-Tech Me UNLIMITED T--AD-A237 559 Tech Memo

P 1204 P 1204

ROYAL AEROSPACE ESTABLISHMENT

Technical Memorandum

February 1991

Application of S1 BYL2 to the AGARD WG18

Compressor Test Cases

by DTICELECTE

W. J. Calvert S J

DEFENSE TECHNICAL NFCRM HTIOH ENTER

91033?3Procurement Executive, Ministry of Defence

Farnborough, Hampshire

UNLIMITED 2 z5' 004

CONDITIONS OF RELEASE0099152 301533

..................... DRIC U

COPYRIGHT (c)1988

CONTROL.LERHIMSO LONDON

..................... DRIC Y

Reports quoted are not necessarily available to members of the public or to commercialotganisations.

UNLIMITED

ROYAL AEIOSPACAE ESTABLISHMIENT

Technical Memorandum P 1204

Received for printing 27 February 1991

APPLICATION OF SIBYL2 TO THE AGARD WGI8 COMPRESSOR TEST CASES

by

W. J. Calvert

SUM UMRY

SIBYL2 is an inviscid-viscous blade-to-blade method for calculating the

detailed aerodynamics and overall performance of compressor blades. It may be

applied either on its own to predict the flow for individual blade sections, such

as the mid span of a linear cascade, or in conjunction with a throughflow calcu-

lation to predict the performance of a complete axial compressor.

A previous AGARD paper by the author described applications of SIBYL2 to

most of the compressor cascade test cases which have subsequently been selected

by AGARD Working Group 18: generally good agreement was obtained. This current

paper presents new predictions for the V2 and ARL SL19 cascades and for the high

speed compressor cases. It is hoped that this will be one of many sets of calcu-

lations for these cases, so that an improved understanding of each case may be

obtained, together with an appreciation of the strengths and weaknesses of

different computational approaches.

Copyright©

Controller IIHSO London1991

This Memorandum is a facsimile of a paper prepared for the 77th Symposiumof the AGARD Propulsion and Energetics Panel on "CFD Techniques for PropulsionApplications" to be held in San Antonio, Texas, USA, 27-31 May., 1991.

UNLIMITED

2

LIST, )F CONTLNTS

Page

I INTRODUCTION 3

2 GENERAL DESCRIPTION 01 N2ETHODS 3

3 COMPRESSOR CASCADE TEST CASES 4

3. 1 V2 Subsonic Compressor Cascade 43.2 ARL SL19 Transo i-, ConmpiassoL Gzscade 6

4 HIGH SPEED COMPRESSOR TEST C25ES 6

4.1 NASA Transonic Rotor 67 74.2 RR 11P9 Subsonic Co,pr(-,sor Stage 74.3 DFVLR Transonic Compres!., :, Stage 8

5 CONCLUDING REMARKS 8

References 9

Illustrations Figures 1-8

Report documentation page inside back cover

Accession For

N TIS GRA&IDTIC TABUnannounced

Dlstribution/

Availability Codos

'Avj',l and/or

D i st Special

T11 P 1204

Application of SIBYL2 to the AGbD 111 8 Ccmpressor Test Cases 3

W J CalvertPropulsion Departnnt

Royal Aerospace Establi,,,.vtPyestock, Farnborough,

Hamrpshire, UK

SUMMARY experimental data of many inporant areas ofhigh speed flows, due primarily to the

lBYL2 is an inviscid-viscous blade-to-blade relatively small sczle of turbcmachineryY, tnod for calculat. Yj the detailed aero- blading. Thus sane aspects of the codesdynamics and overall performance of conpres- cannot be adequately validated. Given thissor blades. I- may rx! applied either on its situatinn, theve is ruch to be gained byown to predict -ne flow for individual blade aplying c.-_iP! to a wide range of test cisessections, such es the mid span of a linear aid by caparing the results from different_az r'ide, or in conjL,!ction with a throughflow codes for the same test cases. An irportantcaLu.ation to predi-tr tl performance of a contributi.a of W.18 is to have identified acoaplete axial coapzessor. suitable sut ot test cases to initiate this.

A previous AGARD paper by tCe author descri- The main code employed in the present work isbed applications of SIBYL to rost of the ti.e RAE SIBML2 method 3. This is an invitscid-copressor cascade test cases whicQ. havc. riscous interact.ion technique for predictingsubsequently been selectea -y AGARD Working the blade-to-blade perfonmnce of axialGroup 18: generally good a;'ement as ccxlpressors. It can be applied either on itsoble-ned. This current paper presents new own to individual blade sections, such as thepri. iitson-s for the V and A&!. SL19 cascades mid span of ". linear cascade, or linked to aanj ior tne high speed conpror cases. It strealine curvature throughflow method in anis hc'e. that this will be one of many sets Sl-S2 system to give a carplete quasi-three-of ca culations for these cases, so that an dinnonal prediction of the flow andimprove undetstanding of each case may be overall perfomance of one or more bladeobtained, together with an appreciation of rows.the strengths and weaknesses of differentcaputational approaches. A brief description of the SIBYL2 and Sl-S2

nethods is given in section 2 of this paper.LIST OF SYMOLS Section 3 the- reviews the result6 fran

previous applications 4 of S1BYL2 to thec blade chord compressor cascade test cases chosen by W.18H boundary layer shape factor 8*/0 and presents further predictions for the V2M Mach numfber high subsonic cascade and new predictions forP static pressure the ARL SL19 transonic cascade. Results froma flow angle, degrees the RAE Sl-S2 system for the tiree high speed,&* boundary layer displacement thickness ccrpressor cases are then given in section 4.0 boundary layer mmentun thickness(0 loss coefficient based on upstream 2. GRAL DESCIPTIO OF METHMS

conditions92 stream tube contraction (upstream! The RAE SIBYL2 method is an inviscid-viscous

downstream) interaction technique to predict the blade-to-blade performance of axial carpressor-.

Subscripts The inviscid part consists of a tine mardhingEuler calculation, based on Ref 5, so it can

conditions at upstream boundary of handle transonic flows: shock waves can becascade or inlet to blade row captured and the losses due to them can be

predicted. The calculation takes place on aconditions at downstream boundary of specified axisymmetric stream surface, and itcascade or exit from blade row includes the effects of rotation and of

varying radius and stream tube thickness in1. INTRMCTICN the axial direction. The viscous calculation

is an integral technique consisting of threeValidation/calibration of carputer codes parts to estimate laminar boundary layeragaList suitable tc ;t cases is a highly development, transition point and turbulentirportant area of Crrutational Fluid boundary layer development respectively.Dynamics which is rightly receiving increased The inviscid-viscous matching prccedureattention, as shown for exarrple by Ref 1. employs mixed modes to allow a valid solutionThe current state-of-the-art for turbo- to be obtained even when there are regions ofmachinery test cases is indicated by the separated flow due to shock wave/boundaryselection made by At3RD Working Group 18 layer interactions or to excessive diffusion.(W318). These cover a useful range of The viscous calculatiois continue in the wakepractical cascades and rotating machines, downstream of the bladed regi on, and aand provide a reasonable amount of data on ccrpressible flow mixing calculation isthe flow fields as well as measurerents of carried out on the downstream boundary tooverall performance. However, as nnt-d hy A. e-in +h e vit Conditio,.q T'.-_Q theWG!8 in Chapter III, neither the experi- method provides both details of the internalmental data nor the flow codes are perfect. flows, including boundary layer development,Also it is not generally possible to obtain and predictions of the blade section c¢erall

TM P 1204

4

pe-formance, such as exit flow angle and specified as an input to the S1-S2 system.lsses. Thev additional losses are modelled by a

cosjtant dcag force in the S1 calculation andIn order to enploy the SIBYL2 methcd to L" U effects are fed to S2 via the normalprecdict the performance of coplete blade SI-S2 linkirg. Although the need for suchrows, it .. ,st be linked with a throughflow input i6 one of the limitations of the S1-$2calculdtion which deteunines the spanwise approach, the ease with which such correc-variations of the flow and ensures that tions can be incorporated in order to tureradial equilibrium is satisfied. A semi- the model for a given ccapressor can be aautoatic SI-S2 procedure for this is descri- distinct advantage in practice. It isbed in Ref 6. This procedure has now been eiphasised, however, that apart front theautomated, with both the SBK'L2 calculation stated end losses and annulus wall blockagefor the blade-to-blade sections ana tae factors no other corrections were employedstreamline curvature cqlculation for the hub- for the cases presented herm.to-tip flow incorporatud into one progrza.ztogether with the blade geometry ana data 3. CCMPRESSOR CASCADE TEST CASESlink rojtines. The streamline curvatureroutine I is an inproved modular version of AGARD working Group 18 selected fourthat previously enployed. It allows curved copressor cascades suitable for inviscid-calculating planes to be bcecified to give &r" viscous blade-to-blade methods:-accurite match to the leading and trailingedges of each blade row. Spanwise m.xing i E/CA-2 V2 High Subsonic Cascadeterms have been added to some versions of theroutine, but these were not included in the ii E/CA-3 CNMERA 115 High Subsonicpresent SI-S2 calculations. Cascade

The basic pi inciples of the SI-$2 interaction iii E/CA-4 DFVIR L030-4 Low Supersonicremain the same. Essentially there are three Cascadeparts to each loop:-

v E/CA-5 ARL SL19 Transonic Cascadei the performance of the blade-to-blade sections is predicted by SIBYL2 - SIBYL2 has previously been applied to thethis involves calculations for up to 9 first three of these and the results aresections on each blade row. given in Ref 4. There was generally good

agreement between the test results andii the hub-to-tip flow throughout the predictions, in terms of both internal flowwhole conpressor is calculated by the S2 details such as blade surface pressurestreamline curvature method, using the distributions and overall perfomance para-pitchwise averaged data from the SI meters such as static pressure ratio anti exitsolutions, flow angle. The only significant problem

area was that increases in losses at highiii the relative flow angles and incidence were not predicted directly, butstream tube geometry from the Sl and S2 relied on the user noticing the high effec-parts of the calculation are compared to tive incidence indicated by the solution andsee whether the solution has converged - specifying revised starting conditions forif not new blade geometry for each the surface boundary layer calculation. Forsection is found by calculating the cases near optinum incidence, deviationintersections between the stream angles were predicted to about ± 10, staticsurfaces from the current S2 solution pressure ratios to ± 4% and overall lossand the blade aerofoils. coefficients to ± 2% of inlet dynamic head.

These levels of accuracy have been confirmedTarget convergence criteria for the complete in applications of the code to other cas-Sl-S2 solution are typically a) Sl and S2 cades, such as the supercritical designsrelative flow angles agreeing to within 0.10, described in Refs 8 and 9.and b) S] and S2 stream tube thicknessvariations agreeing to 0.1%. The number of The test conditions chosen by W.18 for theS1-S2 loops required to achieve this depends CNERA 115 and DFVIR L030-4 cascades are theon the difficulty of the blade section same as, or similar to, those previouslyoperating conditions a,-: th'P nuser of blade studied. Therefore only calculations for therows involved. Awo..._ e.rg, loops are V2 and ARL SL19 cascades will be presentedgenerally sufficient fi. oite or two blade here.rows, but more will be -_-_ded, for exarple,if blade choking i. - nficantly affecting 3.1 Vz Subsonic Crpressor Cascadethe flow distro cion. The solutions fromthe previous 1oo-p provide the starting The V2 cascade is a high canber (56.80), highconditions for calcul~tirnz on the current solidity (2.22) design with 7% thick DCAloop, so that the procedure i-' quite blade profiles. Twelve test conditions wereefficient. For exanple, a solution for tMe selected by WS38, coverina inlet Mach nzbershr iyp- RD s-cgw see sectiol 4.21 requires from 0.3 to 0.85, inlet flow angles fromabout 6 hours CPU on a Stardent 1500 mini- 47.50 to 54.50 and stream tube contractionsupercarputer. ratios from 1.08 to 1.39.

It should be noted that, in order to obtain As noted in Refs 2 and 4, the axial variationrealistic losses towards the blade ends, assured for stream tube thickness between theempirically derived extra losses must be easured upstream and downstream values is

TM P 1204

critcal, but also open to uncertainty. Two to -any reasons. SaTe rf the more ob iousplausiblt. assumptions could be (a) a linear ones are:variation with strx'anwise distance betweenthe two reasuring planes, and (b) constant i errors in the prediction of tians-values upstream and downstream and a linear ition between laminar and turbulent flowvariation between the leading and trailingedge planes of the cascade. For the V2 ii the effect of high free-streamcascade, which has a high aspect ratio of turbulence on the turbulent boundary3.75, the fozmer is more likely to be correct layer - this is not included in theana this is confinred by the cakculations for present calculationsthe test condition with values of M/0./ of0.80/54.5/1.39 (Fig la). Assumption (a) waa iii the general uncertainty in predic-therefore adopted for all subsequent calcula- ting separated turbulent boundary layerstions for this cascade.

iv errors in masuring very thinTne choice of stream tube thickness variation boundary layers - the reasured displace-ig a good example of a situation where there rent thickness is only 0.4 nn at 42%is uncertainty in the experisental data; and chordhence where ccxmparisons against other methodsrun with the sane assunptions would be parti- v errors in measuring unsteady flowscularly useful, with a pneumatic probe

An 89 x 16 calculating grid was erployed for vi errors in measuring reverse flowsall calculations of the V2 cascade, with 51 x with a forward facing probe16 points equally spaced in the axial andtangential directions within the blade row. The level of agreement achieved in this caseOutside the row, the axial spacing was is typical of tnat for the test conditionsincreased by a constant factor to give up- near opti-tn incidence.stream and downstream boundaries for thecalculation damain 0.8 axial chords from the The remaining two operating conditions -blade row. 0.80/52.5/1.12 ai4 0.80/54.5/1.23 - are

stalled, with reasured loss coefficients ofThe twelve W318 test conditions can be over 10%. As previously noted, suchgrouped into three categories: (a) choked conditions can be identified by the fact thatflow, (b) operation near optimun incidence mnxinun blade loading occurs at the leadingand (c) stalled flow. The three choked flow edge. The user can then specify that theconditions are:- 0.85/49.5/1.20, suction surface boundary layer starts as a0.80/47.5/1.17, and 0.80/49.5/1.27. They turiulent layer, with a mramentum thicknessdcmonstrate that, relative to a nominal Reynolds number of about 500. A solutionunchoked condition of 0.80/49.5/1.20, choked with the boundary layer tripped in this wayflow can be produced by small changes in Mv, for the condition 0.80/52.5/1.12 is shown inCL, or Q. SIBYL2 successfully predicts this. Fig id. The results for surface Mach nuwberFor example, the predicted Mach number and suction surface displacemant thicknesscontours, blade s.r.f-ce Mach number distri- agree quite well but the predicted shapebutions and d-velopment of the boundary layer factors for the suction surface are raichdisplacevent Lhiukness and shape factor at higher than deduced from the boundary layer0.85/49.5/1.20 are shown in Fig lb, together meas nts.with the test data. There is good agreerentbetween prediction ard test, except for the In this context it is interesting to considerlevels of shape factor near thi trailing edge the loss produced by applying a compressibleof the suction surface. flow mixing calculation to the reasured

suction surface boundary layer parameters atThe point at 0.80/49.5/1.12 is typical of the S8% chord. At the two stalled flow condi-seven operating conditions near optimzn tions this loss exceeds the total lossincidence with easured loss coefficients deduced from the wake traverses oy about 25%.below 5%. The predicted Mach number Assuming that the suction surface loss shouldcontours, blade surface Mach number distri- be about the sane proportion of the totalbutions, and boundary layer developrent are loss as at the other test conditions (ie 80%)shown in Fig Ic, together with the test data. implies that the maentum thickness should beIt can be seen that the overall performance about half of the stated value. This wouldparameters match quite well, although the obviously be in much better agreerent withdiscrepancy of 1.30 on exit flow angle is the predicted values of marentum thicknesshigher than usual. There is also close and shape factor.agreexent between the predict.d inviscid Machnumbsrs on the blade surfaces and the alues An alternative approach to tripping thededuced from the reasured surface static boundary layer would be to employ a grid withpressures. The agreement between predi tion nuch closer axial spacing near the leadingand test for the suction surface boundary edge and to assume that all boundary layerslayer Aevelopment is reasonable -amlitati- e )rh ,ulent Thi would givS nionnvely, but the predicted values of shape broadly carparable with the Navier-Stokesfactor are higher than the test values in the solutions for the V2 cascade shown in Ref 10.separated flow region near the trailing edge, However, a grid giving better modelling ofwhile the predicted values of displacement the leading edge region would be needed tothickness near mid chord are lower than the obtain grid-independent solutions. Also thetest values. These differences could be due process of laminar separation, transition andTM P 1204

6

turbulent reattachnent for high incidence surface static pressures and the predictedflows needs to be understood better so that inviscid total pressures. The agreementmore realistic models of the viscous flow between prediction and test is generally(either integral boundary layer calculations good. Note that the calculation indicatesor turbulence models) can be developed, the presence of strong shock wave/boundary

layer interactions on both the suction and3.2 ARt SL19 Transonic Compressor CatSade pressure surfaces of the blade. It is

possible that the pre-shock boundary layer onThe ARt SL19 transonic compressor cascade was the pressure surface could be laminar,derived fra a rotor section of a single although transition has been forced upstreamstag-- compressor 11 with a design point of the shock in the SIBYL2 solution, and thispressure ratio of 1.9. It is a thin, high might erplain the discrepancies betweenstagger (56.90), low canber(-.9 0 ) section prediction and test in this region. Thewith a design point inlet Mach nuaber of agreement between measured and predicted1.616. Cascades with similar blades have overall perfotnanoe parameters is reasonablebeen tested in facilities at Detroit Diesel - the predicted unique incidenoe inlet angleAllison (DDA) 2, DFVLR 13 and ONERA 14, and and the exit flow angle are both 10 too high,one test point from each series was selected static pressure rise agrees to about 5%, andby W318. total pressure loss is 10% compared with a

ireasured value of 12%.A guide to the relevance of each operatingcondition can be obtained by computing the The predictions are quite sensitive to thepressure ratio and efficiency of a corres- assumptions made about stream tube thickness.ponding rotor section from the measured For exanple, taking a constant value upstreamoverall cascade performance - for this of the initial shock wave reduces the predic-purpose the rotor section was assumed to have ted inlet flow angle by 0.80, while assumingzero inlet absolute flow angle and constant a linear variation between the plane of thestream surface radius. The results for all upstream static pressure tappings and thethe DDA test points taken at the design inlet downstream traverse plane increases the valueMach number 12 were as follows:- by 0.8'. There is also same uncertainty

about the appropriate blade shape to use forCascade Rotor Rotor the calculations, since the two instrumentedstatic total adiabatic blades in the cascade both differed slightlypressure pressure efficiency fram the design intent and some blade defonr-ratio ratio % ation due to the aerodynamic loads was

observed during testing 15. Using the reasu-1.220 1.086 67.8 red shape of the suction surface blae given1.468 1.180 77.0 in Ref 2 reduces the predicted inlet flow1.672 1.318 81.8 angle by 0.70, whereas the shape for the1.870 1.489 84.0 pressure surface blade is much closer to the2.036 1.674 84.8 design intent.2.097 1.727 83.82.220 1.801 81.1 Once again calculations with a number of2.300 1.868 82.2 different codes would be instructive and

might enable a consensus view to be reachedThis cltarly shows that the points with of the appropriat3 boundary conditions to behigher static pressure ratios are of greater applied.practical interest to the compressor engi-neer, and it is these which the SlBYL2 method 4. HIGH SPEED OCUPRE$SOR TEST CASESwas developed to predict. Of the WG18 testconditions only the DFVLR point falls into As mentioned above, the S1-$2 systemthis category. incorporating SIBYL2 has been applied to all

three high speed compressor test casesFor the SIBYL2 calculation of the DFVLR test selected by W,18 viz:point the stream tube thickness wag assumedto be constant upstream of the cascade i E/QO-2 NASA Transonic Rotor 67leading edge and to reduce linearly withaxial distance to the measured value at the ii E/CO-3 RR HP9 Subsonic Stagetraverse plane 47% of gap axially downstreamof the blade trailing edge. A 119 x 21 iii E/CO-4 DFVIR Transonic Stagecalculating grid was employed for the timemarching part of the calculation, with 81 x The system was kept the same throughout and21 points within the blade row equally spaced similar calculating grids were applied. Ninein both axial and tangential directions. The S1 sections ware used to model each blade rowtest value of inlet tangential velocity was and the grids used 'or the time marching partspecified and the operating condition for the of the S1BYL2 calculations were similar tosolution was chosen such that the main those used for the cascades in section 3pressure rise on the suction surface occurred above. The S2 solutions typically erployedclose to the measured position. 21 srrPminpQ and tere .... Ab.--

calculating planes between the leading andThe predicted Mach number contours, blade trailing edge planes of each blade row.surface Mach number dist ributions and bound-ary layer development for the design blade The inlet total pressure profiles for eachshape are shown in Fig 2, together with the conpressor were based closely on the measuredMach nutbers deduced from the measured data, except that no attempt was made to

TM P 1204

7

achieve zero values of velC'ity at the walls. However, the calculation predicts a rmreSmall amounts of annulus wal.l blockage, about distinct shock in the un-overed passage at1% for the two transonic fans and 2 % for 30% blade height for boti. conditions than wasHP9, were included to allso for this. The reasured with the laser anenrmeter.losses used to model end effects were assuTedto affect the outer regions of each row: the The predicted boundary layer development forloss coefficients were specified to drop from the suction surface indicates a trailing edge15% at the annulus walls, to zero at a separation for the sections over the innerdistance of 1/3 rotor chord from the walls, third of the span, with the shock wave/This assumption is in line with that made for boundary layer interaction near mid chordprevious Sl-S2 calculations of the rotor row becoming significant and causing sate separ-of transonic civil fans 6 16. If mre than ation over the outer half of the blade. Theone stage were being considered, it would interaction is sufficiently strong at theobviously ce necessary either to include the peak efficiency condition to cause copleteeffect of spanwise mixing within the S2 part separation of the boundary layer downstreamof the calculation, or to use a more even of the shock for the tip sections.distribution of extra loss.

4.2 RR HP9 Subsonic Ccnpressor Stace4.1 NASA Transonic ROtor 67

The Rolls-Royce .T9 rig is a single stageRotor 67 is the first stage rotor of a two model of one of t e later stages in a civilstage fan. It was tested as an isolated row HP ccxpressor. it therefore has a muchto provide data for coparison with numerical higher inlet hub/tip ratio (0.84) and lowerpredictions which were free fram circum- rotor aspect ratio (0.9) than the other twoferential variations due to stationary blade W318 carpressor cases. Inlet spoiler ringsrows. The rotor has an aspect ratio of 1.6 are employed to generate a suitable inletbased on average span/mid height chord, and velocity profile. The type "rD" caseat inlet the tip diameter is 514 mm with a selected by WG18 has conventional bladehub/tip ratio of 0.375. The design point profiles, a design stage pressure ratio ofconditior. are a tip speed of 429 m/s, an 1.24, a mass flow rate of 9.1 kg/s and a tipinlet tip relative Mach nutmber of 1.38, a speed of 251.3 m/s at ISA inlet conditions.rotor pressure ratio of 1.63 and a mass flow Tip diameter is 518.2 nram.rate of 33.25 kg/s. Laser anemometry wasextensively used to neasure details of the All three operating conditions - maxinuminternal flcw fields, and radial traverses flow, mid chic and near surge - were studiedwith a combination probe were carried out to using the Sl-S2 system. The predictions atdetenrmine the inlet and exit variations of maxirmum flow and mid chic were straight-total and static pressure, total terperature forward, but the initial attemrpts for theand flow angle. near surge point failed. The high incidences

for the Sl sections nearest the hub causedTwo operating conditions were selected by very high levels of profile loss to beWG18 - one near peak efficiency and the other predicted for this region:, these resulted innear stall. The flow at both conditions was even more arduous flow conditions beingpredicted using the S!-$2 system and same of calculated for the next SI-$2 loop by the S2the results are shown in Figs 3 and 4. Con- part of the system and eventually the SIBYL2sidering firstly the overall performance, the calcu:itions failed to converge. The presentmass flow rate at the pressure ratio for the SI-S2 technique is probably over-sensitive topeak efficiency condition was predited to be local flow problems like this, since there is34.0 kg/s, about 1% below the test value, no all,,i noe for spanwise moverent of lowPredicted adiabatic efficiency was 92.3%, energy ooundary layer flows - inclusion ofccrapared with a test value of 93%. At the sparwise mixing in the S2 calculation shouldmass flow for the near stall point, the help in this respect. Also, the assuaptionpredicted pressure ratio and efficiency were that stream surfaces remain axisymnetric1.727 and 91.6%, ccrrpared with test values of means that the relief of the eidhwall flow1.728 and 90.1%. There is also good agree- conditions due to three-dirensional flows i'ment between the predicted and measured is neglected, a fundanental limitation of theradial profiles of total pressure and axisynretric SI-S2 approach. To enable atenperature at rotor exit, as shown in Fig 3. solution to be obtained for HP9 at the near(The values plotted are relative to the inlet surge mass flow, Sl calculations were carriedconditions at mid span.) out only between 10% and 90% blade heights

(the other Sl-S2 solutions includedThe predicted Mach nutrer contours at the two calculations for sections within 5% height ofconditions for the blade-to-blade sections at calculationsiforesectionstwithine5%theightio

30%,70%and 0% pan romthe ub re sown the walls) with extrapolation used to obtain30%, 70% and 90% span fron the hub are shown the conditions for stream surfaces closer toin Fig 4, together with the predicted blade the walls.surface ach nuirbers and boundary layerdeveloprent. The contours may be ccrpared Predicted Mach number contours and distriblu-

- l w., -- tions and boundary layer developrent for thefrom the test data given in Fig 17 of Ref 2. rotor and stator mid height sections at theThe agreerent is quite good for the outer three operating conditions are shown in Figsections, with the SI-S2 system correctly 5. These demonstrate the range of operatingpredicting the dual shock system noted in Ref incidences which are to be expected at diffe-1" at the peak efficiency condition, and the rent points along a constant speed character-single normal shock wave just detached from istic for a single stage corrpressor - highthe blading at the near stall condition. positive blade loadings at the leading edge

ni P 1204

I8

near surge, roughly zero 1o" %-Inq at mid chic system over-estimates the rotor efficiency byand high negative loadings at 0,,hain flow. about 4% at both conditions, and this resultsTM' predicted boundary layer tPhaviour is in significantly higher pressure ratios beingtypical of blade sections with circular arc oredicted at a given mass flow. The magni-cambez lines. On the suction surface the tude of this discrepancy is disappointing,shape factors are relatively low over most of gi,. n the level of agreement on overallthe chord, indicating attached boundary pxc:"onrance for the other two cases and forlayers. However, there is rapid growth near ota-r transonic fans. In particular thethe trailing edge where the boundary layer SI-S2 system has not identified any reasonhas become too thick to withstand the to tne peak efficiency of the DFVR rotorconstant diffusion associated with the bein) over 5% lower than for NASA rotor 67,circular arc camber line and separates. On which has a similar pressure ratio, tip speedthe pressure surface the boundary layer ar- aspect ratio. It will be interesting toremains attached throughout, apart from a see whether other CFD methods can explainpossible laminar bubble near the leading edge -Lis difference.at the maximum flow point. No interblademeasurements are available for comparison D.~pite the poor agreement on overall per-with these data. founance, the predicted Mach number contours

for the rotor at 18%, 45%, 68% and 89% bladeThere is sorw uncertainty about the rotor heigt,,s ar in close aareement with thoseexit conditions for this case 2, and so the given in Figs 10 and 1i of Ref 2 (See Fig 8).ccmparisions shown between prediction and At peak efficiency the solution correctlytest (Fig 6) are for the circumferentially predicts an oblique leading edge shock nearaveraged conditions downstream of the stator. the rotor tip with a maximin suction surfaceThe main discrepancies are that:- Mach number of about 1.4. At 68% height the

shock is becorming more normal, and by 45%i the temperature rises near the height it is detached from the blade leadingouter casing are overestimated - this edge. The predicted diffusion of the flowcould be due to the relatively crude within the blade passage also matches theallowance for end wall effects, or to test data well. The predicted suctionneglecting spanwise mixing effects in surface boundary layer development sho'4s athe S2 throughflow calculation. gradual progression with blade height,

roughly similar to that for NASA rotor 67.ii the predicted efficiencies at the Near the hub, separation occurs only near thenear surge condition are much higher trailing edge; by about 45% height the shockthan measured - this is probably wave/coundary layer interaction has beccmeassociated with the problems in strorn enough to cause a snall separationpredicting high incidence flows noted in bubble at about mid chord, and the extent ofsection 3.1 above. Improved agreement the separation at the trailing edge hascould be obtained by "tripping" the reduced; at 68% height the mid chord separa-suction surface boundary layers at the tion has inceased and the trailing edgeblade leading edge. separation has disappeared; and at 89% the

shock wave/boundary layer interaction isApart from this the agreement obtained is strong enough to cause complete separationreasonable, particularly when the uncertainty for the rear part of the blade.due to the manufacturing tolerance of 0.750on blade stagger angle is considered. If the At the point nearer surge, the shock wave hasprediction at the maximum flow point is moved forward and it is just detached fromrepeated with the rotor and stator rows the leading edge at 89% height in bothrestaggered open by this amunt, then at the prrxdicted azc measured results. The shocksame exit flow function there are increases be-c.es fitrther detachec at 68% and 45%of 0.9% on mass flow, 0.012 on pressure ratio heights, ard at 18% theie is only a smallana 1.2% on efficiency. region near the leading edge where the flow

is supersonic. The predicted boundary layer4.3 DFVLR Transonic Copressor Stage behaviour is similar to that at peak effici-

ency except that the shock/boundary layerThe DFIVL transonic fan has an inlet tip interaction near the tip is predicted to bediameter of 398 nrm, an inlet hub/tip ratio of slightly weaker and the boundary layer0.5 and a rotor aspect ratio of 1.7. The reattaches before the trailing edge. How-design point pressure ratio is 1.51 at a mass over, gi-en the high effective incidencesflow rate of 17.3 kg/s and a blade tip speed indicated over the inner half of the blade,of 421 T/s. The rotor has been tested with it is likely that the actual suction surfacetwo different stator rows: the laser amo- boundary layers are rather thicker.metry measurements of the rotor flow fieldwere taken with the NACA 65/60 stator row and 5. CW=nMG REMARKSthis is the configuration which has beenstudied using the Sl-S2 system. The SIBYL2 inviscid-vr,_nu in-'ter'ction

method has now been applied to all the highPredictions were carried out at the sane exit speed compressor cascade test cases selectedmass flow functions as the two test condi- by AGARD W18, and, linked with a streamlinetions selected by W918 and the predicted and curvature throughflow method in an SI-S2measured radial distributions of perfonmance system, to all the high speed ccrpressorare ccmpared in Fig 7. The agreement is much cases. The predictions obtained arepoorer than for the other two cllpressors generally in encouraging agreement with theThe main problem seems to be that the S1-S2 eyperimental data, both in terms of the

TM P 1204

9

features of the internal flows and the 7. Dunham, J., et al, "A new turbcioachineryoverall perfom-ance parameters. throughf low program using the streamline

curvature method", Unpublished RAE report,For the copressor cascades the blade 1990.choking/unique incidence angles and deviationangles are predicted to about ±i !°, the 8. Fuchs, R., et al, "Experimentalstatic pressure rise to ± 4% and (less investigation of a supercritical compressorsatisfactorily) the total loss coefficient to rotor blade section", in "Advanoed Technology± 2% of inlet dynamic lead, for operation for Aero Gas Turbine Ccaponents", AGA CPbelow stalling incidence. The increased 421, May 1987, Paper 39.losses which occur at high positive incidenceare not fully predicted autcnatically, but 9. Steinert, W., et al, "Design and testingsuch situations cai be detected by examining of a controlled diffusion airfoil cascade forthe effective incidence predicted for the industrial axial flow ccapressorblade. It is believed that inprovements are applicaticn", AIZE Paper 90-GT-140, 1990.needed both in the viscous mnxelling and inthe numerical resolution of the leading edge 10. Dawes, W. N., "A ca'parison of zero andregion before -his type of flow can be one equation turbulence modelling forpredicted correctly. turbcnmchinery calculations", ASYE Paper

For the copressor test cases, the predicted 90-M-303, 1990.

results for the NASA rotor 67 and for the HP9 11. Wennerstrom, A. J., "Experimental studystage agree well with the measured results. of a high-through-flow transonic axialIn particular, peak efficiency levels are ccpressor stage", Proceedings 6th ISABE,estimated to within about 1%. The results Paris, po 447-457, 1983.for the DFVMR stage are less satisfactory:the rotor efficiency is over-estimated by 12. Fleeter, S., et al, "Experimentalabout 4%, despite close agreement between the investigation of a supersonic compressormeasured and predicted blade-to-blade Mach cascade", ARL TR 75-0208, June 1975.nunsers. Thus the present method fails toexplain the difference of over 5% in measured 13. Tweedt, D. L., et al, "Eqerimentalrotor efficiency between the NASA 67 and investigation of the performance of aDFVIR rotors. It will be interesting to see supersonic conpressor cascade", ASIE Paperwhether other methods are more successful. 88-GT-306, 1988.

The present applications of the RAZ SIBYL2 14. Founnaux, A., et al, "Test results andcode to the WG18 corpressor test cases have theoretical investigations on the ARL 19been valuable, both in evaluating the code supersonic blade cascade", ASME Paperand in gaining further understanding of the 88-GT-202, 1988.test cases. The author would like tocongratulate the Working Group for taking the 15. Schreiber, H. A., and Tweect, D. L.,lead in identifiying this set of cases, and "Experimental investigation and analysis ofhe would encourage others to carry out and the supersonic ccrpressor cascade ARL-2DPC",publish corresponding predictions. DFVIR IB-325-02-87, 1987.

____R__ 16. Calvert, W. J., et al, "Perfonance of acivil fan rotor designed using a quasi-

1. "Validation of Ccrrputational Fluid three-dimmsional calculation system", inDynamics", AGARD CP 437, May 1988. "Turbomachinery - Efficiency Prediction and

2. "Test Cases for Corputation of Internal Imprwement", Proc I Mach E, 1987-6,

Flows in Aero Engine Components", AGARD 17. Pierzga, M. J., and Wood, J. R.,AR275, July 1990. "Investigation of the three-dirensional flow

field within a transonic fan rotor: experi-3. Calvert, W. I., "An inviscid-viscous ment and analysis", ASZE Journal of Eng forinteraction treatment to predict the blade- Gas Turbines and Power, Vol 107, No 2, Aprilto-blade performance of axial corpressors 1985, pp 436-449.with leading edge normal shock waves", ASIEPaper 82-GT-135, 1982. 18. Wadia, A. R., and Beacher, B. F.,

"Three-dimansional relief in turbmachinery4. Calvert, W. J., "Application of an bladingV', ASME Journal of Turbamachinery, Volinviscid-viscous interaction method to 112, No 3, pp 587-598, 1990.transonic carpressor cascades", in "ViscousEffects in Turbamachines", AGARD CP 351, June1983, paper 2. © British Crown Copyright, 1991/vM

5. Denton, J. D., "An inproved time Published with the permission of the Control-marching method for turbomachinery flow ler of Her Britannic Majesty's Stationerycalculation", ASME Journal ot Eng tor Power, office.Vol 105, pp 514-524, 1983.

6. Calvert, W. J., and Ginder, R. B., "Aquasi-three-dimensional calculation systemfor the flow within transonic compressorblade rows", ASME Paper 85-GT-22, 1985.

TM P 1204

iFig 1

. o o o $ o o o o o o o o o o

co

. . .. .- . . . . . . o o o o o o I

9 LLn

-7030

10

a; 0

0 f

0

0 0 a~p; 0

1 0 .0 00 N 0 0 O o o oo 0 1 a 1 0o o o o o 0

0 N o G.

-Y

S - - . . . . . ... . 0 C

TMP 1204

Figs 2&3

2.0 8.0

1.82

1 .7. - .0 -- 0 .0 6

1.5 .0 -- 0.05

3.0 0.03)( )

14 3 3.0 *0.02

0.0

1.0UCTION SURFACE 00• 2 ' o. L .i~

S.7 2.0 - 0.02

0.6 1 .0 C 0.010. Test S187L2 4.0 0

0.4 ALl 07.9 8.9-At2 8.1 9.1 7 0

0.3 Loss 0.123 0 100 20 1 0

0. P2P1 2 178 2.1,70.0

0.2

0.1 1.0 0 01

0.0 , A CH NUMBER DISTRIUT ON PRESSURE SURFACE 0 0C

Fig 2 S1BYL2 prediction For ARL SL19 cascade (symboLs indicate test data)

14 :21 7-I '22 1 .22

TEMPERATURE 1.20 TEMPFRATURE: .20 0RAT IO B. 8 oRATIO 8

.16 o8o

-OO, 1 0

0j

90~095 2/ "'"- .. 95

ADIABATIC A IAB IC

EPPICIENCT 85 EFPIC I ENO 80

7O [0 o TEST0 7 o TEST

1.70 1 .70 -

PRESSURE 1.70 - 1.70 -

RATIO 1.65- PRESSURE o0R .RATIO 1.65

.50 1 .500- 20 C o 60 80 10 0 20 40 60 80 100

PERCENT ANNU-US HEIGHT PERCENT ANNULUS HEIGHT

a. Padiol pro;'Les near peak ePPlcIency b. Radial proPiles near stalL

Fig 3 S1-S2 predictions For NASA tronsonic rotor 67

TMP 1204

Fig 4a

6 SECTION 30%. SECTION 3V1'

./6.0 o.06

1.3- 5.0 0.05

122 4.0 0 041 3.0 - L 0. 03

20 s -0.01

" 0. SUCTION 'URFACE 0.00

0.7

.

0.6 7.0

.0.60-5 5.0- 0.05

0 - 4.0- 0.04

0.3 - 3.0- 0.03-50 .2 - 2.0 - 0.02

1.0 0.01

. PRESSURE SURFACE

- 6 SECT I0 ON 8.0 SECTION 70/.

1 5 - 7.0

14 6 6.0 -0.06

1.3 -T5.0O -0.00

70O,. / 1. 1. 0.F.01~

3.0- 0.03

1.0 - 0.02

0.9- - .. 0.01

0.6 SUCT S.0 0I

3 0.7 - 7.0 o0.6-60 00

0.4 I . -0 0

0.3- 3 0O -. 03

0.2 2 20- 0.02

0.1 1.0 1 0.01

0.0 PRESSURE SURFACE .

qS SECTION 90% SECTION 90' 7

go% co0 06

/ 0 0 05

AS .2k 10 -0.04.0 - 0.0

2.0 0.02

0. SUCTON SURFAC' 0.0

0.07.0

0.5 5.0S 0.05

0.2 2.0 0 .02

0 '01 O. .... .. 0O

0. PRESSURE SURFACE 0.

MACH NUMBER CONTOURS rIACH NUMBER DISTRIBUTIONS BOUNDARY LAYER PARAMETERS

Fig 4o SIBYl2 predictions For NASA rotor 67 (t peak eFFiciency

TMP 1204

Fig 4b

SECTION 30%' .0 SECI10N 30%.

1. 0..0

1.3 50 -10.052- 1.0-- 0.01

:i3.011 .0.03

30 . . SUCTION SURFACE 10.00

0. 7.0

0O 6.0 -0.06

9 0. 5.0 - 0.05q014j 1.0 0

0. 3.C - 0.03

.q022.0- -0.02

0.I 1 .0- 10.0?0.0 PRESSURE SURFACE 00

.6 SECTION 701' 0 SECTION 70%'

1. .000

3 5. 0.05

4.0 -- 0.017.2 3.0[ 0.03

7_ 2.0- -0.02

-~1.0 t1j~ 0.01

0.9 SUCTION SURFACE C.0

06.7

600.71 7.0

0. .0 -0.00

0.54 5.0 -0

0.3-10-00

12, 0.2 2 .0 -C00.2.0 - 1 2

0.011. 0.00

0.0 PRESSURE SURFACE 00

1 6 SECTION 90V 6.0 SECTION 90' -

9 %1 .1 - 6.0 - .006

1.3 5 .0 - *O05

1.2 - 4.0' 0o 04

Vs. 1.13.0- .0.032.0- 0.02

E1.0-1 .0- 00

0.9 I

0.8 SUCTION SURFACE 0.00

14. 0.7-7.

016 6.0-0.04

0.1 1 3.0- -0.03

0.2 2.0- -- 10.02

0.0 1.0 PRESSURE SURFACE 00

MlACH NUMiBER CONTOURS M1ACH NUMiBER DISTRIBUTIONS BOUNDARY LATER PARAM1ETERS

Fig ib S1BYL2 predictions For NASA rotor 67 near staLL

TMP 1204

Fig 5

0000 0 0 0 0 0 0 0 0 00 00 0 0

000 0 0 000 0.000 00000 0 0000 0

O m ~ m

00 0 00000C! ! 0 0 0 00 00 0 0 0 0 000 00 00!

02 0 o Lu0. ( 0 000 0 0 0 00 0 ;C

0 0 D

(f) CD 0

L I -j

a! I I 1 e 1 1

IC)I 0ce U'--0 0 0 0 0 0 0 0 0 0~~LL -- 0 0 0 0 0 - 0 0 0 0 0

0

<fyfIa

LL~~~ I:uzc

TMP 120

Figs 6&7

.11 - 1.11 ,1,11

1.10 -1.10 1.1001 .9TU TErERATURE1 E09 1 T

TEMPEATU0 1.11u .9 0 ~ R~O I.06RATIO 1.04? RTO 0 1.0

1.0, 0 0 0 1.0 -.0

05 1.05 - 1.5

o0 00 1 00

so0- 0 11[ 0 0 9D 0

AZIA91171C 95 ADIBTC 6 ADIABATIC 85 c0-EFFICIENCY 80 0 EFFICIENC - 0

75-75 0 750

7.0 ;0 - 70 PR-SR 1F20 TEST I .

I- 070 "IT1i~ .j 000 1 7607

1.30- 3j 0 1.30-

1.25 ~- 1.25 1.25 0

917U1 0 0 0 PES E0 M1UE0 ,RA I O -T5 RA TIO A115 1 .15

1 TO .0 1.10

1 .05 1.05 1.05

0 20 10 Go 80 too 0 20 10 60 00 1o 0 20 10 60 l 10C

PRCENT AIUSILS I1596T PERCENT ANNULUS HEIGHT PERCENT

AN".LUS 1f0T

a. Radial proFiLes at max FLow b. RadiaL proFiles at mid chic c. RadiaL prOPI es neo' S0.'-e

Fig E S1-S2 predictions For RoLLs-Royce HP9 stage

1 .26 0.

1.22 o

TEMPERATURE' 20 CL T00E1wPA0TUREI.20 - aRATiO I RATIO

1.14 1100 I00

55 95/-,

/o 0 90~CAD IABATIC 05. AD)IABATIC e5 "

07070 9-7S O0

E 75 1.7C - 0EF0N 7 0 /10 6 r --

75 TEST - 0S115E 71\ 0 TEST - STA6E C

- -T EST 00756 1.T0 TT - 0711 0.JS o70 0 l$ O 000

170 0.T T1.5 -.

7 50

00 0000

.55 /000 a00 01.55 -

1.55 1.50

0 0 41 60 so 0 IIIi;6_~ I US HEIGHT PE'-PCENT'ANNLi, ,W1IT

a. RadiaL proFiles near peak eFFiclency D. Radial proFiLes near stall

Fig 7 S-S2 predictions For DFVLR tronsonic stoge

TIMP 1204

Fig 8a

-SECTION 18% 6.0 SECTION 184

2.5' 5.0

1.4 4.0 H 0.04

1. 2.0 1 0.02

1 S0 A .. 01

.6 o.8 SUCTION SURFACE

0..

5.00.5 -4.0 0.040.,q - ~3.0o 000.3 2.0 -0.02

A 0.41 3 .0- 0.010.3O

d 0.00PRESSURE SURFACE

SSECTION 45/ 6 SECTION 45%

1.5 -5.0 -1.4 4.0- 0.04

-5 1.3 3.0 0.037 7 1 .2 3.075 1.1 2.0 - 0 2

1.0 1.0- 0.0. 0.9 - F0, : , 000

0.8 SUCTION SURFACE

0 .7 - .-

0.6 - 5.00.5 1.0 0.0.0.1 3.0 - 0.030.3 2 0 -. 020.2 00

0.l____________ - 0 000.0 PRESSURE SURFACE

1 6 SECTION 68' - 6.0 SECTION 68

68'.1.5 -5.0k1 .4 - .0- -0.041.3 3

1.2 3.0 0.03

11 2.C 0.021.0 .0 - -- 0.0

0.8 . 0.000.8 SUCTION SURFACE

0.6 0

o.o __ _o i J0.5 - .0 00,,0.1 3o 0 10 .0' 0.3 2 .e 0 , ]1.1202 1 0F "OO

0.0 F PRESSURE SURFACE

1.6 SECTION 89% 6.0 SECTION 89489%. 1.5 -5.0

31Or .0.03

1.2-

.2 -. 0 [ -0.0 2

1 2.0

0.021.1 .0

0.9- 0.000.8- SUCTION SURFACE

0.7-05.0

.0.5 10 0.04

.1 3.0 - 0.030.3 2.0 -0.02

00.2 .01

.0,0 PRESSURE SURFACE

MACH NUMBER CONTOURS MACH NUMBER DISTRIBUTIONS BOUNDARY LAYER PARAMETERS

Fig 8o SIBYL2 predictions Por DFVLR rotor it peak eFFiciency

TMP 1204

Fig 8b

.6 SECTION 10%s~ SECT ION 161 -

5.

•. 2r 1[-0.03

2.1 2 - 0.021.0.0 0.011 0.9

,g 0.8 SUCTION SURFACE

0.7 6.00.6 5.06 0.5 1.0 10.037054.0 0.04

3.0 -1-0.030.3

0.2 2.0 -0.02

0.1 .0 0.010.1 . ESR SURFACE 10.00

1 .i SECTION 45. 0 SECTION 45%1.5 51.4

6 13.06 1.2 3 0.03

1. 2 0.021.0 1.0 0.010.90.8 SUCTION SURFACE 0.00

0.7 -6.00.6- 5.0-

0.5 1.0 1 0.04

0.6 3.0 03J0. 2 2-- 0 .0 2

0.1 1,0 0.010.0 L 0O

PRESSURE SURFACE 0.00

1.6 SECTI1II 68%, 6.0 SECTION G84

68% 1.5 5.0-1.0

1.3 i.0 0.041.2 3.0 0.031.1 2.0 0.02

t[ .0 0.01~0.50.8 SUCTION SURFACE 0.00

b.0

0 .6 b .0o;I0.5 10 000.0 0404

0. 3.0 0.03

0.30.21 2.0 000.1 1.0 -0.010.0PRESSURE SURACE 0

89; 1.6 SECTION 89% G.c - SECTION 89%

1.5 5.01.41.3.0 001 2 3 0 -0 . 0 2

r1 .1 2.0 -0."

1.0 1.0 I -0.01

0.8 SUCTION SURFACE 0.00

3 0 .7 .0

j3 0.6 5.0-0.53.0

0.030.4

0.2[ .0 u.02/ A s 1.0-. i

PRESSURE SURFACE 0MACH NUMBER CONTOURS MACH NUMBER DISTRIBUTIONS BOUNDARY LAYER PARAMETERS

Fig 8b SIBYL2 predictions For DFVLR rotor neor surge

TMP 1204

REPORT DOCUMENTATION PAGE

Overall security classification of this page

TJNLIMITED

As far as possibie this p-ge should contain only unclassified information. If it is necessary to enter classified information, the boxabove must be mvked to indicate he classificatiun, e.g. Restricted, Confidential or Secret.

I DRIC Reference 2. Originator's Reference 3. Agency 4. Report Security Classification/Marking(to be added by DRIC) RAE TM P 1204 ReferenceRA TP 24UNLIMIlTED

5 DRIC Code for Originator 6. Originator (Corporate Author) Name and Location

7674300E Royal Aerospace Establishment, Pyestock, Hants, UK

5a Sponsoring Agency's Code 6a. Sponsoring Agency (Contract Authority) Name and Location

N/A N/A

7 Title Application of SIBYL2 to the AGARD WGI8 compressor test cases

7a (For Translatuons) Title in Foreign Language

7b. (For Conference Papers) Title, Place and Date of Conference77th Symposium of the AGARD Propulsion and Energetics Panel on "CFD Techniquesfor Propulsion Applications", to be held in San Antonio, Texas, USA, 27-31 May 1991

8 Author 1. Surname, Initials 9a. Author 2 9b. Authors 3,4 .... 10. Date Pages Refs

Calvert, W.J. February 17__ __ __ __ __ __ __ __ __ __ __ __ __ __ 199,1 1 18

i Contract Ntmber 12. Period 13. Project 14. Other Reference Nos.

15. D.stribution statement

( ) Controlled by -

(b) Special limitations (if any) -

If it is intended that a ccpy of this document shall be released overseas refer to RAE Leaflet No.3 to Supplement 6 ofMOD Manual 4.

16 Descripat, (Key, "mids) (Descriptors marked * are selected from TEST)Axial flow compressors*. Compressor blades*. Turbomachinery*. SIBYL2.Cascades. i_ cases.

17. AbstractSIBYL2 is an inviscid-viscou" blade-to-blade method for calculating the

detailed aerodynamics and overall -rformance of compressoi blades. It may beapPTlied either on its own to predict the flow for individual blade sections, such asthe mid span of a linear cascade, or in conjunction with a throughflow calculationto predict the performance of a complete axial compressor.

A previous AGARD paper by the author described applications of SIBYL2 to mostof the compressor cascade tust cases which have subsequently been selected by AGARDWIorking Group 18! enerall, good gwas obtained. -lis uurreuL paper pre-sents new predictions for the V2 and ARL SL19 cascades an, for the high speed com-pressor cse. It is hoped that this will be one of mar: sets of calculationsfe- these cases, so that an improved understanding of eacl case may be obtained,

. together with an appreciation of the strengths and weaknerses of dizferent compu-tational approaches.

RAE I-orii A 14.3 (p vised October 1980)