AutomaticComputerizedOptimizationInDieCastingProzess

8/7/2019 AutomaticComputerizedOptimizationInDieCastingProzess

1/8

K SIMULATION

Casting Plant & Technology 4/008

Dr.-Ing. Ingo Hahn, Dr.-Ing. Gtz Hartmann,Magma Gieereitechnologie GmbH, Aa-chen, Germany

www.magmasoft.de

Today, hundreds of different gating systems and overow wells can be assessed overnight by automatic optimization. Thecomputer lists all variants, simulates the casting process and automatically nds the optimum variant. (Photo: MAGMA)

Automatic computerized optimiza-

tion in die casting processes

The simulation o die lling and so-

lidication processes was rst applied

in oundries more than 20 years ago.

Although casting simulation is still a

young technology, it has gained wide

acceptance. More than 150 oundries

in Germany use simulation programs.

In addition to larger oundries and di-

rect automotive suppliers, simulation

has become an established practice

also in SME oundries. They use simu-

lation mainly as a tool or testing the

quality o castings or new or modi-

ed equipment [1].

Due to the multitude o actors a-

ecting the quality o castings and

the complex interactions o physics,

metallurgy and casting geometry,

empirical knowledge was the prin-

cipal resource on which optimized

manuacturing engineering relied.Foundry simulation can quantiy

experience and thereore only test

a state, whereas the conclusions

rom the calculations and improve-

ments require the hands o an ex-

pert. Continuous improvement and

optimization is a succession o tri-

als and errors, both in reality and in

simulation.

In recent years, response time o

simulation sotware has improved

and now integrates parallel process

computers. The computing time

needed or one variant o the cast-

ing process to be optimized can thusbe completed within a ew minutes.

Combining simulation sotware with

an optimization program makes is


2/8

SIMULATION K


possible to automatically analyze cal-

culated variants with regard to the

dened target criteria (e.g. low poros-

ity and low rate o returns), to create

new variants and to analyze them in

the same way. The vision o automat-

ic computer-assisted solution-nd-

ing or casting problems has become

reality. A number o examples can be

ound, rst and oremost, in gravity

casting. Here, especially the optimi-

zation o riser technology should be

mentioned, [2], but also the geometry

o runners and gates is increasingly

designed by this technology [2, 3, 4].

This paper looks at examples o the

application o this next generationcasting process simulation in pressure

die casting.

Process development forpressure die casting

A major aspect o process develop-

ment or pressure die casting is the de-

sign o dies or die lling and heating.

Foundry experts can spontaneously

name a long list o optimization tar-

gets that they usually try to reach allat the same time:

to avoid delamination o the melt

in the runner as well as turbulences

and gas porosities in the casting;

to avoid oxides and inclusions;

to achieve simultaneous and uni-

orm lling o cavities in multiple

dies, depending on their location,the pouring system, and the casting

parameters;

to optimize the gate cross sections

or ecient eeding, including thin-

walled sections;

to dimension and position gates

and to design die heating to avoid

shrinkage porosity; to minimize the runner volume to

reduce returns;

to avoid cold laps by improving the

main fow control in the casting.

Optimization with MAGMAfrontier. Based on a choice of start designs with appropriate degrees of freedomand manufacturing restrictions, the optimization algorithm automatically carries out a simultaneous analysis of

several objectives and nds optimized production parameters, such as the location of coolers.

Figure 2

specification,

manufacturing know-how

test results,


test results,


first design

improved design

optimized

design

first parts

improved parts

pattern,

tools

To optimize casting by the conventional approach, a rst method mustbe designed based on experience, the casting inspected and modiedsystematically if problems are encountered. This process is repeated untila satisfactory result is obtained.

Figure 1


3/8

K SIMULATION

4 Casting Plant & Technology 4/008

Additionally, there is the die lie

time, which is aected by the fow

rates in the gate and by the tempera-

ture changes in the pouring cycle.

The shapes o the running system

and o the gate are important pa-

rameters in pressure die casting. The

fow rates in the gate have a critical

impact on the die lling process.

The dened shot curves o the cast-

ing machine must be adapted to the

running system in the permitted

range.

Optimizing the casting pro-cess by one-dimensional

search

Each oundry has a number o

processes that are subject to con-

stant improvement. This is driven

by the necessity to improve the

quality o castings or to cut the costs

o production. Basically, a simple

optimization method is applied: the

one-dimensional search that ater

a number o successive trials, tests

and improvements nally should

produce an acceptable result, al-though it is not known whether this

is the genuine absolute optimum.

Thereore, runner optimization

means that a rst variant o the cast-

ing technique is designed based on

experience, then the casting is in-

spected and i a problem occurs, the

technique is modied (again empir-

ically). This process is repeated until

the quality o the casting and/or the

cost eciency o the process is ac-

ceptable (Figure 1).

Typical o the one-dimensionalsearch is the small number o trials

and the risk o ending up in a dead-

lock. There are numerous examples

rom oundries to prove this: Quite

oten, though, oundries ound a

surprising solution to their prob-

lems ater a radical reorientation.

The one-dimensional search is a

method o optimization that ts hu-

man ways o thinking and acting. It

can work systematically and quickly

because experience and power ocomprehension enable man to ob-

tain a relative optimum ater only a

ew trials.

260

240

220

200

180

0 4 10 2 86

Air contact of melt [ms]

Deviation of filling times [ms]

Duration of air contact of the melt in the die cavities and deviation oflling times for the variants calculated by the optimization algorithm

Figure 5

Three different variants of a running system. The parametric geometrysupports variations of the cross sections of runner and gates at any re-quired position.

Figure 4

Runner CAD of a six-cavity die. The layout in the gure is obtained as aresult of arranging the cavities on as little space as possible.

Figure 3


4/8

K SIMULATION


In oundry practice, the eective-

ness o this optimization approach has

substantially improved in recent years:

Today, trials are oten not carried out

in reality using pattern, die and casted

prototype, but virtually by means osimulating the casting process. At the

same time, the requirements on pro-

ductivity and robustness o the pro-

duction process o high-quality die

cast parts increase constantly. Finan-

cial necessities require accurate plan-

ning o the die concept and the manu-

acturing process. With the increasing

complexity o the die castings, the ap-

plication o experiences to new proj-ects becomes ever more complicated

and thereore questionable. There is

an urgent necessity to eliminate the

trial-and-error actor or the user.

Automatic computerizedoptimization

Automatic computerized optimi-

zation represents a new approach to

solve diculties in the manuactur-

ing process. MAGMASOFT simulation

calculations with varying process pa-

rameters and die layouts are used as

an experimental ground. For this, the

simulation program was integrated

into an optimization loop that runs

without user interaction ater opti-

mization targets and degrees o ree-

dom were dened. Several objectives,

even contradictory objectives, can be

pursued at the same time (e.g. castingquality, productivity, material input).

To have a positive impact on the op-

timization targets, manuacturing

variables (e.g. casting conditions, ma-

terials, die temperature, shot curves,

spray conditions) and geometries (e.g.

runner design, gate dimensioning,

position and dimension o cooling

channels) can be varied. Manuactur-

ing restrictions (e.g. cycle times, die

concept) can be taken into account

(Figure 2).The MAGMArontier optimization

tool is based on genetic algorithms.

The rst generation o variants is de-

ned as DOE (design o experiments)

Air contact of the poured melt (a) before optimization, (b) after geometrical optimization, and (c) after optimi-zation of the shot curve. The duration of air contact during die cavity lling is used as the measure of intensityregarding the tendency to form oxide.

Figure 7

Flow rates in the die at 970 ms (left) and 986 ms (right) after start ofplunger movement

Figure 6


5/8

K SIMULATION


using the example o statistical experi-

mental design rom the mostly very

large number o possible variations.

Several generations are then processed.According to the laws o genetics, posi-

tive characteristics will survive, such

as ew air pockets, reduced shrinkage

porosity or increased yield. The best

possible compromises between in-

dividual objectives are ound in this

way. The quantitative inormation on

the impact o the varied process con-

ditions obtained during optimizationcan also be used or sensitivity studies.

Hence, the oundry expert obtains ar

more inormation rom the optimized

results than only on the part in ques-

tion. The trial and error method is

now transerred to the computer. The

next generation o casting process sim-

ulation makes proposals or optimum

parameter combinations or the bestpossible layout. This will be explained

with the ollowing examples covering

important process development areas

or die casting.

Homogenizing die fillingwith minimized oxideformation

Most small castings are produced in

multiple-cavity dies. To produce prod-

ucts o uniorm quality and to allow

smooth lling o the die, the cavitiesshould all be lled at the same time

and local fow peaks should be avoid-

ed. In this rst example, six castings

are poured simultaneously (Figure 3).

Normally, the courses o the cross

sections o the six runners are de-

signed by the engineer based on expe-

rience and proven tables. In our case,

this is provided by the optimization

program. A parametric three-dimen-

sional model o the geometry is intro-

duced in MAGMASOFT or this pur-pose. Figure 4 shows some possible

designs o the runner system. Taking

advantage o the available symmetry,

Three different variants of the geometry of gates and overow tabs. The

angle around the casting and the extension of the overow tabs are varied.

Figure 9

Cast piston rings; the area at the ring circumference is machined and must be absolutely free of pores.

Figure 8


6/8

K SIMULATION


the calculation o always hal the die

plate is sucient or the simulation.

The lling time or each die cavity

is known ater the simulation. This isthe time span rom when the plunger

begins to move until the complete

lling o the cavity. The target unc-

tion or the optimization program

is to make the deviations between

the lling times o the die cavities as

small as possible by selecting certain

runner cross sections.

Apart rom the homogeneous ll-

ing process o all cavities, oxides in

the castings should also be avoided in

this example. As a result o the simu-lated die lling, the intensity with

which the melt is exposed to air can

also be analyzed as local distribution

across the lled die cavity. The on this

basis calculated reduction o oxide

ormation in the cavities was dened

as the second target unction o the

optimization program.

Figure 5 illustrates the air contact

in the die cavities and the deviationo lling times or 486 calculated vari-

ants. Each point represents one vari-

ant calculated by the optimization

algorithm. Variants rated good are

located in the let lower range o the

diagram, where homogeneous lling

o cavities with the shortest possible

air contact was ound. The blue line

in the diagram marks the best vari-

ants suggested by the target unctions

the optimum, nally, is obtained by

the engineer weighting the two ob-

jectives. The comparative analysis othe variants o an optimization run

provides only qualitative data; a de-

tailed analysis o the simulation re-

sults concerning the target variants is

unavoidable.

The best compromise between

both objectives is represented by the

green variant in the diagram. Figure 6

illustrates the lling o the die cavity

or the appropriate running system

at two dierent times. The cavities

are lled simultaneously this is anexample o a genuine homogeneous

lling. Initially, the detailed analysis

o oxidation ailed to produce a sat-

isactory result, as it can be seen in

Figure 7. The scatter diagram shows

a qualitative reduction o air contact

due to the variation o the geometry

o the running system. However, the

improvement obtained with this vari-

ant is not enough to talk about an

absolutely satisactory result. In a sec-

ond optimization run the shot curve

was varied in addition to the geom-

etry o the runner cross sections. The

additional degrees o reedom avail-

able to the optimization program in

this case were the switchover point

and the speed o the plunger in the

second phase o die cavity lling. The

improvement regarding the tendencyor the ormation o oxide obtained

in 620 simulated variants is shown

in Figure 7. No urther improvements

regarding the deviations o the lling

times in comparison with the purely

geometrical optimization were pos-

sible.

The optimization o the shot curve

may appear to be slightly academic

considering the practical diculties

o setting an exact and reproducible

shot curve; however, the result under-lines the major impact the shot curve

has on the quality o a casting and

thereore also the necessity or the

ounder to be careul about setting a

good and reproducible shot curve.

Minimizing gas porosity

Avoiding gas porosity is one o the

central tasks in designing the die cast-

ing technique. During the whole die

lling process, the melt should ll the

runners and gates completely to avoidthe entrapment o air in the cavities.

Bubbles entrapped in a casting should

leave it as quickly as possible through

an overfow tab. It can be anticipated

that the designs o the gate geometry

and overfow tabs have an essential

impact on the quantity and the routes

o air entering the pouring system.

Problems with gas pores were en-

countered in the critical area o the ring

circumerence during the production

o an ensemble o our piston rings (Fig-ure 8). The task o automatic optimiza-

tion was to eliminate the pore risk. For

this purpose, the geometries o gates

Front view of the ange. Thebolt-on pieces are critical toshrinkage porosity.

Figure 11

Air distribution in the casting after lling before (left) and after (right)optimization

Figure 10


7/8

K SIMULATION


and overfow tabs were to be optimized.

Taking advantage o the available sym-

metry, the pouring system and castings

were modeled the gates and overfow

tabs were designed parametrically. Fig-

ure 9 is a typical illustration o three

variants o this geometry. For each

variant, the simulation shows the dis-

tribution o air in the cavity at the end

o lling. The purpose o optimization

was to keep the air entrapment in the

critical area o the piston ring as low as

possible. 360 variants were simulated.

The result a substantial improvement

is shown in Figure 10. It is possible to

keep the to-be-machined area o the

piston ring ree o pores and conse-

quently ensure high product quality.

Elimination of shrinkageporosity

Shrinkage porosities or cavities

occur where material accumulates

and normally course problems both

with respect to mechanical load o

the casting and or the machining

process. Material accumulations are

required in most die castings or unc-

tional reasons, which makes it impos-

sible to eliminate them completely by

modication o the casting design.

Consequently, it is o undamental

importance to provide the appropri-

ate conditions or eeding. This in-

cludes the design and positioning othe gates through which the casting

can be squeezed ater die lling; ur-

thermore, the solidication can be

controlled to some extent by the sys-

tematic control o the temperature o

the die. The present example is a cast

fange o a hydraulic brake. Shrinkage

cavities due to material accumulation

can orm in the areas o the bolt-on

parts and o the lining groove. These

areas will be machined and thus re-

quire to meet highest requirements odensity (Figure 11).

Figure 12 shows the fange includ-

ing casting technique. In addition

to runner and gate, seven cooling

circuits are available. Consequently,

there are a number o degrees o ree-

dom available or optimization. On

the one hand, the gate geometry can

be changed to acilitate squeezing, on

the other hand, the cooling circuits

can be lled with coolant in dierent

sequences.

The reduction o shrinkage porosi-ty was the target unction in the opti-

mization run. A total o 450 variants

were simulated. Figure 13 illustrates

the improvement o solidication ob-

tained by optimization. The methods

described above cannot completely

avoid shrinkage cavities, but these

cavities can be reduced to a level that

allows the casting to be machined

without damage. In addition, this

optimization run provided inorma-

tion on the cooling water lines thathave an eect on solidication and

on the ones that have not. This inor-

mation is o utmost importance or

Flange with shrinkage porosity (a) before and (b) after optimization. Inthe optimized variant, shrinkage cavities have been eliminated in theto-be-machined areas.

Figure 13

Brake ange with cooling water lines (a) and gate (b). The best possible

lling of eight cooling circuits with coolant is calculated by the optimiza-

tion program. The gate is parameterized and its geometry can be modied.

Figure 12


8/8

K SIMULATION


the layout o a casting technique o

this type. For example, it may be pos-

sible to remove cooling water lines

and consequently reduce the costs o

design.

Optimized squeezing

Shrinkage cavities that are sur-

rounded by thin walls can oten not

be removed by squeezing. One way o

obtaining good quality in spite o this

is local squeezing.

The case described below is that o

dimensioning a squeezing system or

the orced eeding the critical area o

a casting (Figure 14). I they do not

work rom the beginning, such sys-

tems must be optimized by trial and

error, which is expensive. The optimi-

zation program nally nds a variant

that ensures optimal product quality

at the critical material accumulation

with minimum squeezer volume (Fig-

ure 15). It becomes clear how critical a

design o this kind can be and why it

is so dicult to design local squeezer

systems correctly.

Summary

The examples described above il-

lustrate that the automatic optimiza-

tion o casting techniques based on

simulation is also entering the cast-

ing practice. Elimination o trial and

error gives the expert the opportunity

to improve processes at highest qual-

ity and maximum cost eectiveness.

He also learns about the sensitivitieso the impacts and interactions o the

process parameters to the nal cast-

ing quality o process conditions. The

pool o available possibilities has not

been exhausted by ar it is the next

generation o simulating casting pro-

cesses.

References:

[1] Druckgusspraxis (2005) No. 3 (June),pp. 90-94.

[2] Giesserei 91 (2004) No. 10, pp. 22-30.[3] Sturm, J. C.: Optimierung von

Gietechnik und Gussteilen. Sympo-sium on Simulation in Product andProcess Development, November 5-7,2003, Bremen, Germany.

[4] Giesserei 94 (2007) No. 4, pp. 34-42.

Solidication of a die casting takes longest where walls are thick.Besides, shrinkage cavities are bound to form. As this area cannot besqueezed sufciently, local squeezing is a solution.

Figure 14

Illustration of a feeding criterion: right design of a squeezing system

found by computer optimization; no shrinkage cavities are seen in thearea of inspection below the die: left a design of very similar dimensi-

ons in which not all shrinkage cavities were avoided.

Figure 15

AutomaticComputerizedOptimizationInDieCastingProzess

Documents

Transcript of AutomaticComputerizedOptimizationInDieCastingProzess