flowfield measurements about an airfoil with...

42nd Aerospace Sciences Meeting and Exhibit AIAA-2004-0559Jan. 5-8, 2004, Reno, NV

1American Institute of Aeronautics and Astronautics

This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.

FLOWFIELD MEASUREMENTS ABOUT AN AIRFOIL WITHLEADING-EDGE ICE SHAPES

Andy P. Broeren,∗ Harold E. Addy, Jr.,† and Michael B. Bragg‡

University of Illinois at Urbana-Champaign, Urbana, Illinois, 61801NASA Glenn Research Center at Lewis Field, Cleveland, Ohio 44135

∗Research Scientist, Department of Aerospace Engineering, University of Illinois, Member AIAA.†Research Engineer, Icing Branch, NASA Glenn Research Center at Lewis Field, Member AIAA.‡Professor and Head, Department of Aerospace Engineering, University of Illinois, Associate Fellow AIAA.

ABSTRACT

Flowfield measurements were carried out on theupper surface of a GLC-305 airfoil configured withglaze and rime ice-shape simulations. The mean androot-mean-square fluctuation of the streamwise velocitywere measured using a split-hot-film probe at severalchordwise locations. These data were taken at threedifferent angles of attack preceding stall for each iced-airfoil configuration at Reynolds numbers of 3.5×106

and 6.0×106 with Mach numbers of 0.12 and 0.21. Thevelocity measurements confirmed the presence of alarge separation bubble downstream of the ice shapes.The separation bubbles for the glaze ice configurationwere much larger than those for the rime ice case,resulting from the differences in the ice horn geometry.Other than the differences in size, the integralboundary-layer characteristics were very similar.Changes in Reynolds number did not significantlyaffect the separation bubble characteristics. However, alarger Mach number did result in a slightly largerseparation bubble for the glaze ice case at α = 6 deg.The root-mean-square velocity distributions had peakvalues in the separated shear layer, downstream oftransition, that compared well with previous work.

NOMENCLATURE

c Airfoil chord lengthCl Lift coefficientCl,max Maximum lift coefficient, coincident with αstall

Cp Pressure coefficientM Freestream Mach numberRe Reynolds number based on chordu Time-averaged (mean) streamwise velocityurms Root-mean-square velocity fluctuationUe Boundary-layer edge velocityU∞ Freestream velocityx Chordwise distance along airfoily Distance normal to airfoil chord

ydiv Normal location of dividing stream lineysurf Normal location of airfoil surfaceα Airfoil angle of attackαstall Stalling angle of attack, coincident with Cl,max

δ Boundary-layer edge location (thickness)δ* Boundary-layer displacement thicknessθ Boundary-layer momentum thickness

INTRODUCTION

Large, leading-edge ice accretions on airfoils causesignificant performance effects such as reduced lift andincreased drag. The effects are well documented for anumber of ice shape families.1-3 Perhaps lessunderstood, however, are the complex flowfield detailsassociated with the ice accretion/airfoil geometry. Anunderstanding of the flowfield is important forcomputational modeling used to predict the importantperformance losses. Knowledge of the iced-airfoilflowfield is also critical in determining what geometricfeatures of the ice contribute to the performancedegradations and how these may differ for otherairfoils.

Typically, large glaze ice accretions on airfoils arecharacterized by “horns” that protrude some distanceoff the airfoil surface into the oncoming flow. Theseflowfields have been studied in previous work and abasic understanding of the time-averaged characteristicshas been developed. The ice horn causes the flow toseparate and for low angles of attack, the separatedshear layer reattaches some distance downstreamforming a separation bubble aft of the ice shape. It ispossible that the reattached boundary layer,downstream of the bubble may separate again upstreamof the airfoil trailing edge. In this case there are twoseparated flow regions of importance. For large glazeice shapes, the separation bubble has a large, globaleffect on the pressure distribution.3 These are referredto as “long bubbles” as originally described by Tani4

for clean airfoils. Rime ice shapes tend to differ from

American Institute of Aeronautics and Astronautics2

glaze ice shapes in that usually there are no “horns”protruding into the flow. However, flow separationdoes occur because of discontinuities between the iceshape and airfoil surface.3 Depending upon the size ofthe ice shape, these bubbles may have a small, localeffect on the pressure distribution and would thus be“short” using Tani’s definition.

This understanding of iced-airfoil separationbubbles has been developed over the last twenty years(or so) of icing research. The flowfield researchincluded simple experimental methods such as flowvisualization, detailed experimental methods such ashot-wire/film anemometry, or laser velocimetry andcomputational methods. Bragg, Khodadoust andSpring,5 performed split-hot-film anemometrymeasurements in the separation bubble flowfield abouta simulated glaze ice accretion on the leading edge of aNACA 0012 airfoil. Both upper and lower surfaceseparation bubble characteristics were measured. Theoverall results were consistent with the descriptiongiven in the preceding paragraph. These measurementswere performed at a low Reynolds number of 1.5×106

and a Mach number of 0.12. The results showed howthe bubble grew in size as the airfoil angle of attackincreased up to maximum lift. The largest bubblemeasured covered over 30% of the chord at angle ofattack one degree below stall. An important result ofthis work was that the time-averaged separation bubblecharacteristics compared favorably to laminarseparation bubbles that can form on uncontaminated, orclean, airfoils. Khodadoust6 performed laser-Dopplervelocimeter (LDV) measurements on a straightrectangular wing with a leading-edge ice accretionhaving the same geometry. The earlier split-film dataclosely agreed with the non-intrusive LDVmeasurements and CFD calculations. In addition to thequantitative measurements, Khodadoust used a surface-oil method to visualize separation bubble reattachmentfeatures. This work was extended to an iced-swept-wing configuration and employed both LDV andhelium bubble flow visualization.7,8 These data showedthat the leading-edge separation caused a strongspanwise vortex on the swept wing. This alsoconfirmed what had been observed in CFD calculations.

While the time-averaged characteristics of large iceshape induced separation bubbles are fairly well known,the unsteady characteristics are less certain. Bragg etal.5 performed some time-dependent measurements as apart of their study and report a low-frequencyoscillation in the separation bubble flowfield. Gurbackiand Bragg9 have considered unsteady characteristics inmore detail. A NACA 0012 airfoil model instrumentedwith high-frequency response pressure transducers(Kulites) was tested with both rime and glaze ice

simulations. The time-dependent pressure distributionsindicated large-scale flow fluctuations. Spectralanalysis of the data performed later indicated flowfrequencies similar to that measured by Bragg et al.5

A motivation for work in this area is to providedata for improvement of computational results. Severalstudies have noted that the complexity of the flowfieldand potential for large-scale unsteadiness may rendersome computational methods incapable of correctlymodeling the flows. For example, Dunn et al.10 andPan et al.11 both suggest that accurate prediction ofstalling angle and maximum lift for airfoils with largeice shapes requires unsteady, three-dimensionalmethods. Chung and Addy12 suggest that moreflowfield measurements should be made in order tobetter assess the predictive capabilities of current CFDmethods for iced-airfoil calculations.

The present investigation directly addresses thelatter suggestion. Time-averaged flowfield velocitymeasurements were carried out on the upper surface ofa GLC-305 airfoil with both a rime and glaze ice-shapesimulation. Performance measurements for this airfoilwith the simulated ice shapes were carried out inprevious experiments and thus the additional flowfieldinformation provides for a comprehensive data set.13 Akey conclusion from the previous testing was that largechanges in Reynolds number had very little effect onthe iced-airfoil performance. The objectives of thepresent investigation were to measure the time-averaged flowfield velocities on the iced-airfoil uppersurface at several angles of attack for two different iceshapes. Split-hot-film anemometry was used todetermine the streamwise velocity profiles at severalchordwise locations along the airfoil upper surface.The measurements were performed in a pressure tunnelat Reynolds numbers of 3.5×106 and 6.0×106 at Machnumbers of 0.12 and 0.21 to investigate these effects.

EXPERIMENTAL METHODS

All aerodynamic testing was carried out at NASALangley Research Center, using the Low-TurbulencePressure Tunnel (LTPT). The LTPT is a closed-returnwind-tunnel that is principally used for two-dimensional airfoil testing and is described in detail inreferences 14 and 15. It can be operated at stagnationpressures from near vacuum to 147 psia (except 15 to20 psia) and over a Mach number range of 0.05 to 0.40.A heat exchanger and nine turbulence reduction screensare located in the inlet settling chamber. Thecontraction ratio is 17.6:1 and the test sectiondimensions are 36-inches wide by 90-inches high by90-inches long. The tunnel was designed for two-


dimensional airfoil testing with model chord lengths upto 36-inches.14 The freestream turbulence intensitylevels were about 0.1% or less for the operatingconditions used in this investigation.15

The GLC-305 airfoil model had a 36-inch chord by36-inch span and was mounted horizontally across thewidth of the test section. The model was machinedaluminum and had removable leading-edge sections.The leading-edge sections allowed for the various ice-shape simulations to be attached to the airfoil model.Two ice-shape simulations were used in thisinvestigation and are shown in Fig. 1. The contour ofthese ice simulations was determined from smoothedcoordinates of an ice tracing. These simulations werebuilt from a laser-sintering rapid-prototyping methodand had a constant cross-section in the spanwisedirection. More details about the model, ice shapes andtunnel installation for aerodynamic performancemeasurements can be found in reference 13.

The flowfield mapping experiments required theaddition of a significant amount of equipment bothinside and outside of the test section. The existingboundary-layer measurement apparatus was adaptedfrom previous testing.16 The traversing mechanism wasmounted to the east wall of the test-section as shown inFig. 2. A motor equipped with optical encoderfeedback was used to control the vertical position of themeasurement probes. This vertical traverse wascantilevered from the sidewall turntable via a trackplate. The vertical positioning was computer controlledand provided a positioning accuracy of ±0.0003-inches.The track plate allowed the vertical traverse to bepositioned in the chordwise direction. An operatorremotely controlled the chordwise positioning based onfeedback from a potentiometer that was routinelycalibrated during the test. The chordwise position wasset to within ±0.003 x/c. For the majority of chordwiselocations the vertical traverses were performed in adirection normal to the chord line. There were somelocations, near the leading edge, where the surfacecurvature required a small tilt angle. The tilt angle wasalso controlled remotely and measured by visualinspection of a vernier scale on the traverse apparatus.These angles were small (less than 10 deg.) and nocorrection to the streamwise velocity component wasperformed. The spanwise location of the measurementplane was fixed at 25% span (9 in.) from the side wall.The airfoil model had a row of surface static pressureorifices located at this spanwise station. These tapsallowed for direct measurement of the airfoil pressuredistribution directly below the traversing mechanism.

Two measurement probes were used. A TSI, Inc.model 1288 split-film probe was used for all of the dataacquisition. This was a 0.006-inch diameter platinum

Fig. 1 Smoothed ice shapes, after Addy et al.13

Fig. 2 Photograph of boundary-layer measurementapparatus installed in LTPT test section.

Smoothed 16.7-minute Rime Ice Shape 212

Smoothed 22.5-minute Glaze Ice Shape 944


film mounted on a fiber rod. The plane of the split wasoriented normal to the model chord (when the traverseaxis was not tilted). This orientation of the split-filmprovided the magnitude and direction of the streamwisevelocity component. The magnitude (but not direction)of the normal velocity was also obtained with thissensor. In addition, data from a flattened stagnationpressure probe were also acquired. The flattened tip ofthe pressure probe had a width of 0.010 inches. Thesedata were only valid in non-reverse flow regions andwere included only as an additional method to identifythe boundary-layer edge and compare to the split-filmdata in regions where no reverse flow was present. Thepressure probe was also used as part of an electricalcontact circuit. The contact circuit sensed when thepressure probe contacted the airfoil surface. Basedupon this location, the split-film sensor was positionedbetween 0.005 and 0.015-inches above the surface.This offset distance was regularly measured and appliedto the velocity profile data.

Since the purpose of this investigation was tosurvey the flowfields of two iced-airfoil configurationsover several Reynolds and Mach number conditions,time constraints prohibited detailed study of theseparated flowfields. This limited the number ofchordwise locations where the velocity profiles wereacquired. In general, profiles were acquired every 0.03x/c near the leading edge (up to x/c = 0.15), and thenevery 0.05 x/c farther downstream to the expectedshear-layer reattachment region. Downstream of theexpected reattachment region profiles were acquiredevery 0.10 to 0.25 x/c. The spacing of points in thevertical direction was adjusted based on the thickness ofthe separation bubble or boundary layer. The profilesconsisted of 25 to 40 points that had a non-lineardistribution with closer spacing near the airfoil surface.For the glaze ice shape configuration 944, thesevelocity profiles were acquired at α = 0, 4 and 6 deg.for Re = 3.5×106 and 6.0×106 at constant M = 0.12. Asubset of the velocity profiles was also acquired at Re =6.0×106 and M = 0.21. For the rime ice shapeconfiguration 212, the velocity profiles were acquired atα = 6, 8 and 10 deg. for Re = 3.5×106 and 6.0×106 atconstant M = 0.12.

The dual-sensor split-film probe was configured tomeasure the magnitude and direction of the streamwisevelocity component. This method and data reductionprocedures are described by Bragg et al.5 and Spring.17

The calibration of the sensor was performed in a 4-inchdiameter ejector-driven flow facility that was locatedwithin the plenum of the LTPT. The Reynolds andMach number combinations required two separatecalibrations corresponding to two stagnation pressures.During each calibration the plenum pressure was set to

each of these stagnation pressures. In this way, eachcalibration was performed with the appropriate airdensity and minimal correction of the data wasrequired. The calibration data were fit withpolynomials using a least-squares method. Severalcalibrations were performed during the test to ensurethat there was no drift due to sensor contamination orother effects. The velocity data are estimated to haveuncertainties of a few percent of the freestreamvelocity. This estimate is based on the calibrationrepeatability and comparison of the split-film velocitydata to that determined from the pressure probe.

Since the purpose of this investigation was tosurvey the flowfields of two iced-airfoil configurationsover several Reynolds and Mach number conditions,time constraints limited the amount of data the could beacquired during each velocity profile. For eachmeasurement point, 3000 split-film voltage sampleswere acquired at a rate of 1000 samples per second.This limited the bandwidth to a low-pass filter cut-offof 500 Hz. The voltages were temperature corrected (toaccount for differences in temperature during the dataruns versus calibration) and the calibration was applied.The resulting velocities were then corrected for anyminor difference in density between the data run andcalibration. No corrections were applied for potentialprobe support interference effects. While it is likelythat some interference effects were present, no suitablecorrection methods were found that were applicable tothis experiment. It should also be noted thatKhodadoust6 performed non-intrusive LDVmeasurements that compared very well with previoussplit-film measurements for the same iced-airfoilgeometry, further indicating that interference effectswere small.

The mean streamwise velocity (u) and root-mean-square of the velocity fluctuation (urms) were calculatedfrom the time series data. The mean velocity profileswere used to calculate several boundary-layerparameters. These calculations were carried out usingmethods similar to Bragg et al.5 The stagnationstreamlines were defined by the height above the airfoilsurface in each profile where the streamwise velocitywas zero. The dividing streamlines were defined by theheight above the airfoil surface where the integratedmass flow (in the streamwise direction) was zero. Thislocation, ydiv/c, can be expressed mathematically forconstant density flow as

where Ue is the boundary-layer edge velocity. Theedge velocity was selected manually for each profile

0

/

/

=

−∫ c

yyd

U

u surfcy

cy e

div

surf


using custom-written data-reduction software with agraphical user interface. Selection of the edge velocityalso allowed for calculation of the boundary-layerdisplacement thickness,

The momentum thickness was also calculated in theusual way, but with a simple modification for velocityprofiles that may have reverse flow,

RESULTS AND DISCUSSION

The lift performance of the GLC-305 airfoil wasaffectedly differently by the two ice shapes consideredin this investigation. As shown in Fig. 3, the maximumlift penalty for the airfoil with rime shape 212 wasmuch less than for the glaze shape 944. The lift curvesshow that maximum lift was attained at about 11 deg.for the rime shape and at about 7 deg. for the glazeshape. (The lift performance and pressure data weretaken from Addy et al.13 and were acquired without thetraversing mechanism installed in the test section.)This information was used to select the angles of attackat which the flowfield data were acquired. The choiceof 6, 8 and 10 deg. for the rime shape revealed theflowfield development leading up to stall. Theanalogous choices for the glaze shape were 0, 4 and 6deg.

A comparison of the ice geometries in Fig. 1 showsthat the glaze shape had a large “horn” that is orientedat a large angle relative to the airfoil chord line. Incontrast, the rime shape had a geometry that could beconsidered as an extension of the airfoil leading edge.These differences in geometry determined the extent ofseparated flow aft of the ice shape. A comparison ofclean and iced pressure distributions at matched angleof attack is shown in Fig. 4. The measured Cp

distribution in the region of the rime and glaze iceshapes was not smooth owing to the roughness of thesimulated ice. This is particularly evident on the lowersurface for both ice shapes. The clean-airfoil pressuredistribution was marked by a large suction peak ofnearly –4.0 in Cp, followed by a very large adversegradient. In the rime-ice pressure distribution, therewas a suction peak of –2.5 in Cp at x/c ≈ -0.03. Thispressure tap was located near the tip of the ice shape,where a large, local flow acceleration was likely tooccur. The short region of nearly constant pressurefrom x/c ≈ -0.02 to x/c ≈ 0.01 was indicative of a small

Fig. 3 Lift performance comparison of the GLC-305airfoil with and without leading-edge ice-shapesimulations at M = 0.12 and clean data at Re =3.0××106, iced data at Re = 3.5××106, after Addy et al.13

Fig. 4 Effect of ice simulations on surface pressuredistribution at M = 0.12 and clean data at Re =3.0××106, iced data at Re = 3.5××106, after Addy et al.13

∫δ

−

−=δ

c

cy

surf

esurf

c

yyd

U

uc

/

/

1/*

−

−=θ ∫

δ

c

yyd

U

u

U

uc

surf

e

c

cy esurf

1//

/

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

-4.0

-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

Clean, α = 6.0 deg.Rime shape 212, α = 6.0 deg.Glaze shape 944, α = 6.0 deg.

Cp

α (deg.)-8.0 -4.0 0.0 4.0 8.0 12.0 16.0 20.0

-0.8

-0.4

0.0

0.4

0.8

1.2

1.6

CleanRime shape 212Glaze shape 944

Cl


separation bubble. The pressure recovery aft of this“plateau” indicates transition in the separated shearlayer.4 That is, the start of transition occurred at theend of the plateau and start of the recovery regions.The shear-layer reattachment location can beapproximated by the location where the iced-airfoil Cp

intersects the clean airfoil Cp.4,5 In this case, the clean

and iced pressure recovery regions overlap. The split-film flowfield data indicated that reattachment occurredat x/c ≈ 0.05, which was slightly downstream of thisoverlap region. The pressure distribution for the airfoilwith the glaze ice shape indicates that the separationbubble was much larger in this case. The region ofnearly constant pressure on the upper surface extendedto x/c ≈ 0.25. The approximate shear-layerreattachment location, determined by the intersection ofthe clean and iced pressures was located at x/c = 0.60.This compares favorably to the location determinedfrom the split-film flowfield measurements, x/c = 0.53.The divergence of the trailing-edge and lower-surfacespressures also indicated that the airfoil was near stall.

An important concern in conducting the intrusivesplit-film measurements was the effect of the probe andsupport strut on the separated and reattaching flowfield.An analysis of the pressure distributions shows that thiseffect was likely small. Figure 5 shows a comparisonof pressures for various chordwise probe locations.These are for the airfoil with the glaze ice shape at 4deg. angle of attack. The data labeled “No Probe” weretaken from the main chordwise row of pressure tapswithout the traversing mechanism, and the other threesets were taken from the chordwise row of taps locatedat the same spanwise station as the main verticaltraverse strut. These pressures were acquired with thesplit-film probe located within 0.015-inch of the airfoilsurface. As shown in the data, the presence of theprobe and strut caused slightly larger suction pressuresin the “plateau” region of the separation region. Thebeginning of the pressure recovery is not effected bythe probe and strut. However, the recovery gradientbecame larger as the probe was moved farther upstreaminto the bubble region. This had the effect of slightlyreducing the bubble size with the shear-layerreattachment being farther forward by a few percentchord. It should be noted that the faired probe holderwas about 8% chord downstream of the probe tip.Since the probe was angled down toward the airfoilsurface, the probe holder was always much fartherabove the surface. Its presence is observed in thepressure at x/c = 0.40 with the probe tip at x/c = 0.30.Its proximity to the surface caused a local flowacceleration resulting in the more negative Cp at thislocation. This effect was not observed as the probe tipwas located farther upstream with the probe holder

located in the pressure recovery region. This minoreffect of the probe and strut interference wascomparable to that observed by Bragg et al.5

Khodadoust6 also compared split-film velocity data tonon-intrusive LDV data and found very goodagreement, further indicating that interference effectswere small.

Fig. 5 Effect of probe and strut on surface pressuredistribution for the glaze ice configuration 944 at αα= 4 deg. with Re = 3.5××106 and M = 0.12.

Flowfield ComparisonsContour plots of the mean streamwise velocity

provide a good overall illustration of the separated flowpast the ice shape. An example of these data is shownin Fig. 6 for the airfoil with the glaze ice shape at 6 deg.angle of attack. The plot shows how the boundary-layer separated near the tip of the glaze ice horn. Asignificant reverse flow region formed below theseparated shear layer. Reverse flow velocities as highas 40% of the freestream velocity were recorded insidethe bubble. Outside of the bubble, streamwise flowvelocities a factor of 1.6 larger than the freestreamvalue were recorded. A strong shear layer divided theseflow regions. The shear-layer thickness grewdownstream of separation as transition occurred. Novelocity data were acquired upstream of x/c = 0.02directly aft of the ice horn, as indicated in Fig. 6. It islikely that the local flow in this region was very low-speed in either the upstream or downstream direction.The zero-velocity contour line shows that the mean

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

-2.0

-1.8

-1.6

-1.4

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

No ProbeProbe at = 0.30Probe at = 0.20Probe at = 0.10

Cp

x/cx/cx/c


Fig. 6 Contour plot of mean streamwise velocity forthe glaze ice configuration 944 at αα = 6 deg. with Re= 3.5××106 and M = 0.12.

reattachment location was near x/c = 0.53. Thedeveloping turbulent boundary layer downstream ofreattachment did not show evidence of separation up tox/c = 0.95, where the last profile was measured. Whilethe contour plots provide an overall view of thecomplex flowfield, quantitative comparisons of thebubble characteristics are difficult to determine.

The stagnation and dividing streamlinecharacteristics demonstrate more quantitative aspects ofthe separation bubble development. These are plottedin Fig. 7 for the rime ice shape and Fig. 8 for the glazeice shape. Both sets of streamlines indicate the meanshear-layer reattachment location by the intersection ofthe streamline with the airfoil surface. For the airfoilwith the rime shape, the increase in bubble size wasnearly linear with increasing angle of attack, basedupon the movement of reattachment locations. For theairfoil with the glaze shape, the bubble growth wasmuch more non-linear with increasing angle of attack.The bubble approximately doubled in size from α = 0to 4 deg. and then nearly doubled in size from α = 4 to6 deg. The streamlines clearly illustrate the differencein aerodynamic severity between the two different iceshapes. The rime ice shape could be thought of as anextension of the airfoil leading edge. Since it was notsmooth, a small bubble formed at 6 deg. angle of attack.The glaze shape, in contrast, had a large upper surfacehorn that was located downstream of the leading edgeand had a large angle to the oncoming flow. Thisresulted in a much larger bubble at 6 deg. angle ofattack and hence the impending stall. The bubbleformed in the rime ice case at α = 10 deg., was evensmaller than for the glaze case at 6 deg. owing to thisdifference in geometry. These results reinforceprevious performance results about the size andlocation of ice horn features.2,18

The separation bubble reattachment locationsdetermined from the streamline plots are summarized inTable 1. Also shown are the locations determined fromthe simple method of comparing the clean and icedpressure distributions. The data indicate that thepressure distribution method works well as anapproximate method of determining reattachment evenfor these long bubbles. Thus, the data further supportthe conclusion of Bragg et al.5 in this regard.

Fig. 7 Stagnation and dividing streamlines for therime ice configuration 212 at Re = 3.5××106 and M =0.12.

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

0.0

0.1

0.21.61.41.21.00.80.60.40.20.0

-0.2-0.4

u/U∞

y/c

u/U∞ = 0.0 contour line

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.0

0.1

0.2 α = 6 deg.α = 8 deg.α = 10 deg.

Stagnation Streamlines

y/c

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.0

0.1

0.2 Dividing Streamlines

y/c


Fig. 8 Stagnation and dividing streamlines for theglaze ice configuration 944 at Re = 3.5××106 and M =0.12.

Table 1 Comparison of shear-layer reattachmentlocations at Re = 3.5××106 and M = 0.12.

Reattachment Locations, x/cIce Shape andAngle of Attack From Cp’s From Streamlines

Rime, α = 6 deg. 0.04 0.05Rime, α = 8 deg. 0.20 0.18Rime, α = 10 deg. 0.45 0.40Glaze, α = 0 deg. 0.14 0.15Glaze, α = 4 deg. 0.31 0.30Glaze, α = 6 deg. 0.60 0.53

The integral boundary-layer parameters offerfurther comparison of these separation bubblecharacteristics. The displacement and momentumthicknesses are plotted for both shapes in Figs. 9 and10. For the glaze ice shape case (Fig. 9), there was alarge increase in displacement thickness from x/c =-0.02 to 0.02. This trend was also observed by Bragg etal.,5 for a simulated glaze ice shape on a NACA 0012airfoil and their data are plotted in Fig. 9 forcomparison. The plot symbols were matched based onbubble size. The separation bubble length for α = 4deg. of Bragg et al.5 was similar to α = 0 deg. in thepresent data. Likewise, the bubble length for α = 6 deg.of Bragg et al.5 was similar to α = 4 deg. in the presentdata. The agreement in these data for the α = 0/4 deg.case is remarkably good, considering the differences inice shape and airfoil geometry. These differences in

geometry likely caused the divergence in the two datasets downstream of x/c = 0.15, which coincides with thebubble reattachment location. It is likely that thepressure distributions were different downstream of thislocation, thus leading to differences in the turbulentboundary-layer development. For the α = 4/6 deg.comparison, the agreement is also good up to x/c =0.15. Downstream of this location the trend in thepresent data is consistent with the trend for the α = 0deg. case, whereas δ*/c tends to level off for the Bragget al.5 data. The reason for this is not clear, but may berelated to changes in the flowfield close to stall.Maximum lift in both the present data and Bragg et al.5

occurred at 7 deg. Therefore, bubbles of similar sizesoccurred at one degree below maximum lift for Bragget al.,5 but three degrees below maximum lift for thepresent data. Unfortunately, no data downstream of x/c= 0.35 were available in the present case one degreebelow maximum lift (i.e., at α = 6 deg.) to aid in this

Fig. 9 Comparison of integral boundary-layerparameters for the glaze ice configuration 944 at Re= 3.5××106 and M = 0.12.

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.0

0.1

0.2 α = 0 deg.α = 4 deg.α = 6 deg.

Stagnation Streamlinesy/

c

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.0

0.1

0.2 Dividing Streamlines

y/c

x/c

δ* /c

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

α = 0 deg.α = 4 deg. Ref. 5α = 4 deg.α = 6 deg. Ref. 5α = 6 deg.

Displacement Thickness

x/c

θ/c

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.000

0.002

0.004

0.006

0.008

0.010

0.012

0.014

0.016

0.018

0.020Momentum Thickness


interpretation. The data comparison in the α = 0/4 deg.case do suggest that these large separation bubbles mayhave certain integral characteristics that are universallysimilar.

Analogous comparisons were also performedbetween the two data sets for the boundary-layermomentum thickness as shown in Fig. 9. As with thedisplacement thickness, the agreement is good formatched bubble sizes at α = 0/4 deg. upstream of x/c =0.20. Downstream of this location there was somedivergence likely owing to the differences in geometry.Agreement is also good in the α = 4/6 deg. caseupstream of x/c = 0.30, except for the present datapoints at x/c = 0.15 and x/c = 0.25. The location x/c =0.30 approximately corresponds to bubble reattachmentfor both data sets. Bragg et al.5 noted that for their α =4 deg. data, the first local maximum in momentumthickness corresponded to the shear-layer transitionlocation as determined from the surface pressuredistribution. In the present data (at α = 0 deg.), thelocal maximum was located at x/c = 0.10. This is veryclose the transition location estimated from the surfacepressure distribution. The local maxima at x/c ≈ 0.15and x/c ≈ 0.25 for α = 4 and 6 deg. in the present dataalso corresponded to the transition location determinedfrom the surface pressure distribution. Bragg et al.5

also suggested that the shear-layer reattachmentlocation could be determined by the second localmaxima in the θ/c data, if present. This only occurredat x/c = 0.20 for α = 0 deg. in the present data and thiswas downstream of reattachment which occurred at x/c= 0.15.

These integral boundary-layer parameters areplotted for the rime ice shape in Fig. 10. The smallerbubble sizes for this ice shape are clearly indicated bythe much lower values of displacement thicknesscompared to the glaze ice case. Despite the differencesin the magnitude of the values, the trends are verysimilar. There is a steep increase in δ*/c from x/c =-0.017 to x/c = 0.0, this is followed by a reduced slopereaching a maximum value. Given the differences inthe geometry of the ice shapes, the agreement in thesetrends further indicates universal similarities of theseparation bubble characteristics. Analogous trendswere also observed in the momentum thicknesscharacteristics. For example, there are local maxima atx/c = 0.02 and x/c = 0.08, for α = 6 and 8 deg.,respectively. These locations approximatelycorrespond to the start of the pressure recovery regionof the surface pressure distribution, thus indicative ofshear-layer transition. It is not clear from these datawhy a similar local maximum does not occur at α = 10deg.

Fig. 10 Comparison of integral boundary-layerparameters for the rime ice configuration 212 at Re= 3.5××106 and M = 0.12.

Reynolds and Mach Number EffectsThe independent effects of Reynolds and Mach

number on the separated flowfield development wasalso investigated. This was motivated by previousmeasurements that showed little change in theintegrated performance coefficients and pressuredistributions of the iced airfoil for large changes inReynolds number.13 In fact, it was found that changesin Mach number had more of an effect on maximum liftthan did changes in Reynolds number.13 An example ofthis effect is illustrated in Fig. 11. The plot shows howthe maximum lift coefficient decreased slightly as theMach number was increased. This drop in Cl,max wasabout 8% between M = 0.12 and M = 0.28. Acomparison with the lift data in Fig. 3 show virtually nodifference in Cl,max from Re = 3.5×106 to 10.5×106 at M= 0.12. Similar trends were observed in the flowfieldsof the present study. For example, Fig. 12 shows acomparison of stagnation streamlines for three

x/c

δ* /c

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

α = 6 deg.α = 8 deg.α = 10 deg.

Displacement Thickness

x/c

θ/c

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.000

0.002

0.004

0.006

0.008

0.010

0.012

0.014

0.016

0.018

0.020Momentum Thickness


Reynolds and Mach number combinations for theairfoil with glaze ice shape 944. For the α = 4 deg.case, the bubble size is virtually identical across thethree free-stream conditions. The apparent discrepancybetween x/c = 0.12 and 0.24 for the M = 0.21 caseoccurred because no boundary-layer profiles were takenon this interval for this condition. For α = 6 deg., thestagnation streamline for the M = 0.21 case indicatesthat the separation bubble was slightly larger than forthe other conditions. This is consistent with theinterpretation of the pressure distributions in Addy etal.13 and the present study that indicate larger bubblesizes at higher Mach numbers. It is reasonable that thelarger bubble results in airfoil stall at a slightly lowerangle of attack and lower lift coefficients. It is unclearwhat phenomenon causes the larger bubble sizes forhigher Mach numbers.

The integral boundary-layer parameters also showvery little influence of Reynolds and Mach numbers.These are plotted in Fig. 13 for the glaze ice shape caseat α = 6 deg. The larger bubble size observed in Fig. 12for M = 0.21, is partially represented in the δ*/c data atx/c = 0.35. Unfortunately, no displacement thicknessescould be calculated downstream of this locationbecause the edge of the boundary layer was notmeasured. This was simply a limitation of thetraversing mechanism. This effect is not observed inthe momentum thickness data, however, it suffers fromthe same limitations downstream of x/c = 0.35. Thediscrepancy in θ/c values between x/c = 0.12 and 0.24for the M = 0.21 case, is likely due to the lack of datafor these locations.

A comparison of the velocity profiles is shown inFig. 14. The profile at x/c = 0.12 shows the strongreverse flow velocities in the separation bubble region.There was good agreement in the velocity data acrossthe three conditions up to the edge of the shear layer at(y-ysurf)/c = 0.07, where the M = 0.21 velocities wereslightly slower than the M = 0.12 data. The profilesdownstream of this location show how the strength andsize of the reverse flow region decreased up to x/c =0.55 which was close to the mean reattachment locationfor this case. The vertical location above the surfacewhere the M = 0.21 velocity data began to depart fromthe M = 0.12 data moved closer to the surface forprofiles located farther downstream. Although it isdifficult to see in Fig. 14, the profiles at x/c = 0.55 showthat this location was downstream of meanreattachment for the M = 0.12 cases, but upstream ofmean reattachment for M = 0.21. The velocity profilesat x/c = 0.45 and 0.55 also illustrate that thedisplacement thickness at M = 0.21 was likely largerthan for the M = 0.12 cases, since the mean velocitieswere slower. Therefore, it is likely that the

displacement thicknesses downstream of x/c = 0.35would continue to be larger at M = 0.21 than at M =0.12.

Fig. 11 Effect of Mach number on the liftperformance of the GLC-305 airfoil with glaze iceconfiguration 944, after Addy et al.13

Fig. 12 Effect of Reynolds and Mach number onstagnation streamline locations for the glaze iceconfiguration 944.

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.0

0.1

0.2 α = 6 deg.

y/c

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.0

0.1

0.2 = 3.5×106, = 0.12= 6.0×106, = 0.12= 6.0×106, = 0.21

α = 4 deg.

y/c

Re MRe MRe M

α (deg.)-6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

= 10.5×106, = 0.12= 10.5×106, = 0.21= 10.5×106, = 0.28

Cl

Re MRe MRe M


Fig. 13 Effect of Reynolds and Mach number onintegral boundary-layer parameters for the glaze iceconfiguration 944 at αα = 6 deg.

Fig. 14 Effect of Reynolds and Mach number onselected velocity profiles for the glaze iceconfiguration 944 at αα = 6 deg.

x/c

δ* /c

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

= 3.5×106, = 0.12= 6.0×106, = 0.12= 6.0×106, = 0.21

Displacement Thickness α = 6 deg.

Re MRe MRe M

x/c

θ/c

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00.000

0.002

0.004

0.006

0.008

0.010

0.012

0.014

0.016

0.018

0.020Momentum Thickness α = 6 deg.

u/U∞

(y-y

surf

)/c

-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.60.00

0.02

0.04

0.06

0.08

0.10

0.12

= 3.5×106, = 0.12= 6.0×106, = 0.12= 6.0×106, = 0.21

Re MRe MRe M

x/c = 0.12

u/U∞

(y-y

surf

)/c

-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.60.00

0.02

0.04

0.06

0.08

0.10

0.12x/c = 0.30

u/U∞

(y-y

surf

)/c

-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.60.00

0.02

0.04

0.06

0.08

0.10

0.12x/c = 0.45

u/U∞

(y-y

surf

)/c

-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.60.00

0.02

0.04

0.06

0.08

0.10

0.12x/c = 0.55


Turbulence Intensity ResultsThe turbulence intensity contours for the separated

shear-layer regions were found to be in agreement withresults from other separated flows. In this case theturbulence intensity was calculated as the root-mean-square of the fluctuating streamwise velocity andnormalized by the freestream velocity. An example ofthese data is shown in Fig. 15 for the airfoil with theglaze ice shape at α = 6 deg. This turbulence intensitycontour corresponds to the velocity contour alreadydiscussed in Fig. 6. The maximum values, in the rangeof 0.32 to 0.36, occurred in the middle of the separatedshear layer at x/c ≈ 0.30. This region of peakturbulence intensity began just downstream of x/c =0.25, which has already been identified as the shear-layer transition location from the pressure data. Whiledifficult to see in the black and white contour plot, theurms/U∞ levels in the range of 0.28 to 0.32 persistdownstream to x/c = 0.55, the vicinity of reattachment.Downstream of reattachment the turbulence intensitylevels decreased gradually, having peak values in therange of 0.20 to 0.24 at x/c = 0.95.

The trends and values of the turbulence intensitycompare favorably with LDV measurements of theseparation bubble flowfield past a NACA 0012 airfoilwith simulated glaze ice accretion. Khodadoust6

reported peak values of urms/U∞ = 0.34 in the vicinity ofthe separated shear-layer transition location. Thegeneral distribution of the turbulence intensitythroughout the bubble flowfield was also very similar.Khodadoust noted that these values are in the range ofthose reported for separated flows downstream of abackward-facing step. For example, Eaton andJohnston19 state that local turbulence intensity values(urms/U∞) near the center of the reattaching shear layerexceed 0.30. These large values have been attributed tolarge-scale and low-frequency perturbations of the

Fig. 15 Contour plot of turbulence intensity for theglaze ice configuration 944 at αα = 6 deg. with Re =3.5××106 and M = 0.12.

separated shear layer sometimes referred to as“flapping.”20,21 In the case of the backward-facing step,this vertical motion of the shear layer results inmovement of the reattachment location. It is possiblethat similar unsteady characteristics may be present inthe iced-airfoil case. Bragg et al.5 reported a low-frequency component in the spectra of the fluctuatingshear-layer streamwise velocity. Likewise, Gurbackiand Bragg22 identified low-frequency components inthe fluctuating pressure spectra in the separation bubbleflowfield on an iced airfoil. The combined results ofthese studies suggest that a completely time-averagedrepresentation of the flow may mask important details.This has implications for numerical modeling as well.

SUMMARY AND CONCLUSIONS

Flowfield measurements were performed on theupper surface of a GLC-305 airfoil configured withlarge glaze and rime ice-shape simulations. The meanand root-mean-square fluctuation of the streamwisevelocity were acquired using a split-hot-film probe atseveral chordwise locations. These data were taken atthree different angles of attack preceding stall for eachiced-airfoil configuration. The freestream conditionswere Re = 3.5×106 and 6.0×106 at M = 0.12 and 0.21.Integral boundary-layer parameters for these cases werecalculated from the mean velocity profiles.

The velocity measurements confirmed the presenceof large separation bubbles downstream of the iceshapes. For all cases measured, the separated shearlayer reattached to the airfoil some distancedownstream. No evidence of turbulent boundary-layerseparation as far aft as x/c = 0.95 was found. Theseparation bubbles for the glaze ice configuration weremuch larger than those for the rime ice case, resultingfrom the differences in the horn geometry. Other than

x/c0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

0.0

0.1

0.20.360.320.280.240.200.160.120.080.040.00

urms /U∞

y/c


the differences in size, the integral boundary-layercharacteristics were very similar. Analogous trends inthe displacement thickness and momentum thicknesswere also observed and these were consistent with datafrom other experiments. Following the work performedby others, it was noted that local maxima in themomentum thickness distributions correlated well withthe shear-layer transition location. These observationssuggests that these large separation bubbles may havecharacteristics that are universally similar. Changes inReynolds number did not significantly affect theseparation bubble characteristics. However, a largerMach number did result in a slightly larger separationbubble for the glaze ice case at α = 6 deg. This resultwas consistent with previous observations of the airfoilsurface pressure and performance data. The rmsvelocity distributions had peak values in the separatedshear layer, downstream of transition, that comparedwell with previous work. These large values (on theorder of 0.30 to 0.35 times the mean freestreamvelocity), indicate that potentially large-scaleunsteadiness was present in the flowfield.

ACKNOWLEDGEMENTS

The authors wish to acknowledge severalindividuals who made significant contributions to thiswork. In particular, Joe Zoeckler and Carl Blaser fromNASA Glenn played an important role in modifyingand improving the traversing apparatus. Test engineersPam Phillips and Bill Sewall and the experiencedtechnicians of the NASA Langley LTPT deservespecial recognition for their diligent efforts to makethese sophisticated measurements successful. TheLTPT staff played a collaborative role in upgrading thetraverse apparatus and developing the split-filmcalibration facility. Finally, the authors at theUniversity of Illinois were supported, in part, underNASA Grant NCC 3-852 from the John H. GlennResearch Center.

REFERENCES

1. Lynch, F.T., and Khodadoust, A., “Effects of IceAccretion on Aircraft Aerodynamics,” Progress inAerospace Sciences, Vol. 37, 2001, pp. 669-767.

2. Lee, S., and Bragg, M.B., “Investigation of FactorsAffecting Iced-Airfoil Aerodynamics,” Journal ofAircraft, Vol. 40, No. 3, May-June, 2003, pp. 499-508.

3. Bragg, M.B., Broeren, A.P., and Blumenthal, L.A.,“Iced-Airfoil and Wing Aerodynamics,” SAE Paper2003-01-2098, Jun., 2003.

4. Tani, I., “Low Speed Flows Involving SeparationBubbles,” Progress in Aeronautical Sciences,Pergamon, New York, 1964, pp. 70-103.

5. Bragg, M.B., Khodadoust, A., and Spring, S.A.,“Measurements in a Leading-Edge Separation Bubbledue to a Simulated Airfoil Ice Accretion,” AIAAJournal, Vol. 30, No. 6, Jun. 1992, pp. 1462-1467.

6. Khodadoust, A., “An Experimental Study of theFlowfield on a Semispan Rectangular Wing with aSimulated Glaze Ice Accretion,” Ph.D. Dissertation,Dept. of Aeronautical and Astronautical Eng., Univ. ofIllinois, Urbana, IL, 1992.

7. Kerho, M.F., Bragg, M.B., and Shin, J., “HeliumBubble Flow Visualization of the Spanwise Separationon a NACA 0012 with Simulated Glaze Ice,” AIAAPaper 92-0413, Jan. 1992.

8. Bragg, M.B., Kerho, M.F., and Khodadoust, A.,“LDV Flowfield Measurements on a Straight andSwept Wing with a Simulated Ice Accretion,” AIAAPaper 93-0300, Jan. 1993.

9. Gurbacki, H.M., and Bragg, M.B., “UnsteadyAerodynamic Measurements on an Iced Airfoil,” AIAAPaper 2002-0241, Jan. 2002.

10. Dunn, T.A., Loth, E., and Bragg, M.B.,“Computational Investigation of Simulated Large-Droplet Ice Shapes on Airfoil Aerodynamics,” Journalof Aircraft, Vol. 36, No. 5, Sept.-Oct. 1999, pp. 836-843.

11. Pan, J., Loth, E., and Bragg, M.B., “RANSSimulations of Airfoils with Ice Shapes,” AIAA Paper2003-0729, Jan. 2003.

12. Chung, J. J., and Addy, H.E. Jr., “A NumericalEvaluation of Icing Effects on a Natural Laminar FlowAirfoil,” AIAA Paper 2000-0096, Jan. 2000.

13. Addy, H.E. Jr., Broeren, A.P., Zoeckler, J.G., andLee S., “A Wind Tunnel Study of Icing Effects on aBusiness Jet Airfoil,” AIAA Paper 2003-0727, alsoNASA TM-2003-212124, Jan. 2003.


14. Von Doenhoff, A.E., and Abbott, F.T. Jr., “TheLangley Two-Dimensional Low-Turbulence PressureTunnel,” NACA TN 1283, May 1947.

15. McGhee, R.J., Beasley, W.D., and Foster, J.M.,“Recent Modifications and Calibration of the LangleyLow-Turbulence Pressure Tunnel,” NASA TP 2328,July 1984.

16. Spaid, F.W., and Lynch, F.T., “High ReynoldsNumber, Multi-Element Airfoil FlowfieldMeasurements,” AIAA Paper 96-0682, Jan. 1996.

17. Spring, S.A., “An Experimental Mapping of theFlowfield Behind a Glaze Ice Shape on a NACA 0012Airfoil,” M.S. Thesis, Dept. of Aeronautical andAstronautical Eng., Ohio State Univ., Columbus, OH,1987.

18. Kim, H.S., and Bragg, M.B., “Effects of Leading-Edge Ice Accretion Geometry on AirfoilAerodynamics,” AIAA Paper 99-3150, 17th AIAAApplied Aerodynamics Conference, June 28-July 1,1999.

19. Eaton, J.K. and Johston, J.P., “A Review ofResearch on Subsonic Turbulent Flow Reattachment,”AIAA Journal, Vol. 19, No. 9, Sept. 1981, pp. 1093-1100.

20. Simpson, R.L., “Aspects of Turbulent Boundary-Layer Separation,” Progress in Aerospace Sciences,Vol. 32, 1996, pp. 457-521.

21. Driver, D.M., Seegmiller, H.L., and Marvin, J.G.,“Time-Dependent Behavior of a Reattaching Shear-Layer,” AIAA Journal, Vol. 25, No. 7, July, 1987, pp.914-919.

22. Gurbacki, H.M. and Bragg, M.B., “UnsteadyFlowfield About an Iced Airfoil,” AIAA 2004-0562,Reno, NV, January 5-8, 2004.

flowfield measurements about an airfoil with...

Documents