Simulating the Multi-Stage Environment for Single-Stage

THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS345 E. 47th St., New York, N.Y. 10017

The Society shall not be responsible for statements or opinions advanced in papers or discussion at meetings of the gS-GT-187Society or of its Divisions or Sections, or printed in its publications. Discussion is printed only if the paper is published® in an ASME Journal. Authorization to photocopy material for internal or personal use under circumstance not

falling within the fair use provisions of the Copyright Act is granted by ASME to libraries and other users registered withthe Copyright Clearance Center (CCC) Transactional Reporting Service provided that the base fee of $0.30 per pageis paid directly to the CCC, 27 Congress Street, Salem MA 01970. Requests for special permission or bulk reproductionshould be addressed to the ASME Technical Publishing Department.

Copyright'© 1995 by ASME All Rights Reserved Printed in U.S.A.

SIMULATING THE MULTI-STAGE ENVIRONMENT FORSINGLE-STAGE COMPRESSOR EXPERIMENTS

J.M.M. Place and M. A. HowardRolls-Royce plc

Derby, United Kingdom

N. A. CumpstyWhittle Laboratory

University of CambridgeCambridge, United Kingdom

1. ABSTRACT.

The performance of a single-stage low-speed compressor hasbeen measured both before and after the introduction of certainfeatures of the multi-stage flow environment. The aim is to make thesingle-stage rig more appropriate for developing design rules formulti-stage compressors. End-wall blockage was generated by teethon the hub and casing upstream of the rotor. A grid fitted upstreamproduced free-stream turbulence at rotor inlet typical of multi-stagemachines and raised stage efficiency by 1.8% at the design point.The potential field that would be generated by blade rowsdownstream of an embedded stage was replicated by introducing apressure loss screen at stage exit. This reduced the stator hub cornerseparation and increased the rotor pressure rise at flow rates belowdesign, changing the shape of the pressure-rise characteristicmarkedly. These results highlight the importance of features of theflow environment that are often omitted from single-stageexperiments and offer improved understanding of stageaerodynamics.

2. NOMENCLATURE.

c chordP static pressurePo stagnation pressureP,, relative stagnation pressureTG compressor rig configuration with the turbulence grid fitted

upstream of the stage.s rotor blade pitchSSHW single slant hot wireSTG compressor rig configuration with the turbulence grid fitted

upstream of the stage and the pressure loss screen fitteddownstream of the stator.

U wheel speedW time-mean relative velocityx axial coordinate

rotor relative inlet angle(3 rotor relative exit angleAH change in stagnation enthalpy across the rotor at mid-height

WTS total-to-static pressure rise coefficient=P3-P1

2p Um^r.^. total-to-total pressure rise coefficient=

P03—P01

zp Umflow coefficient, = volume flow rate / annulus area

6 rotor wake momentum thicknessp air density

Subscriptsm mid-heightx axial1 rotor inlet measurement station2 rotor exit measurement station3 stator exit measurement station4 measurement station 2.6 stator chords downstream of the

stator trailing edge

Superscripts

pitchwise averaged

spanwise and pitchwise averagedpeak value

3. INTRODUCTION.

Low speed experimental compressors are useful tools for theaerodynamicist. Their mechanical simplicity and low speed meansthe rig is less expensive to build, maintain and modify and can readilybe of large scale. The major aero-engine manufacturers regularly use

Presented at the International Gas Turbine and Aeroengine Congress & ExpositionHouston, Texas - June 5-8,1995

This paper has been accepted for publication in the Transactions of the ASMEDiscussion of it will be accepted at ASME Headquarters until September 30,1995Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/11/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

four stage low speed compressors to assess blade designs prior tohigh speed testing but work is also done on single-stage researchcompressors. Often the aim of the research carried out on such rigsis to improve blade designs for multi-stage machines and, if theresults from them could be relied upon to achieve this, the validationof design philosophies could become still cheaper and quicker. Aconventional single-stage compressor clearly lacks certain features ofthe multi-stage flow environment that may be significant in theireffect on blade aerodynamics, and it might be the case that a bladedesign optimized through a series of single-stage tests may be farfrom the best for a multi-stage machine.

To make the flow field in a single-stage machine morerepresentative, some features of the multi-stage environment can bequite easily introduced to the rig. i) First, it is common practice tofit teeth to the casing and hub end-walls upstream of the stage tothicken the boundary layer and ensure that at stage inlet the axialvelocity profile is typical of that seen by an embedded stage. ii)Second, the levels of free-stream turbulence measured in multi-stagemachines generally are much greater than those found in asingle-stage rig and may have a different length scale; it is supposedthat the behaviour of blade profile boundary layers and thedevelopment of corner separations within blade rows near theend-walls are affected by the free-stream turbulence. Efforts tocorrect this have been made before: Cyrus (1988) introduced aturbulence-generating grid upstream of a test stage (although in theprocess this did alter the stage inlet axial velocity profile quiteradically). iii) The last feature considered in the work described hereis the coupling of the potential field from downstream stages whichaffects the flow in all but the rearmost stage. There is evidence thatthe rearmost stator in a compressor is less able to sustain staticpressure rise than stators further upstream, because a downstreamstage tends to suppress regions of separation.

The performance effects of wakes and streamwise vorticity fromupstream blade rows as they convect through the stage have notbeen accounted for in this experimental programme.

The intention of the work described here is twofold: to modify asingle-stage compressor rig so that it simulates the multi-stageenvironment adequately and also to use the opportunity that asingle-stage machine offers to assess the influence on stageperformance of increased free-stream turbulence and then thepressure field imposed by downstream blade rows. As well asdemonstrating improved simulation of the multi-stage environment,this short series of experiments found significant changes in stageperformance and highlighted features of compressor aerodynamicsthat are inadequately understood.

The paper is structured to provide the reader first with all theinformation (the rig details, measurement techniques and results)necessary for discussion. Following discussion of the observations,the paper then closes with conclusions.

4. EXPERIMENTAL RIG AND MULTI-STAGESIMULATION.

The experiments to demonstrate improved multi-stagesimulation were performed on the Deverson single-stage axialcompressor at the Whittle Laboratory that has been previouslydescribed by McDougall (1988). The addition of an auxiliary fan atrig exit well downstream (6.8 rotor mid-height radii) of the statortrailing edge allows the mass flow rate to be varied independently ofthe research-stage pressure rise. The stage has zero inlet swirl,giving high reaction at the design flow coefficient of 0.55. The

blades are of C4 section with circular arc camberlines and details forthe mid-height aerofoil sections are given in Table 1. Other stageparameters are given in Table 2.

Rotor Stator

Number of blades 51 36

Chord (mm) 111 114

Camber (degrees) 26.5 42.9

Stagger (degrees) 47.9 14.3

Solidity 1.31 0.95

Max thickness/chord 0.10 0.10

ABLE I COMPRESSOR BLADING GEOMETRY AT MID-HEIGH'

Hub / Tip ratio 0.80

Mean radius (mm) 685.8

Rotor rpm 500

Design flow coefficient, 4) 0.55

Design AH/Um2 0.45

Axial spacing / rotor chord 0.45

TABLE 2 C4 STAGE DETAILS.

At the design flow coefficient of 0.55 without multi-stagesimulation features (the Baseline build) the stage operates at anefficiency of 87.8% with a rotor mid-height diffusion factor of 0.44.At this operating point, the Reynolds number based on chord for therotor at mid-height is 324,000 (with a relative inlet Mach number of0.12) and for the stator 215,000. The rotor tip clearance gap was setto 2% of rotor tip chord and the stator hub was sealed.Measurements were taken with the rig in the three configurationsshown in Fig. I and described below.

4.1 Baseline BuildThe Baseline configuration is that described by McDougall and

consists of the stage with teeth positioned far upstream of the rotorleading edge to generate an appropriate axial velocity profile similarto the thick end wall boundary layers observed in multi-stagecompressors. The work done on spanwise mixing by Li (1991) andGoto (1992) used the Deverson Rig in this configuration.

4.2 Turbulence Grid (TG) Build.The TG build was created by fitting a turbulence grid 5.4 rotor

chords upstream of the rotor leading edge (Fig. 1), increasing thefree-stream turbulence intensity and reducing the length scale atrotor inlet. The grid was positioned as far upstream as possible toallow the wakes from the bars to mix out while producing at rotorinlet turbulence characteristics typical of the multi-stage environmentas found by Camp and Shin (1994). A series of experiments hadbeen conducted earlier in a wind tunnel to select a suitable geometryfor the grid, which consisted of 102 radial bars each of 4 mm squaresection steel welded to three 6.4 mm square section circumferentialbars placed at 18%, 53% and 85% of the annulus height. Table 3compares the mid-height rotor inlet turbulence of the Baseline andTG builds with that at mid-height downstream of stator 3 of theCranfield four stage compressor (see Camp and Shin).

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 04/11/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Boundary layer thickenersfar upstream Rotor Stator

II

Flow- I II I

4

Extent of rotating hubBaseline build

Additional end-wall boundary layerthickeners at hub and casing

Flow I I x I I4 V ^3 4

Turbulence grid TG build 100 mm

Flow I I

1 34 4

STG build Downstreampressureloss screen

Dashed lines indicate measurement planes

FIG. 1 EXPERIMENTAL ARRANGEMENTS FOR THE DEVERSON C4SINGLE-STAGE COMPRESSOR

Baseline TG Cranfield 4stage

turbulence 0.3% 3.7% 4.7%intensity

integral length 0.17c 0.06c 0.05cscale

c is the rotor chord at mid-heightTABLE 3 A COMPARISON OF TURBULENCE CHARACTERISTICS.

It was found that the turbulence grid thinned the end-wallboundary layers and so, to ensure that the axial velocity profile atstage inlet was similar for all builds, additional teeth were fitted atthe same axial station as the grid. The incoming axial velocityprofiles measured at stage inlet for the Baseline and TG builds arecompared in Fig. 2 (the profile for the STG build described in section4.3 is the same as that of the TG build). The profile for the Baselineis calculated from the total pressure measured at a singlecircumferential position while that of the TG build was calculatedfrom an area traverse over one pitch of the turbulence grid. For theTG build (and the STG) the wakes from the radial bars were veryshallow at rotor inlet, the velocity deficit being approximately onepercent of the free-stream velocity. The wakes from thecircumferential bars were deeper, and the mid-height bar wake isvisible in the figure, but near the end-walls the velocity profiles forthe three builds are very similar. The maximum difference in rotorincidence between the Baseline and TG build across the span is 0.5degrees at a flow coefficient, 4), of 0.57. Since the changes inincidence are small and limited to a small proportion of the span

(around mid-height where the 0.5 degree maximum occurs) it isreasonable to suggest that differences in performance between theBaseline and TG builds are due principally to changes in theturbulence rather than discrepancies in the inlet axial velocity profile.

1.00 casing

0.80

p Baseline

fl- 0.60 ❑ TG (& STG)

000

0.40

-

0.20

hub0.00

0.00 5.00 10.00 15.00

Turbulence intensitycasing1.00

0.80

0 Baseline

❑ TG (& STG)0.60

o0U IL 0.40

0.20

0.00 1 , hub

0.40 0.60 0.60 1.00 1.20

Axial velocity as a fraction of the peak value, VAxU1x

FIG. 2 STAGE INLET (STATION 1) CONDITIONS FOR THE THREEBUILDS.

4.3 Screen and Turbulence Grid (STG) BuildThe STG build consisted of the TG build with a downstream

pressure loss screen positioned 0.5 stator chords downstream of thestator trailing edge (Fig. 1). The intention was to simulate theaxisymmetric potential effect of a downstream repeating stage on theresearch stage. The static pressure field of a healthy downstreamstage adjusts to incoming distortion in such a way as to tend tosmooth out the non-uniformity. This happens because the localpressure rise across the downstream stage is related to the local inletvelocity. Therefore, a region of low velocity flow in, say, the statorwake at the hub (such as might be caused by a comer separation inthe upstream stage stator) undergoes a greater rise in static pressureacross the downstream stage compared to the fluid at mid-height. Ifthe hub fluid is to reach the static pressure enforced at downstream

3


C

C)C)

0 0.60

C)0)

6)U

C)

0.50Cl,

C)

stage exit (where axial flow sets the static pressure to be constant upthe span) by means of this greater pressure rise then the hub staticpressure must reduce at downstream stage inlet: this is the "sucking"effect caused by a downstream stage and means that, other factorsbeing equal, an embedded stator is less prone to separation than thefinal stator in a compressor.

By considering the upstream static pressure field produced bythe reaction of the stage to variations in the inlet velocity profile, itcan be shown that behaviour equivalent to a stage can be obtainedfrom a pressure-loss screen provided the gradient of the screenpressure loss characteristic is equal to the gradient of the pressurerise characteristic of the stage (see, for instance, Lambie, 1989).(This is true if the gradients of the static-to-static or total-to-staticpressure characteristics are equal). To simulate the correctmagnitude of this phenomenon using a pressure-loss screen requiressome means of selecting the strength of the screen and finding asuitable axial position for the screen relative to the research stage.Wind tunnel experiments established a design that gave the requiredloss: a pressure loss screen was constructed from wire mesh with ablockage of 0.57 followed by 50 mm of paper honeycomb to ensureaxial flow downstream of the screen. Fig. 3 compares thetotal-to-static pressure rise characteristic curve of the stage in STGconfiguration with the total-to-static pressure drop curve of thescreen to show that over the majority of the flow range the gradientis reasonably well-matched.

STG C4 stage pressure rise

-\1

TICt

00.40

Screen pressure drop + 1.165

0.30

FIG. 3 SIMULATING THE PRESSURE RISE OF THE C4 STAGEWITH THE DOWNSTREAM PRESSURE LOSS SCREEN.

In some respects, the pressure-loss screen is similar to anactuator disc and there have been studies to determine theappropriate axial position for an actuator disc modelling a singleblade row (such as Horlock, 1978); these concluded that it should bebetween the leading and trailing edge, further aft at low loading andtowards the leading edge nearer stall. Bearing in mind that anobjective of the experiments was to determine the trends in changesto stage performance as each element of the multi-stage simulationwas added, the front of the screen was fitted in the Deverson rig 0.5stator chords downstream of the stator trailing edge. This may be a

little close for a strictly faithful simulation of a downstream stage atflow rates above design, but was intended to ensure that any changesto the stage aerodynamics would be measurable.

5. MEASUREMENT AND ANALYSIS TECHNIQUES

5.1 Pitot Probe.At stage inlet, pressure measurements were taken 0.6 rotor

chords upstream of the rotor leading edge using a flattened pitotprobe; velocities were calculated by measuring the casing staticpressure and assuming there was no radial variation in staticpressure. This assumption is justified because there is no swirl atinlet and the annulus walls are parallel with constant radii.

5.2 Ordinary Hot Wire.A straight hot wire normal to the mean flow direction was used

upstream of the rotor to measure the stage inlet flow turbulencecharacteristics. The DC signal was logged at 20 kHz with a low passfilter of 10 kHz to avoid aliassing of data. Definitions of turbulenceintensity and integral length scale are those used by Camp and Shin(1994).

5.3 Single Slant Hot Wire.To measure the rotor exit flow field, a single slant hot wire

(SSHW) was traversed radially at the rotor exit measurement station(0.19 chords downstream of the rotor trailing edge at mid-height).For the TG and STG builds the probe was positionedcircumferentially 40% of stator pitch from the suction surface side ofthe stator leading edge. For the Baseline build the circumferentialposition was slightly different, at 28% of stator pitch from thesuction surface side of the stator leading edge. The particularSSHW technique used was described by Goto (1992), adevelopment of the earlier work of Hunter (1979); more details aregiven by Place (1995).

Phase-locked data from the wire was logged as the rotor passesupstream of the probe at a rate such that 110 datapoints were takenin the time that 2.8 rotor passages pass; this was repeated 150 timesto give an ensemble average. Using this technique it is estimatedthat the absolute flow angle can be measured to within one degreeand the pitch angle to within two degrees, although repeatability isbetter than this. Velocity measurements depend on the accuracy ofthe King's law wire calibration as well as the probe calibration, butvelocity magnitudes are estimated to be correct to within 2%.Circumferentially-averaged axial velocities are calculated usingarea-weighting over the central passage measured (i.e. 40circumferential points). The circumferentially-averaged absoluteflow angle is defined as the arc tangent of the ratio of themass-meaned axial and mass-meaned tangential velocity componentsevaluated from the ensemble average at each of the 40 data points(which covers a single rotor passage) at each spanwise position. Inthe absence of a measured distribution of static pressure, thisprocedure gives an approximation to the mixed-out flow angle. Therelative flow angle is found by solving the appropriate velocitytriangle based on the mean axial velocity and absolute flow angle andthe uncertainty in this measurement is +/- 1 degree.

The SSHW technique generally captures the mid-height rotorwake with between fifteen and twenty data points. It is thereforepossible to calculate the rotor wake momentum thicknesses withgood repeatability. Because there is a pitchwise variation in velocityat rotor exit outside the wakes it is not straightforward to choose anedge velocity for the momentum thickness calculation. A consistent

0.40 0.45 0.50 0.55 0.60 0.65

Flow coefficient, 0

Note that the screen pressure drop characteristic has been shifted by 1.165so that it plots conveniently on the same graph as the stage pressure risecharacteristic.

0H

4


and plausible wake momentum thickness calculation method thatcan be automated is required and the following has been found to beeffective. The wake is split into suction side and pressure sidecomponents and then an appropriate edge velocity must be chosenfor each component. The peak velocity is used for the suction side,since this is found to be clear at the edge of this part of the wake.On the pressure side, the edge of the wake is defined arbitrarily asbeing 10% of pitch outboard of the point at which the secondderivative of velocity in the circumferential direction is at aminimum. The wake momentum thickness, 0, is then the sum of thetwo thicknesses calculated for each side of the wake, multiplied bythe cosine of the rotor relative exit flow angle, (3 2, to transform theparameter into the relative streamwise direction. Thetwo-dimensional profile loss coefficient is then given by

4POR 20 c0s2 131

2 PW1 —cobR2cos202

where 0^ is the rotor relative inlet angle and s is the blade pitch.Uncertainty analysis shows that the rotor profile loss coefficientcalculated in this way should be accurate to within 0.3% of 1/2pW, 2 .

5.4 Three Hole Probe.Area traverses were performed over a single stator passage

using a calibrated three hole wedge probe at stator exit, 15% ofstator chord downstream of the stator trailing edge. In combinationwith a pressure transducer the stagnation pressure measured by theprobe in steady flow conditions at a pressure difference similar tothat between the flow at stator exit and atmosphere is estimated tobe correct to within 5 Pascals.

There is a continuing problem of how to measure the stagnationpressure at rotor exit in the stationary frame where the flow isnon-uniform and unsteady. One body of experience suggests that apneumatic multi-hole probe, either aligned with the flow bybalancing the pressures from the side ports or using a fixed probesetting angle and a calibration, allows good comparisons betweenbuilds (e.g., Cyrus, 1986, Howard et al, 1993, and Lewis, 1993). Analternative is to measure the stagnation pressure at stator exitbetween the stator wakes, where the flow is much steadier, and takethis as representative of the rotor exit stagnation pressure. Thistechnique has been regularly used in the Central Institute of AviationMotors, Russia - Savin (1993). Since an objective of this study is tocompare the performance of different rig configurations, rather thandetermine the absolute level of rotor performance, it is reasonable touse the stagnation pressures measured at stator exit as a reliablemeans of assessing changes in rotor pressure rise, especially as thereare no blade rows upstream of the rotor.

5.5 Stage Pressure Rise and Efficiency Characteristics.In order to compute efficiency both stagnation pressure rise and

power input are required. Detailed traverse data were used to obtainstagnation pressure at both inlet and exit because the use of simplermethods were not valid for the build with the downstream screen.For example, stagnation pressure derived from exit static pressurecould not be used to compare all three builds since the exit staticpressure tappings were downstream of the screen for the STG build.The radial and circumferential traverses were carried out at four flowcoefficients at inlet and at three flow coefficients at exit (0=0.43,0.51 and 0.57). The flow at compressor exit (0.15 chordsdownstream of the stator trailing edge and 0.35 chords upstream ofthe screen) is highly non-uniform and is therefore expected to be

sensitive to the form of averaging used. A mixed-out average in thecircumferential direction combined with a mass-average in the radialdirection was chosen after comparison of various methods (seeAppendix 1). The values of average inlet and exit pressure at flowcoefficients other than those traversed could then be obtained fromappropriately defined interpolations. An exit static pressurerepresentative of the whole annulus was obtained by subtracting theaverage exit dynamic head from the stagnation pressure.

The power input was derived from the rotational speed andreadings from a phase-shift torque transducer. The measured torquewas corrected for bearing and windage effects which were obtainedover a range of mass flow rates and drum speeds by measuring thetorque with the blades removed.

Uncertainty analysis suggests that the random error appropriatefor comparison of efficiency between the builds is approximately +/-0.7% which is consistent with the observed scatter. There arehowever significant bias errors additional to this. For instance, dueto small non-axisymmetries at inlet in combination with othersources of uncertainty the inlet mass flow is estimated to be within+/- 0.8% of the correct value. This implies an increase to theuncertainty when comparing the efficiency of this compressor withothers.

6. EXPERIMENTAL RESULTS.

6.1 Stage Performance Characteristics.Fig. 4 shows the stage total-to-static pressure rise and adiabatic

efficiency characteristics for the three builds. The pressure risecharacteristics for the Baseline and TG builds appear similar, butbetween x=0.60 and 4=0.47 the TG build has a higher efficiency,

95.00

pL

90.00U

85.00U

W 80.00

75.00

o.so yin

0 Raseline

2 0.60Cj)

0

0.40

0.40 0.45 0.50 0.55 0.60 0.65Flow coefficient, 0

FIG. 4 C4 STAGE CHARACTERISTICS FOR THE THREE BUILDS.

CU

00UNN


STG

0.8 o00

^=0.51

Contours of W/U r, at intervals of 0.1

FIG. 5 C4 ROTOR EXIT FLOW FIELD FOR THE THREE BUILDS AT0=0.51

c .0.60

0C0UN

- 0.40

0.20

0.00

with a peak value of 90.4% at approximately 4)=0.57. At thisoperating point measurements suggest that the TG build is 1.8%more efficient than the Baseline. This change is so large that asignificant proportion of it is felt to be genuine. Introducing thedownstream screen causes a major change to the shape of thecharacteristic curves and below 4=0.55 the pressure rise is increasedsignificantly relative to the TG build (at 4)=0.51 the total-to-totalpressure rise increases by 3.9%). Peak efficiency was increased to91.4% and has moved from a flow coefficient of 0.57 to 0.53 and soat 0=0.51 the stage efficiency has risen by 3.3% relative to the TGbuild.

The stall behaviour of the stage for the three builds was notinvestigated with detailed time-dependent measurements, but it isapparent that only small changes to the stall point resulted from theintroduction of the turbulence grid and downstream screen. Inpassing it should be noted that changes to the inlet velocity profilewith flow coefficient that are seen by an embedded stage are notreproduced in this single-stage machine: these could be due tochanges in deviation of an upstream stator row as pressure riseincreases and may affect stalling behaviour.

Baseline

^=0.51

PS I SS

0.5 I I' i^;i

0 0 I(, 0. 800C

0 g0

TG

__ O.

p=0.51 03ooo00 Lo. o iii

4-Direction of rotor rotation

Flow field measurements were taken at three flow coefficients,0.57, 0.51 and 0.43, but for brevity comparisons between builds willgenerally be made at q=0.51, the operating point at which differencesbetween the TG and STG builds are greatest.

6.2 Rotor Exit Flow FieldContours of relative velocity for the three builds measured by

the SSHW are depicted in Fig. 5 at =0.51 to give an impression ofthe rotor exit flow field. The wake is thicker near the hub at all flowrates: ammonia and ozalid flow visualization by McDougall (1988)showed that this was the result of separation in the corner region ofthe hub and the suction surface. Stream-sheet twisting is evidentwith the lowest energy fluid being swept over towards the suctionsurface at the hub. The most obvious differences are between the

1.00 -r casing

0.80

0=0.51o Baseline❑ TG:L STG

HError bar

hub

20.00 30.00 40.00 50.00 60.00 70.00

Rotor relative exit flow angle, 13 2

1.00 casing

0=0.510.80

o Baseline❑ TG

C L STG

a0.60

0CO

LL 0.40

0.20

0.00 hub

0.20 0.30 0.40 0.50 0.60

Rotor exit axial velocity, Um

FIG. 6 ROTOR EXIT FLOW CONDITIONS FOR THE THREE BUILDSAT 0=0.51.

6


1.00

0.60

CCD0.' 0.60o

0UCDU_ 0.40

0.20

0.00

Baseline and the other two builds both near the casing and in theform of the corner separation over the lower part of the span. Theintroduction of free-stream turbulence thinned the rotor wake ataround 30% height while it remains similar closer to the hub at about10% height.

The circumferentially-averaged distributions of flow angle andaxial velocity at rotor exit are given in Fig. 6 for all three builds at4=0.51 and it is immediately evident that for the bulk of the spanthere is little difference between the builds. The turbulence gridcauses more fluid to be drawn through between 15% and 35% of theannulus height, implying an improvement in blade performance overthat region of the span, with a reduction near the casing. Thecircumferentially-averaged relative exit flow angle distribution at thesame operating point shows small changes that fall within theuncertainty band. Compared to the Baseline, the TG build has lowerdeviation between 20% and 50% height but is similar elsewhere.The STG build however has a deviation that is lower than the TGbuild over most of the span.

6.3 Rotor Wakes.Increased free-stream turbulence is expected to affect the

chordwise position of transition with consequent effects on thedevelopment of the blade surface boundary layers. A comparison ofthe profile losses calculated from wake momentum thicknesses of theBaseline and TG builds is made in Fig. 7 at 4=0.57 and 4=0.51. Forclarity, the data is only shown between 16% and 87% of the annulusheight, since beyond these bounds the wakes become much thicker(for these to be shown a much larger scale would be required) andtheir edges less distinct. Also near the end walls, where the flow ismore three-dimensional, a two-dimensional momentum thicknessbecomes less appropriate as a measure of blade behaviour. Between39% and 78% height, a region that might be expected to be closestto two-dimensional, the TG build has profile losses about 10%greater than those of the Baseline build at 4)=0.57. Between 10%and 30% span, however, in a region of the wake dominated by thehub corner separation, the increased free-stream turbulence has actedto reduce the wake momentum thickness substantially. These trendshold true at increased loading at 4=0.51 and also near stall at 4=0.43(the latter are not shown), but as the flow coefficient reduces so theTG build wakes become more similar to those of the Baseline.

To conveniently summarise the results, a mean value for theprofile loss is calculated by area-averaging the profile losscoefficients between 39% and 78% annulus height; results for thethree builds are given in Table 4. The STG losses around mid-spanare similar to those of the TG build at 4=0.57 but about 5% greaterat 0=0.51.

Baseline TG STG

0=0.57 2.9% 3.3 % 3.3 %

t =0.51 3.4% 3.7% 3.9%

Uncertainty band is +/- 0.3 %TABLE 4 AREA-AVERAGED ROTOR PROFILE LOSSES AS A

PERCENTAGE OF INLET RELATIVE DYNAMIC HEAD.

6.4 Rotor Pressure Rise.Fig. 8 shows the area-average stagnation pressure rise outside of

the stator wakes at stator exit for the three builds at 4=0.51. Thesedata were circumferentially area-averaged over the part of thefree-stream (i.e., clear of the stator wakes) shown in Fig. 9, and so

1.00

0=0.570.80 0 Baseline

❑ TGL STG

0.6000U

LL 0.40 Error bar

0.20

0.00 I hub

0.02 0.03 0.04 0.05 0.06

Profile loss coefficientc asing

i

T=0.51

o Baseline❑ TGL STG

hub

0.02 0.03 0.04 0.05 0.06

Profile loss coefficientFor clarity, data not shown below 16% span nor above 88% span.

FIG. 7 ROTOR '2D' PROFILE LOSSES FOR THE THREE BUILDS,CALCULATED FROM ROTOR WAKE MOMENTUMTHICKNESSES.

are representative of the rotor pressure rise. The pressure coefficientplotted is defined as the difference between the mean stage inletstagnation pressure and that measured at stator exitnon-dimensionalized by a dynamic head based on the mid-heightwheel speed. The estimated uncertainty in total pressuremeasurement using the wedge probe equates to an uncertainty of0.007 in this pressure rise coefficient value (although repeatabilityhas been found to be better than this). For the Baseline and the TGbuilds the pressure rise around mid-span appears generally similar (adifference of 0.006), although near the end-walls the TG build isdelivering more pressure rise, significantly greater than the errorband. Meanwhile, the STG distribution is similar in shape to that ofthe TG build but with the level of pressure rise increased at all radii:at 40% height the difference in coefficient value is 3.5% of the STGpressure rise at this operating point.


1.00

0.80

m 0.60

00

m 0.40-

0.20

0.00

0=0.51

0 Baseline

❑ TG

STG

Error bar

casing

hub

0.70 0.80 0.90 1.00

P03 —P01

2p UmObtained by area-averaging between the wakes.

FIG. 8 FREE-STREAM STAGNATION PRESSURE AT STATOR EXITAT =0.51.

6.5 Stator Exit Flow Field.Traverses using the wedge probe at stator exit shown in Fig. 9

indicate that the dominating flow feature is the stator suction-surfacehub comer separation which grows steadily as the flow coefficient isreduced (flow reversal was confirmed by oil and dye flowvisualization). The pitchwise resolution of the wedge probe areatraverse at stator exit was much coarser than that at rotor exit usingthe SSHW and so it has not been possible to compare stator wakemomentum thicknesses reliably for the Baseline and TG builds.

The comparison of the stagnation pressure contours measured atstator exit shown in Fig. 10 demonstrates how the wake width haschanged little with the introduction of high free-stream turbulence.Increased free-stream turbulence did not significantly reduce theextent or intensity of the stator corner separation at the hub at4=0.51 or closer to stall.

The downstream screen reduces the severity of the large hubcorner separation in the stator at flow coefficients below 0.57. Themarked reduction in the size of the stator hub corner separations atflow coefficients of 0.51 and 0.43 are shown in Fig. 11. Thestagnation pressure coefficient contours in Fig. 1la at level 0.90show that the rotor is delivering more pressure rise in the free-streamregion for a flow coefficient, 0, of 0.51 in the case of the STG build(as noted above in section 6.4).

Traverse over a single stator passage

0.7500

/ v a=0.57 /^ O a

/^J1 a

— 0.8000 0.8,00

0 .7500

SS PS Direction of rotor rotation

^^ a 0.8500.o •soo

0=0.51

Region of area-averaging to give free-stream P,

(I?

a^=0.43

O

A

P03 P01 AContours of 1 2 at intervals of 0.05

T UmFIG. 9 C4 BASELINE STATOR EXIT FLOW FIELD.

4 ss I PsDirection of rotor rotation

0.8000------ tf ' dJf ----

Baseline: dashed contours %f G Contour interval of 0.1TG: continuous contours

7. DISCUSSION.

Comparison of data from the three builds shows effects of theturbulence grid and the downstream screen that were expected:increased rotor profile loss but reduced rotor corner loss onintroducing representative inlet turbulence; a strong effect on thestator separation from the downstream screen. These, together withthe changes to stage performance that were not anticipated, notablythe increased rotor pressure rise with the downstream screen, arediscussed in the subsequent sections.

P03—p01Contours of

j0 UmFIG. 10 THE EFFECT OF INCREASED FREE-STREAM

TURBULENCE ON THE STATOR WAKE AT 0=0.51

8


/Id

TG4

aC

0.9500 I (^I 0.°500

jVl I o 'I C

^I STGI o ^ a

(B) 4=0.43

X0.85005 p q0 ^^ 0.90

^^P03—P01Contours of

I2pum

TG a at intervals of 0.05

Shading of zones below thea 0.5 contour level

Direction of rotor rotation

080 008 ^J 0.9000

^^

'STG'

a°°_a

1 a

(A) 0=0.51

FIG. 11 THE EFFECT OF THE DOWNSTREAM SCREEN ON THE STATOR EXIT FLOW FIELD.

7.1 The Effect of Turbulence on C4 Staue PerformanceAlthough the increased free-stream turbulence of the TG build

increased the stage efficiency by 1.8% at flow coefficients of 0.57and 0.51 (these figures are arrived at by comparing curves fitted tothe characteristics, Fig. 4), the rotor profile boundary layersthickened and there was no significant improvement in the statorwakes relative to the Baseline build. Increased turbulence of theappropriate length scale is expected to cause earlier transition,allowing the turbulent boundary layer to grow over a longerproportion of the chord and so causing it to be thicker by the trailingedge. The experimental results support this with increased rotorprofile loss (calculated from the wake momentum thickness) aroundmid-span over the whole operating range, Fig. 7 and Table 4.Increased turbulence also provides for more efficient transfer ofenergy from the free-stream down to the fluid near the blade surface,making the boundary layer able to withstand a greater adversepressure gradient before severe separation occurs; this effect is notseen near mid-span, even at the lowest flow coefficient tested. Thewake is thinned substantially for the TG rotor in the region affectedby the comer separation near the hub at 4=0.57, Fig. 7. Thisreduction in size of the hub corner separation by increasedfree-stream turbulence diminishes as flow coefficient is reduced, Fig.7, perhaps because the static pressure rise at 0=0.51 has gonebeyond the limit above which increased turbulence is unable tosubstantially reduce the hub separation.

The increase in efficiency at 4>=0.57 attributable to raised Herturbulence is not accompanied by a greater increase in stagetotal-to-total pressure rise (Fig. 4) for two reasons. There is a smallchange in the rotor inlet axial velocity profile between the builds (andhence the rotor exit axial velocity) which leads to a change in rotorwork. In addition there is an increase in area-averaged rotordeviation of 0.3 degrees (it should be noted that this is well withinthe uncertainty band) for the TG build relative to the Baseline, whichseems consistent with higher profile loss.

The different flow pattern observed in the velocity contour plotsat rotor exit, Fig. 5, results in a radial re-distribution of total pressurewhen the turbulence grid is introduced (see Fig. 8). In the lower half

of the annulus the change in the stagnation pressure levels at statorexit between the wakes can be attributed to a combination of rotordeviation change, rotor loss coefficient change and small differencesin the rotor inlet stagnation pressure profile. Between 30% and 40%height the deviation is about 1 degree lower for the TG build, Fig. 6,consequently work input is increased here and therefore pressurerise. Nearer to the inner wall the TG rotor losses are reducedrelative to the Baseline, Fig. 7, and so, even with similar deviation,the pressure rise is increased.

In the region of the casing (85% to 100% height) the axialvelocity and relative swirl angle shown in Fig. 6 suggest similar workinput for Baseline and TG builds. It is therefore surprising to seesignificant differences in the pressure rise attributed to the rotor asseen in the free-stream at stator exit, Fig. 8. The exit stagnationpressure profile changes are also much larger than the small increaseassociated with the difference in inlet axial velocity profiles that canbe seen in Fig. 2. The only hypothesis that can be offered is that thefree-stream turbulence is mixing out rotor end wall loss morequickly, before it has undergone the full deceleration of the stator. Ithas been shown that for simple mixing a streamwise pressuregradient has a large influence on mixing loss (Denton and Cumpsty,1987) with the loss increasing as mixing takes place afterdeceleration of the flow. If free-stream turbulence increases the rateof mixing, lower losses by stator exit in the free-stream would beexpected.

7.2The Effect of the Downstream Screen on Stage Performance.At a flow coefficient of 0.57 the STG build performance is very

similar to that of the TG and it is only at flow coefficients of 0.55 orless that the downstream screen increases the pressure risesignificantly. Corroborative evidence can be found for this type ofchange and the clearest example is given by Wisler (1980). The firststage of the GE low speed research compressor was tested alone andthen with the three downstream stages fitted: as Fig. 12 shows, theeffect of the downstream stages is to increase the first stage pressurerise at flow coefficients below 0.37 and reduce it at flow coefficientsabove 0.4 (the design flow coefficient was 0.408). These trends are


Deverson C4 single stage compressor GE 4 stage low speed compressor

t

0.80

1

Design 4

Design 4)0.90

Ca)U_a)@00mm

0.80NNQ

O

0

is0~ 0.70

Ca)U_a)00va)a,

yNCl,Q

@O 00c6

I°-

70

60 I stage 1 alone

stage 1 of 4 stage

from Wisler (1980)

L^ 0117

0.50

0.40 0.45 0.50 0.55 0.60 0.65 0.25 0.30 0.35 0.40 0.45 0.50Flow coefficient, 0 Flow coefficient, 0

FIG. 12 THE EFFECT OF DOWNSTREAM STAGES ON THE FIRST STAGE PRESSURE RISE.

similar to those seen on the STG build on the Deverson rig(including the reduced pressure rise at higher flow rates whichremains unexplained), suggesting that the screen has succeeded insimulating important features of the multi-stage flow field.

The effect of the reduction in the stator corner separation onstage performance at 4)=0.51 was calculated by taking thefree-stream stagnation pressure at stator exit (Fig. 8) asrepresentative of the rotor exit pressure and then calculating a statorloss for each build using the average total pressure at stator exit.This showed that 25% of the difference in total-to-total pressure risebetween the TG and STG builds and approximately 1% of the 3.3%change in efficiency is due to the reduction in stator loss near the endwalls. The remainder of the change in pressure rise is due to thestagnation pressure differences seen in Fig. 8. The measuredreduction in rotor deviation of approximately one degree atmid-height corresponds well with the 0.8 degree change that asimple velocity triangle analysis shows would give the measured0.025 increase in rotor pressure rise coefficient at 0=0.51, shown inFig. 8. This change in deviation is much less than the maximumprobable error in rotor exit flow measurement, but the compatibilityof the deviation change measured and the deviation required to givethe increased pressure rise seems persuasive. Unfortunately, areduction in deviation and increase in rotor work appearsincompatible with the torque measurements: at 4)=0.51, the torqueof the STG build was 0.5% less than that of the TG, which is adifference a little greater than the observed scatter in this type ofmeasurement. The measured increase in stage efficiency at thisoperating point is as large as it is because the torque was very similarfor the TG and STG builds. Because the alteration in rotor pressurerise was so surprising a check was performed on the samecompressor rig but with a different compressor stage. This was a50% reaction stage with inlet guide vanes. Three Kiel probes,spaced at 120 degree intervals around the circumference, werepositioned at mid-pitch and mid-height near the stator outlet (Bolger,1994). These probes and the compressor were left untouched whenthe downstream screen was removed: the results showed a smallincrease in stagnation pressure rise with the screen present

downstream of the stator, backing up the trend described in thispaper. However the probes also showed an increase in thenon-axisymmetry of the flow at stator exit with the screen fitted: it isnot known whether this also has been the case for the C4 stage.

Changes to rotor performance with the introduction of thescreen were not anticipated. Rotor pressure rise could haveincreased by the measured difference between TG and STG, whilethe torque remained unchanged, had the rotor losses reduced alongthe span. For instance, reducing the mid-height loss coefficient from3.9% to 1.5% would result in an increase in mid-height pressure riseequal to that measured at stator exit between the wakes. Howeverthe slant hot wire did not measure such a change in rotor loss (whichis greater than the error band) and so it is concluded that the screendid not improve rotor performance by reducing profile loss.

Leaving aside the difficulties posed by the small reduction intorque, there are possible explanations for the change in rotor workwith the downstream screen such as a spanwise re-distribution offlow through the rotor (rather than a deviation change), since thestator passage end-wall blockage had been reduced by the screen.The measured stagnation pressure rise is increased all the way up thespan, Fig. 8, which suggests that it is not a re-distribution that isresponsible. In any case the re-distribution shown by the axialvelocities calculated from SSHW data (see Fig. 6) is insufficient togive this alteration in work.

In the same way that the screen had thinned the stator wakesnear the end-walls, so the screen might conceivably have acted onthe rotor wakes at all radii and reduced the viscous (as distinct fromthe inviscid) component of rotor deviation. However, this is notsupported by the measured rotor wake momentum thicknesses of theSTG build, which are similar to or thicker than those of the TG build(see the profile losses calculated from the wake momentumthicknesses in Table 4).

There is an unsteady, two-dimensional mechanism for thereduction of rotor deviation. Much of the deviation of a C4 profile isinviscid (see Cumpsty, 1989): to follow the curved surface of theaerofoil section, the curvature of the fluid in the blade-to-blade planemust be supported by a cross-passage pressure gradient. Just

10


beyond the trailing edge of an isolated blade the static pressure isnearly uniform in the circumferential direction (since there can be noblade force) and substantial flow curvature cannot be supported.The bulk of the fluid is unable then to follow the blade surface anddeviation results. In a multi-stage compressor the trailing edge of arotor, for instance, is not moving through a circumferentially uniformstatic pressure field, but instead through the potential field of thedownstream stator. At times the rotor trailing edge will be in acircumferentially varying static pressure field with the sense of thegradient such that it will support the rotor turning and reduce therotor deviation. At other times, of course, the reverse will be trueand the effect of the downstream blade row might be to increase thetime-averaged deviation. It is not known whether these effects willcancel out or, if they do not, in which sense they will alter thedeviation and pressure rise. The unsteady mechanism just describedwill be stronger as flow rate is reduced towards stall, since the statorwill be more forward loaded and coupling between rotor and statorwill be stronger. The downstream screen reduces the stator hubseparation and so increases the stator lift and therefore the statorstatic pressure field, through which the rotor passes. If thisphenomenon was sufficient to cause a reduction in rotor deviation ofbetween 0.5 and 1.0 degrees it would imply that the rotor-statorinteraction for this stage is very sensitive to stator performance.Note that at 4=0.57, where the stator hub separation is small andrelatively unaffected by the downstream screen, there is littledifference between the TG and STG builds. This hypothesis couldbe tested by use of a 2D unsteady inviscid code modelling themid-height rotor and stator blade sections.

It should be mentioned that there are references in the literatureto changes to rotor deviation. As an example, Smith (1970)recounts that the measured rotor and stator deviation of blading inthe GE four stage low speed multi-stage compressor was lower thancascade data suggested by more than a degree and attributed this toblade-row interaction effects. Reducing the axial gap between bladerows from 0.27 to 0.07 mean chords produced changes to thecompressor performance that were consistent with the bladedeviation being reduced further by up to 0.5 degrees.

The downstream screen in effect has steepened the slope of thestage pressure rise characteristic, resulting in increased pressure riseat low flow rates and reduced pressure rise at high flow rates. It hasnot been possible to explain this, although it is in part due to thereduced severity of the stator separation. Further progress might bemade by modelling this particular stage using first a 2D and then a3D unsteady code as a means of prompting a theoretical appreciationof the phenomenon.

8. CONCLUSIONS.

It has been shown that simulating the effects of a multi-stageenvironment alters the behaviour of a single-stage compressor rigsignificantly. The effects introduced, in addition to thickening of theend wall boundary layers, were inlet free-stream turbulence and adownstream screen. It follows that single-stage compressor testsmay be inappropriate both for the development of blading and forunderstanding phenomena in compressors unless steps are taken torepresent better the flow conditions that will be met within themulti-stage environment.

Introduction of the turbulence grid caused a large increase instage efficiency, measured as 1.8%. The increased free-streamturbulence thickened the rotor profile wakes by 10% but reduced theseverity of the rotor hub corner separation. The stator corner

separations seemed to be unaffected by increased free-streamturbulence at compressor inlet. Higher stagnation pressure wasdelivered near to the casing at stator exit between the stator wakes.This could be due to the influence of free-stream turbulence on therotor exit end wall flow.

Simulating the potential field of a downstream stage with thescreen reduced the extent of stator separation near the hub end-wall,particularly at flow coefficients below that of peak efficiency. At4=0.51 this change to stator performance accounted for a 1%increase in stage total-to-total pressure rise coefficient and a 1%increase in stage efficiency.

Introducing the downstream screen apparently caused the rotorpressure rise to increase at higher loading; this was achieved withoutincrease to the rotor torque, leading to an increase of approximately2% in stage efficiency at X0.51. No satisfactory explanation forthis can yet be given.

9. ACKNOWLEDGEMENTS.

Thanks are due to Dr. J. P. Longley for his interest andsuggestions during the course of the experimental work that formsthe basis of this paper. The authors are grateful to J. J. Bolger forproviding an independent check of the effect of the screen on rotorwork. Finally, we would like to thank Rolls-Royce PLC, theDepartment of Trade and Industry and the Defence ResearchAgency, Pyestock for both funding and permission to publish thiswork.

10. REFERENCES.

Bolger, J. J., 1994, Private communication.Camp, T. R., Shin, H.-W., 1994, "Turbulence Intensity and

Length Scale Measurements in Multistage Compressors.", ASMEInternational Gas Turbine and Aero-engine Congress andExposition, The Hague, June 1994, Paper 94-GT-4 (to be publishedin the ASME Journal of Turbomachinery).

Cumpsty, N. A., 1989, Compressor Aerodynamics, Longman,UK, page 168.

Cyrus, V., 1986, "Experimental Study of Three-DimensionalFlow in an Axial Compressor Stage", ASME 31st International GasTurbine Conference and Exhibit, Dusseldorf, Paper 86-GT-1 18.

Cyrus, V., 1988, "Effect of the Inlet Velocity Profile in theThree-Dimensional Flow in a Rear Axial Compressor Stage", ASMEJournal of Turbomachinery, Vol. 110, October 1988, pp 434-440.

Denton, J. D., Cumpsty, N. A., 1987, "Loss Mechanisms inTurbomachines", Institute of Mechanical Engineers InternationalConference Turbomachinery - Efficiency Prediction andImprovement, Cambridge, Paper C260/87.

Goto, A., 1992, "Three-Dimensional Flow and Mixing in AxialFlow Compressor with Different Rotor Tip Clearances.", ASMEJournal of Turbomachinery, July 1992, Vol. 114, pp 675-684.

Horlock, J. H., 1978, Actuator Disk Theory - Discontinuities inThermo Fluid Dynamics, McGraw-Hill.

Howard, M. A., Ivey, P. C., Barton, J. P., Young, K. F., 1993,"Endwall Effects at Two Tip Clearances in a Multi-Stage Axial FlowCompressor with Controlled Diffusion Blading.", ASME Journal ofTurbomachinery, October 1994, Vol. 116, pp 635-647.

Hunter, I. H., 1979, "Endwall Boundary Layer Flows andLosses in Axial Turbomachines. ", PhD Thesis, University ofCambridge.

11


Lambie, D., 1989, "Inlet Distortion and Turbofan Engines.',PhD Thesis, University of Cambridge.

Lewis, K. L., 1993, "The Aerodynamics of ShroudedMulti-Stage Turbines.", PhD thesis, University of Cambridge.

Li, Y. S., Cumpsty, N. A., 1991, "Mixing in AxialCompressors: Part II - Measurements in a Single-Stage Compressorand a Duct", ASME Journal of Turbomachinery, Vol. 113, April1991, pp 166-174.

McDougall, N. M., 1988, "Stall Inception in AxialCompressors", PhD Thesis, University of Cambridge.

Place, J. M. M., 1995, "Three-Dimensional Flow in AxialCompressors", PhD Thesis, University of Cambridge (to besubmitted).

Savin, N. M., 1993, Private communication.Smith, L. H., 1970, "Casing Boundary Layers in Multistage

Axial-Flow Compressors. ", Flow Research on Blading, ed. L. S.Dzung, Elsevier, pp 275-303.

Wisler, D. C., 1980, "Core Compressor Exit Stage Study,Volume II - Data and Performance Report for the BaselineConfiguration.", NASA Report No. CR-159498 ,November 1980.

APPENDIX 1. AVERAGING OF EXIT STAGNATIONPRESSURE.

Combined radial and circumferential traverses were used atstator exit (plane 3, Fig. 1) to define both the detailed distribution offlow and the stage pressure rise.

To define the stage pressure rise an average value appropriate tothe whole annulus is desirable (circumferential and radial coverage).This can readily be obtained from a ring of static pressure tappingsdistributed around the circumference for the Baseline and TG buildas described below. This could not be done for the STG build. Thestagnation pressure can be derived from the average of the exitcasing static pressures by adding an average exit dynamic headderived from the known mass flow. For convenience the averagestagnation pressure derived in this way is termed a 'continuityaverage' (since the dynamic head is derived from mass flowcontinuity). Well defined preliminary characteristics were derivedbased on the exit continuity average stagnation pressure for theBaseline and TG builds.

A study has been carried out on both the averaging of stagnationpressure and the definition of interpolation procedures necessary toderive pressure at flows between those traversed. The aim of thisstudy was to develop a procedure which can be used for all threebuilds and which gives similar characteristics to those based oncontinuity average pressure for the Baseline and TG builds.

To choose a suitable means of circumferential averaging, valuesof average stagnation pressure were calculated usingmass-weighting, area-weighting and also using a simple mixingcalculation (the 'mixed-out' pressure). It was found that massaveraging gave efficiency values which were too high close to stallsince in the comer separation regions, where the stagnation pressureis low, there is low mass flux and therefore low weighting.Area-weighting in the circumferential direction gave improvedresults near stall and efficiency values close to those derived from thecontinuity average pressure at a flow coefficient of 0.51 but the rateof change of efficiency with flow rate was much steeper. The bestcompromise was achieved with the mixed-out value of stagnationpressure (this averaging method was also used by Howard et al,1994). In the radial direction mass-weighted averaging is fairlystandard and was used throughout the study. Consequently a

combination of mixed-out circumferentially and mass-averagedradially was used to obtain the compressor characteristics for allthree builds shown in Fig. 4.

12


Documents

Simulating the Multi-Stage Environment for Single-Stage