flow and thermal performance of a gas turbine nozzle guide vane

FLOW AND THERMAL PERFORMANCE OF A GAS TURBINE NOZZLE GUIDE VANEWITH A LEADING EDGE FILLETStephen LynchMechanical Engineering, Virginia TechAdvisor: Karen A. TholeAbstractComplex three-dimensional vortex flows developat the junction of a gas turbine airfoil and its casing(endwall). These flows increase the transfer of heatfrom the combustion gases to the metal parts andcontribute to reduced aerodynamic efficiency. Paststudies have shown that the use of a large fillet at theairfoil-endwall junction can reduce or eliminate theendwall vortex flow pattern. To determine the effectsof a fillet on turbine surface conditions, wall shearstress and heat transfer coefficients were measured onthe endwall of a low-speed linear turbine vane cascade.A shear stress measurement technique was developedand implemented for this study. High-resolutionmeasurements of magnitude and direction providedquantitative information about the endwall flowfeatures with and without a large fillet at the airfoilendwalljunction. The fillet was shown to increaseshear stress magnitude, but reduce turning of the flow atthe endwall. Heat transfer coefficients were alsomeasured with and without the fillet. Results indicatedthat the leading edge fillet changed the distribution ofheat transfer on the vane endwall.IntroductionA major factor in reduced aerodynamic efficiencyand increased part temperatures in a gas turbine engineis a complex three-dimensional flow, known assecondary flow. This flow occurs where an axialturbine airfoil meets the inner or outer casing, known asthe endwall (see Figure 1). In an axial gas turbine,secondary flows result in aerodynamic losses in a vaneor blade stage, which can reduce the engine’s overallefficiency by up to 3%. Secondary flows also tend toconvect hot mainstream gases onto the endwall, whichresult in higher local heat transfer coefficients andincreased metal temperatures. A 25°C (50°F) increasein metal temperature can result in a reduction in partlife by a factor of two.Controlling or eliminating secondary flow couldincrease the aerodynamic efficiency of the engine andreduce the required cooling, or the same amount ofcooling could increase part life expectancy. Whileseveral options to eliminate secondary flow have beensuccessfully investigated, a fundamental understandingof the flow remains elusive. More information aboutthe flow and its interaction with the turbine surfaces isnecessary to advance current engine designs. Thispaper will discuss the influence of a large fillet at thejunction of the endwall and the airfoil, on the surfaceheat transfer and wall shear stress of a modern nozzleguide vane.Literature ReviewSecondary flow is not unique to gas turbines andhas been investigated for symmetric airfoils andcylinders in crossflow. However, a couple of featuresof a turbine cascade are unique. First, the airfoils arenot symmetric and turn the flow through large angles.This results in cross-passage pressure gradients that arenot present for symmetric airfoils or cylinders. Second,the flow is accelerated in the turbine passage by astreamwise favorable pressure gradient. These factors,as well as the three-dimensionality of secondary flow,make it difficult to predict secondary flow developmentand progression.Langston et al. 1 presented one of the firstdescriptions of endwall secondary flow in a gas turbine.Other researchers have presented secondary flowmodels with slight differences, but they all agree on themajor components. The incoming boundary layer onthe endwall has a constant static pressure profile in thespanwise (normal to wall, along span of airfoil)direction. However, the boundary layer has a nonuniformvelocity profile because of the difference invelocity between fluid entrained near the wall and fluidin the mainstream, which corresponds to a non-uniformtotal pressure profile. As the boundary layer stagnateson the airfoil, the total pressure profile becomes aspanwise pressure gradient that drives the flow to theendwall. This turning creates a vortex that splits at thestagnation point and wraps into two legs around theInlet boundarylayerEndwallPassagevortexCounterEndwall vortexcrossflowFigure 1. Secondary flow model presented byLangston 2 .Lynch 1

pressure and suction sides of the turbine airfoil—this isthe horseshoe vortex. The portion of the horseshoevortex on the pressure side, known as the passagevortex, is augmented further by the inherent pressuregradient between airfoils (i.e., the pressure side of oneairfoil faces the suction side of its neighbor). Theportion of the horseshoe vortex that passes to thesuction side, known as the counter vortex, has anopposite sense of rotation to the passage vortex, andtends to orbit the passage vortex as it interacts with thepassage vortex downstream of the stagnation point(refer to Figure 1).Past research has shown that modifications to theleading edge of a gas turbine vane can reduce oreliminate some of the features of the secondary flow.Sauer et al. 3 used an asymmetric leading edge bulb tointensify the counter vortex, which resulted in a 50%reduction in aerodynamic losses at the exit of the vanepassage. Zess and Thole 4 used computational fluiddynamics (CFD) to design an asymmetric leading edgefillet, which they later experimentally tested. Theirresearch indicated elimination of the leading edgehorseshoe vortex and an order of magnitude reductionin turbulent kinetic energy levels associated with vortexdevelopment. Becz et al. 5 studied two bulb designs aswell as an asymmetric elliptical fillet and found thatonly the fillet reduced overall total pressure loss. It alsoslightly reduced airflow turning, which agrees with thewall shear stress results of this study presented later.Other studies have considered the effects ofleading-edge modifications on the thermal environmentof the vane or blade. Shih and Lin 6 performedcomputational studies on two fillet designs with andwithout swirl in the incoming flow profile. Their studyindicated the largest reduction of heat transfer on theairfoil and endwall and the least overall aerodynamicpenalty occurred with no fillet and inlet swirl.Lethander et al. 7 successfully integrated optimizationsoftware with a commercial CFD solver to design arelatively large fillet (with dimensions at the limits oftheir design space) that reduced surface temperatures onthe endwall. Han and Goldstein 8 used naphthalenesublimation to infer heat transfer distributions on theendwall through the heat and mass transfer analogy.Their linear asymmetric fillet, based on the optimaldesign by Zess and Thole, reduced the horseshoe vortexeffects but resulted in increased heat transfer near theleading edge endwall-airfoil junction due to intensifiedcorner vortices. Mahmood et al. 9 performed smokeflow visualization, took total pressure measurements,and obtained Nusselt number distributions on theendwall for four fillet geometries. Their resultsindicated a reduction in the leading edge vortex sizeand lower endwall heat transfer coefficients for all filletgeometries, with a concave elliptical geometry showingthe largest reduction in heat transfer.It is apparent from past studies that a modificationto the airfoil-endwall junction has some positive effectin optimizing the aerodynamic and heat transferenvironment of a gas turbine vane. Fillets have shownthe most promise; thus, this study focused onunderstanding how that geometry influences the flow.To date, none of the studies have presentedmeasurements of heat transfer on the fillet surfaceitself. This surface would be of interest to turbinedesigners, since it would require additional cooling.Furthermore, wall shear stress has not beenexperimentally measured for vanes with airfoil-endwallmodifications. Wall shear stress may be a contributorto total pressure loss in a cascade. Measurements mayprovide further insight into the secondary flowreduction mechanisms of a fillet.Experimental DesignA large closed-loop, low-speed wind tunnel wasused to perform heat transfer and wall shear stressmeasurements on a scaled-up nozzle guide vane testsection. The flow is driven by a 50-hp axial fan, andpasses through a primary heat exchanger to modulatethe overall flow temperature. It enters a splitter section,where flow can be diverted for vane coolingexperiments (not used in this study). The primary flowpasses through several screens and a contractionsection, which can be used to simulate a gas turbineannular combustor. The flow enters the corner testsection, which contains the vanes for this study, andthen passes back into the fan.The corner test section holds two full nozzle guidevanes and a third partial vane connected to a flexiblewall. The sides of the test section contain bleeds toremove the sidewall boundary layers and ensure that theflow around the vanes is periodic. The vane design is athree-dimensional extrusion of the two-dimensionalmidspan airfoil geometry from the Pratt & WhitneyPW6000 turbine engine. The vane is scaled up by afactor of nine for high measurement resolution, and isinstrumented with static pressure taps at 40% of thespan (measured from the bottom endwall) to determinethe static pressure coefficient. A description of theTable 1 PW6000 nozzle guide vane geometry and flowparameters.Actual chord length 6.60 cm (2.6”)Scaling factor 9Scaled chord length (C) 59.4 cm (23.4”)Pitch/chord 0.77Span/chord 0.93Axial chord (C ax )/chord 0.48Inlet Reynolds number 2.2x10 5Flow inlet angle 0°Flow exit angle 78°Lynch 2

turbine vane parameters is given in Table 1.The bottom endwall surface of the corner testsection was manufactured out of a 2.54 cm (1”) thicksheet of Last-A-Foam FR-6706 low-density closed-cellpolyurethane foam, which has a very low thermalconductivity (0.0287 W/m-K). This material was usedto minimize conduction losses for the heat transferstudies. The endwall was instrumented with type-Ethermocouples in various locations throughout the vanepassages, for calibration of the infrared images.The top endwall was made of acrylic, and hadthirteen image ports spaced around the test section. Thelayout of the image ports allowed overlap of infraredimages to create a full endwall temperature map. Eachimage port was sealed with a removable acrylic coverwhen an image was not being taken.Heat Transfer MeasurementsHeatexchangerAxial fanCorner testsection with vanesFigure 2. Schematic of the low-speed recirculatingwind tunnel used in this study.Endwall heat transfer measurements were taken ona constant heat flux plate attached to the endwall andfillet surfaces. The heaters, manufactured byElectrofilm Manufacturing, consisted of a 37 µmcopper layer on top of a 75 µm thick kapton layer, inwhich 25 µm inconel elements were embedded in aserpentine pattern. The endwall heater had fourseparate circuits, three of which could be turned offwhen a fillet geometry was installed over the top ofthem, so as to not introduce error in the determinationof the fillet surface heat transfer. The heaters wereattached to the foam surfaces using double-sided tapeembedded in a very thin layer of silicone adhesive. E-type thermocouples embedded in the foam endwall andfillet surfaces were placed in contact with the bottomsurfaces of the heaters with Omegabond thermalcement.The heater top surfaces were painted with Krylonflat black paint, which has a nominal emissivity of 0.96,and enables good resolution of surface temperatureswith the infrared camera. Small crosses were etchedinto the paint on the endwall heater, for imagepositioning and thermocouple location determination.Three E-type ribbon thermocouples were attached tothe endwall heater top surface with Omegabond thermalcement. The ribbon thermocouples were used to checkthe calibration with the bottom-surface thermocouples.An infrared camera (Flir P20) was used to capturesurface temperatures on the heat flux plate. Infraredradiation emitted from the surface is converted totemperatures by the camera. Calibration of the imagedsurface temperatures was achieved by matchingacquired thermocouple measurements with the knownlocation of the thermocouple in an image. Ameasurement bias between the bottom-mountedthermocouple and the infrared top-surface measurementwas accounted for by a one-dimensional calculation ofthe thermal resistance of the heater. The bias of 0.5°Cwas slightly larger than the uncertainty of a type-Ethermocouple (±0.2°C).Infrared images were taken by sequentiallyremoving the inserts in the imaging locations on the topendwall and placing the camera (in a fixture) over theimaging location. Based on an uncertainty analysis, itwas determined that five images should be taken ateach location and averaged, where each image is alsoan average of 16 frames taken by the camera. Theviewing area of the camera on the heat flux surface was16.6 cm by 22.3 cm (6.5” x 8.78”), which it digitizedonto 240 by 320 pixels. Flir’s ThermaCam Researchersoftware was used to process the images and calibratethem. An in-house routine was developed to assemblethe individual images into a single map of the entireendwall.A particular goal of this study was to analyze theheat transfer on the fillet surface. However, imagingthe fillet is complicated by its three-dimensional nature.Perspective distortion is introduced, in which regions ofthe fillet closer to the camera appear larger than regionsfarther away. To correct this, crosses were etched intoFigure 3. Infrared images of the heat flux surfaceprovide surface temperatures. The images arecalibrated and assembled into an endwall map.Lynch 3

Figure 4. An infrared image of the foil grid used todetermine spatial transformation control points. The basis of oil film interferometry (OFI) is thebehavior of a thin oil film under the influence of shear.black paint on a stainless steel foil grid. The fillet wasMovement of the oil film can be dictated by shear on itswrapped with the foil grid and imaged with the infraredsurface, pressure gradients, gravity, and surface tension.camera. The grid provided control points for spatialIn many cases, such as for this study, shear is thetransformation of the fillet surface in the image. Thedominant force, by several orders of magnitude. Thecamera location was fixed so that subsequent imagesuse of light ray interferometry to measure the height ofwould be at the same location and orientation. Error inthe oil film as it is thinned by airflow can be used withsubsequent camera positioning was less than 4 pixelsa thin-oil film reduction of the Navier-Stokes equations(2.8 mm at the nominal focal distance of the endwall).for the oil, to obtain the shear acting on the oil.An in-house MATLAB code was developed toDerivation of the thin oil film equation consists ofperform the spatial transformations. Thea mass and momentum balance on a differential controltransformations involved some distortion of the imagevolume through which oil convects. A mass balance onoutside of the control points, so only the portion of thethe control volume gives the height of the oil and thefillet surface within the grid was retained from theaverage convective velocity as functions of time andtransformed images. Three-dimensional spatialspace. The x- and z-momentum fluid equations can betransformation enabled a realistic look at the fullsimplified by performing an order of magnitudesurface heat transfer. The capability also existed toanalysis to determine that the Reynolds number for theperform two-dimensional transformations by projectingoil film is much less than one. The inertial terms andthe fillet surface to the endwall.streamwise viscous terms can then be neglected, whichThe input heat to the endwall and fillet surfacessimplifies the x- and z-momentum equations so thatwas calculated by measuring the voltage across a heatershear forces are balanced by pressure gradients andcircuit, as well as the voltage across a precision resistorgravity. The momentum equations can be solved for(1Ω) in series with the circuit, which gave the current.the u and w oil velocities in terms of the pressureThe total power was divided by the area of the circuit togradient, gravity, and airflow shear by applying no-slipobtain the input heat flux. This flux was corrected forboundary conditions at the wall-oil interface, and theconduction and radiation losses, which accounted for amaximum of 0.2% and 21% of the input power,respectively. No correction due to conduction to thevane itself was performed, since the vane was alsoτw , x ∂Uconstructed of low-density closed-cell foam. TheU +remaining convective flux was used with the measured∂ xsurface temperatures to calculate heat transfercoefficients, which were normalized in the form of a∂hStanton number based on inlet mainstream velocity hdx h + ∂ xhSt =(1)ρC UU =1hyu dyp∞where h is the heat transfer coefficient (relatingconvective heat flux to temperature gradients betweenthe wall and the air), ρ is the density of air, C p is thespecific heat of air, and U ∞ is the freestream inletvelocity.Theory of Oil Film InterferometryA measurement technique for wall shear stress hadnot yet been developed in our laboratory. A methodknown as oil film interferometry was selected based onits robustness and simplicity. Holley and Langstonpresented oil film interferometry results for their lowspeedscaled-up turbine blade cascade 10 . They wereable to determine the unique features of endwallsecondary flow (saddle point, separation line).Naughton and Sheplak 11 present a very good descriptionof the fundamentals and implementation considerationsof the oil film interferometry technique.h∫0zFigure 5. Oil film control volume with x-directioncomponents shown.xLynch 4

desired shear at the air-oil interface. These velocitiesare related to the average convective velocities byintegrating over the height of the oil. Combining thevelocity distribution from momentum with the massbalance for the oil yields the thin oil-film equation2 3∂h∂ ⎛ τ⎞w,xh⎜h ⎧∂P⎫+ −⎟⎨ − ρgx ⎬∂t∂x2 3⎝ µ µ ⎩ ∂x⎭⎠(2)2⎛3∂ τ⎞w,zh⎜h ⎧∂P⎫+ −⎟ = 0⎨ − ρgz ⎬∂z2 3⎝ µ µ ⎩ ∂z⎭⎠For representative values pertinent to this study (h = 1µm, oil viscosity ν = 100 cSt, shear stress τ w = 10 Pa, oildensity ρ = 1000 kg/m 3 , dP/dx = 100 Pa/m, g x = 10m/s 2 ), order-of-magnitude analysis on the terms in theparentheses in Eq. (2) shows that the shear stress termis at least two orders of magnitude larger than the otherterms, which are then neglected. Knowledge of theheight of the oil as a function of space and time leads tothe wall shear that acted on the oil. For spatiallyconstant viscosity and shear stress (reasonableassumptions in this study because of the large scale andsmall measurement sizes), the reduced oil film equationcan be solved by separation of variables for the heightof the oil. Such a solution requires measurement of theoil height only at the end of the wind tunnel run, sincethe conditions leading to the final oil film thickness areintegrated over time.Fizeau interferometry provides a means ofmeasuring the height of the oil. Light strikes thesurface of the oil film and is reflected and refracted.The refracted rays are then reflected by the solidboundary at the lower surface of the oil and pass backout of the oil. The phase difference between theinitially reflected and refracted rays will attenuate oraugment the rays, creating interference bands. Thephase difference is related to the height of the oil byλφ ⎛h = ⎜4π⎜⎝n2oil1− n2airsin2θi⎞⎟⎟⎠(3)where λ is the light wavelength, φ is the phasedifference, θ i is the incidence angle of the light, and n oiland n air are the indices of refraction of the oil and air,respectively. Examination of the interferograms relatedthe phase difference to the physical spacing of thebands.The wall friction coefficient reported in this studyhas been normalized by the upstream dynamic pressureand is generally defined asτwCf=(4)1 2ρU2∞~500 nmLow-pressure sodium vaporlamp (monochromatic)θ Interference betweenilight raysAirflowh∆ x, φ = 2πDow Corning200 silicone oilNickel foil substrateFigure 6. Schematic of Fizeau interferometry techniqueon oil film.Implementation of Oil Film InterferometryThe reader is again referred to Naughton andSheplak 11 for an excellent description ofimplementation considerations for the oil filminterferometry method. The major componentsrequired for oil film interferometry are a light source, acamera to capture images, a properly reflective surface,and, most importantly, oil.It is desirable to have a single wavelength of light,to eliminate noise in an interferogram and allow forprecise determination of the spacing of the patterns. Alaser is an ideal source; however, other less-costlysources can work just as well. A low-pressure sodiumvapor lamp, recognizable as a faded yellow street lamp,emits light at two very closely spaced wavelengths(589.0 nm, 589.6 nm). For this study, a low-pressuresodium vapor lamp from ABS Lighting was used toilluminate the oil films.The interferograms are typically on the order of 1mm (0.040”) between bands, so high-resolutionimaging is necessary. A digital camera (Pentax *iSt DSdigital SLR) with a 75-300 mm zoom/macro lens wasFigure 7. A nickel foil patch on the endwall isilluminated in a fixture. Interferograms are analyzedfor shear stress direction and magnitude.Lynch 5

attached to a fixture. The fixture allowed formeasurement of the incident angle of light relative tothe oil film, and consistency in the positioning of thecamera.A highly reflective surface provides goodresolution of interference patterns without lowering thecontrast. Glass, Mylar, and polished steel have all beenfound to be good substrates, but polished aluminum isnot. Nickel and other foils can also work, especiallysince they can be moved to various regions of a largemodel without having to plate the entire model. Forthis study, 0.05 mm (0.002”) thick nickel foil with 14%tungsten (by weight) was cut into 7.6 cm by 3.81 cm(3” x 1.5”) rectangular patches, and adhered to thepainted surface of the heater with a sprayed-on layer ofrubber cement. To ensure that the leading edge of thefoil did not interfere with the boundary layer, severaltests were performed in which the oil film was in thesame geometrical location, but the distance of the filmfrom the leading edge of the foil was varied. In allcases, the variation of measured skin friction was lessthan 9%, which is approximately the uncertainty of theoil film method.The choice of oil depends on the test conditionsand the degree of uncertainty tolerable in the results.Typically, silicone oils (polydimethylsiloxane,commercially known as Dow Corning 200 Fluid) areused because of the relative insensitivity of theirviscosity with temperature, their transparency, and theease of removal from the model. Silicone oils can betailored to a wide range of viscosities by lengtheningthe polymer chain. Although the viscosity is lesssensitive to temperature than petroleum oils, it can stillvary by 2% per °C, which makes determination of theoil viscosity one of the largest terms in the uncertaintyof the shear stress measurement. It is recommendedthat the viscosity of each batch of oil be measured toaccount for variations from the nominal viscosity.Naughton and Sheplak 11 provide a correlation thatrelates a reference measurement of oil viscosity over alarge range of temperatures. For this study, silicone oilin nominal viscosities of 100 cSt, 500 cSt, and 1000 cStwere used for shear stress measurements on the vanecascade endwall. Brian Holley performed referencemeasurements for each nominal viscosity, for thetemperature correction.The test procedure to obtain shear stressmeasurements consisted of several steps. First, a nickelfoil patch was cleaned and sprayed with the thin layerof rubber cement. It was affixed to the endwall in thedesired measurement location, which was recorded sothat the measurements could be converted to vanecoordinates. Next, the nickel surface was cleaned andlightly polished to remove any cleaning fluid film. Asilicone dropper was dipped into the oil and smalldroplets (

St ∞Figure 9. Installation of the linear fillet tested in thisstudy.determine which have the largest effect on reducingendwall heat transfer and aerodynamic loss. To thatend, a test matrix of possible fillet designs wasdeveloped. Based on the success of the large linearprofilefillet of Lethander et al. 7 , this fillet was chosenas a baseline for comparison with elliptical filletdesigns. Elliptical designs might be simpler tointegrally cast into a vane or blade shape withoutcreating stress concentration points where the filletmeets the endwall or airfoil. The fillet of Lethander etal. is suited for parameter studies, since the focus of thatwork was an optimization of the fillet design formaximum thermal benefit.Due to the limits of time and the complexities ofmeasuring heat transfer and shear stress on the nonconformablesurfaces of an elliptical fillet, results haveonly been obtained for the linear fillet. Future workwill include testing for other fillet configurations todetermine the dominant design parameters of a filletthat reduce endwall heat transfer and shear stress.Two complete fillets were machined out of thesame closed-cell foam as the endwall and instrumentedwith type-E thermocouples. One of the fillets was splitto go around the two side vanes. Kapton heaters, withthe same thickness and composition as the endwallheater, were attached to the fillet surfaces with doublesidedtape. The fillets were secured firmly to theendwall and airfoil, and sealed around their perimeterwith silicone.ResultsThe results for measurements of heat transfer onthe endwall, and on the fillet when installed, arepresented first. Shear stress measurements with andwithout the fillet are shown for the endwall only.Heat Transfer ResultsTo provide a baseline for comparison to the filletresults, measurements of endwall heat transfer for the(a)(b)Figure 10. Comparison of endwall heat transfermeasurements for this study (b), to Kang et al. 11 (a).unfilleted nozzle guide vane were taken. Kang et al. 11presented measurements of endwall heat transfer for thePW6000 vane geometry, which were compared to theresults from this experiment to determine that thebaseline case was displaying the correct trends. Figure10 shows contours of Stanton number for Kang et al.compared to this study. High contour values denotehigh heat transfer coefficients; heat can more easilymove into the endwall. Although there are some slightdifferences, particularly in the region directly upstreamof the blades and along the flexible wall connected tothe lower vane (x/C = 0.5, y/C = 0.25), these are mostlikely due to different flow conditions for this study.Specifically, the endwall heater plate had a shorterupstream length, and the inlet boundary layer had adifferent shape due to previous combustor simulatormodifications. It can be seen from the baseline resultsthat the horseshoe vortex near the stagnation region ofthe vane causes increased heat transfer upstream of thevane. Also, the passage vortex, sweeping from thepressure to suction sides of adjacent airfoils, causeshigher heat transfer along the suction side of the vane.Figure 11 shows a three-dimensional view of themeasured heat transfer on an endwall with a linearfillet, compared to measurements on the unfilletedendwall. It is not immediately obvious that the fillethas reduced endwall heat transfer, especially on thesuction side of the fillet. However, the shapes of thecontours have changed, which indicates that the fillet isinfluencing the secondary flow and its interaction withthe surface. Regions of low Stanton number are shiftedmore to the pressure side of the endwall with the linearfillet. Also, the large gradient in heat transfer at thestagnation region of the vane has been somewhatreduced. Overall, the surface areas of the endwall andthe near-wall airfoil region have decreased with theaddition of a fillet. Thus, even though the fillet doesnot dramatically reduce endwall heat transfer, it mayLynch 7

C f= 0.100(a)St∞=hρC Up∞C f= 0.050(b)Inlet flowdirectionFigure 12. Wall shear stress vectors show the saddlepoint (insert) and cross-passage flow.Figure 11. Comparison of endwall heat transfer with(b) and without (a) a large linear fillet.C f = 0.150reduce overall cooling requirements for the nozzleguide vane.Wall Shear Stress ResultsA vector plot of wall shear stress on the endwall ofthe unfilleted vane is presented in Figure 12. Thevectors displayed in the figure were uniformly sampledfrom approximately 400 data points taken over theentire endwall. Several features of the endwall shearstress due to secondary flow are apparent. Sheardecreases as flow approaches the vane and stagnates.Flow washes down the airfoil and moves away from thebase of the airfoil in the horseshoe vortex. The saddlepoint, where the flow diverges around the vane, iswhere the incoming boundary layer separates off of theendwall. Further in the passage, cross-passage flowsweeps from the pressure side of the lower airfoil to thesuction side on the adjacent airfoil. The turning at theendwall due to secondary flow is larger than for thebulk flow away from the wall, which smoothly followsthe airfoils.Figure 13 shows endwall shear stress vectorsoverlaid on the results from the unfilleted vane. In thestagnation region, it appears that the fillet is divertingBlue Linear filletRed No filletC f = 0.020Figure 13. Endwall shear stress with a linear fillet (bluevectors), with the unfilleted results overlaid (red vectors).Lynch 8

50Inviscid predictionwith no filletVane with linear filletVane with no filletcooling requirements, although this conclusion wouldrequire knowledge of the airfoil surface heat transferwith and without the fillet.C p-5-10-15-20-25-30-1 -0.5 0 0.5 1s/CFigure 14. Static pressure coefficient around the vane at40% span, showing the acceleration of the flow on thesuction side due to the fillet.flow around the vane farther upstream. This may meanthat the horseshoe vortex has not been eliminated, buthas been displaced away from the vane. In the passage,the endwall shear for the linear fillet case is larger thanthe unfilleted case, which is unfortunate since wallshear is thought to be a contributor to total pressure loss(aerodynamic efficiency reduction) in a gas turbine.However, the increase is not unexpected since the filletreduces the flow area of the nozzle guide vane passage,causing flow to accelerate more rapidly. The staticpressure coefficient around the vane, shown in Figure14, supports this argument. The more interesting resultfrom the linear fillet endwall shear is the turning of theendwall flow. The linear fillet begins to turn thepassage flow sooner than in the unfilleted case, but theoverall turning when the flow exits the vane passage isreduced. This may indicate a weaker passage vortexstructure. Flowfield measurements should be taken toconfirm this.ConclusionsA wall shear stress method was developed andbenchmarked for this study. High-resolutionmeasurements on the scaled-up nozzle guide vaneprovided quantitative information about the endwallsecondary flow features. When a fillet was added to theairfoil-endwall junction, the shear stress magnitude inthe passage increased, but the overturning due to thepassage vortex decreased, indicating possibly weakersecondary flow.The addition of a fillet was shown to change thelayout of the endwall surface heat transfer coefficients,shifting regions of low heat transfer to the pressure sideof the airfoil. The reduced surface area of the endwallairfoiljunction with a fillet may mean reduced partAcknowledgmentsThe authors gratefully acknowledge the support ofthe National Science Foundation’s GOALI (GrantOpportunities for Academic Liaisons with Industry)program, and the Virginia Space Grant Consortium.Dr. Lee Langston and Brian Holley at theUniversity of Connecticut provided silicone oil andnickel foil, and were instrumental in our developmentof the OFI method.Kaitlin Keim from Virginia Tech assisted in thedevelopment of the OFI method and performed thebenchmarking experiments.Industrial input was provided by Joel Wagner andPeter Tay at Pratt & Whitney Engines in Hartford,Connecticut.References1 Langston, L.S., Nice, M.L., Hooper, R.M, 1977,“Three-Dimensional Flow Within a Turbine CascadePassage,” ASME Journal of Engineering for Power,Vol. 102, pp. 21-28.2 Langston, L.S., 1980, “Crossflows in a TurbineCascade Passage,” ASME Journal of Engineering forPower, Vol. 102, pp. 866-874.3 Sauer. H., Muller, R., Vogeler, K., 2000, “Reductionof Secondary Flow Losses in Turbine Cascades byLeading Edge Modifications at the Endwall,” ASMEPaper 2000-GT-0473.4 Zess, G.A., Thole, K.A., 2002, “Computational Designand Experimental Evaluation of Using a Leading EdgeFillet on a Gas Turbine Vane,” ASME Journal ofTurbomachinery, Vol. 124, pp. 167-175.5 Becz, S., Majewski, M.S., Langston, L.S., 2004, “AnExperimental Investigation of Contoured LeadingEdges for Secondary Flow Loss Reduction,” ASMEPaper GT2004-53964.6 Shih, T.I-P., Lin, Y.-L., 2002, “Controlling Secondary-Flow Structure by Leading-Edge Airfoil Fillet and InletSwirl to Reduce Aerodynamic Loss and Surface HeatTransfer,” ASME Paper GT-2002-30529.7 Lethander, A.T., Thole, K.A., Zess, G.A., Wagner, J.,2003, “Optimizing the Vane-Endwall Junction toReduce Adiabatic Wall Temperatures in a TurbineVane Passage,” ASME Paper GT2003-38940.Lynch 9

8 Han, S., Goldstein, R.J., 2005, “Influence of BladeLeading Edge Geometry on Turbine Endwall Heat(Mass) Transfer,” ASME Paper GT2005-68590.9 Mahmood, G.I., Gustafson, R., Acharya, S., 2005,“Experimental Investigation of Flow Structure andNusselt Number in a Low-Speed Linear Blade PassageWith and Without Leading-Edge Fillets,” ASMEJournal of Heat Transfer, Vol. 127, pp. 499-512.10 Holley, B.M., Becz, S., Langston, L.S., 2005,“Measurement and Calculation of Turbine CascadeEndwall Pressure and Shear Stress,” ASME PaperGT2005-68256.11 Naughton, J.W., Sheplak, M., 2002, “ModernDevelopments in Shear-Stress Measurement,” Progressin Aerospace Sciences, Vol. 38, pp. 515-570.12 Kang, M.B., Kohli, A., Thole, K.A., 1999, “HeatTransfer and Flowfield Measurements in the LeadingEdge Region of a Stator Vane Endwall,” ASME Paper98-GT-173.(b)Lynch 10