Investigation of Turbulent Boundary-Layer Separation Using Laser Velocimetry

D. Modarress ° and D. A. JohnsontNASA Ames Research Center, Moffett Field, Calif.Boundary-layer measurements realized by laser veloeimetry are presented for a Mach 2.9, two-dimensional,shock-wave/turbulent boundary-layer interaction containing an extensive region of separated flow. Meanvelocity and turbulent intensity profiles were obtained from upstream of the interaction zone to downstream ofthe mean reattachment point. The superiority of the laser velocimeter technique over pressure sensors in tur-bulent separated flows is demonstrated by a comparison of the laser velocimeler data with results obtained fromlocal pitot and static pressure measurements for the same flow conditions. The locations of the mean separationand reanachment points as deduced from the mean velocily measurements are compared 1o oil-flowvisualization results. Representative velocity probability density functions obtained in the separated flow regionare also presented. Critical to the success of this investigation were !) the use of Bragg cell frequency shifting and2) artificial seeding of the flow wit h submicron light-scattering particles.IntroductionNTIL recently, numerical solution of the time-averagedconservation equations for the separated flow producedby a shock wave impinging on a turbulent boundary layer wasbeyond the scope of computational fluid dynamics. Now,with the development of better numerical codes and moreadvanced computers, a number of investigators have madepredictions of this type of flow. _s However, still open toquestion is how satisfactorily the turbulence transportproperties are being modeled. To test the various turbulencemodels employed, comparisons with experiments are needed.Unfortunately, the only mean velocity profile data availablefor comparison have been those obtained with pitot and staticpressure probes, which must be considered suspect within andnear the separation region because of the very high turbulencelevels present. The degree of turbulence is such that, over alarge portion of the flow, the velocity component in thestreamwise direction fluctuates in sign. Thus, regardless ofwhether the pressure probes are aligned with the freestreamflow or faced in the backward direction, measurement errorswill result. Also of concern are the flow disturbancesproduced by the probes themselves.The laser velocimeter, because of its nonintrusive nature,presents an attractive alternative in the study of separatedflows. Its importance in general fluid flow research is wellrecognized. 6 The advantages of laser velocimetry overconventional instrumentation become more evident as thecomplexity of the flows increase. For example, in the study ofcompressible turbulent flows, the Reynolds normal and shearstresses can be measured without the signal interpretationdifficulties involved with hot-wire anemometry. In regions ofturbulent separation, forward and reverse instantaneousvelocities can easily be distinguished through frequencyshifting techniques. Such directional information is notachievable with other existing techniques. Also, since it has atruly linear response, no inaccuracies need result when thefluctuations are large compared to the mean value.Presented as Paper 76-374 at the AIAA 9th Fluid and PlasmaDynamics Conference, San Diego, Calif., July 14-16, 1976; submittedAug. 12, 1976; revision received Jan. 14, 1979. Copyright ©American Institute of Aeronautics and Astronautics, Inc., 1976. Allrights reserved.Index categories: Lasers; Boundary Layers and Convective HeatTransfer--Turbulent."NRC Research Associate. Presently Assistant Professor, ARYA-MEHR University of Technology, Tehran, lran. Member AIAA.tResearch Scientist. Member AIAA.Several investigators have demonstrated that the laservelocimeter can provide accurate localized velocity in-formation in high-speed wind tunnels. Favorable meanvelocity comparisons with pitot-tube measurements have beenobtained for turbulent supersonic boundary layers with zeropressure gradient. 7-9The Reynolds normal and shear stresseshave been measured for an undisturbed turbulent boundarylayer 9'_° and for a relatively mild (unseparated case) shock-wave/turbulent boundary-layer interaction. _ For the pointsaway from the wall, fluctuation measurements agreed wellwith hot-wire anemometer measurements.The major shortcoming of the laser velocimeter technique isthat minute particles are required in the detection process.These particles must be large enough to provide the scatteredlight levels required by the detection electronics and yet smallenough to follow the fluid motion. In compressible flows, theparticle trackability requirements can be especially extremebecause of the existence of large spatial velocity gradientsand/or high, turbulent convection velocities. However, theneed for a better understanding of the behavior of these flowsand the potentials of laser velocimetry provide the impetus toovercome this difficulty.Another controversial problem associated with laservelocimetry is the error associated with the velocity biasing, inRef. 12 it is argued that the particle passage through thesensing volume is not independent of the instantaneousvelocity field. The probability of a particle occurrence isdirectly proportional to the magnitude of the instantaneousvelocity vector Iv, I, and moreover, this weighting function isindependent of the concentration of particles. Hence, anymeaningful correction to the readings is done when Iv, I isknown. Experimental results have shown that to correct thedata on the basis of a one-dimensional model would result inan overcorrection and more erroneous results.In the present investigation, a directionally sensitive laservelocimeter system was used to study the interaction of anexternally generated oblique shock wave and a turbulentboundary layer for a freestream Mach number of 2.9. Theinteraction was sufficiently strong to produce an extensiveregion of recirculating flow. Mean velocities and turbulenceintensities in the streamwise direction were obtainedthroughout this interaction region with the velocimetersystem. Representative examples of these mean velocityprofiles are compared to results obtained with pitot and staticpressure probes for the same flow conditions. The ap-proximate locations of the mean separation and reattachmentpoints as determined from the velocity measurements are