ISROMAC-9The 9th International Symposium on Transport Phenomena and

Dynamics of Rotating MachineryHonolulu, Hawaii, February 10-14, 2002

1

STEADY FLUID FLOW INVESTIGATION USING L2F AND PIV

IN A MULTI-PASS COOLANT CHANNEL

KEY WORDS: MULTI-PASS COOLANT CHANNEL, PARTICLE IMAGE VELOCIMETRY, LASER-2-FOCUS VELOCIMETRY, CFD VALIDATION.

ABSTRACTFluid flow and pressure loss in a stationary two-

pass coolant channel system is investigated experi-mentally and numerically. One of the main objectivesof this paper is to describe the capability of differentmeasuring techniques, i.e. L2F versus PIV. Results ofpressure losses are not the objective of this paper. Theinvestigated system has an engine-near lay-out with an180° turn including a turning vane, and has smoothwalls. As a first step, the system is analyzed in non-rotating mode. During future work it will rotate aboutan axis orthogonal to the center-line of the straightpasses. The results shown in this paper demonstratethe effect of rotation with isothermal flow conditionexcluding any buoyancy. Turbulent channel flow witha Reynolds number of 25.000 and 50.000, derivedwith the hydraulic diameter of the first pass, was in-vestigated.

At the test rig at DLR several test models weremounted to investigate pressure drop behavior andfluid motion separately. Numerical investigationsregarding flow structure and pressure drop have beencarried out in order to validate a CFD code against theexperimental data. The numerical solution is com-pared with streamline pictures, obtained from flowvisualization of all walls using oil flow technique, andnear wall PIV results. Mid-span flow distribution isreceived from laser light sheet visualization techniquewith oil fog as indicator and also from PIV.

The results presented in this paper clarify the com-plex flow situation given by the two pass system with

inherent turn. Especially in the bend region and behindthe turning vane appear separation regions, vorticesand a wake with high local turbulence. These verydemanding measuring task represents a benchmarktest case for the different used measuring techniques(L2F, PIV and visualization).

INTRODUCTIONThe aero-engine industry and the power generation

industry operate in a highly competitive market.Therefore high requirements in terms of high technol-ogy development as well as cost and developmenttimes reduction are constant major objectives. On theother hand, environmental and safety constraints arean increasingly stringent necessity, which enforces thedemand for new technologies. Currently the industryrelies on expensive and time consuming rig test pro-grams, whereas the existing numerical design toolshave a number of deficiencies in accurately describingthe complex multi-pass coolant channel flow. Underthese conditions the Institute of Propulsion Technol-ogy is involved in national and European researchprograms aimed to providing the industry with highquality experimental data from the flow field for CFDvalidation.

In past projects the flow behavior in rotating pas-sages was analyzed at DLR using wall pressure meas-urements to obtain the pressure drop. In addition,Laser-2-Focus velocimetry (L2F) was used to obtainflow velocity components and fluctuations. Althoughtime-consuming, this non-intrusive, single-point

Dipl.-Ing. Martin ELFERT

German Aerospace CenterInstitute of Propulsion Technology

Linder Höhe, D-51147 Köln, GermanyFax Number: +49 2203 64395

E-mail address: martin.elfert@dlr.de

Dr.-Ing. Marc P. JARIUS

German Aerospace CenterInstitute of Propulsion Technology

Linder Höhe, D-51147 Köln, GermanyFax Number: +49 2203 64395

E-mail address: marc.jarius@dlr.de

2

measurement technique worked very well withinstraight and smooth duct flows, which generally havea moderate degree of turbulence. However, state of theart, serpentine shaped, multi-pass systems areequipped with ribbed walls, in order to improve heatexchange which is typical in realistic configurations.In this case L2F velocimetry was not able to measureaccurate flow properties due to the increased turbu-lence intensities in the vicinity of the ribs as in thebend region and further downstream. The dividingwall separating the two passages forces the flow into asharp turn which generally results in a flow separation.The flow within the separation bubble itself is veryunsteady.

An important work package within these researchprograms is to provide detailed information of theflow field inside the multi-pass coolant channel usingsingle point measurement techniques such as La-ser 2 Focus (L2F) velocimetry [SCHODL, 1977] andplanar techniques such as Planar Doppler Velocimetry(PDV) [RÖHLE, 2000] or Particle Image Velocimetry(PIV) [ADRIAN, 1991, WILLERT, GHARIB, 1991,RAFFEL et al, 1998]. Modern planar measurementtechniques such as PIV are capable of obtaining com-plete maps of flows even at high turbulence. As a firststep toward applying this technique, a multi-passcooling system is investigated in stationary (e.g. non-rotating) mode using two-component PIV. The newresults are compared with results from L2F, flowvisualization and CFD. The high quality of the ob-tained new results encourage the application of two-component PIV to the rotating system. As a logicalconsequence, the application of three-component PIVwill be necessary to obtain the complete flow fieldinformation within the complex flow passages. Thispaper’s intention is to report on the status of applica-bility of PIV in a multi-pass coolant channel.

NOMENCLATUREdh m Hydraulic diameterf mm Lens focal lengthf# - Lens f-numberν m²/s Kinematic viscosityn rpm SpeedRe - REYNOLDS number, Re=ρv0d/ητ µs Pulse delayv m/s Volumetric absolute velocityv0 m/s Intake velocityx, y, z mm CARTESIAN co-ordinates

Abbreviationst.v. turning vanes.s. suction sidet.e. trailing edge

TEST FACILITY AND INSTRUMENTATIONTest Facility

A large-scale coolant multipass-system rotatingspanwise was investigated during the experiments. Airis supplied in the rig through a rotary sealing assem-bly, a second one is used for venting after the air haspassed through the test section. Both are mounted tothe end of the double hollow shaft. A schematic of thetest rig is given in Figure 1. More detailed informa-tion about the test facility are presented in [RATHJENet al, 1999]. Reynolds number and rotation numbercan be adjusted appropriately to the models inserted.

Test Model GeometryFigure 2 shows a schematic draw of the model.

The test section consists of a leading edge duct (firstpass) with trapezoidal cross-section extending radiallyoutward, a 180° bend with a 90° turning vane and asecond pass with different trapezoidal cross-sectionextending radially inward. The ratio between the hy-draulic diameter of the second pass and the hydraulicdiameter of the first pass is 1.5. The length of eachpass is 9.975 dh, where dh denotes the hydraulic di-ameter of the first passage.

In the bend region the flow is not only directedthrough the 180° u-bend but also perpendicular to thatfollowing the turbine airfoil shape. The test section is12 dh long. The distance between the tip plate and thedivider wall is about 1.875 dh.

The turning vane is mounted in the circumferentialmid-plane of the bend and is 0.12 dh thick. The dividerwall has a thickness of 0.3 dh.

The model has a small plenum chamber and aninlet and outlet length, both are 3 dh long.

Measuring Techniques

L2F – Laser-Two-Focus VelocimetryFor the non-intrusive measurement of flow veloci-

ties, angles and fluctuations inside the rotating ductthe Laser-2-Focus (L2F) technique [BEVERSDORFF,HEIN, SCHODL, 1992] has been applied using a newlydeveloped optical unit to direct the laser beams intothe rotating frame of reference. The optical devicemounted in front of the rotor enables the system tostationary non-triggered signal processing. Figure 3

3

shows a schematic drawing of the optical set-up usedin this kind of application. Following the beam pathfrom the laser to the probe volume, the laser light isemitted by an Ar+ ion laser of 4 Watt power outputwhich is connected to the launching unit mounted tothe test rig by an optical fiber for vibration decouplingpurpose. In this unit, the laser beam is split into twobeams. The rotation prism co-rotating with half thespeed of the rotor focuses the two beams in the rotat-ing frame of reference, saving invariable orientation ofthe light barrier in relation to the rotating duct. Fromthe turning axis the beams are guided by several mir-rors to the measuring point inside the duct finallypassing a plane optical window in the duct wall. Ad-justment of the mirrors enables flow measurements atdifferent lateral and axial duct locations. The twobeams are highly focused in two foci (measuring vol-ume) by means of a traversable focus lens mounted tothe rotor. Traversing of the lens, which is also possibleduring rotation, moves the probe volume within thetest section along a line (Figure 4). The L2F methodis based on time of flight measurements of co-flowingparticles. Small oil droplets with a size less than0.5 µm were added to the flow before entering theshaft. The flow-following behavior of these particleseven at high accelerations has been successfully dem-onstrated and thereby a correlation between particleand flow velocity is assured [SCHODL, 1977]. Thepassing of a particle conveyed by the flow through thelight barrier (i.e. the 2 foci) produces two consecutivelight pulses. The time interval between the signals andthe known distance of the foci give the flow velocity.The light pulses received by the same optic in backscattering in the outer area - the inner part is used forlaunching the laser beams - are imaged at two photomultipliers.

PIV – Particle Image VelocimetryNow, the flow field is measured first in steady

mode by means of PIV. This modern technique is ascientific tool for qualitative and quantitative investi-gations of flow fields. The principle of the PIV withlaser, light sheet optic, laser light sheet and test sectionis shown in Figure 5 and the existing test set-up isshown in Figure 6. PIV is based on the principle tocapture the image of the flow field two times (pulsedelay in the range of 5 µs). In the flow there are tracerparticles (aerosol) with an average size of 0.1 µm.These particles reflect in the laser light sheet, which isformed with a light sheet optic consisting of a systemof one spherical and two cylindrical lenses. It gener-ates a light sheet with a thickness of 1 mm and a di-vergent angle of 24°. The laser light sheet is adjustableand visualizes the front channel as well as the back

channel, due to a mirror. Illumination was provided bya standard, frequency-doubled, double-cavityNd:YAG laser (NewWave, Gemini PIV) with a pulseenergy of up to 120 mJ per pulse at 532 nm. On therecording side, a thermo-electrically cooled, interlinetransfer CCD camera (PCO, 1280 x 1024 pixel reso-lution) with a f = 55 mm, f # 2.8 lens (Nikon) wasused. A bandpass filter with center frequency 532 nmand width 5 nm (FWHM) placed in front of the lensrejected most of the unwanted radiation. The pulsedelay between the laser pulses varied between τ = 6 µs(investigation of mean flow) and τ = 12 µs (investiga-tion of secondary flow). The high resolution CCD1

camera takes two pictures, depending on the pulsinglaser. Each picture has a size of 2.5 MB. The CCDcamera, perpendicularly positioned to the light sheet,is a so-called cross correlation camera having double-frame single-exposure evaluation. Now it is possibleto calculate the value, direction and orientation of theabsolute velocities with the Cross Correlation Func-tion (CCF). 50 pictures at each camera position aretaken to calculate the mean value of the absolute ve-locity. For the secondary flow investigations (cut 1...6)into the first pass from the multi-pass coolant channelone camera positions is adjusted (Figure 7a) and forthe investigations of the mean flow (cut 1 & 2) intothe second pass of the multi-pass coolant channel fivecamera positions are necessary (Figure 7b).

OFV – Oil Flow Visualization (Wall Flow)To analyze the flow phenomena, several kinds of

flow visualization techniques have been applied. Us-ing the oil flow visualization method, wall streamlinepatterns could be achieved within the duct. It is possi-ble to show and localize the boundary layer separa-tion, the reverse flow and vortex regions next to thesurface. To some extent it is possible to draw conclu-sions from the two-dimensional method for applica-tion to the strongly three-dimensional character of theflow field. The duct wall has been previously paintedwith a mixture consisting of oil of appropriate viscos-ity, oil acid and for contrast purpose Titanium dioxide(TiO2) and has been then exposed to the flow for atime period long enough to dry the oil paint. The ductis dividable in two halves and is opened after the testrun for viewing. A detailed description of the range ofapplication is given in [MALTBY, KEATING, 1962].

LFV – Laser Light Sheet Flow Visualization (Flow)To obtain more information about the flow around

the turning vane and the separation zone, a camerasystem has been installed to the Perspex duct observ-

1 Charge Coupled Device

4

ing inside the second pass with the mid plane lightedwith a laser light sheet. In order to visualize the flowstructures, seeding of the flow with oil steam is pro-vided by an oil fog generator. With frame grabbing ona PC-machine the video pictures were captured. Thedigitized pictures have been improved with regard toluminance and brightness.

RESULTSIn the following, some exemplary experimental re-

sults are presented. All results are presented in theform of isolines and vectors. For a better understand-ing of the isolines, red areas represent high velocities,blue areas represent areas with lower velocities. Theshown vectors all have a uniform size to representonly the orientation of the flow field. With regard toPIV processing, standard cross correlation algorithmswere used with an interrogation window size of32 x 32 pixel and a 50 % overlap. The velocity of theintake flow into the first pass amounts tov0 = 15.25 m/s, and consequently the REYNOLDSnumber is about 25.000 for the PIV measurements.

The comparison of the axial velocity component atcut 1 and 2 between L2F and PIV measurement in thesecond pass at the position z = 193 mm is presented inFigure 8. The axial velocity component at the z-position is extracted off the measured PIV results andshows relatively good agreement with the L2F results.The position of the jet and the wake is slightly differ-ent, which could be explained by deviations in modelmanufacturing. Measurements in the region of theseparation bubble were not possible with the L2Fsystem due to too high turbulence.

The measured flow field of the secondary flowwithin the first pass in six cuts (z = 190...240 mm, stepsize 10 mm) for a REYNOLDS number of 25.000 areshown in Figure 9. At a distance of 190 mm from theduct entry (Figure 9a, cut 1) a relatively big andstrong vortex occurs (x = 16.0 mm, y = 13.0 mm). Asmaller vortex exists at the channel too (x = 8.0 mm,y = 5.0 mm). An interpretation of that phenomenon is,that the flow is disturbed due to the inlet distortions,which are still visible at this position. The same resultwas found during CFD simulation where the completeinlet geometry with settling chamber was modeled(Figure 12) at cut 3. The mean value of the secondaryflow velocity is in the range of about 1 m/s. At theposition of cut 2 (Figure 9b, z = 200 mm) the bigvortex moved closer to the side wall (x = 17.5 mm,y = 11.0 mm) and the smaller vortex is already disap-pearing. The mean value of the secondary flow veloc-ity is in the same range like before. If the position of

the light sheet is varied to the positions of cut 3, 4, 5and 6 (Figure 9c, d, e and f) all vortices are disap-peared. The mean value of the secondary flow velocityis increasing more and more from cut to cut and isfinally reaching a value of 13 m/s at cut 6 (Figure 9f)and is becoming now the main stream velocity com-ponent due to the flow turning process. At cut 6 at theleading edge the beginning of the recirculation fromthe upper passage corner can be observed.

The measured flow field of the main flow withinthe second pass in two cuts 2 mm (right, wall section)and 10 mm (left, mid-span) for a REYNOLDS numberof 25.000 are shown in Figure 10. At the mid-spansection (left) a relative stationary separation bubbleoccurs at the upper surface (s.s.) of the t.v.. Behind theseparation bubble exists a relative high flow deviationwith high velocities about 24 m/s close to the wall. Another separation bubble with reverse flow is visible onthe dividing wall. Due to that flow recirculation a jetflow with velocities about 24 m/s exists on the pres-sure side at the t.e. of the t.v.. The thickness of theseparation bubble is decreasing in downstream direc-tion. The form and extension of the separation bubblewith a length of approximately 5 dh is similar to thosefound by visualization (Figure 11). The areas of highvelocity are coinciding further downstream(z = 190 mm) where mixing of both jets leads to anuniform velocity profile. The mean value of the mainflow velocity further downstream is in the range ofabout 13...16 m/s.

At the wall section (right) the mean value of thevelocity is increased to 25 m/s on the upper side of thet.v.. Turning of the flow leads to an impingementeffect to the leading wall of the second leg due to theinclination of the center-lines of both cross-sections.The separation bubble on the dividing wall is muchlarger in their extension than before. At the s.s. of thet.v. appears a boundary layer separation (x = 20 mm,z = 215...220 mm) due to side wall effects. The jetflow at the pressure side of the t.v. is still existing. Butthe mean value of the main flow velocity distributionis decelerated further downstream to a range of about8...10 m/s.

Figure 11 shows the streamline patterns on thesuction side (left) and the pressure side (right) of theduct wall, obtained from oil flow visualization ex-periments, in comparison with a numerical solutionclose to the wall. The streamlines on both walls showthe effect of the 180° turn, the influence of the turningvane and, clearly the extension and thickness of theseparation bubble at the dividing wall.

One capture from the video of the laser light sheetvisualization is shown in Figure 13. Also this tech-

5

nique confirms the already recognized flow phenom-ena.

SUMMARYThis paper describes experimental investigations

of a typical two-pass cooling channel system of aturbine blade with respect to fluid flow phenomenaaffected by geometry effects and flow turning sup-ported by a turning vane. A complex flow situation ispresent showing several kind of separation and vor-tices:(i) At the upper surface of the t.v. due to negative

incidence off the incoming flow,(ii) further downstream at the s.s. of the t.v. due to

too strong curvature,(iii) at the dividing wall due to the sharp edge where

the flow is incapable to follow.The jet region under the t.v. and the wake from the

t.v. are leading to high velocity gradients and corre-sponding high shear stresses at the contact layer pro-ducing very high turbulence levels there. Higher tur-bulence levels than 30 % leads to uncertainties in L2Fresults. With supposed levels of 50 % and more thePIV technique has not a problems. But to give anaccurate value of the turbulence level you will needabout 2000 PIV pictures which need high storagecapacity.

The performed flow visualization and simulationare very helpful in that case to understand the interac-tion of all apparent effects. The presented experimentswere mainly intended to assess the feasibility of ap-plying PIV in a multi-pass cooling channel used forapplied cooling research. The results have shown thatit is possible to apply PIV in a multi-pass coolingchannel. The application of PIV is approved and willbe adopted to the rotating system in the near future.

REFERENCESADRIAN, R. J., 1991, Particle-Imaging Techniques for

Experimental Fluid Mechanics, Annual Review ofFluid Mechanics 23.

BEVERSDORFF, M., HEIN, O., SCHODL, R., 1992, AnL2F-Measurement Device with Image RotatorPrism for Flow Velocity Analysis in Rotating Cool-ant Channels. 80th Symp. on Heat Transfer andCooling in Gas Turbines, AGARD PEP, Antalya,Turkey.

ELFERT, M., HOEVEL, H., TOWFIGHI, K., 1996, TheInfluence of Rotation and Buoyancy on RadiallyInward and Outward Directed Flow in a RotatingCircular Coolant Channel, Proc. 20th ICAS Confer-ence, Sorrento, Italy, 2490-2500

MALTBY, R. L., KEATING, R. F. A., 1962, The Sur-face Oil Flow Technique for Use in Low SpeedWind Tunnels, AGARDograph 70.

RATHJEN, L., HENNECKE, D.K., ELFERT, M., BOCK,S., HENRICH, E., 1999, Investigation of Fluid Flow,Heat Transfer and Pressure loss in a Rotating Multi-Pass Coolant Channel with an Engine Near Geome-try, ISABE-Conference, Florence, ISABE- IS-216

RAFFEL, M., WILLERT, C., KOMPENHANS, J., 1998,Particle Image Velocimetry, Springer Verlag BerlinHeidelberg.

RÖHLE I. 2000, “Doppler global velocimetry”, inRTO Lecture Series 217, Planar Optical Measure-ment Methods for Gas Turbine Components,Neuilly-Sur-Seine Cedex, France.

SCHODL, R., 1977, Entwicklung des Laser-Zwei-Focus-Verfahrens für die berührungslose Messungder Strömungsvektoren, Dissertation, Tech. Univ.Aachen, Germany.

WILLERT, C., GHARIB, M., 1991, Digital ParticleImage Velocimetry, Experiments in Fluids, No.10,Springer Verlag.

6

FIGURES

Inlet Outlet

9.97

5dh

1.87

5dh

0.15dh

12d h

0.3dh

0.12dh1.125dh

3dh

x/dh = 12.975

x/dh = 22.95x/dh = 0

25°

20°

35°

30° axis o

1.5dh

0.3dh

dh = hydraulic diameter o

3∗

4

2

2∗

31

1

Figure 1. Schematics of the test rig at DLR-Cologne Figure 2. Schematic of test model geometry

Figure 3. Optical set-up for the L2F-measurements Figure 4. Measuring lines for the L2F-measur

Figure 5. Principle of Particle-Image Velocimetry Figure 6. Test set-up for stationary PIV-meas

LLiigghhtt sshheeeett ooppttiiccLLaasseerr

AAccrryylliicc

CCCCDD--ccaammeerraa

F

light fiber

n/2 n

mirror

mirror

laser

laser beam seeding

test sectionmultipass-system

inlet outlet

window

focus lens

derotationprismL2F headdetector

Laser

Mirror

Laser light sheet

Flow directionImage plane

Imaging optic

Light sheet optic

Tracer particle into light sheet

Particle image from first light pulse at t0

Particle image from second light pulse at t1

Leading wall

f

f

4∗

∗

u

F

10°

rotationTrailing wall

the first pass

ements

rements

mmooddeell

llooww mmeetteerr

F

a

F

a

7

igure 7. Camera and cut positions. a. Secondary flow invesgation into second pass (cut 1 & 2)

x in mm

vin

m/s

10 15 20 25 300

5

10

15

20

25

30

Wall section, PIVWall section, L2FWall section, L2F, corr.

PIV

L2F

b

igure 8. Comparison of the axial velocity component betwposition z = 193 mm. a cut 1; b cut 2; (according F

b

tigations into first pass (cut 1...6); b. Mean flow investi-

x in mm

vin

m/s

10 15 20 25 300

5

10

15

20

25

30

Mid span, PIVMid span, L2FMid span, L2F, corr.

PIV

L2F

een L2F and PIV measurement in the second pass at theigure 7b)

zin

mm

00

5

10

15

20

25

30 ã

yz

inm

m

00

5

10

15

20

25

30

yz

inm

m

00

5

10

15

20

25

30

y

30 ãa

e

c

b

Figure 9. Secondary flow

`ìí=NI=ò=Z=NVM=ãCut 1, z = 190 mm

x in mm5 10 15 20 25

x in mm5 10 15 20 25

ã

ã

x in mm

zin

mm

0 5 10 15 20 250

5

10

15

20

25

y

x in mm

zin

mm

0 5 10 15 20 250

5

10

15

20

25

30 ã

y

30ãf

d

v

`ìí=RI=ò=Z=OPM=ãCut 5, z = 230 mm

8

i5 10 15 20 25

zin

mm

00

5

10

15

20

25

y

elocity distribution in six cross cuts of the f

`ìí=SI=ò=Z=OQM=ãCut 6, z = 240 mm

`ìí=PI=ò=Z=ONM=ãCut 3, z = 210 mm

`ìí=QI=ò=Z=OOM=ãCut 4, z = 220 mm

`ìí=OI=ò=Z=OMM=ãCut 2, z = 200 mm

5 10 15 20 25

irst pass measured with PIV (Re=25.000)

9

Figure 10. Mean flow velocity distribution in two longitudinal cuts of the second pass measured with PIV(Re=25.000)

x in mm5 10 15 20 25 30 35

50

55

60

65

70

75

80

85

x in mm5 10 15 20 25 30 35

50

55

60

65

70

75

80

85

90

95

100

105

110

115

120

125

130

135

140

145

150

155

160

165

170

175

180

185

190

195

200

205

90

95

100

105

110

115

120

125

130

135

140

145

150

155

160

165

170

175

180

185

190

195

200

205

Mid-spany = 10 mm

Wall sectiony = 2 mm

Pos 1

Pos 2

Pos 3

Pos 4

Pos 5

zin

mm

210

215

220

225

230

235

240

245

v in m/s: 2 4 6 8 10 12 14 16 18 20 22 24

zin

mm

210

215

220

225

230

235

240

245

v in m/s: 2 4 6 8 10 12 14 16 18 20 22 24

10

Pressure side Suction side

Figure 11. Comparison between calculated and visualized wall velocities and streamlines on pressure side (left) andsuction side (right), non-rotating (Re=50.000)

Figure 12. Numerical simulation at cut 3 in first pass Figure 13. Pattern of laser light sheet visualization (bend)