various methods toclean metals travelling through the filling sys-tems of castings will be reviewed here.

2.3.6.1 Dross trap (or slag trap)

The dross trap is used in light alloy and copper-based alloy casting. In ferrous castings it iscalled a slag trap. For our purposes we shallconsider the devices as being one and the same.

It is good sense to include a dross trap in therunning system. In principle, a trap sited at theend of the runner will take the first metalthrough the runner and keep it away from thegates. This first metal is both cold, having givenup much of its heat to the running system enroute, and will have suffered damage by oxideor other films during those first moments beforethe sprue is properly filled.

In the past, designs have been along the linesof Figure 2.37a. This type of trap was sized witha view to accommodating the total volume ofmetal through the system until the down-runnerand horizontal runner were substantially filled.This was a praiseworthy aim. In practice, how-ever, it was a regular joke among foundrymenthat the best quality metal was concentrated inthe dross trap and all the dross was in thecasting! What had happened to lend more thanan element of truth to this regrettable piece offolk-law?

It seems that this rather chunky form of trapsets up a circulating eddy during filling. Drossarriving in the trap is therefore efficientlyfloated out again, only to be swept through thegates and into the casting a few moments later!Ashton and Buhr (1974) have carried out workto show that runner extensions act poorly astraps for dirt. They observed that when the firstmetal reached the end of the runner extension itrose, and created a reflected wave which thentravelled back along the top surface of themetal, carrying the slag or dirt back towards theingates. Such observations have been repeatedon iron and steel casting by Davis and Magny(1977) and on many different alloys in theauthor's laboratory using real-time radiographyof moulds during casting. The effect has alsobeen simulated in computer models. It seems,therefore, to be real and universal in castings ofall types. We have to conclude that this design ofdross trap cannot be recommended!

Figure 2.26b shows a simple wedge trap. Itwas thought that metal flowing into the nar-rowing section was trapped, with no reboundwave from the end wall, and no circulating eddycan form. However, video radiographic studieshave shown that such traps can reflect a back-ward wave if the runner is sufficiently deep.



Also, of course, the volume of melt that they canretain is very limited.

A useful design of dross trap appears to be avolume at the end of the runner that is providedwith a narrow entrance (the extension shown inbroken outline in Figure 2.37b) to suppress anyoutflow. It is a kind of wedge trap fitted with amore capacious end. In the case of persistentdross and slag problems, the trap can be exten-ded, running around corners and into sparenooks and crannies of the mould. If the entrancesection is less than the height of a sessile drop, itwill be filled by the entering liquid, thus being toonarrow to allow a reflected wave to exit. It shouldtherefore retain whatever material enters. Inaddition, depending on the narrowness of thetapered wedge entrance, to some extent thedevice should be capable of filling and pressur-izing the runner in a progressive manner akin tothe action of a gate. This is a useful technique toreduce the initial transient momentum problemsthat cause gates to fill too quickly during the firstfew seconds. This potentially useful benefit hasyet to be researched more thoroughly so as toprovide useful guidelines for mould design.

The device can be envisaged to be useful incombination with other forms of by-passdesigns such as those shown in 2.37d and e.

Slag pockets

For iron and steel castings the term `slag pocket'is widely used for a raised portion of the runnerthat is intended to collect slag. The large size ofslag particles and their large density differencewith the melt encourages such separation.However, such techniques are not the panaceathat the casting engineer might wish for.

For instance the use of traps of wedge-shapeddesign, Figure 2.51, is expected to be almostcompletely ineffective because the circulationpattern of flow would take out any material thathappened to enter. On the other hand, a rectan-gular cavity has a secondary flow into whichbuoyant material can transfer if it has sufficienttime, and so remain trapped in the upper

circulating eddy. This consideration againemphasizes the need for relatively slow flow for itseffectiveness. Also, of course, none of these trapscan become effective until the runner and the trapbecome filled with metal. Thus many fillingsystems will have passed much if not all suchunwanted material before the separation mech-anisms have a chance to come into operation. Afurther consideration that causes the author tohesitate to recommend such traps is that theylocally remove the constraint on the flow of themetal, allowing surface turbulence. Thus thetraps might cause more problems than they solve.

Davis and Magny (1977) observed the fillingof iron and steel castings by video radiography.They confirmed that most slag retention deviceseither do not work at all, or work with onlypartial effectiveness. These authors made cast-ings with different amounts of slag, and testedthe ability of slag pockets sited above runnersto retain the slag. They found that rectangularpockets were tolerably effective only if thevelocity of flow through the runner was below0.4 m sÿ1 (interesting that this is precisely thecritical speed at which surface turbulence willoccur, and so cause surface phases to be tur-bulently stirred back into the bulk liquid). For acasting only 0.1 m high the metal is alreadytravelling four times too fast. For such reasonsthe experience with slag pockets has beensomewhat mixed in practice.

In defence of the historical use of such traps itmust be borne in mind that they were tradi-tionally used with pressurized filling systems,heavily choked at the gate, so that the runnerwas encouraged to fill as quickly as possible,making the trap effective at an early stage offilling. Also, if the choking action was suffi-ciently severe the speed of flow in the runnermay be sufficiently slow to ensure slag entrap-ment. However, this text does not recommendpressurized filling systems mainly because of theproblems that follow from the necessarily highingate velocities.

Perhaps, therefore, the slag trap has come tothe end of its useful life.

Figure 2.51 Various designs of slag pockets: (a) relatively ineffective self-emptying wedge; (b) rectangular trapstores buoyant phases in upper circulating flow.



2.3.6.2 Swirl traps

The centripetal trap is an accurate name for thisdevice, but rather a mouthful. It is also knownas a whirl gate, or swirl gate, which is shorter,but inaccurate since the device is not really agate at all. Choosing to combine the best of bothnames, we can call it a swirl trap. This is con-veniently short, and accurately indicates itsmain purpose for trapping rubbish.

The idea behind the device is the use of thedifference in density between the melt and thevarious unwanted materials which it may carry,either floating on its surface or in suspension inits interior. The spinning of the liquid creates acentrifugal action, throwing the heavy melttowards the outside where it escapes through theexit, to continue its journey into the casting.Conversely, the lighter materials are throwntowards the centre, where they coagulate andfloat. The centripetal acceleration ac is given by:

ac � V 2=r (2.9)

where V is the local velocity of the melt, and r isthe radius at that point. For a swirl trap of50 mm radius and sprue heights of 0.1 m and1 m, corresponding to velocities of 1.5 and4.5 m sÿ1 respectively we find that accelerationsof 40 and 400 m sÿ2 respectively are experiencedby the melt. Given that the gravitational accel-eration g is 9.81 m sÿ2, which we shall approx-imate to 10 m sÿ2, these values illustrate thatthe separating forces within a swirl trap can bebetween 4 and 40 times that due to gravity. Theseare, of course, the so-called `g' forces experiencedin centrifuges.

So much for the theory. What about thereality?

Foundrymen have used swirl traps exten-sively. This popularity is not easy to understandbecause, unfortunately, it cannot be the result oftheir effectiveness. In fact the traps have workedso badly that Ruddle (1956) has recommendednot to use them on the grounds that their poorperformance does not justify the additionalcomplexity. One has to conclude that theirextraordinarily wide use is a reflection of thefascination we all have with whirlpools, and anunshakeable belief, despite all evidence to thecontrary, that the device should work.

Regrettably, the swirl trap is expected to becompletely useless for film-forming alloys whereinclusions in the form of films will be too slug-gish to separate. Since some of the worst inclu-sions are films, the swirl trap is usually worsethan useless, creating more films than it canremove. Worse still, in the case of alloys ofaluminium and magnesium, their oxides are

denser than the metal, and so will be centrifugedoutwards, into the casting! Swirl traps aretherefore of no use at all for light alloys (how-ever, notice that the vortex sprue base, althoughnot specifically designed to control inclusions,might have some residual useful effect for lightalloys since the outlet is central). Finally, swirltraps seem to be difficult to design to ensureeffective action. In the experience of the author,most do much more harm than good.

The inevitable conclusion is that swirl trapsshould be avoided.

The remainder of this section on swirl traps isfor those who refuse to give up, or refuse tobelieve. It also serves as a mini-illustration of thereal complexity of apparently simple foundrysolutions. Such illustrations serve to keep ushumble.

It is worthwhile to examine why the tradi-tional swirl trap performs so disappointingly.On examination of the literature, the textbooks,and designs in actual use in foundries, threemain faults stand out immediately:

1. The inlet and exit ducts from the swirl trapsare almost always opposed, as shown inFigure 2.52a. The rotation of the metal as aresult of the tangential entry has, of course,to be brought to a stop and reversed indirection to make its exit from the trap. Thedisorganized flow never develops its intendedrotation and cannot help to separate inclu-sions with any effectiveness.

Where the inlet and exit ducts are arrangedinthecorrecttangentialsense,thenTrojanetal.(1966) have found that efficiency is improvedin their model results using wood chips inwater. Even so, efficiencies in trapping thechips varied between the wide limits of 50 and100 per cent.

2. The inlet is nearly always arranged to behigher than the exit. This elementary faultgives two problems. First, any floating slag ordross on the first metal to arrive is immedi-ately carried out of the trap before the trap isproperly filled (Figure 2.52c). Second, as wasrealized many years previously (Johnson andBaker 1948), the premature escape of metalhampers the setting up of a properly devel-oped spinning action. Thus the trap is slow todevelop its effect, perhaps never achieving itsfull speed in the short time available duringthe pour. This unsatisfactory situation is alsoseen in the work of Jirsa (1982), who describesa swirl trap for steel casting made frompreformed refractory sections. In this designthe exit was again lower than the entrance,and filling of the trap was encouraged merelyby making the exit smaller than the entrance.



Much metal and slag almost certainly escapedbefore the trap could be filled and becomefully operational. Since 90 per cent efficiencywas claimed for this design it seems probablethat all of the remaining 10 per cent whichevaded the trap did so before the trap wasfilled (in other words, the trap was workingat zero efficiency during this early stage).Jeancolas et al. (1971) report an 80 per centefficiency for their downhill swirl trap forbronze and steel casting, but admit that thetrap does not work at all when only partly full.

3. In many designs of swirl trap there isinsufficient attention paid to providing ac-commodation for the trapped material. Forinstance, where the swirl trap has a closed topthe separated material will collect against thecentre of the ceiling of the trap. However,work with transparent models illustratesclearly how perturbations to the flow causethe inclusions, especially if small, to ebb andflow out of these areas back into the mainflow into the casting (Jeancolas et al. 1971;Trojan et al. 1966). Also, of course, traps ofsuch limited volume are in danger of becom-ing completely overwhelmed, becoming sofull of slag or dross that the flow into thecasting becomes necessarily contaminated.

Where the trap has an open top the parabolicform of the liquid surface assists the con-centration of the floating material in the central`well' as shown in Figure 2.52d. The extra heightfor the separated materials to rise into is usefulto keep the unwanted material well away fromthe exit, despite variations from time to time inflow rate. Some workers have opened out thetop of the trap, extending it to the top of the

cope, level with the pouring bush. This certainlyprovides ample opportunity for slag to float wellclear, with no danger of the trap becomingoverloaded with slag. However, the author doesnot recommend an open system of this kind,because of the instability which open-channelsystems sometimes exhibit, causing surging andslopping between the various componentscomprising the `U'-tube effect between thesprue, swirl trap and mould cavity.

It is clear that the optimum design for theswirl trap must include the features:

(i) the entrance at the base of the trap;(ii) the exit to be sited at a substantially higher

level;(iii) both entrance and exit to have similar

tangential direction, and(iv) an adequate height above the central axis to

provide for the accumulation of separateddebris.

In most situations the inlet will be moulded inthe drag, and the exit in the cope, which is themost marginal difference in level between thetwo. At the high speeds at which the metal canbe expected to enter the trap the metal willsurge over this small ledge with ease, takinginclusions directly into the casting, particularlyif the inlet and outlet are in line as shown inFigure 2.52b. This simplest form of cope/dragparting line swirl trap cannot be expected towork.

The trap may be expected to work somewhatmore effectively as the angle of the outlet pro-gresses from 90±180 degrees. (The 270 degreeoption would be more effective still, except thatsome reflection will show it to be unmouldable

Figure 2.52 Swirl traps showing (a) incorrectopposed inlet and exit ducts; (b) correcttangential arrangements; (c) incorrect low exit;(d) correct high exit.



on this single joint line; the exit will overlay theentrance ports! Clearly, for the 270 degreeoption to be possible, the entrance and exitshave to be moulded at different levels, necessi-tating a second joint line provided by a core oradditional mould part.)

When using preformed refractory sections, orpre-formed baked sand cores, as is common forlarger steel castings, the exit can with advantagebe placed considerably higher than the entrance(Figure 2.52d).

These simple rules are designed to assist thetrap to spin the metal up to full speed before theexit is reached, and before any floating oremulsified less-dense material has had a chanceto escape.

For the separation of particulate slag inclu-sions from some irons and steels, Castings 1991showed that a trap 100 mm diameter in therunning system of moulds 0.1 to 1 m high wouldbe expected to eliminate inclusions of 0.2 to0.1 mm respectively. The conclusion was that,when correctly designed, the swirl trap could bea useful device to divert unwanted buoyantparticles away from ferrous castings.

We have to remember, however, that it is notexpected to work for film type inclusions.Compared to particles, films would be expectedto take between 10 and 100 times longer toseparate under an equivalent field force. Thusmost of the important inclusions in a largenumber of casting alloys will not be effectivelytrapped. Thus the alloys that need the techniquemost are least helped.

This damning conclusion applies to otherfield forces such as electromagnetic techniquesthat have recently been claimed to removeinclusions from melts. It is true that forces canbe applied to non-conducting particles sus-pended in the liquid. However, whereas com-pact particles move relatively quickly, and canbe separated in the short time available while themelt travels through the field. Films experiencethe same force, but move too slowly because oftheir high drag, and so are not removed.

In summary, we can conclude that apartfrom certain designs of by-pass trap, othervarieties of traps are not recommended. Ingeneral they almost certainly create moreinclusions than they remove.

2.3.6.3 Filters

Filters take many forms: as simple strainers,woven ceramic cloths, and ceramic blocks ofvarious types. Naturally, their effectivenessvaries from application to application, as isdiscussed here.

Strainers

A sand or ceramic core may be moulded toprovide a coarse array of holes, of a size anddistribution resembling a domestic colander.A typical strainer core might be a cylinder 30 to50 mm diameter, 10 to 20 mm long, containing10 or more holes, of diameter approximately3±5 mm (Figure 2.53a). These devices are mainlyused to prevent slag entering iron castings. Thedomestic colander is usually used to strainaggregates such as peas. These represent solidspherical particles of the order of 5 mm dia-meter. Thus, when applied to most metal cast-ings, the rather open design of strainers meansthat they can hardly be expected to perform anysignificant role as filters.

In fact, Webster (1967) has concluded thatthe strainer works by reducing the rate of flowof metal, assisting the upstream parts of thefilling system to prime, and thereby allowing theslag to float. It can be held against the topsurface of the runner, or in special reservoirsplaced above the strainers to collect the retainedslag. Webster goes on to conclude that if thestrainer only acts to reduce the rate of flow, thenthis can be carried out more simply and cheaplyby the proper design of the running system.

This may not be the whole story. Thestrainer may be additionally useful to laminizethe flow (i.e. cause the flow to become morestreamlined).

However, whatever benefits the strainer mayhave, its action to create jets downstream of thestrainer is definitely not helpful. The placing of astrainer in a geometry that will quickly fill theregion at the back of the strainer would be a

Figure 2.53 Various filters showing (a) a strainer core(hardly a filter at all); (b) woven cloth or mesh, forming atwo-dimensional filter; (c) ceramic foam and extrudedblocks, constituting three-dimensional filters.


(c)

(a) (b)


great advantage. A geometry to suppress jettingis provided by the tangential filter print to bediscussed later (Figure 2.56). The extruded orpressed ceramic filters with their arrays of par-allel pores are, of course, equivalent to strainerswith a finer pore size. They also benefit from thetangential placement to the oncoming flow aswill be described.

Over the years there has been much workcarried out to quantify the benefits of the useof filters. Nearly all of these have shownmeasurable, and sometimes important, gains infreedom from defects and improvements inmechanical properties. These studies are toonumerous to list here, but include metals of alltypes, including Al alloys, irons and steels. Therelatively few negative results can be traced tothe use of unfavourable siting or geometry ofthe filter print. For positive and reliable results,these aspects of the use of filters cannot beoverlooked. Special attention is devoted to themin what follows.

Woven cloth or mesh

For light alloys, steel wire mesh or glass cloth(Figure 2.53b) is used to prevent the oxides fromentering the casting. Cloth filter material has thegreat advantage of low cost.

The surprising effectiveness of these ratheropen meshes is the result of the most importantinclusions being in the form of films, whichappear to be intercepted by and wrap aroundthe strands of the mesh. Openings in the mesh orweave are typically 1±2 mm; this gives goodresults, being highly effective in retaining filmsdown to this size range. Significantly, it is also aconfirmation of the large size of the majority offilms that cause problems in castings, particu-larly in light alloys.

The use of steel wire mesh is also useful toretain films. The steel does not have time to gointo solution during the filling of aluminiumalloy castings, so that the material of the castingis in no danger of contamination. However, ofcourse, the steel presents a problem of ironcontamination during the recycling of the run-ning system. Even the glass cloth can sometimescause problems during the break-up of themould, when fragments of glass fibre can befreed to find their way into the atmosphere ofthe foundry, and cause breathing problems foroperators. Both materials therefore need carein use.

Some glass cloth filters are partially rigidizedwith a ceramic binder, and some by impregna-tion with phenolic resin. (The outgassing of theresin can cause the evolution of large bubbleswhen contacted by the liquid metal. Provided

the bubbles do not find their way into thecasting the overall effect of the filters is defi-nitely beneficial in aluminum alloys.) Both typessoften at high temperature, permitting the clothto stretch and deform.

A woven cloth based on a high silica fibre hasbeen developed to avoid softening at thesetemperatures, and might therefore be very suit-able for use with light alloys. In fact at thepresent time its high-temperature performanceusually confines its use to copper-based alloysand cast irons. There are few data to report onthe use of this material. However, it is expectedthat its use will be similar to that of the othermeshes, so that the principles discussed hereshould still apply.

Despite the attraction of low cost, it has to beadmitted that, in general, the glass cloth filtersare not easy to use successfully.

For instance, as the cloth softens and stret-ches there is a strong possibility that the clothwill allow the metal to by-pass the filter. It isessential to take this problem into account whendeciding on the printing of the filter. Clearly, itis best if it can be firmly trapped on all of itsedges. If it can be held on only three of its fouredges the vulnerability of the unsupported edgeneeds careful consideration. For instance, eventhough a cross-joint filter may be properly heldon the edges that are available, the filter issometimes defeated by the leading edge of thecloth bending out of straight, bowing like a sail,and thus allowing the liquid to jet past. All thefilters shown in Figure 2.54 show this problem.It may be better to abandon cloth filtration ifthere is any danger of the melt jetting around acollapsing filter.

When sited at the point where the flowcrosses a joint as in Figure 2.54 a greensandmould will probably hold the cloth successfully,the sand impressing itself into the weave, pro-vided sufficient area of the cloth is trapped inthe mould joint. In the case of a hard sandmould or metal die, the cloth requires a shallowprint which must be deep enough to allow itroom if the joint is not to be held apart. Also, ofcourse, the print must not be too deep, other-wise the cloth will not be held tight, and may bepulled out of position by the force of the liquidmetal. Some slight crushing of a hard sandmould is desirable to hold the cloth as firmly aspossible.

A rigidized cloth filter can be inserted acrossthe flow by simply fitting it into the pattern in apre-moulded slot across the runner and somoulding it integral with the mould (Figure 2.54c).However, this is only successful for relativelysmall castings. Where the runner area becomeslarge and the time and temperature become too



high, the filter softens and bows in the force ofthe flow. Even if it is not entirely pulled out ofposition it may be deformed to sag like a fencein a gale, so that metal is able to flow over thetop. This is the reason for the design shown inFigure 2.54e. The edge of the filter crosses thejoint line, either to sit in a recess accuratelyprovided on the other half of the mould, or if theupstand is limited to a millimetre or so, to besimply crushed against the other mould half.(The creation of some loose grains of sand is oflittle consequence in the running system, as hasbeen shown by Davis and Magny (1977); loose

material in the runner is never picked up by themetal and carried into the mould. The authorcan confirm this observation as particularly truefor systems that are not too turbulent. Thelaminizing action of the filter itself is probablyadditionally helpful.)

If the filter is introduced at an earlier stage ofmanufacture during the production of a sandmould, it can be placed in position in a slot cutin the runner pattern. When the sand is intro-duced the filter is automatically bonded intothe mould (Figure 2.54d). Again, an upstandabove the level of the joint may be useful

Figure 2.54 Siting of cloth filters (a) in the mould joint; (b) in a double crossing of the joint; (c) in a slot moulded acrossthe runner; (d) in a slot cut in the runner pattern; (e) with an additional upstand across the joint plane to assist sealing.


(a) (b)

(c) (d)

(e)


(Figure 2.54e). In any case, when using filtersacross runners, it also helps to arrange for theselvage (the reinforced edge of the material) ofthe cloth to be uppermost to give the unsup-ported edge most strength; the ragged cut edgehas little strength, letting the cloth bend easily,and allowing some, or perhaps all, of the flow toavoid the filter. All the cloth filters used asshown in Figure 2.54 are defeatable, since theyare held only on three sides. The fourth side isthe point of weakness. Failure of the filter by theliquid overshooting this unsupported edge canresult in the creation of more oxide dross thanthe filter was intended to prevent! Increasing thetrapped area of filter in the mould joint cansignificantly reduce the problem.

Geometries that combine bubble traps (orslag or any other low density phase) are shownin Figure 2.55 for in-line arrangements, and forthose common occasions when the runner isrequired to be divided to go in opposite direc-tions. With shallow runners of depth of a fewmillimetres there is little practical difference inwhether the metal goes up or down through thefilter. Thus several permutations of these geo-metries can be envisaged. Much depends on thelinks to the gates, and how the gates are to beplaced on the casting. In general, however,I usually aim to have the runner exit from thefilter below the joint.

Cloth filters are entirely satisfactory wherethey can be held around all four sides. This is thecase at the point where gates are taken verticallyupwards from the top of the runner. This isa relatively unusual situation, where, instead ofa two-parted mould, a third mould part forms abase to the mould and allows the runner systemto be located under the casting. Alternatively, aspecial core can be used to create an extra jointbeneath the general level of the casting.

Another technique for holding the filter onall sides is the use of a `window frame' of strongpaper or cardboard that is bonded or stapled tothe cloth. The frame is quickly dropped into itsslot print in the mould, and gives a low cost rigidsurround that survives sufficiently long to beeffective.

Ceramic block filters

Ceramic block filters of various types intro-duced in about 1980 have become popular, andhave demonstrated impressive effectiveness inmany applications in running systems.

Unfortunately, much that has been writtenabout the mechanisms by which they clean upthe cast material appears to be irrelevant. This isbecause most speculation about the filtrationmechanisms has considered only particulate

inclusions. As has become quite clear over recentyears, the most important and widespreadinclusions are actually films. Thus the filtrationmechanism at work is clearly quite different,and, in fact, easily understood.

In aluminium alloys the action of a ceramicfoam filter to stop films has, in general, not beenrecognized. This is probably because the filmsare so thin (a new film may be only 20 nm thick,making the doubled-over entrained bifilm stillonly about 40 nm) that they cannot be detectedwhen wrapped around sections of the ceramicfilter. This explains part of the curious experi-ence of finding that a filter has cleaned up acasting, but on sectioning the filter to examine it

Figure 2.55 Uses of glass cloth filtration (a) for an in linerunner; (b) for transverse runners.


(a)

Glass cloth

(b)


under the microscope, not a single inclusion canbe found.

The contributing effect, of course, is that thefilter acts to improve the filling behaviour ofthe casting, so reducing the number of inclu-sions that are created in the mould during thefilling process. This behaviour was confirmedby Din et al. (2003) who found only about10 per cent of the action of the filter was theresult of filtration, but 90 per cent was theresult of improved flow.

A further widespread foundry experience isworth a comment. On occasions the quantity ofinclusions has been so great that the filter hasbecome blocked and the mould has not filledcompletely. Such experiences have caused someusers to avoid the use of filters. However, in theexperience of the author, such unfortunateevents have resulted from the use of poor frontends of filling systems (poor basin and spruedesigns) that create huge quantities of oxidefilms in the pouring process. The filter hastherefore been overloaded, leading either to itsapparently impressive performance, or its fail-ure by blockage. The general advice given tousers by filter manufacturers that filters willonly pass limited quantities of metal is seen to beinfluenced by similar experience. The author hasnot found any limit to the volume of metal thatcan be put through a filter without danger ofblockage, provided the metal is sufficiently cleanand the front end of the filling system isdesigned to perform well.

Thus the secret of producing good castingsusing a filter is to team a good front end of thefilling system together with the filter. (If theremainder of the filling system design is good, thiswill, of course, help additionally.) Little oxide isthen entrained, so that the filter appears to dolittle filtration. However, it is then fully enabled toserve a valuable role as a flow rate control device.The beneficial action of the filter in this case isprobably the result of several factors:

(i) The reduction in velocity of the flow(provided an appropriately sized cross-section area channel is provided down-stream of course). This is probably thesingle most important action of the filter.However, there are other important actionslisted below.

(ii) Reduces the time for the back-filling of thesprue, thereby reducing entrainment defectsfrom this source.

(iii) Smooths fluctuations in flow.(iv) Laminizes flow, and thus aids fluidity a

little.(v) The freezing of part of the melt inside the

filter by the chilling action of the filter (as

predicted for Al alloys in ceramic foamfilters by computer simulation and foundby experiment by Gebelin and Jolly 2003)may be an advantage, because this may actto restrict flow, and so to reduce deliveryfrom the filter in its early moments. Thesubsequent re-melting of the metal as morehot metal continues to pass through thefilter will allow the flow to speed up to itsfull rate later during filling. (Interestingly,this advantage did not apply to preheatedceramic moulds where the preheat wassufficient to prevent any freezing in thefilter).

There are different types of ceramic block filter.

(i) Foam filters made by impregnation ofopen-cell plastic foams with a ceramicslurry, squeezing out the excess slurry, andfiring to burn out the plastic and developstrength in the ceramic. The foam structureconsists of a skeleton of ceramic filamentsand struts defining a network of intercon-necting passageways.

(ii) Extruded forms that have long, straight,parallel holes. They are sometimes referredto as cellular filters.

(iii) Pressed forms, again with long, straight butslightly tapered holes. The filters are madeindividually from a blank of mouldable clayby a simple pressing operation in a two-partsteel die.

(iv) Sintered forms, in which crushed andgraded ceramic particles are mixed with aceramic binder and fired.

In all types the average pore size can be con-trolled in the range 2±0.5 mm approximately,although the sintered variety can achieve at least2±0.05 mm. Insufficient research (other thanthat funded by the filter manufacturers!) hasbeen carried out so far to be sure whether thereare any significant differences in the perfor-mance between them. An early result of Khanet al. (1987) found that the fatigue strength ofductile iron was improved by extruded cellularfilters, but that the foam filters were unpredict-able, with results varying from the best to theworst. Their mode of use of the filters was lessthan optimum, being blasted by metal in theentrance to the runner, and with no back pro-tection for the melt. (We shall deal with theseaspects below.) The result underlines the prob-able unrealized potential of both types, andreminds us that both would almost certainlybenefit from the use of recent developments. Ingeneral, we have to conclude that the publishedcomparisons made so far are, unfortunately,often not reliable.



For aluminium alloys the results are lesscontroversial, because the filters are highlyeffective in removing films which have, ofcourse, a powerful effect on mechanical prop-erties. Mollard and Davidson (1978) are typicalin their findings that the strength of Al±7Si±Mgalloy is improved by 50 per cent, and elongationto failure is doubled. This kind of result is nowcommon experience in the industry.

For some irons and steels, where a highproportion of the inclusions will be liquid, mostfilter materials are expected to be wetted by theinclusion so that collection efficiency will behigh for those inclusions. Ali et al. (1985) foundthat for alumina inclusions in steel traversing analumina filter, once an inclusion made contactwith the filter it became an integral part of thefilter. It effectively sintered into place; despitethe fact that both inclusion and filter are solid atthe temperature of operation, they behave asthough they are `sticky'. This behaviour is likelyto characterize many types of inclusion at thetemperature of liquid steel.

In contrast with this, Wieser and Dutta(1986) find that whereas alumina inclusions insteel are retained by an alumina filter, even up tothe point at which it will clog, deoxidation ofsteel with Mn and Si produces silica-containingproducts that are not retained by an extrudedzirconia spinel filter. These authors also testedvarious locations of the filter, discovering thatplacing it in the pouring basin was of nouse, because it was attacked by the slag anddissolved!

Although these results might have beeninfluenced by the rapid flow rates that appear tohave been used in this work, it is a warning thatfiltration efficiency is likely to be stronglydependent on inclusion and filter types. Ali et al.(1985) confirms this strong effect of velocity,finding only at very low velocities measured inmm sÿ1 was a high level (96 per cent) of filtra-tion achieved in steel melts.

Block filters are more expensive than clothfilters. However, they are easier to use and morereliable. They retain sufficient rigidity to mini-mize any danger of distortion that might resultin the by-passing of the filter. It is, however,important to secure a supply of filters that aremanufactured within a close size tolerance, sothat they will fit immediately into a print in thesand mould or into a location in a die, withminimal danger of leakage around the sides ofthe filter. Although all filter types haveimproved in this respect over recent years, thefoam filter seems most difficult to control, theextruded is intermediate, whereas the pressedfilter exhibits good accuracy and reproducibilityas a result of it being made in a steel die; residual

variation seems to be result of poor control ofshrinkage on firing.

Leakage control

It is essential to control the leakage past thefilter. There are various techniques.

(i) A seating of a compressible gasket ofceramic paper. This approach is useful whenintroducing filters into metal dies, where thefilter is held by the closing of the two halvesof the die. The variations in size of the filters,and the variability of the size and fit of thedie parts with time and temperature, whichwould otherwise cause occasional crackingor crushing of the rather brittle filter, areaccommodated safely by the gasket.

(ii) Moulding the filter directly into an aggre-gate (sand) mould. This is achieved simplyby placing the filter on the pattern, andfilling the mould box or core box withaggregate in the normal way. The filter isthen perfectly held. In greensand systems orchemically bonded sands the mould materi-al seems not to penetrate a ceramic foamfilter more than the first pore depth. This isa smaller loss than would be suffered whenusing a normal geometrical print. However,the technique often requires other measuressuch as the moulding of the filter into aseparate core, or the provision of a loosepiece in the pattern to form the channel onthe underside of the filter.

There are other aspects of the siting of filters inrunning systems that are worth underlining.

(i) Siting a filter so that some metal can flow by(into a slag trap for instance) prior topriming the filter is suggested to have theadditional benefit that the preheat of thefilter and the metal reduces the primingproblem associated with the chilling of themetal by the filter (Wieser and Dutta 1986).An example is seen in Figure 2.56d.

(ii) The area of the filter needs to be adequate.There is much evidence to support the factthat the larger the area (thereby giving alower velocity of flow through the filter) thebetter the effectiveness of the filter. Forinstance, if the filter area is too small inrelation to the velocity of flow then the filterwill be unable to retain foreign matter: theforce of the flow will strip away retainedfilms like sheets from the washing line in ahurricane; particles and droplets will followa similar fate.



(a)

(b)

(c)

(d)

Larger area forslower exit flow

‘Well’ to protectexit from filter

.�� # () !�� . �� . �� . �� '


(iii) Many filter placements do not distribute theflow evenly over the whole of the filtersurface. Thus a concentrated jet is unhelpful,being equivalent to reducing the active areaof the filter. The tangential placement of afilter can also be poor in this respect, since theflow naturally concentrates through thefarthest portion of the filter. This is coun-tered by tapering the tangential entranceand exit flow channels as illustrated inFigure 2.57b. The provision of a bubble trapreduces the effectiveness of the taper, but thepresence of the trap is probably worth thissacrifice. (If the trap is not provided, bubblesarriving from entrainment in the basin orsprue gather on the top surface of the filter.When they have accumulated to occupyalmost the whole of the area of the filter thesingle large bubble is then forced through thefilter, and travels on to create severe prob-lems in the mould cavity. The trap is expectedto be similarly useful for the diversion of slagfrom the filter face during the pouring ofirons and steels.)

Mutharasan et al. (1981) find that the efficiencyof removal of TiB2 inclusions from liquid alu-minium increased as the velocity through thefilter fell from about 10 mm sÿ1 to 1 mm sÿ1.Later, the same authors found identical beha-viour for the removal of up to 99 per cent ofalumina inclusions from liquid steel (Muthar-asan et al. 1985). However, it is to be noted thatthese are extremely low velocities, lower thanwould be found in most casting systems. In thework by Wieser and Dutta (1986) on the filtra-tion of alumina from liquid steel, somewhathigher velocities, in the range 30±120 mm sÿ1,are implied despite the use of filter areas up toten times the runner area in an attempt to obtainsufficient slowing of the rate of flow. Even these

flow velocities will not match most runningsystems. These facts underline the poverty of thedata that currently exists in the understandingof the action of filters.

Wieser and Dutta go on to make the inter-esting point that working on the basis of pro-viding a filter of sufficient size to deal with theinitial high velocity in a bottom-gated casting,the subsequent fall in velocity as the casting fillsand the effective head is reduced implies that thefilter is oversize during the rest of the pour.However, this effect may be useful in counteringthe gradual blockage of the filter in steel con-taining a moderate amount of inclusions.

Use of filters in running systems

In general the correct location for the filter isnear the entrance to the runner, immediatelyfollowing the sprue. The resistance to penetra-tion of the pores of the filter by the action ofsurface tension is an additional benefit, delayingthe entry into the filter until the sprue has atleast partially filled. The frictional resistance toflow through the filter once it is operationalprovides a further contribution to the reductionin speed of the flow. This frictional resistancehas been measured by Devaux (1987). He findsthe head loss to be large for filters of area onlyone or two times the area of the runner. Heconcludes that whereas a filter area of twice therunner area is the minimum size that is accept-able for a thick-section casting, the filter areahas to be increased to four times the runner areafor thin-section castings. The pressure dropthrough filters is a key parameter that is notknown with the accuracy that would be useful.Midea (2001) has attempted to quantify thisresistance to flow but used only low flow velo-cities useful for only small castings. A slightimprovement is available with Lo and Campbell(2000) who study flow up to 2.5 m sÿ1. Even so,at this time the author regrets that it remainsunclear how these measurements can be used ina design of a running system. A clear workedexample would be useful for us all.

The filter positioned at the entrance to therunner also serves to arrest the initial splash ofthe first metal to arrive at the base of the sprue.At the beginning of the runner the filter is ide-ally positioned to take out the films createdbefore and during the pour. The clean liquid canbe maintained relatively free from further con-tamination so long as no surface turbulenceoccurs from this point onwards. This conditioncan be fulfilled if:

(i) The melt proceeds at a sufficiently lowvelocity and/or is sufficiently constrained

Figure 2.57 (a) Concentration and reverse flow in a foamfilter; (b) tapered inlet and exit ducting to spread flow.


(a)

(b)


by geometry to prevent entrainment (thisparticularly includes the provision to elim-inate jetting from the rear face of the filter).A low velocity will be achieved if the cross-sectional area of the runner downstream ofthe filter is increased in proportion to thereduction in speed provided by the filter.

(ii) Every part of the subsequent journey for theliquid is either horizontal or uphill. Thecorollary of this condition is that the base ofthe sprue and the filter should always be atthe lowest point of the running system andthe casting. This excellent general rule is akey requirement.

Tangential placement

Filters have been seen to be open to criticismbecause of their action in splitting up the flow,thereby, it was thought, probably introducingadditional oxide into the melt. There is some truthin this concern. A preliminary exploration of thisproblem was carried out by the author (Din andCampbell 1994). Liquid Al alloy was recorded onoptical video flowing through a ceramic foamfilter in an open runner. The filter did appear tosplit the flow into separate jets; a tube of oxideforming around each jet. However, close obser-vation indicated that the jets recombined about10 mm downstream from the filter, so that air wasexcluded from the stream from that pointonwards. The oxide tubes around the jetsappeared to wave about in the eddies of the flow,remaining attached to the filter, like weedattached to a grill across a flowing stream. Thestudy was repeated and the observations con-firmed by X-ray video radiography. The work wascarried out at modest flow velocities in the regionof 0.5 to 2.0 m sÿ1. It is not certain, however,whether the oxides would continue to remainattached if speeds were much higher, or if the flowwere to suffer major disruption from, for instance,the passage of bubbles through the filter.

What is certain is the damage that is done tothe stream after the filter if the melt issuing fromthe filter is allowed to jet into the air. Loper andco-workers (1996) call this period during whichthis occurs the spraying time. This is so serious aproblem that it is considered in some detail below.

Unfortunately, most filters are placed trans-verse to the flow, simply straight across a runner(Figure 2.56a) and in locations where the pressureof the liquid is high (i.e. at the base of the sprue orentrance into the runner). In these circumstances,the melt shoots through a straight-through-holetype filter almost as though the filter was notpresent, indicating the such filters are not parti-cularly effective when used in this way. When afoam filter suffers a similar direct impingement,

penetration occurs by the melt seeking out theeasiest flow paths through the various sizes ofinterconnected channels, and therefore emergesfrom the back of the filter at various randompoints. Jets of liquid project from these exitpoints, and can be seen in video radiography. Thejets impinge on the floor of the runner, and on theshallow melt pool as it gradually builds up,causing severe local surface turbulence and socreating dross. If the runner behind the filter islong or has a large volume, the jetting behaviourcan continue until the runner is full, creatingvolumes of seriously damaged metal.

Conversely, if the volume of the filter exitchannel is kept small, the volume of damaged meltthatcanbeformedisnowreducedcorrespondingly.Although this factor has been little researched, it iscertain to be important in the design of a goodplacement for the filter. Loper et al. (1996) realizedthis problem, describing the limited volume at theback of the filter as a hydraulic lock, the word lockbeing used in a similar sense to a lock on an inlandwaterway canal.

Figure 2.56b shows an improved geometrythat enables the back of the filter to be coveredwith melt quickly. Figure 2.56c shows animproved technique, placing the block filtertangentially to the direction of flow. The tan-gential mode has the advantage of the limitationof the exit volume from the filter, and providinga geometrical form resembling a sump, or lowestpoint, so that the exit volume fills quickly. Inthis way the opportunity for the melt to jet freelyinto air is greatly reduced so that the remainderof the flow is protected. A further advantage ofthis geometry is the ability to site a bubble trapover the filter, providing a method whereby theflow of metal and the flow of air bubbles can bedivided into separate streams. The air bubbles inthe trap are found to diffuse away gradually intosand moulds. For dies, the traps may need to belarger.

An additional benefit is that the straight-through-hole extruded or pressed filters seem tobe effective when used tangentially in this way.A study of the effectiveness of tangential place-ment in the author's laboratory (Prodham et al.1999) has shown that a straight-through-porefilter could achieve comparable reliability ofmechanical properties as could be achieved by arelatively well-placed ceramic foam filter (Sirrelland Campbell 1997).

Adams (2001) draws attention to the impor-tance of the flow directed downwards throughthe filter. In this way buoyant debris such asdross or slag can float clear. In contrast, withupward flow through the filter the buoyantdebris collects on the intake face of the filter andprogressively blocks the filter.



The tapering of both the tangential approachand the off-take from the filters further reducesthe volume of melt, and distributes the flowthrough the filter more evenly. In the absenceof these wedge-like features, only the far side ofthe filter carries the main flow, whereas the sidenearest the upstream end is redundant, experi-encing a circulating flow in the reverse direction(Figure 2.57).

Direct pour

Sandford (1988) showed that a variety of toppouring could be used in which a ceramic foamfilter was used in conjunction with a ceramicfibre sleeve. The sleeve/filter combination wasdesigned to be sited directly on the top of amould to act as a pouring basin, eliminating anyneed for a conventional filling system. In addi-tion, after filling, the system continued to workas a feeder. This simple and attractive systemhas much appeal.

Although at first sight the technique seems toviolate the condition for protection of the meltagainst jetting from the underside of the filter,jetting does not seem to be a problem in thiscase. Jetting is avoided almost certainly becausethe head pressure experienced by the filter is solow, and contrasts with the usual situationwhere the filter experiences the full blast of flowemerging from the base of the sprue.

Sandford's work illustrated that without thefilter in place, direct pour of an aluminium alloyresulted in severe entrainment of oxides in thesurface of a cast plate. The oxides were elimi-nated if a filter was interposed, and the fall afterthe filter was less than 50 mm. Even after a fallof 75 mm after the filter relatively few oxideswere entrained in the surface of the casting. Thetechnique was further investigated in somedetail (Din et al. 2003) with fascinating resultsillustrated in Figure 2.58. It seems that underconditions used by the authors in which the meltemerging from the filter fell into a runner barand series of test bars, some surface turbulencewas suffered, and was assessed by measuring thescatter of tensile test results. The effect of thefilter acting purely to filter the melt was seen tobe present, but slight. The castings were found tobe repeatable (although not necessarily free fromdefects) for fall distances after the filter of up toabout 100 mm, in agreement with Sandford.Above a fall of 200 mm reproducibility was lost(Figure 2.58a, b).

This interesting result explains the mix ofsuccess and failure experienced with the directpour system. For modest fall heights of 100 mmor so, the filter acts to smooth the perturbations

to flow, and so confers reproducibility on thecasting. However, this may mean 100 per centgood or 100 per cent bad. The difference wasseen by video radiography to be merely thechance flow of the metal, and the consequentialchance location of defects.

The conclusion to this work was a surprise. Itseems that direct pour should not necessarily beexpected to work first time. If the techniquewere found to make a good casting it should beused, since the likelihood would be that allcastings would then be good. However, if thefirst casting was bad, the site of the filter andsleeve should simply be changed to seek a dif-ferent pattern of filling. This could mean a siteonly a few centimetres away from the originalsite. The procedure could be repeated until a sitewas found that yielded a good casting. Thelikelihood is that all castings would subse-quently be good.

However, the technique will clearly not beapplicable to all casting types. For instance, it isdifficult to see how the approach could reliablyproduce extensive relatively thin-walled pro-ducts in film-forming alloys where surface ten-sion is not quite in control of the spread of metalin the cavity. For such products the advance ofthe liquid front is required to be steady, repro-ducible and controlled. Bottom gating in such acase is the obvious solution. Also, the techniqueworks less well in thicker section castings wherethe melt is less constrained after its fall from theunderside of the filter. Figure 2.58c illustratesthe fall in reliability of products as the diameterof the test bars increases above 20 mm. Equi-valent results would be expected for the increaseof plate sections above about 10 mm.

Even though the use and development of thedirect pour technique will have to proceed withcare, it is already achieving an important placein casting production. A successful applicationto a permanent moulded cylinder head casting isdescribed by Datta and Sandford (1995). Suc-cess here appears to be the result of the limited,and therefore relatively safe fall distance.

Flow rate data through the filter/sleevecombination is necessary to predict the pourtimes of castings. Such data has been measuredby Bird (1989). His results presented here (Fig-ure 2.58d) have been rationalized to apply to50 mm diameter ceramic foam filters, and relateto Al±Si alloys cast at 720 �C. Clearly, filters ofdifferent sizes will pass correspondingly more orless melt per second proportional to their areas,assuming their thickness and pore sizes aresufficiently close. The pores' diameters ofapproximately 1 and 2 mm in Figure 2.58d referto the `pore per inch' categories 20 ppi and10 ppi respectively.



A recent development of the direct pourtechnique is described by Lerner and Aubrey(2000). For the direct pouring of ductile ironthey use a filter that is a loose fit in the ceramicsleeve. It is held in place by the force of impin-gement of the melt. When pouring is completethe filter then floats to the top of the sleeve andcan be lifted off and discarded, avoiding con-tamination of remelted material.

Sundry aspects

1. The dangers of using ceramic block filters inthe direct impingement mode is illustrated bythe work of Taylor and Baier (2003). Theyfound that a ceramic foam filter placedtransversely in the down-sprue worked better

at the top of the sprue rather than at its base.This conclusion appears to be the result ofthe melt impact velocity on the filter causingjetting out of the back of the filter. Thus thehigh placement was favoured for the reasonsoutlined in the section above on the directpour technique. This result is unfortunate,because if the filter exit volume had beenlimited to a few millilitres (a depth beneaththe filter limited to 2 or 3 mm) the lower sitingof the filter would probably have performedin the best way.

2. It is essential for the filter to avoid thecontamination of the melt or the meltingequipment. Thus for many years there ap-peared to be a problem with Al±Si alloys thatappeared to suffer from Ca contamination

Figure 2.58 Direct pour filtration showing (a) the reduced reliability as the fall increases with and without a filter inplace; (b) the interpretation of `a'; (c) the reduced reliability as diameter of test bars is increased; (d) the rate offlow of Al±Si alloys at 720 �C by direct pour through a 50 mm diameter filter (Bird 1989, courtesy of Foseco).


00 12.5 25 50 100 200 400 800

5

0

20

40

30 ppi Filter20 ppi Filter10 ppi FilterNo Filter

elon

gatio

n

Wei

bull

mod

ulus

Fall height after the filter, mm

tens

ile s

tren

gth

60(a) (b)

(c) (d)

12.5

Fall height after the filter, mm

00

20

40

60

Rel

iabi

lity

fact

or (

Wei

bull

mod

ulus

)

80

100

80025

Effect of filter

Direct pourfiltration

No filter Effect offlow

50 100 200 400

H

100

Elongation

Tensile strength

Rel

iabi

lity

fact

or (

Wei

bull

mod

ulus

)

Rel

iabi

lity

fact

or (

Wei

bull

mod

ulus

)

80

60

40

20

0

10

8

6

4

2

00 5 10 15 20

Section diameter, mm

25 30

1.5

1.0

Flo

w r

ate

(kg

s–1)

0.5

00 50 100

Head height above filter (mm)

Pore diameter2 mm

Pore diameter1 mm

150


from an early formulation of the filterceramic. This problem now seems to beresolved by modification of the chemistry ofthe filter material. In addition, modern filtersfor Al alloys are now designed to float, sothat they can be skimmed from the top of themelting furnace when the running systemsare recycled. This avoids the costly cuttingout and separation of spent filters fromrecycled rigging to avoid them collecting ina mass at the bottom of the melting furnace.

Interestingly, the steel gauzes used for Al alloysdo not contaminate the alloy entering the cast-ing. This is almost certainly the result of thealloy wrapping a protective alumina film overthe wires of the mesh as the meniscus passesthrough. However, the steel will dissolve later ifrecycled via a melting furnace of course.

The recent introduction of carbon-based fil-ters for steel does add a little carbon to the steel,but this seems negligible for most grades.Whether the use of such filters will be suitablefor ultra-low carbon steels is being decided asI write.

Xu and Mampaey (1997) report the addi-tional benefit of a ceramic foam filter in animpressive 12-fold increase in the fluidity of greyiron poured at about 1400 �C in sand moulds.They attribute this unlooked-for bonus to theeffect of the filter in (i) laminizing the flow andso reducing the apparent viscosity due to tur-bulence, and (ii) reducing the content of inclu-sions. One would imagine that films would beparticularly important.

Summary

So far as can be judged at this time, among themany requirements to achieve a clean casting,the key practical recommendations for thecasting engineer can be summarized as:

(i) Do not allow slag and dross to enter thefilling system. This task is best solved byeliminating the conical pouring basin andsubstituting an offset stepped basin.

(ii) Use a good early part of the filling system toavoid the creation of additional slag ordross that may block the filter.

(iii) Use filters together with a buoyant phasetrap. The bubble trap described earliershould also work as a slag trap. Thepresence of the filter significantly aids theseparation of the two fluids. Where parti-cularly dirty metals are in use, the trap will,of course, require the provision of sufficientvolume and height on its upstream side toaccommodate retained material, allowing

slag and dross to float clear, and leaving thefilter area to continue working withoutblockage.

(iv) Avoid the great danger of by-passing thefilter by poor printing. Mould-in the filter ifpossible.

(v) Provide protection of the melt at the exitside of the filter, by rapid fill of this volumewith liquid metal. A useful geometry toachieve this is the tangential placement ofthe filter, followed by a shallow well thatcan be quickly filled.

2.3.7 Practical calculation of the filling system

In view of all the information listed under Rule 2in the previous part of this chapter, this sectionattempts to gather this together, to see how wemight achieve a complete, practical solution to afilling system design. The ability to design aquantitative solution, yielding precise dimen-sions of the filling channels at all points, is a keyresponsibility, perhaps the key responsibility, ofthe casting engineer.

Naturally, computers are beginning to havesome capability of optimizing the design offilling systems (McDavid and Dantzig 1989;Jolly et al. 2000). Even so, until the time that thecomputer is fully proficient, it will be necessaryfor the casting engineer to undertake this duty.The complication of the procedure is not to beunderestimated (if it were easy the procedurewould have been developed years ago). Manyfactors need to be taken into account. This shortoutline cannot cover all eventualities, but willpresent a systematic approach that will be gen-erally applicable.

2.3.7.1 Background to the methodingapproach

If a computer package is available to simulatethe solidification of the casting, it is best to carrythis out first. Most software packages are suf-ficiently accurate when confined to the simula-tion of solidification (it is the filling simulation,and other sophisticated simulations such as thatof stress, strain and distortion that are moredifficult, and the results often less accurate). Asolidification simulation with the addition of nofilling or feeding system will illustrate whetherthere are special problems with the casting.Figure 2.59 illustrates the formal logic of thisapproach. It lays out a powerful methodologythat is strongly recommended.

If in fact there are no special problems it isgood news. Otherwise, if problems do appearfor a long-running part, and if they can beeliminated by discussion with the designer of the



component at this initial stage, this is usually themost valuable strategy. Such actions ofteninclude the shift of a parting line, or the coring-out of a heavy section or boss. The purpose of amodest one-time design change is to avoid, so

far as possible, the ongoing expense for the life ofthe product of special actions such as the provi-sion of chills or feeders, or an extra core, etc.

If, despite these efforts, a problem remains,the various options including additional chills,feeders or cores will require detailed study tolimit, so far as possible, the cost penalties. Thefollowing section provides the background forthe next steps of the procedure.

2.3.7.2 Selection of a layout

First, it will be necessary to decide which way upthe part is to be cast (this may be changed laterin the light of many considerations, includingproblems of core assembly, desirable fillingpatterns, subsequent handling and de-gatingissues, etc.). If a two-part mould is to be used,the form of the casting should preferably bemainly in the cope, allowing gating at the lowestparts of the casting. This may prove so difficultthat a third box part may be selected throughwhich the running system could be sited underthe casting. If some solution to the challenge oflowest point gating cannot easily be found, therisk of filling the casting at some slightly higherpoint may need to be assessed. Some fillingdamage might have to be accepted for somecastings. Even so, it is unwelcome to have tomake such decisions because the extent of anysuch damage is difficult to predict.

A heavy section of the casting needs specialattention. This may most easily be achieved byorienting this part of the casting at the top andplanting a feeder here. Alternatively, other con-siderations may dictate that the casting cannot beoriented this way, so arrangements may have tobe made to provide chills and/or fins to this sec-tion if it has to be located in the drag.

When a general scheme is decided, includingthe approximate siting of gates and runners, theprovision of feeders, if any, and the location ofthe sprue, a start can be made on the quantifi-cation of the system.

2.3.7.3 Weight and volume estimate

The weight of the casting will be known, or canbe estimated. This is added to an estimate of theweight of the rigging (the filling and feedingsystem) to give an estimate of the total pouredweight. Dividing this by the density of the liquidmetal will give the total poured volume.Unfortunately, of course, the weight of the rig-ging is clearly not accurately known at this earlystage because it has not yet been designed.However, an approximate estimate is nearlyalways good enough. Although a revised valuecan always be used in a subsequent iteration of

Figure 2.59 Methoding procedure for computersimulation.


Start

Simulatesolidification

of componentas designed

Changedesign

Changedesign

possible?

Changesolidificationconditions

Simulatesolidification

OK? Inherent problem?

OK?Yes

Yes

No

No

No

No

YesYes

Success FundamentalRe-think


the rigging design calculations to obtain anaccurate value for the weight of the rigging,after some experience an additional iterationwill be found to be hardly ever necessary.

2.3.7.4 Selection of a pouring time

The selection of a pouring time is always aninteresting moment in the design of a fillingsystem for a new casting.

A common concern is how can productionrates be maintained high if metal velocities inthe filling system need to be kept below thecritical 0.5 m sÿ1? Fortunately, this is not usuallya problem because the time to fill a casting isdependent on the rate of mould filling measuredas a volume per second, and can be fixed at ahigh level. At the same time the velocity throughthe ingate can be independently lowered simply(well, simple in principle, but perhaps harder inpractice!) by increasing the area of the gate.These considerations will become clearer as weproceed.

When faced with a new design of casting, thefirst question asked by the casting engineer is`How fast should it be filled?'

Sometimes there is no choice. On a fastmoulding line making 360 moulds per hourthere is only 10 seconds for each complete cycle,of which perhaps only 5 seconds may be theavailable time for pouring. (Although it is worthkeeping in mind that even here, as a last resort,the pour time might be doubled if two pouringfacilities were to be installed.)

When there is a choice the pour time canoften be changed between surprisingly largelimits. One factor is sometimes the rate of rise ofthe metal in some sections. The surface of an Alcasting becomes marked with striations due tothe passage of transverse unzipping waves at avertical rise velocity below 60 mm sÿ1 (Evanset al. 1997). Considerations that control thechoice of rate of metal rise in steel castings(Forslund 1954, Hess 1974) indicate that thesefactors have yet to be properly researched. Inpractice, a common rate of rise in a steelfoundry making castings several metres talland weighing several tonnes is 100 mm sÿ1,although, with an improved design of fillingsystem this rate might be reduced. A furtherlimit to the fill time of a steel casting is thepossible collapse of the cope when subjected toradiant heat of the rising metal for too long.This problem is reduced by generous ventingof the top of the mould via a top feeder forinstance, and is further reduced by the practiceof providing a white mould coat based on amaterial such as alumina or zircon, thusabsorbing much less of the incident radiation.

A slow rate of rise in the mould cavity canlead to transverse unzipping waves, butalthough they can leave their witness on thesurface of the oxidized surface of the castingthey are usually harmless to its internal struc-ture. However, if the alloy has an extra strongsurface oxide, or is partially freezing because ofa cool pour, the waves lead to such severe sur-face horizontal laps that the casting is usuallynot repairable. Types of geometries where thisproblem is most often seen are illustrated inFigure 2.60. A hollow cylinder cast on its side isa common casualty (2.60a) because of the sud-den increase in area to be filled, reducing therate of rise as the metal reaches the top of thecore. The problem is also found on the uppersurfaces of tilt castings if the rate of tilt is tooslow (2.60b).

Alternatively another constraint on a choiceof pour time is the consideration that it may benecessary to fill the mould before freezing startsin its thinnest section (or, more usefully, its

Figure 2.60 Common lap problems at low rates of riseof liquid surface in the mould (a) in a horizontal pipe orcylinder; (b) on the cope surface of a tilt casting.


Sudden enlargementof area + decreasednet head

(a)

(b)

Surface lapdefects


smallest modulus). Thus an idea of the timeavailable can be gained from the Figure 2.61.(Readers are recommended to generate suchdiagrams for themselves for special castingconditions, using embedded thermocouples todetermine the freezing times versus modulusrelations, e.g. for cast iron in zircon shellmoulds, or aluminium-based alloys in invest-ment moulds at 100 �C, etc.)

Clearly filling at too slow a rate does bring itsproblems. However, many castings are filledvery much faster than necessary, and there arebenefits to a reduction in this speed. For thisreason, having made a choice of an approximatefill time for the casting, it is instructive to con-sider whether this time could be doubled, or evendoubled again. It is surprising how often this is apossibility. Whereas the experienced foundry-man will hesitate to extend the pouring time of afamiliar casting, his experience will be basedusually on a poor filling system. Such systemsgenerate problems such as slopping and surging,

and splashes ahead of the main body of melt.The cooling and oxidation of these splashesprior to the arrival of the main flow causes themto be imperfectly assimilated on arrival of themain body of liquid, resulting in the appearanceof a `cold' lap. Thus fill rate or temperature isincreased in an effort to avoid this apparentproblem, usually resulting in worsening theproblem. The provision of a good filling systemis not subject to such problems: the advancingliquid front keeps itself together, and so keepsitself warm. The result is that pouring tempera-ture can often be reduced and pouring timesextended without penalty.

In general, if there is a wide choice of time, forinstance somewhere between 5 and 25 seconds,it is strongly recommended to opt for the max-imum time, giving the minimum fill rate. Thisis because, compared to the faster fill rate, theselection of the slower rate reduces the crosssection area of all parts of the filling system, inthis case by a factor of 5. This is economically

Figure 2.61 Freezing times for plates indifferent alloys and moulds.



valuable, giving a great boost to yield. The fillingchannels shrink from appearing `chunky' toappearing like needles (with the confident buterroneous predictions by all experiencedonlookers that such systems will never fill). Inaddition, there is the benefit that the slimmerfilling system actually works better, improvingthe quality of the casting by giving less room forthe metal to jump and splash. Random scrapfrom pouring defects is thereby reduced. Theseare important benefits.

On the arrival of a completely new design ofcasting, the choice of the time to fill the mouldcan sometimes be impressively arbitrary, withperhaps no-one in the foundry having any clearidea on the time to use. Nevertheless, a valuethat seems reasonable can be tried, and canalways be modified on a subsequent trial.

The important fact to remember is that pro-vided the pouring basin is kept filled to at leastits designed level, the filling time is not allowedto vary by chance as in a hand-cut runningsystem, and is not under the control of thepourer, but remains accurately under the con-trol of the casting engineer.

2.3.7.5 Fill rate

Having selected a fill time, the average fill rate is,of course, simply the total poured weight divi-ded by the total time in a convenient unit suchas kg sÿ1. This requires to be converted to theaverage volume fill rate by dividing by thedensity of the liquid metal, giving a value in suchunits as m3 sÿ1.

Even this value cannot be used directly. Thisis because the filling system has to be sized totake the significantly higher rate of flow at thebeginning of the pour. The average fill rate is,of course, less than the initial fill rate becausethe high initial rate is not maintained. Themetal slows as the mould fills, the fill ratefinally falling to zero if the metal level inthe mould finally reaches the same level as thatin the pouring basin. To make allowance forthis effect, it is convenient to assume that theinitial fill rate is a factor of approximately 1.5times higher than the average fill rate. Thisfactor is actually precisely correct if the castingis a uniform plate with its top level with thepouring basin, as shown in Appendix 1. How-ever, in general, the factor is not particularlysensitive to geometry, as can be demonstratedby such exercises as checking the fill times ofextreme examples such as a cone filled via itstip compared to it inverted and filled viaits base.

The initial flow rate, Q, preferably in unitsm3 sÿ1, is the value to be used for defining the

size of all of the remaining features of the fillingsystem that we require to calculate.

Incidentally, for a given volume flow rate, themould will fill in the same time whether alumi-nium or iron is poured (Galileo would haveknown this). Thus the system described belowapplies to all metals and alloys, perhaps to thesurprise of many of us who have unwittinglyaccepted the dogma that each metal and alloyrequires its own special system.

Later, when the first mould is poured, thefilling should be timed with a stopwatch as acheck of the running system design. The actualtime should be within 10 per cent of the pre-dicted time. In fact the agreement is often closerthan this (Kotschi and Kleist 1979), to theamazement of doubters of casting science!

After the first casting is produced, it may beclear that it needs a casting rate either slower(allowing some solidification during pouring) orfaster (to avoid cold lap-type defects). Thesemodifications to the rate can be easily andquickly carried out by minor adjustment (usuallyonly millimetres of changes to dimensions arerequired) to the size of the filling channels. Again,it is useful to emphasize that such changes remainunder the control of the casting technologist (notthe pourer).

2.3.7.6 Sprue (down-runner) design

Now that an initial rate of pouring has beenchosen, how can we achieve it accurately, lim-iting the rate of delivery of metal to precisely thischosen value? Theoretically it can be achievedby tailoring a funnel in the mould of exactly theright size to fit around a freely falling streamof metal, carrying just the right quantity(Figure 2.13). We call this our down-runner, orsprue.

The theoretical dimensions of the sprue cantherefore be calculated as follows. If a stream ofliquid is allowed to fall freely from a startingvelocity of zero, then after falling a height h itwill have reached velocity v. The height h alwaysrefers to the height to the melt surface in thepouring basin. This zero datum is one of thegreat benefits of the offset basin compared tothe conical basin (the starting velocities cannever be known with any accuracy whenworking with a conical basin). Thus we have

v � �2gh�1=2

To obtain the sprue sizes it is necessary to realizethat the low velocity v1 at the top of the spruemust be associated with a large cross-sectionalarea A1. At the base of the sprue the highermetal speed v2 is associated with a smaller area



A2. If the falling stream is continuous it is clearthat conservation of matter dictates that

Q � v1A1 � v2A2 � v3A3 etc:

where subscript 3 can refer to any downstreamlocation for the local values of the area of thestream and its velocity (for instance the area andvelocity at the gates). Since the velocities arenow known from the height that the melt hasfallen (neglecting any losses at this stage), and Qhas been decided, each of the areas of the fillingsystem can now be calculated.

In nearly all previous treatises on runningsystems the important dimension of the sprue forcontrolling the precise rate of flow has beenassumed to be the area of the exit. This part of thesystem has been assumed to act as `the choke',regulating the rate of flow of metal throughoutthe whole running system. It is essential to revisethis thinking. If the sprue is correctly designed tojust touch the surface of the falling liquid at allpoints, the whole sprue is controlling. There isnothing special about the narrowest part at thesprue exit. We shall continue this concept so faras we can throughout the rest of the filling sys-tem. If we achieve the target of fitting thedimensions of the flow channels in the mould justto fit the natural shape of the flowing stream, itfollows that no one part is exerting control. Thewhole system is all just as large as it needs to be;the channels of the filling system just touch theflowing stream at all points.

Even so, after such features as bends andfilters and other complications, the energy lossesare not known precisely. Thus there is a sense inwhich the sprue (not just its exit, remember) isdoing a good job of controlling, but beyond thispoint the precision of control may be lost tosome extent after those features that introduceimponderables to the flow. (In the fullness oftime we hope to understand the features better.Even now, computers are starting to makeuseful inroads to this problem area.)

Thus, as long as the caster pours as fast aspossible, attempting to fill the pouring basin asquickly as possible, and keeping the basin fullduring the whole of the pour, then he or she willhave no influence on the rate of filling inside themould; the sprue (the whole sprue, remember)will control the rate at which metal fills themould.

For most accurate results it is best to calcu-late the sprue dimensions using the formulaegiven above, and using the alloy density toobtain the initial volume flow rate Q.

However, for many practical purposes wecan take a short cut. It is possible to construct auseful nomogram for Al assuming a liquid

density of 2500 kg mÿ3 and for the dense alloysbased on Fe and Cu assuming a liquid densityaround 7500 kg mÿ3 (Figure 2.62). Thus areas ofsprues at the top and bottom can be read off,and the sprue shape formed simply by joiningthese areas by a straight taper. Using the dia-gram it is simple to read off areas of the sprue atany other intermediate level if it is desired toprovide a more accurately formed sprue havinga curved taper. Recall that the heights aremeasured in every case from the level of metal inthe pouring basin, regarding this as the zerodatum.

The nomogram is easy to use. For instance ifwe wish to pour an aluminium alloy casting atan average rate of 1.0 kg sÿ1, corresponding, ofcourse, to an initial rate of 1.5 kg sÿ1, Figure 2.62is used as follows. The 1.5 kg sÿ1 rate with adepth in the basin (the top level down to thelevel of the sprue entrance) of 100 mm, and asprue length of 200 mm (total head height to thetop of the melt in the basin of 300 mm), then itsentrance and exit areas can be read from thefigure as approximately 440 mm2 and 250 mm2

respectively. Remember from section 2.3.2.3that it is advisable to increase the area of theentrance by approximately 20 per cent to com-pensate for errors, particularly the error intro-duced if the sprue shape is approximated to astraight taper. Thus the final sprue entranceshould be close to 500 mm2.

As a check on the nomogram read-outs for ouraluminium alloy casting, we can now calculate thedimensions numerically using the equations givenabove. At 1.0 kg sÿ1 average fill rate, corre-sponding to an initial rate 1.5 kg sÿ1, assuminga liquid density of 2500 kg mÿ3, we obtain aninitial volume flow rate Q� 1.5/2500� 0.6�10ÿ3 m3sÿ1. We can calculate that the fallsof 100 mm and 300 mm are seen to cause themelt to accelerate to a velocity of 1.41 and2.45 m sÿ1, giving areas of 424 and 245 mm2

respectively. These values are in reasonableagreement with those taken directly from thenomogram.

The cross-section of the filling system can, ofcourse, be round or square, or even some othershape, provided the area is correct (we areneglecting the small corrections required as aresult of increased drag as sections deviate fur-ther from a circle). However, in view of makingthe best junction to the runner, a slot sprue andslot runner are strongly recommended for mostpurposes. (Multiple sprues might be useful toconnect to a number of runners. Several suchsprues would be expected to work better thanone large sprue as a result of improved con-straint of the metal during its fall as shown inFigure 2.63.)



If we were to choose a slot-shape for thesprue, convenient sizes might be in the region of7� 70 mm2 entrance and 5� 50 mm2 exit. (Iftwo sprues were used as shown in Figure 2.63these areas would, of course, be halved.)

2.3.7.7 Runner

Taking a simple turn from the sprue exit into therunner, the runner will have dimensions 50 mmwide by 5 mm deep. The inside radius of thisturn should be at least approximately 1 or 2times the thickness of the channel, thus we shallchoose 10 mm.

However, the melt will be travelling close to2.45 m sÿ1. The problem remains, `How to getthe speed down from this value, nearly five timestoo high, to a mere 0.5 m sÿ1 at the gate withoutcausing damage to the flow en route?' This is thecentral problem for the design of a good fillingsystem. This central problem seems in the past

either to have been overlooked, or to havesolutions proposed that do not work. At thispoint we need to appreciate the possible solu-tions with some care.

We only have a limited number of strategiesfor speed reduction. At this stage of the devel-opment of the technology the options appear toinclude

(i) Filtration.(ii) A number of right-angle bends in succes-

sion (Jolly et al. 2000).(iii) A by-pass runner design acting in a surge

control mode, calculated to introduce themelt through the gate at the correct initialrate (the rate increases later of course whenthe surge container is full).

Considering our first option. If the filter reducesthe flow rate by a factor of 4 or 5 (computersimulation might assist to provide a better

Figure 2.62 Nomogram giving approximate sprue areas (mm2) for light and dense metals as a function of flow rate andhead height.


10

Tota

l met

al h

ead

h (m

)5

4

3

2

0.50.4

0.3

0.2

0.1

0.05

0.1 1

1 10

Flow rate (m3s–1 and kg s–1)

100

10 100Alkg s–1

kg s–1 Cu or Fe

1

10–4 10–3 10–2 10–1m3s–1

1010m

mm

m2

100

100

mm

mm

2

100

1000

mm

mm

2

10,0

010

,000

mm

mm

2

10m

m2

100

mm

2

1000

mm

2

10,0

00m

m2

*�� $ +�� .��


figure) then the runner exit from the filter wouldrequire to be increased in area by a factor offour or five. We shall choose a value of 4 to givea margin of safety, helping to ensure that therunner was properly filled and slightly pressur-ized. Its area would then be 4� 245 and so closeto 1000 mm2. The dimensions of the slot runnerafter the filter might then be 10� 100 mm. Thisis a rather large width that would be in danger ofcollapse because of mould expansion if a highmelting point material such as an iron or steelwere to be cast in a silica sand mould. Therunner would be expected to survive for an Alalloy.

However, it would perhaps be more con-venient if the runner were divided into tworunners of 10� 50 mm. Much depends on thelayout of the mould and the filling system. Tworunners might be more conveniently filled from

two separate sprues, possibly exiting from theopposite ends of a suitably modified basin, com-plete with a central pouring well and verticalsteps either side as illustrated in Figure 2.63.The problem in this case is the expense of twofilters.

The second option, using a succession of rightangle bends, is only recommended if a goodcomputer simulation package for flow in narrowfilling systems is available to test the integrity ofthe flow (most simulation packages do not pre-dict flows accurately, most cannot cope with thinsections, and most cannot cope with surfacetension). If a proven software package to simu-late flow is available, a reasonable solution canbe found largely by trial and error along the linesof the development described by Jolly et al.(2000). This approach is not described furtherhere.

The third option can be a good solution. Anapproximate procedure is as follows. The speedof flow into a by-pass is assumed to be constant(this is clearly an overestimate, but thereforeerrs on the side of safety) at Q� 0.6�10ÿ3 m3 sÿ1. If the by-pass volume is positionedabove the level of the runner, rising up toheight H (Figure 2.37), where H is perhaps atleast 20 or 30 mm or more above the height ofthe bottom of the casting, the gradual filling ofthe by-pass trap will cause the metal in the gateto experience a gradually higher filling pressure.At the point at which the overflow is filled, thepressure comes on to the gate from the fullheight of the metal in the pouring basin. At thisinstant the casting should be filled to somedepth at least 20 or 30 mm above the gate, sothat any jetting into the mould when the fullfilling rate comes into effect will be to someextent suppressed. (The precise depth to sup-press completely the formation of bifilmsremains to be researched.)

If we assume an approximate model (awfullyprimitive, but better than nothing) that the areaAo of the overflow is sufficiently large to ensurethat the head of liquid it contains rises at thecritical velocity Vcrit, then we can assume theliquid in the gate will follow its rise at a roughlysimilar rate. Thus we can define the area Ao ofthe overflow required for this to happen Q/Vcrit.

It remains to work out what height H isrequired for the overflow.

If negligible metal enters the gate comparedto that entering the overflow we have themajority of the flow rate Q entering the over-flow of volume AoH. Thus the time required tofill the overflow is t�AoH/Q. Furthermore, ifthe velocity Vcrit through the gate area A3

remains roughly constant (even though the areaAc of the base of the casting is starting to be

Figure 2.63 An arrangement for a single basin and twosprues leading to two runners.



filled to a depth h) the statement for volumeconservation is Vcrit�A3�Vc�Ac. Also, the aver-age velocity of rise in the casting is given byVc� h/t. Thus the time required to fill thecasting to a height h above the runner (neglect-ing the relatively trivial amount contained in thegate) is given by t� hAc/Vc�A3. Equating thesetwo estimates for times gives

AoH=Q � hAc=Vc � A3

Rearranging to give the height of the overflowwe obtain finally

H � �Q=Ao��h=H��Ac=A3��1=Vcrit�As mentioned earlier, it is sensible to arrange theoverflow to be a cylinder and connected tan-gentially to the runner. In this way, by avoidingunnecessary turbulence and filling more pro-gressively, a better quality of metal is preservedfor future recycling within the foundry.

The careful sizing of overflows to suppressthe early jetting of melt through the gates isstrongly recommended. To the author's knowl-edge, the technique has been relatively little usedso far. More experience with the technique willalmost certainly lead to greater sophistication inits use.

2.3.7.8 Gates

In general, it is essential that the liquid metalflows through the gates at a speed lower thanthe critical velocity so as to enter the mouldcavity smoothly. If the rate of entry is too high,causing the metal to fountain or splash in themould cavity the battle for quality has probablybeen lost.

The gate should enter at the very base of thecasting, if possible at right angles onto a thinsection, as has been described earlier. Gatingdirectly across a flat floor of a casting is to beavoided if possibleÐa thin jet of metal skatingacross a flat surface is a recipe for mouldexpansion defects of various sorts that will spoilthe surface of the casting. The casting will also beat risk from the formation of an oxide flow tubethat may constitute a serious internal dis-continuity in the casting. If directed at rightangles against a core, higher velocities can betolerated, since the thin section in the castingeffectively acts as an extension of the runnersystem, helping to spread and thus reduce thevelocity before the melt arrives in a section largeenough to allow the melt room to damage itself.Thus gating onto a core is often useful providing,of course, that we have succeeded to design thefilling system to remain free from entrained air.

For our example casting, the velocity at thebase of the sprue is 2.45 m sÿ1. Thus to achieve0.5 m sÿ1 through the gate(s), and if no friendlycore is conveniently sited, we shall require anexpansion of the area compared to that of thebase of the sprue by a factor A2/A3� 5approximately. In terms of the gating ratiomuch loved by the traditionalists among us, weare using 1 : 1 : 5 for this casting. The use of thissize of gate assumes that we are gaining noadvantage from a by-pass runner design. If agood by-pass design could be devised anacceptable ratio might then become 1 : 1 : 1effectively easing subsequent cut-off, and redu-cing any possible problems of hot-spots orconvection at this location.

Sometimes the by-pass cannot be provided.Even if available, it may be useful to use boththe by-pass and the enlarged gate until such atime that our understanding of filling systemsmakes it clear that such belt-and-braces solu-tions are not required.

Where the gates form a T-junction with thecasting, the maximum modulus of the gatesshould be half of that of the casting (if, as will benormal, no feeding is planned to be carried outvia the gates). Thus, in general, the thickness ofthe gates needs to be less than half the thicknessof the wall. This forces the shape of the gatesto be usually of slot form. If made especiallythin in alloys of good thermal conductivity, thegate can sometimes be usefully employed to actas a cooling fin soon after the filling of thecasting.

The other major consideration that must notbe overlooked is the problem of the transverse,or lateral, velocity of the melt in the mouldcavity as it spreads away from the ingate. Thiscan easily exceed the critical velocity despite thevelocity in the gate itself being correctly con-trolled. In this case a single gate may have to bedivided to give multiple gates as described insection 2.3.2.6.

The area of the gate required to reduce thegate velocity to below the critical velocity, andthe limitations of its thickness, sometimes dic-tates a length of slot significantly longer thanthe casting. In this situation there is little choicebut to revise the design of the filling system,selecting a correspondingly longer fill time sothat the gate can be shortened to fit the lengthavailable. Alternatively, a by-pass runner designmay be the solution. If a solution cannot befound, the conclusion has to be accepted thatthe casting as designed cannot be made so as toenjoy reliable properties and performance. Aserious discussion with the designer will prob-ably be required.



Rule 3

Avoid laminar entrainment ofthe surface film (the non-stopping,non-reversing condition)

3.1 Continuous expansion ofthe meniscus

If the liquid metal front continues to advance atall points on its surface, effectively, continuingto expand at all points on its surface, like aprogressively inflating balloon, then all will bewell. This is the ideal mode of advance of theliquid front.

In fact, we can go further with this interestingconcept of the requirement for continuous surfaceexpansion. There is a sense in that if surface is lost(i.e. if any part of the surface experiences con-traction) then some entrainment of the surfacenecessarily occurs. Thus this can be seen to be anall-bracing and powerful definition of the condi-tion for entrainment of defects, simply that sur-face must not be lost. Clearly, surface is effectivelylost by being enfolded (in the sense that the foldnow disappears inside the liquid, as has been thecentral issue described under Rule 2), or by simplyshrinking (leading to folding) as described below.Thus in a way this condition `Avoid loss of sur-face' can be seen to supersede the conditions ofcritical velocity or Weber number. It promises tobe a useful condition that could be recognized innumerical simulation, and thus be useful forcomputer prediction of entrainment.

In practice, however, an uphill advance of theliquid front, if it can be arranged in a mouldcavity, is usually a great help to keep the liquidfront as `alive' as possible, i.e. keeping themeniscus moving, and so expanding and creat-ing new oxide.

While the surface is being continuouslyexpanded when filling the mould the casting hasthe benefit that the older, thicker oxide is con-tinuously being displaced to the walls of themould where it becomes the skin of the casting.Thus very old and very thick oxide does notnormally have a chance to form and becomeentrained. In fact, one of the great benefits of agood filling system is to ensure that the olderoxides on the surface of the ladle or pouringbasin etc. do not enter the mould cavity. Wheninside the mould cavity, the continued expan-sion of the surface ensures that the surface oxideis brought into contact with the mould surface,and so becomes the skin of the casting (and notentrained inside the liquid where it would con-stitute a defect). For instance, in the tilt pouringof aluminium alloy castings, the filling of a newcasting can be checked by dropping a fragmentof paper on the metal surface as the tilt com-mences. This marker should stay in place,indicating that the old skin on the metal wasbeing retained by the runner, so that only cleanmetal underneath could flow into the mouldcavity. If the paper disappears into the runner,the runner is not doing its job. In a way, the useof the tea-pot pouring ladle and bottom-pourladles common in the steel casting industry arein response to their special problems, in whichthe high rates of reaction with the environmentat such high temperatures encourages the sur-face oxide to grow from microscopic to mac-roscopic thickness, to constitute the familiarslag layer.


Problems arise if the front becomes pinned bythe rate of advance of the metal front being tooslow, or if it stops or reverses. Loss of surfacearea by enfolding bifilm defects can then occurin two ways:

(i) If the liquid front stops, a thick surface filmhas chance to form. This may become sothick and strong that the front can no longerre-start its advance if pressure increases toencourage flow a little later. This thick filmmay be subsequently entrained as the gen-eral advance re-starts, so that metal over-flows and submerges it. As the new metalrolls over the old film a new fresh oxide islaid down over the old thick oxide, formingour familiar double film. This can constitutea large geometrical defect, sometimes in theform of a vertical tubular crack, and some-times a large horizontal crack extendingacross the whole casting, or as a horizontallap around its complete perimeter. Suchbifilm defects are characterized microscopi-cally by asymmetrical components; onebeing the thick underlying, stationary film,and the other the younger thinner filmprovided by the meniscus that rolls upagainst it (Figure 3.1).

(ii) If the liquid front reverses, the shiny,swelling front of the liquid experiences abrief moment as it is flattened, prior toreversing its curvature in the oppositedirection. The flat form has slightly lessarea, so that the small excess of surface maybe entrained by random folding. This canform tiny but insidious surface cracks.

Both of these actions occur in various waysduring casting. We shall consider them in detailin this section.

3.2 Arrest of vertical progress

A sudden increase in cross-sectional area of thecasting, such as the extensive horizontal areas atB and C in Figure 3.2 tend to bring the generaladvance of the liquid front practically to a stop.Such interruptions to the advance of the frontare likely to result in lap-type defects at b andboth at c1 and c2.

The film on the melt thickens while the frontis stopped. It can be submerged later if metalbreaks through at some point and flows over it.During the process of submerging the film, thenewly arriving metal rolls over the thicker oxide,so rolling in place its own, newer oxide film. Inthis way an asymmetric double oxide layer iscreated, with dry sides facing dry sides so that

the double layer forms an unbonded interface asa crack. This process can result in a major defect,often spanning the casting from wall to wall.

Another good name for this defect is an oxidelap. (It is to be distinguished from the solidifi-cation defect that often has a superficiallysimilar appearance, called here a cold lap.

Figure 3.1 An unstable advance of a film-forming alloy,showing the formation of horizontal laps as the interfaceintermittently stops and re-starts by bursting through andflooding over the surface film.

A

B

c1 c2

C

a

b

Figure 3.2 Steady filling via the bottom gate is interruptedbecause of overflow to the heavy section A, and fillingof extensive horizontal surfaces B and C, leading to thedanger of lap defects at distant, apparently unrelatedregions of the casting a, b, c1 and c2.

Rule 3. Avoid laminar entrainment of the surface film 103


The method of treating these defects is quitedifferent. A cold lap can be cured by increasingthe casting temperature, whereas increasing thetemperature is likely to make an oxide lapworse.)

3.3 Waterfall flow

Instead of a large horizontal defect, a tubular oreven cylindrical defect that I call an oxide flowtube can form in several ways. If the liquid fallsvertically, as a plunging jet, the falling stream issurrounded by a tube of oxide (Figure 3.3).Despite the high velocity of the falling metalinside, the oxide tube remains stationary,thickening with time, until finally surroundedby the rising level of the metal in the mouldcavity. This rising metal rolls up against theoxide tube, forming a double oxide crack.Notice the curious cylindrical form of this crackand its largely vertical orientation. The arrest ofthe advance of the front in this case occurred bythe curious phenomenon that although themetal was travelling at a high speed parallel tothe jet, its transverse velocity, i.e. its velocity atright angles to its surface, was zero. It is the zerovelocity component of a front that allows theopportunity for a thick oxide skin to develop.

These oxide flow tubes are often seen aroundthe falling streams of many liquid metals andalloys as they are poured. The defects are alsocommonly seen in castings. Although occa-sionally located deep inside the casting wherethey are not easily found, they are often clearlyvisible if formed against the casting surface,especially in Al alloys in permanent moulds.

Even with the best design of gravity-pouredsystem, the rate of fill of the mould may be farfrom optimum at certain stages during the fill.For instance, Figure 3.2 shows a number ofcommon geometrical features in castings thatcause the advance of the liquid metal to come to

a stop. The heavy section filled downhill at Awill cause the metal front to stop at point a,possibly causing a lap-type defect at a point onthe casting well away from the real cause of theproblem. In an uncharacteristic lapse of rigour,the author often refers to this problem as the`waterfall' effect. This always occurs if the liquidfalls into a recess. Until the recess is filled, theremainder of the liquid front cannot advance.There are several reasons for avoiding any`waterfall' action of the metal during the fillingof the mould.

(i) A cylindrical oxide flow tube forms aroundthe falling stream itself. If the fall is from areasonable height, the tube is shed fromtime to time, and plunges into the meltwhere it will certainly contribute to severerandom defects. The periodic shedding ofoxide flow tubes into the melt is a commonsight during the pouring of castings. Severalsquare metres of oxide area can be seen to beintroduced in this way within a minute or so.

(ii) The plunging jet is likely to exceed thecritical velocity. Thus the metal that hassuffered the fall is likely to be impaired bythe addition of randomly entrained bifilms.

(iii) As the melt rises around the tube, support-ing it to some degree and reducing theheight of the fall, the flow tube remains inplace and simply thickens. As the generallevel of the melt rises around the tube, thenew oxide rolls up against the surface of thecylinder, forming the curious cylindricalbifilm that acts as a major cylindrical crackaround a substantially vertical axis.

(iv) During the period of the waterfall action,the general rise of the metal in the rest of themould will be interrupted, causing an oxideto form across the whole of the stationarylevel surface. Thus a major horizontal lapdefect may be created.

Defects 1 and 2 above are the usual fragmentedand chaotic type of bifilm. Defects from 3 and 4are the major geometrical bifilms.

Waterfall problems are usually easily avoidedby the provision of a gate into the mould cavityat every low point in the cavity. Occasionally,deep recesses can be linked by channels throughthe mould or core assembly; the links beingremoved during the subsequent dressing of thecasting.

3.4 Horizontal stream flow

If the melt is allowed to spread without con-straint across a horizontal thin-walled plate,

Oxide flow tube defect from a fall

Figure 3.3 Waterfall effect leading to a vertical oxideflow tube (among other defects).



gravity can play no part in persuading the flowto propagate on a broad stable front, as wouldhappen naturally in a vertical or sloping plate(Figure 3.4). The front propagates unstably inthe form of a river bounded by river bankscomposed of thickening oxide. This mode offlow occurs because the oxide at the flow tip isthin, and easily broken, allowing the front toadvance here, but not elsewhere, where theoxide on the front is allowed to thicken, sorestraining any advance. This is a classicalinstability situation leading to a kind of den-dritic advance of a front. The meanderingadvance leads to a situation where its sinuousoxide flow tube is sealed into the casting as theliquid metal finally arrives to envelop it.

The avoidance of extensive horizontal sec-tions in moulds is therefore essential for repro-ducible and defect-free castings. Any horizontalsections should be avoided by the designer, orby the caster tilting the mould. The tilting of themould is more easily said than done with mostof our automatic moulding and casting lines, andrepresents a serious deficiency in much of ourstandard foundry equipment. This deficiencyneeds to be addressed in future equipment. Incontrast, a tilting facility is easily provided, and,in principle, can be programmed into the fillingprocess by some casting techniques such sometilt casting machines, and in the CosworthProcess, where the mould is held in a rotatablefixture during casting. The flow across suchinclined planes is therefore progressive, if slow,

but the continuous advance of the front at allpoints assists the aim of keeping the meniscus`alive'.

A fascinating example of a flow tube can bequoted from observations of uphill flow in anopen channel driven by a travelling magneticfield from a linear motor sited under the chan-nel. When used to drive liquid aluminium alloyuphill, out of a furnace and into a higher-levelreceiver, the travelling melt is seen to flow insideits oxide tube. When the magnetic field is swit-ched off, the melt drains out of the oxide tubeand back into the furnace, the tube collapsingflat on the bottom of the channel. However,when the field is switched on again, the sameoxide tube magically refills and continues topass metal as before. Clearly the tube has con-siderable strength and resilience. It is soberingto think that such features can be built into ourcastings, but remain unsuspected and almostcertainly undetectable. Clearly, the castingmethods engineer requires vigilance to ensurethat such defects cannot be formed.

The vertical oxide flow tube is probably morecommon than any of us suspect. The examplegiven below is simply one of many that could bedescribed.

Figure 3.5 illustrates the bronze bell hungoutside the railway station in Washington DC.Horizontal weld repairs record for all time thefatal hesitations in the pouring process that ledto the horizontal oxide bifilms that would haveappeared as horizontal laps. The vertical weldrepairs record the passage of the falling streamsthat created the vertical flow tubes, the oxidelaps that led to cracks through most of thethickness of the casting. This is a commonsource of failure for bells, nearly all of which aretop-poured through the crown. The renownedLiberty bell (the only survivor of three attempts,all of which cracked) reveals a magnificentexample of a flow tube defect that starts at thecrown, curves sinuously around and over theshoulder, and finally falls vertically. Althoughthere are many examples of bells that exhibitthese long cracks, it is perhaps all the moresurprising that any bells survive the top-pouringprocess. It seems likely that in the majority ofcases of bells of thicker section the oxide flowtube is not trapped between the walls of themould to create a through-thickness pair ofparallel cracks. In such thicker sections the tubeis more likely to be detached and carried away,crumpling into a somewhat smaller defect thatcan be accommodated elsewhere. It is to behoped that the new resting place of the defectwill not pose any serious future threat to theproduct. Clearly, the top pouring of castings is arisky manufacturing technique.

Oxide flow tube defectsfrom horizontal filling

Figure 3.4 The filling of a rectangular box type casting,illustrating the progressive advance of the front thatcharacterizes the filling of vertical walls. The horizontaltop, however, fills unpredictable meanders of river-likeflows, leading to horizontal oxide flow tubes.



Oxide flow tubes are common defects seen ina wide variety of castings that have been filledacross horizontal sections or down slopingdownhill sections. The deleterious flow tubestructures described above that form when fill-ing downwards or horizontally are usuallyeliminated when filling vertically upwards, i.e. ina counter-gravity mode. The requirement thatthe meniscus only travels uphill is sacred.However, even in this favourable mode of fill-ing, a related oxide lap defect, or even a cold lapdefect, can still occur if the advance of themeniscus is stopped at any time, as we have seen.

3.5 Hesitation and reversal

If the meniscus stops at any time, it is commonfor it to undergo a slight reversal. Minorreversals to the front occur for a variety ofreasons. Some of these are discussed below.

(i) A reversal will practically always occur whena waterfall is initiated. This occurs becauseat the point of overflow, the liquid will be ata level slightly above the overflow, dictatedby the curvature of its meniscus, i.e. for aliquid Al alloy it will be about 12.5 mmabove the height of the overflow since thisis the height of the sessile drop. However,immediately after the overflow starts, thegeneral liquid level drops, no longer sup-ported by the surface tension of the menis-cus. In the case of liquid aluminium alloythis fall in general level of the liquid will beperhaps about 6 mm, just enough to flattenthe more distant parts of the meniscusagainst the rest of the mould walls.

(ii) Hesitations to an advancing flow will oftenbe accompanied by slight reversals becauseof inertial effects of the flow. Momentumperturbations during filling will cause slightgravity waves, the surface therefore experi-encing minor slopping and surging motion,oscillating gently up and down.

These minute reversals of flow flatten the oxi-dized surface of the meniscus. When advancing,the meniscus adopts a rounded form, but whenflattened, the oxidized surface now occupies lessarea. A fold necessarily develops, wrinkling thesurface, endangering the melt with the possibil-ity of the entrainment of this excess oxide oncethe melt is able to continue its advance. Thefolding in of a small crack attached to the sur-face of the casting is illustrated in Figure 3.6.Such shallow surface cracks occurring as aresult of hesitation and/or reversal of the frontare common in aluminium alloys, and arerevealed by dye penetrant testing.

It is instructive to estimate the maximumdepth that such oxide folds might have. Fol-lowing Figure 3.6, if the front of the liquid inFigure 3.6a is a cylinder of radius r, the peri-meter of the quarter of a cylinder is pr/2, so thatthe maximum length of excess surface if the meltlevel now drops a distance r is pr/2ÿ r� r((p/2)ÿ1)� r/2 approximately. The radius of the menis-cus r is approximately 6 mm for liquid aluminium(as a result of the total height of a sessile dropbeing approximately 12 mm), giving the excesslength 3 mm. If this is folded just once to create abifilm, its potential depth is therefore found to beapproximately 1.5 mm.

If the melt continues its downward oscillationthe defect can be straightened out as shown in

Figure 3.5 Nick Green (tall,handsome and young) and the author(less tall, not so good looking, andsignificantly older) inspect the fine8000 kg American Legion FreedomBell outside the railway station inWashington DC, unfortunatelyspoiled by welds in an attempt torepair cracks caused by horizontaloxide laps and vertical oxide tubedefects (photograph courtesy ofauthor's wife).



Figure 3.6c. Alternatively, if the bifilm createdin this way holds itself closed, possibly becauseof viscous adhesion (i.e. the trapped liquid metaltakes time to escape from between the films) orpossibly as the result of other forces such as Vander Waals forces, then there is the danger thatadditional folds may be created on each oscil-lation cycle.

In fact, many of these defects are not as deepas the maximum estimate of 1.5 mm for severalreasons: (i) the melt surface may not drop the fulldistance r; (ii) the film may be folded more thanonce, creating a greater number of shallowerfolds; and (iii) the fold-like crack may hinge to lieflat against the surface of the casting. The action

of internal forces as a result of flow of the liquidmay be helpful in this respect. For these reasonssuch defects are usually only a fraction of a mil-limetre deep, so that they can often be removedby grit blasting. Only relatively rarely do theyreach the maximum possible depth approaching1.5 mm. Even so, for castings requiring totalintegrity, that may be designed for conditions ofservice involving high stress or fatigue, theseminor oscillations of the front are very realthreats that are best avoided.

The ultimate solution, as we have emphasizedhere, is that the melt should be designed to bekept on the move, advancing steadily forwardsat all times.

Figure 3.6 The creation of a bifilm crack by the reversal of the front, causing the meniscus to flatten and enfold in theexcess surface area. Surface cracks of the order of a millimetre depth can be formed in this way.


(a) (b)

(c) (d)

<1.7 mm


Rule 4

Avoid bubble damage

Entrainment defects are caused by the foldingaction of the (oxidized) liquid surface. Some-times only oxides are entrained, as doubled-overfilm defects, called bifilms. Sometimes thebifilms themselves contain small pockets ofaccidentally enfolded air, so that the bifilm isdecorated by arrays of trapped bubbles. Much,if not all, of the microporosity observed incastings either is or has originated from a bifilm.Sometimes, however, the folded-in packet of airis so large that its buoyancy confers on it a life ofits own. This oxide-wrapped bubble is a massiveentrainment defect that can become importantenough to power its way through the liquid andsometimes through the dendrites. In this way itdevelops it own distinctive damage pattern inthe casting.

The passage of a single bubble through anoxidizable melt is likely to result in the creationof a bubble trail as a double oxide crack, a longbifilm, in the liquid. Thus even though the bub-ble may be lost by bursting at the liquid surface,the trail remains as permanent damage in thecasting.

The bubble trail occurs because the bubble isnearly always attached to the point where itwas first entrained in the liquid. The enclosingshroud of oxide film covering the crown of thebubble attempts to hinder its motion. However,if the bubble is sufficiently buoyant, its buoy-ancy force will split this restraining cover.Immediately, of course, the oxide re-forms onthe crown, and splits and re-forms repeatedly.In this way the bubble progresses by its skinsliding around the bubble, gathering together ina mass of longitudinal pleats under the bub-ble as a trail that leads back to the point atwhich the bubble was first entrained as apacket of gas.

The structure of the trail is a kind of col-lapsed tube. In section it is star-like but with acentral portion that has resisted complete col-lapse because of the rigidity of the oxide film(Figure 4.1b). This is expected to form anexcellent leak path if it joins opposing surfacesof the casting, or if cut into by machining. Inaddition, of course, the coming together of theopposite skins of the bubble during the forma-tion of the trail ensure that the films makecontact dry side to dry side, and so constituteour familiar classical bifilm crack.

Poor designs of filling systems can result inthe entrainment of much air into the liquidstream during its travel through the filling basin,during its fall down the sprue, and during itsjourney along the runner. In this way dozens oreven hundreds of bubbles can be introducedinto the mould cavity. When so many bubblesare involved the later bubbles have problemsrising through the maze of bubble trails thatremain after the passage of the first bubbles.Thus the escape of late-arriving bubbles ishampered by the accumulation of the tangleof residual bifilms. If the density of films issufficiently great, fragments of bubbles remainentrapped as permanent features of the casting.This messy mixture of bifilm trails and bubblesis collectively christened `bubble damage'. Inthe experience of the author, bubble damageis probably the most common defect in castings,but up to now has been almost universallyunrecognized.

Bubble damage is nearly always mistaken forshrinkage porosity as a result of its irregularform, usually with characteristic cusp-likemorphology. When seen on polished sections,the cusp forms that characterize bubbledamage are often confused with cusps that are


associated with interdendritic shrinkage poros-ity. However, they can nearly always be dis-tinguished with complete certainty by theirdifference in size. Careful inspection of thedendrite arm spacing will usually reveal thatcusps that would have formed around dendritesas the residual interdendritic liquid is suckedinto the dendrite mesh are usually up to tentimes smaller than cusps that are caused by thefolds of oxide in bubble trails (Figure 4.1b).Clearly, the two are quite distinct and totallyunrelated.

Bubble damage is commonly observed justinside and above the first (or sometimes the last)ingate from the runner (Figure 4.1a). The largebubbles have sufficient buoyancy to escape upthe first ingate, but smaller bubbles can be car-ried the length of the runner, to appear throughthe farthest ingate. Alternatively, they can evenbe carried back once again if there is a backwave. This non-uniform distribution associatedsometimes with first and sometimes with someother ingate position is a common but not uni-versal feature of bubble damage. This is becausethe presence of cores, and sometimes strong

flows of metal inside the mould cavity can causethe bubble path to deviate a long way froma direct vertical path to the surface. Highlyindirect paths are commonly observed in videoradiography studies. Nevertheless, the commonfeature of bubble damage is its non-uniformdistribution.

If the bubbles completely escape the remainingtrails can float around, finally settling somedistance from their source. Irregular masses ofoxides in odd corners of castings have beenpositively identified as groups of tangled bubbletrails. The bubbles have moved on and escaped,but their trails have remained in suspension. Theyhave broken free from their moorings (thepoint at which the bubble was first entrained)and have travelled, tumbling and ravelling asthey go, carried by the sweeping and circulat-ing flow of the liquid during the filling process.Texan founders will recognize an analogy withtumbleweed.

Another common feature of bubble damageis the entrapment of small bubbles just under thecope skin of the casting (Figure 4.1c). They areprevented from escaping only by the thickness

Figure 4.1 (a) Pattern of bubble damage in a casting; (b) trails invisible in radiography are usually visible ontransverse sections; (c) small entrained air bubbles do not have sufficient buoyancy to break the double oxide barrierto escape to the atmosphere.

Rule 4. Avoid bubble damage 109

Transverse section ofbubble trail

Dendrites

Oxide skinof casting

Oxide skinof bubble

Air

Liquidmetal

Shrinkage porositygrown from the

bubble trail

(b)

(a)

(c)


of the oxide skin on the casting and their ownoxide skin. Both these films require to be bro-ken. (This is achieved by larger bubbles becauseof their stronger buoyancy forces, but not bysmaller bubbles. The dividing line between largeand small bubbles seems to be in the region of5 mm diameter for many light and dense alloys.)Such bubbles, sitting only a double thickness ofoxide depth under the top skin of the casting arecommonly broken into when shot blasting, oron the first machining cut. These too are com-monly observed in video radiographic studies.

Close optical examination of the interiors ofbubbles and bifilms in an aluminium alloycasting often reveals some shiny dendrite tipscharacteristic of shrinkage porosity. This addsto confusion of identification, because shrink-age cavities will often form, expanding anexisting bifilm, unfurling and opening it, andfinally sucking one or both of its films into thedendrite mesh. Subsequently only fragments ofthe originating oxides will sometimes be foundamong the dendrites. This process has beenobserved in video radiographic studies of cast-ings. An unfed casting has been seen to draw inair bubbles at a hot spot on its surface. Thebubbles floated up in succession, but the laterbubbles became trapped by dendrites. Assolidification progressed, shrinkage caused theair bubbles to gradually convert to shrinkagecavities. The perfectly round and sharp radio-graphic images were seen to become `furry' andindistinct as the liquid meniscus was sucked intothe surrounding mesh of dendrites. Finally, thedefect resembled an extensive shrinkage cavity;its origin as a gas bubble no longer discernible.

Other real-time radiography has shownbubbles entrained in the runner, and sweptthrough the gate and into the casting. Theupward progress of one bubble in the region of5 to 10 mm diameter appeared to be arrested,the bubble circulating in the centre of the cast-ing, behaving like a balloon on a string. Thestring, of course, being the bubble trail acting asa tether. Other bubbles of various sizes up toabout 5 mm diameter in the same casting wereobserved to float to the top of the casting,coming to rest under the oxide skin of the copesurface. These bubbles had clearly broken freefrom their tethers, probably as a result of theextreme turbulence during the early part ofthe filling process. The central bubble wasmarginally just too small to tear free from itstrail. In addition, it may have lost some buoy-ancy as a result of loss of oxygen during its rise,or perhaps more likely, it ascended as far as itdid because of assistance from the force of theflow of the melt. When this abated higher inthe mould cavity, its buoyancy alone was

insufficient to split its oxide skin, so that itsupward progress was halted.

Where many bubbles have passed throughan ingate into the mould, a cross-section of theingate will reveal some central porosity. Theseare the bubble trails, pushed ahead of thegrowing dendrites, and so concentrated in thecentre of the ingate section. Close examinationwill confirm that this porosity is not shrinkageporosity, but a mass of double oxide films, thebubble trails. In Al alloys they appear as a seriesof dark, non-reflective oxidized surfaces inter-leaved like the flaky, crumpled pages of an oldsepia-coloured newspaper.

In some stainless steels the phenomenon isseen under the microscope as a mixture ofbubbles and cracks. (A remarkable combina-tion! Without the concept of the bifilm such acombination would be extremely difficult toexplain.) In these strong materials the highcooling strain leads to high stresses that open upthe double oxide bubble trails.

In grey iron cylinder heads the bubbles andtheir trails are coated not with oxide but with alustrous carbon film. The carbon film appearsto be somewhat more rigid than most oxidefilms, and so resists to some extent the completecollapse of the trail, and retains a more opencentre. In effect, the bubbles punch holesthrough the cope surfaces of the casting, so thattheir trails form highly efficient leak paths.

The bubble trail is usually a collapsed, ornearly closed, tube. However, completely openbubble trails have been observed by Divandariin pressure die castings (Figure 4.2). In thisprocess the very high injection velocities, of theorder of 10 to 100 times the critical velocity forentrainment, naturally entrains considerablequantities of air and mould gases. These extra-ordinary conditions are perhaps better descri-bed in terms of atomization and emulsificationof the air and the metal. The very high pressure(up to 100 MPa, or 15 000 psi) applied duringcasting is mainly used to compress theseunwanted gases to persuade them to take up theminimum volume in the casting. If, however, thedie is opened before the casting is fully solidi-fied, as is usual to maximize productivity, theentrained bubbles may experience a reduction intheir surrounding mechanical support, allowingthe bubbles to expand under their immenseinternal pressure. At the same time, of course,their bubble trails will also be re-inflated. Suchopen bubble trails in pressure die-cast compo-nents are expected to be serious sources ofleakage, particularly when broken into bymachining operations. In this case the problemis greatly reduced (although perhaps never quiteeliminated) by sacrificing some productivity,



allowing the castings to solidify more com-pletely before opening the die.

4.1 Gravity-filled running systems

In gravity-filled running systems the require-ment to reduce bubbles in the liquid streamduring the filling of the casting calls for offsetstepped basins, or other advanced filling sys-tems. The conventional conical or funnel-shaped pouring basin cannot be permitted. Therequirement also demands properly engineeredand manufactured sprues. The sprue is requiredto be tapered, the taper calculated to match, orvery slightly compress, the natural form of thefalling stream; the stream naturally narrowsduring its fall because of its acceleration underthe action of gravity. By tailoring the shape ofthe sprue to the natural shape of the stream themelt has the best chance to avoid the entrain-ment of air. Parallel or reversed taper sprues arenot recommended. They may be permitted onlyif special precautions are adopted such as theprovision of a filter and bubble trap combina-tion in the entrance to the runner, as close aspossible to the sprue exit.

It is mandatory that the taper of the spruecontains no perturbations to upset the smoothfall of the liquid metal. Thus it must be well-fitting with the pouring basin, and accurately

matched in size and alignment at mould or diejoints; no steps, ledges, or abrupt changes indirection are permissible (a typical sprue mis-match across a mould joint is shown inFigure 4.1). Also no branching or joining of otherducts, runners, gates or sprues is allowable. Allsuch features (unfortunately especially commonin investment castings) have the potential for theintroduction of air into the stream, or theuncontrolled premature escape of droplets anddribbles of liquid into other parts of the mouldcavity The dividing of sprues might becomeallowable at some future date when such featureshave been properly researched by accuratecomputer simulation and video radiography.

At one time it was mandatory that each spruehad a sprue well at its base. The well wasthought to facilitate the turn of the metalthrough the right angle bend into the runnerwith minimum turbulence. All the work on welldevelopment had been based on water models,and the standard runner had been one of largearea, the expanded area specially selected with aview to slow the flow. However, more recentwork in the author's laboratory using bothwater models and liquid metals observed byvideo radiography has demonstrated that, atbest, the well is no better than no well at all forsuch large area runners, and at worst, causesconsiderable extra turbulence. The excellentwork by Isawa (1993) noted that even the bestdesigns of wells that he was able to optimizeintroduced hundreds of bubbles that took 2 to4 seconds to clear (Figure 4.3).

Thus the new designs of filling systemsincorporate no well at the base of the sprue. This

Figure 4.2 Re-inflated bubbles in a Zn alloy pressuredie casting (Divandari and Campbell 2003).

Figure 4.3 Water model of bubbles entrained by surfaceturbulence in a well (Isawa and Campbell 1994)showing the decrease of bubbles with time in differentwell designs, extrapolated to the time for the last bubble,tLB, for runners twice the area of the sprue exit.


1 mm

0 1 2 3 4 5

tLBtLB

tLBRange of bestwell designs

Time (s)

Num

ber

of b

ubbl

es p

er c

m3

Nowell

1000

100

10

1


departure from tradition is possible onlybecause the new filling systems are characterizedby runners of approximately the same area asthat of the sprue exit (Section 2.3.2.5). It is to benoted that the traditional choice of sprue exit/runner/gate ratios of 1 : 2 : 2 and 1 : 2 : 4 etc. areautomatically bad. The runner is too large to fillcompletely, regrettably ensuring bubble damageproblems.

An additional beneficial consequence of theavoidance of a well is the addition of friction tothe liquid provided by the additional solidsurface of the mould at the point impacted bythe metal as it turns the corner, so slowing thevelocity of the melt to the greatest extent. If thesprue/runner junction is nicely formed, bubblesare formed for precisely zero seconds. Thisawkward way of making a simple statement thatno bubbles are formed is deliberate. It empha-sizes the contrast with filling systems that havebeen accepted as conventional up to now.Nowadays it is not necessary to accept a designthat introduces any bubbles at all.

It is mandatory that no interruption to thepour occurs that leads to the lowering of themelt in the pouring basin below the minimumdesign level. If the sprue entrance is unpressur-ized in this way air will enter the running system.In the worst instance of this kind, if the basinlevel drops to the point that the sprue entrancebecomes uncovered this has to be viewed as adisaster. A provision must be made for thefoundry to reject automatically any castings thathave suffered an interrupted pour, or slow pour

that has allowed the basin to empty to a levelbelow the designed minimum level.

To be safe, it is worth ensuring that basins areprovided that are at least twice if not four timesthe required minimum depth to keep the spruefilled, and ensuring that the pourer keeps wellabove the minimum level. In this way the castingmay run a little faster, but air will be excludedand bubble damage avoided.

4.2 Pumped and low-pressure fillingsystems

Pumped systems such as the Cosworth Process,or low-pressure casting systems into sand mouldsor dies, are highly favoured as having thepotential to avoid the entrainment of bubbles if,and only if, the processes are carried out underproper control. The reader needs to be awarethat good control of a potentially good processshould not be assumed; it is required to bedemonstrated.

For instance, for pumped systems, bubblescan be released erratically from the interior wallof a tube launder system, especially if it is notcleaned with regular maintenance, as well asfrom the underside of a badly designed dis-tribution plate used in some counter-gravitysystems such as the early variants of the lowpressure sand (LPS) system (Figure 4.4c).

Although low-pressure filling systems can,in principle, satisfy the requirement for thecomplete avoidance of bubbles in the metal,

Figure 4.4 Bubbles introduced by defectivecounter-gravity systems. (a) Leak in a risertube of a low pressure die casting machine;(b) the dangerous ingestion of massivebubbles when the melt level is too low;(c) bubbles entrapped by dross and poordesign features of some pumped systems,that are released erratically and thusdamage castings in a non-reproducible way.


(a) (b)

Bubbles trapped atledges formed bydross

Bubbles lodgedunder flatdistribution plate

Improved designof distribution plate

(c)

pump


a leaking riser tube in a low-pressure castingmachine can lead to a serious violation(Figure 4.4a). The stream of bubbles from a leakin a defective riser tube will float directly up thetube and enter the casting. Unfortunately, thisproblem is not rare. Thus regular checks forsuch leakage, and the rejection of castings sub-jected to such consequent bubble damage, willbe required.

The other major problem with conventionallow-pressure delivery systems as used for lightalloy casting where the melt is contained withina pressurized vessel is that the topping up of thepressure vessel itself usually damages the qualityof the melt. The uncontrolled fall first from thefoundry transfer ladle, then down a chute, andfinally into the melt is an unsatisfactory transferprocess introducing much bubble damage intothe metal (Figure 2.49).

The Griffin Process counter-gravity processfor the production of steel wheels for rail rollingstock makes an interesting comparison. In thiscase, of course, the pressurized furnace containsliquid carbon steel. The large density differencebetween the steel and buoyant defects suchas bubbles, bubble trails and other entrainedoxides encourages such materials to float out

relatively quickly, so that the topping up of thefurnace does not necessarily introduce perma-nent damage; by the time the mould is casta good quality of steel has developed. In thisway a high-integrity safety-critical productcan be routinely produced. Even so, one canimagine that the deoxidation practice, leavingdifferent amounts of Si, Mn and Al, plus otherssuch as Ca, could influence the flotation timesignificantly.

For aluminium alloys, however, the near-neutral buoyancy of the introduced defectsmeans that very few have time to float out, and ofthe remainder, not all are subsequently removedby the filter, if any, at the entrance to the mouldcavity. Even the prior use of rotary degassingunits cannot be relied on to effect a completetreatment of the melt.

In fact, in low-pressure casting units (seeFigure 2.49) it is difficult to see how enclosedpressure vessels can be made to deliver liquidalloy of good quality. Much emphasis has beenplaced on the precise control over delivery ratesand volumes for such units. The quality of thedelivered melt, however, can only remain farfrom optimum. The use of such technologycannot be recommended at this time.



Rule 5

Avoid core blows

5.1 Background

When sand cores are surrounded by liquidmetal, the heating of the sand and its bindercauses large volumes of gas to be generated inthe core. Normally, the core will be designed sothat the gas can escape through the core printsand so be dissipated in the mould. In this waywe hope that the pressure inside the core isprevented from rising to high levels. In somecircumstances, however, the pressure of gas inthe cores may rise to such a level, higher thanthe pressure in the liquid, with the result that abubble is forced out into the melt. It is blowninto existence. Blow defect is therefore a goodname for this type of gas pore. Bubbles formedin this way are of large size, and so highlybuoyant. They rise through the metal leavingoxidised bubble trails in their wakes.

This is, of course, another form of bubbledamage as has been discussed under Rule 4.However, it is sufficiently distinct that it benefitsfrom separate consideration.

For instance bubble damage arising fromsurface turbulence in the filling system is gen-erated by the high velocities in the front end ofthe system (in the basin, sprue or runner). Thehigh shear stresses in the melt ensure that thebubbles are chopped mainly into small sizes,in the range 1 to 10 mm diameter. Some of thesmaller bubbles have been observed in videoradiographic studies to coalesce in the gate.These coalesced bubbles float quickly, beforeany significant solidification has taken place,and so burst at the liquid surface and escape.Bubbles smaller than about 5 mm diameter haveonly a tenth of the buoyancy of the 10 mmbubbles, and cannot split the oxides that bartheir escape (Figure 4.1c). If they succeed to

reach the top of the casting they thereforeremain trapped at a distance only a doubleoxide skin depth beneath the surface of thecasting.

Turning now to the quite different type ofbubble given off by the outgassing of a core,these bubbles are large. In irons and steels thesingle core blow bubble is about 13 mm diam-eter. In light alloys the effective bubble diameteris approximately 20 mm (Figure 6.22, Castings2003). Although these large bubbles have highbuoyancy, they are not produced immediately.The timing of their eruption into the meltdetermines the kind of defect that is formedin the casting. If, in relatively thick sections,the bubble detaches prior to any freezing, therepeated arrival of bubbles at the surface of thecasting can result in repeated build-up of bub-ble skins, forming a puff-pastry of the multipleleaves of oxide, known as an exfoliation defect(Figure 5.1b). More usually, the core takes timeto warm up, and takes further time to build upits internal pressure, thus allowing time for somefreezing to occur. Thus by the time the bubble isfinally forced into existence, it rises to sit underthe frozen layer of solid (Figure 5.1a).

Once a core has blown its first bubble, addi-tional bubbles are easily formed, since thebubble trail seems usually to remain intact, andkeeps re-inflating to pass an additional bubblealong its length. (The effect is interestinglysimilar to the re-inflating of the oxide tube withmetal as described in Section 3.4.) The bubblescontain a variety of gases, including watervapour, that are aggressively oxidizing tometals such as aluminium and higher meltingpoint metals. Bubble trails from core blowsare usually particularly noteworthy for theircharacteristically thick and leathery double


oxide skin, built up from the passage of manybubbles. This thick skin is part of the reasonwhy core blows result in such efficient leakdefects through the upper sections of castings.The reason is that they are, of course, auto-matically connected from a cored volume of thecasting, and often penetrate to the adjacent core(since little solidification will usually haveoccurred between cores to stop it). Alter-natively, in thicker sections, they travel to thevery top of the casting.

After the emergence of the first blow into themelt, the passage of additional bubbles con-tributes to the huge growth of some blowdefects. Often the whole of the top of a castingcan be hollow. The size of the defect cansometimes be measured in fractions of metres.

Blows can form from moulds. Whereasfounders are familiar with the problem of blowsfrom cores, blows from moulds are rarely con-sidered. In fact this is a relatively commonproblem (even though this section remainsentitled `core blows' as a result of commonusage). The huge volumes of gas that are gen-erated inside the mould have to be considered.

They need room to expand and flow. Any visitorto an iron or steel foundry will be impressedwith the jets of flame issuing from the joints ofmoulding boxes. Effectively the gases andvolatiles will be fighting to get out. It is prudenttherefore to provide them with escape routes,since escape via the liquid metal in the mouldcavity can spell disaster for the casting.

The build-up of back-pressure inside themould cavity, leading to incompletely filledmoulds, is most easily dealt with by the provi-sion of one or more whistlers. These are narrow,pencil-shaped vents through the cope.

The escape of gases entrapped in the mouldcavity is made more difficult by the applicationof mould coats, so that pressures can be doubled(Ohnaka 2003), making the provision of whis-tler vents more necessary.

The build-up of pressure can be even moresevere in moulds that are enclosed in steel boxes,and which are sat on a steel plate or on aconcrete floor. The gases are relatively free toescape from the cope, but gases attempting toescape from the drag are sealed in by the over-lying liquid in the mould cavity (Figure 5.3). Theproblem is enhanced if the casting is a tight fit inthe moulding box, as is usually the case ofcourse, since the casting engineer is always try-ing to get as much value as possible out of eachbox (Figure 5.2). In fact the build up of pressureinside sand moulds crammed into tight-fittingsteel boxes has, in the author's experience, con-tributed to a number of spectacularly defectivecastings, and in one instance, to a casting thatpersistently refused to fill its mould because theback-pressure of gases rose so high.

Figure 5.1 (a) A core blowÐa trapped bubblecontaining core gases evolved after some solidification;(b) an exfoliated dross defect produced by copious gasfrom a core blow prior to any solidification.

Figure 5.2 Casting in a close-fitting steel box on anunvented flat steel plate, showing blows from anupwardly oriented feature on the lower part of the mould.

Rule 5. Avoid core blows 115


The build-up of pressure in the drag has beenobserved by the author to lead to severe blowdefects in metre square flat plates of a bronzealloy, particularly towards the far end of theplate where the condensation of volatiles drivenahead of the melt adds to the amount alreadyavailable from the sand binder (Figure 5.3). Theprovision of woven nylon vent tubes throughthe drag was quite inadequate; the enormousquantities of gas simply overwhelmed thispainstaking but useless provision. The dragneeded to vent from the whole of its lower surfacearea by standing the mould clear of the ground,or standing it on a deeply ridged base plate.

It is worth commenting on the curious butcommon provision of whistler vents through thetop of the mould in an effort to eliminate `gas'porosity in the casting; when the founder sees acore blow he will often apply a mould vent.Regrettably, this action is totally misguided. Amoment's reflection reveals the self-evident factthat the porosity (i.e. the entrained air bubbles)is already in the metal, and the metal itselfwould have to rise up the vent to eliminate theporosity from the casting. The error in thinkingarises because of the confusion between gasentrained in the melt, and gas entrapped in themould cavity. When these are separated intotheir logically separate categories, confusiondisappears, and the correct remedial action canbe identified.

Although blows can be formed above flatplates as described above, it is to be noted thatthey form much more easily from upwardpointing features of cores or moulds (Figure 5.2).The effect is the upside-down equivalent of thedroplet of water detaching from the tip of a sta-lactite. Thus the removal of upwardly pointingfeatures, or the inverting of the whole casting, isoften a useful tactic.

Considerable volumes of water vapour aregiven off from clay-based core repair and mouldrepair pastes. This is because the clay containswater of crystallization, so that even afterthoroughly drying the core repair at 100 or200�C, the water bound in the structure of theclay remains unchanged, only being released ata high temperature, in the region of perhaps600�C. Thus the water is released only when theclay contacts the liquid metal. This is particu-larly unfortunate, because the clay is composedof such fine particles that it is substantiallyimpermeable, preventing the escape of the waterinto the core or mould, so that the water isforced to boil off through the metal. Repair ofcores with clay-based pastes therefore generallyleads automatically to blow defects. The wideuse of core repair pastes illustrates that thisdanger is little known. The use of such materialsis to be avoided unless followed by baking at atemperature that can be demonstrated to avoidthe generation of blows in the melt.

Figure 5.3 A large flat platecasting with an enclosed drag.Volatiles are driven ahead andcondense in the cooler distantmould, exacerbating defects atthe far end.



The generation of blows off chills is the resultof an almost identical process. When a blockmetal chill is placed in a bonded aggregatemould, the pouring of the metal causes a rapidoutgassing of the volatiles in the aggregate/binder mixture. The volatiles, particularly watervapour, are driven ahead of the spreading liquidmetal, and condense on any cold surface, suchas a metal chill. When the liquid metal finallyarrives and overruns the chill the condensatesboil off. Since the chill is impermeable, thevapour is forced to bubble through the melt.

To demonstrate that a chill, a core, orassembly of cores, does not produce blows mayrequire a procedure such as the removal of all orpart of the cope or overlying cores, and taking avideo recording of the filling of the mould. Ifthere are any such problems, the eruption ofcore gases will be clearly observable, and will beseen to result in a boiling action, creating a frothof surface dross that would of course normallybe entrapped inside the upper walls of the cast-ing. A series of video recordings might be foundto be necessary, showing the steady develop-ment of solutions to a core-blowing problem,and recording how individual remedies resultedin progressive elimination of the problem. Thevideo recording requires to be retained by thefoundry for inspection by the customer forthe life of the component. Any change to thefillingrateofthecasting,orcoredesign,orthecorerepair procedure, would necessitate a repeat ofthis exercise.

For castings with a vertical joint where acope cannot be conveniently lifted clear toprovide such a view, a special sand mould maybe required to carry out the demonstration thatthe core assembly does not cause blows from thecores at any point. This will have to be con-structed as part of the tooling to commission thecasting. This will have to be seen as an invest-ment in quality assurance.

5.2 Prevention

By far the best solution to the evolution of gasesfrom cores is the use of a sand binder for thecore that has little or no evolution of gas asthe core becomes hot. This would represent aperfect solution. The best hopes here are theinorganic binders that contain no water ofcrystallization. However, the few binders thathave so far been developed to meet this criterionare usually not satisfactory in other ways. Theperfect core binder has yet to be developed!

In the meantime, one of the best actions toavoid blows from cores (or more occasionallyfrom moulds) is to increase the permeability of

the core by the use of a coarser aggregate orby the use of venting. Since the core print isusually the area where all the escaping gas has toconcentrate, a simple hole through the length ofthe print makes a huge impact on the problem,as has been shown previously by the author(Castings 2003). For some aluminium alloycastings this can be a complete solution. How-ever, of course, if the vent hole can be continuedto the centre of the core this is even better.The further provision of easy escape for gasesthrough the mould and out to the atmosphere isnecessary for copper-based and iron and steelcastings where the outgassing problem becomessevere because of the higher temperatures.Many readers will have seen the impressive jetsof burning carbon monoxide issuing from ventsin moulds of iron and steel castings for manyminutes after pouring. Figure 5.4 illustrates asuccession of improved venting techniques.

For low-volume production involving themaking of cores by hand, a vent can be providedalong a curved path through a core by laying a

Figure 5.4 Venting of a core, illustrating progressivelyimproved techniques.


Core

Vacuum

Mould

Vent to core centre

Vent to atmosphere

Vent to vacuum

Vent throughprint


waxed rope inside while it is being made. Thecore is subsequently heated to melt the wax sothat the rope can be pulled out. The concernsthat the wax itself, now percolated into the core,would add to the volatiles and so counter anybenefit is, thankfully, unfounded. The provisionof the vent is an overwhelming benefit.

For more delicate low-volume work, theauthor has witnessed long curved cores foraerospace castings being drilled by hand, usinga drill bit fashioned from a length of piano wireheld in a three-jaw rotary chuck driven by asmall motor. The tip of the high-carbon-steelwire is hammered flat, and ground to a sharppoint shaped like an arrow head. The core isdrilled by hand in a series of straight lengths, thepiano wire drill buzzing quickly through thecore. Each hole is targeted to intersect the pre-vious hole, the straight holes emerging on thebends, where the openings are subsequentlyplugged by a minute wipe of refractory cement(Figure 5.5). The complete vent is checked toensure that it is continuous, and free from leaks,by blowing smoke in to one of the vent open-ings, and watching for the smoke to emergefrom the far opening. Only when the smokeemerges freely at the far end, and from no otherlocation, is the core accepted for use. It is thenstored in readiness for mould assembly.

Occasionally, instead of an opening to theatmosphere, it is necessary to link the outsideopening to a vacuum line. This is relativelycommon practice in gravity die (permanentmould) casting, increasing the efficiency of the

extraction of gases from a resin-bonded sandcore. However, the evolution of volatiles fromthe binder creates problems by condensing assticky resins and tars in the vacuum line, so that,for long production runs, regular attention isrequired to avoid blockage, often dictating thetiming of the withdrawal of the tooling formaintenance.

The reader is advised caution with regard tothe application of a vacuum line to aid venting.The author once tried this on an extensive thin-section core with a single small area printaround which was poured liquid stainless steelat 1600�C. The resulting rapid build-up ofpressure was so dramatic that it blew off thevacuum connection with a bang! However, thediscerning reader will notice the extreme cir-cumstances described here, and rightly concludethat in this case the author was testing thepatience of Providence.

The prevention of blows from condensationon chills is widely known and generally wellapplied. The chill should normally be coatedwith a ceramic wash or spray that is afterwardsthoroughly dried to give an inert, permeable andnon-wetted surface layer. The effective perme-ability of the surface can be further enhanced byproviding deep V-grooves in a criss-cross pat-tern. The grooves are bridged to some degree bythe action of the surface tension of the melt, sothat the bottoms of the grooves act as surfacevents, tunnelling the expanding vapours tofreedom ahead of the advancing melt. Addi-tionally, the V-grooves are thought to enhancethe effectiveness of the chilling action byincreasing the contact between the casting andthe chill.

If there is an option, it is far better to arrangethat the core vents through prints that aredirected vertically upwards (Figure 5.6). This isbecause as the melt rises in the mould, thevolatiles migrate through the core ahead of themetal, concentrating in the last part of the core.If the core is vented at its base this is a potentialdisaster. The volatiles are too far from the print,and will continue to be pushed ahead, finallybeing pushed into the form of an eruption ofbubbles from the top of the core. This problemcan be reduced by covering the core with liquidmetal as quickly as possible. Venting from thebase is then given its best chance.

Even so, a print allowing outgassing from thetop of the core is ideal. If a vent cannot beprovided up the centre of the top print, a topprint is still valuable, even though it may con-tain no central vent, because the volatiles willtravel up the core surface. This can be seen oncore prints that emerge from the tops ofaluminium alloy castings. The melt is seen toFigure 5.5 Drilled holes to vent a narrow circular core.



flutter, trembling against the side of the coreprint as gas rushes up in the form of mini waves,causing ripples to radiate out across the surfaceof the melt.

The provision of soft ceramic paper gaskets,preferably with a central hole, shaped like awasher, placed on the end of core prints is anexcellent provision for the escape of gases. Thissimple remedy prevents the melt from flashingover the end of the print to block the vent

(Figure 5.6). The compressible washer allowsfor the sealing of the core print against ingressby liquid metal, but allows closure of the mouldwithout danger of the crushing of the core.

Finally, if the core can be covered quicklywith liquid metal, and the pressure in the metalquickly raised to be at all times greater than theinternal pressure generated inside the core, thenbubble formation will be suppressed. Thussimply filling the mould faster is often a quickand complete solution. The provision of anadditional top feeder to increase hydrostaticpressure needs care, since if the feeder has largevolume the delay in the rise of pressure to fill itmay be counter-productive. If feeding of thecasting is not really required, the sprue andpouring basin can provide the early pressurisa-tion that is needed, it would be better to leavewell alone and not be tempted to provide a topfeeder.

For those interested in quantifying some ofthe problems of core outgassing and the effectof sizes of vents and temperatures, etc., theprevious volume `Castings (2003)' derives anapproximate analytical formula to describe thephysics of core blows. Eventually, it is hoped,we can look forward to the day when computersimulation will provide an accurate descriptionof each core and mould, allowing in detail forthe effect of intricate geometries and the com-plicated effect of rate of filling that are some-times encountered. A welcome start has beenmade by Maeda and colleagues (2002) whodemonstrate a computer simulation of the flowof gas through an aggregate core. Perhaps wecan now look forward to such studies becominga commonplace feature of the design of a newcasting.

Figure 5.6 Vertical upwards venting, preferably with asoft print, is ideal. However, the addition of a centralvent hole through the core print, or even down into thecentre of the core, would be even better.


(a)

(b)

(c)

Bad

Good

Best

Compressibleceramicpaperwasher


Rule 6

Avoid shrinkage damage

6.1 Feeding systems design background

Before getting launched into this section, weneed to define some terms.

There is widespread confusion in parts of thecasting industry, particularly in investment cast-ing, between the concepts of filling and feeding ofcastings. It is essential to separate these twoconcepts.

Filling is self-evidently the short periodduring the pour, and refers to the filling ofthe filling channels themselves and the fillingof the mould cavity. This may only lastseconds or minutes.Feeding is the long, slow process that isrequired during the contraction of the liquidthat takes place on freezing. This processtakes minutes or hours depending on the sizeof the casting. It is made necessary as a resultof the solid occupying less volume than theliquid, so the difference has to be providedfrom somewhere. This contraction on solid-ification is a necessary consequence of theliquid being a structure resembling a randomclose-packed array of atoms, compared tothe solid, which has denser regular closepacking in a structure known as a crystallattice.

Figure 6.1 illustrates the three separate shrink-age problems that occur during the cooling of ametal.

1. The liquid grows in density as it cools.However, this simple thermal contraction inthe liquid state is not usually a significantproblem because most of the superheat (the

temperature above the liquidus) of a melt isusually lost during or quickly after pouring.

2. The main problem is the contraction onsolidification. This is around 3% for manysteels, but over 6% for Al alloys (Table 6.1).This is the contraction requiring to be fed bya feeder. Its principal action is simply that ofa reservoir (there are other important func-tions of feeders that we shall consider later).

3. The subsequent contraction in the solid stateremains a problem for the patternmaker. We

Figure 6.1 Schematic illustration of three shrinkageregimes: in the liquid; during freezing; and in the solid.


shall not concern ourselves with the pattern-maker's problems in this section.

We shall concentrate our attention on the mainproblem, the contraction on freezing as listed in2 above. To allow ourselves the luxury of somerepeated emphasis and further definitions : toprovide for the fact that extra metal needs to befed to the solidifying casting to compensate forthe contraction on freezing, it is normal toprovide a separate reservoir of metal. We shallcall this reservoir a feeder, since its action is tofeed the casting, i.e. to compensate for the soli-dification shrinkage (obvious really!). In muchcasting literature the reservoir is known, non-obviously, as a riser, and worse still, maybe confused with other channels that commu-nicate with the top of the mould, such asvents, or whistlers, since metal rises up theseopenings too. The author reserves the nameriser for the special kind of feeder described inSection 2.3.2.7, that is connected to the side ofthe casting via a slot gate, and in which metalrises up at the same time as it rises in the mouldcavity.

It is most important to be clear that the filling(sometimes called the running) system is notnormally required to provide any significant

feeding. The filling system and the feedingsystem have two quite distinct roles: one fillsthe casting, and the other feeds the shrinkageduring solidification. (On occasions it is possibleand valuable to carry out some feeding via thefilling system, but this requires the special pre-cautions that are described later.)

The main question relating to the provisionof a feeder on a casting is `Should we have afeeder at all?' This constitutes Rule 1 for feed-ing. This is a question well worth asking, and weshall return to it later. Just for the moment weshall assume that the answer is `Yes'. The nextquestion is `How large should it be?'

There is of course an optimum size. Figure 6.2aillustrates a section of a feeder on a plate castingin which the required shrinkage volume is justnicely concentrated in the feeder. This is thesuccess we all hope for. However, success isnot always easily achieved, and Figure 6.2 b, cand d show the complication posed by the dif-ferent shrinkage behaviour of different alloys.The pure Al and the Al±12Si alloy are bothshort freezing range, and contrast with theAl±5Mg alloy which is a long freezing rangematerial.

Some additional points of complexity in theoperation of feeders in real life need to beemphasized.

(i) The Mg-containing alloy in Figure 6.2d willalmost certainly contain some fine, scat-tered microporosity that will have acted toreduce the apparent shrinkage cavity.

(ii) The complicated form of the pipe in Al±12Si alloy almost certainly reflects thepresence of large oxide films that wereintroduced by the pouring of the castings.These large planar defects fragment boththe heat flow and the mass flow in thefeeder, and the short freezing range andsurface tension conspire to round off thecavities in the separated volumes of liquid.In addition, the oxide, together with thesolidifying crust on the top surface ofthe feeder also has some strength andrigidity, again complicating the collapse ofthe feeder top, and influencing the shape ofthe shrinkage pipe as it, and its associatedoxide skin, gradually expands downwards.These effects are additional reasons for the20 per cent safety factor often used forthe calculation of feeder sizes. Feeders oftendo not have the simple carrot-shapedshrinkage pipe predicted by the computer.Figure 6.2e gives a further excellent exam-ple of the action of flow from a feederdiverted and fractured by the presence oflarge bifilms.

Table 6.1 Solidification shrinkage for some metals

Metal Crystalstructure

Meltingpoint�C

Liquiddensity(kg/m3)

Soliddensity(kg/m3)

Volumechange(%)

Ref.

Al fcc 660 2368 2550 7.14 1Au fcc 1063 17 380 18 280 5.47 1Co fcc 1495 7750 8180 5.26 1Cu fcc 1083 7938 8382 5.30 1Ni fcc 1453 7790 8210 5.11 1Pb fcc 327 10 665 11 020 3.22 1Fe bcc 1536 7035 7265 3.16 1Li bcc 181 528 ± 2.74 4,5Na bcc 97 927 ± 2.6 4,5K bcc 64 827 ± 2.54 4,5Rb bcc 39 1437 ± 2.3 4,5Cs bcc 29 1854 ± 2.6 4,5Tl bcc 303 11 200 ± 2.2 2Cd hcp 321 7998 ± 4.00 2Mg hcp 651 1590 1655 4.10 3Zn hcp 420 6577 ± 4.08 2Ce hcp 787 6668 6646 ÿ 0.33 1In fct 156 7017 ± 1.98 2Sn tetrag 232 6986 7166 2.51 1Bi rhomb 271 10 034 9701 ÿ 3.32 1Sb rhomb 631 6493 6535 0.64 1Si diam 1410 2525 ± ÿ 2.9 2

References: 1, Wray (1976); 2, Lucas (quoted byWray, 1976); 3, This book; 4, Iida and Guthrie (1988);5, Brandes (1983).

Rule 6. Avoid shrinkage damage 121


Figure 6.3 shows the results of Rao et al. (1975),who investigated the feeding of a simple platecasting in Al±12Si alloy by planting on succes-sively larger feeders. Interestingly, when the dataare extrapolated backwards to zero feeder sizethe porosity is indicated to be approximately 8per cent, which is close to the theoretical 7.14 percent solidification shrinkage for pure aluminium(Table 6.1) and may indicate that 1 per cent or so

of thermal contraction due to superheat mayhave contributed to the total shrinkage contrac-tion. At a feeder modulus of around 1.2 times themodulus of the casting, the casting is at its mostsound. The residual 1 per cent porosity is prob-ably dispersed gas porosity (i.e. gas precipitatedinto dispersed microscopic bifilms so as to openthem). As the feeder size is increased further thesolidification of the casting is now progressivelydelayed by the nearby mass of metal in the fee-der. Thus while this excessive feeder is no dis-advantage in itself, the delay to solidification ofthe whole casting increases the time available forfurther precipitation of hydrogen as gas porosity.However, it is clear from this work that anundersized feeder will result in very serious por-osity. In contrast, an oversize feeder causes lessof a problem, increasing porosity slightly bythe opening of bifilms and thereby reducingmechanical properties. In addition, of course,the oversize feeder does adversely influence thefreezing time (important for cycle time in per-manent moulds) and reduces the metallic yieldthus adversely influencing the economics!

Figure 6.4 generalizes the finding of Raoand colleagues, to show the expected relationbetween gas and shrinkage porosity. Clearly, asthe feeder size is increased, the optimum feeder

Figure 6.2 Cross-section of (a) simple plate casting,nicely fed, with all of its shrinkage porosity concentratedin the feeder; (b) 99.5Al; (c) Al±12Si; (d) Al±5Mg;(e) radiograph of Al±12Si alloy feeder (courtesyFoseco 1988).

300

200

100

0

mm

(b)

(e)

(a)

(c) (d)

.�� ) & "�� #5,%� � �'?��

�� '�� ; �� ,( *��


size is hardly changed by the amount of gas insolution. However, as the gas content increases,the minimum level of porosity that can beachieved steadily rises, although never usuallyexceeds 1 or 2 per cent, compared to the 7 or8 per cent contribution of shrinkage. Clearly, itis more important to deal with the shrinkageproblems than with gas problems in castings.(This conclusion might raise the eyebrows ofpractised foundry people. It needs to be keptin mind that most of what previously hasbeen generally described as `gas' in castings, hasactually been entrained air bubbles as a result ofour poor filling systems.)

Where computer modelling is not carried out,the following of the seven feeding rules by theauthor is strongly recommended. Even whencomputer simulation is available, the seven ruleswill be found to be good guidelines. For thecomputer itself, following Tiryakioglu's reducedrules constitutes the most powerful logic and isrecommended, although the same rules alsoconstitute a useful check for those who deter-mine feeder sizes by pen and paper.

In addition to observing all the requirementsof the Rules for feeding, the use of all fivemechanisms for feeding (as opposed to onlyliquid feeding) should also be used to advan-tage. This will be found to be especially usefulwhen attempting to achieve soundness in anisolated boss or heavy section where the provi-sion of feed metal by conventional techniquesmay be impossible. However, a reminder of the

attendant dangers of the use of solid feeding arepresented later.

6.1.1 Gravity feeding

As opposed to filling uphill (which is of coursequite correct) feeding should only be carried outdownhill (using the assistance of gravity).

Attempts to feed uphill, although possible inprinciple, can be unreliable in practice, and maylead to randomly occurring defects that have allthe appearance of shrinkage porosity. In castingsof modest size feeding uphill appears to be suc-cessful as will be discussed in Section 6.4 `Activefeeding'. In many castings, particularly largercastings, problems occur when attempting tofeed uphill because of the difficulties caused bytwo main effects: (i) adverse pressure gradient asdiscussed below; and (ii) adverse density gradientleading to convection as dealt with in Chapter 7.

The atmosphere is capable of holding upseveral metres of head of metal. For liquidmercury the height is approximately 760 mm,being the height of the old-fashioned atmos-pheric barometer of course. Equivalent heightsfor other liquid metals are easily estimatedallowing for the density difference. Thus forliquid aluminium of specific gravity 2.4 com-pared to liquid mercury of 13.9, the atmospherewill hold up about (13.9/2.4)� 0.76� 4 m ofliquid aluminium.

While no pore exists, the tensile strength ofthe liquid will in fact allow the metal, in prin-ciple, to feed to heights of kilometres, since inthe absence of defects the liquid can withstandtensile stresses of thousands of atmospheres.The liquid can, in principle, hang up in a tube,its great weight stretching its length somewhat.However, the random initiation of a singleminute pore will instantly cause the liquid to`fracture', causing such feeding to stop and gointo reverse. The liquid in the tube will fall,finally stabilizing at the level at which atmos-pheric pressure can support the liquid. Thus anyheight above that supportable by one atmo-sphere is clearly at high risk.

Moreover, there is even worse risk of attempt-ing to feed against gravity. If there is a leak pathto atmosphere, allowing atmospheric pressureto be applied in the liquid metal inside thesolidifying casting, the melt will then fall fur-ther, the action of gravity tending to equalizelevels in the mould and feeder. Thus, if the fee-der is sited below the casting, the casting willcompletely empty of residual liquid. Regrett-ably, this is an efficient way to cast porouscastings and sound feeders.

Clearly, the initiation of a leak path toatmosphere (via a double oxide film, or via a

Figure 6.4 Generalized relation between gas andshrinkage as feeder size is increased in terms of themodulus ratio.


Total shrinkage + gas porosity

Por

osity

(%

)

Zero gasporosity

High gasporosity

‘Pure’ shrinkagecontribution

‘Pure’ gascontribution Intermed

8

7

6

5

4

3

2

1

00 0.5 1.0 1.5

mf /mC


liquid region in contact with the surface at a hotspot) is rather easy in many castings, making thewhole principle of uphill feeding so risky that itshould not be attempted in circumstances whereporosity cannot be tolerated. It is a pity that thecomforting theories of pushing liquid uphill byatmospheric pressure or even hanging it fromvast heights using the huge tensile strength ofthe liquid cannot be relied upon in practice.

For most purposes, the only really reliableway to feed is downhill, using gravity.

6.1.2 Computer modelling of feeding

Good computer models have demonstratedtheir usefulness in being able to predict shrink-age porosity with accuracy. A simulation usinga reliable modelling software package shouldnow be specified as a prior requirement to becarried out before work is started on making thetooling for a new casting. This minor delay willhave considerable benefits in shortening theoverall development time of a new casting, andwill greatly increase the chance of being `rightfirst time'.

However, at this time many computer simu-lations are inadequate for other reasons thatrequire to be recognized. For instance theseinclude:

(i) no allowance for the effect of thermalconduction in the cast metal (rare);

(ii) no allowance for the important effects dueto convection in the liquid (common);

(iii) neglect of, or only crude allowance for, theeffect of the heating of the mould by theflow of metal during filling (rare);

(iv) no capability of any design input. Thusgating and feeding designs will be requiredas inputs (universal at this time).

For the future, it is to be expected that softwarepackages will evolve to provide intelligentsolutions to all these requirements. Examples ofa good start in this direction are shown byDantzig and co-workers (Morthland et al. 1995and McDavid and Dantzig 1998). In themeantime, it remains necessary to use computermodels with some discretion. For instance in thework by the Morthland team they warn that theresults are specific to the feeding criterion used.If a more stringent temperature gradient cri-terion were used (for instance 2 K cmÿ1 insteadof 1 K cmÿ1) the feeder would have been larger.

The above approaches to the optimization ofthe feeding requirements of castings haveinvolved the use of numerical techniques such asfinite element and finite difference methods.Ransing et al. (2003) propose a geometrical

method based on an elegant extension of theHeuvers circle technique. This technique isdescribed later in the section describing the feedpath requirements for feeding (Section 6.2Feeding Rule 5).

6.1.3 Random perturbations to feeding patterns

In aluminium castings, flash of approximately1 mm thickness and only 10 mm wide has beendemonstrated to have a powerful effect on thecooling of local thin sections up to 10 mm thick,speeding up local solidification rates by up toten times (see Section 6.5.3). The effect is muchless in ferrous castings because of their muchlower thermal conductivity.

Thus for high conductivity alloys flash has tobe controlled, or used deliberately, since, inmoderately thick sections, it has the potential tocut off feeding to more distant sections. Theerratic appearance of flash in a production runmay therefore introduce uncertainty in thereproducibility of feeding, and the consequentvariability of the soundness of the casting. Flashon thick sections is usually less serious becauseconvection in the liquid in thick sections con-veys the local cooled metal away, effectivelyspreading the cooling effect over other parts ofthe casting, giving an averaging effect over largeareas of the casting. In general however, it isdesirable that these uncertainties are reduced bygood control over mould and core dimensions.

The other known major variable affectingcasting soundness in sand and investment cast-ings is the ability of the mould to resist defor-mation. This effect is well established in the caseof cast irons, where high mould rigidity is acondition for soundness. However, there is evi-dence that such a problem exists in castings ofcopper-based alloys and steels. A standard sys-tem such as statistical process control (SPC), orother techniques, should be seen to be in placeto monitor and facilitate control of such chan-ges. Permanent moulds such as metal or gra-phite dies are relatively free from such problems.Similarly many other aggregate mouldingmaterials are available that possess much lowerthermal expansion rates, and so produce cast-ings of greater accuracy and reproducibility.Many of these are little, if any, more expensivethan silica sand. A move away from silica sandis already under way in the industry, and isstrongly recommended.

The solidification pattern of castings pro-duced from permanent moulds such as gravitydies and low pressure dies may be considerablyaffected by the thickness and type of the die coatwhich is applied. A system to monitor and



control such thickness on an SPC system shouldbe seen to be in place.

For some permanent moulds, pressure die-casting and some types of squeeze casting thefeeding pattern is particularly sensitive to mouldcooling. After the development and acceptanceof the casting, any further changes to coolingchannels in the die, or to the cooling sprayduring die opening, will have to be checked toensure that corresponding deleterious changeshave not been imposed on the casting. Thequality of the water used for cooling alsorequires to be seen to be under good control ifdeposits inside the system are not to be allowedto build up and so cause changes in the effec-tiveness of the cooling system with time.

6.1.4 Dangers of solid feeding

It is often possible to make a casting withoutfeeders despite a large feeding demand. Because,in favourable conditions, the casting can col-lapse plastically, the shrinkage volume is merelytransferred from the inside to the outside of thecasting. Here, if the volume is distributed nicely,the shrinkage will cause only a negligible andprobably undetectable reduction in the size orshape of the casting.

If the outside shrinkage is not distributed sofavourably, but remains concentrated in a localregion, a surface sink is the result.

When operating without feeders, a secondpossibility is the formation of shrinkage pores,grown from initiation sites (almost certainlybifilms), so that solid feeding immediately fails.It seems that such events tend to be triggered byrather large, rather open bifilms, whose sizemight be measured in centimetres.

A further possibility of much smaller bifilmswill be common, but not easily perceived. Ifthe melt has a distribution of small, possiblymicroscopic, bifilms, these will be unfurled tosome extent by the reduced pressure in the unfedregion thus being converted from crumpledcompact features of negligible size to flat thinextensive cracks. Thus although the casting maycontinue to appear perfectly sound in the unfedregion, and solid feeding declared to be a com-plete success, the mechanical properties of thispart of the casting will be reduced. In particular,although the yield strength of the region will behardly affected, that part of the casting willexhibit reduced strength and ductility.

If the localized shrinkage problems are evenmore severe, the distribution of small bifilmswill develop further. After unfurling to becomeflat cracks, additional reduction of pressure inthe liquid will open them further to becomevisible microporosity. The pores may even grow

to such a size that they become visible onradiographs.

Thus in view of the action of a feeder topressurize the melt and so help to resist theunfurling and even the inflating of bifilms, thebifilms, are still present, but simply remain outof sight. Using a domestic analogy from homedecoration, there is a very real sense in whichadding feeders to castings is almost literally`papering over the cracks'.

6.1.5 The non-feeding roles of feeders

Feeders are sometimes important in other waysthan merely providing a reservoir to feed thesolidification shrinkage during freezing.

We have already touched on the effect thatfeeders can have on the metallurgical quality ofcast metal by helping to restrain the unfurlingand opening of bifilms by maintaining apressure on the melt. This action of the feederto pressurize the casting therefore helps tomaintain mechanical properties, particularlyductility.

A further key role of many feeders, however,is merely as a flow-off or kind of dump. Manyfilling system designs are so poor that the firstmetal entering the mould arrives in a highlydamaged condition. The presence of a generousfeeder allows some of this metal to be floatedout of the casting. This role is expected to behindered, however, in highly cored castingswhere the bifilms will tend to attach to cores intheir journey through the mould.

In general, experience with the elimination offeeders from Al alloy castings has resulted in thecasting `tearing itself apart'. This is a clear signof the poor quality of metal probably resultingmainly from the action of the poor runningsystem. The inference is that the casting is full ofserious bifilm cracks. These remain closed, andso invisible, while the feeder acts to pressurizethe metal. If the pressurization from the feeder isremoved the bifilms will be allowed to open,becoming visible as cracks. This phenomenonhas been seen repeatedly in X-ray video radio-graphy of freezing castings. It is observed thatgood filling systems do not lead to the castingtearing itself apart, even though the absence of afeeder has created severe shrinkage conditions.In this situation the casting shrinks a little more(under the action of solid feeding) to accom-modate the volume difference.

The action of the feeder to pressurize the meltduring solidification is useful in further ways.Both summarizing and thinking further we have:

(i) As we have seen, pressurization raisesmechanical properties, particularly ductility.



(ii) Pressurization together with some feedinghelps to maintain the dimensions of thecasting. Although the changes in dimen-sions by solid feeding are usually small, andcan often be neglected, on occasions thechanges may be outside the dimensionaltolerance. A feeder to ensure the provisionof liquid metal under some modest pressureis then required.

(iii) Pressurization can delay or completelyprevent blow defects from cores.

In summary, providing the filling system designis good so as to avoid creating large bifilms,and provided the solidification rate is suffi-ciently fast to retain the inherited population ofbifilms compact, castings that do not requirefeeders for feeding should not be provided withfeeders.

6.2 The seven feeding rules

Although the conditions for feeding were origi-nally listed as six rules (Castings 1991), at thattime the basic first rule was implicitly assumed.Only since has it been recognized as havingsufficient importance to be listed as a separateRule, bringing the total to seven. The originallyoverlooked first rule is `Do not feed (unlessnecessary)'.

The great literature on the feeding of castingsis mainly concerned with two feeding rules: Thefirst is The feeder must solidify at the same timeas, or later than, the casting. This is Chvorinov'sheat-transfer criterion.

The second and widely understood and well-used Rule, usually known as the volume criterionis as follows: The feeder must contain sufficientliquid to meet the volume contraction require-ments of the casting.

However, there are additional rules that arealso often overlooked, but which define addi-tional thermal, geometrical and pressure criteriathat are absolutely necessary conditions for thecasting to freeze soundly.

The junction between the casting and the feedershould not create a hot spot, i.e. have a freezingtime greater than either the feeder or the casting.This is a problem, which, if not avoided, leads to`underfeeder shrinkage porosity'. The junctionproblem is a widely overlooked requirement. Itoften overrides the Chvorinov requirement,making the feeder size calculated by the condi-tion stipulated by the Great Master to beinsufficient.

There must be a feed path to allow feed metalto reach those regions that require it. Thiscommunication criterion appears so self-evident

it is understandable why this criterion has beenoften overlooked as part of the overall logic.Nevertheless it does have a number of geomet-rical implications which are not self-evident, andwhich will be discussed.

There must be sufficient pressure differentialto cause the feed material to flow, and the flowneeds to be in the correct direction (obviouswhen spelled out!).

There must be sufficient pressure at all pointsin the casting to maintain the dimensionalaccuracy of the casting and to suppress theformation and growth of cavities. The reductionof the rate of unfurling of bifilms is alsoan important and largely unrecognized roleof the feeder, being a largely invisible contribu-tion to the mechanical properties of the castingalloy.

It is essential to understand that all of theabove criteria must be fulfilled if castings are tobe produced that require soundness, accuracyand high mechanical properties. The readermust not underestimate the scale of this prob-lem. The breaking of only one of the rules mayresult in ineffective feeding, and a defectivecasting. The wide prevalence of porosityin castings is a sobering reminder that solutionsare often not straightforward. Because thecalculation of the optimum feeder size istherefore so fraught with complications, isdangerous if calculated wrongly, costs moneyto cast on, and more money to cut off, thecasting engineer is strongly recommended toconsider whether a feeder is really necessary atall. This is our first question. You can see howvaluable it is to ask this. We shall start withthis rule.

Feeding Rule 1:Do not feed

(unless necessary)

Rule 1 is perfectly applicable to most thin-walledcastings. In fact the addition of a feeder to a thin-walled casting will often impair the casting,causing misruns as a result of the feeder fillingpreferentially to the casting or simply delaying thefilling and pressurization of the casting itself.This rule was mentioned in Castings 1991 butwas not given the status of a Rule. This was anoversight. It is probably the most important ruleof all. For instance if a feeder is incorrectly sized,violating any one of the subsequent Rules, theconsequences are so serious in some cases that itis likely that the casting would have been betterwith no feeder at all.



Probably 50 per cent of small and medium-sized castings do not need to be fed. This isespecially true as modern castings are beingdesigned with progressively thinner walls. Infact, as we have already mentioned, the siting ofheavy feeders on the top of thin-walled castingsis positively unhelpful for the filling of thecasting, since the slow filling of the feeder delaysthe filling of the thin sections at the top of thecasting, with consequent misruns.

As a general rule, therefore, it is best to avoidthe placing of feeders on thin-walled castings.The low feed requirement of thin walls can bepartly understood by assuming that of the total7 per cent solidification shrinkage in an alumin-ium casting, 6 per cent is easily provided alongthe relatively open pathway through the grow-ing dendrites. Only about the last 1 per cent ofthe volume deficit is difficult to provide. Thus ifthis final percentage of contraction on freezinghas to be provided by solid feeding, moving thewalls of the casting inwards, this becomes, atworst, 0.5 per cent per face, which on a 4 mmthick wall is only 20 mm. This small movement iseffectively unmeasurable since it is less than thesurface roughness. If this deficit does appear asinternal porosity then it is in any case ratherlimited, and normally of little consequence incommercial castings. (It may require someattention in castings for safety-critical andaerospace applications.)

The other feature of thin-walled castings isthat considerable solidification will often takeplace during pouring. Thus the casting is effec-tively being fed via the filling system. The extentto which this occurs will, of course, vary con-siderably with section thickness and pouringrate. If the section thickness (or rather, modu-lus; see below) of the filling system is similar tothat of the casting, then feeding via the fillingsystem might be a valuable simplification andcost saving. This important and welcome benefitto cost reduction is strongly recommended.

Of the remainder of castings that do suffersome feeding demand, many could avoid the useof a feeder by the judicious application of chillsor cooling fins. The general faster freezing of thecasting might then allow the provision of suffi-cient residual feeding via the filling system asindicated above. Minor revisions, opening uprestrictions to the feed path along the length ofthe filling system may provide valuable (andeffectively `free') feeding from the pouring basin.

However, this still leaves a reasonable num-ber of castings that have heavy sections, isolatedheavy bosses, or other features which cannoteasily be chilled and thus need to be fed. Theremainder of this section is devoted to gettingthese castings right.

Feeding Rule 2:The heat-transfer requirement

The heat-transfer requirement for successfulfeeding can be stated as follows: the freezingtime of the feeder must be at least as long as thefreezing time of the casting.

Nowadays this problem can be solved bycomputer simulation of solidification of thecasting. Nevertheless it is useful for the reader tohave a good understanding of the physics offeeding, so that computer predictions can bechecked, since many computer simulations arenot especially accurate at the present time, andmuch of the basic input data are not wellknown. Also, of course, computer time could beusefully avoided in sufficiently simple cases. Inthis chapter we shall concentrate on approacheswhich do not require a computer.

We have seen in Chapter 4 that the freezingtime of any solidifying body is approximatelycontrolled by its ratio (volume)/(cooling surfacearea), known as its modulus, m. Thus theproblem of ensuring that the feeder has a longersolidification time than that of the casting issimply to ensure that the feeder modulus mf islarger than the casting modulus mc. To allow afactor of safety, particularly in view of thepotential for errors of nearly 20 per cent whenconverting from modulus to freezing time, it isnormal to increase the freezing time of thefeeder by 20 per cent, i.e. by a factor of 1.2. Thusthe heat-transfer condition becomes simply

mf > 1:2 mc (6.1)

It is important to notice that the modulus hasdimensions of length. Using SI units it isappropriate to use millimetres. (Take care tonote that in French literature the normal unitsare centimetres, and in the USA at the presenttime, a confusing mixture of millimetres, centi-metres and inches, to the despair of all thosepromoters of the welcome logic of the units ofthe Systeme International. It is essential there-fore to quote the length units in which you areworking.)

The modulus of a feeder can be artificiallyincreased by the use of an insulating or exo-thermic sleeve. It can be further increased by aninsulating or exothermic powder applied to itsopen top surface after casting. Recent develop-ments in such exothermic additions haveattempted to ensure that after the exothermicreaction is over, the spent exothermic materialcontinues in place as a reasonable thermalinsulator. These products are constantly beingfurther developed, so the manufacturer's cata-logue should be consulted when working out



minimum feeder sizes when using such aids.However, as a guide as to what can be achievedat the present time, a cylindrical feeder in aninsulating material is only 0.63D in diametercompared with the diameter D in sand. Thisparticular insulated feeder therefore has only40 per cent of the volume of the sand feeder.Useful savings can therefore be made, but have,of course, to be weighed against the cost of theinsulating sleeve and the organizational effort topurchase, store, and schedule it, etc. However, afurther benefit that is easily overlooked from theuse of a more compact feeder is the fasterpressurization of thin sections that may aidfilling, and so reduce losses due to occasionalincomplete filling of mould cavities, and thefaster pressurization of cores to reduce thechance of blows.

When working out the modulus of the cast-ing it is necessary to consider which parts are ingood thermal communication. These regionsshould then be treated as a whole, characterizedby a single modulus value. Parts of the castingthat are not in good thermal communicationcan be treated as separate castings. For instance,castings of high thermal conductivity such asthose of aluminium- and copper-based alloyscan nearly always be treated as a whole, sincewhen extensive thin sections cool attachedthicker sections and bosses, the thin sections actas cooling fins for the thicker sections. Con-versely, of course, the thick sections help tomaintain the temperature of thinner sections.The effect of thin sections acting as cooling finsextends for up to approximately ten times thethickness of the thin section.

However, for castings of low thermal con-ductivity materials such as steel and nickel-basedalloys (and surprisingly, the copper-based Al-bronze), practically every part of the casting canbe treated as separate from every other. Thus acomplex product can be dealt with as anassembly of primitive shapes: plates, cubes,cylinders etc. (making allowance, of course, fortheir common mating faces, which do not countas cooling area in the modulus estimate).

Table 6.2 lists some common primitiveshapes. Familiarization with these will greatlyassist the estimation of appropriate feedermodulus requirements.

Feeder Rule 3:Mass-transfer (volume) requirement

At first sight it may seem surprising that whenRequirement 2 is satisfied then the volumerequirement is not automatically satisfied also.

However, this is definitely not the case. Althoughwe may have provided a feeder of such a size thatit would theoretically contain liquid until afterthe casting is solid, in fact it may still be too smallto deliver the volume of feed liquid that thecasting demands. Thus it will be prematurelysucked dry, and the resulting shrinkage cavitywill extend into the casting.

Figure 6.5 illustrates that normal feeders arerelatively inefficient in the amount of feed metalthat they are able to provide. This is becausethey are themselves freezing at the same time asthe casting, depleting the liquid reserves of thereservoir. Effectively, the feeder has to feed bothitself and the casting. We can allow for this inthe following way. If we denote the efficiency eof the feeder as the ratio (volume of availablefeed metal ) / (volume of feeder, Vf,) then thevolume of feed metal is, of course, eVf. Since theliquid contracts by an amount a during freezing,then the feed demand from both the feeder andcasting together is a(Vf�Vc), and hence:

eVf � a Vf � Vc� � (6.2)

or:

Vf � eÿ a� �Vc (6.3)

For aluminium where a� 7 per cent approxi-mately (see Table 6.1 for values of a for othermetals), and for a normal cylindrical feeder ofH� 1.5D where e� 14 per cent, we find:

Vf � Vc (6.4)

i.e. there is as much metal in the feeder as in thecasting! This is partly why the yield (measuredas the weight of metal going into a foundrydivided by the weight of good castings delivered)in most aluminium foundries is rarely above50 per cent. In fact, yields of 45 per cent are com-mon. Metal in the running system, and scrapallowance will reduce the overall yield of goodcastings even further. The economic benefits ofhigher-yield casting processes such as counter-gravity casting, in which metallic yields of80 to 90 per cent are common, appear compel-lingly attractive, especially for high-volumefoundries.

For steels the value of a lies between 3 and4 per cent, depending on whether solidificationis to the body-centred-cubic or face-centred-cubic structures. For pure Fe±C steels the fccstructure applies above 0.1 per cent carbonwhere the melt solidifies to austenite. For a� 4and e� 14 per cent, Equation 6.3 gives:

Vf � 0:40Vc



and for steel that freezes to the bcc structure(delta ferrite) with a� 3, and using a feeder of14 per cent efficiency we have:

Vf � 0:27Vc

Thus, compared to Al alloys, the smaller solid-ification shrinkage of ferrous metals reduces thevolume requirement of the feeder considerably.For graphitic cast irons the value reduces evenfurther of course, becoming approximately zeroin the region of 3.6 to 4.0 per cent carbonequivalent. Curiously, a feeder may still berequired because of the difference in timing

between feed demand and graphite expansion,as will be described later.

The interesting reverse tapered feeder(Figure 6.5c) has been promoted for many years(Heine 1982, Creese and Xia 1991) and iscurrently widely used for ductile iron castings.

Even so, the reader needs to be aware that inthe opinion of the author, Figure 6.5 may not beas accurate as we would like. At this time, theextent of the uncertainties is not known fol-lowing the recent work of Sun and Campbell(2003). This investigation of the effect of posi-tive and negative tapers on the efficiencies offeeders, found that the reverse tapered feeder

�� ) # /��

%�� )� ��

5&&I ! � �� ;��

7��

D�

�(,<C? M M

��.�

D�

�(,<C?

�

�($((?

�0� �� @>?

D

H

,(�

�(,<C?

�

�($((?

,=��

��(,%%?

��

��($,/?

$(�

�($((?

��

�($$$?

�� 0� ��

�

�($=(? M M

��

D

�

�(=((? M M

*�� < +�� 1�� ,$2


(Figure 6.5c) appeared to be less efficient thanparallel sided cylindrical feeders, or even feederswith a slight positive taper. These doubts are anunwelcome sign of the extent of our ignoranceof the best feeder designs at this time.

Whether the size of the feeder is dictated bythe thermal or volume requirement is related tothe geometry of the casting. Figure 6.6 shows atheoretical example, calculated neglecting non-cooling interfaces for simplicity. Curve A is theminimum feeder volume needed to satisfy thethermal condition mf� 1.2mc; and curve B isthe minimum feeder volume needed to satisfy thefeed demand criterion based on 4 per centvolume shrinkage and 14 per cent metal utiliza-tion from the feeder. Figure 6.6 reveals thatchunky steel plates up to an aspect ratio of about6 or 7 length to thickness are properly fed by afeeder dictated by freezing time requirements.However, thin section steel plates above thiscritical ratio always freeze first, and so require afeeder size dictated by volume requirements.

In fact the shape of the shrinkage pipe in thefeeder is likely to be different for each of theseconditions. For instance, the feeder efficienciesshown in Figure 6.5 are appropriate for thefeeding of chunky castings because the con-tinuing demand of feed metal from the castinguntil the feeder itself is almost solid naturallycreates a long, tapering shrinkage pipe, resem-bling a carrot.

In the case of the more rapid solidification ofthin castings, the relatively large diameter feederneeded to provide the volume requirement willgive a shallowly dished shape in the top of thefeeder, since the feed metal is provided early,before the feeder itself has solidified to any greatextent. The efficiency of utilization of the feederwill therefore be expected to be significantlyhigher, as confirmed by Figure 6.6.

Research is needed to clarify this point. In themeantime, the casting engineer needs to treat thepresent data with caution, and conclusions fromFigures 6.5 and 6.6, for instance, have to be viewedas illustrative of general principles rather thannumerically accurate. Clearly, it is desirable toachieve smaller, more cost-effective feeders. Thechange of feeder efficiency depending on whetherfreezing or volume requirements are operatingrequires more work to clarify this uncertainty. Inthe meantime, this problem illustrates the powerof a good computer simulation to avoid thenecessity for simplifying assumptions.

A further use of feeders where the castingengineer requires care is the use of blind feederssited low down on the casting. The problems arecompounded if such low-sited blind feeders areused together with open feeders placed higher. Itmust be remembered that during the early stagesof freezing the top feeder is supplying metal tothe blind feeders as well as the casting. The blindfeeders have to be treated as though they are an

Figure 6.5 Metal utilization of feeders of various forms moulded in sand. The (a) cylindrical and (b) hemisphericalheads have been treated with normal feeding compounds; (c) efficiency of the reverse taper heads depends on detailedgeometry (Heine, 1982, 1983); (d) exothermic sleeve (Beeley, 1972). Metal utilization for ductile iron plates with(e) cylindrical sand feeder; (f) insulating feeder; and (g) cruciform exothermic feeder (after Foseco 1988).


(e) (f) (g)

14%

20%

(a) (b) (c) (d)

67%10%to40%


integral part of the casting. The size of the topfeeder needs to be enlarged accordingly. Theblind feeders only start to operate independentlywhen the feed path from the top feeder freezesoff. This point occurs when the solidificationfront has progressed a distance d/2, where d isthe thickness of the thickest casting sectionbetween the top and the blind feeders. Thus thevolume of the blind feeder is now reduced by thed/2 thickness layer of solid that has alreadyfrozen around its inner walls.

If this caution were not already enough, afurther pitfall is that the thickness of solid shellinside the blind feeder may now exceed thelength of the atmospheric vent core, creepingover its end and sealing it from the atmosphere.

It is therefore prevented from breathing, and isunable to provide any feed metal.

There are therefore subtleties in the operationof blind feeders that make success illusive. It iseasy to make a mistake in their application, andthe correct operation of the atmospheric ventis not always guaranteed, so it is difficult torecommend their use on smaller castings. Forlarger castings, where the feeder size is large, thecollapse of the top of blind feeders is morepredictable, and they become more reliable.

Whereas the size of feeders for alloys such asthose based on alloys that shrink in a conven-tional fashion on freezing are straightforward tounderstand and work with, graphitic cast ironsare considerably more complicated in theirbehaviour. They are therefore more complicatedto feed, and the estimate of feeder sizes subjectto more uncertainty.

The amount of graphite that is precipitateddepends strongly on factors that are not easy tocontrol, particularly the efficiency of inocula-tion. In addition, the expansion of the graphitecan lead to an expansion of the casting if themould and its container are not rigid. This leadsto a larger volume of casting that requires to befed, with the danger that the feeder is nowinsufficient to provide this additional volume.Shrinkage porosity as a result of mould dilationis a common feature of iron castings. One of theways to reduce this problem is to use very dense,rigid moulds in rigid, well-engineered boxes.Furthermore, the expansion of the graphite canbe accommodated without swelling the castingby allowing the residual melt to exude out of thecasting and into the feeder. The provision of asmall feeder is therefore essential to the pro-duction of many geometries of small iron cast-ings, even though subsequent examination ofthe feeder indicates, mysteriously, that no liquidhas been provided by the feeder.

Ductile cast irons commonly use a reverse-tapered feeder such as that shown in Figure 6.5c.If the feeder remained full for too long, its topwould freeze over, preventing the delivery ofany liquid (recall the coffee-cup experimentFigure 6.10). It is logical therefore to encouragethe feeder to start feeding almost immediately,concentrating the action of shrinkage through-out the casting all on the small area at the top ofthe feeder. In this way the level of liquid in thetop of the feeder falls quickly, becoming sur-rounded by hot metal, so very soon there is nodanger of it freezing over. For this to happen itis essential that no feeding continues to beprovided via the running system. This wouldkeep the feeder full for too long. Thus it isnecessary to design the ingate to freeze quickly.The feeder then works well.

Figure 6.6 Feeder volume based on a feeder moulded insand, and calculated neglecting non-cooling interfaces forsimplicity. Curve A is the minimum feeder volume tosatisfy mf� 1.2 mc; and curve B is the minimum volume tosatisfy the feed demand of 4 per cent volume shrinkage and14 per cent utilization of the feeder.



This feeding technique, although at this timeused exclusively in the ductile iron industry sofar as the author is aware, would be expected tobe applicable to a wide variety of metals andalloys.

Feeding Rule 4:The junction requirement

The junction problem is a pitfall awaiting theunwary. It occurs because the simple act ofplacing a feeder on a casting creates a junction.As we have seen from Section 2.3.2.6, a junctionwith an inappropriate geometry will lead to ahot spot. The hot spot may cause a shrinkagecavity that extends into the casting.

The range of simple T-junctions was shownin Figure 2.32. Clearly the problem junctions arethose with 1 : 1 ratios of the thickness of theupright to that of the horizontal. If we assumethat the feeder when planted on the casting is akind of T-junction, and if we further assumethat the ratios of thickness discussed so far forsimple plates are also valid for ratios of modulusof shaped castings, we can, with some justifica-tion, extend the T-junction findings to identifythe 1 : 1 ratio of feeder modulus to castingmodulus is a problem.

The simplest example illustrating the prob-lem clearly is that of the feeding of a cube. Thecube casting has the reputation of being notor-iously difficult to feed. This is because thecasting technologist, carefully following Rule 2,calculates a feeder of 1.0 or perhaps 1.2 timesthe modulus of the cube. If the cube has sidelength D, then the feeder of 1 : 1 height to diam-eter ratio works out to have a diameter of 1.2D.Thus the cube appears to require a feeder ofrather similar volume sitting on top. However,the cube and its feeder are now a single compactshape that solidifies as a whole, with its thermalcentre in the centre of the new total cast shape,i.e. approximately in the centre of the junction.The combination therefore develops a shrinkagecavity at the junction, the hot spot between thecasting and feeder. When the casting is cut offfrom the feeder, the porosity that is found isgenerally called `under-feeder shrinkage poros-ity.' This rather pompous pseudo-technical jar-gon clouds the clear conclusion that the feeder istoo small.

Returning to our junction rules; to avoidcreating a hot spot we need to ensure that thefeeder actually has twice the modulus of thecasting. Thus the cube should have had a feederof side length 2D. The shrinkage cavity wouldthen have been concentrated only in the feeder.

However, in some cases the junction problemcan be avoided. The simplest solution is not toplace the feeder directly on the casting so as tocreate a junction. It happens that this rule is noteasily applied to a cube because there is noalternative site for it.

However, in the case of a plate casting, thereare options. The feeder should not be placeddirectly on the plate, but should be placed on anextension of the plate.

The general rules to solve the junction pro-blem are therefore as follows:

1. Appendages such as feeders and ingatesshould not be planted on the casting so asto create a T- (or an L-) junction (althoughthe L-junction is rather less detrimental thanthe T-junction). They are best added asextensions to a section, as an elongation toa wall or plate, effectively moving the junc-tion off the casting.

2. If there is no alternative to the placing of thefeeder directly on the casting, then to avoidthe hot spot in the middle of the junction, theadditional requirement that the feeder mustmeet is, if a T-junction,

mf42mc (6.5)

and if an L-junction

mf41:33mc (6.6)

The value of the constants is taken from Sciama(1974).

Note that no safety factor of 1.2 has beenapplied to these feeder sizes. This is because theshrinkage cavity does not occur exactly at thegeometric centre of the freezing volume trappedat the thermal centre; the cavity `floats' to thetop, and the feed liquid finds its level at the baseof the isolated region. Thus the final shrinkagecavity is naturally displaced above the junctioninterface, giving a natural `built-in' safety factor.

Note that we have assumed that the feeder isabove the casting, so as to feed downwardsunder gravity. This is the recommended safeway to use feeders. If the feeder (or large gate)were placed below the casting, gravity wouldnow act in reverse, so that any shrinkage cavitycaused by the junction would float into thecasting (effectively, the residual liquid metal inthe casting drains into the feeder). This actionillustrates one of the dangers of attempting tofeed uphill. Conditions in which feeders mightbe used to feed uphill are discussed later inSection 6.4.

Although it is not a good idea to make thefeeder any larger than is really required, if it is



only marginally adequate, the tail of porosityseen in Figure 6.2 may on occasions just enterthe casting, and may therefore be unacceptable.This necessitates the application of a safetyfactor, giving a feeder of larger size on average,but still just acceptable even when all the vari-ables are loaded against it. It is common to usethe factor 1.2.

Feeding Rule 5:Feed path requirement

There must be a feed path. It is clearly no usehaving feed metal available at one point on thecasting, unable to reach a more distant pointwhere it is needed. Clearly there has to be a waythrough.

In a valuable insight, Heine (1968) has drawnattention to the fact that the highest-modulusregions in a casting are either potential regionsfor shrinkage porosity if left unfed, or may befeeding paths. He recommends the identificationof feed paths that will transport feed metalthrough castings of complex geometry, such asthe hot spots at the T-junctions between plates.(He also draws attention to the fact that certainlocations are never feed paths. These includecorners or edges of plates, or the ends of barsand cylinders.)

The various ways to help to ensure that feedpaths remain open are considered in this section.

Directional solidification towards the feeder

If the feeder can be placed on the thickest sec-tion of the casting, with progressively thinnersections extending away, then the condition ofprogressive solidification towards the feeder canusually be achieved.

A classical method of checking this due toHeuvers can be used in which circles are inscri-bed inside the casting sections. If the circle dia-meters increase progressively towards the feederthen the condition is met (Figure 6.7). Lewis andRansing (1998) draw attention to the fact thatHeuvers' technique is only two-dimensional,and that the condition would be more accu-rately represented in three dimensions by aprogressive change in the radius of a sphere,effectively equivalent to the progressive increasein casting modulus towards the feeder. Thefundamental reason for tapering the casting inthis way is to achieve taper in the liquid flowpath (Sullivan et al. 1957). For convenience weshall call this the modulus gradient technique.

Failure to provide sufficient modulus gra-dient towards the feeder can be countered in

various ways by (i) either re-siting the feeder,or (ii) providing additional feeder(s), or(iii) modifying the modulus of the casting.Ransing describes a further option, (iv) in whichhe proposes a change in heat transfer coeffi-cient. The latter technique is a valuable insightbecause it is easily and economically computedby a geometrical technique, and so contrastswith the considerable computing reserves andeffort required by finite element and finite dif-ference methods. If R is the radius of theinscribed sphere, the local solidification time t isproportional to R2/h where h is the heat transfercoefficient at the metal/mould interface. Thus atlocations 1 and 2 we have h2� h1(R1/R2)2. Thisrelation allows an estimate of the change of hthat is required to ensure that the freezing timeincreases steadily towards the feeder. Ransinguses a change of 10 per cent increase of solidi-fication time for each different geometrical sec-tion of the casting (i.e. for feeding via a thinsection to a thick distant section he increasesthe freezing time of the thin section over that ofthe thick by 10 per cent). The technique can beusefully employed in reverse, in the sense thatknown values of h produced by a chill canquickly be checked for their effectiveness indealing with an isolated heavy section. In everycase the target is to eliminate the local hotspot, and ensure a continuous feed path backto the feeder. This simple technique has ele-gance, economy and power and is stronglyrecommended.

The casting modulus can be modified byproviding either a chill or cooling fin to speed

Figure 6.7 Use of Heuvers circles to determine the amountof attached padding (Beeley 1972) and the use of detached(or indirect) padding described by Daybell (1953).



solidification locally, or by providing extrametal to thicken the section and delay solidifi-cation locally. The provision of extra metal onthe casting is known as padding. The addition ofpadding is most usefully carried out with thecustomer's consent, so that it can be left inposition as a permanent feature of the castingdesign. If consent cannot be obtained then thecaster has to accept the cost penalty of dressingoff the padding as an additional operation aftercasting.

Occasionally this problem can be avoided bythe provision of detached, or indirect, paddingas shown in Figure 6.7. Daybell (1953) wasprobably the first to describe the use of thistechnique. The author has found it useful in theplacement of feeders close to thin adjacent sec-tions of casting, with a view to feeding throughthe thin section into a remote thicker section.

The principle of progressive increase ofmodulus towards the feeder, although generallyaccurate and useful, is occasionally seen to benot quite true. Depending on the conditions, thisfailure of the principle can be either a problem ora benefit, as shown below. (Even so, the Ransingtechnique described above is unusual since itsuccessfully takes this problem into account.)

In a re-entrant section of a casting the con-fluence of heat flow into the mould can cause ahot spot, leading to delayed solidification at thispoint, and the danger of local shrinkage por-osity in an alloy that shrinks during freezing,such as an Al alloy. Alternatively, in an alloythat expands such as a high carbon-equivalentcast iron, the exudation of residual liquid intothe mould as a result of the high internal pres-sure created during the precipitation of graphitecan cause penetration of the aggregate mouldmaterial, with unwelcome so-called burned-onsand. Such a hot spot can occur despite anapparent unbroken increase in modulus throughthat region towards the feeder. This is becausethe simple estimation of modulus takes noaccount of the geometry of heat flow away fromcooling surfaces; all surfaces are assumed to beequally effective in cooling the casting. Such ahot spot requires the normal attention such asextra local cooling by chill or fin, or additionalfeed via extra padding or feeders.

The failure of the modulus gradient techni-que can be used to advantage in the case offeeder necks to reduce the subsequent cut-offproblem. Feeders are commonly joined to themain casting via a feeder neck, with the modulusof the neck commonly controlled to be inter-mediate between that of the casting and thefeeder; the moduli of casting, neck and feederare in the ratio 1.0 : 1.1 : 1.2 (Beeley 1972).However the neck can be reduced considerably

below this apparently logical lower limit,because of the hot spot effect, and because of theconduction of heat from the neighbouringcasting and feeder that helps to keep the neckmolten for a longer period than its modulusalone would suggest. This point is well illu-strated by Sciama (1975), and his results aresummarized in Figure 6.8. The results clearlydemonstrate that for steel, feeder necks can bereduced to half of the diameter D of the feeder,providing that they are not longer than 0.1D.The higher thermal conductivity copper- andaluminium-based materials can have necksalmost twice as long without problems.

By extrapolation of these results towardssmaller neck sizes, it seems that a feeder neck insteel can be only 0.25D in diameter, providing itis no more than approximately 0.03D in length.Similarly, for copper and aluminium alloys the0.25D diameter neck can be up to 0.06D long.These results explain the action of the Wash-burn core, or breaker core, which is a wafer-thincore with a narrow central hole, and which isplaced at the base of a feeder, allowing it to beremoved after casting by simply breaking it off.In separate work the dimensions of typicalWashburn cores is recommended to be a thick-ness of 0.1D and a central hole diameter of 0.4D(work by Wlodawer summarized by Beeley1972). The hole size and thickness appear to bevery conservative in relation to Sciama's work.However, Sciama may predict optimistic resultsbecause he uses a feeder of nearly 1.5 times themodulus of casting, which would tend to keepthe junction rather hotter than a feeder with a

Figure 6.8 Effect of a constricted feeder neck onsoundness of steel, aluminium bronze, and 99.5Alcastings. The experimental points by Sciama (1975)denote marginal conditions.



modulus of only 1.2 times that of the casting (itwould be valuable to repeat this work using amore economical feeder). Also, of course, con-servatism may be justified where feeding con-ditions are less than optimum for other reasonsin the real foundry environment.

The aspect of conservatism because of thereal foundry environment is an interesting issue.For instance, the feeder neck could, in theextreme theoretical case, be of zero diameterwhen the thickness of the feeder neck core waszero. The reductio ad absurdum argument illus-trates that an extreme is not worth targeting,especially when sundry debris in suspension,such as metal dendrites or rigid bifilms, couldclose off a narrow aperture.

Minimum temperature gradient requirement

Experiments on cast steels have found that whenthe temperature gradient at the solidus (i.e. thetemperature at which the final residual liquidfreezes) falls to below approximately 0.1±1 K mmÿ1 then porosity is observed even in well-degassed material. Although there is muchscatter in other experimental determinations, itseems in general that the corresponding gra-dients for copper alloys are around 1 K mmÿ1

and those for aluminium alloys around2 K mmÿ1 (Pillai et al. 1976). It seems thereforethat the temperature gradient defines a criticalthreshold of a non-feeding condition. As theflow channel nears its furthest extent, andbecomes vanishingly narrow, it will becomesubject to small random fluctuations in tem-perature along its length. This kind of tem-perature `noise' will occur as a result of smallvariations in casting thickness, or of density ofthe mould, thickness of mould coating, block-age or diversion of heat flow direction by ran-dom entrained films etc. Thus the channel willnot reduce steadily to infinite thinness, but willterminate when its diameter becomes close tothe size of the random perturbations.

There has been some discussion about theabsolute value of the critical gradient for feedingon the grounds that the degree of degassing, orthe standard of soundness, to which the castingwas judged, will affect the result. These arecertainly very real problems, and do help toexplain some of the wide scatter in the results.

Hansen and Sahm (1988) draw attention to amore fundamental objection to the use of tem-perature gradients as a parameter that mightcorrelate with feeding problems. They indicatethat the critical gradient required to avoidshrinkage porosity in a steel bar is five to tentimes higher than that required for a plate, andpoint to other work in which the critical

gradient in a cylindrical steel casting is a functionof its diameter. Thus the concept of a singlegradient which applies in all conditions seemsto be at fault. If this can be confirmed, whichseems likely, then its use will require to bere-thought.

Feeding distance

It is easy to appreciate that in normal conditionsit is to be expected that there will be a limit tohow far feed liquid can be provided along a flowpath. Up to this distance from the feeder thecasting will be sound. Beyond this distance thecasting will be expected to exhibit porosity.

This arises because along the length of aflow channel, the pressure will fall progres-sively because of the viscous resistance to flow.(This effect was covered in more detail inCastings 2003.) When the pressure falls to acritical level, which might actually be negative,then porosity may form. Such porosity mayoccur from an internal initiation event (such asthe opening of a bifilm), or from the drawinginwards of feed metal from the surface of thecasting, since this may now represent a shorterand easier flow path than supply from the moredistant feeder.

There has been much experimental effort todetermine feeding distances. The early work byPellini and his co-workers (summarized byBeeley (1972)) at the US Naval MaterialsLaboratory is a classic investigation that hasinfluenced the thinking on the concept of feed-ing distance ever since. They discovered that thefeeding distance Ld of plates of carbon steelscast into greensand moulds depended on thesection thickness T of the casting: castingscould be made sound for a distance from thefeeder edge of 4.5T. Of this total distance, 2.5Tresulted from the chilling effect of the castingedge; the remaining 2.0T was made sound by thefeeder. The addition of a chill was found toincrease the feeding distance by a fixed 50 mm(Figure 6.9). They found that increasing thefeeder size above the optimum required toobtain this feeding distance had no beneficialeffects in promoting soundness. The feedingdistance rule for their findings is simply:

Ld � 4:5T (6.7)

Pellini and colleagues went on to speculate thatit should be possible to ensure the soundness ofa large plate casting by taking care that everypoint on the casting is within a distance of 2.5Tfrom an edge, or 2.0T from a feeder.

Note that all the semi-empirical computerprograms written since have used this and the



associated family of rules as illustrated inFigure 6.9 to define the spacing of feeders andchills on castings. However, the original datarelate only to steel in greensand moulds, and

only to rather heavy sections ranging from50±200 mm. Johnson and Loper (1969) haveextended the range of the experiments down tosections a thickness of 12.5 mm and have re-analysed all the data. They found that for plates,the data, all in units of millimetres, appeared tobe more accurately described by the equation:

Ld � 72m1=2 ÿ 140 (6.8)

and for bars:

Ld � 80m1=2 ÿ 84 (6.9)

where m is the modulus of the cast section inmillimetres. The revised equations by Johnsonand Loper have usually been overlooked inmuch subsequent work. What is also over-looked is that all the relations apply to cast mildsteels in greensand moulds, not necessarily toany other casting alloys in any other kinds ofmould.

In their nice theoretical model, Kubo andPehlke (1985) find support for Pellini's feedingdistance rules for steel castings, but it is a con-cern that no equivalent rule emerged forAl±4.5Cu alloy that they also investigated.

In fact, colleagues of Flinn (1964) foundthat whereas the short-freezing-range alloysmanganese bronze, aluminium bronze, and 70/30cupro-nickel all had feeding distances thatincreased with section thickness, the long-freez-ing-range alloy tin bronze appeared to react inthe opposite sense, giving a reduced feedingdistance as section thickness increased. (Thenominal composition of this classical long-freezing-range material is 85Cu, 5Sn, 5Zn, 5Pb. Itwas known among traditional foundrymenas `ounce metal' since to make this alloy theyneeded to take one pound of copper to which oneounce of tin, one ounce of zinc, and one ounce oflead was added. This gives, allowing for smalllosses on addition, the ratios 85 : 5 : 5 : 5.)

Kuyucak (2002) reviews the relations forestimating feeding distance in steel castings, andfinds considerable variation in their predictions.This makes sobering reading.

Jacob and Drouzy (1974) found long feedingdistances, greater than 15T, for the relativelylong-freezing-range aluminium alloys Al±4Cuand Al±75i±0.SMg, providing the feeder is cor-rectly sized.

All this confusion regarding feeding distancesremains a source of concern. We can surmisethat the opposite behaviour of short- and long-freezing-range materials might be understood interms of the ratio pasty zone/casting section.For short-freezing-range alloys this ratio is lessthan 1, so the solidified skin of the alloy is

Figure 6.9 The famous results by Pellini (1953) for (a)the temperature distribution in a solidifying steel bar; and(b) the feeding distances for steel plates cast in greensand.



complete, dictating feeding from the feeder, andthus normal feeding distance concepts apply.

For the case of long-freezing-range materialswhere the pasty zone/casting section ratio isgreater than 1, and in fact might be 10 or more,the outer solid portions of the casting are farfrom solid for much of the period of solidifica-tion. The connections of liquid through to theouter surface will allow flow of liquid from thesurface to feed solidification shrinkage. Inaddition, the higher temperature and lowerstrength of the liquid/solid mass will allowgeneral collapse of the walls of the castinginwards, making an important contribution tothe feeding of the inner regions of the casting bythe `solid feeding' mechanism. It is for this rea-son that the higher conductivity, and lowerstrength alloys of Al and Cu can be character-ized by practically infinite feeding distances,particularly if the alloys are relatively free frombifilms. Internal porosity simply does notnucleate, no matter how distant the castinghappens to be from the feeder; the outer walls ofthe casting simply move inwards very slightly.

Thus although the general concept of feedingdistance is probably substantially correct, atleast for short-freezing-range alloys, and parti-cularly for stronger materials such as steels, itshould be used, if at all, with great caution fornon-ferrous metals until it is better understoodand quantified. In summary it is worth notingthe following:

1. The data on feeding distances have beenderived from extensive work on carbon steelscast in greensand moulds. Relatively littlework has been carried out on other metals inother moulds.

2. The definition of feeding distance is sensitiveto the level of porosity that can be detectedand/or tolerated.

3. It is curious that the feeding distance isdefined from the edge of a feeder (not itscentreline).

4. The quality of the cast metal in terms of itsgas and oxide content would be expected tobe crucial. For instance, good quality metalachieved by the use of filters and gooddegassing and casting technique (i.e. witha low bifilm content) would be expected toyield massive improvements in feeding dis-tance. This has been demonstrated byRomero et al. (1991) for Al-bronze. Berryand Taylor (1999) report a related effect,while reviewing the benefit to the feedingdistance of pressurizing the feeder. This workis straightforwardly understood in terms ofthe pressure on the liquid acting to suppressthe opening of bifilms.

A final note of caution relates to the situationwhere the concept of feeding distance applies toan alloy, but has been exceeded. When this hap-pens it is reported that the sound length is con-siderably less than it would have been if thefeeding distance criterion had just been satisfied.If true, this behaviour may result from the spreadof porosity, once initiated, into adjacent regions.The lengths of sound casting in Figure 6.9a areconsiderably shorter than the maximum lengthsgiven by Equations 6.5 to 6.7, possibly becausethe feeding distance predicted by these equationshas been exceeded and the porosity has spread.Mikkola and Heine (1970) confirm this unwel-come effect in white iron castings.

Other parameters (criteria functions)

In a theoretical study of the formation of por-osity in steel plates of thickness 5 to 50 mm, withand without end chills, Minakawa et al. (1985)investigated various parameters that might beuseful in assessing the conditions for the onset ofporosity in their castings. They looked at G, thetemperature gradient along the centreline of thecasting at the solidification front, and the frac-tion solid fs along the centreline. Neither ofthese was satisfactory. However, they did findthat the parameter G/V1/2 suggested by Niyamaet al. (1982) correctly assessed the difficulty ofproviding feed liquid under the various condi-tions of their work, where V is the velocity ofadvance of the freezing front. In plate-likecastings the value of G drops to low levels in thecentre of the plate, and at the same time Vincreases because the front accelerates along thecentre of the plate, reaching its highest velocity,requiring feed metal at the highest rate. It thuscreates the largest pressure drop to drive thisflow. To obtain sound castings, therefore, theyfound that the value of G/V 1/2 has to be atleast 1.0 K s1/2 mmÿ 3/2.

It would be valuable to know whether thisparameter is similarly discriminating for othercasting alloys, particularly the high-thermal-conductivity alloys of aluminium and copper.

In another theoretical study Hansen andSahm (1988) support the usefulness of G/V1/2 forsteel castings. However, in addition they go onto argue the case for the use of a more complexfunction G/V1/4 VL

1/2 where VL is the velocity offlow of the residual liquid.

They proposed this relation because theynoticed that the velocity of flow in bars was fiveto ten times the velocity in plates of the samethickness, which, they suggest, contributes tothe additional feeding difficulty of bars com-pared to plates. (A further contributor will bethe comparatively high resistance to collapse



that is shown by bars, compared to the effi-ciency of solid feeding in plates as will be dis-cussed later.) They found that G/V1/4VL

1/25 acritical value, which for steel plates and bars isapproximately 1 K s3/4 mmÿ7/4. Their parameteris, of course, less easy to use than that due toNiyama, because it needs flow velocities. TheNiyama approach only requires data obtainablefrom temperature measurements in the casting.

Feeding Rule 6:Pressure gradient requirement

Although all of the previous feeding rules maybe met, including the provision of feed liquidand a suitable flow path, if the pressure gradientneeded to cause the liquid to flow along the pathis not available, then feed liquid will not flowto where it is needed. Internal porosity maytherefore occur.

A positive pressure gradient from the outsideto the inside of the casting will help to ensurethat the feed material (either solid or liquid)travels along the flow path into those parts ofthe casting experiencing shrinkage. The variousfeeding mechanisms (to be discussed later) areseen to be driven by the positive pressure such asatmospheric pressure and/or the pressure due tothe hydrostatic head of metal in the feeder. Theother contributor to the pressure gradient, thedriving force for flow, is the reduced or evennegative pressure generated within poorly fed

regions of some castings. All of these drivingforces happen to be additive; the flow of feedmetal is caused by being pushed from the out-side and pulled from the inside.

Figure 6.10 illustrates the feeding problemsin a complicated casting. The casting divideseffectively into two parts either side of the bro-ken line. The left-hand side has been designed tobe fed by an open feeder Fl and a blind feederF2. The right-hand side was intended to be fedby blind feeder F3.

Feeder Fl successfully feeds the heavy sectionS1. This feeder is seen to be comparatively large.This is because it is required to provide feedmetal to the whole casting during the earlystages of freezing, while the connecting sectionsremain open. At this early stage of inter-connection of the whole casting, the top feeder isalso feeding both blind feeders, of course.

Feeder F2 feeds S5 because it is providedwith an atmospheric vent, allowing the liquid tobe pressurized by the atmosphere as in the coffeecup experiment illustrated below, so forcing themetal through into the casting. (The reader isencouraged to try the coffee cup experiment.)

The identical heavy sections S3 and S4 showthe unreliability of attempting to feed uphill. InS4 a chance initiation of a pore has created afree liquid surface, and the internal gas pressurewithin the casting happens to be close to1 atmosphere. Thus the liquid level in S4 falls,finding its level equal to that in the feeder F2.(If the internal gas pressure within the casting

Figure 6.10 (a) Castings withblind feeders, F2 is correctlyvented but has mixed results onsections S3 and S4. Feeder F3 isnot vented and therefore doesnot feed at all. The unfavourablepressure gradient draws liquidfrom a fortuitous skin puncturein section S8. See text forfurther explanation. (b) Theplastic coffee cup analogue: thewater is held up in the upturnedcup and cannot be released untilair is admitted via a puncture.The liquid it contains is thenimmediately released.



had been much less than an atmosphere, thenthe level in S4 would have been correspond-ingly higher.) The surface-initiated pore in S2has grown similarly, equalizing its level exactlywith that in the feeder F2, since both surfacesare subject to the same atmospheric pressure.In Section S3, by good fortune, no pore initia-tion site is present, so no pore has occurred,with the result that atmospheric pressure viaF2 (and unfortunately also via the puncture bythe atmosphere at the hot spot in there-entrant section S2) will feed solidificationshrinkage here, causing the section to be per-fectly sound.

Turning now to the right-hand part of thecasting, although feeder F3 is of adequate size tofeed the heavy sections S6, S7 and S8, itsatmospheric vent has been forgotten. This is aserious mistake. The plastic coffee cup experi-ment (Figure 6.10) shows that such an invertedair-tight container cannot deliver its liquidcontents. The pressure gradient is now reversed,causing the flow to be in the wrong direction,from the casting to the feeder! The detailedreasoning for this is as follows. The pressure inthe casting and feeder continues to fall asfreezing occurs until a pore initiates, eitherunder the hydrostatic tension, or because of abuild-up of gas in solution, or because of theinward rupture of the surface at a weak pointsuch as the re-entrant angle in section S8. Thepressure in section S8 is now raised to atmos-pheric pressure while the pressure in the feederis still low, or even negative. Thus feed liquid isnow forced to flow from the casting into thefeeder as freezing progresses. A massive porethen develops because feeder F3 has a large feedrequirement, and drains section S8 and thesurrounding casting. The defect size is worsethan that which would have occurred if nofeeder had been used at all!

Section S6 remains reasonably sound becauseit has the advantage of natural drainage ofresidual liquid into it. Effectively it has been fedfrom the heavy section S8. The pressure gra-dient due to the combined actions of gravity,shrinkage and the atmosphere from S8 to S6 ispositive. The only reason why S6 may displayany residual porosity may be that S8 is a ratherinadequate feeder in terms of either its thermalrequirement or its volume, or because the feedpath may be interrupted at a late stage.

Section S7 cannot be fed because there is nocontinuous feed path to it. S9 is similarly dis-advantaged. This has been an oversight in thedesign of the feeding of this casting. In a sandcasting it is likely that S7 and S9 will thereforesuffer porosity. This will be almost certainly truefor a steel casting, but less certain if the casting

is a medium-freezing-range aluminium alloy.The reason becomes clear when we consider aninvestment casting, which, if a high mouldtemperature is chosen, and if the metal is clean,will allow solid feeding to operate, allowing thesections the opportunity to collapse plastically,and so become internally sound, provided thatno pore-initiation event interrupts this action.Solid feeding is often seen in aluminium alloysand castings, but rarely in steel sand castingsbecause of the greater rigidity of the solidifiedsteel, which successfully resists plastic collapsein cold moulds.

The exercise with the plastic coffee cup showsthat the water will hold up indefinitely in theupturned cup until released by the pin causing ahole. The cup will then deliver its contentsimmediately (but not before!). Blind feeders aretherefore often unreliable in practice because theatmospheric vent may not open reliably. Suchfeeders then act to suck feed metal from thecasting, making any porosity worse.

If a blind feeder is provided with an effectiveatmospheric vent, then the available atmos-pheric pressure may help it to feed uphill. Themaximum heights supportable by one atmos-phere for various pure liquids near their meltingpoints are:

However, as we have seen, feeding uphill is notaltogether reliable, and cannot be recommendedas a general technique. To restate the reasonsbriefly, this is because any initiation of ashrinkage or gas pore, or any inward rupture ofthe casting surface, will release the internalstress of the casting, removing the pressure dif-ference between the casting and feeder. With thepressures in casting and feeder equalized, themetal level in the casting will fall, and that inthe feeder will rise so as to equalize the levels ifpossible. The result is a porous casting. Blindfeeders that are placed low on the casting can beunreliable in practice for this reason.

This loss of pressure difference cannot occurif the feeder is placed above the general level ofthe casting so that feeding always takes placewith the assistance of gravity. Feeder F2 inFigure 6.10 would have successfully fed sectionsS2 and S4 either if it was taller, or if it had beenplaced at a higher location, for instance on thetop of S4.

It is clear that F3 may not have fed section S8if the corner puncture occurred, even if it had

Mercury 0.760 m (barometric height)Steel 1.48 mZinc 1.58 mAluminium 4.36 mMagnesium 6.54 mWater 10.40 m



been provided with an effective vent, becausethe pressure gradient for flow would have beenremoved. A provision of an effective vent, andthe re-siting of the base of the feeder F3 to theside of S8, would have maintained the sound-ness of both S6 and S8 and would have pre-vented the surface puncture at S8. S7 and S9would still have required separate treatment.

The conclusion to these considerations is:place feeders high to feed downhill. This is ageneral principle of great importance. It is ofsimilar weight to the general principle discussedpreviously, place ingates low to fill uphill. Theseare fundamental concepts in the production ofgood castings.

Feeding Rule 7:Pressure requirement

The final rule for effective feeding is a necessaryrequirement like all the others. Sufficient pres-sure in the residual liquid within the castingis required to suppress both the initiation andthe growth of cavities both internally andexternally.

This is a hydrostatic requirement relating tothe suppression of porosity, and contrasts withthe previous pressure gradient requirement thatrelates to the hydrodynamic requirements forflow (especially flow in the correct direction!)

A fall in internal pressure may cause a varietyof problems:

1. Liquid may be sucked from the cast surface.This is particularly likely in long-freezing-range alloys, or from re-entrant angles inshorter-freezing-range alloys, resulting ininternal porosity initiated from, and con-nected to, the outside.

2. The internal pressure may fall just suffi-ciently to unfurl, but not fully open thepopulation of bifilms. The result will be anapparently sound casting but poor mechani-cal properties, particularly a poor elongationto failure. (It is possible that some so-called`diffraction mottle' may be noted on X-rayradiographs.)

3. The internal pressure may fall sufficiently toopen bifilms, so that a distribution of fineand dispersed microporosity will appear. Themechanical properties will be even lower.

4. The internal shrinkage may cause macro-shrinkage porosity to occur, especially if thereare large bifilms present as a result of poorfilling of the casting. Properties may now be indisaster mode and/or large holes may appearin the casting. Figure 6.11 illustrates pressure

loss situations in castings that can result inshrinkage porosity. Figure 6.12 illustrates thecommon observation in Al alloy castings inwhich a glass cloth is placed under the feederto assist the break-off of the feeder aftersolidification. Bafflingly, it sometimes ap-pears that the cloth prevents the feeder fromsupplying liquid, so that a large cavity appearsunder the feeder. The truth is that doubleoxide films plaster themselves against theunderside of the cloth as the feeder fills. Thehalf of the film against the cloth tends to weldto the cloth, possibly by a chemical fusingaction, or possibly by mechanical wrappingaround the fibres of the cloth. Whateverthe mechanism, the action is to hold backthe liquid above, while the lower half of the(unbonded) double film is easily pulled awayby the contracting liquid, opening the voidthat was originally the microscopically thininterface of air inside the bifilm.

5. If there is insufficient opportunity to openinternal defects, the external surface of thecasting may sink to accommodate the inter-nal shrinkage. (The occurrence of surfacesinks is occasionally referred to elsewhere inthe literature as `cavitation'; a misuse oflanguage to be deplored. Cavitation properlyrefers to the creation and collapse of minutebubbles, and the consequent erosion of solidsurfaces such as those of ships' propellers.)

Often, of course, the distribution of defectsobserved in practice is a mixture of theabove list. The internal pressure needs to be

Figure 6.11 Pressure loss situations in castings leading tothe possibility of shrinkage porosity.


Primaryshrinkagepipe

2. Shrinkage cavity due to level (i.e. pressure) criteria

1. Shrinkage cavity due to isolation of liquid


maintained sufficiently high to avoid all of thesedefects.

Finally, however, it is worth pointing outthat over-zealous application of pressure toreduce the above problems can result in a newcrop of different problems.

For instance, in the case of long-freezing-range materials cast in a sand mould, a highinternal pressure, applied for instance to thefeeder, will force liquid out of the surface-linkedcapillaries, making a casting having a `furry'appearance. Overpressures are not easy to con-trol in low-pressure sand casting processes, andare the reason why these processes often strug-gle to meet surface-finish requirements. In castiron castings the generation of excess internalpressure by graphite precipitation has beenshown by work at the University of Alabama(Stefanescu et al. 1996) to lead to exudation ofthe residual liquid via hot spots at the surface ofthe casting, leading to penetration of the sandmould. Later work on steels has shown analo-gous effects (Hayes et al. 1998).

In short-freezing-range materials the inwardflow of solid can be reversed with sufficientinternal pressure. Too great a pressure willexpand the casting, blowing it up like a balloon,producing unsightly swells on flat surfaces(Figure 6.13).

Pressurization of cast iron castings

The successful feeding of cast irons is perhapsthe most complex and challenging feeding taskcompared to all other casting alloys as a result

of the curious and complicating effect of pres-sure. The effects are most dramatically seen forductile irons.

The great prophet of the scientific feeding ofductile irons was Stephen Karsay. In a succes-sion of chattily written books he outlined the

Figure 6.12 The apparent blockingof feed metal by a glass clothstrainer in an Al alloy casting bythe action of bifilms collected onthe underside of the cloth.

Figure 6.13 A comparison between the external size andinternal shrinkage porosity in a casting as a result of(a) moderate pressure in the liquid, and adequately rigidmould, and (b) too much pressure and/or a weak mould.


Feeder

Top half of bifilmmechanically wrapped/weldedto fibres of glass cloth,presenting barrier tofeed metal

Casting

Glass cloth

Unbonded lower halfof bifilm sucked intothe casting, leadingto ‘under-feeder’porosity

Fragments of lower half ofbifilm sucked into dendritemesh, and thereby dispersed

Solidification shrinkage

Liquid shrinkage

Liquid shrinkage

Mould dilation

Solidification shrinkage


principles that applied to this difficult metal(see, for instance, Karsay 1992 and 2000). Hedrew attention to the problem of the swellingof the casting in a weak mould as shown inFigure 6.13, in which the valuable expansion ofthe graphite was lost by enlarging the casting,causing the feeder to be inadequate to fill theincreased volume. He promoted the approach ofmaking the mould more rigid, and so betterwithstanding stress, and at the same time reduc-ing the internal pressure by providing feedersthat acted as pressure relief valves. The feeder,after some initial provision of feed metal duringthe solidification of austenite, would back-fillwith residual liquid during the expansion of thesolidification of the eutectic graphite. The finalstate was a feeder that was substantially sound.(Occasionally, one hears stories that suchsound feeders would be declared to be evidentlyuseless, having apparently provided no feedmetal. However, their removal would immedi-ately cause all subsequent castings to becomeporous!).

The reproducibility of the achievement ofsoundness in ductile iron castings is, of course,highly sensitive to the efficiency of the inocula-tion treatment, because the degree of expansionof graphite is directly affected. This is notor-iously difficult to keep under good control, andmakes for one of the greatest challenges to theiron founder.

Roedter (1986) introduced a refinement ofKarsay's pressure relief technique in which thepressure relief was limited in extent. Some reliefwas allowed, but total relief was prevented bythe premature freezing of the feeder neck. In thisway the casting was slightly pressurized,elastically deforming the very hard sand mould,and the surrounding steel moulding box (if any).The elastic deformation of the mould and itsbox would store the strain energy. The sub-sequent relaxation of this deformation wouldcontinue to apply pressure to the solidifyingcasting during the remainder of solidification.Thus soundness of the casting could beachieved, but without the danger of unac-ceptable swells on extensive flat surfaces.

For somewhat heavier ductile iron castings,however, it has now become common practiceto cast completely without feeders. This hasbeen achieved by the use of rigid moulds, nowmore routinely available from modern green-sand moulding units. Naturally, the swelling ofthe casting still occurs, since, ultimately, solidsare incompressible. However, as before, theexpansion is restrained to the minimum by theelastic yielding of the mould and its container,and distributed more uniformly. Thus the wholecasting is a few per cent larger. If the total net

expansion was 3 volume per cent, this corres-ponds to 1 linear per cent along the threeorthogonal axes, so that from a central datum,each point on the surface of the casting would beapproximately 0.5 per cent oversize. This uni-form and very reproducible degree of oversize isusually negligible. However, of course, it can becompensated, if necessary, by making the pat-tern 0.5 per cent undersize.

The use of the elastic strains to re-applypressure is strictly limited because such strainsare usually limited to only 0.1 linear per cent orso. Thus only a total of perhaps 0.3 volume percent can be compensated by this means. This is,as we have seen above, only a fraction of thetotal volume change that is usual in a graphiticiron, and which permanently affects the size anddimensions of the casting. The judgement offeeder neck sizes to take advantage of such smallmargins is not easy.

With the steady accumulation of experiencein a well-controlled casting facility, the castingengineer can often achieve such an accuracy offeeding that even such a modest gain is con-sidered a valuable asset. Even so, the reader willappreciate that the feeding of graphitic irons isstill not as exact a science and still not as clearlyunderstood as we all might wish.

6.3 The new feeding logic

6.3.1 Background

Much of the formal calculation of feeders hasbeen of poor accuracy because of a number ofsimplifying assumptions that have been widelyused. TiryakiogÆlu has pioneered a new way ofanalysing the physics of feeding, having, inaddition, the good fortune to have as a criticaltest his late father's exemplary experimentaldata on optimum feeder sizes determinedmany years earlier (E. TiryakiogÆlu 1964). Thereader is recommended to the original papersby M. TiryakiogÆlu (1997±2002) for a completedescription of his admirable logic. We shallsummarize his approach only briefly here, fol-lowing closely his excellent description(TiryakiogÆlu et al. 2002).

As we have seen in Rules 2 to 4, an efficientfeeder should (i) remain molten until the portionof the casting being fed has solidified (i.e. thesolidification time of the feeder has to be equalto, or exceed, that of the casting), (ii) containsufficient volume of molten metal to meet thefeeding demand of that same portion of thecasting, (iii) not create a hot spot at the junctionbetween feeder and casting. An optimum feederis then defined as the one with the smallest



volume, for its particular shape, to meet thesecriteria. A feeder that is less compact or that hasless volume than the optimum feeder will resultin an unsound casting.

The standard approaches to solve these prob-lems have usually been based on the famous ruleby Chvorinov (1940) for the solidification timeof a casting:

t � BV

A

8>: 9>;n

(6.10)

where B is the mould constant, V is the castingvolume, A is the surface area through whichheat is lost, and n is a constant (2 in Chvorinov'swork for simple shaped castings in silica sandmoulds). The V/A ratio is known as the modulusm, and has been used as the basis for a numberof approaches to determining the size of feedersfor the production of sound castings, as describedin Section 6.2 Feeding Rule 2.

Despite its wide acceptance, Chvorinov'sRule has limitations because of the underlyingassumptions used in deriving the equation. As aresult of these limitations, the exponent, n,fluctuates between 1 and 2, depending on theshape and size of the casting, and the mould andpouring conditions. One of the reasons for thisanomaly is that Chvorinov's Rule originally didnot take the shape of the casting into con-sideration. A new geometry-based model(TiryakiogÆlu et al. 1997) proved that the mod-ulus includes the effect of both casting shapeand size. These two independent factors wereseparated from each other by the use of a shapefactor k where

t � B 0k1:31V 0:67 (6.11)

and B0 is the mould constant. The shape factor,k, is the ratio (the surface area of a sphere ofsame volume as the casting)/(the surface areaof the casting). In Equation 6.11, V assesses theamount of heat that needs to be dissipated forcomplete solidification, and k assesses the rela-tive ability of the casting shape to dissipate theheat under the given mould conditions.

k � As

A� 4:837V2=3

A(6.12)

where As is the surface area of the sphere.Adams and Taylor (1953) were the first to

consider mass transfer from feeder to casting.They realized that during solidification, a massof aVc needs to be transferred from the feeder tothe casting (a is fraction shrinkage of the metal).However, as TiryakiogÆlu (2002) explains, theirdevelopment of the concept unfortunately

introduced errors so that the final solution wasnot accurate.

Moreover, the lack of knowledge about theeffect of heat transfer between a feeder and acasting has led researchers to the mindset ofconsidering the feeder and casting separately. Inother words, almost all feeder models have beenbased on calculation of solidification times forthe feeder and casting independently, and thenassuming that the same solidification char-acteristics will be followed when they are com-bined. However, it should be remembered thatthe feeder is also a section of the mould cavityand the solidifying metal does not know (orcare) which section is the casting and which isthe feeder.

The objective of the foundry engineer whendesigning a feeding system is to have the thermalcentre of the total casting (the feeder±castingcombination) in the feeder. In fact, all threerequirements for an efficient feeder listed (i) to(iii) above can be summarized as a singlerequirement: the thermal centre of the feeder±casting combination will be in the feeder. Thisnew approach, which treats the casting±feedercombination as a single, total casting, con-stitutes the foundation of TiryakiogÆlu's newapproach to characterize heat and mass transferbetween feeder and casting.

6.3.2 The new approach

Let us consider a plate casting that is fed effec-tively by a feeder. Knowing that the solidifica-tion contraction of the casting is aVc, thisvolume is transferred from the feeder to thecasting, resulting in the final volume of thefeeder being (Vfÿ aVc). Solidification contrac-tion of the feeder is ignored since it does notchange the heat content of the feeder. If thefeeder has been designed according to the rulesfor efficient feeding, the last part to solidify inthis combination is the feeder. In other words,the thermal centre of the casting±feeder combi-nation (the total casting) is in the feeder.Therefore the solidification time of the totalcasting is exactly the same as the feeder, andboth have the same thermal centre.

This is not true for the casting, however. Thethermal centre of the casting is also in the feeder,but its solidification time may or may not beequal to that of the total casting. Hence

k1:31t V0:67

t � k1:31f Vf ÿ aVc� �0:67 (6.13)

where subscript t refers to the total casting. Sofar we have ignored the heat transfer betweenthe casting and the feeder. Using optimumfeeder data obtained by systematic changes



in feeder size for an Al±12wt%Si alloy(TiryakiogÆlu 1964) the solidification times ofcasting and feeder are compared in Figure 6.14aand total casting and feeder in Figure 6.14b.Figure 6.14a shows the (tcÿ tf) relationshipwhen mass transfer is taken into account andheat transfer is ignored. It should be kept inmind that the scatter in Figure 6.14a is not dueto experimental error since all values were cal-culated. Figure 6.14b shows the relationshipbetween the solidification times of feeder andtotal casting (feeder�casting combination).Although the agreement in Figure 6.14b isencouraging, at low values the error is up to30%. However solidification time should beidentical for the feeder and the total casting. Theerror is due to neglect of the heat exchangebetween feeder and casting. Mass is transferredfrom the feeder to the casting throughout thesolidification process. Since solidification takesplace over a temperature range, subtracting aVc

from Vf adjusts for mass exchange completely.For heat exchange however, this treatmentassumes isothermal conditions, and therefore isnot sufficient. Hence the feeder solidificationtime needs to be adjusted for the heat exchangewith the casting. We can treat this heat exchangeas if it were superheat extracted from/given tothe feeder. For the superheat model, we willuse the model by E. TiryakiogÆlu (1964) for itssimplicity and its independence from actualpouring temperature. Equation 6.13 can now berewritten as:

k1:31t V0:67

t � k1:31f Vf ÿ aVc� �0:67exDTf (6.14)

where x is a constant dependent on thealloy (0.0028�Cÿ1 for Al±Si eutectic alloy(TiryakiogÆlu 1964) and 0.0033�Cÿ1 for Al±7%Si(TiryakiogÆlu et al. (1997b)), DTf is the tem-perature change (rise or fall) in feeder because ofthe heat exchange, and can be easily calculatedusing Equation 6.14. The sum of change in heatcontent of the feeder and casting is zero (heatlost by one is gained by the other). Therefore:

C Vf ÿ aVc� �DTf � C Vf � aVc� �DTc � 0

(6.15)

where C is the specific heat of the metal. Hence

DTc � Vf ÿ aVc� �DTf= Vf � aVc� � (6.16)

The total solidification time of the casting cannow be written as

tc � k1:31c Vf � aVc� �0:67exDTc (6.17)

The solidification times of feeder and castingcan now be compared. This comparison is

Figure 6.14 (a) Comparison of calculated solidificationtimes of (a) casting and feeder; (b) total casting andfeeder; (c) total casting (or feeder) versus casting afteradjustment to account for heat transfer between feeder andcasting (TiryakiogÆlu et al. 2002).


0

3000

6000

9000

12000

15000(a)

(b)

(c)

0 3000 6000 9000 12000 15000

tc /B (mm2)

t f/B

(m

m2 )

0

3000

6000

9000

12000

15000

0 3000 6000 9000 12000 15000

tt/B (mm2)

t f/B

(m

m2 )

0

3000

6000

9000

12000

15000

0 3000 6000 9000 12000 15000

tc /B (mm2)

t t/B

(m

m2 )


presented in Figure 6.14c, which shows a prac-tically perfect fit and a relationship that can beexpressed as:

tt � tf � atc (6.18)

The data for Al±Si alloy shown in Figure 6.14cgives a� 1.046.

In a separate exercise, using the data for steelby Bishop and co-workers (1955) assuming x of0.0036�Cÿ1 for steel (TiryakiogÆlu 1964), a simi-lar excellent relationship is obtained where a isfound to be 1.005 (TiryakiogÆlu 2002). Thus thesolidification time of optimum-sized feeders inthe feeder±casting combination was found to beonly a few per cent longer than that of castingsboth for Al±Si alloy and steel castings.

We can conclude that for an accuratedescription of the action of a feeder, both massand heat transfer from feeder to casting duringsolidification have to be taken into accountsimultaneously. Previous feeder models thataccount for mass transfer assume that thetransfer takes place isothermally and at thepouring temperature. This previous assumptionoverestimates the additional heat brought intothe casting from the feeder. The new modelincorporates the effect of superheat and is basedon the equality of solidification times of feederand total casting.

The requirements for efficient feeders: (i)solidification time; (ii) feed metal availability;and (iii) prevention of hot-spot at the junction;can be combined into a single requirement whenthe casting±feeder combination is treated as asingle, total casting. The three criteria reduce tothe simple requirement: `The thermal centre ofthe total casting will be in the feeder'.

The disarming simplicity of this conclusionconceals its powerful logic. It represents theideal criterion for judging the success ofa computer model of a casting and feedercombination.

6.4 Active feeding

Most feeding systems on castings are passive.They work by themselves without outsideintervention. (Even those counter-gravity sys-tems to which pressure is applied to enhancefeeding are not considered to be `active' in thesense discussed below.)

There has recently been introduced a novelsystem of feeding in which a side feeder ispressurized and is thus encouraged to feed uphill(Figure 6.15). This approach to feeding wasdeveloped for use with automatic mouldingwhere moulds could not be inverted through180 degrees after pouring. The concept appearsto have proved useful within the context of fast,automatic greensand moulding for aluminiumalloy castings in a vertically parted mould,where the application of the pressurizing tubescan be automated, and where the casting sizesand weights are limited. Current problems withrepeatability may be the teething problems ofthe new technique that remain to be completelysolved by additional effort.

The counter-gravity-filling of such mouldshas been a considerable advance in terms ofattaining a high quality of metal in the mouldcavity. It has also been useful because the spacetaken by the down sprue is now released asadditional moulding area.

Figure 6.15 Active feeding in a vertically parted automatic greensand moulding machine.


Conventional gravity feeding

+P

Active, pressurized feeding


The provision of small, compact feeders hasbeen a similar benefit, saving mould space,although this is countered to some extent by theneed for a direct line of access for the pressurizingtube to the top of the feeder. Greensand mouldshave been shown to develop useful temperaturegradients in thin-walled aluminium alloy castings(Rasmussen 1995). This natural gradient, theconsequence of a good bottom-gating technique,is exploited by the bottom feeder.

It seems that the danger of convectionreversing these advantages is small if the cast-ings are of limited wall thickness and weight,which is the case for most casting produced onvertically-parted automatic moulding machines.At the time of writing, the limits are not yetknown. Clearly, at some point of increasing wallthickness and casting weight an extendedsolidification time will encourage the develop-ment of convection, and feeding uphill in suchcircumstances will become problematic, if notactually impossible.

Thus although active feeding has beeninvestigated by computer simulation and shownto have attractive advantages, its application tosections of 15 mm thickness in Al alloys (Hansenand Rasmussen 1994) seems likely to be close to,if not actually over, a limit at which convectionwill start to undo the benefits.

6.5 Freezing systems design

In this section on feeding, we are of coursemainly concerned with the action of chills andfins to provide localized cooling of the casting.In this way we can assist directional solidifica-tion of the casting towards the feeder, thusassisting in the achievement of soundness. Thisis one of the important actions of these chillingdevices. It is, however, not the only action, asdiscussed below.

Chills also act to increase the ductility andstrength of that locality of the casting. FromCastings 2003 the proposal is that this occursbecause the faster solidification freezes in thebifilms in their compact form before they havea chance to unfurl. (Recall that the bifilms arecompacted by the extreme bulk turbulenceduring pouring and during their travel throughthe running system. However, they subse-quently unravel, opening up in the mould cavitywhen conditions in the melt become quiet onceagain.)

The interesting corollary of this fact is that ifchills are seen to increase ductility and strengthof a casting, it confirms that the cast material isdefective, containing a high percentage ofbifilms. Another interesting corollary is that if a

casting alloy can be cast without bifilms, chillingshould not increase its properties. This rathersurprising prediction is fascinating, and, if true,indicates the huge potential for the increase ofthe properties of cast alloys. All castings with-out bifilms are therefore predicted to haveextraordinary ductility and strength. It alsoexplains our lamentable current condition inwhich most of us constantly struggle in ourfoundries to achieve minimum mechanicalproperties for castings. Some days we win, otherdays we continue to struggle. The message isclear, we need to focus on technologies for theproduction of castings with reduced bifilmcontent, preferably zero bifilm content. Therewards are huge.

Another action of chills is to straightenbifilms. This action occurs because the advancingdendrites cannot grow through the air layerbetween the double films, and so push the bifilmsahead. Those that are somehow attached to thewall will be partially pushed, straightened andunravelled by the gentle advance of grains. Thiseffect is reported in Castings (2003). Thusalthough a large percentage of bifilms will bepushed ahead of the chilled region, concentrating(and probably reducing the properties) in theregion immediately ahead, some bifilms willremain aligned in the dendrite growth direction,and so be largely perpendicular to the mould wall.

The overall effects on mechanical propertiesof the pushing action are not so easily predicted.The reduction in density of defects by the chillwill raise properties, but the presence of occa-sional bifilms aligned at right angles to thesurface of the casting would be expected to beseverely detrimental. These complicated effectsrequire to be researched. However, we canspeculate that they seem likely to be the cause oftroublesome edge cracking in the rolling of castmaterials of many types, leading to the expenseof machining off the surface of many alloysbefore rolling can be attempted. The superbformability of electroslag remelted compared tovacuum arc remelted alloys is almost certainlyexplained in this way. The ESR process pro-duces an extremely clean material because oxidefilms will be dissolved during remelting underthe layer of liquid slag, and will not re-form inthe solidifying ingot. In contrast, the relativelypoor vacuum of the VAR process ensures thatthe lapping of the melt over the liquid meniscusat the mould wall will create excellent doubleoxide films. If considerable depths of the surfaceare not first removed `oxide lap defects' will openas surface cracks when subjected to forging orrolling.

Although, as outlined above, the chillingaction of chills and fins is perhaps more



complicated than we first thought, the chillingaction itself on the rate of solidification is welldocumented and understood. It is this thermalaspect of their behaviour that is the subject ofthe remainder of this section.

6.5.1 External chills

In a sand mould the placing of a block of metaladjacent to the pattern, and subsequentlypacking the sand around it to make the rest ofthe mould in the normal way, is a widely usedmethod of applying localized cooling to thatpart of the casting. A similar procedure can beadopted in gravity and low-pressure die-castingby removing the die coat locally to enhance thelocal cooling rate. In addition, in dies of alltypes, this effect can be enhanced by the inser-tion of metallic inserts into the die to providelocal cooling, especially if the die insert is highlyconductive (such as made from copper alloy)and/or artificially cooled, for instance by air, oilor water.

Such chills placed as part of the mould, andthat act against the outside surface of the castingare strictly known as external chills, to distin-guish them from internal chills that are cast in,and become integral with, the casting.

In general terms, the ability of the mould toabsorb heat is assessed by its heat diffusivity.This is defined as (KrC)1/2 where K is thethermal conductivity, r the density, and C thespecific heat of the mould. It has complex unitsJ mÿ 2 Kÿ1 sÿ1/2. (Take care not to confuse withthermal diffusivity defined as K/rC, and nor-mally quoted in units of m2sÿ1.) From the roomtemperature data in Table 6.3 we can obtainsome comparative data on the chilling power ofvarious mould and chill materials, shown inTable 6.4.

It is clear that the various refractory mouldmaterialsÐsand, investment and plasterÐareall poor absorbers of heat, and become worse inthat order. The various chill materials are all ina league of their own, having chilling powersorders of magnitude higher than the refractorymould materials. They improve marginally,

Table 6.3 Mould and metal constants

Material Meltingpoint(�C)

Liquid±solidcontraction

Specific heat(J kgÿ1Kÿ1)

Density(kg mÿ3)

Thermal conductivity(Jm Kÿ1 sÿ1)

(%) Solid Liquidm.p.

Solid Liquidm.p.

Solid Liquidm.p.

20�C m.p. 20�C m.p. 20�C m.p.

Pb 327 3.22 130 (138) 152 11680 11020 10678 39.4 (29.4) 15.4Zn 420 4.08 394 (443) 481 7140 (6843) 6575 119 95 9.5Mg 650 4.2 1038 (1300) 1360 1740 (1657) 1590 155 (90)? 78Al 660 7.14 917 (1200) 1080 2700 (2550) 2385 238 ± 94Cu 1084 5.30 386 (480) 495 8960 8382 8000 397 (235) 166Fe 1536 3.16 456 (1130) 795 7870 7265 7015 73 (14)? ±Graphite ± ± 1515 ± ± 2200 ± ± 147Silica sand ± ± 1130 ± ± 1500 ± ± 0.0061 ±Investment(Mullite) 750 ± ± 1600 ± ± 0.0038 ± ±Plaster ± ± 840 ± ± 1100 ± ± 0.0035 ± ±

References: Wray (1976); Brandes (1991); Flemings (1974)

Table 6.4 Thermal properties of mould and chill materials at approximately 20�C

Material Heat Diffusivity(K�C)1/2

(Jmÿ2 Kÿ1sÿ1/2)

Thermal DiffusivityK/�C(m2sÿ1)

Heat Capacityper unit volume�C (JKÿ1mÿ3)

Silica sand 3.21� 103 3.60� 10ÿ9 1.70� 106

Investment 2.12� 103 3.17� 10ÿ9 1.20� 106

Plaster 1.8� 103 3.79� 10ÿ9 0.92� 106

Iron (pure Fe) 16.2� 103 20.3� 10ÿ6 3.94� 106

Graphite 22.1� 103 44.1� 10ÿ6 3.33� 106

Aluminium 24.3� 103 96.1� 10ÿ6 2.48� 106

Copper 37.0� 103 114.8� 10ÿ6 3.60� 106



within a mere factor of 5, in the order steel,graphite, and copper.

The heat diffusivity value indicates the actionof the material to absorb heat when it is infi-nitely thick, i.e. as would be reasonably wellapproximated by constructing a thick-walledmould from such material. When a relativelysmall lump of cast iron or graphite is used as anexternal chill in a sand mould, it does notdevelop its full potential for chilling as indicatedby the heat diffusivity because it has limitedcapacity for heat.

Thus although the initial rate of freezing of ametal may be in the order given by the abovelist, for a chill of limited thickness its coolingeffect is limited because it becomes saturatedwith heat; after a time it can absorb no more.The amount of heat that it can absorb is definedas its heat capacity. We can formulate the usefulconcept of volumetric heat capacity in termsof its volume V, its density r and its specificheat C:

Volumetric heat capacity � VrC

In the SI system its units are J Kÿ1. Theresults by Rao and Panchanathan (1973) on thecasting of 50 mm thick plates in Al±5Si±3Cu

reveals that the casting is insensitive to whetherit is cooled by steel, graphite or copper chills,provided that the volumetric heat capacity ofthe chill is taken into account.

These authors show that for a steel chill25 mm thick its heat capacity is 900 J Kÿ1. Achill with identical capacity in copper would berequired to be 32 mm thick, and in graphite36 mm. These values led the author to conclude(Castings 1991) that copper may therefore notalways be the best chill material. However, usingsomewhat more accurate data, copper is found,after all, to be best. These results are presentedin Figure 6.16 showing the relative heat capa-cities and diffusivities. It is clear that for similarthicknesses of a block chill, copper is alwaysmost effective whether limited by heat diffusiv-ity or heat capacity.

Figure 6.17 illustrates that the chills areeffective over a considerable distance, the lar-gest chills greatly influencing the solidificationtime of the casting even up to 200 mm (fourtimes the section thickness of the casting)distant. This large distance is perhaps typical ofsuch a thick-section casting in an alloy of highthermal conductivity, providing excellent heattransfer along the casting. A steel casting wouldrespond less at this distance.

Figure 6.16 Relative diffusivities(ability to diffuse heat away if alarge chill) and heat capacities(ability to absorb heat if relativelysmall) of chill materials.


Cu

1500

1000

500

0

4

3

2

1

0

Hea

t cap

acity

MJ/

k–1 m

–3

Hea

t diff

usiv

ity M

J2 m

–4 k

–2 s

–1

Finite chills(heat capacity limited)

Infinite chills(heat diffusivity limited)

Graphite

Fe

Al

SandInvestment

Plaster


The work by Rao and Panchanathan revealsthe widespread sloppiness of much presentpractice on the chilling of castings. Generalexperience of the chills generally used in foun-drywork nowadays shows that chill size andweight are rarely specified, and that chills are ingeneral too small to be fully effective in anyparticular job. It clearly matters what size ofchill is added.

Computational studies by Lewis and collea-gues (2002) have shown that the number, sizeand location of chills can be optimized by com-puter. These studies are among the welcome firststeps towards the intelligent use of computersin casting technology.

Finally, in detail, the action of the chill is noteasy to understand. The surface of the castingagainst the chill will often contract, distortingaway and thus opening up an air gap. Thechilled casting surface may then reheat to suchan extent that the surface remelts. The exuda-tion of eutectic is often seen between the castingand the chill (Figure 6.18). The new contactbetween the eutectic and chill probably thenstarts a new burst of heat transfer and thus rapidsolidification of the casting. Thus the history ofcooling in the neighbourhood of a chill may be asuccession of stop/start, or slow/fast events.

6.5.2 Internal chills

The placing of chills inside the mould cavitywith the intention of casting them in place is aneffective way of localized cooling. The simple

method of mixtures approach (Campbell andCaton 1977) indicates that to cool superheatedpure liquid iron to its freezing point, and freez-ing a proportion of it, will require various levelsof addition of cold, solid iron depending on theextent that the added material is allowed to melt(Table 6.5). These calculations take no accountof other heat losses from the casting. Thus fornormal castings the predictions are likely to beincorrect by up to a factor of 2. This is broadlyconfirmed by Miles (1956), who top-pouredsteel into dry sand moulds 75 mm square and300 mm tall. In the centre of the moulds waspositioned a variety of steel bars ranging from12.5 mm round to 25 mm square, covering arange of chilling from 2 to 11 per cent solidaddition. His findings reveal that the 2 per centsolid addition nearly melted, compared to thepredicted value for complete melting of 3.5 percent solid. The 11 per cent solid addition causedextensive (possibly total) freezing of the castingjudging by the appearance of the radial grain

Figure 6.17 Freezing time of a plate 225� 150� 50 mm inAl±5Si±3Cu alloy at various distances from the chilled endis seen to decrease steadily as the chill is approached, andas the chill size is increased (Rao and Panchanathan1973).

Figure 6.18 Al±Si eutectic liquid segregation byexudation at a chilled interface of an Al±Si alloy.



structure in the macrosections. He found 5 percent addition to be near optimum; it had areasonable chilling effectiveness but causedrelatively few defects.

In the case of the higher additions, where theheat input is not sufficient to melt the chill, thefusing of the surface into the casting has to bethe result of a kind of diffusion bonding process.This would emphasize the need for cleanness ofthe surface, requiring the minimum presence ofoxide films or other debris against the chillduring the filling of the mould. If Miles had useda better bottom-gated filling technique he mayhave reduced the observed filling defects fur-ther, and found that higher percentages werepractical.

The work by Miles does illustrate the prob-lems generally experienced with internal chills.If the chills remain for any length of time in themould, particularly after it is closed, and moreparticularly if closed overnight, then condensa-tion is likely to occur on the chill, and blowdefects will be caused in the casting. Blows arealso common from rust spots or other impuritieson the chill such as oil or grease. The matchingof the chemical composition of the chill and thecasting is also important; mild steel chills will,for instance, usually be unacceptable in an alloysteel casting.

Internal chills in aluminium alloy castingshave not generally been used, almost certainlyas a consequence of the difficulty introduced bythe presence of the oxide film on the chill. Thisappears to be confirmed by the work of Biswaset al. (1985), who found that at 3.5 per cent byvolume of chill and at superheats of only 35�Cthe chill was only partially melted and retainedpart of its original shape. It seems that over thisarea it was poorly bonded. At superheats above75�C, or at only 1.5 per cent by volume, the chillwas more extensively melted, and was useful inreducing internal porosity and in raisingmechanical properties. The lingering presence ofthe oxide film from the chill remains a concernhowever.

The development of a good bond between theinternal chill and the casting is a familiar prob-lem with the use of chapletsÐthe metal devicesused to support cores against sagging because ofweight, or floating because of buoyancy. A onepage review of chaplets is given by Bex (1991).To facilitate the bond for a steel chaplet in aniron or steel casting the chaplet is often platedwith tin. The tin serves to prevent the formationof rust, and its low melting point (232�C) andsolubility in iron assists the bonding process.

The bond between steel and titanium insertsin Al alloy castings has been investigated inJapan (Noguchi et al. 2001) who found only a10 mm silver coating was effective to achieve agood bond, although even this took up to5 minutes to develop at the Al±Ag eutectictemperature 566�C. Attempts to achieve a bondwith gold plating and Al±Si sprayed alloy werelargely unsuccessful.

It seems, therefore, that internal chills inaluminium alloys might be tolerable to tackleporosity problems in castings that are difficultto tackle by other techniques. However, theoxide film remains an ever-present danger. Itwill persist as a double film (having acquired itssecond layer during the immersion of the chill)and so pose the risk of leakage or crack for-mation. Such risks are only acceptable for lowduty products.

Brown and Rastall (1986) use the non-bonding of heavier aluminium inserts in alumi-nium castings to advantage. They use a castaluminium alloy core inside an aluminium alloycasting to form re-entrant details that could noteasily be provided in a pressure die cast product.Also, of course, because the freezing time isshortened, productivity is enhanced. The inter-nal core is subsequently removed by dis-assembly or part machining, or by mechanicaldeformation of the core or casting.

6.5.3 Fins

Before we look specifically at fins on castings, itis worth spending some time to consider theconcepts involved in junctions of all typesbetween different cast sections. Figure 2.34shows a complete range of T-junctions betweenwalls of different relative thickness. When thewall forming the upright of the T is thin, it actsas a cooling fin, chilling the junction and theadjacent wall (the top cross of the T) of thecasting. We shall return to a more detailedconsideration of fins shortly.

When the upright of the T-section hasincreased to a thickness of half the casting sec-tion thickness, then the junction is close tothermal balance, the cooling effect of the fin

Table 6.5 Weight percentage of internal chills in purecast iron

Calculatedaddition (%)(Campbelland Caton1977)

Observedaddition (%)(Miles, 1956)

Result

3.0 Chill completely melted3.5 '2 Chill reaches melting

point, but does not melt7.0 50% of melt is solidified10.5 '11 100% of melt is solidified



balancing the hot-spot effect of the concentra-tion of metal in the junction. Kotschi and Loper(1974) were among the first to evaluate junc-tions and highlighted this special case.

By the time that the upright of the T hasbecome equal to the casting section, the junctionis a hot spot. This is common in castings.Foundry engineers are generally aware the 1 : 1T-junction is a problem. It is curious thereforethat castings with even wall thickness are said tobe preferred, and that designers are encouragedto design them. Such products necessarily con-tain 1 : 1 junctions that will be hot spots. How-ever, because the 1 : 1 thickness junction is suchan intractable problem, Mertz and Heine (1973)suggest that it should be fed from its end, alongits length, and thereby used as a feeding path. Infact, they go further and recommend generousconvex radii for the fillets and the planting of apad on the far side of the T to maximize thefeeding distance along the junction, as illu-strated in Figure 6.19.

When finally the section thickness of theupright of the T is twice the casting section, thenthe junction is balanced once again, with thecasting now acting as the mild chill to counterthe effect of the hot spot at the junction. Wehave considered these junctions merely in theform of the intersections of plates. However, wecan extend the concept to more general shapes,introducing the use of the geometric modulusm� (volume)/(cooling area). It subsequentlyfollows that an additional requirement when afeeder forms a T-junction on a casting is that thefeeder must have a modulus two times themodulus of the casting. The hot spot is thenmoved out of the junction and into the feeder,with the result that the casting is sound. This isthe basis behind Rule 4 for feeding discussed inSection 6.2.

Pellini (1953) was one of the first experi-menters to show that the siting of a thin `para-sitic' plate on the end of a larger plate couldimprove the temperature gradient in the largerplate. However, the parasitic plate that he usedwas rather thick, and his experiments were car-ried out only on steel, whose conductivity ispoor, reducing useful benefits.

Figure 6.20 shows the results from Kim et al.(1985) of pour-out tests carried out on 99.9%pure aluminium cast into sand moulds. Thefaster advance of the freezing front adjacent tothe junction with the fin is clearly shown. (As anaside, this simple result is a good test of somecomputer simulation packages. The simulationof a brick-shaped casting with a cast-on finshould show the cooling effect by the fin. Somerelatively poor computer algorithms do nottake into account the conduction of heat in the

casting, thus predicting, erroneously, theappearance of the junction as a hot spot.)

Creese and Sarfaraz (1987) demonstrate theuse of a fin to chill a hot spot in pure Al castings

Figure 6.19 (a) T-junction with normal concave filletradius; (b) marginal improvement to the feed path alongthe junction; (c) convex fillets plus pad that doublesfeeding distance along the junction; and (d) practicalutilization of a T-junction as a feed path (Mertz and Heine1973).

Figure 6.20 A T-junction casting in 99.9 Al byKim et al. (1985) showing successive positions of thefreezing front.



that were difficult to access in other ways. Theycast on fins to T- and L-junctions as shown inFigure 6.21. The reduction in porosity achievedby this technique is shown in Figure 6.22. Forthese casting sections of 50 mm there was noapparent difference between fins of 2.5 and3.3 mm thickness so these results are treatedtogether in this figure. These fins at 5 and 6 percent of the casting section happen to be close tooptimum as is confirmed later below. The rea-son that they conduct away perhaps less effec-tively than might be expected is because of theirunfavourable location at 45 degrees to two hotcomponents of the junction.

Returning to the case where the upright of theT is sufficiently thin to act as a cooling fin, onefurther case that is not presented in Figure 2.34is the case where the fin is so thin that it does notexist. This, you will say, is a trivial case. Butthink what it tells us. It proves that the fin canbe too thin to be effective, since it will haveinsufficient area to carry away enough heat.Thus there is an optimum thickness of fin for agiven casting section.

Similarly, an identical argument can be madeabout the fin length. A fin of zero length will

have zero effect. As length increases, effective-ness will increase, but beyond a certain length,additional length will be of reducing value. Thusthe length of fins will also have an optimum.

These questions have been addressed in apreliminary study by Wright and Campbell(1997) on a horizontal plate casting with asymmetrical fin (Figure 6.23). Symmetry waschosen so that thermocouple measurementscould be taken along the centreline (otherwisethe precise thermal centre was not known sothat the true extension in freezing time may nothave been measured accurately). In addition thehorizontal orientation of the plate was selectedto suppress any complicating effects of convec-tion so far as possible. The thickness of the finwas B.H and the length L.H where B and L aredimensionless numbers to quantify the fin interms of H, the thickness of the plate. From thisstudy it was discovered that there was an opti-mum thickness of a fin, and this was less thanone tenth of H. Figure 6.23a interpolates anoptimum in the region of 5 per cent of thecasting section thickness. The optimum lengthwas 2H, and longer lengths were not effective(Figure 6.23b). For these conditions the freezing

Figure 6.21 T- and L-junctions in pure aluminium cast inoil-bonded greensand. The shape of porosity in thesejunctions is shown, and the region of the junction used tocalculate the percentage porosity is shown by the brokenlines. The position of fins added to eliminate the porosity isshown. Results are presented in Figure 6.22.

Figure 6.22 Results from Creese and Sarfaraz (1987,1988) showing the reduction in porosity as a result ofincreasing length of fins applied as in Figure 6.21.



time of the casting was increased by approxi-mately ten times. Thus the effect is useful.However, the effect is also rather localized,so that it needs to be used with caution.

Eventually, non-symmetrical results for a chillon one side of the plate would be welcome.

Even so, the practical benefits to the use of afin as opposed to a chill are interesting, even

Figure 6.23 The effect of a symmetrical fin on the freezing time at the centre of a cast plate of 99.9Al alloy as a functionof the length and thickness of the fin (averaged results of simulation and experiment from Wright and Campbell 1997).


1.2

1.0

0.8

0.6

0.4

0.2

00 0.1 0.2 0.3 0.4 0.5 0.6 0.7

99.9 Al

L /H

L

H

B

1

2

4

Thermocouple

100

100

Fin thickness B /H

Rel

ativ

e so

lidifi

catio

n tim

e t f

/t o

1.2

1.0

0.8

0.6

0.4

0.2

00 1 2 3 4

Fin length L /H

Rel

ativ

e so

lidifi

catio

n tim

e t f

/t o

B /H0.75

0.5

0.4

0.2

0.1

0.05

(a)

(b)


compelling. They are:

1. The fin is always provided on the casting,because it is an integral part of the tooling.Thus, unlike a chill, the placing of it cannotbe forgotten.

2. It is always exactly in the correct place. Itcannot be wrongly sited before the making ofthe mould. (The incorrect positioning of achill is easily appreciated, because althoughthe location of the chill is normally carefullypainted on the pattern, the application ofthe first coat of mould release agent usuallydoes an effective job in eliminating all tracesof this.)

3. It cannot be displaced or lifted during themaking of the mould. If the chill liftsslightly during the filling of the tooling withsand the resulting sand penetration under theedges of the chill, and the casting of addi-tional metal into the roughly shaped gap,make an unsightly local mess of the castingsurface. Displacement or complete fallingout from the mould is a common danger,sometimes requiring studs to support the chillif awkwardly angled or on a vertical face.Displacement commonly results in sandinclusion defects around the chill or can addto defects elsewhere. All this is expensive todress off.

Figure 6.24 A comparison of the action of chills and cooling fins in aluminium bronze alloy AB1 (Wen, Jolly andCampbell 1997).


500

H=

50

L

B

B

L

Fin

500

H=

50

Chill

Thermocouple

0.02

0.040.12

0.2

0.4

1

2

4

0.2

0.08

0.12

FIN CHILL

B /H B /H1.0

0.8

0.6

0.4

0.2

00 0.1 0.5 1 5 10

Rel

ativ

e so

lidifi

catio

n tim

e t f

/t o

Length L /H

(a)

(c)

(b)


4. An increase in productivity has been reportedas a result of not having to find, place andcarefully tuck in a block chill into a sandmould (Dimmick 2001).

5. It is easily cut off. In contrast, the witnessfrom a chill also usually requires substantialdressing, especially if the chill was equippedwith v-grooves, or if it became misplacedduring moulding, as mentioned above.

6. The fin does not cause scrap castings becauseof condensation of moisture and othervolatiles, with consequential blow defects,as is a real danger from chills.

7. The fin does not require to be retrieved fromthe sand system, cleaned by shot blasting,stored in special bins, re-located, counted,losses made up by re-ordering new chills,casting new chills (particularly if the chill isshaped) and finally ensuring that the correctnumber in good condition, re-coated, anddried, is delivered to the moulder on therequired date.

8. The fin does not wear out. Old chills becomerounded to the point that they are effectivelyworn out. In addition, in iron and steelfoundries, grey iron chills are said to `losetheir nature' after some use. This seems to bethe result of the oxidation of the graphiteflakes in the iron, thus impairing the thermalconductivity of the chill.

9. Sometimes it is possible to solve a localizedfeeding problem (the typical example is theisolated boss in the centre of the plate) bychilling with a fin instead of providing a localsupply of feed metal. In this case the fin isenormously cheaper than the feeder.

This lengthy list represents considerable costsattached to the use of chills that are not easilyaccounted for, so that the real cost of chills isoften underestimated.

Even so, the chill may be the correct choicefor technical reasons. Fins perform poorly formetals of low thermal conductivity such as zinc,Al-bronze, iron and steel. The computer simu-lation result in Figure 6.24 illustrates for therather low thermal conductivity material, Al-bronze, that there are extensive conditions inwhich the chill is far more effective.

The kind of result shown in Figure 6.24would be valuable if available for a variety ofcasting alloys varying from high to low thermalconductivity, so that an informed choice couldbe made whether a chill or fin was best in anyparticular case. These results have yet to beworked out and published.

Fins are most easily provided on a joint lineof the mould, or around core prints. Sometimes,however, there is no alternative but to mould

them at right angles to the joint. From a prac-tical point of view, these upstanding fins onpatternwork are of course vulnerable todamage. Dimmick (2001) records that fins madefrom flexible and tough vinyl plastic solved thedamage problem in their foundry. They wouldcarry out an initial trial with fins glued onto thepattern. If successful, the fins would then bepermanently inserted into the pattern. In addi-tion, only a few standard fins were found to besatisfactory for a wide range of patterns; a fairlywide deviation from the optimum ratios did notseem to be a problem in practice.

Sarfaraz and Creese (1989) investigated aninteresting variant of the cast-on fin. Theyapplied metal fins to the pattern, and rammedthem up in the sand as though applying a nor-mal external chill, in the manner shown inFigure 6.21. The results of these `solid' or `cold'fins (so called to distinguish them from theempty cavity that would, after filling with liquidmetal, effectively constitute a `cast' or `hot' fin)are also presented in Figure 6.22. It is seen thatthe cold fins are more effective than the cast finsin reducing the porosity in the junction castings.This is the consequence of the heat capacity ofthe fin being used in addition to its conductingrole. It is noteworthy that this effect clearlyoverrides the problem of heat transfer across thecasting/chill interface.

The cold fin is, of course, really a chill of ratherslim shape. It raises the interesting question, thatas the geometry of the fin and the chill is varied,which can be the most effective. This questionhas been tackled in the author's laboratory (Wenand colleagues 1997) by computer simulation.The results are summarized in Figure 6.24.Clearly, if the cast fin is sufficiently thin, it ismore effective than a thin chill. However, fornormal chills that occupy a large area of thecasting (effectively approaching an `infinite' chillas shown in the figure), as opposed to a slimcontact line, the chill is massively more effectivein speeding the freezing of the casting.

Other interesting lessons to be learned fromFigure 6.24 are that a chill has to be at leastequal to the section thickness of the casting to bereally effective. A chill of thickness up to twicethe casting section is progressively more valu-able. However, beyond twice the thickness,increasingly thick chills show progressivelyreducing benefit.

It is to be expected that in alloys of higherthermal conductivity than aluminium bronze, afigure such as Figure 6.24 would show a greaterregime of importance for fins compared tochills. The exploration of these effects for avariety of materials would be instructive andremains as a task for the future.



The business of getting the heat away from thecasting as quickly as possible is taken to a logicalextreme by Czech workers (Kunes et al. 1990) whoshow that a heat pipe can be extremely effectivefor a steel casting. Canadian workers (Zhang et al.2003) explore the benefits of heat pipes for

aluminium alloys. The conditions for successfulapplication of the principle are not easy, however,so I find myself reluctant at this stage to recom-mend the heat pipe as a general purpose techniquein competition to fins or chills. In special circum-stances, however, it could be ideal.



Rule 7

Avoid convection damage

7.1 Convection: the academicbackground

Convection is the flow phenomenon that arisesas a result of density differences in a fluid.

In a solidifying casting the density differencesin the residual liquid can be the result of dif-ferences in solute content as a consequence ofsegregation. This is a significant driving forcefor the development of channel defects knownas the `A' and `V' segregates in steel ingots andas freckle trails in nickel- and cobalt-baseinvestment castings. The name `freckles' comesfrom the appearance of the etched componentsthat shows randomly oriented grains in thechannels that have been partly remelted in theconvecting flow and detached from their origi-nal dendrites. These defects are discussed earlierin Castings (2003) and are not discussed furtherhere. Although, for many reasons, channeldefects are unwelcome, they are usually not lifethreatening to the product.

Convection can also arise as a result ofdensity differences that result from temperaturedifferences in the melt. There have been num-erous theoretical studies of the solidification oflow melting point materials in simple cubicalmoulds, of which one side is cooled and theother not. The resulting gentle drift of liquidaround the cavity, down the cool face and upthe non-cooled face, changes the form of thesolidifying front. A schematic example is shownin Figure 7.1. These are interesting exercises, butgive relatively little assistance to the under-standing of the problems of convective flow inengineering systems.

The results due to Mampaey and Xu (1999)who studied the natural convection in anupright cylinder of solidifying cast iron showed

that the thermal centre of the liquid masswas shifted upwards, and graphite nodules inspheroidal graphite irons were transported bythe flow. Such studies reflect the gentle action ofconvection in small, simple shaped, closed sys-tems; the kind of action one would expect to seein a cooling cup of tea. These facts have lulled usinto a state of false security, assuming convec-tion to be essentially harmless and irrelevant.We need to think again.

7.2 Convection: the engineeringimperatives

Convection was practically unknown as animportant factor in shaped castings until theearly 1980s. Even now, it is not widely knownnor understood. However, it can be life anddeath to a casting, and has been the death of anumber of attempts to develop counter-gravitycasting systems around the world. Most workersin this endeavour still do not know why theyfailed. The Cosworth Casting Process nearlyfoundered on this problem in its early days, onlysolving the problem by its famous (infamous?)roll-over system.

Thus convection is not merely a textbookcuriosity. The casting engineer requires to cometo terms with convection as a matter of urgency.The problem can be of awesome importance,and can lead to major difficulties, if notimpossibilities, to achieve a sound and saleablecasting.

Convection enhances the problems of uphillfeeding in medium section castings, makingthem extremely resistant to solution. In factincreasing the amount of (uphill) feeding byincreasing the diameter of the feeder neck, for


instance, makes the feeding problem worseby increasing the opportunity for convection.Many of the current problems of low-pressurecasting systems derive from this source.

In contrast, having feeders at the top of thecasting, and feeding downwards under gravity iscompletely stable and predictable, and givesreliable results.

The instability of convective problems isworth emphasizing. Because the heavy, coolliquid overlays the hotter less dense liquid, thesituation is metastable. If the stratified layers ofliquid are not disturbed there is a chance thatthe heavy liquid will remain wobbling aroundon the top, and may solidify in place withoutincident. However, a small disturbance mayupset the delicate balancing act. Once started,the cold melt will slip sideways, plungingdownwards to the bottom, and the hot liquidsurge upwards, so that a convective circulationwill quickly establish. In practice therefore, anumber of castings may be made successfullyif the metastable equilibrium is not disturbed,

but, inexplicably, the next may exhibit massiveremelted and unfed regions.

Triggers to initiate the unstable flow couldarise from many different kinds of uncontrolledevents. A significant trigger could be an eventsuch as the rising of a bubble from a core blow,as a result of an occasionally ill-fitting of a coreprint, leading to the chance sealing of the corevent by liquid metal.

Momchilov (1993) gives one of the very fewaccounts of the exasperating randomness ofconvection problems. He found that with the useof two riser tubes from a furnace containingliquid metal into one die cavity, successive cast-ings could be observed to have completely dif-ferent internal temperature histories. The firstcasting might be fine. However, the subsequentcasting would suffer a die temperature inexplic-ably overheating by 120�C and the temperaturein the furnace simultaneously dropping by 65�C.These are powerful and important exchanges ofheat between the die and the crucible below.These changes caused the second casting to bepartially remelted.

The use of twin riser tubes by Momchilovraises an important feature of convection.Convective flows require to be continuous, as ina circulation. Thus in the case of two riser tubesinto one die cavity, the conditions for a circularflow, up one tube and down the other, are ideal.It is likely that Momchilov would have solvedhis problem, or at least greatly reduced it, sim-ply by blocking off one of the tubes.

The elimination of ingates in this way to solveconvection problems in counter-gravity fedcastings should be considered as a standard firststep. This was found to be a useful measure inthe early days of the Cosworth Process when itoperated merely as a static low-pressure castingprocess. (The later development of the roll-overconcept represented a welcome total solution.)

The only other description of the problems ofconvection ever discovered by the author comesfrom a patent by Rogers and Heathcock (1990).They fall foul of convection during the attemptto make an aluminium alloy cylinder blockcasting in a counter-gravity filled permanentmould. They found that as the mould heated upthe problem became worse, and the rate offlow of the convection currents increased. Themicrostructure of the casting was unacceptablein the area affected by convection. They dealtwith the problem by providing strong coolingjust above the ingates. This solution clearlythreatened the provision of feed metal while thecasting was solidifying, and so was a riskystrategy. There is no record that the patent wasever implemented in production. Perhaps con-vection secured another victim.

Solid

Solid

Cooledsideof mould

Cooledsideof mould

Convectingliquid

Liquid(no convection)

(b)

(a)

Figure 7.1 Solidification in 2-D box, of which only theright-hand side is cooled. (a) planar front in the case ofno convection; (b) the distortion caused by convective flow.



Castings that employ a third mould part tosite the running system under the casting are atrisk of convective effects causing the melt tocirculate up some ingates and down others viathe casting above and the runner underneath.This is especially dangerous if the runner is aheavy section. Pressurizing the runner with anadequate feeder is a way of maintaining the netupward movement of metal required for thefeeding of the casting, thus reducing the dele-terious effects of the convection to merely thatof delaying freezing. In this case the worst thathappens is the development of a locally coarserstructure.

Investment casting often provides numerousconvective loops in wax assemblies, as a result ofattaching the wax patterns at more than onepoint to increase the strength of the completewax assembly. A typical wax assembly for thecasting of polycrystalline Ni-base turbine bladesis illustrated in Figure 7.2a. The central uprightis surrounded by six blades (only two are shownin the section), so that in addition to its heavysection designed to act as a feeder, it is kept evenhotter by the presence of the surrounding bladecastings that prevent loss of heat by radiation.Conversely, of course, the blades cool quicklybecause they can radiate heat freely to the coolsurroundings. A convective loop is therefore setup, with hot metal rising up the central feeder,and falling through the cooling castings. Thefinal grain structure seen on the etched compo-nent reveals the path of the flow (Figure 7.2b).The casting is designed to have fine surfacegrains nucleated by the cobalt aluminate addi-tion to the primary coat of the mould. However,because additional hot metal enters the mouldcavity from the top after the chill grains areformed, the original chill grains are remelted.The convective flow sweeps down through thecasting, becoming a concentrated channel as itexits the base of the blade. The very narrowsection of the trailing edge of the casting is notpenetrated, and so escapes remelting, as does thelarge region in the bottom right that the flowhas missed.

Very large blades for the massive land-basedturbines for power generation are sometimescast horizontally. In this case each end of thecasting is subject to convective problems as isseen in Figure 7.3.

The cutting of convective links in waxassemblies is recommended, and cries out forwide attention in most current investmentcasting operations. The strengthening of waxassemblies by wax links inadvertently providesconvective links and should be avoided. Ceramicrods can provide strengthening, or, if wax con-nections are used, they should be plugged with a

Figure 7.2 (a) Lost wax assembly of six Ni-base turbineblades around a central feeder, showing the expectedconvective loops; (b) an etched blade, showing theremelting of the fine surface grains created by thecobalt-aluminate nucleant in the mould surface, and thesubsequent growth of coarse grains that define the flowpath.

Rule 7. Avoid convection damage 159

(a)

(b)

100

50

0

mm


ceramic disc to avoid metal flow. These simplemodifications to the wax assembly will com-pletely change the mode of solidification of thecastings, allowing for the first time an accurateunderstanding of filling and feeding effects.

Other problems in sand castings are illus-trated in Figure 7.4. Gravity die (permanentmould) castings are less prone to these problemsbecause of their more rapid rate of heatextraction by the metal mould. For castings inmetal moulds the sections have to be con-siderably larger before convection starts to be athreat. The aspect of the relative times for soli-dification and convection damage are dealt within more detail in the section below on castingsection thickness.

It is evident that many computer predictionsof heat flow and the feeding of castings will bequite inadequate to deal with convection prob-lems, since it is usual to consider the lossof heat from castings simply by conduction.Clearly, thicker sections in a loop will cool morequickly than the computer would predict, sinceconvection allows them to export their heat.Conversely of course, thin sections in the sameloop will suffer the arrival of additional heatthat will greatly delay their solidification. Infact, if the hot section has an independent sourceof heating, such as the electrical heating pro-vided in many counter-gravity systems, thesections in the loop can circulate for ever.The computer would have particular difficultywith this.

Even so, the greater speed and sophisticationof computing will eventually provide the pre-dictions containing the contribution of convec-tion that are so badly needed. It is hoped thatfuture writers and founders will not need tolament our poor abilities in this area.

7.3 Convection damage and castingsection thickness

If the solidification time of the casting is similarto the time taken for convection to becomeestablished, extensive remelting can be causedby convective flows. Serious damage to themicro- and macro-structure of the casting canthen occur. The time for convection to startappears to be in the region of 1 or 2 minutes. In3 or more minutes convection can become wellestablished, causing extensive remelting and amajor redistribution of heat in castings.

Castings that freeze in a time either shorterthan 1 minute, or longer than perhaps 10 minutes,are expected to be largely free from convectionproblems as indicated below.

Thin section castings are largely free fromconvection difficulties. They can therefore befed uphill simply because (i) the viscous restraintof its nearby walls makes any convective ten-dency more difficult, and (ii) more rapid freez-ing allows convection less time to develop andwreak damage in the casting. Thus instability is(i) suppressed and (ii) given insufficient time,respectively, so that satisfactory castings canbe made.

Conversely, thick section castings arealso relatively free from convection problems,because the long time available before freezingallows the metal plenty of time to convect,re-organizing itself so that the hot metal floatsgently into the feeders at the top of the casting,and the cold metal slips to the bottom. All thisactivity occurs and is complete before any sig-nificant amount of solidification has occurred.Thus the system reaches a stable conditionbefore damage can be caused. Once again,castings are predictable.

Figure 7.3 Horizontal orientation of alarge investment-cast turbine blade,illustrating convective loops in the rootand shroud. The flows convey heat fromthe cylindrical feeders, remeltingregions of the casting.


Ceramicplugs Support

Feeder

RunnerFilter


In what can only be described as a perverseact of fate, convection does its worst in the mostcommon sizes of castings, the problem emergingin a serious way in the wide range of inter-mediate section castings. These include theimportant structural castings such as auto-motive cylinder heads and wheels, and the largerinvestment cast turbine blades in nickel-basedalloys amongst many others. Convection canexplain many of the current problems withdifficult and apparently intractable feedingproblems with such common products. Theconvective flow takes about one or two minutesto gather pace and organize itself into rapidly

flowing plumes. This is occurring at the sametime as the casting is attempting to solidify. Theflows cut channels through the newly solidifiedmaterial, remelting volumes of the casting.

The channels will contain a coarse micro-structure because of their greatly delayedsolidification, and in addition may containshrinkage porosity if unconnected to feed metal.This situation is likely if the feeders solidifybefore the channels as undoubtedly happens onoccasions, because the channels derive theirenergy for flow from some other heat source,such as a very heavy section low down on thecasting, or the ingate attached to the riser tubeof a counter-gravity system for instance.

For conventional gravity castings that re-quire a lot of feed metal, such as cylinder headsand blocks, and which are bottom gated, buttop fed, this will dictate large top feeders,because of their inefficiency as a result of beingfurthest from the ingates, and so containing coldmetal. This is in contrast to the ingate sections atthe base of the casting that will be nicely pre-heated. The unfavourable temperature regime isof course unstable because of the inverted den-sity gradient in the liquid, and thus leads toconvective flow, and consequent poor predict-ability of the final temperature distribution andeffectiveness of feeding. It is the standard legacyof bottom filling: the favourable filling condi-tions leading to the worst feeding conditions.Life never was easy for the casting engineer.

The upwardly convecting liquid within theflow channels usually has a freezing time closeto that of the preheated section beneath, whichis providing the heat to drive the flow. In thecase of many low-pressure systems, the metalsupply system is artificially heated, leading to aconstant heat input, so that the convectingstreams rising out of these regions never solidify.This is what happened to the Cosworth systemin the early days of its development. When themould and casting (which should by now havebeen fully solidified) was hoisted from thecasting unit, liquid poured from the base ofthe mould, emerging from such remelted chan-nels to the amazement of onlookers who hadassumed that after the appropriate length oftime for solidification, no liquid could possiblystill be present, and present in such quantity.

When removing a convecting casting froma counter-gravity filling system in this way,the draining of liquid from the interdendriticregions leaves regions in the casting that appearconvincingly like shrinkage defects, and areusually confused as such.

The convection of hot metal up (and, ofcourse, the simultaneous movement of coldmetal down) the riser tube of a low-pressure

Figure 7.4 Encouragement of thermal convection by(a) side feeding; (b) bottom feeding; (c) itselimination by top feeding.

Rule 7. Avoid convection damage 161


casting unit (Figures 7.5 and 7.6) delays thefreezing of the casting in the mould above, andcan lead to a significant reduction in pro-ductivity. The author is aware of a casting beingmade on a low-pressure machine whose freezingtime kept increasing as the melt was subjected toincreasingly thorough rotary degassing treat-ment. It seems that each rotary degassingtreatment reduced the amount of bifilms insuspension. As the effective viscosity of the meltwas progressively reduced in this way the con-vection increased, extending the time taken forthe casting to solidify. Thus clean metal is free toconvect, whereas melt with an internal semi-

rigid lattice of bifilms will be more resistant toflow.

7.4 Countering convection

Solutions to the problems of convection aresummarised as follows:

(i) The inversion of the mould after castingeffectively converts the preheated bottom ingatefilling system into a top-feeding system, thusgaining a really efficient feeding system.

Furthermore, of course, the massive technicalbenefit of the inversion of the system to take thehot metal to the top, and the cold at the bottom,confers stability on the thermal regime. Con-vection is eliminated. For the first time, castingscan be made reliably without shrinkage porosity.

The massive productivity and economicbenefit of this technique follows because themould now contains its liquid metal all belowthe entry point, so that it can be detached fromthe casting station without waiting for thecasting to freeze (which is of course the standardproductivity delay suffered by most counter-gravity casting processes). In this way cycletimes can be reduced from about 5 minutes to1 minute. This is a powerful and reliable systemused by such operations as the Cosworth Pro-cess and an increasing number of other pro-cesses at the present time. We can hope thattechniques involving roll-over immediately aftercasting will become the norm for most castingsin the future.

(ii) Tilt casting processes (where the roll-overis used during castingÐactually to effect thefilling process) can also satisfy the top-feedingrequirement. However, in practice many geo-metries are accompanied by waterfall effects, ifonly by the action of the sliding of the metal inthe form of a stable, narrow stream down thesloping side of the mould. Thus meniscus con-trol is, unfortunately, often poor. Where thecontrol of the meniscus can be improved toeliminate entrainment problems, tilt castingtechniques are valuable. Ultimately, if the tilt iscontrolled to perfection, a kind of horizontaltransfer of the melt can be achieved, as discussedin Section 2.3.3. This system does not seemdifficult or costly to attain, and is to be recom-mended strongly.

(iii) Cut convective loops. Explore the elimi-nation of ingates on counter-gravity feeding ofcastings. The widespread convective loops ininvestment castings wax assemblies should becut by the wider use of ceramic supports andstops.

Figure 7.5 Convection driven flow within a solidifyinglow pressure casting.

Figure 7.6 Remnants of the convective plumes in acasting, defining regions of coarse structure andporosity.



Rule 8

Reduce segregation damage

At regions in which the local cooling rate of thecasting changes, such as at a change of section,or at a chill, or at a feeder, it is to be expectedthat a change in composition of the casting willoccur. In many alloy systems such variations areso slight as to be negligible. Such problems aretherefore normally neglected. However, thereare alloy systems that are particularly prone tosuch severe segregation problems that the cast-ing may be scrapped (if detected) or (if not cor-rected) can threaten its performance in service.

There are many good solidification texts thatdeal with the problem of segregation, so that it ishardly necessary to treat the subject in anylength here. Thus the various types are simplylisted as a reminder of the extent of the problem,and the many forms it can take.

Microsegregation is an unavoidable con-sequence of normal solidification in whichsolutes are concentrated (if the distributioncoefficient is less than one, as is usual) in theresidual liquid between dendrites. This inter-dendritic segregation can be cured, because itcan be re-distributed by diffusion back into thedepleted, rather pure, centres of the dendrites byhomogenization heat treatment; the diffusiondistances achievable during heat treatment areof the same order as the inter-dendritic spacing.Such heat treatments are usually carried out attemperatures close to the melting point of thealloy, and require up to several hours to achievea reasonable re-distribution of solutes.

When macroscopic flow occurs duringsolidification, the contents of these microscopicregions may be dispersed or may be concen-trated in distant regions of the casting depend-ing on whether the pattern on flow diverges orconverges respectively. Macrosegregation is theresult. This re-organization of the pattern of

chemical elements in the casting involves dis-tances vastly greater than can be cured by sub-sequent heat treatment. Heat treatments timesof perhaps the age of the earth (i.e. geologicaltime scales) may cure it but is not recommended.Macrosegregation, if it occurs, is unfortunatelytherefore, for all practical purposes, a perma-nent feature. Some of the various types ofmacrosegregation are described below.

In the casting of steel ingots the segregationof impurities into the head of the ingot, gen-erally known as positive or normal segregation,has been so bad that it has been necessary to cut-off and discard the top of the ingot. This hasrepresented a massive loss to the efficiency ofthe steel industry over the years, and was themain driving force for the development of con-tinuous casting in the 1960s, and now almostuniversal use for the casting of steel.

A well-understood type of segregation is so-called inverse segregation. Because this is theperfectly normal segregation to be expected inconditions of dendritic solidification the authorprefers to call it simply dendritic segregation. Inthis case the partitioned solute is segregatedpreferentially to the face of the mould, especiallyif this is a chill mould (Figure 8.1a). A similareffect will occur, of course, at the junction with athinner section that will act as a chill. The effectis shown in Figure 8.1b. The distribution ofalloying elements in this figure is particularlydisturbing, because in the case, for instance, ofthe high strength Al±4.5%Cu alloy, usuallychosen for highly stressed applications, thedeviations in chemistry can easily be outside theallowed specification of the alloy. The regionsof the casting high in solute will normally beexpected to be extra strong, possibly evenbrittle. The regions depleted in solute will,


conversely, be weak. What gives even greatercause for concern is that these sharp differencesin mechanical properties are sited so close to thechange in section where any stress will normallybe concentrated (Figure 8.1b).

Naturally, in a complex thermal field, andwhere the geometry of the casting is requiring acomplex distribution of residual liquid to feedshrinkage, these chemical variations can be

complex in distribution, and not always easilypredicted, except perhaps by sophisticatedcomputer simulation.

Another highly visible and severe type ofinverse segregation occurs when the residualliquid, concentrated against the surface of thecasting, actually penetrates the surface, toemerge as exuded segregated liquid on the sur-face of the casting. Such exudations are oftenlow melting point eutectics. The historicallyfamous example is `tin sweat' on bronze castingsand ingots, in which beads of tin-rich eutecticoccur on the surface of the copper-based alloy.The driving force for such exudations is some-times the general contraction of the casting as itcools, and sometimes the internal pressure gen-erated by the precipitation of gas from solution.An example of an Al±Si eutectic exudationagainst a chill surface is common in Al±Sialloys. In this case the chill first causes the meltto freeze quickly. However, the surface of thecasting then contracts away from the chill,whereupon it reheats and melts, allowing thesegregated residual liquid to bleed into the airgap that has opened up between the casting andthe chill (Figure 6.18).

In heavy steel castings and in steel ingots, thedistribution of carbon is significantly increasedunder the feeder. This occurs simply because inthe feeder the last liquid to solidify is high incarbon (and other elements such as phosphorusand sulphur) but this liquid is still being sup-plied to the casting. The liquid is finally suckedinto the casting to compensate for the shrinkageaccompanying the solidification of the casting(Figure 8.2). There has been much research intothis problem, so that actions to reduce theproblem are now reasonably well understood.An early description is given by Flemings (1971)in which he draws attention to the complicatingfact that, depending on whether the flow fromthe feeder is converging or diverging, the seg-regation can be positive or negative respectively.

Figure 8.1 (a) Dendritic segregation pattern, concentrating solute against a chilled face. (b) The analogous patternproduced by a reduction in section thickness acting as a cooling fin.

Figure 8.2 (a) The positive segregation pattern under afeeder. (b) The positive segregation pattern under acooling fin, but negative close by inside the fin. Theseextremes are both close to the vulnerable change in section.


Upper limit

Lower limit

Chill

(a)

%Cu

4

0

Al– 4Cu casting Al– 4Cu castingThin fin extension

Upper limit

Lower limit

%Cu

(b)

4

0

Positivesegregation

Positivesegregation

Positivesegregation

Negativesegregation

(a)

(b)


One of the easiest actions to reduce the problemis merely to make the feeder somewhat oversizeso that the remaining liquid is less concentratedat the time it is demanded by the casting. Again,it is sufficiently complicated to require a com-puter solution for an ultimate quantitativedescription. It is perhaps sufficient for most ofus to be aware that the problem exists, andcheck to see how important it may be.

Other positive segregation derives from theflow of the liquid, and is driven almost purely bygravity. A well-known example of gravity seg-regation of the liquid is the concentration ofcarbon, and other light elements such as sulphurand phosphorus, in the tops of large ferrouscastings. In contrast, tool steel castings sufferfrom the concentration of heavy elements suchas tungsten and molybdenum at the base of thecasting.

Gravity segregation of the solid can occur byequiaxed crystals in the melt that sediment tothe bottom of solidifying castings leading tothese lower regions being generally more purebecause they are composed of some of the firstsolid to solidify. Some additional purificationduring freezing may occur because of thedivergence of flow of residual liquid through

this zone. The overall effect is known as negativesegregation.

Strong concentrations of segregated solutesand inclusions are found in channel segregates,that are once again a feature of larger, or slowlycooled castings. In steel ingots these are the oncefamiliar `A' and `V' segregates, nowadays muchless prominent in continuously cast steels. InNi-base superalloys they are commonly knownas freckle defects.

As we have seen above, when extensive and/or intensive, all such changes in composition ofthe casting may cause the alloy of the casting tobe locally out of specification. If this is a seriousdeviation, the coincidence of local brittleness ina highly stressed region of the casting mightthreaten the serviceability of the product. Thepossibility of such regions therefore needs to beassessed prior to casting if possible, anddemonstrated to be within acceptable limits inthe cast product. Otherwise, techniques toreduce the segregation may need to be imple-mented. This is probably easier said than done.One approach would be to attempt to cool thecasting locally with chills or fins so as to achievea more even temperature distribution through-out the casting.

Rule 8. Reduce segregation damage 165


Rule 9

Reduce residual stress (the `no water quench'requirement)

9.1 Introduction

Action to reduce internal stress can be awe-somely important. Unfortunately, it seems thatin general, the engineering community has notbeen made aware of the central importance ofthis factor in the manufacturing of engineeringcomponents. All manufacturedcomponentscon-tain internalstress,oftenhigh.Theproblemis thatthis very real danger is invisible.

The problem is widespread, and not confinedto metal products. A common example we haveall seen is the high stress revealed by the maze ofcracks around the plug hole in some plasticwash basins. In this case the stress has beenrelieved by cracking, probably aided and abet-ted by the soaps and detergents that encouragecrack growthÐperhaps to be known as liquidsurfactant embrittlement, analogous to liquidmetal embrittlement or stress corrosion crackingin metals.

There are those metallurgists within theindustry, some eminent, and whose opionionson other matters I respect, that have taken issuewith me. They have argued that the presence ofresidual stresses, particularly those from quench-ing, are actually irrelevant since the wholecomponent is in balance with its own stresses.The question of balance is certainly true. How-ever, this argument overlooks the fact that thedistribution of stress is usually far from uni-form, and parts of the component may be nearto their failure stress even prior to the applica-tion of any service stress. Usually, as we shallsee, the major tensile stress is in the centre, and itis this part of the component that fails firstunder tensile load.

Admittedly, not all components are necessa-rily endangered by internal stress. Indeed, the

stress can be beneficial in some cases (see someexamples in Castings 2003). However, the majorrisk is that the stress may not be beneficial. Itmay add to the service stress and so promotepremature failure at only low service stress, tothe bewilderment of the designer who imagineshis component material to be inert. Because ofthe complexity of some castings, and the com-plexity of the state of stress, it is usually not easyto estimate the magnitude of either the internalresidual stress or its precise action. Often, how-ever, it is at least equal to or exceeds the yieldstress. Thus it is not trivial. In fact at this levelit will dominate all other designed loads in afatigue condition, and certainly lead to earlyfailure. It is ignored at our peril.

This section takes a look at the wide spectrumof stresses in castings, and attempts to clarifythose that are important and which shouldbe controlled, from those that can be safelyneglected.

9.2 Residual stress from casting

In an aggregate mould, castings are cooledrelatively slowly, so that the final internal stressin the product will normally be relatively low,and can often be neglected. It is true that thedimensions of the casting will often be changedby stress during cooling, but on shaking outfrom the mould the final, residual, stress will notnormally be high. Some examples are given inCastings (2003). In addition, the distortions thathave arisen during cooling in the mould areusually extremely reproducible. This is a con-sequence of the reproducible conditions of pro-duction, in which the mould is the sametemperature each time, and the metal is the same


temperature each time, so that the final shape isclosely similar each time. This reproducibility isprobably greater than for any other castingprocess.

This repeatable regime is not quite so wellenjoyed by the various kinds of die-casting,particularly gravity die (permanent mould)casting, as a result of many factors, but in par-ticular the variability of mould size and shape asa result of variation of mould temperature. Thesomewhat faster cooling, particularly because ofthe earlier extraction of the casting from themould, is an additional factor that does notfavour low final stress.

In general, internal stress remaining from thecasting process is rarely high enough to betroublesome but we cannot always be compla-cent about this. The ability to predict stressesusing computer simulation will be invaluable tomaintain a cautious watch for such dangers.

Ultimately, however, particularly for alumi-num alloys, the stresses from casting are usuallyeliminated by any subsequent high temperaturesolution heat treatment.

9.3 Residual stress from quenching

The final stresses in the component are dictatedby the final stages of this treatment, which isnormally a quench, and normally into water.Thus the major problems of internal stressesand distortion of the casting are usually createdat this moment. Furthermore, the stresses arenot significantly reduced by the subsequentageing treatment. The temperatures for ageingtreatments are too low to lead to stress relief.

It is unfortunate that many heat treatmentsrequire a quenching stage, intended to cool thecasting sufficiently quickly to freeze solutes in toa solid solution, thereby preventing them fromprecipitating. If the quench is slow some solutemay be lost by precipitation from solution, thusmaking it unavailable for subsequent hardeningreactions, so that the final strength of the cast-ing is reduced. This reasoning has driven thequest by metallurgists for quenching rates to beas fast as possible.

The problem has been that all such researchby metallurgists to optimize heat treatments hasbeen carried out on test bars of a few millimetresin diameter that represent no problem to coolquickly. The outside and inside of the bars is inexcellent thermal communication, and the highthermal conductivity of most metals ensuresthat the cooling throughout the section isessentially uniform. Thus the world's standardson heat treatment often dictate water quenchingto obtain the highest material properties.

Quite clearly, the problem of larger compo-nents, or certain components of special geome-trical complexity in which uniform cooling is animpossibility, has been overlooked. This is amost serious oversight. The performance of thewhole component may therefore be underminedby the application of these techniques that havebeen optimized by work on small test bars, andwhich therefore are inappropriate, if not actu-ally dangerous, for many large and complexcomponents.

This is such a common problem, that when atroubled casting user telephones me to saywords to the effect `My aluminium alloy castinghas broken. What is wrong with it?' this is such aregular question that my standard, and rathertired, reply now is `Do not bring the casting tome. I will tell you now over the telephone why ithas failed. It has failed because it has beenpoured badly and therefore contains bifilms thatreduce its strength. However, in addition, youhave carried out a solution heat treatmentaccompanied by a water quench.' The caller isusually stunned, incredulous that I know that hehas water quenched his casting, and asks howI know. My experience is this: in all my lifeinvestigating the causes of failure of perhapshundreds of Al alloy castings, only one failedbecause of serious embrittlement caused as aresult of the alloy being outside chemical spe-cification. All the rest failed for only two rea-sons; (i) weakening by bifilms, together with(ii) massive internal stresses that have loadedthe already weakened casting close to its failurestress even before any service stress was applied.I have to record, with some sadness, that all thestandard and costly investigations by metallur-gists into the chemical specification, the metal-lurgical structure, the mechanical properties andother standard metallurgical tests, are nearlyalways irrelevant. It underlines the importanceof understanding the new metallurgy of castmetals in which the residual stresses and bifilmstogether play the dominating roles in the per-formance of engineering components, particu-larly cast engineering components.

The key role of internal stress in the failure ofcastings (and other components such as forg-ings) is explained in Figure 9.1. The stress e isgiven by

e � a � DT (9.1)

where a is the coefficient of thermal expansion,and DT is the temperature change experiencedby the part.

Equation 9.1 explains why not all shapes andsizes of castings necessarily suffer a problem.Compact or small castings, and those for which

Rule 9. Take action to reduce residual stress 167


the quenchant can easily reach all parts, areoften not seriously affected, because DT isnecessarily small.

The actual magnitude of the strain e is salu-tary to estimate. For aluminium a is about20� 10ÿ6 Kÿ1 and the temperature fall during aquench is approximately 500 K. The strainworks out to be therefore approximately 1 percent. For steels a is approximately 14� 10ÿ6,and the temperature change for quench frommany heat treatments in the region of 900 K,again giving a strain close to 1 per cent. Sincethe yield (or proof ) strain is only approximately0.1 per cent, these imposed quench strains areabout ten times the yield strain, and can there-fore be seen to take the component well into theplastic deformation range.

Parts that are particularly susceptible includelarge, thick section castings, where the heat ofthe interior takes time to reach the outside of thecasting, giving high DT. Ingots or other block-type products can be seriously stressed for thisreason. Direct chill (continuously cast) ingotsaluminium alloys are severely cooled by water,but are often over 300 mm diameter. While sit-ting on the shop floor awaiting further proces-sing the strong 7000 series alloy ingots havesometimes been seen to explode like bombs. (Asan aside, the length of time taken before theingot decides to fail is curious and interesting.It seems likely that the failure under the highinternal stresses is initiated from one of the largebifilms that is expected to be entrained duringthe turbulent start of casting. The gradual pre-cipitation of hydrogen into the bifilm will gra-dually increase the pressure in the bifilm crack,encouraging it to extend as a stress crack. The

hydrogen may be already in solution in themetal, or may be gradually accrued by reactionwith water vapour in the atmosphere duringstorage, especially if the bifilm is connected tothe exterior of the ingot surface, allowing directpenetration of water vapour. Other penetratingcontaminants may include air to cause addi-tional internal oxidation, or fluxes, or traces ofchlorine gas, or sulphides from greases, to act assurface active additions to reduce the surfaceenergy of the metal and so further encouragecrack growth. Research to clarify these possi-bilities would be valuable.)

Other varieties of castings that are suscep-tible to damaging levels of residual stress includethose that are hollow, with limited access for thequenchant into the interior parts of the casting,and which also have interior geometrical fea-tures such as dividing walls and strengtheningribs (Figure 9.1). This latter series of geometricalrequirements might seem to eliminate mostcastings. Perhaps surprisingly therefore, the listof castings that fulfils these requirements israther long, and includes such excellent exam-ples as automotive cylinder heads and blocks,and housings for components such as com-pressors and pumps. When immersed in thewater quench, the water attempts to penetratethe entrances into the hollow interior of thecasting. However, because the casting is origin-ally above 500�C, any water that succeeds inentering will convert almost instantaneously tosteam, blowing out any additional water that isattempting to enter. The result is that the out-side of the casting cools rapidly, whereas theinterior can cool only at the rate that thermalconduction will conduct the heat along the tor-tuous path via interior walls of the casting to theouter surfaces.

The rate of conduction of heat from theinterior to the exterior of the casting can beestimated from the order of magnitude relation

x � Dt� �1=2 (9.2)

where x is the average diffusion distance, D isthe thermal diffusivity of the alloy, and t is thetime taken. The thermal diffusivity is defined as

D � K=rC (9.3)

where K is the thermal conductivity, about200 W mÿ1 Kÿ1 for aluminium, the density r isabout 2700 kg mÿ3 and the specific heat C isapproximately 1000 J kgÿ1 Kÿ1. These valuesyield a value for the thermal diffusivity D closeto 10ÿ4 m2 sÿ1. (The corresponding value forsteel is approximately 10ÿ5 m2sÿ1.) Equation 9.1is used to generate Figure 9.2 in which the

Typical ~100 mm path fordiffusion of heat from thecentre of the casting duringa quench

Figure 9.1 Schematic representation of a hollow castingwith small ports to the outside world and internal walls,such as a cylinder head, illustrating the long diffusion pathfor heat during a quench, together with the high internaltensile stress that might result in failure.



distance for diffusion of heat out of a productindicates the approximate boundaries of saferegimes constituting conditions in which suffi-cient time is available for the diffusion of heatfrom the interior during the quench. The time ofcooling in different quenchants over the criticalrange of approximately 500�C down to 250�Cis provided by results such as that shown inFigure 9.3. These results were obtained by sitinga thermocouple in the centre of a 10 mm wall ofan Al±7Si±0.4Mg alloy casting. Similar resultswould be valuable for ferrous materials.

For a solid aluminium bar of 20 mm dia-meter, or a solid plate of 10 mm thickness,Figure 9.3 indicates that quenching in water willreduce the temperature from 500 to 250�C inabout 5 seconds. Substituting 5 s in Equation 9.2shows that on average heat will have travelled20 mm in this time. The 20 mm bar or 10 mmplate will both therefore enjoy a reasonablyuniform temperature so that minimal stress willbe generated.

If, when quenching castings following hightemperature heat treatment, the time for coolingthe outer sections of castings is shorter than thetime required for heat to diffuse out frominterior sections, the outer parts of the castingcool to form a rigid frame. However, the innersections will cool and contract later, but by thattime unfortunately experiencing the restraint ofthe outside rigid sections. Thus the interiorsections go into tension, and the outer parts intocompression.

As stated above, this situation is common insuch castings as automotive cylinder heads,whose links between the internal sections andthe outside world are via tortuous routes aroundthe water jackets. The total distance that heatnow has to diffuse is of the order of 100 mm.

However, the walls of the casting are 10 mm orless, so that the cooling of the exterior of thecasting will again occur in a time of the order of5 seconds. However, Figure 9.3 indicates thatapproximately 100 seconds is required for theheat to diffuse the 100 mm distance from theinside to the outside, so the interior of the cast-ing will be expected to experience high tensilestress.

If the cylinder head casting had been sub-jected to quenching in a blast of air, Figure 9.3indicates that cooling will now take a leisurely100 seconds or so. Thus sufficient time isavailable for the internal sections to lose theirheat to the outside so that the casting maintainsa reasonably uniform temperature during thequench. The generation of high internal stress isavoided.

The author has personal experience ofquenching complex cylinder heads into water,and has suffered the consequences of banana-shaped castings that required to be straightenedwith a 50 000 kg press specially bought-in torectify the damage. Those were the castings thatdid not crack in the quench itself (the internalcracks inside the water jackets were often diffi-cult to locate). In addition, castings failed byfatigue in service after only short lives. Theintroduction of polymer quenching eliminatedthe problem. As explained in Castings (2003)there are a number of polymers that can be used.One commonly in use is a solution of polyalk-ylene glycol in water. The polymer precipitatesout of solution at 73�C, and so deposits over thesurface of the hot castings, forming a sticky,viscous layer. The layer is resistant to boiling, sothat a vapour blanket is avoided, and a steady,uniform flow of heat from the casting into thewater is achieved. When the casting finally

Figure 9.3 Quench rates in a 10 mm thick Al plate castingin a variety of quench media.


Tim

e ta

ken

for

exte

rior

of c

astin

g to

coo

l th

roug

h th

e cr

itica

l tem

pera

ture

ran

ge (

s)

1 10 100 1000Average distance for diffusion of heat during quench (mm)

Lowstress

Dangerof

high stress

Thre

shol

d fo

r ste

elTh

resh

old

for A

l allo

ys

1000

100

10

1

.�� - # �� '

500

400

300

200

100

01 10 100 1000

Time (s)

Tem

pera

ture

(°C

)

Water30%polymer

Forcedair

Still air


cools below 73�C the polymer goes back intosolution.

However, the polymer quench was not along-term practical solution for production foran automotive product. It had to be cleanedout of internal cavities where it could lodge,becoming concentrated and carbonized duringthe subsequent ageing treatment. In addition,any residual core sand in such locations wouldbe effectively cemented into place to causedamage later in the life of the engine when itfinally became dislodged.

In contrast to its use with automotive cast-ings, polymer was excellent for aerospace cast-ings where the extra trouble to clean eachcasting individually did not outweigh the benefitof superb heat treatment response and reducedinternal stress.

Air quenching was, however, a completesolution for automotive castings. It was low costand quickly and easily implemented in a seriesproduction environment. The castings retainedtheir accuracy, and quench failures and fatiguefailures disappeared. We were able to restoreproductivity and profitability (and get ourmoney back for the press).

Figure 9.4 illustrates how the overall strengthof a casting can be reduced by a heat treatmentdesigned to increase the strength of its material.Figure (a) shows the stress±strain curve for thealloy, and the imposition of 1 per cent; tensilestrain on the inner parts of the casting as a resultof a water quench. This quench strain results ina quench stress close to the failure stress of the

material. If no ageing treatment is carried outthis stress is locked into the component for therest of its life. Naturally, it has little residualstrength left, and is likely to fail on the firstapplication of a stress in service.

However, after an ageing treatment in whichthe yield strength of the alloy was intended to bedoubled, the situation is shown in Figure 9.4(b).Assuming the benefit of a small amount of stressrelief (the amount indicated in the figure may berather generous), the residual quench stress isonly slightly lower; substantially unchanged. Ifadditional service stress in tension is applied tothe central parts of the casting, the residualtensile stress in these parts is effectively a startingpoint for the additional loading. Thus, effec-tively, the new stress±strain curve for the com-ponent is shown in (c). It is clear that the newoverall stress±strain response has been reducedcompared to the original unheat-treated mate-rial; as a result of our lengthy, complex andexpensive heat treatment the component iseffectively weakened.

In summary, the residual stress in aluminiumalloy castings quenched into water in this way iswell above the yield point of the alloy. Evenafter the strengthening during the ageing treat-ment, the stress remains at between 30 and 70 percent of the yield stress, with a useful workingapproximation being 50 per cent. Thus theuseful strength of the alloy is reduced from itsunstressed state of 100 per cent, down to around50 per cent. This massive loss of effectivestrength makes it inevitable that residual tensile

Failurestress

Quenchstress

0.1%proofstress

0.1%proofstrain

1%appliedstrain

0.1% Strain e

0.1PS

(a) (b) (c)

0.1PSStr

ess,

r

Minimal stressrelief duringageing treatment

Residualstress

New effectivebaseline

Figure 9.4 Evolution of the stress/strain curve of an Al alloys as heat treatment progresses (a) after quench; (b) afterageing to double the 0.1 per cent proof stress; and (c) the final effective stress/strain curve showing propertieseffectively less than the as quenched properties.



stresses are a significant cause of casting failurein service, particularly fatigue failure, since theresidual stress is always generously above thefatigue limit of the alloy.

For many castings, the use of a boiling waterquench has been demonstrated to be of negli-gible help in reducing the stresses introduced bywater quenching (Castings 1991). Thus althoughthe rate of quench is certainly reduced by theuse of hot or boiling water the results are notalways reliable. This is almost certainly theconsequence of the variability of the vapourblanket that forms around the hot casting.The blanket forms and disappears irregularly,depending on many factors including the precisegeometry of the part, its inclination during thequench, and the proximity of other hot castings,etc. In addition, from a practical point of view, ahot water quench is not cheap to install, run ormaintain.

Turning now to steels; in contrast to thebehaviour of Al alloys, the thermal diffusivityD is approximately ten times lower, of the orderof only 10ÿ5. The reader can quickly show thatthe corresponding distances to which heat canflow are 7 mm in 5 seconds but only 30 mm in100 seconds. For a given rate of quench there-fore, steels will suffer a higher residual stress(Figure 9.2). Nevertheless, they are much moreable to withstand such disadvantages, havinghigher strength, but more particularly, higher

elongation. Thus although the final internalstress is high, the steel product is nowhere nearthe failure condition experienced by the alumi-nium alloy casting. The aluminium alloy castingexperiences about 1 per cent imposed elongationbut has only a few per cent, perhaps even lessthan 1 per cent elongation prior to failure. Thusit can fail actually in the quench, or early inservice. In contrast, the steel casting has ten ortwenty times greater elongation (as a result pri-marily of its reduced bifilm content). Thusalthough the 1 per cent or so of imposed quenchstrain resulting from unequal cooling may resultin 1 per cent or so of distortion of the product,its condition is far from any dangerous condi-tion that might result in complete failure, sinceenormously greater strain has to be imposed toreach the failure condition (Figure 9.5).

The above statements are so important theyare worth repeating in different words foradditional clarity. The rapid quenching of steelsfor metallurgical purposes (such as the stabili-zation of austenite for Hadfield Manganesesteel) is not usually a problem. The reason isthat most steels are particularly clean, becauseof the rapidity with which entrainment defectsare deactivated and/or detrained after pouring.The result is that steels typically have an elon-gation to failure of 40 or 50 per cent. In contrast,most Al alloys (and probably most Mg alloys)do not enjoy this benefit; suffering from a high

Figure 9.5 A comparison of stress/strain curves of an Al alloy and a steel, illustrating the relatively dangerouscondition of the Al alloy after a quench.


Ultimate stress

Quenchstress

Quenchstress

Ultimate stress

Approx 1%quench strain

Failure strainfor steel

Failure strain for Al alloy

Steel

Al alloy


density of bifilms they typically achieve less thana tenth of this ductility. Thus the application of1 per cent strain takes the aluminium alloy closeto, or even sometimes in excess of its failurestrain. For steels, even though the 1 per centstrain applied by the quench will take the partinto the plastic region, causing huge stresses, thesteel remains safe; its greater freedom frombifilms permits it to endure enormously greaterextension before it will fail (Figure 9.5).

For the future, the production of Al alloyswith low bifilm concentration promises to offerductilities in the range of that of steels. Already,good foundries know that high strengths to-gether with elongations of 10 to 20 per cent areachievable, if good care is taken.

Slower quenching techniques are safer, al-though, of course, the strength attained by theheat treatment is somewhat reduced. Even so,the reduced mechanical strength when usingslower and more controlled quenches such as apolymer or a forced air quench is more thancompensated by the benefit of increased relia-bility from putting unstressed (or more accu-rately, low-stressed) castings into service. Thusthe casting designer and/or customer needs tospecify somewhat reduced mechanical strengthand hardness requirements in order to gain asuperior performance from the product. Thereductions of strength and hardness are expected

to be in the range 5 to 10 per cent, but theimprovement in casting performance can beexpected to be approximately 100 per cent. Theseare huge benefits to be gained at no extra cost.

9.4 Distortion

Residual stresses in castings are not only seriousfor parts that require to withstand stress inservice. They are also of considerable incon-venience for parts that are required to retain ahigh degree of dimensional stability. This pro-blem was understood many years ago, beingfirst described as early as 1914 in a modelcapable of quantitative development by Heyn.The model of a three-bar casting is shown inFigure 9.6. The internal stresses are representedby two outer springs in compression, each car-rying half of the total load of internal com-pressive stress, and an inner spring in tensioncarrying all of the internal tensile stress. If oneof the surfaces of the casting is machined away,one of the external stresses is removed. It ispredictable therefore that the casting will deformto give a concave curvature on the machined sideas illustrated in the figure.

The distortion of castings both before andafter machining is a common fault, and typicalof castings that have suffered a water quench.

Elastic model representing theequilibrium state of stress

Compression

Compression

Tension

Casting

After machining away one part of cast surface

Relief of internal tension by internal fracture

(a)

(b)

(c)

=

=

=

Figure 9.6 Heyn's (1914) modelof the balance of internal stressesafter rapid cooling: (a) thequenched casting showing highinternal tensile stress andrelatively low external compressivestress; (b) the distortion of thecasting after one side is machinedaway; and (c) the condition ofinternal tensile failure.



Once again, it is a problem so frequentlyencountered that I have, I regret, wearied ofanswering the telephone to these enquiries too.After all, it is difficult to understand how acasting could avoid distortion if parts of it arestressed up to or above its yield point.

For light alloy castings in particular a moregentle quench, avoiding water (either hot orcold), and choosing polymer or air will usuallysolve the problem instantly. As mentionedbriefly above, such polymer performs wellfor aerospace castings but is expensive andmessy, whereas air is recommended as beingclean, economical and practical for high volumeautomotive work. Otherwise, stress relievingcastings by heat treatment prior to machiningis strongly recommended (Castings 2003). Ineither case, of course, some fraction of theapparent strength of the product has to besacrificed.

9.5 Heat treatment developments

Although not strictly relevant to the question ofreducing residual stresses, it is worth empha-sizing the newer developments in heat treat-ments that give approximately 90 per cent ormore of total attainable strength, but with muchreduced stress and greatly reduced cost. Thereduced cost is always an attention-grabbingtopic, and materially helps the introduction oftechnology that can deliver an improved product.

Figure 9.7 illustrates the progression ofrecent developments in heat treatment of Alalloys where the problem of stress is central.

The traditional full heat treatment of aprecipitation-hardened alloy, that constitute thebulk of cast structural components at this time,consists of a solution treatment, water quenchand age as illustrated in (a). The treatmentresults in excellent apparent strength for thematerial, but is energy intensive in view of thelong total times.

Illustration (b) shows how the traditionaltreatment can be reduced significantly in mod-ern furnaces that enjoy accurate control overtemperature, thereby reducing the risk of over-heating the charge because of random thermalexcursions. An increase in temperature by 10�Cwill allow, to a close approximation, an increasein the rate of treatment by a factor of 2. Thustimes at temperature can be halved. These ben-efits are cumulative, such that a rise of 20�C willallow a reduction in time by a factor of2� 2� 4, or a rise of 30�C a reduction in time ofa factor 2� 2� 2� 8, etc. Both (a) and (b)require separate furnaces for solution and age-ing treatments (if long delays waiting for the

T(°

C)

500

200

500

200

500

200

500

(d)

(c)

(b)

(a)

200

Solution

Water or polymerquench

Water or polymerquench

Age

Air quench

Air quench

Cool in still air

Figure 9.7 A progression of precipitation heat treatmentdevelopments for Al alloys. (a) A traditional fulltreatment, giving excellent apparent properties buttaking between 12 and 24 hours; (b) shortened treatmentgiving nearly equivalent result; (c) the use of air quenchto reduce time, energy, and residual stress; (d) an ultimateshort and simple cycle.



solution furnace to cool to the ageing tempera-ture are to be avoided). Thus floor spacerequirement is high. Floor space requirement isincreased further by the quench station, and, if apolymer quench is used, by a rinsing tankstation.

The reader will appreciate that the tinyadditional energy required by the higher tem-perature is of course completely swallowed bythe huge savings in overall time at temperature.

If an air quench is used to gain the benefits ofreduced residual stress, the additional benefitsto the overall cycle time are seen in Figure (9.7c)because the quench can now be interrupted andthe product transferred to the ageing furnacealready at the correct temperature for ageing,saving time and reheat energy. Additional ben-efits include the fact that the air quench isenvironmentally friendly; the castings are notstained by the less-than-clean water; the con-veyor is straightforward to build and maintain;there is no mechanism required for loweringinto water that normally results in complex andrusting plant. As we have repeatedly empha-sized, the products from this type of furnacehave somewhat lower apparent strengths andhardness, but greatly improved performance inservice.

Figure 9.7d shows an ultimate system thatmight be acceptable for some products. Theageing treatment is simply carried out by inter-rupting the air quench slightly above the normalageing temperature, and allowing the part tocool in air (prior to final rapid cooling by fans ifnecessary). This represents a kind of naturalageing process in which no ageing furnace isrequired. Strengths will suffer somewhat, butthe lower costs and simplicity of the process maybe attractive, making the process suitable forsome applications.

9.6 Epilogue

Although the strength of the material will belowered by a slower quench, the strength ofthe component (i.e. the failure resistance of thecomplete casting acting as a load bearing part)in service will be increased.

If water quench is avoided with a view toavoiding the dangers of internal residual stress,it is common for the customer to complainabout the 5 per cent or so loss of apparentproperties. In answer to such understandablequestions, an appropriate reply to focus atten-tion on the real issue might be `Mr Customer,with respect, do you wish to lose 5 per cent or50 per cent of your properties?'

In the experience of the author, a numberof examples of castings that have been slowlyquenched, losing 5 or 10 per cent of theirstrength, are demonstrated to double their per-formance in service (Castings 2003).

Finally therefore, it remains deeply regrett-able, actually a scandal, that many nationalstandards for heat treatment continue to specifywater quenching. This disgraceful situationrequires to be remedied. In the meantime theauthor deeply regrets having to recommend thatsuch national standards be set aside. It is easyfor the casting supplier to take refuge in the factthat our international and national standardson heat treatment often demand quenching intowater, and thereby avoid the issue that such aproduction practice is risky for many compo-nents, and in any case provides the user with acasting of inferior performance. However, theethics of the situation are clear. We are notdoing our duty as responsible engineers and asmembers of society if we continue to ignorethese crucial questions. We threaten the per-formance of the whole component merely tofulfil a piece of metallurgical technology thatfrom the first has been woefully misguided.

The fact is that our inappropriate heattreatments have been costly to carry out, andhave resulted in costly failures. It has to beadmitted that this has been nothing short of acatastrophe for the engineering world for thepast half century, and particularly for thereputation of light alloy castings, not to mentionthe misfortune of users. As a result of theunsuspected presence of bifilms they have suf-fered poor reliability so far, but as a result of theunsuspected presence of residual stress this hasbeen made considerably worse by an unthinkingquest for material strength that has in factreduced component performance.



Rule 10

Provide location points

This Rule, provide location points, is addedsimply because the foundry can accomplish allthe other 9 Rules successfully, and so producebeautiful castings, only to have them scrappedby the machinist. This can create real-life dramaif the castings have been promised in a just-in-time delivery system. This Rule is added to helpto avoid such misfortunes, and allow all partiesto sleep more soundly in their beds.

Before describing location points, their logi-cal precursors are datum planes. We need todecide on our datums first.

10.1 Datums

A datum is simply a plane defining the zero fromwhich all dimensions are measured. For a castingdesign it is normal to choose three datum planes atright angles to each other. In this way all dimen-sions in all three orthogonal directions can beuniquely defined without ambiguity.

In practice, it is not uncommon to find acasting design devoid of any datum, there beingsimply a sprinkling of dimensions over thedrawing, none of the dimensions being neces-sarily related to each other. On other designs thedimensions relate with great rigour to eachother and to all machined features such asdrilled holes, etc., but not to the casting. In yetother instances that the author has suffered,datums on one face have not been related todatums on other views of the same casting. Thusthe raft of features on one face of the castingshifts and rotates independently of the raft offeatures on the opposite face.

Figure 10.1a shows a sump (oil pan) for a dieselengine designed for gravity die-casting in an alu-minium alloy. The variations in die temperature

and ejection time result in variability of the lengthof the casting that are well known with this pro-cess, and not easily controlled. The figure showshow the dimensioning of this part has made thepart nearly unmanufacturable by this method.Three fundamental criticisms can be made:

1. The datum is at one end of the product. If thedatum had been defined somewhere near thecentre of the part, then the variabilityproduced by the length changes of the castingwould have been approximately halved.

2. There is only one feature on the componentwhose location is critical; this is the dipstickboss. If the boss is slightly misplaced then itfouls other components on the engine. It willbe noticed that the dipstick boss is at the farend of the casting from the datum. Thusvariability in length of the casting will ensurethat a large proportion of castings will bedeemed to have a misplaced boss. If thedatum had been located at the other end ofthe casting, near to the boss, the problemwould have been reduced to negligible pro-portions. If the datum had been chosen as theboss itself, the problem would have disap-peared altogether, as in Figure 10.1b.

3. The datum is not defined with respect to thecasting. It is centred on a row of machinedholes, which clearly do not exist at the time thecasting is first made and when it is first requiredto be checked. Depending on whether themachinist decides to fix the holes in the centreof the flange, or relate them by measurement tothe more distant dipstick boss, or to the centreof the casting averaged from its two ends, orany number of alternative strategies, the drilledholes could be almost anywhere in or evenpartly off the flange!


Figure 10.1b shows how these difficulties areeasily resolved. The datum is located against theside of the dipstick boss, and hence is fixed in itsrelation to the casting and goes some way tohalving errors in the two directions from thisplane. It also means that the dipstick boss itselfis now impossible to misplace, no matter howthe casting size varies; all the other dimensionsare allowed to float somewhat because move-ment of other points on the casting is not a

problem in service. This part can then be pro-duced easily and efficiently, without trauma toeither producer or customer!

In summary, the rules for the use of datums(partly from Swing 1962) are:

1. Choose three orthogonal datum planes.2. Ensure the planes are parallel to the axes of

motion of the machine tools that will be usedto machine the part (otherwise unnecessarycomputation and opportunity for error isintroduced).

3. Fix the planes on real casting features, suchas the edges of a boss, or the face of a wall(i.e. not on a centreline or other abstractconstructional feature). Choose casting fea-tures that are:(a) critical in terms of their location, and(b) as near the centre of the part as possible.

10.2 Location points

Location points are those tiny patches on thecasting that are used to locate the casting pre-cisely and unambiguously in three dimensions.They are required by the toolmaker, since he canconstruct the tooling with reference to them; thefounder, to check the casting once it has beenmade; and the machinist, who uses them tolocate the casting prior to the first machiningoperations. These features therefore integratethe manufacture of the product, ensuring itssmooth transmission as it progresses fromtoolmaker to founder to machinist.

Whereas the casting datums are invisibleplanes, defining the concept of a zero in thedimensional space in and around a casting,the tooling points are real bits of the casting.The datums are the software, whereas the tool-ing points are the hardware, of the dimensioningsystem. It is useful, although not essential, forthe datums to be defined coincident with thetooling points.

Location points are known by several dif-ferent names, such as machining locations(which is a rather limiting name) or pick-uppoints. On drawings, TP for tooling point isused as the common abbreviation for thedrawing symbol. Although in practice I tend touse all the names interchangeably it is proposedthat `location points' describes their functionmost accurately and will be used here.

Before the day of the introduction of locationpoints in the Cosworth engine-building opera-tion, I was accustomed to a complex cylinderhead casting taking a skilled man at least2 hours to measure to assess how to pick up the

(b)

(a)

.�� ' �� '

�� '�� ; �� ,( *��


casting for machining. The casting was repeat-edly re-checked, re-orienting it slightly withshims to test whether wall thicknesses wereadequate, and whether all the surfaces requiredto be machined would in fact clean up onmachining. After the 2 hours, it was common tosee it, our most expensive casting, dumped onthe scrap heap; no orientation could be found toensure that it was dimensionally satisfactoryand could be completely cleaned up onmachining. All this changed on the day when thenew foundry came on stream. A formal systemof location points to define the position of thecasting was introduced. After this date, nocasting was subjected to dimensional checking.All castings were received from the foundry andon entering the machine shop were immediatelythrown on to a machine tool, pushed up againsttheir location points, clamped, and machined.No tedious measurement time was subsequentlylost, and no casting was ever again scrapped formachining pick-up problems.

It is essential that every casting has definedlocations that will be agreed with the machinistand all other parties who require to pick up thecasting accurately.

For instance, it is common for an accuratecasting to be picked up by the machinist usingwhat appear to be useful features, but whichmay be formed by a difficult-to-place core, or apart of the casting that requires some dressingby hand. Thus although the whole casting hasexcellent accuracy, this particular local feature issomewhat variable in location. The result is acasting that is picked up inaccurately, and doesnot therefore clean up on machining. As a resultit is, perhaps rather unjustly, declared to bedimensionally inaccurate.

The author suffered precisely this fate afterthe production of a complex pump body castingfor an aerospace application that achievedexcellent accuracy in all respects, except for asmall region of the body that was the site wherethree cores met. The small amount of flash atthis junction required dressing with a handgrinder, and so, naturally, was locally ground toa flat, but at various slightly different depthsbeneath the curved surface of the pump body.This hand-ground location was the very site thatthat machinist chose to locate the casting. Theresult was disaster. Furthermore, it was noteasily solved because of the loss of face to themachinist who then claimed that the locationoptions suggested by the foundry were incon-veniently awkward. The fault was not his ofcourse. The fundamental error lay in notobtaining agreement between all parties beforethe part was made. If the location point used bythe machinist really was the only sensible option

for him, the casting engineer and toolmakerneeded to ensure that the design of the corepackage would allow this.

Ultimately, this Rule is designed to ensurethat all castings are picked up accurately, andconveniently if possible, so that unnecessaryscrap is avoided.

Different arrangements of location points arerequired for different geometries of casting. Someof the most important systems are listed below.

10.2.1 Rectilinear systems

1. Six points are required to define the positionof a component with orthogonal datumplanes that is designed for essentially recti-linear machining, as for an automotivecylinder head or block. (Any fewer pointsthan six are insufficient to define the positionof the casting, and any more than six willensure that one or more points are potentiallyin conflict.)

On questioning a student on how to use a six-point system to locate a brick-shaped casting,the reply was `Oh easy! Use four points aroundthe outside faces and one top and one bottom.'This shows how easy it is to get such conceptswildly wrong!

In fact, the six points are used in a 3, 2, 1arrangement as shown in Figure 10.2. The sys-tem works as follows: three points define plane A,two define the orthogonal plane B, and onedefines the remaining mutually orthogonalplane C (Figure 10.2). The casting is then pickedup on a jig or machine tool that locates againstthese six points. Example (a) shows the basic useof the system: points 1, 2 and 3 locate plane A;points 4 and 5 define plane B; and point 6defines plane C. Planes A, B and C may be thedatum planes. Alternatively, it is often just asconvenient for them to be parallel to the datumplanes, but at accurately specified distancesaway.

Clearly, to maximize accuracy, points 1, 2 and3 need to define a widely based triangle, andpoints 4 and 5 similarly need to be as widelyspaced as possible. A close grouping of thelocations will result in poor reproducibility of thepickup of the casting; tiny errors in the positionor surface roughness of the tooling points wouldbe magnified if they were not widely spaced.

Example (b) shows an improved arrangementwhereby the use of a tooling lug on the long-itudinal centreline of the casting allows thedimensions along the length of the casting to behalved. The largest dimension of the casting isusually subject to the largest variability, sohalving its effect is a useful action.

Rule 10. Provide location points 177


Example (c) is a further development of thisidea, creating lugs that serve the additionaluseful purpose of allowing the part tobe clamped immediately over the points ofsupport, and off the faces that require to bemachined. In practice the lugs may be existingfeatures of the casting, or they may be additionsfor the purpose of allowing the casting to bepicked up for inspection or machining. Thisconcept is capable of further development, usinglugs arranged on all the centrelines of the castingso as to halve the errors in all directions.

10.2.2 Cylindrical systems

Most cylindrical parts do not fall nicely into theclassical six-point location systems as describedabove for rectilinear components. The errors of

eccentricity and diameter both contribute to arather poor location of the centre using thisapproach. The unsuitability of the orthogonalpick-up system is analysed nicely by Swing(1962).

In fact, the obvious way to pick up a cylinderis in a three-jaw chuck. The self-centring actionof the chuck gives a useful averaging effect onany out of roundness and surface roughness, andis of course insensitive to any error of diameter.

In classical terms, the three-jaw chuck isequivalent to a two-point pickup, since it definesan axis. We therefore need four more points todefine the location of the part absolutely. Threepoints abutting the jaws will define the plane atright angles to the centre axis, and one finalpoint will provide a `clock' location. Figure 10.3shows the general scheme.

Another location method that is occasionallyuseful is the use of a V block. This is a way ofensuring that a cylindrical part, or the roundedge of a boss, is picked up centrally, averagingerrors in the size and, to some extent, the shapeof the part. The method has the disadvantagethat errors in diameter of the part will cause thewhole part to be shifted either nearer to or away

Figure 10.3 The use of a three-jaw self-centring chuckfor casting location and clamping.


6

4

2

3

3

3

5

4

4

5

4

1

1

1

1

5

2

2

2

5

6

6

6

3

(c) Use of toolinglugs for clamping

(d) Jig or fixtureattached to mark-outtable or machine tool

(a) Basic location system

(b) Halving of length errors

Plane B

Plane A

Plane C

.�� # �� 9�� +�� '�� *�� ' �� "�� 1�� +�� '


from the block, depending on whether the dia-meter is smaller or larger. (The reader canquickly confirm these shifts in position withrough sketches.)

A widely used but poor location technique isthe use of conical plugs to find the centre of acored hole. Even if the hole is formed by themould, and so relatively accurately located, anyimperfection in its internal surface is difficultto dress out, and will therefore result in mis-location. If a separate core forms the hole thenthe core-positioning error will add to the overallinaccuracy of location. Location from holes isnot recommended.

It is far better to use external features such asthe sides of bosses or walls, as previously dis-cussed. These can be more easily cast andmaintained clean.

10.2.3 Trigonal systems

For some suitable parts of triangular form, suchas a steering gear housing, a useful and funda-mentally accurate system is the cone, grooveand plane method (Figure 10.4).

10.2.4 Thin-walled boxes

For prismatic shapes, comprising hollow, box-like parts such as sumps (oil pans), the pickupmay be made by averaging locations defined onopposite internal or external walls. This is amore lengthy and expensive system of locationoften tackled by a sensitive probe on themachine tool, that then calculates the averageddatum planes of the component, and orients thecutter paths accordingly. This technique isespecially useful where an average location isdefinitely desirable, as a result of the castingsuffering different degrees of distortion of itsrelatively thin walls.

The tooling points should be defined on thedrawing of the part, and should be agreed by(i) the manufacturer of the tooling, (ii) the caster,and (iii) the machinist. It is essential that allparties work from ONLY these points whenchecking dimensions and when picking the partup for machining.

For maximum internal consistency betweenthe tooling points, all six should be arranged tobe in one half of the mould, usually the fixed orlower half, although sometimes all in the cope.The separation of points between mould halves,or having some defined from the mould andsome from cores, will compromise accuracy.However, it is sometimes convenient and correctto have all tooling points in one internal core, oreven one half of an internal core (defined fromone half of the core box) if the machining of the

part requires to be defined in terms of its inter-nal features.

Clearly therefore, the location points arerequired to be actual cast-on features of thecasting. This point cannot be over-emphasized.It is not helpful, for instance, to define a loca-tion feature as a centreline of a bore. This invi-sible feature only exists in space (perhaps weshould say `free air'). Virtual features such ascentrelines have to be found by locating several(at least three) points on the internal as-castsolid surface of the bore, and its centre therebycalculated. Clearly, these `virtual' or `free-air'so-called location points necessarily rely on theirdefinition from other nearby as-cast surfaces.These ambiguities are avoided by the directchoice of as-cast location features.

These features need to be cast nicely, withoutobscuring flash, or burned-on sand, and defi-nitely should not be attacked by enthusiasticfinishers wielding an abrasive wheel.

It is essential that the location points are notmachined. If they are machined, the circum-stance poses the infinitely circular question`What prior datums were used to locate andposition the casting accurately to ensure that themachining of the machining locations (fromwhich the casting would be picked up formachining) were correctly machined?' Unfor-tunately, such indefensible nonsense has itscommitted devotees.

Figure 10.4 A plan view of a steering housing for a car,showing a flat, groove and cone location system.


Cone

Flat

Vee groove


In general it is useful if the tooling locationson the casting can remain in position for the lifeof the part. It is reassuring to have the toolingpoints always in place, if only to resolve disputesbetween the foundry and machine shop con-cerning failure of the part to clean up onmachining. It is therefore good practice to try toavoid placing them where they will be eventuallymachined off. Using existing casting featureswherever possible avoids the cost of additionallugs, and possibly even the cost of subsequentremoval if their presence on the final product isnot allowed.

The definition of the six-point locations,preferably on the drawing, prior to the manu-facture of the casting, is the only method ofguaranteeing the manufacturability of the part.The method allows an integrated approach rightfrom the start of the creation of the tooling,because the patternmaker can use the toolingpoints as the critical features of the tooling inrelation to which all measurements will bedefined. The foundry engineer will know how topick up the casting to check dimensions afterthe production of the first sample castings. Themachinist will use the same points to pick up thecasting for machining. They all work fromthe same reference points. It is a commonlanguage and understanding between design,manufacture, and inspection of products. Dis-putes about dimensions then rarely occur, or ifthey do occur, are easily settled. Casting scrapapparently due to dimensioning faults, or faultypickup for machining, usually disappears.

This integrated manufacturing approach isrelatively easily managed within a single inte-grated manufacturing operation. However,where the pattern shop, foundry and machinistare all separate businesses, all appointed separ-ately by the customer, then integration can bedifficult to achieve. It is sad to see a well-designed six-point pick-up system ignoredbecause of apparent cussedness by one memberof the production chain. The industry and itscustomers very much need purchasing andmanufacturing policies based on the adoption ofintegrated and fundamentally correct systems.

10.3 Location jigs

Figure 10.2d illustrates a basic jig that is designedto accept a casting with a six-point location sys-tem. The jig is simply a steel plate with a series ofsmall pegs and blocks. It contrasts with manycasting jigs, which are a nightmare of construc-tional and operational complexity.

Our simple jig is also simple to operate. Whenplacing the casting on the jig, the casting can

be slid about on locations 1, 2 and 3 to defineplane A, then pushed up against locators 4 and5 to define plane B, and finally slid along loca-tors 4 and 5 until locator 6 is contacted. Thecasting is then fixed uniquely in space in relationto the steel jig plate. It can then be clamped, andthe casting measured or machined. The sixlocations can, of course, be set up and fixed inthe machine tool that will carry out the firstmachining operation.

After the first machining operations it is nor-mal to remove the casting from the as-cast loca-tions and proceed with subsequent machiningusing the freshly machined surfaces as the newlocation surfaces. McKim and Livingstone (1977)go on to define the use of functional datums whichmay become useful at this stage. They aremachined surfaces that normally relate to featureslocating the part in its intended final application.

Other jigs can easily be envisaged forcylindrical and other shaped parts.

10.4 Clamping points

During machining the forces on the casting canbe high, requiring large clamping loads toreduce the risk of movement of the casting.Clamping points require to be thought aboutand designed in to the casting at the same timeas the location points. This is because theapplication of high clamping loads to the cast-ing involves the risk of distortion of the casting,and of spring-back after release of the clamps atthe end of machining. Flat machined surfacesare apt to become curved after machiningbecause of this effect.

The great benefit of using tooling lugs asshown in Figure 10.2c can therefore be appre-ciated. The location point and the clampingpoint are exactly opposed on either side of thelug. In this way the clamping loads can be high,without introducing the risk of the overall dis-tortion of the casting.

Further essential details of the design of theclamping action include the requirement for theaction to move the part on to, and hold itagainst, the location point.

For softer alloys that are easily indented, theclamp face needs to be 5±10 mm in diameter,similar to the working area of the tooling point.Even so, a high clamping load will typicallyproduce an indentation of 0.5 mm in a soft Alalloy, decreasing to 0.2 mm in an Al alloyhardened by heat treatment, and correspond-ingly less still in irons and steels.



10.5 Mould design: the practical issues

The problem for the casting engineer is toachieve a successful design of the mould. Thisproblem is not to be underestimated, since itrequires the simultaneous solving of a list ofissues including

(i) The design of the mould and core assemblycan be a problem in itself. It is notuncommon to find that it is impossible toassemble the cores because some shapefeature of neighbouring cores has beenoverlooked. It is all too easy to stumbleinto such pitfalls in a complex core assem-bly. When the first set of cores are madefrom the new patternwork, with its shiningnew varnish and paintwork, the discoveryof such `passing problems', where one corewill not pass another and so fit into theassembly, are greeted with embarrassmentand dismay.

The other common problem for the castingengineer and toolmaker is the design of theassembly so that cores fit, in logical order, onlyinto the drag if possible (Figure 10.5). Cores inthe cope are not usually an option for horizon-tally parted greensand moulds, since, if the sandstrength is not high, they are in danger of fallingout of their prints when the cope is turned overand closed onto the drag prior to casting. Glu-ing cores into the cope is possible in the case ofstrong chemically bonded sand moulds. How-ever, gluing takes time and is therefore costly,and introduces the danger that any excess gluemay cause a blow hole defect in the casting if itcontacts the metal. In addition, glue applied to acore print may prevent the core from venting,leading to a rather different form of blow defectfrom the core itself. The use of glues shouldtherefore be avoided if at all possible.

It is common for complex core assemblies tobe assembled at a separate station sited off themould assembly line. Core assembly can then beaccurate since the assembly is built up in a jig.The cores are designed to be lifted by the jig,transferring from the assembly station andlowered into the mould as a complete package.Castings that require lengthy core assemblytimes are not thereby allowed to slow the cycletime of the moulding line.

(ii) The filling system. The provision of a goodfilling system, and its integration with the restof the mould and core system is sometimesnot easy, and in some cases the additionaltrouble or expense to provide a good fillingsystem is by-passed. (The minefield of poor

castings and high scrap rates is always enteredfor apparently good reasons.) The fillingsystem design forms the major part of thisbook. It is mandatory reading. Its rules arerecommended to be followed in all cases.

(iii) The feeding system. Naturally, followingthe first rule for feeding, it is clearly best iffeeders can be completely avoided. How-ever, if they are considered to be necessary,it is usually not a problem to place feedershigh on a casting. Thus the provision offeeders rarely involves difficulties of moulddesign. One of the key issues is to place thefeeders so that they are easy to cut off ormachine away subsequently.

(iv) The avoidance of infringement of any of the10 Rules. For instance, convection consid-erations might force the issue of rotating themould through 180 degrees after filling. Thisaction usually confers other benefits andmakes integration of the filling and feedingsystems powerfully effective and economic.It is a strategy to be recommended.

However, sometimes the solution to all theseissues is not straightforward. For instancemuch time may be spent attempting to solve

Figure 10.5 (a) Simple cake core and drag assembly;and (b) a cope and drag with side cores, all located in thedrag; (c) an apparently lower-cost alternative to (b),but resulting in possible loss of dimensional control.



the issues with the casting oriented in onedirection, only to realize that such an orienta-tion involves insoluble problems. The casting isthen turned upside down and the exercise isrepeated in the hope of a better outcome. Suchexperiences are the day to day routine of thecasting engineer.

Furthermore, the complexity of the issues isnot easily solved at this time by computer. Therehave been many such attempts, but it is fair tosay at this stage such efforts have not beendeveloped to such a degree that the professionalcasting expert is offered any significant help.However, of course, we may look forward to theday when the computer can provide usefulsolutions.

10.6 Casting accuracy

There is, of course, very little reason to go togreat lengths regarding the provision of castinglocation points if the casting is hopelessly inac-curate in other ways. This section draws atten-tion to the general problems associated withcasting accuracy.

Any casting that we make is, in common withall other manufactured products, never quiteperfect in terms of size and shape. To allow forthis, tolerances are quoted on engineeringdrawings. So long as the casting is within tol-erance, it will be acceptable.

Some reasons for the casting being out oftolerance include elementary mistakes like thepatternmaker planting the boss in the wrongplace. This leads to an obvious systematic errorin the casting, and is easily recognized and dealtwith by correcting the pattern. It is an exampleof those errors that can be put right after thefirst sample batch of castings is made andchecked.

Another common systematic error in castingsis the wrong choice of patternmaker's contrac-tion allowance. The contraction of the castingduring cooling in the mould is often of the orderof 1 or 2 per cent. However, it depends stronglyon the strength of the mould. For instance, in anextreme case, a perfectly rigid mould will fix thecasting size; in such a situation the castingsimply would have to stretch during coolingsince it would be prevented from taking itsnatural course of contracting. To summarize,the choice of contraction allowance prior to themaking of the first casting is often not easy, andis often not exactly right. This point is taken upat length in Castings 2003, with recommenda-tions on how to live with the problem.

Other errors are less easily dealt with. Theseare random errors. No two nominally identical

castings are precisely alike. The same is true forany product, including precision-machinedparts. The ISO Standard (1984) for casting tol-erances indicates that although different castingprocesses have different capabilities for preci-sion, in general the inaccuracies of castings growwith increasing casting size, and the standardtherefore specifies increasing linear tolerances aslinear dimensions increase. (Nevertheless it isworth pointing out that the corresponding per-centage tolerance actually falls as casting sizeincreases.) Other work on the tolerancing ofcastings suggests that the ISO standard is still inits early days, and has considerable potential forfurther improvement (Reddy et al. 1988).

Because of the effects of random errors beingsuperimposed on systematic errors, it is ofcourse risky to attempt to correct the pattern-maker's error by moving the boss into anapparently correct location simply after theproduction of the first trial casting. Figure 10.6illustrates that the random scatter in positionsmight mean that the boss appeared to be in the

Figure 10.6 Statistical distribution of casting dimensions(a) before, and (b) after pattern development. Basedon Osborn (1979).



correct place first time if the casting happened tobe number 1 in Figure 10.6, or might have beenover twice as far out of place, compared to itsaverage position, if the first casting had beennumber 2. A sample of at least two or threecastings is really needed, and preferably ten or ahundred. The mean boss location and its stan-dard deviation from the mean position can thenbe known and the appropriate actions taken.

At the present time it will be of little surprise tonote that such exemplary action is not commonin the industry. This is because companies are notgenerally equipped with a sufficient number offast, automated three-coordinate measuringmachines. As such standards of measurementbecome more common, so the attainments interms of accuracy of castings will increase.

Finally, as we have seen, the ISO Standardgives the general trend of increase in the size ofrandom errors as casting size increases. How-ever, the casting designer and engineer requiremuch more detailed knowledge of the sources ofindividual contributions to the final total error.The remainder of this chapter is an examinationof these contributions. The reduction of theseerrors allows the production of castings that areconsiderably more accurate than the minimumaccuracy requirements of the InternationalStandard Organization document.

10.7 Tooling accuracy

Tooling is taken to include the pattern and itscoreboxes, or the die, and any measuring orchecking jigs and gauges.

The problem of constructing the pattern,allowing correctly for the contraction and dis-tortion of the casting, has already been dis-cussed and will not be dealt with here.

Patterns used in sand casting, and dies usedin die-casting, are subject to wear, so that thecasting gradually becomes oversize. Conversely,the tooling of many processes is also subject tobuild-up problems associated with the deposi-tion of small amounts of mould aggregate andbinder on the surface of the tooling, and thegradual accumulation of release agents that maybe used, causing parts of the casting to becomeundersize.

Distortion is another problem. Wood is auseful and pleasant material for pattern con-struction. It is easily worked, light to handle,and easily and quickly repaired or modified asnecessary. Even so, it is not a contender forreally accurate work because of its tendency towarp. A good patternmaker will attempt toreduce such movement to a minimum by thecareful use of ply and the alignment of the grain

of the wood, together with strengthening bat-tens. The use of various stabilized woods andsynthetic wood-like materials has also helpedconsiderably (Barrett 1967). Nevertheless theultimate stability in tooling is only achieved withthe use of metal, or cast resin that is properlysupported in metal frames. Cast-resin patterns,especially when cast into aluminium alloyframes for strength and rigidity, are usuallyextremely reliable. However, some resin systemssuch as polyurethanes tend to suffer from theabsorption of solvents from the chemical bin-ders in the sand, and so suffer swelling anddegradation (Gouwens 1967). Cast-resin pat-terns that are backed with wood frames are notreliable; the warpage of the wood distorts theinternal resin shape, usually within a month or so.After a year the tooling is seriously inaccurate,so that cores produced from such equipmentwill not assemble properly.

The working temperature of tooling affectsthe casting size directly; a warm pattern will givea slightly larger casting. If we consider an epoxyresin corebox cast into an aluminium alloyframe, the box will largely take its size from thetemperature of the metal frame (i.e. not theinternal lining of epoxy resin). If the tempera-ture at the start of the Monday morning shift is10�C, and if the returning sand creeps up to30�C by the end of the morning, then for a500 mm long casting the 20�C temperature risewill cause the castings to grow by 20� 20�10ÿ6� 500� 0.2 mm. This is not large in itself,but when it is added to other random variablesthe uncertainty in the final casting lengthbecomes increasingly out of control.

Anderson (1987) emphasized the importantrequirement that for the most accurate work thepattern or die should be utilized as an adaptivecontrol element in production. Thus it needs tobe built in such a way that it can be modified toproduce the required size and shape of thecasting. The use of patterns split transverselyacross their major length is common. The priorinsertion of a spacer in this split allows thespacer to be removed and replaced by some-thing thinner or thicker as necessary. Suchsimple techniques involve only modest extraexpense during the construction of the patternbut are a reassurance against the possibility ofmajor expensive rebuilds later.

10.8 Mould accuracy

Mould accuracy depends strongly on the mouldassembly method. Usually, a mould assemblysimply involves two parts; a cope and a drag.This makes for maximum accuracy. On other



occasions the mould assembly can be compli-cated, requiring many parts, and requiringmuch discussion between the pattern shop,foundry, and casting designer to find anappropriate solution. Accuracy can nowbecome illusive and troublesome.

As a general rule, it is useful to ensure thateven in the most complicated of mould assem-blies, the design of the assembly consists prin-cipally of a drag and a cope. In addition, allparts should be interrelated via a single mouldpart, which will normally be the drag. In thiscase the assembly of cores will be in the drag,with each core located separately to prints in thedrag, and the final operation will simply consistof closing with the cope. The cope also shouldhave features that contact and locate directly onthe drag as the key mould part.

Such simple rules are easily forgotten. Onecan see many ambitious castings that have failedto achieve dimensional acceptability because themould assembly has resembled a random heapof assorted blocks: one layer of cores totteringupon another, with finally the cope perchedprecariously on top. Clearly, the details of thecasting formed in the cope bear no constantrelation to the details formed in the drag.

The simplest form of construction of mouldsthat was widely popular consisted of a drag with acope in the form of a `cake core' as shown inFigure 10.5a. Accuracy was excellent. However,this simple construction did not allow for theplacement of any useful bottom-gated runningsystem, and the top pouring that had to beaccepted as a consequence might have been mar-ginally acceptable for some rough and ready greyiron castings in greensand moulds but did not givegood results for the majority of casting alloys.

Figure 10.5b shows a simple type of cope±drag arrangement with side cores, all located inthe drag, apart from side core S2, which rests onS1. It was judged that the small accumulation oferrors in the positioning of S2 would be accept-able in this case. If the positioning of S2 hadbeen critical, it could still have been located inthe drag by stepping the contour of the dragappropriately to form a convenient location.

Figure 10.5c shows how it would have beeneasy to have saved some sand by abandoningthe deep drag construction, and having a coreassembly that consisted simply of a pile of cores,rather than a proper cope and drag. The overallaccuracy of the casting now suffers from theaccumulation of errors introduced by the inter-mediate side cores S1, S2 and S3. There will, forinstance, be a poor match between the top andbottom features of the casting.

The dimensional problems that arise in set-ting cores are examined by Skarbinski (1971). In

general it needs to be said about the accurateprinting of cores that the core print should bedesigned assuming that the core will be pro-duced with errors in its size and shape, and willhave to fit into a mould which will also havesuffered some distortion during its manufacture.The print also sometimes has to restrict themovement of the core during mould fillingbecause of buoyancy, and yet may also have toallow the relative movement of the core in itsprint to permit the thermal expansion of thecore. All this is a seemingly impossible task toachieve accurately. However, it is usually solv-able by applying the following simple rules:

1. The print requires tolerance where it needs tofit (i.e. must not be made size-for-size, thatwould have produced an interference fit. Theonly exception to this is the heights of coreswhere cores are stacked one on top of theother, since in this case the accumulation oferrors requires to be kept to a minimum).

2. The print requires clearance where it is notrequired to fit, and where expansion clear-ance is required.

Rules often appear pedantic or even pedes-trian when they are spelled out! However, theapplication of the rules involves much work thatis seldom expended on the design of pattern-work. Although some prints are easily andquickly designed, others require lengthy agon-ized consideration resulting in compromises thathave to be carefully assessed. Every printrequires such detailed design. It is attentionto details such as these that makes the differencebetween the inadequate and the excellentcasting.

Nevertheless, these problems are eliminated ifthe use of the core can be avoided altogether.(The situation is reminiscent of Feeding Rule 1:Avoid using a feeder if possible. The usefulequivalent rule for cores, that makes an addi-tional rule for mould assembly, is:

3. The number of cores should be reduced to aminimum, moulding as much as possibledirectly in the cope and drag.

Not only does the application of this principlereduce dimensional errors, but also the additionof each core involves considerable extra toolingcost, and an additional cost in the production ofthe casting, sometimes approaching the costof the production of a cope or drag. The addi-tion of a core between a cope and drag repre-sents the third piece of sand to be added to thetwo original mould halves; thus the costs at thisstage may increase by 50 per cent for theaddition of the first core. At other times a small



core can save money by avoiding extra com-plexity of the tooling. Each case needs separateevaluation.

A high hidden addition to total casting costsresults from the use of cores. These difficult toassess costs arise from the accumulation of anumber of minor operations, most of which areusually overlooked. For instance, the core needsto be scheduled, made (perhaps on a capital-intensive core-blowing machine), deflashed,stored on special racks (taking up valuable floorspace), retrieved from storage, transported tothe moulding line, and then correctly assembledinto the mould. Errors arise as a result of theincorrect core being made or transferred, orsufficient are broken in storage or transit tocause the whole process to be repeated. Alter-natively, its assembly into the mould gets for-gotten at the last moment! Cores are thereforealmost certainly more expensive than mostfoundry accounting systems are aware of. (Thecosts of chills, and of scrapped castings, aresimilarly illusive and not widely understood.)

A further use of cores, in addition to theirobvious purpose in providing detail that cannotbe moulded directly, is that the running systemcan often be integrated behind and underneaththem, the main runner and gates being locatedbeneath a base, side or end core. This is avaluable facility offered by the use of a core andshould not be overlooked. In a number ofcastings the addition of a core may be for thesole purpose of providing a good running sys-tem. Such a core is often money well spent!

The problem with the automation of core-assembly systems is finding the core again afterit has been put down on, for instance, a con-veyor or a storage rack. This is a difficult job fora robot, since extreme accuracy is required, andthe cores are often of extreme delicacy. Clearly,one method of solving this problem is never toput the cores down in the first instance.

Schilling (1987) succeeded in developing thisconcept with a unique system of making andassembling cores in which the cores are notreleased from one half of the opened coreboxuntil the other half of the core has already beenlocated in the core-assembly package. In thisway the cores are assembled completely auto-matically and with unbelievable precision. Coresare located to better than 0.03 mm, allowingthem to be assembled with clearances which areso small that the cores could not be assembledby hand. In fact the cores are sprung into placewith interference fits. The rigorous applicationof this technique means that castings need to bedesigned for the process, since the assemblyof each core is by vertical placement over theprevious core. For instance, any threading of

cores in through holes in the sides of other cores,such as often occurs with port cores through thewater jacket core of a cylinder head casting, isnot possible. This disadvantage will limit thetechnique to partial application, loading somebut not all cores of a cylinder head, for instance.Even this would be an important advance.

A final note in this section relates to cope-to-drag location. This is, of course, of primaryimportance. Failure to achieve good locationresults in a mis-match defect. Mis-match is alateral location error, and not to be confusedwith the vertical precision with which cope anddrag meet, which is normally of the order of�0.05 to 0.10 mm.

In foundries using moulds contained inmoulding boxes, however, mis-match is unfor-tunately all too common and is usually the resultof the use of worn pins and bushes that are usedto locate the boxes. Southam (1987) analyses theeffect of the errors involved in the pin and bushlocation system. These are numerous and ser-ious. The pin-to-bush clearance is typically0.25 mm, and given an apparently acceptableadditional wear of 0.35 mm, he finds that thetotal possible mis-match between cope and dragmoulds is as much as 1.5 mm.

He proposes, therefore, a completely differ-ent system, in which pins and bushes are elimin-ated. The cope and drag boxes are simplyguided by wear blocks fixed to the outside edgesof the box. These slide against two guides on thelong side of the box, and one guide against thenarrow side of the box during moulding andclosing operations. The boxes are held againstthe guides by light spring pressure, or by pneu-matic cylinders. The system appears deceptivelysimple, but actually requires a certain amount ofgood engineering to ensure that it operatescorrectly on mould closure, as Southamdescribes. Although Southam calls his methodthe three-point registration system, it is in realitya classical six-point location system, since heuses a further three points to locate the drag in aparallel plane to the cope during closure.

The ability to locate cope to drag with neg-ligible error has a number of benefits thatSoutham lists. The maintenance and replace-ment of worn pins and bushings is a foundrychore and expense that is eliminated. (In factanyone who has not experienced the problem ofcarrying out such an operation in a jobbingfoundry will have a problem to comprehend theawesome scale of the task, because of the hun-dreds of pins and bushes, and the relentless wearproblem, requiring the operation to be repeatedat regular intervals despite the multitude ofpressing problems elsewhere in the foundryenvironment.) Instead, only three guides on the



closer and three each on the cope and dragpattern need to be checked, and the effect ofwear of these parts on mis-match is minimalbecause the resultant displacement is largelyself-compensating.

For the case of precision core packages, thesand mould is not contained in a box, and thus hasto be located directly witha sand-to-sand location.Since this is defined from the patternwork, thelocation relates perfectly to the casting details, andmismatch is therefore not possible.

10.9 Summary of factors affectingaccuracy

Some of the many factors that control theaccuracy of the final casting have been dealtwith above, and some are planned for inclusionin Castings processes to follow this publication.We shall therefore content ourselves here with abrief summary:

1. Pattern inaccuracy.2. Mould inaccuracy.3. Mould expansion and/or contraction because

of temperature and pressure.4. Casting expansion because of precipitation of

less dense phases such as graphite or gases.5. Casting contraction on freezing causing local

sinks.6. Casting contraction on cooling leading to

(a) different overall casting size, dependingon the constraint by the mould, and (b)distortion if unevenly constrained or un-evenly cooled.

7. Casting overall change of size on heat treat-ment or on slow ageing at room temperature.

8. Casting distortion if unevenly cooled by aninappropriate quenchant or too rapid quenchfrom heat-treatment temperature.

9. Casting distortion caused by shot blasting.This effect has not been dealt with previously.

The compressive stresses introduced into thesurface by a peening effect can lead to thedistortion of the casting as reported by Kaschand Mikelonis (1969). The effect is widelyused in the sheet metal industry to induce thecontrolled forming of curved surfaces; air-craft wing panels are formed from flat sheetsin this way; the flat product gradually curvesaway, becoming convex towards the direc-tion of the impingement of the shot. Con-trolled shot peening is also used to increasethe fatigue resistance of castings as discussedby Lawrence (1990) and O'Hara (1990).

10.10 Metrology

Even if it were possible to produce an absolutelyaccurate casting, it would not be possible toprove it! This apparently curious statement isthe consequence of errors that occur duringmeasurement. Inexact measuring of the castingwill cause apparent random deviations in thedimensions of the casting. Svensson and Villner(1974) point out this problem, and work outthe influence of measuring accuracy on theapparent dimensional accuracy of the casting.Table 10.1 is based on their work.

It is clear that even if the casting has dimensionsthat are quite correct, even careful measurementwill introduce a certain amount of apparent error,and careless measurement will, of course, introduceeven more. These errors have been a traditionalproblem within the industry but the introduction oflarge-size three-dimensional coordinate measuringmachines has significantly helped.

Even so, problems still remain. For instance,Swedish workers point out that for small dimen-sions, and where high accuracy is required, thesurface roughness will influence the apparentaccuracy of the casting. Thus a change in thesurface finish from 75 mm to 200 mm will give anincrease of one tolerance grade in the ISO system.

Table 10.1 Limits of accuracy of measurements

Measuring equipmentand range(mm)

Accuracy ofmeasurement(�mm)

Mean dimension(mm)

Accuracy(%)

ISO tolerancegrade IT

Steel tape >1000 1 1000 0.10 132000 0.05 125000 0.02 11

Steel rule 500±1000 0.5 500 0.10 121000 0.05 11

Vernier calliper 0±500 0.1 50 0.20 10100 0.10 10200 0.05 9500 0.02 9



The surface finish influences the measurementand location processes in other ways. For instance,the modern touch probes, which locate dimensionson the casting with the most delicate of contactpressure, effectively only measure to high spots,thus biasing the measurements in one direction:exterior dimensions on the casting are measuredoversize, and cored holes appear undersize.

Results from mark-out equipment using ascribing line tend to give more averaged results,since minor surface irregularities are cut through.

Similarly, when castings are clamped on totheir location points, the small area of the contactpoints, typically 5 mm diameter, and the highloads which can be exerted by the clamps, ensurethat the locating jig point actually indents thesurface of the casting by up to 0.25 mm for somealuminium alloy sand castings. Harder materialssuch as cast irons will, of course, indent less. Allsurface irregularities are effectively locallysmoothed and averaged in this operation. Theindentation effect sets an upper limit to theaccuracy and repeatability with which castingscan be picked up for measurement or machining.

A traditional method of checking the profile ofa casting is by the use of template gauges. Theseare typically sheets of metal that have been cut tothe correct contour. On applying them to thecasting, the contour on the casting can be seen tobe correct or not, depending on the clearance thatcan be seen between the two. This is an analoguetechnique that can no longer be recommended inthese modern times. The gauges are expensive tomake. They are also subject to wear, and thus needto be checked regularly and occasionally replaced.However, what is much more serious, they aredifficult to use in any effective way. This is becausein practice the contours never match exactly. Theproblem for the user then is how inaccurate canthe contour of the casting be allowed to becomebefore remedial action must be taken?

Theuseof `go/no-go'gaugesremoves thematterof judgement. However, the gauges are againsubject to wear, and thus require the costand complexity of a calibration system. Morefundamentally, their use is similarly not helpful interms of providing useful data to assist processcontrol.

All these difficulties can be removed by the useof a much simpler technique: the use of simplegoalpost fixtures that straddle the casting and areequipped with one or more spring-contact probes,such as dial gauges. The readings from the gaugesare read and recorded. The operation becomeseven simpler with the use of digital read-outdevices (Figure 10.7). Linear transducers areeasily fitted and operated, and give an immediatenumerical signal of the degree of inaccuracy.

The goalpost would be calibrated and storedon a standard casting, and thereby always be seento be in calibration by being set to zero in thisposition. (For calibration away from the zero,other readings can be obtained by the insertion ofslip gauges under the probe.)

The use of digital electronic read-out in thisway allows its incorporation into data-loggingand quality-monitoring systems, such as statis-tical process control. By watching the trends ona daily or weekly basis, the gradual drifts incasting dimensions can be used to predict, forinstance, that tooling wear will reach a level thatwill require the tooling to be replaced in threeweeks' time. Such prior warning allows theappropriate action to be planned well inadvance.

Figure 10.7 Comparison of checking techniques for themonitoring of the size and shape of castings by:(a) template, with the casting and template sat on abaseplate; (b) an equivalent analogue measurement usingspring dial gauges; and (c) digital measurement usinglinear displacement transducers, with the casting locatedon a six-point jig. The six-point locations for the goalpostframe on the jig are omitted for clarity.



Appendix

The 1.5 factor

Experimental results for side-gated 99.8Al platecastings plotted in Figure A1 show that castingtime tc may be estimated for the plates and othercastings from an equation;

tc � 1:5� casting volume/initial casting rate

(1)

This is equivalent to

initial fill rate/average fill rate � 1:5 (2)

These experimental results gives support tothe value of 1.5 chosen by previous authors,particularly those of the British Non-ferrousResearch Association (now no longer with us)researching for the UK Admiralty (ShipDepartment 1975).

Exploring the 1.5 factor further by a theore-tical approach is not quite so straightforward,but an attempt is outlined below.

Consider Figure A2, The velocity at the baseof the sprue is given by

V2 � �2gH�0:5

If the area of the base of the sprue is A2 andthe mould cavity is of uniform area AC theinitial velocity of rise in the mould will begiven by

Vi � �AC=A2� � �2gH�0:5

Similarly, at some later instant, when the melthas reached height h, the net head driving thefilling is now reduced to (Hÿ h) so that the rateof rise is now

100

10

1

0.11000100

Pourin

g time

(s)

10

Pou rin g rate per casting weight (ml/s .kg)

200 mm height100 mm height

Tot al casting

t = 1.5 × Vol ume/Initial pouring r ate

Figure A1 Experimental demonstration ofthe relation between initial and average fillingrates (Data from Runyoro and Campbell1992).


V � �AC=A2� � �2g�H ÿ h��0:5

Substituting dh/dt for the rate of rise V, re-arranging and integrating between the limits oftime t� 0 at h� 0, and t� tc at h� b, we find thecasting time (the time to fill the mould) tc isgiven by

tc � �Am=A2� � �2=g�0:5 �H0:5 ÿ �H ÿ b�0:5�Now writing simple definitions of the initial rateof casting Qi and the average rate of casting Qav insuch units as volume of liquid per second, definedby the appropriate velocity times the area, wehave

Qi � A2 � �2gH�0:5

Qav � Am � b=tc

It follows that

Qi=Qav � 2H0:5�H0:5 ÿ �H ÿ b�0:5�=b

This solution to the filling problem is interest-ing. There are various combinations of H andb that can fulfil the conditions defined bythe equation. For instance, if H� b, thenQi/Qav� 2, which is actually an obvious resultmeaning simply that the average is half of thestart and finishing rates.

On the other hand, Qi/Qav� 1.5 only whenb� 0.89H. This represents an intriguing result,indicating that for most castings the top of thepouring basin is on average only about 10 per centhigher than the height of the casting. Thus itseems the factor 1.5 is quite fortuitous, andresults simply from the geometry we happen toselect for most of the castings we make. If, ingeneral, we were to raise (or lower) our pouringbasins in relation to the tops of our castings, thefactor would have to be revised.

However, all is not so bad as it seems.Figure A3 shows that the factor 1.5 does notchange rapidly with changes in relative height ofbasin, varying over reasonable changes in basinheight of b/H from 85% to 95% from roughly1.55 to 1.60. These changes are of the sameorder as errors arising from other factors suchas frictional losses, etc. and so can be neglectedfor most practical purposes.

The Bernoulli equation

Daniel Bernoulli represents the revered name inflow. He published his equation in 1738 in oneof the first books on fluid flow. This magnificentresult is the one used for all descriptions of flowin pipes and channels. Whole books are devotedto its application.

There are of course, excellent examples of thepower of Bernoulli's equation. Sutton (2002)made good use of the equation to describe the

Safe working range

Minimum fill level used for calculation

Vel V

Vel Vi

Area A2

Area Am

Vel V2

h

b

H

Figure A2 Schematic view of the filling ofa uniform casting.

2.0

1.5

0 0.5 1. 01.0

(b/H )

Q i

Q av

Figure A3 The relation between the initial and averagefill rates for a uniform casting as a function of the relativeheights of the casting and the pouring basin.

Appendix 189


pressures along a long runner, explaining theearly partial filling of gates at different positionsalong the runner, and thus resulting in part-filled castings. He used the equation in itssimplest form, derived from a statement ofconservation of energy along a flow tube asillustrated schematically in Figure A4:

p1=rg � v21=2g � z1 � p2=rg � v2

2=2g � z2

� constant

where all the component terms of this equationhave units of length, conveniently metres. Forthis reason each term can be regarded as a`head'. Thus p/rg� pressure head, v2/2g�kinetic or velocity head, and z� potential orelevation head.

In application to the running system used bySutton (Figure A5) at location 1 the heightabove the centreline of the runner is 0.5 m, thekinetic head at this point is zero because themelt has zero downwards velocity, and the eleva-tion head z is considered zero because the run-ner is horizontal. At point 2, the pressure headrequires to be known, since this is the pressureraising the melt level in the vertical ingate. Theelevation is zero once again, and the velocityhead is close to 0.25 m, easily deduced from thetotal fall height and allowing for a small lossfactor of 0.70 (probably overestimated, since Ithink this should be more like 0.80 or even 0.85)as a result of the turn at the base of the sprue.Thus the Bernoulli equation becomes

0:5� 0� 0 � p2=rg � 0:25� 0

Thus p2/rg� 0.25 metre

It is not necessary to find p2 alone; the wholeterm is the height distance. Thus this answerwould be the same for aluminium or iron.

Sutton found that because of this kinetichead, ingates were filling before the runner wasfully filled. The first impression in his multi-impression mould was only about 200 mmabove the runner so that metal entered themould cavity under only about 50 mm net head.The result was a premature dribble into thecavity that quickly froze. The arrival of melt atthe intended full flow rate a few seconds laterwas too late to remelt and thus assimilate thefrozen droplets. An apparently mis-run castingwas the result.

In general however, the application of theBernoulli equation to filling systems is not quiteso straightforward as has sometimes beenassumed. There are various reasons for this.

1. In general, Bernoulli's equation relates tosteady state flow. However, of course, infilling systems most of the interest necessarilylies in the priming of the flow channels. Inthis situation the surface tension of theadvancing meniscus can be important, asenshrined in the Weber number. If thepriming is not carried out well, the castingis likely to suffer severely.

Hydraulic gradient

(Free liquid level)

Energy gradient

v12 /2 g

v22 /2 g

p1 /qg

z1

z2

p2 /qg

2

1

Figure A4 A pictorial representation of the termsof the Bernoulli equation.

1

2

h

h1 = 0.5

Figure A5 An example of the use of the Bernoulliequation by Sutton (2002) to calculate the rise of metalin a vertical gate.

190 Appendix


2. The surface tension of liquid metals is overten times higher than that of water, and evenhigher still compared to most organic liquids.Thus pressures due to surface tension havebeen neglected and are neglectable for suchcommon room temperature liquids on whichmost flow research has been conducted. Theadditional pressure generated because of thecurvature of the meniscus at the flow front,and the curvature at the sides of a flowstream affect the behaviour of metals inmany examples involved in the filling ofmoulds. For instance at the critical velocitythat is targeted in mould filling, the effectsof surface tension and flow forces are equal.At velocities lower than this, surface tensiondominates.

3. The presence of the oxide (or other thin, solidfilm) on the surface of an advancing liquidis a further complication, and is not easilyallowed for. The flow adopts a stick-slipmotion as the film breaks and re-forms. Theadvance of the unzipping wave is a classicinstance that could not be predicted by apurely liquid model such as that described byBernoulli.

4. The frictional losses during flow, that can beexplicitly cited in Bernoulli's equation, areknown to be important. However, in general,although they are assumed to be known, theyhave been little researched in the case of theflow of liquid metals. Furthermore, it isunfortunate that most of the research to datein this field has used such poor designs offilling systems that the existing figures arealmost certainly misleading. The losses needto be confirmed by new, careful, accuratestudies, supplemented by accurate computersimulation together with video X-ray radio-graphy of real flows.

5. The presence of oxide films floating about insuspension is another uncertainty that cancause problems. The density of such defectscan easily reach levels at which the effectiveviscosity of the mixture can be very muchincreased (although it is to be noted thatviscosity does not appear explicitly in theBernoulli equation). The suppression ofconvection in such contaminated liquids iscommon. Flow out of thick sections and intovery thin sections can be prevented comple-tely by blockage of the entrance into the thinsection.

From the above list it is clear that the applica-tion of Bernoulli is more accurate for thickersection flows where surface effects and internaldefects in the liquid are less dominant. As fillingsystems are progressively slimmed, and casting

sections are thinned, Bernoulli's equation has tobe used with greater caution.

As a result of the problems of the applica-tion of Bernoulli to the priming of the fillingsystem, it has been relatively little used in thisbook because the concentration of effort hasfocused on the control of the priming of thesystem. The subsequent flow of the systemwhen completely filled, as nicely described byBernoulli, is, with the greatest respect to theGreat Man and his magnificent equation, muchless important.

Rate of pour of steel castingsfrom a bottom-pour ladle(Figures A6, A7, A8)

All three parts of the nomogram have to beused in conjunction to obtain the time of pourof a casting. The reduced scale illustration ismerely to show the intended arrangement of thethree components of the nomogram. The fol-lowing is an example of how the nomogramis used.

A ladle contains 5000 kg of steel, from whichwe wish to pour a casting of total weight1250 kg. Thus we follow the arrows from thesestart points to the junction A. From here ahorizontal line connects to the next figure,where we select a pouring nozzle for the ladleof 60 mm diameter. At this junction B we dropa vertical line down to intersect with the linedenoting that our ladle is about 1.5 m internaldiameter. From this junction C we continuewith a parallel line to the family of sloping lines,to find that our casting will pour in approxi-mately 23 seconds.

Interestingly, the reader can check that thenext 1250 kg casting in line (now starting witha ladle of 5000ÿ 1250� 3750 kg will be foundto pour in about 29 seconds, and the next in34 seconds, and the next in 77 seconds, as theladle progressively empties.

Running system calculation record

A typical work sheet record (Table A1) for thecalculation of a filling system. Successive itera-tions may be required after initial trials becauseof unforeseen reasons that can be noted andused for the tuning of subsequent filling systemdesigns.

The system below can, of course, be usedmost effectively in simple spread-sheet format.

Appendix 191


Table A1 Running system record

Casting name

Part number

Customer Alloy

Design 1signed Date

Design 2signed Date

Design 3signed Date

Casting weight�Mc kg

Rigging system weight�Mr kg

Total poured weight kg

Mc�Mr�M

Select fill time� t s

Average mass flow rate�M/t kg sÿ1

Liquid metal density� � kg mÿ3

Average volume flow rate�M/t� m3 sÿ1

Initial volume flow rate Q� 1.5M/t� m3 sÿ1

Height of liquid in basin� h m

Velocity into sprue entrance V1� (2gh)1/2 m sÿ1

Area of sprue entrance�Q/V1 m2

Total casting height�H m

Velocity at base of sprue V2� (2gH)1/2 m sÿ1

Area of sprue base A2�Q/V2 m2

Radius of turn at sprue base R� (A2)1/2 m

Area of runner�A2 or possibly up to 1.2A2 m2

Select critical velocity Vc� 0.5 to 1.0 m sÿ1 m sÿ1

Total area of gates A3�A2 �V2/Vc m2

Min fill depth of basin� h m

Basin depth� 2h to 4h for safety m

Basin volume for 1 s response�Q m3

Basin sides, if square, for 1 s response� (Q/h)1/2 m


50 100

200

500 1k 2k 5k 10k

20k

50k

100k

200k

300k

10 20 40 60 100Discharge time (s)

200 400600 1000 2000

0.5

1.0

Mean

inte

rnal

diam

eter

of la

dle (

m)

1.5

2.0

2.5

Wei ght of metal in ladl e prior to pour (kg)

120

100

8060

15

5040

3025 20

Nozzle

diam

(mm)

50

100 15

0 200 30

0 400

500 10

00 1500 20

00 3000 40

0050

00 10k 20

k 30k

40k

50k 10

0k

150

k

Wei ght of metal to b

e poured fro

m ladle (kg)

AB

C

Figures A6, A7 and A8 Rate of delivery of steel from a bottom-pour ladle.

Ap

pen

dix

193


Appendix A9

Design methodology for investmentcastings (Figure A9)

The wax assembly is required to be strong towithstand the shell-making process, so thatgood bottom-filling systems are not easilyapplied to investment castings. Here the mainsupport column forms a central feeder, filledupwards because of the ceramic disc at the baseof the pouring basin. The basin is sharp edgedand offset from the sprue. The sprue is cut fromsheet wax, is only a few millimetres thick and sois relatively flimsy, and not expected to conferany strength on the whole assembly. The sprueturns and expands under the filter to maximizecoverage. The filter is tangentially placed, withsome scope for the sideways displacement of

bubbles or other buoyant material into a trap.Small `feet' at the ends of the radial runnershelp the assembly to stand upright duringprocessing. Other small extensions from thetops of the castings can form vents for dewax-ing and/or escape of entrapped gas in themould during pouring. The breaking open ofthe shell into the top of the vents will allowmost effective venting. Connection of the ventsto the top of the pouring basin can be used tostrengthen the assembly if necessary, andplugged with fireclay to prevent the accidentalingress of metal. Most of the Rules are fulfilledby this design. Convection is controlled by theavoidance of circular flow paths and rapidcooling of the casting because of their thinsection and their placement around the outsideof the assembly where they can radiate awayheat effectively.

h1

h2

max

min

Figure A9 Investment casting, design methodology.

194 Appendix


References

Adams A; 2001 Modern Castings 91 (3) 34±36.Adams C M and Taylor H F; 1954 TAFS 61 686±693.Ali S, Mutharasan R and Apelian D; 1985 Met Trans

16B 725±742.Anderson G R; 1985 Patt-Tech 85 Conference, Oxford,

UK.Anderson G R; 1987 TAFS, 95 203±210.Ashton M C and Buhr R C; 1974 Phys. Met. Div.

Internal Report PM-1-74-22 Canada Dept EnergyMines and Resources.

Askeland D R and Holt M L; 1975 TAFS 83 99±106.Barrett L G; 1967 TAFS, 75 326±329.Bates C and Wallace J F; 1966 TAFS, 74 174±185.Beck A, Schmidt W, and Schreiber O; 1928 US Patent 1,

788, 185.Beeley P R; 1972 Foundry Technology Butterworths,

London.Berry J T and Taylor R P; 1999 TAFS 107 203±206.Bex T; 1991 Modern Casting November p 56 only.Bird P G; 1989 Foseco Technical Service Report

MMP1.89 `Examination of the factors controllingthe flow rate of aluminium through DYPUR units'.1 March 1989.

Biswas P K, Rohatgi P K and Dwarakadasa E S; 1985 BrFoundryman 78 511±516.

Bishop H F, Myskowski E T and Pellini W S; 1955 TAFS63 271±281.

Bossing E; 1982 TAFS 90 33±38.Bromfield G; 1991 The Foundryman July 261±265.Brown N and Rastall D; 1986 European Patent Appli-

cation No 87111549.9 filed 10.08.87. pp 9.Campbell J; 1991 Castings Butterworth Heinemann.Campbell J and Caton P D; 1977 Institute of Metals

Conference on Solidification Sheffield, UK, 208±217.

Campbell J and Isawa T; 1994 UK Patent GB 2,284,168 B.(Filed 04±02±1994).

Chvorinov N; 1940 Giesserei, 27 177±186, 201±208,222±225.

Clegg A L and Das A A; 1987 Br Foundryman, 80137±144.

Cox B M, Doutre D, Enright P and Provencher R; 1994TAFS 102 687±692.

Creese R C and Safaraz A; 1987 TAFS 95 689±692.Creese R C and Safaraz A; 1988 TAFS 96 705±714.Creese R C and Xia Y; 1991 TAFS 99 717±727.Cunliffe E; 1994 PhD Thesis. University of Birmingham,

UK.Datta N and Sandford P; 1995 3rd AFS Internat

Permanent Mold Casting of Aluminum Conferencepaper 3 pp 19.

Davis K G; 1977 AFS Int Cast Metals J, March, 23±27.Davis K G and Magny J -G; 1977 TAFS, 85 227±236.Daybell E; 1953 Proc Inst Br Foundryman 46 B46-B54.Din T and Campbell J; 1994 University of Birmingham

UK unpublished work.Din T, Kendrick R and Campbell J; TAFS 2003 111

paper 03-017.Durville P H G; 1913 British Patent 23,719.Eastwood L W; 1951 AFS Symposium on Principles of

Gating 25±30.Emamy M, Taghiabadi R, Mahmudi M and Campbell J;

2002 Statistical study of tensile properties of A356aluminum alloy, using a new casting design. ASM 2ndInt Al Casting Technology Symposium 7±10 October2002 Columbus Ohio.

Evans J, Runyoro J and Campbell J; 1997 SP97 4thDecennial International Conference on SolidificationProcessing, Sheffield, 74±78.

Flemings M C; 1974 Solidification Processing McGrawHill.

Forslund S H C; 1954 21st Int Foundry Congress,Florence, paper 15.

Gebelin J -C and Jolly M R; 2003 Modeling of Casting,Welding and Solidification Processing Conference X,Destin USA.

Gebelin J -C Jolly M R and Jones S; INCAST 13 (11)(Dec) 22±27.

Geskin F S, Ling E and Weinstein M I; 1986. TAFS 94155±158.

Gouwens P R; 1967 TAFS 75 401±407.Grote R E; 1982 TAFS 90 93±102.


Groteke D; 2002 Internat J Cast Metals Research 14(6)341±354.

Halvaee A and Campbell J; 1997 Trans AFS 105 35±46.Hansen P N and Rasmussen N W; 1994 BCIRA Inter-

national Conference, University of Warwick, UK.Hansen P N and Sahm P R; 1988 Modelling of casting,

welding and advanced solidification processes IV(Giamei A F; Abbaschian G J, editors) The Mineral,Metals and Materials Society.

Hayes K D, Barlow J O, Stefanescu D M and Piwonka T S;1998 TAFS 106 769±776.

Hedjazi D, Bennett G H J and Kondic V; 1975. BrFoundryman 68 305±309.

Heine R W; 1982 TAFS 90 147±158.Heine R W and Uicker J J; 1983 TAFS 91 127±136.Hess K; 1974 AFS Cast Metals Research J March 6±14.Heyn E; 1914 J Inst Metals 12 3.Hiratsuka S, Niyama E, Anzai K, Horie H and Kowata T;

1966 4th Asian Foundry Congress 525±531.Hirt C W; 2003 www.flow3d.comHodjat Y and Mobley C E; 1984 TAFS 92 319±321.Huang H, Lodhia A V and Berry J T; 1990 TAFS 98

547±552.IBF Technical Subcommittee T571; 1969 Br Foundryman

62 179±196.IBF Technical Subcommittee TS71; 1971 Br Foundryman

64 364±379.IBF Technical Subcommittee T571; 1976 Br Foundryman

69 53±60.IBF Technical Subcommittee T571; 1979 Br Foundryman

72 46±52.Isawa T; 1993 PhD Thesis University of Birmingham

Department of Metallurgy and Materials Science`The control of the initial fall of liquid metal ingravity filled casting systems'.

Isawa T and Campbell J; 1994 Trans Japan Foundry-men's Soc 13 (Nov) 38±49.

ISO Standard 8062; 1984 CastingsÐSystem of Dimen-sional Tolerances.

Itamura M, Murakami K, Harada T, Tanaka Mand Yamamoto N; 2002 Int J Cast Met Res 15167±172.

Itamura M, Yamamoto N, Niyama E and Anzai K; 1995Proc 3rd Asian Foundry Congress, eds Lee Z H,Hong C P, Kim M H. 371±378.

Jacob S and Drouzy M; 1974 Int Foundry Congress 41Liege, Belgium. Paper 6.

Jeancolas M and Devaux H; 1969 Fonderie 285 487±499.Jeancolas M, Devaux H and Graham G; 1971 Br

Foundryman 64 141±154.Jirsa J; 1982 Foundry Trade J Oct 7, 520±527.Johnson S B and Loper C R; 1969 TAFS 77 360±367.Johnson W H and Baker W O; 1948 TAFS 56 389±397.Jolly M J, Lo H S H, Turan M, Campbell J and Yang X;

2000 Modeling of Casting, Welding and AdvancedSolidificationProcessingIXAachen,editorsSahmPR,Hansen P N, Conley J G.

Kahn P R, Su W M, Kim H S, Kang J W and Wallace J F;1987 TAFS 95 105±116.

Karsay S I; Ductile Iron Production 1; state of the art1992. Published by QIT Fer et Titane Inc 1992 and`Ductile Iron; the essentials of gating and risering

system design' revised 2000 published by Rio TintoIron & Titanium Inc.

Kasch F E and Mikelonis P L; 1969 TAFS 77 77±89.Kim M H, Moon J T, Kang C S and Loper C R; 1993

TAFS 101 991±998.Kim M H, Loper C R and Kang C S; 1985 TAFS 93

463±474.Kim S B and Hong C P; 1995 Modeling of casting,

welding and advanced solidification processes VIITMS 155±162.

Kono R and Miura T; 1975 Br Foundryman 69 70±78.Kotschi T P and Kleist O F; 1979 AFS Int Cast Metals J

4 (3) 29±38.Kotschi R M and Loper C R; 1974 TAFS 82 535±542.Kotzin E L; 1981 Metalcasting and Molding Processes.

American Foundrymen's Soc., Des Plaines, Illinois,USA.

Kubo K and Pehlke R D; 1985 Met Trans 16B 359±366.Kunes J, Chaloupka L, Trkovsky V, Schneller J and

Zuzanak A; 1990 TAFS 98 559±563.Kuyucak S; 2002 TAFS 2002 110.Lai N-W and Griffiths W D; 2003 University of

Birmingham, UK. To be published.Laid E; 1978 US Patent Application 16 Feb 1978

Number 878,309.Latimer K G and Read P J; 1976 Br Foundryman 69

44±52.Lawrence M; 1990 Modern Castings February 51±53.Lerner Y and Aubrey L S; 2000 TAFS 108 219±226.Lewis R W and Ransing R S; 1998 Met and Mat Trans B

29B(2) 437±448.Lewis R W, Ransing M R and Ransing R S; 2002

Internat J Cast Metals Research 15(1) 41±53.Lin H J and Hwang W-S; 1988 TAFS 96 447±458.Lo H and Campbell J; 2000 Modeling of Casting,

Welding and Advanced Solidification Processes IX EdsSahm P R, Hansen P N, Conley J G. pp 373±380.

Loper C R; 1981 TAFS 89 405±408.Loper C R, Javaid A and Hitchings J R; 1996 TAFS 104

57±65.MaedaY,NomuraH,OtsukaY,TomishigeHandMoriY;

2002 Int J Cast Met Res 15 441±444.Mampaey F and Xu Z A; 1999 TAFS 107 529±536.Manzari M T, Lewis R W and Gethin D T; 2000

`Optimum design of chills in sand casting process'Proc IMECE 2000 Int Mech Eng Congr Florida, USADETC98/DAC-1234 ASME pp 8.

McDavid R M and Dantzig J; 1998 Met Mater Trans29B 679±690.

McKim P E and Livingstone K E; 1977 TAFS 85491±498.

Mertz J M and Heine R W; 1973 TAFS 81 493±495.Mi J, Harding R A and Campbell J; Int J Cast Met Res

2002 14(6) 325±334.Midea A C; 2001 TAFS 109 41±50 and Foundryman 2003

(March) 60±63.Mikkola P H and Heine R W; 1970 TAFS 78 265±268.Miles G W; 1956 Proc Inst Br Foundryman 49 A201±

210.Minakawa S, Samarasekera I V and Weinberg F; 1985

Met Trans 16B 823±829.Mollard F R and Davidson N; 1978 TAFS 86 479±486.

196 References


Momchilov E; 1993 J Mater Sci Technol, Institute forMetal Science, Bulgarian Academy of Sciences, Sofia.1(1) 5±12.

MorthlandTE,ByrnePE,TortorelliDAandDantzigJA;1995 Metall Mater Trans B 26B 871±885.

Mutharasan R, Apelian D and Romanowski C; 1981J Metals 83(12) 12±18.

Mutharasan R, Apelian D and Ali S; 1985 Met Trans16B 725±742.

Nguyen T and Carrig J F; 1986 TAFS 94 519±528.Nikolai M F; 1996 TAFS 104 1017±1029.Noguchi T, Kano J, Noguchi K, Horikawa N and

Nakamura T; 2001 Int J Cast Met Res 13 363±371.Nyamekye K, An Y-K, Bain R, Cunningham M,

Askeland D and Ramsay C; 1994 TAFS 102 127±131.O'Hara P; 1990 Engineering, February, 41±42.Osborn D A; 1979 Br Foundryman 72 157±161.Pellini W S; 1953 TAFS 61 61±80 and 302±308.Pillai R M, Mallya V D and Panchanathan V; 1976

TAFS 84 615±620.Prodham A, Carpenter M and Campbell J; 1999 CIATF

Technical Forum.Rabinovich A; 1969 AFS Cast Met Res J March 19±24.Rao G V K and Panchanathan V; 1973 Cast Met Res J

19(3) 135±138.Rao T S V and Roshan H Md.; 1988 TAFS 96 37±46.Rao G V K, Srinivasan M N and Seshadri M R; 1975

TAFS 83 525±530.Rashid A K M B and Campbell J; 2004 Oxide defects in

a vacuum-investment-cast Ni-base turbine blade.Accepted for publication in Met and Mat Trans 2004.

Rasmussen N W, Aagaard R and Hansen P N; 2002Internat J Cast Metals Research 14(6) 325±383.

Reddy D C, Murty S S N and Chakravorty P N; 1988TAFS 96 839±844.

Rezvani M, Yang X and Campbell J; 1999 TAFS 107181±188.

Richins D S and Wetmore W O; 1951 AFS Symposiumon Principles of Gating 1±24.

Roedter H; 1986 Foundry Trade J Int Sept 1986 6.Rogers K P and Heathcock C J; 1990 US Patent

5,316,070 date 31 May 1994.Romero J M, Smith R W and Sahoo M; 1991 TAFS

99, 465±468.Ruddle R W; 1956 `The Running and Gating of Sand

Castings' Monograph and Report Series No 19,Institute of Metals, London.

Ruddle R W and Cibula A; 1956±57 J Inst Metals 85265±292.

Runyoro J; 1992 PhD Thesis `Design of the GatingSystem', University of Birmingham.

Runyoro J, Boutorabi S M A and Campbell J; 1992TAFS 100 225±234.

Runyoro J, Boutorabi S M A and Campbell J 1992 TransAFS 100 225±234.

Runyoro J and Campbell J; 1992 The Foundryman85(April) 117±124.

Safaraz A R and Creese R C; 1989 TAFS 97 863±870.Sandford P; 1988 The Foundryman (March) 110±118.Sandford P; 1993 TAFS 101 817±824.Sciama G; 1974 TAFS 82 39±44.

Sciama G; 1975 TAFS 83 127±140.Sciama G; 1993 TAFS 101 643±651.Schilling H; 1987 Patent PCT WO 87/07543.Schmidt D G and Jacobson A E; 1970 TAFS 78 332±337.Sexton A H and Primrose J S G; 1911 The Principles of

Ironfounding The Technical Publishing Co., London.Ship Department Publication 18; 1975 Design & Manu-

facture of Nickel-Aluminium-Bronze Sand CastingsMinistry of Defence (Procurement Executive) Foxhill,Bath, UK.

Sirrell B and Campbell J; 1997 TAFS 105 645±654.Skarbinski M; 1971 Br Foundryman 44 126±140.Southam D L; 1987 Foundry Manage Technol 7 34±38.Stahl G W; 1961 TAFS 69 476±469.Stahl G W; 1963 TAFS 71 216±220.Stahl G W; 1986 TAFS 94 793±796.Stahl G W; 1989 `The gravity tilt pour process'; Proc AFS

Int Conf: Permanent Mold Castings, Miami. Paper 2.Steel Founders Society of America (Anon) 2000 Foundry

Trade J May 40±41.Stefanescu D M, Giese S R, Piwonka T S, Lane A M,

Barlow J and Pattabhi R; 1996 TAFS 104 1233±1264.Sullivan E J, Adams C M and Taylor H F; 1957 TAFS 65

394±401.Sun L and Campbell J; 2003 TAFS paper 03-018 in press.Sutton T; The Foundryman 2002 95 (June) (6) 223±231.Suzuki S; 1989 Modern Castings, October, 38±40.Svensson I L and Dioszegi A; 2000 Modelling of Casting,

Welding and Advanced Solidification Processes IX EdsSahm P, Hansen P and Conley 102±109.

Svensson I and Villner L; 1974 Br Foundryman 67277±287.

Swift R E, Jackson J H and Eastwood L W; 1949 TAFS57 76±88.

Swing E; 1962 TAFS 70 364±373.Taylor K C and Baier A; 2003 Casting Plant�Technol

Int 123 (2) 36±46.Taghiabadi R, Mahmoudi M, Emamy Ghomy M and

Campbell J; 2003 Materials Science and Technology19 497±502.

TiryakiogÆlu E; A Study of the Dimensioning of Feedersfor Sand Castings, PhD Thesis, University ofBirmingham, UK, 1964.

TiryakiogÆlu M, TiryakiogÆlu E and Askeland D A; 1997aInt J Cast Met Res 9 259±267.

TiryakiogÆlu M, TiryakiogÆlu E and Askeland D A; 1997bTAFS 907±915.

Tiryakioglu M, Tiryakioglu E and Campbell J; 2002Internat J Cast Metals Research 14(6) 371±375.

Trojan P K, Guichelaar P J and Flinn R A; 1966 TAFS74 462±469.

Turner G L; 1965 Br Foundryman 58 504±505.Turner A and Owen F; 1964. Br Foundryman, 57 55±61,

and 355±356.Villner L; 1969 Br Foundryman 62 458±468. Also pub-

lished in Giesserei 1970 57(27), 837±844 and Cast MetRes J 1970 6 (3) 137±142.

Wall A J and Cocks D L; 1980 Br Foundryman 73292±300.

Ward C W and Jacobs I C; 1962 TAFS 70 332±337.Webster P D; 1967 Br Foundryman 60 314±319.

References 197


Wen S W, Jolly M R and Campbell J; 1997 Proc 4thDecennial Int Conf on Solidification Processing,Sheffield. Eds Beech J and Jones H. 66±69.

Weins M J, de Botton J L S and Flinn R A; 1964 TAFS72 832±839.

Wieser P F and Dutta I; 1986 TAFS 94 85±92.Wright T C and Campbell J; 1997 TAFS 105 639±644

and Modern Castings June 1997 (see also T Dimmick2001 Modern Castings March 91 (3) 31±33).

Xu Z A and Mampaey F; TAFS 105 853±860.Yang X and Campbell J; 1998 Int J Cast Met Res 10

239±253.

Yang X, Dai X, Campbell J and Wood J; 2003 MaterialsScience and Engineering A354(1±2) 315±325.

Yang X, Din T and Campbell J; 1998 Int J Cast Met Res11 1±12.

Yang X, Jolly M R and Campbell J; 2000 AluminumTrans 2(1) 67±80 and Modelling of Casting, Weldingand Advanced Solidification Processing IX 2000 EdsSahm P, Hansen P and Conley 420±427.

Zadeh A H and Campbell J; 2002 TAFS paper 02±020.Zhang C, Mucciardi F, Gruzleski J, Burke P and Hart M;

2003 AFS Trans paper 010-03.

198 References


Index

Accuracy, 182±6

casting, 182±3, 186

tooling, 183

metrology, 186

mould, 183±6

Additive contamination, 6

Aerofoil section venting, 76

Aggregate moulding:

move away from silica sand, 124

penetration by liquid metal, 134

stability of cast products, 166±7

Al±Mg alloy, 122

Al±Si alloy, 121±2

Alotech Process, 4, 74±5

Atmospheric pressure, 123

AOD, see Argon

Argon:

argon±oxygen-decarburization, 5

degassing of Al, 5

shroud, 26±7

Asymmetric bifilm, 103

Barometric heights for liquid metals, 139

Basalt, 16

Basin (pouring):

conical, 25±7, 111

design depth, 30±1, 112, 188±9

offset, 27

offset step (weir), 28, 111

Bernoulli's theorem, 49, 189±91

Bifilms:

action as cracks, 78

asymmetric, 103

content in steels versus Al alloys, 171

controlling morphology of shrinkage, 121

definition, 11

effect on melt viscosity, 162

entrainment with bubbles, 78

initiation of pores, 1, 5, 125

flotation by rotary degassing, 5, 162

opening agents, 1, 125, 140, 146

populations, 2, 125±6, 140

size, 83

unfurling, 125, 140, 146

Blowholes, see Blows and Bubbles

Blows:

from chills, 117, 150, 155

from cores, 114±9, 126, 158

from core adhesive, 181

from core repairs materials, 116

from moulds, 115

pressurisation to suppress blow, 128

trigger for convection, 158

Bond development with cast-in inserts, 150

Bottom filling, bottom gating, 16

Bottom teemed ladle, 30±3, 102, 191±3

Boxless moulds, 63

Breaker core see Washburn core

Bronze, 12, 16, 54, 69, 116, 164

Bubble:

blow defect, 114

carbon monoxide, 5

core blows, 114±9

entrained, 4, 20, 78, 108

trap, 85

Bubble damage, 26, 40, 44, 78,

108±13, 114

definition, 108

non-uniform distribution, 109

Bubble trails:

closed, 4, 20, 44, 108, 114

detached, 109

open (re-inflation by pressure reduction), 110

temporary re-inflation by core blow, 114

Burned-on sand, 134

Bush, see basin

By-pass, 58, 100±1


Calcium silicide, 2

Calculation:

filling system design, 93±101, 192

record, 192

Carbon boil, 5

Carbon monoxide, 5, 117

Cast iron feeding, 128

Cavitation (pompous and incorrect term for

surface sink)

Centrifugal casting, 54, 55, 76

Centrifugal trap, see Swirl trap

Ceramic tube running system, see preformed

channels, 36, 41

Chills:

bifilm redistribution effects, 146

ceramic wash coat, 118

cooling effect, 127, 133, 146±50

feeding distance effect, 135±7

segregation promotion by chills, 163±4

segregation reduction by chills, 165

Choke:

sprue exit, 23±4, 36±7, 98

runner entrance, 41

Chromium alloyed cast iron, 23

Chvorinov, 126

Clamping points, 177±8

Clay-based core repair pastes, 116

Cobalt aluminate, 159

Coffee cup experiment, 138±9

Cold shut, see Lap, cold

Computer simulation

critical tests, 145, 151

flow simulation, 100

limitations, 124, 160, 181±2

outgassing pressure simulation, 119

segregation prediction, 164±5

solidification optimisation, 133, 149, 155

solidification simulation, 94, 123, 124, 127, 130,

145, 146

stress prediction, 167

surface turbulence prediction, 102

Condensation on chills, 150, 155

Conductivity, thermal, 147, 168

Contact pouring, 27

Continuous casting:

DC (direct chill) Al casting, 7, 168

Ni-base horizontal, 3

roll casting analogue, 54

Convection, 101, 124 (twice), 146, 152, 157±62

Cooling fin, see Fin

Copper-based alloys, 12, 164

feeding distance, 136, 137

fin effect, 52, 128 (twice)

mould rigidity and casting soundness, 124

Core:

adhesive, glue, 181

assembly, 181, 183

assembly automation, 185

assembly jig, 181

blows, see Blows

cake core, 184

costs, 184±5

elimination if possible, 184

print design, 184

repair pastes, 116

Costs, 10,

Cosworth process:

accuracy of pick-up, 176

angled filling possibility, 105

bubble entrainment danger, 112

convection control/roll over, 105, 157, 158, 162

counter gravity control, 74±75

critical velocity experience, 10

gating design, 52

grain refinement experience, 6

metal handling and quality, 4 (twice), 12

Counter gravity casting, 9, 10, 19, 49, 72±5, 145, 157±61

yield benefit, 128, 145±6

Cracks (edge), 146

Criteria functions, 137

Critical fall height, 11, 13, 33, 72

Critical ingate velocity, 17±19, 191

Critical temperature range during quenching, 169

Crystal lattice, 120

Cycle time for gravity die, 122

Cylinder heads and blocks, 110, 161, 168±9, 176,

177, 185

Dam, 29

Datum planes, 175±6

Degassing:

Al alloys, 5

chemical fixation, 4, 5

Dendrite arm spacing, 109, 163

Dendritic advance of flow front, 105

Deoxidation of steel, 5, 86, 113

Die castings (USA, see Pressure die castings)

Die, see also Gravity, Low Pressure, Pressure Die

coat, 124

cooling, 125

Diffraction mottle, 140

Diffusion bonding, 150

Diffusion distance:

heat, 168

solute, 163

Diffusivity:

heat, 147±148

thermal, 147, 168±169

Dilation of mould/casting, see Mould rigidity

Dimensional control, 126 (twice), 170, 172±173

Directional solidification, 133, see also Heuvers

Discharge coefficient, 28±29

Distortion:

casting, 172±173, 186

shot blasting, 186

tooling, 183

Dross, 117

200 Index


Dross trap, 59, 78±79, 125

Ductile iron, 2, 23, 48, 86, 92, 131, 141±142

Ductility, 146

Dump, see Dross trap

Durville, 69

Dye penetrant test, 11, 106

Economics, see Costs

Edge cracking of strip and plate, see Cracks

Ejection time control of casting size, 175

Electromagnetic inclusion control, 82

Electromagnetic pump, 10, 12

Electromagnetic linear motor, 105

Elevation head, 190

Entrainment, 11, 108

Exfoliation defect, 114

Eye-dropper ladle, see Snorkel

Fatigue, 170, 171, 186

Feeder:

annular, 42

blind, 138±139

cast iron uncertainties, 128, 131

definition as reservoir, 120

distance, 135±7

efficiency, 128

hydrostatic pressure delay, 126±7

hydrostatic pressure effect, 119, 125, 127, 137,

139, 145

hydrostatic pressure relief, 142

junction effect, 51, 132±3

neck, 134, 142

non-feeding roles, 125

optimum, 142±5

oversize, 122

pasty zone/casting section ratio effect, 137

pressure relief valve effect for ductile iron, 142

pressurization, 137, 141±42

reverse tapered for ductile iron, 129±32

safety factor, 127, 132±133

segregation problems, 164

sleeves insulating and exothermic, 127

top, 18

under feeder porosity, see also Undersize, 141

under feeder segregation, 164±5

undersize, 122, 132, 143

up-runner and side gate, 61

Feeding mechanisms, 123

solid, 123, 125

Feeding rules, 123, 126±42

1. Do not feed, 126

2. Heat transfer requirement, 127

3. Volume requirement, 128

4. Junction requirement, 132±3

5. Feed path requirement, 133±8

6. Pressure gradient requirement, 138±40

7. Pressure requirement, 140±2

Feeding techniques:

active, 123, 145±6

gravity, 123±124, 132, 139, 158

uphill dangers, 123±4, 132, 138±9, 146, 157

via filling system, 121, 127, 131

Ferro silicon, 2

Filling and feeding definitions, 120, 121

Filling system definition, 16

Filling system design, 93, 192

Fill rate:

bottom pour ladle, 191±3

control, 74

selection, 60±61, 63, 95±7

Films, see also Bifilms

graphitic (lustrous carbon), 23, 110

effect on flow, 191

silicate, 23

Filters, 7±8, 82±93, 113

blockage, 34

ceramic, 85

cloth, 83±5, 140

extruded, 83, 86

feeder removal, 140

flow rate, 7, 43

fluidity benefits, 93

foam, 86

freezing and Remelting, 86

packed bed type, 7

pressed, 83, 86

pressure drop, 89

tangential, 90±91

sintered, 86

speed reduction, 99

steel mesh, 83±5, 93

strainers, 82±83

with reverse tapered sprue, 37, 43

Fins:

advantages, 154±155

effect of conductivity of alloy, 128, 151, 155

general, 101, 124, 127, 133, 146, 150±6

solid fins, 152, 155

Flash, see also Fins, 124

Flaskless, see boxless

Flow:

channel, 60, 161

front instability, 14

priming of channels, 190

Flow-off:

action of feeder, 125

runner see by-pass

Flow tube, see Oxide flow tube

Fluidity:

controlled filling, 63

enhancement, 64

Forging, 146

Fraction solid fS, 137

Free-riders, 59

Frictional losses during flow, 191

Furan resin binder, 63

Furnaces melting/holding, 3

Index 201


Galileo, 97

Gas:

bubble, confusion with shrinkage, 110

see hydrogen

solution, 1, 5

supersaturated solution, 2

Gating:

area, 47±48

bottom, 13

central, 63

direct and indirect, 20±21, 47, 59±63

direct onto a core, 47±8, 101

external, 63

horn, 55

joint line, 13

knife, 54

multiple, 49±50

pencil, 54

ratio, 24, 41, 43, 48, 101

top, 13

touch (kiss), 53

vertical fan, 55

Go/no-go gauges, 187

Grain refinement; Ni-based investment castings, 159

Grain size control by temperature, 17

Graphitic surface films, 23

Gravity casting, 16, 72±73, 111±12

Gravity die:

application of local chill, 147

convection, 160

runners, 44

sump casting pick-up/location, 175

vacuum venting, 118

variability of cast products, 167, 175±6

Grey iron, 23, 48, 54

Griffin Process, 5, 113

H Process, 64

Heat:

capacity, 147±9

diffusivity, 147±8

pipe, 156

Heat treatment to impair strength, 170

Heat treatment developments, 173±4

Hesitation to flow advance, 106

Heuvers circles, 124, 133

Hexachlorethane, 5

Holding furnaces, 3

Homogenization heat treatment, 163

Horizontal stack moulding, see H Process

Horizontal transfer casting, 68, 71

Horizontal velocity:

in basin, 28±9

in mould, 49±50

Horizontal cylinder casting, 95

Hot spot, 134, 141, 151, see also Junctions

Hoult, Fred, 64

Hydraulic cement, 16

Hydraulic jump, 12, 22

Hydraulic lock, 90

Hydrogen embrittlement, 2

Hydrostatic pressure due to depth, 75

Inclusions:

alumina in steel, 89

control, 78±93

flotation, 4

macroinclusions, 12

non-metallic, 1

sand, 84

TiB2 in Al, 89

Ingate porosity, 110

Ingate velocity, 10

Ingot pouring, 25

Inoculation of cast iron, 2, 131, 142

Inorganic binders, 117

Insert bonding, 150

Integrated:

foundry design, 74

manufacturing, 176, 180

Interdendritic liquid, 109

Interrupted pour defect, 112

Interrupted pour technique, see Two-stage pour

Inversion after casting, 64, 157±8, 162, 181

Investment:

casting, 41, 77, 120, 124, 159±60

methodology, 194

materials, 147

Iron Bridge, 16

Jet-like flow:

At chokes, 41

At constriction, 48, 54

At ingate, 23, 48, 54, 55, 58, 100±1

At filters, 82, 90

fast runner jet, 14, 21±2

microjetting, 14, 77

plunging, 12, 16, 104

Jet streams, see Flow channel

Jewellery castings, 76, 77

Junction:

hot spot, 126, 133 (twice)

thermal effects, 50±3, 150±6

Kinetic head, see Velocity head

Ladle

bottom pour (bottom teem), 30±3

lip pour, 30

Laid, Erik, 68

Lateral velocity in moulds, 49, 101

Lap defects, 95

cold, 95, 103±4

horizontal, 105

oxide, 103±4

vertical, 105

202 Index


Leak defects, 110, 115

Level pour, 68

Linear displacement transducer (LDT), 187

Liquid metal tensile strength, 123

Loam, 54

Location points, 175±80

Six point system, 177

Location jigs, 180, 187

Loss of area of surface, 11, 102±3

Low pressure casting, 12, 19, 73, 112±13

Low pressure sand (LPS) technique, 112

Lustrous carbon, 110

Machining of castings:

clamping points/tooling lugs, 177±8, 180

distortion after machining, 172

location of castings, 175±77

machining jigs, 178, 180

machining locations, see Location Points

Magma, 63

Mechanical properties, 146

Melting furnaces, 3

Metrology, 186

Mercury barometric height, 123

Metal/mould reaction, 63

Methoding, see Filling system design

Microjetting, 14

Microporosity, see Porosity

Mismatch, 185

Misrun, 17, 53, 72, 190

Modulus, 127±8, 132, 133, 143, 151

Mould:

box (flask), 42, 115

box registration system, 185

chemically bonded, 42

coat, 115, see also Die coat

cooling, see Die cooling

design, 180±2

dilation, see Mould rigidity

greensand, 42

hardness, see Mould rigidity

investment, 41

multi-impression, 190

penetration by liquid metal, 134, 141

release agent on pattern, 154

rigidity for casting soundness, 124, 131, 141±2

three-part, 41, 159, 181

Nickel-based superalloys, 2, 128, 158±61

Nodularizing treatment for ductile iron, 2

Non-metallic inclusion, see Inclusion

Off-set basin, see Basin

Off-set sprue, see Vortex runner

Oil pan casting, 175, 179

Out-gassing pressure of moulds and cores, 75, 117

Ounce metal, 136

Overspill, 32

Oxide flow tube, 72, 101, 104±6

Oxide lap, see Lap defects

Padding, 134

Partly solid flow, 63

Passing problems in core assembly, 181

Pasty flow, see Partly solid flow

Pasty zone/casting section ratio, 136

Patternmaker's shrinkage (solid state), 64, 120, 182

Permeability, 75, 117

Permanent mould, 124, see also Gravity die

Pin and bush mould box location, 185

Plaster investment, 147

Porosity:

entrained air bubbles, 123

gas, 122±123

initiation from bifilms, 125

macroporosity, 140

microporosity, 108, 122, 125, 140

shrinkage, 123, 125

surface initiated, 135, 140

Potential head, 190

Pouring basin, see basin

Pouring heights, 4

Pouring time, see Filling rate

Preformed running system, see Ceramic tube

Pressure die casting, 11, 77, 110, 125

Pressure gradient in runner, 49

Pressure head, 190

Pressurization of casting, 125, 140±1

Pressurized running system, 21±23

Priming of the running system, 190

Primitive shapes, 128±29

Quality assurance:

records (video), 117

statistical process control (SPC), 124±5, 187

traceability viii

Quenching:

distortion, 172±173

stress, 167±72

Quenching media:

air, 169±72

hot water, 171

polyalkylene glycol, 169±70

water, 168±171

Ratio, see Gating ratio

Reduced pressure test, 1, 5,

Release agent to aid strip mould from pattern, 154, 183

Remelting, 149 (twice), 158±60, 164

Residual stress, see Stress

Response time of a basin, 30

Reversal of flow advance, 102, 106

Reverse tapered feeder, see Feeder

Reynold's number, 20

Ribs; use to aid filling, 14

Rigidity of moulds, see Moulds

Index 203


Riser:

(unhelpful name for Feeder, see 121)

up-runner, 61, 121

Rotary degassing of Al, 5±6, 113

Roll casting, 54

Rolling (of strip or plate etc), 146

Roll-over, see Inversion after casting

Runner:

area, 43

back-wave, 22, 109

dams, 64

slot, 46, 99±100

taper, 44±5, 58

vortex, 68

Sand erosion, 17, 34

Sand inclusions, 17

Scrap rates, 17

Segregation, 163±5

Semi-solid, see Partly solid

Sequential filling, 64

Sessile drop heights, 4, 9, 10, 33

Shot blasting distortion, 186

Shot control filling, 12

Shrinkage:

liquid state, 120

solidification, 120±1, 127±32

solid state linear contraction, see Patternmaker's

Shrinkage porosity, see also Porosity

bubble damage confusion, 108±10

from undersize feeder, 128

interdendritic, 109

pipe, 130

under-feeder shrinkage, 126, 132±3, 140

SI (Systeme International) units, 127

Side pour, see Level pour

Silica sand thermal properties, 147

Silica tube sprue, 27

Silicate films, 23

Sink and float, 4

Sink in surface, 125, 140

Slag in runners, 20, 43±4, 82

Slag pocket, 79

Slag trap, see Dross trap

Sleeking, 54

Slide valve in runner, 66

Solid feeding, see Feeding

Snorkel ladle, 66

Spinner, see Swirl traps

Spraying time, see Jet from filters

Sprue:

base, 39

definition, 13

design, 97±99

flow rate control, 98±9

general, 33±9

height reduction, 21

multiple, 37±39, 98

taper and reverse taper, 37, 111

silica tube, 27

slot, 38, 46, 99

Squeeze casting, 12, 125

Stack moulding horizontal, see H Process

Stahl, 70

Statistical process control (SPC), 124

Steels:

bottom-teemed ladles, 25, 30, 191±3

carbon, 12, 164±5

castings, 5

embrittlement, 2

fins (poor cooling effect), 52, 128

ingot pouring, 25

mould rigidity and soundness, 124

segregation, 163±5

stainless, 12, 110, 118

tool steels, 165

use of filters, 86

use of sand-moulded sprues, 35±6

Stress, 166±74

Stopper, 33, 39

Strainer, see Filters

Strickling, 54

Strontium, 6

Subsurface porosity, 109±10, 114

Sump casting, 175, 179

Surface:

cracks, 106±7

initiated internal porosity, 135

penetration by liquid metal, 134

sink, 125

swell, 141±142

Surface tension:

control of gates, 53

controlled filling, 75, 191

Surge control systems, see also By-pass, 55±9, 65, 99

Swell, see Surface swell

Swirl traps, 67, 80

Tea pot ladle, 102

Temperature gradient:

adverse, 61

critical, 135

solidification front G, 137

Template gauges, 187

Tensile strength of liquid metals, 123

Tilt casting, 9, 14, 69±72, 95, 102, 162

Tin sweat on bronzes, 164

Thermal:

conductivity, 147

diffusivity, 147, 168

Thin-walled castings, 126±7

Tool steels, 165

Tooling definition, 183

Tooling points (TP), see Location Points

Tooling lugs, 177±8

Top pouring, 16

204 Index


Traceability, see Quality systems

Transverse travelling (unzipping) waves, 72, 95, 191

Transverse velocity in moulds, see Lateral velocity

Transverse velocity to flow front, 104

Turbulence:

`turbulence-free', 9,

surface turbulence, 12

Two stage filling, 65±66

Under riser porosity, see Under feeder porosity

Under riser segregation, see Under feeder segregation

Unstable advance of flow front, 105

Unpressurized running system, 21±3

Unzipping wave, see transverse travelling wave

Vacuum:

assisted filling, 76

casting general, 75

casting Ni base alloys, 2, 14

degassing, 5

Van de Waals forces, 107

Velocity head, 190

Velocity of solidification front V, 137

Velocity of liquid flow VL, 137

Vena contracta, 40

Vents:

aerofoil filling, 76

atmospheric links for feeders, 131, 138±40

blocked vent leading to blow, 158, 181

drilled, 118

misplaced efforts reduce porosity, 116

mould and core outgassing, 75, 115, 117

nylon woven, 116

smoke test, 118

wax, 118

whistler, 121

Viscous adhesion, 107

Viscous liquid casting, 16, 162, 191

Volcano, 63

Vortex, 24, 25±6, 36, 38, 66±8

centrifugal action, 59

dump, 58±59

sprue, 41, 66±67

runner, 41, 68

well, 67±68, 80

Washburn core, 134

Water models, 111

Waterfall effects, 13, 50, 104, 162

Wax assembly, 159

Weber number, 102, 190

Well, 39, 111±112

Wetting and non-wetting, 75

Wheels:

Automotive, aluminium alloy, 161

Railroad, steel, 5

Whirl gate, see Swirl trap

Whistlers, see Vents

Wrought processes, see Forging and Rolling

X-ray video radiography, 125, 191

ZA alloys, 12

Index 205


The 10 Rules of Castings - ایران مواد

Documents

Transcript of The 10 Rules of Castings - ایران مواد