USAAMRDL-TR-75-31A

I - /

INTEGRAL ENGINE INLET PARTICLE SEPARATOR

Volume I-- Technology Program

General Electric CompanyAircraft Engine GroupCincinnati, Ohio 05215

luly 1975

Approved for public release;•distribution unlimited.

Prepared for

EUSTIS DIRECTORATE

U. S. ARMY AIR MOBILITY RESEARCH AND DEVELOPMENT LABORATORY

Fort Eustii, Va. 23604

Otis5 While SllC1-i2

Doc Buff senl 0

rFllATIo: . .. EUSTIS DIRECTORATE POSITION STATEMENT

SBy ............ 'Me work reported herein represents the first known systematic in

vestigation of the design variables that influence the performanceWEt of integral inlet particle separators.

It is expected that the design procedures and performance data de-

scribed herein will be used in the design of integral inlet particle

separators for future Army aircraft gas turbine engines.

Appropriate technical personnel of this Directorate have reviewed

this report and concur with the conclusions and reconmendations

contained herein.

Mr. David B. Cale of the Propulsion Technical Area, Technology

Applications Division, served as Project Engineer for this effort.

DISCLAIMERS

The findings in this report are not to be construed as an official Department of 'he Army position unless sodesignated by other authorized documents.

When Government drawings, specifications, or other data are used for any pupose other than in connectionwith a definitely reiatcd Government procurement operation, the United States Government thereby incurs noresponsibility nor any obligation whatsoever; and thp fact that the Government may have formulated, furnished.or in any way supplied the said drawings, specificat;ons, or other data is not to be regarded by implication orotherwise as in any manner licensing the holoer or any other person or corporation, or conveying any rights orpermission, to manufacture. use, or sell any patented invention that may in any way be related thereto.

Trade names cited in this report do not constitute an official endorsement or approval of the use of suchcommercial hardware or software.

DISPOSITION INSTRUCTIONS

Destroy this report when no longer needed. Do not return It te the originator.

I

(I

UP UNCLASSIFIEDSECURITY CL.ASSIFICATION Or THIS PAGE foyh-. 0).i. FnI.'ed)

REPqnUDOCUMENTATION PAGE BEOE OPE TID~NG O2 GOVT ACCESSION NO 3 RECIPIENT':,CA, ALOG NUMBER

_½ USAAMnRDLrTR-75-3lA _ _ __ _ _

Roer uffyP0

BenrdFjhttcW, 'N

A. PRFO MIN OR ANIATI N AETAND CL AD S EPI.'LGA L ME TP OIET TS

I5 CONTROLLIN OORC NAMET AND ADDRESS

I. MONITORMING ORAGNCIZATIN NAM A N ADDRESS(, dI.,., 5,, oI.,M GAOfie S SCRIT CA SS (of fh.PRJC TASK

-Genorra lectric Copany URELASSRKUIFIED BER

Appi Droedforat public

U..Am1i Mb & Laboease itory bto unlitmited.pI ef

Fort Eustis inhq 2'4624• 109

SScrollAparticed Trajeblctrylepara;ditrionEfcenyuet pcfctions uniitd

Volume I of this7 raepr dscie threusofa3-nhpogm

to design and evaluate integral engine inlet separators in 2, 5, and15 lb/sec sizes. The separators are designed to remove sand, dust,and single foreign objects from the air entering a turboshaft engine.

DD FORM 14734 EDITION OF I NOV~ 6!S OBSOLETEUNLSIED (otne)i;:

OD SECURITY CLASSIFICATION OF THIS PAGE (It7,.,, D.11 EI.,elod)

• UNCLASSIFIED

SECURITY CLASSIFICATION OF THIS PAGE(WIh.., fl.I. FtI.-..)

ý " Program effort included:

Definition of turboshaft engine environment in terms of amountsand types of ingested materials.

"k-2 Mechanical and aerodynamic design trade-off studies of integralengine separators followed by full-scale model performance testsof 2 and 15 lb/sec separators.-)

"Detailed design and procurement of 2, 5, and 15 lb/sec separators,and design and procurement of improved components. ->

")A ,Model tests of the separators to obtain detailed aerodynamic and"off-design performance data.

(5) Preparation of a separator design guiae for integral engine separ-ators in the 2 to 15 lb/sec size range. -

"•-) !The design guide (including environment definition) is presented inVolume II of this report. Volume I documents all other aspects of the[program.

UNCLASSIFIEDSECURITY CLASSIrICATION OF THIS PAGE("h#n Data FWt...d)

I -- - _-_

PREFACE

This final reDort contains the results of work done to satisfy the require-ments'oT leiIntegral Engine Inlet Protection Device Technology Program.The program was conducted by General Electric under Contract DAAJ02-73-C-0004 with the Eustis Directorate, U.S. Army Air Mobility Research andDevelopment Laboratorj, Fort Eustis, Virginia. It was carried out over aperiod of 30 months in six phases.

Phase I - Definition of Environment

Phase IIA - Design Trade-Off Definition

Phase IIB - Inlet Scroll Configuration

Phase III - Detailed Separator System Design

Phase IV - Model Hardware and Instrumentation

Phase V - Experimental Evaluation

Phase VI - Preparation of PreliminaryDesign Guide

Several people who contributed to the program should be recognized. Theliterature s4.eirch, which was most of the work of Phase I, was conducted byT.L. De Young. The design and procurement of the models, and the conductof the test program, were almost entirely done by J.A. Ebacher. Nearlyall of the particle trajectory analysis effort was contributed by G.E.Lepine. The aerodynamic data reduction and analysis was done by G.L. Allen.The success of the program was due largely to these contributors.

3

4-

3

'4I

g4i

TABLE OF CONTENTS

Page

PREFACE .. . ........................................................ 3

LIST OF ILLUSTRATIONS .. ............................................. 6

LIST OF TABLES. .. ................................................... 9

INTRODUCTION .. .. ................................................... 10

INDEX TO TECHNICAL DATA .. .......................................... 101

DEFINITION OF ENVIRONMENT .. ........................................ 126

PRELIMINARY DESIGN. .. .............................................. 14

Engine and Separator Systems. .. .......................... 14Subsystems and Their Integration. .. ...................... 19Aerodynamic Losses. .. .................................... 23

AParticle Separation .. .................................... 23

MODEL DESIGNS. .. ................................................... 26

2 Lb/Sec Axial Separator Design .. ........................ 26X2 Lb/Sec Inlet Scroll Separator Design. .. ................ 385 Lb/Sec Separator Design .. .............................. 4515 Lb/Sec Separator Design. .. ............................ 50

TEST FACILITY AND EQUIPMENT .. ...................................... 59

TEST PROCEDURE .. .. ................................................. 62

'T, TEST RESULTS .. .. ................................................... 63

2 Lb/Sec Axial Separator .. .............................. 632 Lb/Sec Inlet Scroll Separator. .. ...................... 685 Lb/Sec Separator .. .................................... 7315 Lb/Sec Separator .. .................................... 84

SEPARATOR PERFORMANCE COMPARISON. .. ................................ 98

CONCLUSIONS .. ...................................................... 106

RECOMMENDATIONS .. .................................................. 106

LIST OF SYMBOLS .. .................................................. 107

'~fl ,5

A~

LIST OF ILLUSTRATIONS

Figure Page

1 T700 Turboshaft Engine With Inlet Separator . . . . . .. 15

2 T700 Type Inlet Separator Schematic ...... . ..... 16

3 Exploded View of Major Components ...... ....... 16

4 2 Lb/Sec Turboshaft Engine With Inlet Scroll Separator . . 17

5 Scroll Separator Tangential Inlet,............. 20

6 g Forces on a Particle ................. 25

7 2 Lb/Sec Engine With Separator ............. 27

8 T700 Type Separator Flow Path for 2 Lb/Sec Engine ..... 29

9 2 Lb/Sec T700 Type Separator With PTO Shaft Canted Forward 30

10 2 Lb/Sec T700 Type Separator With PTO Shaft Between SwirlVanes and Scavenge Air Vanes .............. 31

11 2 Lb/Sec T700 Type Separator With PTO Shaft Canted Aft . . 32

12 2 Lb/Sec Inlet Separator Model .............. 35

13 2 Lb/Sec Model Flow Path With Hidden Splitter Lip . . . . 37

14 2 Lb/Sec Model With Hidden Splitter Lip-Predicted WallStatic Pressures. .. . . .... . ... 39

15 5 Lb/Sec Model Predicted Wall Static Pressures . .... 40

16 2 Lb/Sec Scroll Separator Model (Reference Figure 17). . . 42

17 2 Lb/Sec Scroll Separator Cross Section .......... 42

18 Scroll Separator Pressure Loss Variation With Flow Area. . 1'3

19 Total Pressure Loss Through Offset Bends ......... 44

20 Frictional Resistance in a Smooth Pipe .......... 45

21 Aero Analysis of 5 Lb/Sec Flow Path, Showing Stations . . 47

22 2 Lb/Sec Separator Predicted Wall Mach Number. . . . . . . 48

6

LII

LIST OF ILLUSTRATIONS - Continued

Figure

23 5 Lb/Sec Thinned Design 3 Vane .4..9...... .*9

24 15 Lb/Sec Separator Model. . .............. 53

25 Vane Pressure - Side Slope . .............. 55

26 Design 3 Separator Swirl System ............ 57

27 Inlet Separator Test Cell .. . . ............ 60

28 2 Lb/Sec Hidden Splitter Modification ......... . 65

29 2 Lb/Sec Scroll Separator Inlet Rework ...... . . . . 70

30 2 Lb/Sec Scroll Separator Core APT Profiles ........ 71

31 Wall Static Pressure for 5 Lb/Sec Separator Model - Vane-less .. . . . . . . . . . . . . 78

32 5 Lb/Sec Separator Model Wall Static Pressure With SwirlVanes. 0" 79

33 Comparison of Rainstep Swirl Angle Prediction With TestData .. . . . . . . . . . . . . 8o

34 Core Swirl Angle With No Deswirl Vanes...... ... 81

35 15 Lb/Sec Separator Exit Radial Profiles at Design Air-flows.. . . . . . . . . . ... 86

36 C-Spec Efficiency vs Swirl Level ............. 87

K 37 AC Coexse Efficiency vs Swirl Level ........... 88

38 Core Pressure Loss vs Swirl Level............. 89

39 15 Lb/Sec Model Flow Path Modification ...... . . .. 90

4o Swirl Vane Clocking Relative to Deswirl Vanes.. ..... 91

41 AC Coarse Separation Efficiency vs Pressure Loss ..... 92

7Si llaq emu ••m•mnm mnm mm ~a• em •mmgn • pwumm~

LIST OF ILLUSTRATIONS -Continued

Figure Page

42 C-Spec Efficiency vs Pressure Loss. .. . . .. 931

43 15 Lb/Sec Design 3 Trailing-Edge Extension. . . . .. 94

44 AC Coarse Efficiency vs Scwnrenge Flow Ratio 3-15 Lb/Sec Model 97

45 15 Lb/Sec Core Pressuree-,o s s..... . 97

46 2 Lb/Sec Core PressureeLo s ... ..... 99

RN4T' 5 Lb/Sec Core Pressure Loss. ..... *,@ 10

48 2 Lb/Sec ScrollPressure Lsoss...... 101

49 5 Lb/Sec Scrt-ll Prei "ure Loss ** * * ***102

5 0 15 Lb/Sec Scroll Pressure Loss...........* * 10

51 Separator Exit Radial Profiles At Design Airflow . . . . . 0

52 AC Coarse Collection Efficiency vs Scavenge System Airflow

8K

LIST OF TABLES

Table Page

1 Index to Technical Data for 2-Lb/Sec, 5-Lb/Sec, andS•15-Lb/Sec Separator o.. . . . . . . . l

2 T700 Accessory Requirements... : . ...... . . 21

il system Requirements and Volumes ...........

4 Total Pressure Loss Breakdown - T700 Type Separator 33

S5 T700 Swirl Vane - Cascade Section Properties*....... 57

6 Design 3 Swirl Vane - T700 Size ............ 58

7 2 Lb/Sec Separator Testing Summary .... ........ 64

8 2 Lb/Sec Inlet Separator Test Results - Pressure Loss and 6

AC Coare Efficiency.................. 66

9 Foreign Object Ingestion Test Results - 2 Lb/Sec Model. . 67

410 2 Lb/Sec Scroll Separator Testing Summary ..... 69

"" 11 5 Lb/Sec Separator Testing Summary . ..... ..... 74

I 12 5 Lb/Sec Inlet Separator Test Results - Pressure Lossand AC Coarse Efficincy .. .. 75

13 AC Coarse Sizes .................... 7614 Foreign Object Ingestion Test Results -5 Lb/Sec Model. 76

15 5 Lb/Sec Separator Testing Summary. .......... 82

4'.• V 16 Foreign Object Ingestion Test Results . . . . . . . . .e 95

A 17 15 Lb/Sec Inlet Separator Test - Pressure Loss and AC- Coarse Efficiency... . 96

j,N1

i 9ap mm tm~qu

INTRODUCTION

The adverse environments in which Army equipment operates impose severe)enalties upon gas turbine engine performance, reliability, and life ex-pectancy. Ingested sand, dust, foliage, rain, salt spray, and single for-eign objects can cut off or appreciably reduce an unprotected engine'sterm of service. Numerous studies and development programs have demon-strated that engine inlet protective devices significantly reduce the dam-aging effects of these various forms of ingested materials. No program,however, has collected the large amount of detailed data relating to theenvironments encountered by gas turbine engines in the field or under con-trolled laboratory conditions. Phase I of the Integral Engine Inlet Pro-tection Device Technology Program vas designed to accomplish that objectivein order to define a representative operating environment in which an in-let protection device is expected to function. The results of Phase Iare piesented in Volume II.

INDEX TO TECHNICAL DATA

To make this final report as useful as possible, an index to the technicaldata contained in both volumes of the report is included as Table 1.Table 1 serves a secondary purpose by indicating the parameters whichare important to the design of an inlet particle separator.

10

* - ~1

p H

.14 w to I1- - I IH

1-4 4-4f4 4

0 f -4 to toI I CO4 toI 4

I to (71 to to( to4 a,

Ho (N C\? 0'0 ~ ~~ ~ H 1- I44 '.

rA- ý - Nto 07 Hý 44 1-4 1-4 1- 14 C'

o- to U.0 '\ tX\ to to4 I. aH

HH 1- H 1 1-4 H HLO'.

'.3 - - H ('

tto

Hl H I0C0 cc0' to to

to~r Htot*

to. H H r -

-I N II N NIH-O Ce\ H- HH-H, H H H H H *- H-

g 0H H H HH H H H H

00

0 H

to H C0 ýo r-I I'.4

H- HtI4

H) -H'H .$4 '. HaH

0N *P . - 0 H ý

-r~'H HH H3 ý d aW-7 -H H H 4 a Hl SH H4 H H

H

01 00U)- f1-4

DEFINITION OF ENVIRONMET

SAND AND DUST

In the helicopter environment at low altitude, especially takeoff andlanding over land, sand and dust are always present. They do not generallycause the catastrophic types of failure which often accompany the ingestionof large single objects or masses of foliage, but they continuously erodean engine's mechanical structure and its performance. The MIL-E-5007C sp•Lcification dust concentration, 1.5 milligrams/ft 3 , is less than nearly allsea level and altitude samples which are found in the literature. Most ofthe recorded concentrations are between 1.5 and 15 mg/ft 3 .

However, the site and size distribution of MIL-E-5007C speci'ication dust

(C--Spec) are typical of most terrain samples from various locations in theworld. As the altitvde of the helicopter over the ground increases, thedust tends to become finer but usually not as fine as AC coarse.

FOLIAGE AND RAGS

Foliage or rag ingestion is an important problem since the engine inletis blocked and the engine muzt be shut down until the item is removed.However, the item does not usually cause appreciable engine damage. Whenthe ingested foliage or rag Ls removed, the aircraft becomes operationalagain. Therefore, there are very few statistics to indicate .the magnitudeof the foliage ingestion probikm, which is probably much larger than avail-able data indicates.

SINGLE FOREIGN OBJECTS

The ingestion of large single objects is the most formidable aircraft en-vironmental problem, since nuts, bolts, safety pins, and other large ob-jects which are left in the aircraft inlet system or dz~opped on the runwayusually cause extensive engine damage. Birds as well as ice which breaksoff the fuselage or inlet system, also are important problems.

WEATHER

Weather is important tc the aircraft engine's environment primarily becauseof ice ingested in large chunks which cause serious engine damage. Rainis not as damaging as ice, but it can cause engine mechanical damage andperformance loss. Concenration of rain in inlet air at altitude has beenrecorded at about 8% of th., inlet airflow by weight. The engine is re-quired by Specification MIL-E-5007C to operate at concentrations of 5%.

12

-COMBINATIONS OF ELEMENTS

Combinations of various elements of the environment are considered to berelatively unimportant. For instance, an ingested bolt is no more harmfulif 4t is accompained by dust. Wet dust may be less harmful than dry dust,but its importance is removed mainly because wet dust is hard to stir upon the ground. Wet foliage may be more harmful than dry. Tests with wetand damp foliage are therefore recommended.

Rk

13

g__

PRELIMINARY DESIGN

ENGINE AND SEPARATOR SYSTEMS

The T700 integral separator design was selected as the baseline configura-tion for scaling and designing integral inlet separators in the 2 to 15lb/sec range because it is an intermediate size, and it repreoents the lat-est end most complete integration of accessory, structural and separatorrequirements. Cross sections and visual descriptions of the T700 separatorare shown in Figures 1, 2, and 3.

The T700 was also a convenient choice since the engines used in the 2, 5,and 15 lb/sec designs were selected to be front drive engines with frontmounted accessories. The principles of the separator design apply equallywell for other engines such as rear drives in this size range. In fact,the designs become simpler when some of the subsystem functions such aspower takeoff (PTO) and front drive do not have to be integrated into thedesign. The engine configurations were based upon engine designs previous-ly studied by General Electric. The 15-lb/sec turboshaft engine is ascaled-up version of the T700 engine, while the 2 and 5 lb/sec engineswere based upon a combination of scaling downward plus additional modifica-tions which include: single-stage turbines for both the 2 and 5 lb/secengine, no axial compressor stages for the 2 lb/sec engine, and two axialstages for the 5 lb/sec engine. Although the separator designs are shownon these engines, they are capable of being used on engines of differentmakes, sizes, configurations and installations.

Direct scaling of the T700 separator was not possible for the 2 lb/sec en-gine. Therefore, a T700 type separator design was used as a baseline con-

' figuration, but other separator designs were also investigated. The mostpromising alternate configuration was a scroll separator. This separator,shown in Figure 4, was especially applicable to the 2 lb/sec engine. Somestudies of the scroll design in the 5 lb/sec size were started but weredropped when they indicated that the overall volume was larger than it waswith the scaled T00 type separator. Alternate design studies were madefor both integrated oil tanks and separate oil tanks, for the scrollseparator.

Installation requirements are important in prescribing the separator con-figuration, especially in the 2 lb/sec size. For the scroll separator,the drive shaft is not coaxial with respect to the air inlet as it is inthe T700 separator. This results in a different installation setup andseparate points of attachment for the inlet. Although the engines onwhich this separator design was based are Drincipally for aircraft in-stallations with air inlet from the front, they could be used for otherapplications such as tanks, other ground vehicles, and auxiliary powerunits. The 2 lb/sec scroll separator is especially versatile and flexible

14

IJ -j

15-

U __

"SEPARATEDMATERIALEXTRACTION

COMPRESSORi ! i '• • INLET AIR

ENGINE ,INLET AIR

Figure 2. T700 Type Separator Schematic.

'4I

MAIN FRAME COLLECTION SCROLL CASE

SWIRL FRAME FRONT FRAME

Figure 3. Exploded View of Major Components.

16ii 1•

~KrA

P f"

%

41

Figure 4. 2 Lb/Sec Turboshaft Engine With Inlet Scroll Separator.

17

74I1

.............

for a variety of different installations, because of the small inlet duct.With a rotatable type elbow inlet, air can be introduced at any angularlocation over a 360-degree arc.

Another configuration of the scroll separator is illustrated in Figure 5..fill In this design the inlet air is introduced tangentially into the scroll.

The elbow collector has been replaced by a collector located furtheraround the scroll at a location where the scroll turns the incoming air

rk- sufficiently to separate out the large foreign objects. Pressure drop inthe inlet elbow is thus eliminated, providing an improvement of approxi-K- mately 5 in. H 0 in pressure drop.

SUBSYSTEMS AND THEIR INTEGRATION

To obtain efficient utilization of space and weight, the separators weredesigned so that they could be integrated 'ith other engine subsystems.The lube system, accessories and accessory gearbox, power-takeoff (PTO),anti-icing systems, structure of the front frame, aero-thermo systems, and

Sinlet protective devices are engine subsystems that have a direct in-fluence upon each other and upon the engine. Integration considerationswere:

1. Accessories and Accessory Power. Accessories, accessory gear-boxes (AGB's), and accessory power requirements were consideredin the separator studies. Although they are not a part of thebasic separator design, the size of the accessory power loadand the size and location of the PTO shaft influence separatordesign. The PTO shafting must be the smallest diameter pos-sible to have minimal effect on separator performance. Forimproved performance, the 2 lb/sec design was changed to havethe PTO shaft go through the swirl vanes rather than throughthe scavenge air vanes as in the T700 separator. This wasaccompanied by a reduction in the size of the struts in thecompressor inlet and in the scavenge air exhaust.

In establishing the three different size separators, the PTOshafting, gearing, and splines were sized on the basis of assumedstarter loading and then checked against the total assumed

accessory power load when running. The accessories for theeng4ines are not directly or equally scalable. Some accessor-ies, such as temperature and pressure sensing accessories, arethe same size or nearly the same size regardless of enginerating. Other accessories, such as the blower and lube pump,are more nearly scalable. For the design studies, accessoryrequirements were scaled or estimated using T700 accessoriesand ratings as a reference. Table 2 gives some of the require-ments.

A 19

I _ _ _ _ _ _ [

LiUCD

-J L

-J J --I-

-j V)FLL~l 0

C)u

OcL) a

LU I=

Z LA

C>0:4Lit 4C)~ IJ wCF- )V42: L

20-

TABLE 2. T700 ACCESSORY REQUIREMENTS

Engine Size (Airflow, ib/sec)Accessory Requirement 2 5 15 T700

AGB HP 20 ho 80 ýq

Customer HP 8 15 30 15

Starter Torque as % T700 25 55 130 100

iSizes as % T700 60 80 115 100

Accessory requirements for the scroll separator design weresomewhat different from those for the T700 type separator.

For instance,a. Both have a completely modular accessory package and

gearbox so that the gearbox with accessories can bemounted or removed as a module. For the scroll separa-tor, however, the gearbox is also a support structurefor the main power drive shaft, which must be removedbefore the gearbox can be taken off.

b. The scroll type separator does not provide for a powertakeoff shbaft (PTO) or bevel gearing. Only spur gearingis used.

t-,• 2. Oil Tank and Oil Cooling (Oil to Air Heat Exchange). TheT700 integr-al seperator incorporates a number of features whichwere retained for all the scaled T700 type separator designs.

P •These are:" a. Integral oil tank.

b. Supplemental oil to air cooling by running oil throughthe scavenge air vanes.

V,;c. Thermal compensation of inner and outer separatorstructural parts by running cooling oil through thescavenge air vanes and around the inner and outer struc-

d. Deaeration of oil by introducing at top and letting itflow over the inner wall of the oil tank.

TabOil flows3 and oil tank capacities estimated are Listed in

21

TABLE 3. OIL SYSTEM REQUIREMENTS AND VOLUMES

Engine Size (Airflow, Ib/sec)Item 2 5 15 T700

Oil Tank Volume (gal, oil only) 1.2 1.5 2.1 1.7

Usable Oil (gal) .3 .4 .6 .5

Oil Flow (gal/min) 2.5 3.3 4.5 4.o

Dwell Time (see) 20 20 20 20

Expansion Air Space (U vol) .... .. 10-25

For the scroll separator design, supplemental cooling ofoil with air is possible by one or more of the following:

a. Integral oil tank.b. Running oil lines around the outside wall o -L. scroll -

especially at the inJet.c. Passing oil through structural vanes in front of the

compressor.d. Separate.ý oil-to-air heat exchanger.

For the scroll separator, both integral and separate oiltanks were considered, and there was an air/oil deaeratorinside the oil tank to assure good deaeration of the oil.

3. Anti-Icing. Anti-icing of swirl vanes and leading edge ofsplitter nose was incorporated into the scaled separator de-signs as it is for the T700 separator.

Anti-icing of the scroll separator for 2 lb/sec size isless of a problem. It can be accomplished by hot oilin tubing, surrounding scroll with oil, and/or oil throughstructural vanes leading to compressor.

t4. Structural Requirements. The structural requirements for thefront of the engine are similar in nature and complexity tothose for the T700 engine. The TT00 type separator providesa satisfactory structure without adversely affecting engine

22

Oe

performance. For the 2 lb/sec size, the vanes must be madelonger and larger to carry the structural loads as well as

X to provide necessary internal flow areas for ducting oil.For the scroll separator in the 2 lb/sec size, the only

structural vanes are those at the compressor inlet.

AERODYNAMIC LOSSES

f The reference that has been used for all of the axial separator losses isthe T700 design. In the initial designs for all core airflows, two prin-cipal scaling rules were followed:

1. All dimensions were scaled so that T700 Mach numbers weremaintained.

2. T70C turning in vane cascades was retained.

The above rules could be followed for the 5 and 15 lb/sec designs by scal-ing the T700 design, but the 2 lb/sec design was defined along the insidediameter by the bearing and gearing requirements of the engine, which pre-vented a direct scale of the T700 separator.

Since turning and passage characteristics of the T700 have been retained,A the initial T700 separator cascade and turning loss of 5.6 in. H 0 was used

2for all the T?00 type separators. The friction losses were taken alsofrom T700 test. results. The initial T700 separator friction loss of 3.5in. H2 0 was corrected for changes in the ratio of scrubbed area to flow

area and for Reynolds number.

PARTICLE SEPARATION

A major objective of any separator study is the determination of particleseparation efficiency. The level of efficiency in the preliminary designstages of a program, prior to availability of experimental data, can beobtained by tracing various particles through a specific flow field. Thisparticle tracing or particle trajectory analysis is based upon the classi-

* cal equations of motion of a particle modified by aerodynamic drag.The equations of motion are:

"V r = dr/dt (1)

Ve = r (de/dt) (2)

V = dz/dt (3)

(dVr/dt) (v6/r) + C(Ur -Vr) (4)

23

dV6/dt = C(Ue - V0) - (VrV )/r (5)

dVz/dt = C(U2 - V2) (6)

where

r, e, Z are particle radial, circumferential, and axial position com-ponents;

Ur, Ue, UZ are fluid radial, circumferential, and axial velocity com-ponents;

Vr, Ve, Vz are particle radial, circumferential, and axial velocity

components; and

t is Time.

This analysis provides a general overview of the separation problem. Par-I ticle Trajectory Analysis is discussed more fully in Volume II, Appendix B.

SFigure 6 shows the variation of the particle g-force (V2/r) with averageradius. Once the separation force is determined by comparing the particletrajectory to the flow paths, relative performance can be defined. Thistype of comparison was made in the trade-off study phase of the program.

-2I1'

22

000,00 C-

00

000,09 -

000'09

w 000 0

O00010C CD___-

CD.

*i-

0001 Lo L

44 425D

-r-

MODEL DESIGNS

2 LB/EEC AXIAL SEPARATOR DESIGN

During Phase II of the program, a T700 type separator was designed for the2 lb/sec turboshaft engine shown in Figure 7. Direct scaling of the T700separator does not satisfy the mechanical requirements for the 2 lb/secengine because the inside dimensions of the separator become too small tocontain the front bearings and sump. Also, the vanes are not structurallystrong enough to carry the engine loads, and they are not large enoughinternally for functional uses such as porting engine oil or housing aPTO shaft or accessory lines. The T700 type separator can be used in the2 lb/sec engine if it is made larger in diameter and longer than a directscale with airflow would dictate. The integral separator design shownincorporates these changes. It has longer and larger swirl, deswirl,and scavenge vanes to carry structural loads and to have enough internalport area for ducting anti-icing air through the swirl vane and oil throughthe scavenge vanes. A comparison of a scaled T700 separator flow path andthe 2 lb/sec model flow path is shoin in Figure 8. The figure demonstratesthe large divergence of the design from the direct scaled T700 separator.

In the 2 lb/sec design shown in Figure 7, the PTO shaft goes through theswirl vanes instead of the scavenge vanes in order to reduce its adverseeffect upon engine performance. The PTO shaft and splines were kept assmall as possible to keep vane size to a minimum. Alternate studies were"carried out on separator configurations with the PTO shaft located between

i �the swirl and scavenge vanes, and 'rith it passing through the swirl and de-swirl vanes. These configurations are shown in Figures 9, 10, and 11,

Preliminary studies on oil flow, oil tank volume, and oil cooling were alsocarried out for the 2 lb/sec separato- and compared to the T700 require-ments. The studies showed that approximately 2.5 gal/min of lube oi' mustbe supplied to the bearing sumps and gearbox of the 2 lb/sec engine. Thisis appreciably higher oil flow than indicated by direct scaling. This alsohad a direct effect on the size of the oil tank since the T700 oil dwelltime requirement of 20 seconds was used. In addition, the increased oilflow requires proportionately more supplemental cooling from the air due tothe higher ratio of oil flow to fuel flow.

Because they become relatively more important in the 2 lb/sec engine, spe-cial attentio'. was paid to the accessories and accessory gearbox. A 60%scale of T70C accessories was selected for the 2 lb/sec engine. However,actual sizes were not considered critical during the initial part of theprogram since the AGB is mounted externally to the separator.

26

MR,5 m:.

GEARCASE

- 4-- - LUBE PUMP

Figure 7. 2 Lb/Sec En~gine With Separator.

27

-NM

GEARCASE SOE

'-IT

*via

, mm:

tI)

CN.

04-4

04

I 1E4

4E-4

Co 44

LL0.Qý Cl

(n i

jj-LI V

29

0

P4J

14.

p.4

4 )

C14

300

'44

Cd

.Z

0)

ca)c

Cd

IE-4U)

oC-4>

90

rX4

31

4-4

U)

~ ~E-4

0.

323

Aerodynamic Design of 2 lb/sec Axial Separator

SE • Mechanical constraints on the 2 Ib/sec separator made scaling of inside

diameter and length impossible. T700 Mach numbers and casc~de turningwere retained, and turning losses were assessed at 5.6 in. HO. The im-pact of size on scrubbing losses was significant since the c~annel aspectratio changed substantially as the inner channel wall was forced outward.The average value of the surface area to cross-sectioned area ratio in-creased by 39% for the 2 lb/sec design. Also, the friction coefficient in-creased by 15% due to Reynolds number changes. Friction lose thereforeI •was about 154% of the T700 level which was 5.4 in. H 0. Summation ofturning and friction losses gave an estimate of l.0 2 in. H20 for thepressure loss of the 2 lb/sec axial separator. The total pressure lossbreakdown can be found in Table 4. In the Phase V tests, a loss of 11.1in. H 0 was achieved.

2

TABLE 4. TOTAL PRESSURE LOSS BREAKDOWN - T700 TYPE SEPARATOR

Total Pressure Loss - In *2Loss Mechanism T700 2 lb/sec 5 lb/sec 15 lb/sec

(1972)

Centerbody Friction 1.3 2.0 1.4 1.2

Splitter Friction 1.0 1.5 1.1 0.9

4 Swirl Vane Friction o.4 0.6 0.4 0.4

Deswirl Vane Friction 0.8 1.3 0.8 0.8

Cascade and Turning 5.6 5.6 5.6 5.6

9.1 11.0 9.3 8.9

2 Lb/Sec Model Features (Figure 12)

/The model was made up of four modules. One module is the bellmouth and bul-let nose, which are separable and are constructed primarily of fiberglass.The swirl vane cascade and its flow path make up a second discrete unit,which is all aluminum. The vanes are mounted in a manner that allows rel-atively simple modification of swirl angle within the range of +100. Thedeswirl vanes are mounted in a similar manner to the swirl vanes, allowingthe same flexibility for changes. With the inner and outer flow paths,they form another independent unit. Thf. vanes and outer flow path includ-ing the splitter nose are aluminum. As with the swirl and deswirl vanes,

33

the scroll vanes are attached to the adjacent sections of flow path andcomprise another independent unit. The scroll is also detachable and ismade of fiberglass. The flexibility of the modular design allowed modifi-cation to local contours or overall flow area to be made inexpensively andrapidly during the test phases of the program.

2 Lb/Sec Model Redesign

The flow path for the model tested in Phase V of the program is shown inK Figure 13. Forward of the rainstep, the flow path is the same as that of

the original 2 lb/sec model. Aft of the rainstep, the splitter lip ishidden behind the inner flow path wall. The Phase II swirl vanes wereused for Phase V. Design of the core flow path aft of the rainstep wasdirected toward achieving small diffusion on the hub and in 'he deswirlcascade. Te core side of the splitter lip was made bellmouto-like to pre-vent entrance losses. This feature was taken from the 5 lb/sec model de-sign, which had the lowest core total pressure loss, 3.7 in. H 0, of allmodels tested with swirl and deswirl vanes removed.

34

- .l••, ? #• • - •£z .• , 'e " >: % 4 * ••

SEPARHUB

Z FROM AEýo 'o . R1

0- 50l 1.5101.00 1 C. 0

-1.50 .5-2.00 2.120- 2.500 2.480

.050 R -3.00 2.890

- 3.50 3.300-3.622 -

S-4l.o0 3.465

F1 s -435.480-4.30 3.530

2 -4.50 .500

.090 -5.00 5.595

-5.50 5.235

70-6.00 .090-6.50 3.005

DETAIL "A' -7.00 o 2.970

5PLITTER WOE-- -7.375 2.97010- t IZE

L ___ 52~~IL 'RING--\

- I-3 22

S•5•-2792" 0 ' R i N G

-5t O

7 o.8381 _________

RE97

Rj R, --

--- 4 300 -

Figure 12. 2 Lb/Sec Inlet Separator Model.

\3

SEPARATORFLOWP.ATH_____ ____

HUB SPUITTER OUTER ______ _BUiLET NO SEýLLMOUTH CCLLECTO-R ,CRD-.

R, R, R, 1L ROMSEPAATOR INLET R,~ __5_ ATION B C.

~~~00O~~~~~~ _____ -- . .50 Q4'o80__-_1__________ _____ 3.?5 Z. 9707 3.250 4"-0

________ 2.475 4.285 4-- 0 2.941 -3.91ro 50 3,.501.825 J2.G95 4.Z85 .o.85 5.004 -75. -00

.00 2.120 _ 3.010 4.285 5.50 a.732 1 4.10___ 90" 1020500 2j.4890 J3.370 4.285 6.00 1___ 2.539 4.23Z -T0 5. ., 04

00 .890 - 4.285 _.50 L 2.2IG8 4.5"Z2 20* L3 300 .80050 3.300 -- 4.285 ".0 __ .87 4.1:10 i35 8622 7 .8 75C 1.3o2 1 5.4 8: 150" .7

00 3.465 - 4.22 5 7j.90 1 coo 1 6b.5 1 65 .063.00 4. 147 -L- .5

~50 i3.530 --- 4.147 195, 1228-S I 3. 500 4.095 2 10' 1314.

00 i 3.595 - 4.005 22 139050 5.235 3.965 2.0' 1474

"13- .950 25 ' 55(o00 3.090 _3.950 270* 1 2(

50 3.005 _ - .90285 .71ro

~0 2.370 - 3.950 300- 1.7?

75 2.970 - .950 31 - -1(o64_______330V 195r

34 203j

93r0 1300l 2 1251

NG-

A750 -, 4.50---------.5

85281 - 175'

4 YoLLoVw 5rRurs '5

ILAA

30005

7.9935 N.9-- (cuio. DIAI

____ - --- - Oi

0 .8513 R

RO REF

8.00 j - 4.865 E

I It

I

I-.

C.0

0 V0

*NI

ý,PX~w ACT)v~..?O PHM

37Ia

1w

Figure 14 shows the calculated wall static pressure distribution of the2 lb/sec model, which is comparable to that of the 5 lb/sec model shown inFigure 15.

Since tests of the 5 lb/sec model indicated that its deswirl vanes per-

formed adequately, the same type of design was used in the 2 lb/sec model.

2 LB/SEC INLET SCROLL SEPARATOR DESIGN

In evaluating separator configurations for the 2 lb/sec turboshaft engine,the small inlet area permitted a number of arrangements that would not beas practical in larger size engines. One result was the inlet scrollseparator shown in Figure 4. This separator has approximately the samevolume as the T700 type separator for the 2 lb/sec engine but requires aninlet air opening of only 4.36 in. diameter. Although the inlet diameteris small, it is over 50% larger than the diameter of the inlet to the com-pressor, so the overall pressure drop in the separator is not excessive.

In the scroll design, a plenum chamber is added at the inlet elbow to col-lect and trap a rcximum amount of major-size foreign objects, preventingthem from going through the separator and scavenge blower. Scavenge air istaken from this collector and from several locations along the scroll.Scavenge air openings are sized so that only smaller particles like sandcan go through the openings. The only exception is the second bleed-offopening, which is at the bottom of the separator. It is iarge enough totake nuts, bolts, and similar objects if they should get past the collec-tion trap. Cleanout covers to remove trapped objects are located at theinlet elbow (collection trap) and at the second bleed-zff opening.

The inlet to a scroll separator is adaptable to a variety of differentinstallations because of the small inlet size. Inlet air can be in-troduced from any direction by designing the inlet so that it can rotatethrough 3600 or so that it can be removed sc that air can be brought inS tangentially. An engine design showing tangential introduction of air isshown in Figure 5.

Thie scroll separator can be used with an integral oil tank, a separate oiltank in the same relative location ana space, or a separate oil tank lo-cated elsewhere on the engine. Some additional cupplemental cooling ofthe oil and anti-icing of the scroll are obtained from the integral oiltank arrangement. It is necessary, however, to have other mi'-ans for coolingthe oil si.nce there is not enough fuel to do the entire job. Methodsinvestigated include:

1. Wrapping coils of oil around the inlet scroll.

2. Using an oil-to-air heat exchanger.

3. Cooling oil by passing through the scroll struts which ,,relocated in front o' the compressor.

38

L'it

r4 H

1 0 1

'-4

-t 0

39.

MaM

0 U)

'.40

CMi CM

CMO

CMi

CM,

CM I _i _ _ _ _

I L'404

CY) ~ ~ ~~ CM J -JU 4'jC

C)i

-1u

40~

N?_ _ _

__

WIZ"The scroll separator design al o can be adapted to other configurations,such as would result from reversing the locations of the gearbox and oiltank.

Anti-icing is less of a problem with the scroll separator than the T700types since there are no swirl or scavenge vanes and only one set ofstruts. It is estimated that proper location of oil tubing plus an inte-gral oil tank will provide sufficient anti-icing capability.

Aerodynamic Losses for Inlet Scroll Separator

A prelirinary assessment was made of the total pressure loss through the2 lb/sec scroll separator. It was found that the design presented nFigures 16 and 17 had a pressure loss of about 13.5 inches of water.

The losses can be assigned to three primary sources:

1. Friction throughout the system.

2. Turning through the inlet to the scroll.

3. Turning through the scroll and into the core.

The latter is both the largest, as shown in Figure 18, and due to a lack ofdirectly applicable data, the most uncertain.

Losses due to both the inlet turn into the scroll and the turning throughthe scroll and into the core were modeled as offset bends. The bends hadR/d ratios which varied from 1.0 to 1.5 and duct Mach numbers which variedfrom 0.25 to 0.30. Those losses were interpolated from the data presentedin Figure 19. Friction losses were based on Prandtl's universal law ofinfriction for smooth pipes as shown in Figure 20. The losses due to turningand friction are shown broken down in Figure 18. It shows that the varia-tion of total pressure loss with percentage design is a compromise betweenfrontal area and pressure loss, since an increase in duct size tends toincrease frontal area but decreases duct Mach number and, therefore,losses.

41

low~

BELIMOUTIIMOUNTING FLANGE(RABBETED)

FACILITY SCAVENGEELBOWFLANGE

COLLECTIONSLOT

90' CONSTANT DIABEND - DIA =4.35 IN,(MEAN) 8ENDRADIUS/DIAMETER =2.0

DIFFUSEREBWSMASBELLMOUTH MOUNTING TRANSITIONSLOT // FLANGE TO ALLOW BELLMOUTH SECTIONCUTBACK TO BE MOUNTED HERE ALSO.

900

AREAANUSPASSAGE

OSTN2.0 IN2 / 300 OSTN

(0.46 IN GAP AEx 4.35 IN wIDEY ASG

0.46 IN.0.20 IN.GAP HIGHTGAP HEIGHT

4.0 INý 3600 4.87 INý

Figure 16. 2 Lb/Sec Scroll Separator Model (Reference Figure 17).

45 -V814900 - 3.40 WN

1,ý350 -5.05 INý

1800 - 6.6 2N

41 2700 - 10.10 IN?2

3150 - 11.70 INý 2250 - 8.40 11Z

360o- 13.46 INý'

Figure 17. 2 Lb/Sec Scroll Separator Cross Section.

42

25 _ _ ____

LOSCOMPONENT1. TURNING THROUGHSCROLL &

20 INTO CORE2. TURNING THROUGH INLET

o XNTO SCROLL3. FRICTION

*15

____________________-0DUCT FLOW AREA CHANGE PERCENT+2+0

Figure E~ Scroll Separator Pressure Loss Variationx With Flow Area.

A 43

DATA TAKEN FROM SAE AEROSPACE APPLIED D LTHERMODYNAMICS MANUAL-SECOND EDITION, PAGE 62

0.4

,',• =6D 3DCHOKE LINE

0.2 1L =" R/D =3.0

00.2 0.4 0.6 0.8 1.0

BEND ENTRANCE MACH NUMBER

0.4rc

'l 0.3

0.2 CHOKE LINE

CL0 SL RID =2.0

0.1l0 0.2 0.4 0.6 0.8 1.0

BEND ENTRANCE MACH NUMBER

0.7

0.6crjI' o" CHOE LINE

C " 0.5 -

L i6D3DQ- 0.4 - I _

R/D :1.0

0.3

0 0.2 0.4 0.6 0.8 1.0

BEND ENTRANCE MACH NUMBER

Figure 19. Total Pressure Loss Through Offset Bends.

44

Al ___

ýI NR... .

TAKEN FROM SCHLICHTING, BOUNDARY LAYER THEORYMcGRAW-HILL BOOK COMPANY, 1968, PAGE 562

AP/q = X Lid'½N

r • PRANDTL's UNIVERSAL LAW OF FRICTION FOR SMOOTH PIPESý I/, = 2.0 LOG (RED A) - 0.8

4.0

3.0

2.0

1.5 -

1.0

0.8 _ K0.7

105 2 4 6 8 106 2 4 6 8 107

REYNOLDS NUMBER (R = Od/v)

Figure 20. Frictional Resistance In a Smooth Pipe.

7IV

5 LB/SEC SEPARATOR DESIGN

Mechanical Design

The 5 lb/sec integral separator is almost a direct scale-down of theT700 separator. The flow path, however, incorporates these changes:

1. The splitter lip is hidden behind the inner flow path wall.

2. The scavenge air vanes are increased to approximately 80%scale of T700 separator to obtain sufficient internal area forporting oil and structural strength. This reduces tne number

7f •of scavenge vanes from 15 to 14.

3. There is a slight increase in size of PTO strut over thedirectly scaled strut.

Other changes to the 5 lb/sec engine are:

1. Increase in size of gearbox and accessories to a larger sizethan direct scale.

2. Increase in volume of oil tank required to store oil to alarger capacity than direct scale.

45

- TIT,

3. Increase in amount of supplemental cooling required from airdue to the relatively higher percentage of oil flow requiredfor a 5 lb/sec over a T700 engine.

The flow path for the model is shown in Figure 21. Because of the largebeneficial effect found in the 2 lb/sec vaneless model when it was testedwith a "hidden" splitter lip, the hidden splitter lip was also incorporatedinto the 5 lb/sec model. The vaneless floi' oath forward of the rainstep iscalled the "design 3" flow path. The design of the core flow path aft ofthe rainstep was directed toward prevention of large adverse wall staticpressure gradients. Figure 15 shows that the pressure gradients are rea-sonable and that they should not cause large total pressure losses. As acomparison, Figure 22 shows the pressure distribution calculated for the2 lb/sec model with hidden splitter lip. That configuration was testedonly to evaluate its effect on separation efficiencies. The large adversepressure gradient on the hub wall aft of the rainstep (indicated by Figure22) is the probable reason for an 11.2 in. H2 0 core total pressure l-,ss.The large gradient caused by hiding the splitter lip on the 2 lb/see modelwas eliminated on the 5 PPS model, and relatively low loss resulted.

The swirl vane for the 5 lb/sec model is shown in Figure 23 as design 3A.It is the same design 3, tested during Phase II in the 15 lb/sec model, ex-cept that its thickness was reduced in an effort to produce a beneficialeffect on swirl frame loss.

The scroll separator design was considered for this size of engine, andseveral design studies were initiated. The studies were stopped, however,when they indicated that the frontal area and separator size would be"larger for this design than for a scaled T700 separator.

Aerodynamic Design

For the 5 lb/sec design, which is almost a direct scale of the T700, theratio of scrubbed area to flow area is virtually unchanged from the T700value. Therefore, the only first-order factor which causes the loss to bedifferent from the T700 is the friction coefficient, which is invcrsely pro-portional to Reynolds number raised to the 0.2 power. On this basis the 5lb/sec separator has 3.7 in. H 20 friction loss. Summation of turning andfriction losses yields the overall inlet pressure loss, which is 9.3 in.H 20. This compares with the 9.1 in. H2O loss for the T700 separator.Second-order effects of size have been left out of this estimate since theyare sufficiently small that their impact will be indistinguishable frommodel construc+ion and surface finish effects. The breakdown of lossescan be found in Table 4. During the Phase V tests, a core loss of 8.3 in.H20 was achieved.

46

L))

c'J

- I 0

< b

4-1

P4-

cJiCL 0)

F- 44S 0

rr.iCL-

F F-

00 Ln00

.00

U- C)

~1 471LFIn

Pq~~ ~ ~ ~ V -.-ý1 -l2=4 miIi M__ ,__ cJý

K _ _ _

14

100<1

LU -J

=> = -

o) 0: LO

481#11c

c\J

CD

Icc/

Cl)

C-1IL C~i Ucw

60

r1)

449

15 LB/SEC SEPARATOR DESIGN

Mechanical Desi n

The first design studies on the scaled T700 separator were made for the 15lb/sec engine because the larger separator was most directly scalable. Theinitial studies included a scaling up of flow path dimensions, with theonly change being to move the scroll scavenge vanes forward to maintainthe same spacing to the splitter lip as for the T00 separator. The T700engine was then photographically enlarged, and the engine separatorflow path was compared with the flow path obtained by actually drawing theT700 flow path to the scaled-up dimension. Comparison of the flow pathswas excellent, and the conclusion was reached that the 15 lb/sec separatordesign could be scaled directly from the T700 separator. The only changesto the flow path of the separator were:

1. Scavenge air vanes were moved forward to maintain the samedistance from the inner leading edge of the vanes to theleading edge of the splitter lip as for the T700 design.

1 2. Bumping of the flow path in the compressor inlet duct for thePTO bearing is necessary in the T700, but it is not necessaryfor the 15 lb/sec engine.

r! Other minor changes are:

1. Oil tank volume and oil flow can be reduced below that indicatedby a direct-scale-up from the T700.

2. Accessories and AGB can be reduced below that indicated by adirect scale-up from the T700.

The scroll separator design was not evaluated for this size engine becauseits overall size was estimated to be excessive in comparison to a scaledT700 separator design.

Aerodynamic Design

For the 15 lb/sec design, which scaled directly from the T700, the ratio ofscrubbed area to flow area is unchanged from the T700 value. Therefore,the only first-order factor which causes the loss to be different fromthe T700 is the frictioni coefficient, which is inversely proportional toReynolds number raised to the 0.2 power. Summation of turning and frictionlosses yields the overall inlet pressure loss of 8.9 in. H 0. This com-pares with the 9.1 in. H 20 loss for the T00 separator. Second-ordereffects of size have been left out of this estimate since they are expectedto be sufficiently small that their impact will be indistinguishable frommodel construction and surface finish effects. The breakdown of lossescan be found in Table 4. During Phase V tests, a core loss of 8.6 in, H2 0was achieved.

50

Vi

15 Lb/Sec Model Features (Figure 24)

The basic construction was modular with the same number of units and flex-ibility of assembly as the 2 lb/sec model. Because of the increased size,the 15 lb/sec model had more extensive use of fiberglass and less aluminum,but all vanes and the splitter lip were aluminum. The fiberglass used insaneL bearing areas of bellmouth, bulletnose, and scroll. cover was hard-coated to reduce erosion. The coating used was a cF.rbide-filled epoxy.

IF Design 3 Swirl System for the 15 Lb/Sec Separator

Based on the test results and analysis of Phases II and III, the swirl sys-tem for the Phase V 15 lb/sec model was redesigned.

Features of the redesigned swirl system, called design 3, are:

1. Increased swirl vane pressure side slope near the vane leadingedge. Figure 25 shows the pressure side slope, defined as the

angle between the pressure side surface and the engine center-line, of four different swirl vane designs. The T700 design 1had 1.5% better C-Spec efficiency and 3.5% better AC Coarse ef-ficiency than the T700 design 2 vane at the same swirl levels.This led to the design 3 vane, also shown in Figure 23.

2. Improved solidity camber and angle of attack selection. TheT700 design 2 separator has about 1 in. H20 less loss than theT700 design 1 separator at the same swirl angle, and the T700design 1 swirl vane has the highest ratio of deviation angleto camber angle of any of the swirl systems tested. Aerodynamicdifferences between these two cascades are:

a. T700 design 1 has a much lower solidity than the T700design 2. The T700 design 1 12-vane cascade has the lowestsolidity tested.

b. T700 design 1 has a -12' angle of attack versus 00 to +50for the T700 design 2.

c. T700 design 1 has the highest camber angle of any cascadetested.

For a swirl angle at the rainstep of 360, the design 3 solidity4was chosen as 2.10 at the I.D., varying to 1.30 at the OD. Cam-

ber angle was chosen as a constant 38.20 compared with camberangles as high as 640° for the T700 and 530 for the 15 lb/secPhase II swirl systems. Angle of attack for the design 3vane was chosen as +40 at the 360 swirl level setting.

'; 5~1

j 51

-.A . 9 3- .099 : ==5 ..-. &O 6 - 4.414 -OAS -

L~-:z;~5&: A I- __ 9 - 2~0=2-.J-Th= L-0II 98 5.05, - -b2f' N.( i. _ -

-2.J4 -1 11 =03 ~ 0 .42-2~ -10555 95 1100 A4S ~ 04-04 -911 01 j-9 A 2 _ oI

0.7- ý- .O - Il &ý0 ~ 62. .4. 99 04 64,0 940 -1 4 80074

t'f - LP jPx - -Lass5 ~531 12. 5ý 8 564 70 915.ZA - ~ 9 42 ~ -~ZOI 5 ±f. .3) IiyG, G75cl -4. 5r,-~?. 85 50 .l t9~o

7q - 4031 4(..04I-50 51 0.558A 12_31150 -7kO 4la 08 --0 1.65 LL]GIs

5. 5 Z .s S-Nsc. -755 No a- ____ -

041M G. 0 C.

4.7 M !420 305

- 0.4 Z

-109 __

M.3D SWIRL VANE 5Ee FT

-MAN7 \HIJSFLOWFAI

REP

Figure 24. 15 Lb/Sec Separator Model.

53

S PftEEDING PAGHffBLANK..NO?1T1D'I _ _ _-ILA'

W i-77

- - -ro

---------- W-.~49 REF 70 ACRO 07

2____ CCUi EC TIQ SCJf%

6fl.1360 VMA

F'~ 2.41,

~,0,9. 71C O ~0 A

* 1054 2.07

.0*1 DTAI AALTERNJATE 511. ýTURCL 55II2( SPUTTER NOSE -CiRCULAIZ PLA~TE

so1 -~,O I0±tME 4 CHANWE~L~CR £Q,3J)RI~55 ~o so

tHIMI~~~~~ SOiiZI4TRUTS EQ Sp

I ~21 4OLLtOW~221 .ATr PLATES

SEE DETAI LA

EXFLOWM~~~h SPTCR VI4 *

5"I__L

*J. I5C~~NEI ro '% *-k n7i1IZ J-

z"9 Z 4z)

-.-

41.0

" 325

30

1/)

LU

- ILiU

LIU

U)

10 , El[ T700 DESIGN 2S/A 15 LB/SEC INITI'4L (T700 IYPE)

] <> DESIGN 3 SWIK. VAN E

u 20) 1 )

10 20 30 40 50 60 70 80

PERCENT SWIRL SYSTEM LENGTH

Figure 25. Vane Pressure-Side Slope."I 5

'4

3. Reduced vane chord and area ruled flow path. Pressure dropwas reduced 2.6 in. H 0 in the 15 lb/sec model when 12 swirlvanes were removed from the 24-vane cascade. A similar resultoccurred in the T700 separator, although a swirl level reductionaccompanied the swirl cascade solidity reduction. This pres-sure loss decrease is not explained by a reduction in swirlvane skin friction alone. To identify the reason for loss re-duction, an analysis was made of the loss due to diffusion andmixing of the swirl vane wakes between the swirl vanes and thedeswirl vanes. The result of the analysis for the T700 sepa-rator is that a difference of 1.8 in. H20 results from thisdifNusion/mixing process. The 1.8 in. H2 0 reducticn plus the0.5 in. H20 reduction in basic cascade loss gives a computed2.3 in. H20 pressure drop reduction, versus a measured 2.9 in.H2 0 reduction, Based on this analysis, swirl vane axiallength was made 62% of the T700 design 1 vane length, and theflow area aft of the vane was designed with less diffusion thanthe T700 design 1 vane. The increased axial distance betweenthe design 3 vane trailing edge and the rainstep providedbetter mixing of the vane wakes prior to diffusion aft of therainstep.

The design 3 flow path shape is shown in Figure 26. The effectsof vane blockage and turning angle were considered in the flowpath design, which was chosen to give a constant or accelerat-ing area through the cascade. As shown in Figure 26, a con-stant area flow path would give too large a local slope at therainstep, probably causing excessive separation aft cf therainstep. For this reason, the compromise flow pnth was chosento lower the rainstep slope and give an area diffusion of 4.8%just ahead of the rainstep. No change was made to the outerwall of the flow path. Capability existed in the 15 Lo/secmodel to shi ft. the outer wall axially if necessary to improveseparation efficiency.

Table 5 lists the geometric paramete~rs of the T700 design Iswirl cascade, which is the cascade that was extrapolatedfor the 15 lb/sec Phase II model. Table 6 lists the sameparameters for design 3. The table sho'vs that the design 3vane has constant camber, chord, stagger. and thickness,thus making it a practical vane shape ftat can be easilycontrolled in manufacturing. The d-sign 3 maximum thicknesswas chosen to allow the vane to be cast if desired.

Prediction of separator performance for design 3 was basedlargely on Phase II performance of the 15 lb/sec model. Pres-sure loss was expected to be reduced somewhat from that of theoriginal Phase II configuration because of similarity to theT700 desigm 2, which achieved the lowest losses in the T700 size.

i64'

1PA

DESIGN 3SWIRL VANE

~f7~ '~IIZIIVANESECTIONCHOSEN COMPROMISETo REDUCE CURVATURE

- l235~(12021)AT RAINSTEP

J.NE OF CONSTAN't '.AEA

C'INSTANT AREA ORACCELERATING FLOW PATH

Figure 26. Design 3 Separator Swirl System.

I ~TABLE 5. T700 SWIRL VANIES - CASCADE SECTION PROPERTIES

Section Radius (in.)______J

Parameter 3.5 14.0 14.5 5.0 5.5 (b.C U. .0

Chord - in, 1~.80T 2.352 2.856 3.274 3.6144 3.966 4.237 4.511

Stagger Angle -36002' 33002' 30050' 2802'-:1 26053* 25o211 2h40o9 2 301 3 'deg min

£Thickness (max) -in. .206 .220 .250 .277 .298 .315 .317 .318

Leading Edge -in. .103 .105 1.06 .i06 .107 .107 .108 lo09

Trailing Edge -in. .045 o050 o048 o048 o047 .047 .046 o046

Solidity 1.97 2.25 2.142 2.50 2.53 2.52 2.79 2.76

T1 .1114 .094 o089 o085 .082 .079 .075 .071

Coamber -deg 6140 580 59-50 514-50 51.0~' 52.0 52.00 53-00

57

Jp

TABLE 6. DESIGN 3 SWIRL VANE - T700 SIZE

i Parameter Hub Pitch Tip

Radius - in. 4.000 5.536 6.575

Camber - deg 38.20 38.20 38.20

Stagger Angle - deg 33.10 33.10 33.10

LE Angle of Attack - deg 4.0° 4.0O 4.00

Chord - in. 2.978 2.978 2.978

Thickness (max) - in. .436 .436 .436

T 14.6% 14.6% 14.6%

Solidity 2.13 1.54 1.30

J

5

TEST FACILITY AND EQUIPMENT

FACILITY

The separators were tested at General Electric, Lynn, Mass., in a facilityhaving electric-motor-driven fans to •upply separately controlled main andscavenge airflows. Figure 27 shows the test cell.

SAND FEEDING EQUIPMENT

As illustrated in Figure 27, typical sand feedi-, iýquipment was used for thetests. The sand was poured into a high-capacity feeder hopper, wire 3twas weighed. The auger feed system can be adjusted by ýontrolling thevariable-speed motor to provide the-desired sand concentration.

The sand from the hopper is dumped into a feeder funnel, where it is en-

trained by shop air and transported to the ingestion nozzles. The'nozz esare fixed to provide localized or uniform sand ingestion.

PRESSURE MEASURING EQUIPMENT

All pressure measurenent probes were manufactured as specialty equipmentfor the model. These probes were connected to water manometcr panelswhich were continually monitored. When data was taken, a visual reading

4i was made and recorded on a log sheet. The manometer panel in the separatorfacility supplied a total of 36 measurement points plus 4 reference points.

INSTRUMENTATION

The following instramentation was used for all tests:

1. Airflow Measurement

a. Inlet Bellmouth - measures total airflow

(1) Four OD Statics

b. Scavenge Orifice - measures scavenge airflow

(1) Upstream Static (1)(2) Downstream Static (1)

c. Sling psychrometer - for wet and dry bulb temperature.

59

LLI

ULJ LLJ z

LLI

LLJ

LLJ LLJ

.n U-

Cd

V)

LLJ -j= <0- ro

PE4

< Licn U.

60

2. Internal Aerodynamic Measurement

a. Cobra .irobe - measures swirl angle and total pressures

(1) Swirl Cascade Discharge - radial traverse capability(2) Core Inlet

(a) Radial traverse capability(b) Circumferential traverse capability

b. Kiel Probe (Measures Total Pressure)

(1) Used at any location which accepts the cobra probe

c. Separation Efficiency

(1) Scavenge sand was collected by barrier filter(2) Feeder for sand to be ingested(3) Ormond beam balance to weigh both feeder and filter

Separation efficiency can be defined in different ways, two of which arecommonly used. The definition based on weight is

1 (WEIGHT OF SAND SCAVENGED)/(WEIGHT O' SAND AT SEPARATOR INLET) (1)

The second definition recognizes a penalty for scavenge airflow. It is

Tic = 1 - (SAND CONCENTRATION AT IGV)/(SAND CONCENTRATION AT SEPARATORINLET) (8)

where concentration is (Sand Flow)/(Airflow). The two efficiencies are re-lated by the equation

= - (1 - rw) (SCAVENGE AIRFLOW)/(COMPRESSOR AIRFLOW) (9)

6 6

I61

TEST PROCEDURE

The following tests were conducted with all models:

1. Vaneless Flow Path Evaluation (without swirl or deswirl vanes)

a. Measure core pressure drop.b. Measure pressure drop at rainstep.c. Measure C-Spec efficiency.d. Measure AC Coarse efficiency.

2. Swirl Vane Evaluation (without deswirl vanes)

a. Measure swirl angle and total pressure loss at the rain-step over an extent of 400.

b. Measure AC Coarse efficiency.C. Measure C-Spec efficiency.

3. Deswirl Vane Evaluation

a. Measure core pressure drop.b. Measure AC Coarse efficiency.c. Measure C-Spec effi ciency.d. Photograph dye traces of separator flow path.

4. Model Tuning - Review test results from 3 and modify modelif necessary to improve model performance. Modifications were:

a. Swirl vane solidity changes and restagger.b. Deswirl vane solidity changes and redesign.c. Flow path changes.d. Scavenge vane solidity changes.

Tests I through 4 were run at maximum airflow and 16.5% scavenge flow. Alltests nad swirl angle data obtained with a cobra probe. A Kiel probe wasused to obtain pressure drop at the scroll exit and core inlet.

The tuned configurations which were established by this test program weretested at 15 different airflow combinations. Measurements were made at100%, 90%, 80%, 60% and 40% flow with about 20%, 16.5% and 10% scavengeflow.

62

TEST RESULTS

2 LB/SEC AXIAL SEPARATOR

The results of the Phase II test program are given in Table 7. Becauseof the relatively high separator pressure loss through the 2 lb/sec model,several modifications of it were tested. The main results seen in Table 7are:

1. In the change from Configuration II to V, Configuration V hadgood AC Coarse separation efficiency, but C-Spec efficiencyS~dropped about 4.7%.

2. When the splitter lip was moved forward (Configuration III), corepressure loss increased about 4 in. H20, but separation ef-ficiency on C-Spec improved only 1%.

3. The vaneless Configuration VII had 5.6 in. H20 pressure drop,indicating that the basic vaneless core flow path of the 2 lb/secmodel accounts for over half the core pressure drop found inConfiguration I and nearly half of the loss in Configuration II.

4. Since trajectory analysis work showed the importance of closedline of sight relative to the core inlet, the 2 lb/sec axialseparator was modified as shown in Figure 28 to hide the splitterlip. This new splitter lip was run in a vaneless configuration,and bhe results are shown in Table 7, Configuration VIII. Sepa-ration efficiency on C-Spec sand increased from 66.2 to 78.9%.On AC Coarse dust, separation efficiency increased from 15.3 to58.2%. These large increases in efficiency came at the ex-pense of a 5.6 in. H 0 increase in core pressure lo~ss

125. For Phase V tests the flow path was redesigned to reduce diffu-

sion on the inner core wall. The off-design tests were con-ducted on Configuration X of Table 7. Results of the Phase Voff-design tests on the 2 lb/sec axial separators are given inTable 8, which shows that the redesigned model had losses con-sistent with design predictions and excellent separation per-"formance due primarily to the hidden splitter lip.

In addition to off-design testing, foreign object ingestiontests were run on the 2 lb/sec model. Listed in Table 9 arejthe 48 items ingested.

63

N~ 'r ~ Ar .(n 0' 3' to N ýtoto to 1to to t- to O0 N a a

0 to V\ \D0 NQ '0 N 0 H4 N m~ It I5 0..- t to to to to to to c o a, a,

0 Nv '0 -r . (\ - \0 tV cj

It/.- Ht ` H- H- I H H H H. H- H- H

''4 m r- )C

;4 0 co 0 0 0 0N0 o 0 a% '0 0

C't ' ' t to to N N HL, NN t- to N- to 14 to No w

N NN N N 0 N, HINo t

to to0 to '0 '0 Ho ao N to toN

P. Ccca~ 0. WA '0r 0 ý 10 C w' '-0 \0 I

o CI N-4 N N N- N N- H H N

E--a' C; '0 )

8b ut N, '0 ' I N v N H to 11 'A0 4N 'I

>0 '0l N0 ' ,C 0 N 0 a, c \0J) -H 14 HH 4 H N H- - - -

to to '0 tCl. 4 'N4 '' a' N t- t (n t

(1) -= N N H H- .4I 0- - - -

H0 '

0 a it 0N Cl ( N 41 'l\ CaCj 1

a P) 0(L) 0 r-4 +0 00 16OG

OH rI - H, 01 1 H H~ 00) H+ CZ +CO+l

C. C ~CO C * I.

H 0 HN Hk r I 4' N WNr - f

4 b

(v r4 4-ý ~ ~ 4ý r ýr

_____(nmM_ ( m4

R WIM1 7 R

RAINSTEP PLITTER LIP

• ~ ~~OR i,.•NAL CONF IGURATIlON'

•'•"• 'HI•DE14 SOLITTER LIP CONFIGUI(ATION

2K

Figure 28. 2 Lb/Sec Htidden Splitter Modification.

SOf theý itenis listed in Table 9, one i/4-in, nut was not separated from the

core airstream by the 2 lb/sec separator. All other items were eithertrapped on the swirl vanes, trapped on the scroll vanes, or completely sep-arated and lodged in the facility filter. As a result of this test., it isobvious that some changes to the POD test procedure must be made when test-ing separators in the 2 lb/sec size range. During testing done at designpoint airflow, it was necessary to stop during the tests and clear thescroll vanes of lodged material in order to maintain design point scavengeflow.

i•Tests which simulated ingesting POD during engine start-up were also con-

ducted. The items ingested are also listed in Table 9. All items placedin the model bellmouth were scavenged, All items placed on the top of themodel bullet nose passed into the core airstream. It should be noted thatit is very unlikely that a foreign object will inadvertently balance it-self on the top of a 2 lb/sec engine bullet nose.

An attempt was made to evaluate the performance of the separator in remov-ing water. Because of difficulties in accounting for water that evaporatesduring a water ingestion test, 2 pounds of water was poured directly intothe separator at design point core flow and zero scavenge flow. Verylittle of the water went into the scavenge system.

65

V..

I A CJ O H \ 0 t-1 Co\ Oj\ C~P m~ Ul -~j -

1-4~c AA IA H '0 .- c- H t- .. - 0 Co IA C0 o0 c t- 00 O C o t-- 65 a) t C0 CAD t- co w to

to'.0 U-\ 0 C\ . A H C H 0\ Co t-- (X) 0\

we 0 '0 c 0 0 '.0 H HCo H w U W W CO (1 Co H-

oý C. -! 00 CA 0

0- 0. o- '.0 O\ \D 0 0.D0-

W) a, C( 0 o m H t- H- O\ co t- m' \D0 \0 mY .--

O)U \ 0 0 tý- CD) tl- '.0 I IA\ H H- H -I COD Co)

0OH0 H- o - CUl N H rH H- H H- H- Hi H H- 0 0 0l

1- ~ CU CUj 0 H 0 NU \0 Wo ~ ~ C .

CM I ) IA CMj 0

0 0

Ox~~ ~ 0U0 C U C U ' 0 t- N 0 '. o A C~, . - '1 I '0o ,l C Co ) .. Co C o . 4 I

4CU CI HH-HI H H- H

H 4 0

o k 0 p . AU r4 0q r0 AU Co r0 Co 0r40 H 0 C) ~ 0 O O 0- r4 H CU G\co H - t- 0 Co \D 0r \.0 c Ci ON Co

02I CP\ 0~~ U H H C H HC CU H H CUt H L H CU0 HY 0E- OD OD -t 0 CC 0 0 _t -

CU H\ \10 U0

0)

Co 0 \ O Coj 1- L- \0 Co OD \.0 COD C.) \ H IA Co o\

M:'. x 0. x t 1- b rA IA IA CU CUý HA 4 1 Co

%Ot H H H H H H- H 0 -0I

ii' 0

p H) CU\ \.4 IA '000 o 0 H CU ('\ 00 IA "Cý~~ ~ ~ ~ ~ ~ ~ H H) H:Y HC H- t ý oN Cj It -C

v) ) : (7 cp cp t- t t- Lr\ Lr66r

IS, A Ir 4 A r r 4 4 A 48

jaw,.•

TABLE 9. FOREIGN OBJECT INGESTION TEST RESULTS - 2 LB/SEC MODEL

Nuw.ber NumberIagested Ingested

Item at Design Pt at Start-Up

#10 Nut, 961OMPo1 or equivalent 10 6(PO8, P19, P28)

1/h in. Nut, 9610M50P02 or equiva- 10 6lent (P09, P29)

#10 Bolt, J643P05 (o.h38 in. shiank) 10 5

Lockwire 0.032 in. Dia (two strands 10 9twisted together 1 in. long)

1/4 in. Socket (normal socket head 3 2from a socket 'wrench)

1/8 in. Allen Wrench 3 4

Rag, 12 in. x 12 in. Cotton, Wool 2 0or Linen

67

7757: 1 7

Grass was ingested during the FOD evaluation, with the result that the ma-Jority of the grass clogged the model swirl vanes.

2 LB/SEC INLET SCROLL SEPARATOR

1 Summary

The 2 ib/sec scroll separator tests evaluated the basic design concept of

the scroll configuration plus five modifications. Test results listed inTable 10 indicate that the scroll design concept (shown in Figure 16) haspotential. The 84.9% separation efficiency on C-Spec sand and the 78%efficiency on AC Coarse establish the feasibility of the scroll design.Testing of design modifications highlighted important performance effectsand pointed the way to improving the basic separation efficiency levels.Scroll separator pressure drops listed in Table 10 are two to three timesaxial separator losses. Approximately 35% of this loss is computed to bein the duct between the separator exit at station zero and the AP measur-ing plane. This large duct loss is due to the swirling flow in the duct,and indicates that installation of deswirl vanes could make the scrollseparator pressure losses similar to axial separator pressure losses.The table shows that a drop of 10.9% on C-Spec sard occurred when the 9g0inlet elbow was removed to make Configuration II. The next separatormodification, Configuration III, was to remove the larger of the two in-ternal scavenge vanes, forming a 4.5-in. collection slot. Although thishad potential, it allowed particles to bounce back into the core airstreamand probably starved the elbow collector of much of its share of thescavenge flow. The net effect was a performance decrease of 11.6% onC-opec sand from the original system.

For Configuration IV, all internal scavenge slots were closed so that allof the scavenge air (except for some leakage) was drawn from the elbowscavenge slot. This approach with 27.1% decrement in C-Spec separationefficiency does not appear to be fruitful.

Configuration V returned the scavenge circuit to the original design andhad a substantial modification (shown in Figure 29) to the core inlet in

San effort to reduce the pressure loss. The goal cf reduced pressure dropS was achieved with a 21.4 in. H 20 improvement at the expense of 9.9% inAC Coarse dust separation efficiency. It may be possible to regain thelost efficiency by rotating the inlet elbow 900 to make its rotation inthe same direction as the basic scroll. The basis for a modification ofthis kind is that the inlet elbow tended to force sand into the area wherecore inlet modification removed what bad been a barrier.

68

co E

Iii -4 S N I 0

0C (7 toA

I O Nl N ' 0

J 10 0 ~ '0 A I '

H' -' H

(1) Ha0 4

E- no Na, 0

04 H 4 : '

a)

E 01 0 10 10 a0 '

Cr)

0 0 ' CQ r" a~ C

O.,4 W." 44 I C

> -'A p 0 C

4I--* 0 0' cc H-i 0 ý uO

'-. xC' V) -4D 0 CS

C- H HPH

i-I 05N C') 0'' 0 S H tCS AC S F-1

H . I>

H 6H9

7 7, F

AEROZERO

/'RIG A REWORK (BATH TUB OPAIN)

INLET

CORE FLOW

SCAVENGE FLOW

Figure 29. 2 Lb/Sec Scroll Separator Inlet Rework.

After testing Configuration V, the bathtub drain core inlet shown inFigure 29, an elbow collection slot cutback was tested. Figure 16 showsthe collection slot cutback, which is Configuration VI. Configuration VI,compared with Configuration V, had a 25% increase in pressure drop, a 2.3%increase in AC Coarse efficiency, and a C-Spec efficiency of 79.9%. TheC-Spec efficiency reflects the net impact of the bathtub drain andthe cutback elbow slot. Because the bathtub drain provides a more openpath to the core inlet, it decreased C-Spec efficiency relative to Con-figuration I.

The Configuration VI calculated loss downstream of station zero was 0.8 in.H 0 less than Configuration V due to the reduction in residual swirl2'

angle. On this basis the increased elbow slot added 3.4 in. H 20 pres-Ssure loss to the Configuration V level. The large loss due to the elbowslot is a result of bhe slot to main passage arec' ratio whicl isconsiderably larger than the scavenge flow ratio. Since the slot lipsin the scroll separator did not have very large leading edge radii, eachslot probably contributed to the overall pressure loss. As an be seen inFigure 30, the cutback slot raised pressure losses across the entire an-nulus, and the bathtub drain caused a major improvement in the cen-ter annulus loss when compared to Configuration I. This suggests that alarge portion of the loss is caused by separation due to the Collectionslot, which is mixed by the time it reaches the measuring plane. The core

70

100 _______•°° 'o' I • <'41.4

S•37.032. 0

1 ~~~90______

1 , 80

S0 BASIC SCROLL + 900 ELBOWS APT 18.8 IN. AT 1.3 LB/SEC

V A BASIC SCROLL + 900 ELBOWBATHTUB DRAIN CORE

4 _________INLET, 60

APT = 10.5 IN. AT 1.3 LB/SEC

SVI CONFIGURATION V WITH50 CUTBACK ELBOW SLOT,.i.' 5 APT = 13.1 IN. AT 1.3 LB/SEC

40

r 1 30

20

10 _-_ _---_ __

--- 44.65 10 Is 20 25 30 35

CORE APT -IN. H2 0

Figure 30. 2 Lb/Sec Scroll Separator Core APT Profiles.

A• 71

.-- --

inlet bathtub drain reduced the entrance loss by increasing the entranceradius of curvature and also by increasing the flow effective area in theentrance region.

Phase IIB, the scroll inlet portion of the program, demonstrated feasi-bility of a scroll type separator in the 2 lb/sec size. The test re-suits indicate that a redesigned 2 lb/sec scroll separator should beabout 24% larger in flow area than the present configuration to achievean acceptable 5 to 8 in. H20 pressuze loss. With no improvements, ef-ficiency levels would be 85-90% on C-Spec. and 80-85% on AC Coarse. Withdesign improvements suggested by Phase IIB results, higher collection ef-ficiency levels could be achieved.

Trajectory analysis activity carried out uLdIer Phase III essentiallymatched test results from the vaneless axial type separators. With somemodification, the trajectory analysis program could be adapted to analyzethe scroll separator geometry. At present, the trajectory analysis treatsaxisymmetric flow paths with swirl vanes. The scroll separator is notaxisymmetric, but could be modeled as axisymmetric over several shortintervals. Using the modified trajectory analysis, scroll collectionslot geometry and location could be changed to improve collectionefficiency.

At present, the scroll entrance elbow is in a plane perpendicular to thescroll plane. Turning the elbow to be in the same plane as the scroll,with the turn rotation- the same as the scroll rotation, would improve thecollection efficiency. The sand would enter the elbow uniformly, migrateto the outside of the elbow, and continue along the outside of the scroll.The present elbow orientation is such that sand leaving the elbow is con-centrated to the rear of the scroll, not the OD of the scroll where thescavenge slots are located. Present model geometry does not allow thiselbow change without significant model modification.

A suggested program for future improvement of the scroll concept for a2 lb/sec separator is:

1. Modify trajectory analysis program to analyze the scrollseparator.

S2. Analyze the scroll with the turned elbow at 1.3 lb/sec and2 lb/sec airflows.

3. Change the existing model by reorienting the entrance elbow,relocating the collection slots, and changing the slot con-tours in sequence to assess the impact of each change.

4. Measure the pressure drop and collection efficiency of eachchange.

S1 5. For the best configuration tested - design, build, and test aset of deswirl vanes at both 1.3 lb/sec and 2.0 lb/sec-airflows.

72

5 LB/SEC SEPARATOR

The 5 lb/sec model had the highest swirlless separation efficienciesachieved during the program, while the swirlless core total pressure losswas equal to the lowest. This shows that the hidden splitter lip designneed not have a large pressure loss penalty. Fig-re 21 shows the 5 lb/secmodel flow path which included a hidden splitter lip in conjunction with adesign 3 forward flow path.

Installation of swirl ranes in the 5 lb/sec separ For, Configuration II,Table 11, produced little improvement in separation efficiency over theperformance of the vaneless flow path. Table 11 shows a C-Spec increasefrom 90.0 to 92.8% and an AC Coarse improvement from 74.2 to 77.4% atthe expense of a core pressure loss increase from 3.7 to 7.0 in. H 0.

The reduced diameter outer vall of the 5 lb/sec separator adversely af-fected the vaned separator efficiency, and a similar result was obtainedfrom the 15 lb/sec model, Cofigurations IX and X, Table 12. In the15 lb/sec separator the penalties for the change were 9.0% on C-Spec and4.8% on AC Coarse dust. The reduced diameter outer wall flow path wasutilized in the 5 lb/sec separator in an effort to keep the core pressureloss low.

During the 5 lb/sec Phase V tests, an anomaly was noted in the test data.Testing at the design point gave 79.8% efficiency on AC coarse dust, whichwas inconsistent with the 10% and 20% scavenge flow result at 100% coreairflow. It was found that the batch of AC Coarse sand used at the de-sign point airflow was different from that used for all other flow condi-tions. When the sand batch used for the off-design conditions was run atthe design conditions, the effic-iency increased from 79.8 to 81.3%.This indicates that a variation of at least 1.5% in AC coarse can resultfrom using different batches of AC coarse sand that are within AC coarsespecification limits. Listed in Table 13 is the size distribation fromboth the sand that gave 79.8% efficiency and the sand that gave 81.3%efficiency.

The Phase V off-design tests were conducted on Configuration IV of Table11. The results are given in Table 12.

/ VTrajectory analysis of the 5 lb/sec separator hidden splitter lip con-figuration indicated almost 100% separation efficiency. Since the swirl-less model test results were 70.7% on AC coarse dust and 89% on C-Spec,it seemed possible that some of the inefficiency was due to sand bouncingoff the scroll vanes. Therefore, the scroll vanes were moved aft 4 in.from their original axial location so that any sand that did bounce wouldhave much less chance of bouncing into the core stream. Test results ofthe 4 in. att configuration gave 65.5% on AC coarse and 86.7% on C-Spec,

which was a significant drop in efficiency rather than an improvement.

73

%-ýO N

r0 tol m~ V\ 0 N 0 10' (O' a' \ to to

Ic 1- tl\ 't 0 \'O 0 to Ia'1 a, a', a'1 a G\ to

cc- to c'J t C\J 1\

4 -4 H r-4H

0 0-

to\ to

t~ to~ to to fI

0~~

N N '0 a) C1 0G 0'

EN 0 H\ 4-t t t

'0'0 ' NN N

--

f04 4 ) H'

p H ý H t4 HW

Q 0'

cs 0 0 * *

-440 ~.0,-

W '0 4 4

c co co Hý w 0 E z.I al I

-4 w = + .I )o Hý In H

0 tb cl0 > 0 - I

d4 1 0H 00

~474

ci2 H ý cna o N

IrW Y L\ 0'0 r-i 0% HD t l- -.r 00 Z ý- _:t Cfl t 0 O~jf72 0 t- t) t (\J OD t- It- t- t\J t 0 t - t\O

0\ H- O c t-L\ '0 0\ co (Y) t- CM '0 \DH

Cl\ 0 Cl LfOD 0o *D c* OD I- .o CC O. .o .- .

4P

P.i *di C\J H H Cl C O O ' \ . Clj Cl Cl1 H~ H H-

-- H H \ 0 - \ G

co 02 Hb C% H 00 \ t 00 OD \0 C t- C OHr \ -CDm) U \ t ý - - ~ ý \ c' - 0 a

(n~-I\ W\ L ~ ~ ~ C Hl C l \ Cl (\J ( \J H \ ('U

VV)P4 - H H r Cl j H C O H- ~ H H- H- H H-

U- PLC _ _ _ _ _ _ _ _ _ _ _

.3- 1 75)IQ~K:, 0

TABLE 13. AC COARSE SIZES

Batch Test Results

6ize Range AC Coarse(kicrons) 79.8% 81.3% Spec

0-5 13.8% 13.7% 12+2%

5-10 10.2 10.2 12÷3

10-20 12.6 13.5 14+3

20-40 24.4 22.9 23+3

40-80 30.5 28.8 30+3

80-200 7.9 1o.6 9+3

TABLE 14. FOREIGN OBJECT INGESTION TEST RESULTS - 5 LB/SEC MODEL

No. No. Separated No. SeparatedObject Ingested at 5 lb/sec at 2 lb/sec Idle)

#10 Nut 10 10 4

i/4 in. Nut 10 10 6

#20 Bolt 10 9 9

Lockwire 10 7 8

1/4 in. Socket 3 3 2

1/8 in. Allen Wrench 3 2 1

76

The reduced performance when the scroll vanes were moved aft 4 in. wasprobably due to lower velocities in the scavenge stream at the splitterlip. Also, the splitter lip stagnation streamline moved aft 2 to 3 in.(as measured by a handheld tuft wand). The reverse flow forward of thestagnation point and around the splitter lip leading edge may have carriedwith it sand that had already been separated. The conclusion is that thereverse flow forward of the stagnation point or lower scavenge inlet veloci-ties when the scroll vanes are 4 in. aft causes a larger inefficiency thanthe bouncing of particles off the scroll vanes into the core when they are

~ in their normal location.

When the 5 lb/sec separator without deswirl vanes was tested with and with-out swirl vanes, swirl angle and wall static pressure were measured. Theyare compared with design prediction. in Figures 31, 32, 33 and 34. Wallstatic pressures for the vaneless model are shown in Figure 31. The pre-diction agrees with test results to within 10% of a velocity head exceptat Z = 5.4 on the inner wall of the flow path. The larger deviation of22.5% of a velocity head which occurs there is probably caused by slightseparation in the diffusing region aft of the rainstep. Measured innerwall static pressures are lower than predicted due to friction losses whichwere not accounted for in the potential flow aerodynamic design computerprogram. Figure 32 shows wall static pressures for the model with swirlvanes and without deswirl v-aes. Agreement is within 10% of a velocityhead for all points. Both Figures 31 and 32 show that the predicted staticpressures along the walls are in good agreement with the measured testre3ults.

Swirl angle at the rainstep axial location is presented in Figure 33. Thetrend of the test results is in very good agreement with predictions, butthe absolute level is high by about 2 deg. The higher messured swirl levelmay be due to the inaccuracies in hardware construction. A large boundarylayer on the outer flow path caused by the adverse pressure gradient andthe conservation of angular momentum results in a larger measured swirl orhooking over of the curve in the region near the outer wall.

Figure 34 shows a compar.Lon of swirl angles at the core measuring station.Agreement is v fin 1 degree except at the hub. The high measured swirlin the tip _3 probably caused by losses. The hub measured swirl in-dicates that angular momentum is not conserved along the hub flow lath.

0' Foreign object ingestion test results on configuration VII of Table 11 are

I listed in Table V'

During the Idle Foreign Object Test, 13 of the items not separated also did

not go into the core. These items laid on the bottom of the modelouter wall since there vas not enough air velocity to drag the items intothe scro3l. Grass ingested during these tests passed through the vanelessseparator into the core, --using no change in separator pressure loss. Thegrass probably would have accumulated on the IGV's of an eligine's compres-sor.

77

15.0 - , -

14.5 _ _

Q TEST DATA /

14.00

0 OD TEST DATAOID 0^_i LOWER

____SPUITTER 0 013.5 LIP - ____ ___ _ ___

V)

L. '0' TEST DATA -TEST DATAt -____

13.0 -____

SPLITTER

EDGE-*

12.5PAIN NI STEP

12.0

$1TEST DATA-1

-14 -12 -10 -8 -6 -4 -2 0 +2

Z FROM AERO ZERO - IN.

Figure 31. Wall Static Pressure for 5 Lb/Sec Separator Model -Vaneless.

78

13.5

13.0

-13.5

12.5

0QOD

* OD (TESTDA)12.0 IlD DAA - ____

*ID (TEST 6LIA)

A LOWER SPLUTTER LIP

11.5 1-I RAINSTEP

-14 -12 -10 -8 -6 -4 -2 0 +2Z FROM AERO ZERO - IN.

f Figure 32. 5 Lb/Sec Separator Model Wall Static Pressure With'Swirl Vanes.

979

_ ~~~~~100 __ _ _ __ _ _ _ _ _

1 ~~90____

80

~~ 2 i 70___F~jI O=~____ ____TEST DATA

ANALYTICAL PREDICTION

50 -

40_ _ __ _

It ~ ~~~30__ _ __ _ _

1 j 20 BOUNDARY LAYERPROIL

10

020 25 30 35 40450

SWIRL ANGLE AT THE RAINSTEP -DEGREES

Figure 33. Comparison of Rainstep Swirl Angle Prediction With Test Data.

80

1 1 10070

70

60S

4 0 -

CAFD PREDICTION

TEST DATA

10

'40

025 30

SWIRL ANGLE AT CORE MEASURING STATION - DEGREES

Figure 34. Core Swirl Angle With No Deswirl Vanes.

81

N 0ý

a, C6' ( ý 00 a,0 W '

s: C7\ a, to to Nýto to a , a to

0t rH 1- 0 to C., a' 0

co toN 1-4 0~ tO 0l HI0o a'4 to 0t 0 -

HH1- H Iq 1-' H 1.4 H- 14

.- ~E-'\0 Cl\ It toN m' 'N(N~N 0

0 k to 'N to to 0' ('N r, ' (N OWa 0 'Lk 0 I II

r- H

;4a O0 ao I I II0 JN a' I 01 N r C I\ I\ tI Ir I IH- A H -4

HH

to OH tit H 0 oHq *-.- H ' ) CN

E-O 0 H ~ ~ I

M cn oO\ w0 m0 to 0

a)4 0 A. wk92 N .r92

0 (D

r4 ir482 C\ l

A __________>

~~0-

to a' a' a't

>o to a a'0'

-~~ 1-4 1-4 ~ 4-

to I-Cl

C ~ C02 I toii

I -' -4 1.4-

UNN

1 -4 - '

V 0 \0 o D1\. 0 0 \0 1

CoC

-3 -4

2:44

0* *01~-o02 to a)tN ' 0 C "

'0~~~~ r.''0 t 0

* .4-

vi 0 n b

N N1 N- tN o 4

- to )"" ' ,HH o to :: N ~ =1

0 - 0)0ýc

'V) f )t0 0c 0 02coE40ýý "-1a 0 0

c' 'H0 4 - -4 a4 I 4I 00W

, ~ ~ ~ t g o~0 (-H

0l "(-CN 0) b (1)

Q 1.-4 m4 0V c-o ---. 2 2>2 cu j. o) 0 0 cc

0) 0+0

O.-) u 00

83' 40

15 LB/SEC SEPARATOR

Swtrlless 15 Lb/Sec Model

Four vaneless flow paths were tested. The status flow path, ConfigurationIlI, had the best efficiency and also the highest total pressure loss.Table 15 Configuration IV, shows that the design 5 hub, which was con-toured to minimize hub separation just aft of the rainstep, reduced tot&lpressure '&sses and decreased separation efficiency. When the splitterlip was roveC forward relative to the rainstep, Configuration V, the impactwas quite different from a similar test with the vaned 2 lb/sec model.In tha, case, pressure drop increased by 30% and efficiency improved by1.0 aeid 0.5 points for C-Spec and AC Coarse dusts, respectively. The15 T.o/sec model performance showed negligiole change in pressure drop andin C-Spec efficiency and a 2.2-point decrease in AC Coarse efficiency.Neither of these configurations showed strong evidence that appreciableperformance gains could be achieved through changes in the deswirl region.Th,.refore, -be status deswirl flow path was reinstalled when the design 3swirl section flow path was used. This combination had 12% less pressureloss, equivalent C-Spec separation efficiency end 4% poorer AC Coarseefficiency than the original swirlless status flow path.

Figure 35 is a comparison of core, radial pressure drop profiles for thestatus and the status plus design 3 swirl flow paths. Also included forreference are the status systems with 24 and 12 swirl vanes. Figure 35shows that both the hub and tip total pressure losses improved sub-stantially in both swirlless configurations. The swirlless profiles areessentially flat at about 2 in. H2 0 AP fol: the center 70% of the ch,',neland have boundary layers which cover about 15% of the flow at both theouter and inner walls. When vanes are installed in the status flow path,the 55-60% area shows little additional loss, but the hub and tip lossesincrease by 5 - 3 in. H2 0.

"Design 3 Swirl System Test Results

The 15 lb/sec model was run with the complete design 3 swirl system at bothnominal and 60 open stagger angles. The results are listed in Table 15.Fi F~gure 36 shows that the model's performance on C-Spec sand was substan-

4 tially better (1.9 points) than that of the 12-vane status system at agiven swirl level. Its peirformance was almost equal to that of the 24-vanestatus system. Figure 37 shows that the best efficiency demonstratedon AC Coarse in the 15 lb/sec size was achieved with the design 3 swirlsystem at nominal stagger angle. The pressure drop of these swirl vanes

'i-.,th the status deswirl vanes is halfway between the 24- and 12-vane statusswirl cascade performance. The design 2 swirl cascade produces a differ-ent swirl profile from the status cascade. Therefore, it was probablymismatched with the deswirl cascade, so a deswirl cascade redesigncould be expected 4o move the design 3 performan-e toward the 12-vanestatus levels shown in Figure 38.

'84

Utilizing the 15 lb/sec model's capability for rapid inexpensive designchanges, twelve different configurations were tested aimed at improvingand understanding separator pressure loss.

As shown in Table 15, the first configurations tested involved a filledouter flow p~th and a step diffuser flow path as shown in Figure 39. Thdtheory of these two configurations was to delay and control separaticoi onthe outer wall of the separator. After running the entire configuTationshown in Figure 39, the 1. 8 -in. ring was removed and the test war rerunwith just the step diffuser. Neither configuration lowered pressure loss

4.significantly, but as can be seen in Table 15, a drastic reduction inS~separation efficiency occurred as a result of the change.

The next two configurations involved putting boundary layer trip wires atthe separator entrance along the ID and OD. These trips were meant tocreate turbulence in the boundary layers and thus lower the swirl section

losses. The test results show that the trips actually raised the pressureloss.

A series of tests was run with the swirl vanes "reclocked" relative tothe deswirl vanes. These tests were most encouraging since a aet pressureloss reduction of 0.3 in. H2 0 was reai zed. Figure 40 shows that the totalvariation in loss was 0.5 in. H2 0 for this 18-vane system.

Figures 41 and 42 show + t the installation of the design 3 swirl vanes in"the status flow path improved separation efficiency as a functionof pressure loss. One effect of this change vas the "high gain" of theswirl system. The higher velocity through the cascade caused higher swirllevels at constant swirl vane stagger angle. This caused an increase in

/ core pressure loss. The stagger angle of the swirl vanes were, therefore,reduc.ed to return the separator to the same absolute level of pressure loss.

" II

ii

85

TIP -100

80 __

70-

60

LzJ

S50

0 STATUS (24 SWIRL VANES)40 •12 VANE SWIRL FRAMEO] STATUS VANELESS

A DESIGN 3 SWIRL AND STATUS DESWIRLVANELESS

30

I ,

'20

10

0 5 10 15 20 25

CORE APT - IN. H2 0

Figure 35. 15 Lb/Sec Separator Exit Radial ?rofiles at Design Airflows.

86

•.96 . g xs

Ica

94 -'~i 0- cs , Vn..

fT

92

A-

i "-

U--90 _ _.1,__ -

lU_- T700• u <>2 24-VANE MODEL

81 0 18-VANE LCSF MODELK C-' ~88 ~---'

15 LB/SECSy STATUS TEST0 12 SWIRL VANES

6 DESIGN 3 (18 SWIRL VANES)_ _ _ _D DESIGN 3 (60 STAGGER)

86 'I, ýEA2 LB/SEC

I"7STATUS TEST

I013 SWIRL VANES

84 .... _ I

30 32 34 36 38 40 42SWIRL LEVEL DEGREES

Figure 36. C-Spec Efficiency vs Swirl Level.

87V1

88 ____

~84

¶ C3

CD0U 0T70

<80 ~24-VANE MODEL012-VANE MODEL018-VANE LCSF MODEL

________ ________ __ DESIGN 3 (18 SWIRL VANES)78 tDDESIGN 3 (60 STAGGER)

OSTATUS TEST013 SWIRL VANES

76 - _ _ _ _ _ _ _ _ _ _ _ _

30 3343 840 42

SWIRL LEVEL - DEGREES

Figure 37. AC Coarse Efficiency vs Swirl Level.

88

~~ir

[118

T70016 < - I02-VANE MODEL

1.. 024-VANE MODELu064-• ENGINE 002 HARDWARE

,,, - 018-VANE LCSF MODEL

12

0""15 LB/SEC

•I° OSTATUS TEST

• ¢12 SWIRL VANES

r ADES0GN233(18 SWIRL0VANES)

10ESIN 3(0SAGR

8 6.e~2 LBSEC

OSTATUS TESTN1 12 SWIRL VANES

I~WSG 30 32(63 30 42AGG 44

SWIRL LEVEL -DEGREES

Figure 38. Core Pressure Loss vs Swirl Level.

S89

LOLE

LL.

-j-LLI

-44

ILL <

00 ca

0

rZ4

(.010

WA-L

7t4

90

1 00

0

00C)o1

'-4

0

'-4

ol~ c C

0~~ H H

91- ~

, ft-p "

00

ca 03

CD.

(A (A V) 0) 0U.) (A + 41)= ) ... 1±1 tdF- 0 ry p

F- I ) CU

Z: + +

-j CD CD cDu u w C~CD CDCD I DH< cc t ><o C(- I-- H- - Lii0u -

V, Vo 0) 0V-)Ccr o-C o (0 1-<-:

r-. to0 ýV

Q" CD~ CD CD CD L) CD

LLJui LJ -J LJ -J l

CMt C)\J t

-A)N31JIJJ3 3SdiYOO W

92

L 0

=D V

-J

C)"

LI)+

-U 0C-1 Lo LJ-

LI + +

LA ) >-) = = Q:) =wI w- W - )Q

+ <

.C\J ul) (n) () V) V) 1/) 0U) Lii LLI ii P

CC O 0 0

0-V) +LLJ

'*1 UI) V

< Le)

C\I " UJ LLC.(

Zoo< 0 0 C> *f ____________4

0O

* p -+ )Ix

:C1

CDL~

~' Ico CI con

% A913143101fl 0dS-3

' 93

Figure 43 shows the triangular trailing-edge extension to the swirl vanewhich resulted in an improvement of 0.8 point in the separation efficiencyof both C-Spec and AC Coarse dusts at constant pressure drop. Unlike the

Sflow path change, the trailing-edge modification resulted in no significantI change in swirl level, as shown in Table 15.

"- __ VANE INTERSECTIONWITH FLOW PATH /

M/

/ t• RAINSTEP

4 j jVANE INTERSECTIONX4 WITH FLOW PATH

/i Figuire 43. 15 Lb/Sec Design 3 Traillng-Edge Extenbion.

The 15 lb/sec model was finally tested in a swirlless configuration with a

hidden splitter lip. Results, listed in Table 15, show a 8.9% increasein C-Spec efficiency due to the hidden splitter lip at a 1.0 in. H2 0

j pressure loss penalty. AC Coarse efficiency improved 4.1% due to thehidden splitter lip. These improvements, found by comparing Configuration

-~ III to XVII in Table 15, are greater than the normal trade-off of pressureloss for efficiency achieved by swirl vane stagger angle change.

94

A

Foreign object ingestion results are listed in Table 16 for the swirllesshidden splitter lip configuration.

TABLE 16. FOREIGN OBJECT INGESTION TEST RESULTS - 15 LB/SEC MODEL

No. No. Separated No. Separated atItem Ingested at '5 Lb/Sec 6 Lb/Sec (Idle)

#10 Nut 10 7 9

1/4 in. Nut 10 9 9

#10 Bolt 10 9 10

Lockwire 10 8 9

1/h in. Socket 3 3 3

1/8 in. Allen Wrench 3 3 1

Off-Design Testing

Required parametric off-design tests were carried out on Configuration VIIof Table 15. Results are listed in Table 17 and plotted in Figures 44and 45. Figure 44 shows increasing efficiency with increasing scavengeflow. The rate cf gain, however, diminishes as the scavenge flow approaches20% of the core flow. This indicates that there is a trade-off of ef-ficiency against scavenge power for levels of about 20% scavenge flow. Thelines of constant core flow appearing in Figure 44 show decreasing separa-

tion performance with decreasing flow, This is to be expected since theg-field created by the swirl system decreases with reduced velocity.

Figure 45 shows the effect of varying core flow on the core total pressureloss -or three different levels of scavenge flow. While there appearsto be about a 1 to 1.3 in. H2 0 penalty in going from 10 to 16.5% scLvengeflow at constant core flow, the change from 16.5 to 20% seems to have onlya small (<0.3 in. H20) penalty. Therefore, for operation between 16.5aad 20% scavenge flow, the results of these tests indicate that scavengeflow should be based primarily on a tradeo-off of separation efficiency vsscavenge power requirements.

95q _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ __ _ _ _ _ _ _

L- 0 \ 0 i \ ý \Dt- -Q 0\ -- Y)

(m H t-00 \ 0 0 H\ Lr\ 0- 00 CC) t- CU

Hý Hý CO 1- 11; O\ "0\ Cý cl 41% CC)~~ HKU H\ \ - 0 \0 0 C\ \D 0 0\ \0 H O \- 8?

HU (V ( t - l HU 0\ 0 -i r-1 H~ Cq H (M

m) H~a~ 0 t-\ L\0 0- 0'\ " 0 C \ C) H 0\ \ 002 W-' H H H H H H H H H H U

02\ C\) Lr) Cr\ G\~ 0Y mr 0 H- HC m~ Lf '\ o\OC~- \j Hr 0r t- 0f .4 O\ 0Y 0f 0o 0 \ O' H r 0 0OHA .4 Itto' ~ U \ O A \ 0 '

ca H HH

E-1 ~ CU 0 \ C 0 ~ O\ CU t- 0 \O 0 0 C\ 0~ r-i \U t-coo)Ho r-- "00 HY "0\ 0' 0 "0\ 0 Ho co C\ C) t0 0o0) 0)~ H4 H- H CUi H H- HU Hf H U H U

02

02

02 0 0- 0Y 0 CUN (Y) CU kO 0 0 O'\ CU H\ -r t-

HC~. Ht H -H H H HI N r ri ( H H H- -l 0

g- 0 4) H - C Y -4 O - L' t- C - I' C

0 4)- CU a\ f CO 0t \ CO (' 0 fli 0 CO - "0 CM 0 C\U (to t 0U CUr- H C U1 r H Hr- H- H H-

HH 02 mt Itt \0 t-- 0CC) C\ .4 H CUj CU _:t LI C

HC) )~- H H H H H- Hi H- P4

1 96

__ -7- --

182.5

DESIGN 3 SWIRL VANESAT NOMINAL STAGGER

80.0

0~77.5

OREFLOW REF. TABLE 17)' / 0O 15 LB/SEC

75.0 ,y--Z "13.5 LB/SEC( ([]yl 0 12.0 LB/SEC( /0 9.0 LB/SEC

'0 6.0 LB/SEC

72.57.5 10.0 12.5 15.0 17.5 20.0

SWs/W1 - %

Figure 44. AC Coarse Effiziency vs Scavenge Flow Ratio - 15 Lb/Sec Model.

12 SCAVENGE

0 APPROX = 20%

"10 -0 APPROX = 16.5%

>APPROX = 10%

8

0

5.0 7.5 10.0 12.5 15.0 17.5

wcore -LB/SEC

Figure 45. 15 Lb/Sec Core Pressure Loss.

97

-' - " ' " ' ' ol (mm mw i m m m m # m • wm mm m } o m ( lm" m m m m m• 4

m mm ml mIm •m• m mtj

SEPARATOR PERFORMANCE COMPARISON

PRESSURE LOSS

Figures 45, 46, and 47 show the core pressure loss as a function of coreflow and scavenge flow for the 15, 2, and 5 lb/sec separators respectively. ICore loss increased appreciably with the core flow for each separator.Scavenge flow affected loss insignificantly for the 2 and 5 lb/sec separa-tors. For the 15 lb/sec separator, core loss increased as the scavengeflow increased from 10% to 16.5%. Core loss increased less as scavengeflow increased from 16.5% to 20%.

The scroll pressure losses for the 2, 5, and 15 lb/sec models are shown onFigures 48, 49, and 50 respectively. The figures show that there is adifference in trend between the 9 lb/sec separator and the trend for the2 and 15 lb/sec separator. In the 2 lb/sec and 15 lb/sec separators thescro'l total pressure loss increases with increasing scavenge flow. Pres-sure loss in the 5 lb/sec separator, however, remains essentially un-

1 changed when scavenge flow increases at constant core flow.

Figure 51 shows the core pressure loss radial profile of the various 2

lb/sec configurations tested. A 50% reduction in swirl cascade solidity,shown as squares, resulted in a general decrease in core total pressureloss with respect to the status profile. With a 5' stagger increase inthe swirl cascade, the profile remained generally below status levels.Forward movement of the splitter lip generated a change in shape of theprofile, with the hub and tip experiencing greater deterioration thanmidstream. Since it caused only small (+l point in C-Spec or +0.5 pointin AC Coarse) efficiency improvements, this change was not consideredfurther. Running of a vaneless flow path indicated that the greatestimpact of blading on losses occurs in the tip region. These losses areprobably partly due to hub losses which show up downstream at the tip dueto secondary flows in the deswirl cascade. They are probably also causedby the scavenge entrance - splitter lip - deswirl vane tip portion of theflow path.

OFF-DESIGN PERFORMANCE

Figure 52 shows AC coarse efficiency as a function of scavenge flow ratiofor the 2 and 5 lb/sec separators. The 2 lb/sec model with 1.5 degreesgreater swirl and a higher g-field due to its smaller size consistentlygave higher separation efficiency than the 5 lb/sec model. At scavengelevels above 20%, the 2 lb/sec separator showed little difference in separa-tion efficiency between 100% and 40% core flow. The 5 lb/sec model didnot demonstrate this characteristic and had at least a 3-point difference

98

1f

over the same range of core flows, with the highest core flow showing thehighest efficiency.

Figure 52 shows that the separation efficiency of the 5 lb/sec model in-creases with increasing scavenge flow. The rate of gain, however, dimin-ishes as the scavenge flow approaches 20% of the core flow. This indicatesthat there is a trade-off of efficiency against scavenge power for levelsof scavenge flow above 20%. The lines of constant core flow appearing inFigure 52 show decreasing separation performance with decreasing flow. Thereason is probably that the g-field created by the swirl system decreaseswith reduced velocity.

16

14

12

1 0

S8 --II

K ~41S CAVENGE

0 APPROX = 20%

2 APPROX = 16.5%

Q APPROX = 10%

.5 1.0 1.5 2.0 2.5Wcore - LB/SEC

Figure 46. 2 Lb/Sec Core Pressure Loss.

99

12101

10 01

I ~8 ,

O:

•" 6

!I'1 --- E

SCAVENGELl APPROX 20%

21 AAPPROX = 16.5%

QAPPROX = 10%

2 3 4 5 6

Wcr LB/SEC

Figure 47. 5 Lb/Sec Core Pressure Loss.

4 I 100

AP7

- /24 1

22

20-

18-

SCAVENGE

C3 APPROX = 20%16, APPROX = 16.5%

Q APPROX 1 10%

:1o 14

2- 12<]

1 16

20 .5 1 .0 1.5 2.0W - LB/SEC

co-e

Figure 48. 2 Lb/Sec Scroll Pressure Loss.

101

12

SCAVENGE

0APPROX = 20%

10 AAPPROX =16.5% ____

Q APPROX =10%

6

4_ _

011 23 4 5 6

Wcore -LB/SEC

Figure 49. 5 Lb/Sec Scroll Pressure Loss.

102

25.0

22.5 -

SCAVENGE

"APPROX = 20%

20.0 Z% APPROX = 16.5%

0 APPROX = 10%

:j 117.5 _

C• 15.0 ' _

4 I •-fI-• 12.5

10.0

4 ~~~~~~7.5___________

-p- 5.0

2.5* 5.0 7.5 10.0 12.5 15.0 17.5

Wcore - LB/SEC

Figure 50. 15 Lb/Sec Scroll Pressure Loss.

103

- ____- -- ----

100-

90

80

70

60 ______________

400 STATUS

25 3

CURE ~P I.H0BLOAMIN

N7VI

F 104

-__0 -

! 0 5_ a1520253

COR AP IN0 0 BEO VMIENT

202

10 -• 104

1 90.0

I I

85.0 .. . .87.

5

Lu 82.5;5

~~ 5- BSE OE77,5

80.5

LB/SEC--S2CLMODEL2- I -. MODEL

77.5

0 0 100%

90

72.5

10.0 12.5 15.0 17.5 20-.0 22.5 25.0

W scavenge/Wcore -%

Fi~re 2.AC Coarse Collecti~on Effici•ency vs Scavenge System Ai[rflow.

105

z ),

4.... i • ,• r • . . • r • .

CONCLUSIONS

1. The program established the feasibility of swirlless inlet particleseparators. More development is required to achieve their potential.

2. The 80% separation efficiency goal of the program was exceeded bythe 5 and 2 lb/sec models by about 1.5% and 3%. The 15 lb/secmodel, which has a lower g field than the smaller models, had about78% separation efficiency.

3. Sand concentrations of Specification MIL-E-5007C are lower than thosefound around operational helicopters.

4. Sand size distributions for operational helicopters lie betweenSpecification MIL-F-5007C ;nd AC Coarse distributions.

5. The method of injecting sand into a separator inlet can have a strongeffect on separation efficiency.

6. The program established the feasibility of inlet scroll separators.

RECOMMENDATIONS

Continue to conduct a development program that makes use of the existingseparator models. Systematically vary the following features to obtainimprovements in separator performance:

1. Improve collection efficiency and reCuce collector scrollpre:3sure loss by varying the separator outer wall shape.

2. Improve collection efficiency and reduce collector scrollpressure leas by systematically varying scroll vane solidity,twist, and camber.

3. Improve collection efficiency and reduce scavenge flow re-quirements for swirlless separator configurations by varyingthe ratio of inner flow path maximum diameter to splitterlip diameter while outer flow path and scavenge vane designsremain fixed.

4. Investigate the inlet parameters of sand injection to determinetheir influence on collecticn efficiency; to establish engineinstallation design rules and to develop appropriate componenttest sand injection techniques.

106

LIST OF SYMBOLS

!2a acceleration, ft/sec2

A area, ft 2

c chord length, in.

c loct:l sand concentration, mg/ft3

c5 average sand concentration, mg/ft 3

parameter - 7r Cd pVI8pZD

Cd drag coefficient

id differential

D diameter, In.

DcD critical particle diameter, micron' i cr

1 e unit vector

I f function

fa depth of erosion, milst a

,fd normalized severity factor

Sg acceleration due to gravity, ft/sec2

his insentropic head, ft lb/lb

1 length, in.

SL axial distance downstream from splitter lip, in.

m particle mass, lb

M Mach number

F speed- rpm

N specific speeds

P pressure, lb/in2jQ volume flow, ft 3 /sec

107

1 Ne g~

LIST OF SYMBOLS - Continued

R radius, in,

SR, outer wall radius at the separator inlet, in.

Re Reynolds number

SF scale factor

t time, sec

t maximum vane thickness, in*! m

u gas velocity, ft/sec

U blade section tangential velocity, ft/sec

v particle velocity, ft/sec

vn particle velocity normal to impacted surface, ft/sec

v t particle velocity tangential to impacted surface, ft/sec

V relative particle velocity, ft/sec

V0 air circumferential velocity, ft/sec

W mass -•irflow, lb/see

z axial distance downs.°rear 1 rom compressor station zero, in

Z particle shape factor

0 •!particle incidence angle, deg

02 particle rebound angle, deg

i APT total pressure loss , lb/in?

TIseparation efficiency

, nc separation efficiency based on sand concentration

T1w separation efficency based on sand weight

108 t4 ' _

LIST OF SYMBOLS - Continued

Tv vane passage efficiency

1P air viscosity, lb sec/ft 2

P air density, lb/ft 3

ZAt sum of areas, in?

stream function

I2( ) partial differential

% weight fraction of sand with particle diameter less than DP

SUBSCRIPTS

a refers to air

c refers to core

g refers to gas

2.. refers ts local conditions

m indicates maximum

o refers to average conditions

p refers to particle

! I r radial coordinate

s refers to sand

t indicates stagnation conditions

V refers to relative velocity

Z axial coordinate

E circumferential coordLae2e

8175-757

109

~i I