JVis(2017)20:695–710 DOI10.1007/s12650-016-0405-3 REGULAR PAPER Gholamhossein Taherian • Mahdi Nili-Ahmadabadi • Mohammad Hassan Karimi • Mohammad Reza Tavakoli Flow visualization over a thick blunt trailing-edge airfoil with base cavity at low Reynolds numbers using PIV technique Received:14July2016/Revised:25September2016/Accepted:21October2016/Publishedonline:28November2016 (cid:2)TheAuthor(s)2016.ThisarticleispublishedwithopenaccessatSpringerlink.com Abstract Inthisstudy,theeffectofcuttingtheendofathickairfoilandaddingacavityonitsflowpattern isstudiedexperimentallyusingPIVtechnique.First,bycutting30%chordlengthoftheRisoairfoil,athick blunt trialing-edge airfoil is generated. The velocity field around the original airfoil and the new airfoil is measuredbyPIVtechniqueandcomparedwitheachother.Then,addingtwoparallelplatestotheendofthe new airfoil forms the desired cavity. Continuous measurement of unsteady flow velocity over the Riso airfoil with thick blunt trailing edge and base cavity is the most important innovation of this research. The results show that cutting off the end of the airfoil decreases the wake region behind the airfoil, when separation occurs. Moreover, adding a cavity to the end of the thickened airfoil causes an increase in momentum and a further decrease in the wake behind the trailing edge that leads to a drag reduction in comparison with the thickened airfoil without cavity. Furthermore, using cavity decreases the Strouhal number and vortex shedding frequency. Keywords Thick blunt trialing-edge airfoil (cid:2) Velocity field (cid:2) Base cavity (cid:2) PIV (cid:2) Wind tunnel 1 Introduction Airfoilswiththickenededgesareusedinsubsonicflowtoresolvemanyoftheproblemsencounteredinthe conventional airfoils. In the conventional airfoils, the trailing edge ends to a sharp point. However, for certainapplications(e.g.,awinginhighattackangles),anairfoilwithathicktrailingedgeisrequired.The gradual reduction of sharpness at the downstream of the maximum thickness section of an airfoil creates a strong positive pressure gradient at its low-pressure side, leading to an untimely flow separation. This positive pressure gradient can be somewhat reduced using a thickened edge. Thus, the recovered pressure can be partially transferred to the separation region produced by the airfoil. Improving the aerodynamic performanceofairfoilswiththickenededgeswouldrequirecertaindrag-reducinginstruments(Javarashkian and Lotfi 2008). Mathey (2008), Heskestad and Olberts (1960), and Zhang et al. (2015) studied numerically, and Bourgoyne et al. (2000), Nakano et al. (2007), and Gerontakos and Lee (2008) studied experimentally the flow characteristics of an airfoil to understand the vortex shedding mechanism. Cooperman et al. (2010) numerically and experimentally studied the characteristics of UCD-38-095 airfoilwithblunttrailingedge.Intheexperimentalpart,liftanddragforcemeasurementswerecarriedoutin DavisaeronauticalwindtunnelatReynoldsnumbersof333,000and666,000.Bothfreeandfixedtransition G.Taherian(cid:2)M.Nili-Ahmadabadi(&)(cid:2)M.H.Karimi(cid:2)M.R.Tavakoli DepartmentofMechanicalEngineering,IsfahanUniversityofTechnology,Isfahan84156-83111,Iran E-mail:[email protected] 696 G.Taherianetal. conditions were studied. The wind tunnel results are compared with computational predictions obtained in OVERFLOW, a Reynolds-averaged Navier–Stokes solver using structured overset grids. Baker and Dam (2008) numerically and experimentally studied the FB-3500-1750 airfoil with blunt trailing-edge and used certain instruments to reduce the drag and increase the lift for improving the performanceofthisairfoil.TheyusedapyramidalbalanceforthemeasurementoftheliftanddragatDavis aeronauticalwindtunnel.Toreducedrag,theyimplementedtwosplitterplatesaswellasopenandmoderate cavities.Theirresultsshowedthatusingthesplitterplatescouldreducedragbyupto50%.Althoughabase cavityreduceddragbyupto25%,italsointroduceddrasticfluctuationsinlift.Usingamoderatecavitynot only improved the drag-reducing performance of the splitter plate, but also limited the unsteady vortex shedding. Caietal.(2008)conductedanumericalstudyonthesinusoidaledgeeffectinaflowstoppingairfoil-like body with a thick edge to determine how it affected the aerodynamic characteristics of this body. They examinedtheeffectofthesinusoidaledgewavelengthondragcoefficient,vortexshedding,andfrequency. Their resultsshowed thatusingaspecific wavelengthwould reducedrag morethan 30%ascompared with the drag produced by a smooth-edged airfoil. In this case, the smaller separation region resulting from educeddragwouldleadtoafurtherreductioninthefrequencyofoscillationsinsidetheseparationregionas well as increased flow vorticity. By measuring pressure on the upper surface of an airfoil and implementing the particle image velocimetry (PIV) technique, Aravind and Al-Garni (2009) studied the flow behavior over a two-dimen- sional bluff body combined to a base cavity with various shapes. For this purpose, they used four types of cavities.Theirexperimentswiththemoderatecavityshowedthattworotationalregionsadjacenttothebase cavitywerecreated.Moreover,dramaticvariationswereobservedinthelengthoftherecirculationregionas well as the mean flow field in the presence of the base cavity. Ramjee et al. (1986) conducted theoretical and experimental studies on the NACA0012 airfoil with a thick edge. They produced the thick-edged airfoil by cutting off the trailing edge of the airfoil at the following distances: 5, 10, and 15% of the chord length as measured from the end of the airfoil. They calculatedtheinducedliftanddragforcesinawindtunnelatanglesbetween0(cid:3)and20(cid:3)andRe = 400,000. Experiments were conducted in a closed-circuit open-jet wind tunnel with 1.5 m diameter test section at different air speed. Each of the models was attached to a semiautomatic mechanical six component wire balance. The results showed that lift and drag both increased as a result of increasing the trailing edge thickness. DeshpandeandSharma(2012)empiricallystudiedtheeffectsofconnectingtrapezoidalprismaticblocks withdifferentsizestothethickendofanairfoilforthepurposeofeliminatingCarmenvortices.Theyused the hot wire apparatus as well as flow detection techniques to conduct their experiments. Their results revealed that using trapezoidal prismatic blocks somewhat weakened the Carman vortices formed at the back of the airfoil. However, pressure distribution measurements showed that drag was reduced by a mere 3–4%. Olsman(2010)studiedtheaerodynamiceffectsoftwotypesofcavityontheNACA0018airfoil.Byflow visualization experiments, they observed vortexes instabilities which produce oscillating forces. Donellietal.(2011)numericallyinvestigatedtheeffectsofapplyingsuctionalongtheinternalpointsof cavity on the stability of the trapped vortex. They used Fluent software for the numerical simulation. Gregorio and Fraioli (2008) studied the effect of using circular cavity on a thick blunt trailing-edge airfoilwithflowsuction,injection,orcombinationofthetwo.Thepressuredistributionalongtheairfoiland its cavity was obtained using PIV technique. The results showed that using cavity without any suction and injection causes the vortex not to be trapped and again, vortex shedding phenomenon to be generated. However, using suction or injection could control the flow separation in a specified range. Thaodoetal.(2010)studiedflowpastablunt-edgedtwo-dimensionalNACA0015sectionandthesame section with various base cavity shapes and sizes at high Reynolds numbers using the unsteady Reynolds- averagedNavier–Stokes(URANS)approachwiththerealizablek–eturbulencemodel.Itisobservedthatthe size of the cavity has more influence on the periodic trailing-edge flow than its shape does. ZhdanovandEckelmann(1994)studiedtheeffectofacavityattherearedgeofarectangularplanebody with a semicircular leading edge and a blunt rear edge. The frequency of the vortex behind the model was measured by a single-wire probe placed in the external region of the wake and was recorded by a Nicolet 660frequencyspectrumanalyzer.Theresultsshowthattheuseofthecavitydecreasesvelocityfluctuations. Kotsonisetal.(2014)studiedinfluenceofcirculationaroundasymmetricairfoilwitharoundedtrailing edge. Flow control is achieved by the use of dielectric barrier discharge plasma actuators placed at the Flowvisualizationoverathickblunttrailing-edge... 697 trailing edge of the airfoil. Time-resolved particle image velocimetry is used to elucidate the topology and dynamicalresponseofthewakeflowundertheinfluenceofactuation.Flowfieldmeasurementsindicatethe successful manipulation of the Kutta condition enabled by the plasma actuator. The actuator enhanced the mixing of the wake near the trailing edge while reducing the dominant shedding frequency. ShannonandMorris(2006)studiedthevelocityfieldinthenearwakeregionofanasymmetricbeveled trailing edge to determine the flow mechanisms responsible for the generation of trailing edge noise. Two component velocity measurements were acquired using particle image velocimetry. The small scale tur- bulence was found to be dependent on the phase of the vortex shedding process implying a dependence of the broadband sound generated by the trailing edge on the phase of the vortex shedding process. Inthispaper,first,theRisoairfoilwithasharptrailingedgeisexperimentallystudiedbyPIVtechnique to obtain the flow field around it. Flow structure and vortex dynamics after the separation point are investigated. Then, by cutting off the trailing edge, a thick blunt trailing-edged airfoil is created and experimentallystudiedbyPIVtobecomparedwiththeresultsoftheoriginalairfoil.Finally,abasecavityis addedtothebackofthethickblunttrailing-edgedairfoilanditseffectsontherelevantflowparametersare obtained. 2 PIV test setup Anopencircuitblowerwindtunnelisusedwithacross-sectionalareaof8 9 8 cm2andlengthof1 m.This wind tunnel consists of an air suction blower, a motor from Siemens Company, an inverter from Siemens Company,anairinletduct,andahoneycombmesh.Thehoneycombmeshisusedforestablishingauniform airflowandtheinverterforchangingtheacmotorspeedwhich,inturn,changestheflowvelocity.Theflow viscosity is 3.6 9 10-5 (kg/m s) and free stream velocity changes from 1 to 3 (m/s). Restriction of increasing laser power, camera frame rate, and rate of particles is reason for selection of the test settings, such as flow rate and velocity of free stream. The PIV method is implemented via a laser-wave company continuous green light laser (wave- length = 532 nm;power = 8 W)andahigh-speedPCOcamerausedformeasuringtheflowvelocity.The laser works at continuous mode. Cool water vapor droplets with the rate of 0.5 kg/h and the size of 1–3 microns are uniformly injected into the air. The density of particles is about 3–6 per 8 9 8 pixel2 interrogation area for different measurements. The high-speed PCO camera (Model 1200HS) is used for taking photos of these droplets at a shooting speed of 1800 frames per second. Here, the shooting camera speed is obtained based on flow velocity in the range of 1–3 m/s and maximum search window is equal to 64 9 64 pixel2 with 50% overlap. In other words, particle displacement during the time interval between two consecutive images should be less than about 50% of the search window. Here, 50% of the search windowisequalto32 9 32pixel2inwhicheachpixelisequivalentto50microns,andconsequently,50% of the search window becomes equal to 1.6 9 1.6 mm2. Therefore, maximum velocity becomes equal to 2.88 m/s which is obtained based on particle displacement between two consecutive images (1.6 mm) divided by its time interval (1800-1 s). The laser power is obtained by trial and error, and should be enough to light the particles during the exposure time of the camera that is less than the inverse of camera speed (fps). A VIS–NIR coated cylindrical lens from Edmund Company is placed atthe front ofthe laser to project the laser light onto a plane. The effective focal length of the lens is 20 mm and its dimension is 12.5 9 25 mm2. Figure 1 shows the schematic of the test rig of PIV. TheRisoairfoilwithachordlengthof5 cmandspanof8 cmisplacedinsidethewindtunnelwiththe blockage ratio of 12.5%. Maximum thickness of the airfoil is 1 cm. To produce a thick blunt trailing-edge airfoil, the trailing edge of the Riso airfoil is cutoff at a distance of 30% of chord length from its trailing edge. In the last stage, two 2 mm plates are connected to the end of the thick blunt trailing-edge airfoil to formabasecavitybehindtheairfoil.Figure 2showsthementionedprofilesusedinthisstudy,including:(1) the Riso airfoil, (2) the thick blunt trailing-edge airfoil, and (3) the thick blunt trailing-edge airfoil with a base cavity profile. Inthisstudy,theoriginalairfoilisplacedinsidethewindtunnelandflowpatternsbehindthemodelare measured at different Reynolds numbers and different angles of attack via PIV technique. Then, the same process is repeated for the thick blunt trailing-edge airfoil with and without the base cavity, respectively. Sincetheflowislaminar,thetimescaleoftheflowpatternisthecordlengthdividedbytheflowfreestream 698 G.Taherianetal. Fig.1 SchematicofthetestregioninPIV Fig.2 aRisoairfoil,bthickblunttrailing-edgeairfoil,andcthickblunttrailing-edgeairfoilwithabasecavityprofile velocity.Inthisstudy,theflowvelocityis1–3 m/sandthecordlengthis50 mm,sothetimescaleis0.05 s which is one order of magnitude larger than the imaging time sequence. 3 Image processing In this research, velocity vectors are obtained by the PIV lab1.32 software in which FFT algorithm is used for image processing. This algorithm has a high accuracy, especially for flow fields with rotating and stretching movements as well as pure translation. In fact, FFT is a multi-stage method in which the interrogation window is reproduced at each stage to be capable of capturing rotating and stretching movements (Gilbert and Johnson 2003). Here, 32 9 32 pixel2 interrogation window for the first stage, 16 9 16forthesecondone,and8 9 8forthelastoneareused.Ineachstage,50%overlapisapplied;thus, the computation error is 1 pixel. By considering the blockage length of the model for image calibration, each pixel is equivalent to 50 microns. 4 Uncertainty and validation UncertaintyinPIVmethodcanbedividedintotwomainparts:first,thegeneralerrorsduetotracerparticles lag and their not completely following from the flow pattern, and the second one, the error of image processing. The uncertainty of PIV image processing methods was investigated by McKenna and McGillis (2002). They noted that the uncertainty increases by increasing the size of interrogation window. On the other hand, increasing the size of interrogation window causes the spatial resolution of velocity field to decrease. Their calculations showed that for 32 9 32 pixel2 interrogation window and FFT algorithm, the Flowvisualizationoverathickblunttrailing-edge... 699 Fig.3 Flowpatternbehindthecylinder Table1 ComparisonofStrouhalnumberbetweenthisworkand(Rahmanetal.2007) Error% Strouhal(Rahmanetal.2007) Strouhal(currentwork) Reynoldsnumber 0.9 0.218 0.22 1500 1 0.2178 0.22 2000 2 0.2176 0.222 2500 uncertaintyofvelocitymeasurementis1.93%(McKennaandMcGillis2002).ToverifythePIVtestresults, StrouhalnumberiscalculatedfromPIVtestresultsatthreeReynoldsnumbersof1500,2000,and2500,and then compared with the results of Rahman et al. (2007). Figure 3 shows the raw image captured from this experiment. Table 1 shows that the maximum error is 2%. The error is calculated from the following equation: Strouhalðcurrent workÞ(cid:3)StrouhalðRahman etal:2007Þ Error%¼ : ð1Þ Strouhal ðcurrent workÞ In addition, the experiments were repeated for three times and results show that the difference between them is less than 5%. 5 PIV results for the original airfoil First, the flow around the original airfoil was studied. Next, the flow around the airfoil with a thick blunt trailing-edgewascomparedwiththataroundtheoriginalairfoil.Finally,theeffectofaddingthebasecavity to the thick blunt trailing-edge airfoil was studied. Figure 4showstheinstantaneousvelocityvectorsobtainedduringaspecifictimeintervalfortheoriginal airfoil. The temporal distance between two consecutive images is 0.007 s. The selected time sequence should be enough to encompass the period of flow and to show the entire of flow pattern. Since theflow is laminar, the time scale of the flow pattern is the flow length scale which is the cord length divided by the flowfreestreamvelocity.Inthisstudy,theflowvelocityis1 m/sandthecordlengthis50 mm,sothetime scaleis0.05 swhichisoneorderofmagnitudelargerthantheimagingtimesequence.Angleofattackand Reynolds number are 5(cid:3) and 2150, respectively. A dark region is created under the airfoil resulting from refraction as laser light hits the airfoil, rendering image processing and velocity measurements impossible within this region. The red regions actually show the airfoil and its underneath area excluded from the images. Analysisofthedarkregionundertheairfoilcanbemadepossiblebyprojecting anotherlaseronto thisregion.Asobservedinthisfigure,theflowexperiencedseparationuponpassingovertheairfoil,creating a recirculation region with vortex shedding on the airfoil. First, a vortex is formed within the separation region neartheupper shear layer. This vortex rotates in thecounterclockwisedirection and markedwith a whitearrowintheimages.Initially,thevortexseparatedfromtheshearlayerandmoveswiththeflow,and then, it disappears within a short time after passing the airfoil. However, while the first vortex is 700 G.Taherianetal. Fig.4 Instantaneous velocity vectors over the original airfoil at various instances at Re=2150, U=1m/s, and attack angle=5(cid:3) disappearing, a second vortex rotating clockwise (the red arrow) starts to form at the trailing edge near the lower shear layer. This vortex also separated from the lower shear layer and moves with the flow before disappearing completely. The direction of motion of this vortex is perpendicular to the airfoil surface. The abovephenomenoniscontinuouslyrepeatedataspecificfrequencyandiswhatwecall‘‘vortexshedding’’. Flowvisualizationoverathickblunttrailing-edge... 701 Fig.5 Instantaneousvelocityvectorsoverthethickblunttrailing-edgeairfoilatvariousinstancesandRe=2150,U=1m/s, andattackangle=5(cid:3) 6 PIV results for the thick blunt trailing-edge airfoil Keeping Reynolds number and attack angle constant (Re = 2150 and attack angle = 5(cid:3)), the PIV tests are performed again for the thick blunt trailing-edge airfoil. Figure 5 shows the instantaneous velocity vectors obtainedforthiscase.The time framebetweentwo consecutive imagesis0.0056 s.As seen,noseparation occursontheairfoilsurface,andonlytwovorticesareformedbehindthickblunttrailingedge.Infact,using this thick blunt trailing-edge airfoil causes the location of the separation region to move from the airfoil surface to the back of the airfoil. No further vortex shedding is observed at this point and, as said before, only two vortices are formed at the upper and lower parts of the airfoil with counter clockwise and clockwise rotations, respectively. FlowstudyathigherReynoldsnumber(Re = 3220)onthethickblunttrailing-edgeairfoilshowsthatas the Reynolds number increases, the vortices (unlike the Re = 2150 case) no longer remain in a constant 702 G.Taherianetal. location,butmoveataspecificfrequencyalongtheflowasexplainedhere.First,acounterclockwisevortex isseparatedfromtheuppershearlayerandstartsmovingwiththeflowbeforedisappearingtowardstheend oftheseparationzone.Asecondvortex(rotatingclockwise)isimmediatelyseparatedfromthelowershear layer, again moving with the flow until it vanishes. This is periodically repeated at a specific frequency. Figure 6 shows the instantaneous velocity vectors for a vortex shedding time period at Re = 3220 and attack angle of 5(cid:3). The time between two consecutive images is 0.009 s. As observed, in the first two images, theuppervortex,rotatingcounterclockwise,ismoving inthesamedirectionasthefreestream.In thethirdimage,thisvortexisdisappearingandasecondvortexisbeingformedatthelowerpart,rotatingin aclockwisedirection.Inthefourthimage,thelowervortexisfullyformedandismovinginthedirectionof the free flow. In the fifth image, this vortex is also disappearing and a third upper vortex is being formed. Finally,thelastimageshowsthatthelowervortexfullyvanishesandtheuppervortexisfullyformed,and, thus, the cycle is completed. 7 Velocity distributions behind the trailing edge In Fig. 7, five sections at the back of the airfoil trailing edge have been specified along which velocity distributions are plotted. These five sections are located at 10, 20, 30, 40, and 50% of the chord length (5 cm) from the trailing edge. Figure 8showsthemeanvelocitydistributionsalongthefivespecifiedsectionsattheattackangleof5(cid:3) andtheReynoldsnumberof3070.Themeanvelocitydistributionsareaveragedbytakingthemeanvalues forseveralvortexsheddingperiods.Thecoordinateorigincoincideswiththeupperpointofthediagram.As can be observed, at the initial sections (10 and 20% of the chord), there are two minimum points in the diagram. These points appear due to the formation of two vortices within the separation region. These vortices are the same as those formed during the vortex shedding (in the previous section) which occurred near the upper and lower shear layers. As we move further from the trailing edge, the effects of these two vorticesdiminishandthetwominimumpointsareconsequentlymerged.Asthedistanceofthesectionfrom theendpointisfurtherincreased,thisrecirculationregiongraduallyvanishes.Thecloseregionbetweenthe black line (1 m/s) and the lower part of the velocity curves shows the recirculation or wake region. InFig. 8,thereisamaximumpointabovetheairfoil.Asshowninthisfigure,theflowvelocityincreases uptoacertainpointandthendecreases.Ascanbeseen,suchapointdoesnotexistundertheairfoil.Asthe flow is separated on the airfoil, the free flow above the shear level is accelerated and flow velocity at the uppershearlayerincreases.Therefore,themaximumpointisduetothisaccelerationintheproximityofthe upper shear layer. This phenomenon can be explained as follows: the flow first accelerates while passing over the upper airfoil surface before encountering a positive pressure gradient. However, upon flow sepa- ration,fullpressurerecoverydoesnotoccur.Therefore,weexpecttheflowvelocityintheupperpartofthe separated region to exceed that of the free air flow velocity. The same situation does not exist under the airfoil, since velocity does not increase on the lower surface of the airfoil. Figure 9showstheflowvelocityprofilesatvarioussectionsofthethickblunttrailingedgeairfoilforthe Reynolds number of 2150 and attack angle of 5(cid:3). The chord length for this airfoil is 3.5 cm. As seen, the maximumvelocitypointthatexistsinFig. 9nolongerappearsandtheflowvelocityremainsconstantupon reaching its maximum value. As was mentioned above, the flow over the thick blunt trailing edge airfoil, positionedata5(cid:3)angleofattack,doesnotundergoseparationandthisiswhyamaximumvelocitypointis not observed in this case. As the angle of attack increases, we observe a recurrence of separation and the reappearing of this maximum velocity point. Comparison between Figs. 8 and 9 shows that by getting furtherfromthetrailingedge,thewakeregionbehindtheblunttrailingedgeairfoildecreasesmorethanthat behind the original airfoil. It indicates that the wake region behind the thick-edged airfoil is disappearing faster than that behind the original one. 8 PIV results for the thick blunt trailing-edge airfoil with base cavity As mentioned above, deploying a base cavity is one way of reducing drag in an airfoil with a thick blunt trailingedge.Inthisstudy,weaddabasecavitytotheairfoiluponthickeningitsedge,andstudytheeffects thereof on the flow over the airfoil. Flowvisualizationoverathickblunttrailing-edge... 703 Fig.6 Instantaneous velocity vectors for flow over the thick blunt trailing-edge airfoil at various instances for Re=3220, U=1.5m/s,andattackangle=5(cid:3) 704 G.Taherianetal. Fig.7 Sectionsbehindtheairfoilformeasurementofflowvelocitydistribution Fig.8 Velocitydistributionatvarioussectionsoftheoriginalairfoilata5(cid:3)angleofattackandU=1m/s Figure 10showstheinstantaneousvelocityvectorsforthethickblunttrailing-edgeairfoilconnectedtoa base cavity (angle of attack = 5(cid:3) and Re = 2150). Similar to the case without base cavity, no separation occurshereeither.Itisobviouslyobservedherethattheseparationregionbehindtheairfoilhaschangedits shape into a pointed triangle, and that the separation region is much smaller than that observed when no cavity was used. In the absence of a cavity, two distinct vortexes appeared behind the thick trailing edge, whereas here, no such vortices are visible. Duetotheuseofdark-coloredblades,theflowinsidethecavitycouldnotbeobserved.Thereasonwhy darkbladesareusedisthattheypreventlightreflectionwhichwoulddistorttheimages.Thus,intheabsence of direct observation from inside the cavity, we can conclude that the two vortices move into the cavity when a base cavity is connected to the airfoil. As the Reynolds number increases, the momentum also increases, leading to the instability of these vortices.Thus,thevorticesexitthecavityandbecomeobservable.Althoughthesevorticesaremuchsmaller thanthoseobservedwhennocavityisused,theyare,nevertheless,observableandmoveperiodicallyalong theflowline.Figure 11showstheinstantaneousvelocitiesforoneperiodofoscillation.Inthefirstimage,a vortex is being formed at the upper edge, and in the second image, this vortex is being separated from the