7/30/2019 visousflow-

1/10

Y . ElazarR. P. Shreeve

Turbopropulsion Laboratory,Department of Aeronautics and Astronautics,Naval Postgraduate School,Monterey, CA 93943

Viscous Flow in a ControlledDiffusion Compressor CascadeWith Increasing IncidenceA detailed two-component LD V mapping of the flow through a controlled diffusioncompressor cascade at low Mach number ( ~ 0.25) and Reynolds number of about7x 1 0 5 , at three inlet air a ngles from design to near stall, is reported. It w a s foundthat the suction-side boundary layer reattached turbulent after a laminar separationbubble, and remained attached to the trailing edge even at the highest incidence, atwhich losses were 3 to 4 times the minimum value for the geometry. Boundary layerthickness increased to fill 20 percent of the blade passage at the highest incidence.Results for pressure-side boundary layer and near-wake also are summarized. In formation sufficient to allow preliminary assessment of viscous codes is tabulated.

IntroductionComputational Fluid Dynamics (CFD) has enabled the development of rational methods for designing axial compressorblade elements. Several centers (Sanger, 1983; Hobbs andWeingold, 1984; Dunker et al., 1984, for example) have reported methods for designing "controlled-diffusion (CD )"blading. The approach is to use an in viscid flow compu tation(in either the direct or indirect m ode) to arrive a t a blade shape

over the surface of which the adverse pressure distribution issuch that the computed boundary layer remains attached. Experimental results have clearly demonstrated (Hobbs andWeingold, 1984; Dunker et al., 1984; Sanger and Shreeve,1986) that the design technique works, but also have shownthat the CD blade shapes can sustain a wider range of stall-free incidence than can Double Circular Arc (DCA) bladeshapes designed for the equivalent design point conditions.Since stall-margin is extremely important in the design of military aircraft gas turbine engines, it is fundamentally importantto be able to predict the off-design and stalling behavior ofCD .blading during the compressor design process. Thus efficient computational methods are required that correctly describe flows containing viscous effects such as are shown inFig. 1. Figure 1 illustrates schematically features that may ormay not all be present within a given blading at a given inletair condition. However, whether or not there exists a leadingedge separation "bubble," and where (and over what length)transition occurs in the boun dary layer, will determine whether(or where) separation will occur in the adverse pressure gradienton the suction side of the blade. For a computer code to predictcorrectly the off-design losses and onset of stall of a givenblading, the code must incorporate the ability to predict anddescribe separation bubbles and the process of transition. The

Contributed by the International G as Turbine Institute and presented at the34th International Gas Turbine and Aeroengine Congress and Exhibition,Toron to, Ontario, Canada , June 4-8,1989 . Manuscript received at ASME Headquarters January 17, 1989. Paper N o. 89-GT-131.

The purpose of the present investigation was to obtain datawith which to assess the ability of emerging codes to predictthe off-design behavior of (particularly) CD blading.The work follows the earlier investigation of Sanger andShreeve (1986) wherein the design and experimental evaluationof a particular CD stator cascade was reported. The reportedexperimental results included probe surveys to obtain loss coefficients, blade surface pressure d istributions, and blade surfaceflow observations using the china-clay technique, at seven incidence angles over the useful range of the blading. Guidedby the previous results, in the present study, a two-componentLaser-Doppler Velocimeter (LDV) system was used to obtaina detailed m apping of the flow througho ut the blading at design

T R A N S I T I O N

S E P A R A T I O N ! ? )

S E P A R A T I O NB U B B L E

Fig. 1 Probable viscous flow features2 5 6 / V o l . 1 1 2 , A P R I L 1 9 9 0 Tr ansact i ons o f the A SM E

Copyright 1990 by ASMEDownloaded 11 Jan 2013 to 202.3.77.222. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_U

7/30/2019 visousflow-

2/10

Table 1 Blade coord inate s , cascade geome t ry and tes t condi t ions

LNumber of Test Blades = 20Number of Inlet Guide Vanes = 59kFig . 2 C a s c a de w i nd tunne l se c t i on

and two higher incidence angles toward stall . Measurementswere obtained of the outer (inviscid) flow through the bladepassage, the suction-and pressure-side viscous layer growth,and the initial wake development. In the present paper, theexperiment itself is described and results for the suction-sideviscous behavior (which provide the most critical test of acomp utational description) are presented and discussed. Otherresults are briefly reviewed. Sufficient information is given intables to enable a preliminary code assessment to be made byapplication to the geometry of the experiment. A completeaccount of the experiment and the results was given by Elazar(1988).Test Faci l i ty and Instrumentation

The subsonic cascade wind tunnel at the Naval Pos tgradu ateSchool was used in the present study. The configuration ofthe wind tunnel, the test cascade and operating instrum entationwas the same as was reported by Sanger and Shreeve (1986).The geometry of the test section is shown in Fig. 2. A 0.64 cmPlexiglas window was located in the removable tunnel frontsidewall in the position shown. Adjustment of the inlet airangle ((${) required adjustment of the lower endwall (0WL)

7/30/2019 visousflow-

3/10

Table 2 LDV system characteristics and estimated measurement uncertaintiesLASER T ype

ManufacturerNom i na l P oweiW A V E L E N G T HBl usGreenF R I N G E S P A C I N GBlueGreenF O C A L L E N G T HN U M B E R O F F R I N G E SH A L F A N G L EM E A S U R I NG V O L U M ELengthdiameterF R E Q U E N C Y S H I F T

IncrementsS I G N A L P R O C E S S O RT ypeManufacturer

Argon-IonLcxel1.5 watts

488.0 run514.5 nm

4.51 n m4.76 urn762 . 0 m m283.1

Z 5 m m133.0 um

2 k H z - 4 0 k H z1,2,5,10 MH z

CounterTSI Mod 1990COUNT E R S E T T I NGSFil ters As RequiredGain As Required

Mode CoincidentSingle Measure/Burstcyc l e / bu r s t 8Com par i son 4%

ItemX

X

Yy

z

p.p

'armT.KLX

d fV

Descr ipt ionBlade- toBlade(Passage)Distance f romBlade SurfaceVerticalDistance f romBlade SurfaceS panwi sePitch, Roll , Yawof LDV SystemP l enum P r essu r eP r essu r eAtm. PressureP l enum T em p .

L DV Coun t e rClockBeam Half AngleFocal LengthWavelengthFr inge SpacingParticle Velocity

MethodMill ing M achineFJec. ReadoutMil l ing MachineFlee. ReadoutMil l ing MachineScaleMil l ing MachineScaleMil l ing MachineElec. Readou tElectronic Leve lSperry Model 45WaterManometerScanivalveTransducerMercuryBarometerI ron Constan anThermocouple

0 3 3 %0 . 6 5 %

Uncer tainty0.25 mm

0. 05 m m

0.25 mm0.05 mm

1.25 mm0. 1

25 Pa12 Pa35 Pa0.14 C

1 n-sec20 arcsec(0.2%)7 . 60 m m0.1K,0 . 3%@ l O m / s@ 100 m / s

54-3210

-1-2-3-4-5-6-7-8-9

-10-11-12--13-14-15-16-17-18

- ( j $ 2.6921.722

M 0 9 1 9 t-WTggJTF $ 5 ) 0.000 f

- i 1 1 1 1 r -5 -4 -3 - 2 - 1 0 1 T 1 1 r 1 1 1 j 2 3 4 5 6 7 8 9 10X(cm)

F ig . 3 Cascade passage geom etry and LDV survey stations

Laser Velocimeter SystemOptics and Data Acquisition. A four-beam, two color TSlModel 9100-7 LDV system was used. The system operated inthe backscatter m ode. Characteristics of the system are given

Power, Water

Optical Bench

Lexel4WattArgon-Ion Laser

Collimator

Color Separator

Lens (762 mm )

Beam Expander

Beam Spacers (2)Photomultiplier (blue)

Photomultiplier (green)Steering Modules

Frequency Shifter (blue)Frequency Shifter (green)BeamsplitterPolarizer (blue)

Beam DisplacerBeamsplitterPolarizer (green)

F ig . 4 Components of the LDV system

in Table 2. The components of the system are shown in Fig.4. The laser was a Lexel, 4-watt argon-ion laser. A prism andmirror system separated the two colors from which two beamsof each color were generated, equidistant from an optical axis.The beams, and fringe pattern in the measurement volume,were arranged vertical/horizontal or at 45 , depending on thelocation of the measurement. Two Bragg cells shifted the frequency of one beam in each pair to allow measurements inregions of reversed flow. After the beam expander, the finallens produced a focal measurement volume near the midspancenter plane of th e wind tunnel test section, 762 mm from thelens. Two photo-de tectors collected light scattered by particlespassing the measurement volume via the same optics. Two TSlModel 1990 counter-type signal processors received the outputfrom the photo-detectors and produced digital voltages proportional to the Doppler frequencies of detected events. Thesedata were transferred by way of an interface to a Hewlett-Packard Model HP 1000 computer operating with the TSldata acquisition software program DRP3.

Traverse System. The laser and optics, mounted on anoptical bench, were supported on the bed of a large millingmachine installed on the laboratory floor next to the windtunnel. A view of the installation is given in Fig. 5. The m illwas aligned and leveled such that the manual x, y, z traversemotions of the mill table moved the measurement volume ofthe LDV system horizontally (blade-to-blade), vertically (normal to the leading-edge locus), and fore-and-aft (spanwise)respectively. While the traverse was actuated manually, the xandy coordinates were indicated on digital electronic readouts,which were hand-recorded. The optical bench also could be2 5 8 / V o l . 112, APR IL 1990 Tra nsa ct ions of the AS ME

Downloaded 11 Jan 2013 to 202.3.77.222. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_U

7/30/2019 visousflow-

4/10

yawed to direct the optical axis 3.5 0 with respect to thespanwise direction.Seeding. Olive oil droplets were used. The particle generator incorporated a TSI Model 9302 atomizer driven froma shop-air supply. Oil droplets carried by the controlled airpressure were injected by way of a 6.4 mm wand through theplenum of the wind tunnel below the inlet guide vanes (Fig.2). Seeding only near the center plane of the flow beneath thecorrect pair of guide vanes resulted in a distribution of seed

material across the 7th-8th blade space with no contaminationof the window, and no detectable disturbance to the inlet flow.Preliminary tests were conducted to optimize the seed particledistribution. An air pressure of 40 psi was selected to obtainan average particle size of 0.9microns with standard deviation0.45 microns. At this setting, 95 percent of the particles weresmaller than 1.8 microns. These characteristicswere consideredto be satisfactory by the criteria proposed byDurst et al. (1976)or Dring (1982).Positioning. A special jig, described by Elazar (1988),was manufactured, which attached to the trailing edges ofblades 7 and 8 at midspan. The jig contained six holes, each0.033 cm in diameter, on which the measurement volume couldbe targeted. When themeasurement volume was on target, thefour beams were clearly visible on the back wall of the test

section, and X, Y coordinates of the measurement point werethen known to the accuracy of the jig.Measurement Uncertainty. The estimated measurementuncertainties also are given in Table 2. The uncertainties inthe parameters K, L and Awere given by the manufacturer.Fringe spacing (df ) was calculated following Durst et al. (1976)as df = Alsin K. The particle velocity uncertainty (..:iVI V) wascalculated using

Fig. 5 LDV system Installation

where ..:it = instrument clock uncertainty (I nanosec), and t= 8d l V = time to cross eight fringes. Since the particles donot follow the flow precisely, the uncertainty in the velocitycan be larger than the uncertainty in the particle velocity. Theuncertainty in positioning the measuring volume on the bladesurface, which gave an equal uncertainty in the displacementof the measurement from the surface in boundary layer surveys, was shown by experiment to be less than 0.05 mm.

Experimental ProcedureProgram Overview. A preliminary experiment was conducted in which the test blading was removed from the testsection and the inlet and outlet endwalls were set straight at

13 = 40 0 A flat plate was installed in the center of the testsection and the boundary layer development on the flat platewas measured to verify all aspects of the LDV measurements.Good agreement of the velocity profiles with the Blasius solution was obtained where the boundary layer was laminar,and agreement with other published measurements was seenwhere the boundary layer was turbulent (Elazar, 1988). TheCD blading was then installed and detailed measurements wereobtained in turn at inlet air angles 131 = 40 0 (the design condition), 131=43.4 0 (about twice minimum loss) and 131 =46 0(near to stall) as determined by Sanger and Shreeve (1986). Ateach inlet air angle, measurements were made to map specificregions of the flow; namely, the inlet flow field 0.3 chordsupstream of the blade leading edge (Station 1 in Fig. 3); thepassage between the 7th and 8th blades (Stations 2-15 in Fig.3); the boundary layers on the suction and pressure sides ofthe 7th blade (nominally Stations 3-15 in Fig. 3); and the wakedownstream of the 7th blade (Stations 16-19 in Fig. 3).Setup Procedure at Each Ai r Angle. Wind Tunnel FlowVerification. At each inlet air angle, the inlet and outlet wallsand inlet guide vanes were adjusted to obtain near-uniform

wall static pressure (in the blade-to-blade direction) both upstreamanddownstream of the test cascade (Sanger and Shreeve,1986). The blade-to-blade flow angle distribution was measured by surveys with a United Sensor Model DA-125 probe at1.8 blade chords upstream, and by the LDV system at 0.3chords upstream. The probe measurements provided verification of the inlet flow conditions, which were surveyed extensively in earlier studies by McGuire (1983), Koyuncu (1984)and Dreon (1986).Probe results, spanning four blade passages at design inletair angle, were reported by Sanger and Shreeve (1986). LDVsurvey results, 0.3 chord lengths upstream of the passage be-

50

0 li 00 l i 00 l i 00 li 00 l i 00 li 00 li 00 li 00 li 0D li 0D l i 0D li 0D li 0D l i 0D li 0l i 00.535

-1.0

-0.5xS

0.0

40 45fJ (deg)

lOp- 40 0 l i P_43.4 0 0 P= 46 01Fig. 6 Inlet flow field at Station 1

-1.0 DDroto0-0.5 00x a0S 000.0 III0000

0.50.8 0.9 1.0 1.1 1.2VIV ret

Journal of Turbomachlnery APRIL 1990, Vol. 1121259Downloaded 11 Jan 2013 to 202.3.77.222. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

7/30/2019 visousflow-

5/10

F ig . 7 Suction surface boundary layer and blade wake measurem entsto verify two-dimensionalitytween blades 7 and 8 (as far as the window w ould allow), areshown in Fig. 6.

Optical System Alignment. Proper beam crossing andfringe patterns were verified at intervals while mapping theflow at a fixed air angle. A complete optical system alignmentwas carried out for each inlet air angle. It was preferred toorient the crossing pairs of beams at about 45 to the flowvelocity vector. Th us, at the lower stations (1-5 in Fig. 3), thepairs of beams were oriented vertical/horizontal whereas atstations 6-19, the pairs were rotated through 45. The inletflow and lower passage were surveyed without Bragg cellsinstalled. The Bragg cells were installed and frequency shiftingwas used at all other stations.TraversePositioning. For inlet, passage and w ake surveys,the alignment jig was attached to blades 7 and 8 and the tablewas adjusted to target the measurement volume on a specifictarget hole. When the pattern of four beams was clearly andevenly visible on the back wall of the test section, the X, Ycoordinates of the target hole were entered into the digitalelectronic readouts. The readouts then indicated the displacement of the measuring volume in the X, Y coordinate systemwhose origin lay at the center of curvature of the trailing edgeof the 7th blade.For boun dary layer surveys, first the optical table was yawed3.5 horizontally with respect to the table of the millingmachine. The measuring volume was positioned at the requiredlevel (Y displacement), using the alignment jig. The measuring volume was then traversed horizontally ( X displacement)until the four beams crossed on the blade surface. The crossingwas observed by eye and a n un certainty of less than 0.05 mmwas achieved with ease. The digital readouts were set to zero.The x and y coordinates for displacements normal to the wallwere calculated from the local surface slope.

Survey Procedures at E ach Air Ang le. Inlet Flow and Tunnel Reference Conditions. An LDV survey was made at Station 1 (Fig. 3). The results at each inlet air angle are shown inFig. 6. Despite the proximity of the survey to the blades, thevelocity and flow angle were found to be uniform to within1 percent and 0.75, respectively. In Fig. 6, the localvelocity (V) is shown normalized to a reference velocity (K ref)obtained from tunnel reference parameters at the time of recording the measurement. Such referencing removes the effectsof temporal variations in wind tunnel supply and ambientatmospheric conditions. It can be shown, using Appendix Cof Elazar (1988), that for small variations in supply and ambient atmospheric conditionsV, = C^TdP,0-PaVPa (1)

where C is a constant for a fixed test geometry (inlet air angle).The value of C at each air angle was calculated as the averagevalue of Vx/\jTm{Pm-P^/Pa for the points in the inlet flowsurvey, ^//velo cities measured thereafter were normalized withrespect to the inlet flow reference velocity (V re[ ), calculatedusing equation (1) with the tunnel and ambient conditions atthe time of the measurement.Passage, Boundary Layer and W ake Surveys. Passage velocities were measured at approximately 30 locations acrosseach of the 14 stations (Fig. 3) from the leading to the trailingedge. Boundary layer surveys were made at predeterminedintervals locally normal to the blade surface, beginning atStation 15 in Fig. 3 (in line with the center of curvature) andat each station toward the leading edge until the boundarybecame too thin. B oth pressure and suction surfaces of the 7thblade were mea sured. W ake surveys were made at four stationsdownstream of the 7th blade (Stations 16-19 in Fig. 3).Verification of Periodicity and Two-Dim ensionality. Inview of the unusually large number of test blades (20), periodicity between adjacent blade passages was to be expectedand had been demonstrated for the central passages in earlierstudies (Sanger and Shreeve, 1986). Periodicity was verified inthe present study by making LDV surveys downstream of the7th and 8th blades. The blade wakes were indistinguishablefrom each other (Elazar, 1988). Two-dimensionality also wasexamined since the wind tunnel was not equipped with wallsuction. Previous studies (Sanger and Shreeve, 1988) had shownthat conditions far downstream of the blade trailing edges wereindependent of spanwise displacement for 5 cm from m id-span. H owever, conditions w ithin the blade passage could notbe examined at that time. In the present study, LDV surveyswere made in the wakes and through the boundary layers atthe trailing edge of the 7th blade at midspan and at 2.54cm from midspan. Wake and suction surface boundary layerresults are shown in Fig. 7.

Data Reduction. The LDV data were acquired using program DRP3 on the H P 1000 compu ter, stored on floppy disksand transferred to the mainframe IBM 370-3033 for processing.Tunnel reference measurements were recorded and enteredmanually. Separate computer program s were written to reduce,tabulate and plot results from passage, boundary layer andwake surveys. Blade surface pressure m easurements, acquiredand' recorded using the HP 9845 controller, also were transferred to the mainframe for processing.

Results and DiscussionInlet Flow. The inlet flow measurements are shown in Fig.6. At each inlet air angle, the inlet velocity (V tet) was set toapproximately 85 m/s (Mach number = 0.25). The measuredfree-stream turbulence level, the result of the many inlet guide

2 6 0 / V o l . 112, APR IL 1990 Transa c t ions o f the A S MEDownloaded 11 Jan 2013 to 202.3.77.222. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_U

7/30/2019 visousflow-

6/10

i . 0 .5-

P-40< P - 43.4 ' P = 46 'F ig . 8 Flow at the passage entrance (Station 2) and exit (Station 15)

F ig . 9 Measured variation in AVDR

\ \

. ; ; : - ---

- P

^

- 4 0 -

^ H T r^r-

p -

" - - - -

4 3 . 4 ' " -

~"~"- - . ----'.

P-

B

4 6

^

0 50 10[% CHORDFig. 10 Effect of inlet air angle on the edge velocity distributio n

vanes (59) far upstream, was 1.4 0.2 percent at all conditions.The average flow angle measured, using the LDV system 0.3chord lengths ahead of the blading (Fig. 6), was consistentwith measurements made using a pneumatic probe 1.8 chordlengths upstream; namely, 40.0, 43.4 and 46.0. At the upstream station, the flow angle varied less than 0.1 across oneblade space.Flow Through the Passage. From blade-to-blade surveysat 14 stations, from the leading to trailing edge, the streamlinepaths through the passage were calculated by integrating the

volume flow rate . Similarly, the local volume flow to the inletvolume flow for one passage gave the distribution of the AVDRassuming negligible effect of compressibility. It was found(Elazar, 1988) that no sudden changes occurred in the geometryof the streamlines as the incidence was increased. Rather, aprogressive shift occurred away from the suction side of thepassage. This orderly change in behavior is evident in Fig. 8,which shows the velocity and flow angle variations at the entrance (Station 2) and exit (Station 15) of the passage. Thedistribution of the AVDR from the reference inlet station todownstream of the blading is shown for the three inlet airangles in Fig. 9. The magnitude of the variation (to a maximumvalue of 1.05 just downstream of the blade) was consistentwith previous results (Sanger and Shreeve, 1986) which gaveoverall values of 1.04-1.06 based on probe survey measurements well downstream (1.7 chord lengths) and well upstream(1.8 chord lengths) of the blading. Based on the results in Fig.9, a linear variation of the AVDR from 30 percent chordupstream through the blade passage should be used for codeassessment.Also derived from the passage surveys was the velocity atthe edge of the viscous layer as the suction and pressure surfaceswere approached. The results are shown in Fig. 10. The edgevelocity was taken either as the maximum value of the velocity(where gradients were large and the boundary layer was thin,on the suction side near the leading edge, for example) or asthe velocity where the turbulence level departed from its free-stream value (where gradients were small and the boundarylayer was well defined, on the pressure side near the trailingJourna l o f Turbomach inery APRIL 1990, V ol . 112 /261

Downloaded 11 Jan 2013 to 202.3.77.222. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_U

7/30/2019 visousflow-

7/10

From Surface Pressure MeasurementsD Suction Side A Pressure Side

Fig. 11 Comp arison with edge velocity inferred from surface pressures

edge, for example). The curves in Fig. 10 are shown withoutdata points for clarity in comparing results at increasing incidence. Figure 11 shows, for comparison, the edge velocityderived from measurements of surface pressure, with the assumption that p ressure was constant through the viscous layerand that the flow was incompressible; namely, U/VK{=Vd-cy .

It is clear from Fig. 11 that while the two m easures of edgevelocity agreed quite well at design incidence (j3i = 40), theydeparted from each other as the incidence was increased, notably on the suction side of the blade. In particular, the surfacepressures indicated higher peak velocities than were measuredusing the LDV system. Since a large transverse com ponent ofacceleration would be present locally, the ability of the seedparticles to track the streamlines must be questioned. However,it also is noted that the first com putations by Sha mroth, usinga Navier-Stokes code applied to the specified geometry andtest conditions (Shamroth, 1987), also did not predict the peaksuction pressures, which were measured pneumatically. Asshown in Fig. 12 however, excellent agreement was obtainedbetween the predicted and measured pressures at all incidenceangles, except near the suction surface leading edge. Earliercomparisons of measured pressures with inviscid code calculations (Sanger and Shreeve, 1986) had shown disagreementtoward the trailing edge, bu t had tended to confirm the extremesuction p eaks.Suction-Side Boundary Layer. Illustrative examples of theresults of detailed boundary layer surveys are shown in Fig.13. Equivalent results to those shown for /3,=46 were ob-

aA ^ ^

1 P - 4 6 ' |

50iCHORD

a

" V saa-a**1

IP-43.4-1

- - o _ _ B ~ - u _ r x j . n o50

5 CHORD

| (3 - 40 |

50% CHORD

From Surface Pressure Measurementsn Suction Side A Pressure Side

F ig . 12 Measured and calculated surface pressure distribution s

tained at /3, =40 and fr =43.4. The figure shows results forthe velocity components normal ( v) and parallel ( u) to thesurface, and the turbulence level.The presence of a separation bubble, detected earlier usingthe china-clay technique (Sanger and Shreeve, 1986) was confirmed by the LDV measurements. However, no informationon the flow within the separation bubble was obtained sinceno particles were detected within the region. The absence ofdata near to the w all in surveys at Stations 3 and 4 in Fig. 13was the result of detecting no observable events. It can be seenat Station 3 that the particles were found only in the outerpart of the shear layer over the bubble, consistent with theearlier suggestion that they might not follow the rapid acceleration around the leading edge of the b lade. The presence ofa negative component of velocity normal to the surface wasconsistent with the flow tending toward reattachment downstream of the bubble. If the absence of negative componentsof velocity no rmal to the wall was taken to indicate the completion of reattachment, this point moved from about 15 percent to 30 percent to 40 percent of chord as the air inlet anglewas increased from 40 to 43.4 to 46, respectively. Downstream of the completion of reattachment, the normal component of velocity was negligibly small. The peak turbulencelevel in the shear layer over the bubble reached over 16 percent,decreased during reattachment and became steady at 8-9 percent in the boundary layer to the trailing edge.It was noted that the maximum turbulence level moved gradually away from the wall during reattachment to be at about45 percent of the bounda ry thickness at the trailing edge. Qualitatively, similar trends to those illustrated in Fig. 13 werefound in the results at /3, = 40 and /3, =43.4. Thus the effectof increasing incidence was to progressively delay reattachmentfrom the separation bubble, and to generate a progressivelythicker viscous layer. However, the viscous layer remainedattached all the way to the trailing edge even at the largestinlet air angle. It is probable th at the appearance of separationoccurring just ahead of the trailing edge, when using the china-clay technique (Sanger and Shreeve, 1986), was the result ofgravitational effects on the oil mixture with which the surfacewas wetted, in a region of low dynamic pressures. Considerableeffort was made to verify the present LDV observations, in-

262/Vol .112,APRIL1990 Transa c t ions o f the A S MEDownloaded 11 Jan 2013 to 202.3.77.222. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_U

7/30/2019 visousflow-

8/10

0. 2

7 "t - y i -a0. 2

h -< ^H i ) i .

I 1(- H .

I < i :

i i iV 'V I i -I * = ) :I - ^ H

I -= 1 1 ; 1i= 1 t 1 Li - * i .

i - ^ i .-

0 . 2 ,Ac

n

1 -i .

. . '

)n 1

1

K-" 1H - ' 1

20 TURBULENCE (%)F ig . 13 Results of boundary layer surveys onthe suction surface at0 = 46eluding verification of both mean velocity and fluctuation levels, by conducting redundant measurements within the bladewake with a hot-wire (Baydar, 1988). Excellent agreement be-

g i l i i = ! :50

5 CHORD0.050,040.03' 0.020.010.00

_;

-." ,-"' & ' ". . . A * - *

505 CHORD

P - 4 0 " * Q - 4 3 . 4 " Q = 4 6 F ig . 14 Suction-side boundary layer growth

tween hot-wire and LDV was obtained. It also is noted thatwhereas regions of instantaneous backflow were measured inthe near-wake using LDV, and a time-averaged backflow wasrecorded at Station 16 at (3 = 40, no backflow was detectedat Station 15 at all three inlet air angles.The growths in the boundary layer thickness and displacement thickness and the distribution of the shape factor on thesuction side of the blade are shown in Fig. 14. The thicknesswas defined as the point at which the velocity componentparallel to the surface reached 99 percent of the edge velocity.The displacement thickness was calculated by integration towhere the edge velocity occurred. Downstream of the bubble,the layer was seen to grow steadily. Each increment in flowincidence gave approximately equal increments in boundarylayer growth. At the highest incidence, the suction-side bo undary layer occupied more than 20 percent of the blade passageat the trailing edge. At each incidence, the shape factor reacheda minimum after reattachment and then increased to the trailing edge. The results at the design angle were different thanthose at the higher incidence angles, and gave larger values(closer to separation!) near the trailing edge. The largest valuereached was 2.25 at Station 13, 10 percent of chord ahead ofthe trailing edge. At each angle, the shape factor decreasedslightly from this point aft, to the trailing edge. The differencesbetween design and off-design incidence angles suggest a nonlinear behavior near the leading edge of the blade. This issupported by the measurements of the edge velocity shown inFig. 10. Results for the integral properties at each axial measurement station are given in Table 3.

Pressure-Side Boundary Layer. Similar data were obtainedfor the pressure-side boundary layer velocity and turbulenceprofiles, thickness, and integral properties. The velocity andturbulence profiles (Elazar, 1988) indicated that the pressure-side boundary layer began laminar and became turbulentthrough natural transition. The pressure-side layer was muchthinner, resulting in fractionally larger uncertainties in thederived integral thicknesses. Results for the integral prop ertiesare given in Table 3. (Data are included at three o ther stationsbetween 6 and 9.) The bounda ry layer thickness, displacementJourna l of T u r b o m a c h i n e r y APRIL 1990, V ol . 112 /263

Downloaded 11 Jan 2013 to 202.3.77.222. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_U

7/30/2019 visousflow-

9/10

IIryo

ao =

cs

O N%

0 0!7NNo

ro

sf ?in

csS

C SDO!7 \

oo

vAffmocSp -

r-

arS1o

r ir*io

3so

nm

A3oaseno

CSo

oc-n

A

OO**enmr' l

mosnmm

r^iT I -o

sOOOm

g

ITS

SroO

cviJr-OOenTJ -o

i no

socS

8 >

O 0

>o

Osv nN

eno

soo

sooO si nX )

mm8

mo

Ar $:i\

rio

I Sr-o

sooyCTv

s>nO NmCS^v>=

*MACSo

T fCSOOo

so

1CSocSn?i

3s noo

O 0o>S*r* i

o

x0O

e no

ni Ocs1

srSmOso

8o

rtcS1

v ir~O8omo\r~ *

"o

^rS\o lrSCSo

r~O 0o-o

IICO.

ua

n>

Peu