THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 345 E. 47th St., New York, N.Y. 10017 98hGT-481 The Society shall not be responsible for statements or opinions advanced In papers or discussion at meetings of the Sodety or of its Divisions or Sections, or printed in its publications. Discussion is printed only it the paper Is published in an ASME Journal. Authorization to photocopy for internal or personal use is granted to libraries and other users registered with the Copyright Clearance Center (CCC) provided S3/article or $4/page is paid to CCC, 222 Rosewood Dr., Danvers, MA 01923. Requests for special permission or bulk reproduction should be addressed to the ASME Technical Publishing Department. All Rights Reserved Copyright 0 1998 by ASME . Printed in U.S.A. III IlIlIllIllIllIll 111 111 BREAK SEPARATED-FLOW TRANSITION PART 1 - EXPERIMENTAL METHODOLOGY AND MODE CLASSIFICATION Ansa Hatman and Ting Wang Department of Mechanical Engineering Clemson University Clemson, SC 29634-0921, USA ABSTRACT This first part of the paper presents techniques implemented for the experimental investigation of the transition mechanism in 2-D pressure driven separated boundary layer flows over a flat plate at inlet free-stream turbulence intensities ranging from 0.3 to 0.6% and imposed adverse pressure gradients ranging from K = - 0.68 x10 6 to - 6.25 x10-6 . The structure and behavior of the separation bubble were investigated for various flow conditions. The separated•flow transition modes were identified and classified. The distribution and strength of the adverse pressure gradient were obtained by varying the test section outer wall divergence angle. Specific methods identifying the main parameters that characterize separated-flow transition are introduced and issues regarding measurements of reverse flow are discussed. The methods implemented in determining the separation point, maximum displacement location, the unsteady reattachment region, the start and end of transition, etc. are described. NOMENCLATURE Cp En(f) H12 = 1 - ( 1J.../1J...0)2; pressure coefficient I-D power spectrum of u frequency shear layer thickness = 81/82 ;shape factor .(WU.,2 )(dUeddx): pressure gradient parameter 1-8 geometrical pressure gradient parameter, eqn. (2) local pressure gradient parameter, eqn. (3) = (41. - ts); separation bubble length Rex .U.,x/v ; local Reynolds number Rea .U...,LTN; transition length Reynolds number Reg, .U.,81/v ; displacement thickness Reynolds number Re81 Re g2 .U.,82/v ; momentum thickness Reynolds number KO Kim& Regax =Re82T Re82r integral time scale lii free-stream turbulence intensity, u71.1.„ mean velocity in streamwise direction u' r.m.s. streamwise velocity fluctuation V mean cross-stream (normal) velocity v' r.m.s. cross-stream velocity fluctuation -uv Reynolds shear stress YD Yo coordinate in streamwise direction dividing streamline zero velocity line Clack 8 81 62 O • • X ▪ • boundary layer thickness displacement thickness momentum thickness divergence angle separation angle turbulent intermittency integral length scale dissipation time scale kinematic viscosity Utica= LT max MD • RI u'ron - Real t 0 = (xi - ic,); transition length peak value maximum displacement location reattachment first reattachment (long bubbles) separation onset of transition end of transition mid-transition point, where us reaches maximum local free stream conditions inlet conditions INTRODUCTION Presented at the International Gas Turbine & Aeroengine Congress & Exhibition Stockholm, Sweden — June 2—June 5,1998 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/a To predict the aerodynamic and heat transfer characteristics of a body immersed into a flow stream, the boundary layer development along its surface must be accurately determined; this development is influenced by the laminar-to-turbulent transition, which causes a drastic increase in both skin friction and convective heat transfer. In engineering applications, extended transitional boundary layers can be encountered. For example, it is common that 20 80% of the boundary layer developed on the gas turbine blades or vanes are in transitional regime. Especially in the low pressure turbine of an aircraft engine crusing at high altitudes, when the air density is reduced and the Reynolds numbers become very low, extended transitional boundary layer is a norm, and boundary layer separation may occur. Accurate predictions of both aerodynamic and heat transfer performance are important for turbine design. Transition can be affected by different parameters such as pressure gradients, free-stream turbulence intensity, acoustic waves, flow unsteadiness, surface vibration, surface curvature, surface roughness, wall temperature, etc. Each of these parameters can influence the start and length of the transition and the structure of the transitional boundary layer. A generally accepted classification (Mayle, 1991) shows that there are four important modes of transition in a boundary layer: natural transition, bypass transition, separated-flow transition, and reverse transition. The natural transition begins with weak instabilities (infinitesimal disturbances). Through the amplification of the TollmienSchlichting waves and nonlinear vorticity breakdowns, it becomes a fully turbulent flow. The bypass transition is caused by large (finite) disturbances in the external flow (e.g. high free-stream turbulence intensity) or on the surface (e.g. roughness)• and completely bypasses the Tollmien-Schlichting mode of instability. For large streamwise adverse pressure gradients, the formation of a transitional separation bubble is likely to occur. The detached laminar shear layer is inherently unstable and promotes the growth of disturbances leading to an earlier laminar-to-turbulent transition. Then, turbulent mixing due to the increased entrainment in the rapidly growing turbulent part of the separated shear layer causes the shear layer to reattach. The separated-flow transition may or may not involve linear instability of the Tollmien-Schlichting type. At high Reynolds numbers, the extent of such a separation bubble is small, on the order of 1% of the chord (Tani, 1964), and the slight change in the pressure distribution produced by the dead air region has a negligible effect on the forces acting on the surface. With an increase of incidence angle or a reduction in speed or Reynolds number, the shear layer may fail to reattach, and the short bubble may burst to form either a long bubble, or an unattached free shear layer. The pressure distribution associated with a "long" bubble is different from that of the inviscid flow, and the forces acting on the airfoil are therefore changed. Bubble bursting creates a significant increase in drag and an undesirable change in pitching moment. Separation is the generic name given to a broad class of flow phenomena. The most important feature characterizing this class is that the flow becomes detached from the body surface allowing a region of reverse flow to develop between the body and the separated shear layer. Fig. 1 presents a conventional view of a pressure driven separated boundary layer development. A dividing streamline between the shear layer and the recirculating region starts at the separation point x s and makes an angle y with the surface. The height of the separation region, yD, increases almost linearly downstream of the separation point, reaches a maximum Transition Turbulent .1.11 0.1 4 separated boundary layer free-stream velocity distribution Laminar , unseparated bo ndary layer free-stream velocity distribution XR dividing streamline ge of separated separated boundary layer shear .,,,a11141111110111111 lay w ge of unseparated bound la er ano reverse flow vortex zero velocity line turbulent reattachment dividing streamline Fig. 1 Conventional view of a 2-D laminar boundary layer separation: (a) streamwise free-stream velocity distribution; (b) boundary layer development; and (c) velocity profiles along the separation bubble. above the center of the reverse flow vortex and collapses with the reattachment process. Due to this actual displacement of the shear layer, the rate of growth of the separated boundary layer thickness 8(x) is considerably higher than that of an attached boundary layer, and the free-stream velocity distribution is locally modified. In a decelerated boundary layer, the retarded flow in the boundary layer does not posses sufficient momentum and energy along the body surface to overcome the pressure rise downstream, and finally, the fluid particles near the wall are brought to rest. The main stream, itself decelerating, is unable to energize the fluid in the boundary layer, and an almost stagnant fluid region develops near the wall. The abrupt thickening of the near-wall reverse flow region accompanied by significant values of the velocity component normal to the wall leads to the reverse-flow-vortex breakaway and subsequently to a strong interaction with the free-stream flow. Separation provides a mechanism whereby vorticity can be transported into the inviscid flow field; whereas, in attached flows the vorticity is confined within the boundary layer. There are too many published studies related to boundary layer separation to be cited here. Empirical data on the formation of the transitional separation bubbles, under different experimental 2 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/abou F.> conditions, have been obtained by many authors, such as: Gault (1955), Caster (1967), Horton (1968), Malkiel and Mayle (1995) or Hazarika and Hirsch (1995), etc. The influence of the free-stream turbulence intensity on separated flow behavior was studied by Roberts (1980). Significant computational work on separated flow transition was undertaken by Smith (1987), Brown et al. (1988), and Pauley et al. (1990), to name a few. However, not many studies specifically investigated the interaction between separated flow and laminar-turbulent transition. Therefore, to improve the physical understanding of the separated-flow transition process, detailed measurements were performed in two-dimensional boundary layers over a flat plate at zero incidence. The main work focuses on investigating the complex interaction between laminar-turbulent transition and flow separation. The results of this study will serve as a reference for future studies under more realistic gas turbine conditions, such as on airfoil surfaces under elevated free-stream turbulence. x Flat Plate Test Surface Outer Wall Fig. 2 Test section geometry Table 1 Synopsis of experimental flow conditions U.,0 IK671 xpl iml 1 xD2 im) _ _a r iisj_ 1 BASELINE 20.55 EXPERIMENTAL PROGRAM To investigate the effects of the streamwise pressure gradients on transition, the nondimensional pressure gradient parameter: (1) K= (v/U2)(dUadx) was used to characterize the experimental flow conditions. The desired K distribution was obtained by adjusting the geometry of the test section in the wind tunnel. Theoretically, by assuming onedimensional inviscid flow and ignoring the effect of the boundary layer growth, the free-stream velocity distribution can be derived from mass conservation. A constant value of K parameter can be kept simply by setting the flexible outer wall of the test section into a wedge shape. When the boundary layer separation occurs, the resulting pressure distribution is very different from the theoretical one. For constantK favorable pressure gradient and low adverse pressure gradient cases without separation, the fine adjustment of the outer wall for eliminating the effect of the boundary layer growth can be applied. For separated boundary layers, the adjustment for a constant-K flow is no longer feasible. Consequently, the configuration with the outer wall straight in a wedge-shape was adopted for all the experimental cases. The pressure gradient parameter K evaluated from the test section geometry is designated as the geometrical pressure gradient parameter KG and it is determined by the value of the inlet freestream velocity, the test section outer wall divergence angle 0, and the inlet test section width L: (2) KG = tan / UL. CASE 1.1 CASE 1.2 CASE 1.3 CASE 1.4 CASE 1.5 CASE 1.6 CASE 1.7 CASE 1.8 CASE 1.9 16.25 15.90 8.25 5.30 8.85 9.50 4.90 4.06 4.39 CASE 11.1 CASE 11.2 CASE 11.3 CASE 11.4 CASE 11.5 16.20 11.60 7.05 3.55 2.38 1 Series 1 -0.91 0.7 -0.69 0.48 -1.26 0.23 -2.07 0.48 -1.24 0.48 -1.16 0.70 -2.13 0.23 -2.57 0.23 -3.38 0.70 Series 11 -0.92 0.70 -1.28 0.70 -2.10 0.70 -4.17 0.70 -6.22 0.70 0 [o] 0 0 8.1 6 5.7 6 6 8.1 5.7 5.7 8.1 1.60 1.60 1.60 1.60 1.60 8.1 8.1 8.1 8.1 8.1 The parameters were selected based on the results obtained in the preliminary studies, so that transitional separated boundary layers could be produced for various types of bubbles. A systematic method of generating various types of transitional separated boundary layers was applied by varying the free-stream velocity and the outer wall divergence angle. The flow rate was adjusted through a variable frequency motor controller. To avoid the wind tunnel unsteadiness in the low-speed cases, an additional inlet fan damper was used to control the flow rate. Table 1 shows the geometrical parameters for the baseline case (zero pressure gradient) and the fourteen separated boundary layer experimental cases. The tests in Series I were conducted in a flow configuration with constant divergence angle from xrn to the end of the test section. For the tests in the Series II, the adverse pressure gradient was reduced downstream of x02. For each experimental case, boundary layer single hot-wire measurements were taken on a test surface that was instrumented with pressure taps for pressure distribution measurements. Among the many tested conditions, three representative cases. CASE 1.9, 11.1, and 11.3, were selected to be repeated using the cross-wire of a three-wire probe for detailed turbulent shear stress measurement. The behavior of separated flows is sufficiently different from unseparated flows to warrant special attention to experimental techniques and the data reduction procedure used. Flow qualification studies must be carefully done to assure that the time averaged properties of the flow field are satisfactorily two-dimensional in KG is a convenient parameter to be used for indicating the global pressure gradient strength, however the actual K deviates locally, more or less, from the geometrical determined constant value. To control the relative locations of the separation point and the onset of transition, the laminar boundary/ layer was allowed to develop in zero pressure gradient conditions from the leading edge to some location, xDI, as shown in Fig. 2. The adverse pressure gradient was applied downstream of this location. For several sets of data, the effect of the flow conditions at reattachment on the separation bubble's structure was also investigated by suppressing the wedge angle downstream of xD2 location to reduce the adverse pressure gradients. The experimental flow conditions for each case were obtained by combining different overall streamwise pressure gradients 1CG (different Qand U.„0) and different locations for the start and the end of the adverse pressure gradient region, xDi and xD2. 3 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/a the test section. The two-dimensionality was checked by evaluating spanwise agreement of wall pressure measured at several streamwise locations. The experiments were conducted at low free-stream turbulence intensity, with the inlet value varying from Tu = 0.3 at high velocity cases to 0.6 % at lowest velocity case. Ambient conditions that included temperature, barometric pressure and humidity were also measured for corrections of air properties. Heating and Cooling Circulating System Honeycomb thid Test Facilities A 2-D, open circuit, blowing type wind tunnel shown in Fig. 3 was used. Before entering the test section, the air was drawn through a filter box, forced through two grids, a honeycomb, a heat exchanger, a screen pack and a contraction nozzle. The flow rate was adjusted by a variable frequency motor controller and inlet damper to obtain wind tunnel speeds from 1 m/s to 25 m/s. The vertical test wall is a rectangular surface 2.4 m long and 0.92 m high. The test section had a width of 0.15 in and a large aspect ratio of 6 that helped reduce edge effects and achieved twodimensionality for the boundary layer flow in the center-span of the test wall. The boundary layer suction was applied through a double slot bleeding at the ellipsoidal shaped leading edge of the test wall, so a near zero thickness boundary layer could be achieved at the theoretical x = 0 streamwise location. A longitudinal slot was cut along the centerline of the test section's outer wall that allowed the velocity measurements to be obtained by traversing the hot-wire probe along the test wall centerline through the slot at any streamwise location. Other than the measuring location, the remaining slot was covered with rubber bands. The test surface was instrumented with 48 pressure taps, 1 mm in diameter and flush with the surface. The pressure taps arrangement is shown in Fig. 4. Flow Rate Heat Control Exchanger Valve Fig. 3 Plane view of the low-speed 2-0 boundary layer wind tunnel facility Flow direction Equal spacing: 50 Equal spacing: 30 fc&gicpacing: 25 Equal spacing: 30 Equal cpacirtg' 50 ual spacin :150 ,20011' Instrumentation The boundary layer measurements were obtained using hot-wire anemometry. First, a single hot-wire was used to measure the mean streamwise velocity and ruts values for all 15 cases investigated. A single TM hot wire, model 12185-TI.5, operating in constant temperature mode at an overheat ratio of around 1.8 was used. The single hot-wire measurements served as a guidance for the multiplewire measurements because the single wire could better approach the wall, to within 0.1 nun. The cross-stream Reynolds stresses and the cross-stream velocity component were determined for three cases using the 2.5 gm tungsten cross-wire arrangement of a custom-designed three-wire probe, which is smaller than a commercial one (the spacing between the 'X' array is 0.3 mm). The third wire of this probe is a 1.2 gm platinum wire, which will be used for future temperature measurements in similar separated flow conditions. The details regarding the development and qualification of the three-wire sensor were presented by Shome (1991) and Wang et al. (1996). Attention was paid to the calibration procedure, especially in the low-velocity range below 1 m/s, where the calibration equation shows large deviation from King's Law. A polynomial calibration equation was used to reduce the uncertainty of low velocity measurements in the near-wall region and at the interface between the shear layer and the separation bubble. The hot-wire probes were positioned by a traversing equipment that consisted of a stepper motor mounted on a slide and was controlled by the data acquisition program. For all the probes, a TSI Model IFA 100 system was used as a constant temperature anemometer. A sampling frequency of 2 kHz for 20 seconds acquiring time was used. In the preliminary studies, it was determined that the mean and fluctuating velocity components maailwiwww)66 6' „ tow 0350 500 650 All holes- 4) Ini R011 1100 All dimensions are in mm F g.4 Layout of static pressure taps on the test surface converged to within I% for sampling durations of about 10 seconds even for very low flow speeds. The wall static pressure measurements were performed using the pressure taps on the test surface, which were connected to a pressure transducer through a 48-channel Scanivalve system. Pressure Measurements Time averaged static pressure distribution provided a way of determining the mean free-stream velocity distribution in the test section. Surface pressure measurements were used to roughly determine the bubbles location and its extent. The static pressure measurements were used to evaluate the pressure coefficient: C p=2(p-p.,)/(pU.,02) and the pressure gradient parameter K = (v/I.J... 2)(dUn/dx). The difficulty in determining the pressure gradient parameter K resides in the proper evaluation of the velocity gradient (d1.1 0„/dx) from discrete data points. Besides the geometrical pressure gradient parameter, KG, given by equation (2), a local pressure gradient parameter. Kieed, given by equation (3), was defined. The local value of the pressure gradient parameter at any location situated between two consecutive measured locations x(i-l) and x(i) was determined as: Kic,cat = v (U(i) - U(i-1)1 / [x(i) - x(i-1)) / U(i) / U0-0 (3) 4 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/about K w"! follows the local rapid pressure variations associated with the separation and reattachment processes, while the K G g eometrical parameter serves as a reference for the comparison between different cases. Velocity Measurements For all fifteen cases investigated, the mean velocity (U) and the streamwise velocity fluctuation (u') were acquired using a single hot-wire probe. Some specific problems arose for the hot-wire measurements in the recirculatin g region of the separated flow because the hot-wire sensor is not directionally sensitive. Since in the reverse flow and reattachment re gions of the separated flows instantaneous flow direction is changing, the hot-wire sensor is not an accurate velocity measuring tool. The hot-wire sensor's insensitivity to the flow direction would result in distorted velocity profiles, as shown in Fig. 5. The actual magnitude of the reverse-flow velocity could not be accomplished by simply inspectin g the profiles obtained with a single hot wire. The region containing reverse flow is unstead y in both streamwise and cross-stream directions; this results in a finite minimum value at the location (y = yo) of the theoretical zero velocity value. Therefore, even if the hot wires are not sensitive to the flow direction, the near-wall local minimum could be treated as the average zero velocity location yo. Due to the unsteadiness of the reverse-flow region, and the possible rectifyin g effect of the hot-wire measurements in the low speed flow, the velocity values below yc, are subject to large uncertainty. Since in this study the recirculation region was very thin, a more accurate measurement, such as using the flying-wire technique was difficult and was not employed. The uncertainty in the mean velocity profiles near the wall makes any indirect method unreliable. As a result, the boundary layer integral parameters could not be accurately determined, and the classical Causer procedure for determinin g the skin friction coefficient of an attached turbulent boundary layer can not be satisfactorily used in the separated flow re gion. Furthermore, it is likely that such a probe disturbs the flow to some degree. Mandel (1994) observed that a single wire probe noticeably chan ged the velocity distribution only when it was positioned very close to the separation location. In this case, the free stream was observed to accelerate about 5%, and the size of the bubble decreased about 2%. At no other location was the flow significantly sensitive to the sensor intrusion. It must be mentioned that in Malkiel's experiment, the separation bubble dimension was one order of magnitude smaller than the bubble's dimension analyzed in the present study. Therefore, it was reasonable to expect that the present results were less affected by the separated-flow sensitivity to the sensor intrusion than those of Malkiel and Mayle (1995). 8 a) (b) Separated boundary layer mean velocity profile: (a) typical velocity profile for separated boundary layer flow; (b) velocity profile measured using hot-wires. YD UN. dy 0 the resulting displacement thickness can be derived in the following manner: 8 1 =58 (1 — UM.) dy =IY° (1 — UN.) dy 8(1 0 = y D — IYD (U/U.,) dy + (1 — U/U.) dy = yp + f8 (I — U/U.) dy YD The Csplacement thickness is therefore at least as large as the recirculation region, which may be greater than the thickness of the separated shear layer: h = (5 - y o). The momentum thickness of the entire velocity profile, including the zone of recirculatin g flow, results primarily from the separated shear layer. Since (U/U.) is on the order of 10-2 in the zone of recirculating flow, it can be neglected, and the momentum thickness can be estimated as: , 62 d WUJ I — UN.) dy =. 1YD WU.,01 — UN) dy + + ja UN.(1 — U/U.) dy a lb UN(1 — UN.) dy As a resul , accurate information of the velocity profile in the reverse flow region is not necessary for calculating the boundary layer integral parameters; 81 and 82 can be evaluated by considering only the portion of the velocity profile outside the recirculation region. This topic was extensively discussed by Schmidt and Mueller (1989). integral Parameters INTRODUCTION TO A NEW APPROACH IN ANALYZING SEPARATED-FLOW TRANSITION The effect of the separation bubble on velocity distribution downstream of laminar separation results in the development of a recirculation region that causes a significant chan ge in the displacement thickness. Downstream of the reattachment point, the rate of change in displacement thickness is much smaller. As shown in Fig. 5 (a), the height of the dividing streamline (ye), which determines the normal extent of the recirculating flow, is defined as the line of zero massflux. By assuming that on an average basis the net mass flow of the recirculating region is zero: The main objective of this research was to provide significant physical insight about laminar-to-turbulent transition in separated flows. At present, a reliable method for predictin g the characteristics of separation bubbles or a fully satisfac tory theory to explain the bubble bursting process is yet to be developed. Furthermore, the difference between the transition process in boundary layers which are laminar at separation and those which are already transitional at separation has not been addressed. The present approach to analyzing the separate-flow transition is to identify key parameters for describin g the phenomena of mutual interactions involved in the transition process ' in the separated shear layer. 5 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/a PRIMARY SEPARATED FLOW TRANSMON MODES Separated-Flow Transition Modes The experimental results of Hatman (1997), presented in part in this paper, led to a positive identification of three primary modes of separated-flow transition. A detailed discussion of these modes will be presented in part 3 of this paper, and the supporting data are presented in part 2. However, a brief summary is presented here, to make the data discussion easy to follow. It was observed that the separated boundary layers which separate in a laminar state behave differently from those which are already transitional or highly unstable (pre-transitional) at separation. In this study, the separation bubble types are defined by the flow regime at the separation location. A laminar separation bubble is defined as a separation that occurs in the laminar or pre-transitional flow, x s <xt. A transitional separation bubble is defined as a separation that occurs downstream of the onset of transition, x s>xt. At least three primary separated-flow transition modes were positively distinguished. In this study they are defined as transitional separation mode, laminar separation - short bubble mode, and laminar separation - long bubble mode. The location of the separation point relative to the onset of transition is used to distinguish between transitional separation and laminar separation modes. The differentiation between the separated-flow transition modes also relies on the relative locations of the maximum displacement and mid-transition points, as shown in Fig. 6. The transitional separation mode occurs when the boundary layer separation takes place at relatively high Reynolds numbers and low adverse pressure gradient strength. The onset of transition takes place prior to boundary layer separation and develops mostly as natural transition. The transitional separation bubble can be accompanied by vortex shedding. The laminar separation - short bubble mode occurs at moderate Reynolds numbers and mild adverse pressure gradients with the onset of transition induced downstream of the separation point by inflexional instability at the location coincidental with that of the maximum displacement of the shear layer. It is characterized by a quick transition completion due to a complex interaction between the separated shear layer and the reverse flow vortex. The laminar separation - short bubble mode is characterized by distinctive vortex shedding. When the laminar boundary layer separation takes place at low Reynolds numbers and strong adverse pressure gradients, the shear layer may fail to remain reattached and the laminar separation long bubble mode is likely to occur. Similar to the bubbles in the short mode, the onset of transition in the long bubble mode is also induced downstream of the separation point by inflexional instability. The high fluctuations associated with the midtransition point lead to a local reattachment-like behavior. The transition is forced to occur at low Reynolds numbers by the ejection process, but due to a deficit of turbulence production, the transition completion is considerably delayed. The laminar separation long bubble is not accompanied by vortex shedding. The passage from the transitional separation mode to the laminar separation modes take place gradually through a succession of intermediate stages dominated by one mode of transition or another. The mode of transition is mainly determined by the conditions at separation; however, the conditions at reattachment may influence the length of the separated shear layer. A shear layer in laminar separation - long bubble mode can be "forced" short when the external flow configuration is such that the adverse pressure gradient downstream of first reattachment is suppressed. transitional separation mode xt < Xs Xternax = XMD Xt INTERMEDIA1E STAGES laminar separation - dominant transitional separation mode behavio xs < xt < XMD Xu'max = XMD Xs XMD XR xthilas XT Xs dominant laminar separation mode xs < xt < MAD Xtimax = XR laminar separation short bubble mode xt = XmD XmD XR Xt Xdmax XT Xdmax = XR U I xs Xs XIND XR XlifILIX XmD XR Xt laminar separation "forced" short bubble xt - xmD laminar separation long bubble mode xt = XMD Xl1.1113X = XR1 XT = XR1 - xR /OA Xs XMD XR1 XR Xt Ulnas XT XMD XR1 XR Xt &in= XT Fig.6 Separated-flow transition modes Table 2 presents a classification and the main characteristics of the primary separated-flow transition modes and of the intermediate stages observed in the present study. In literature, the laminar short bubble mode represents the most studied case, and it is considered to be representative for the separate-flow transition. The transitional mode has been ignored or unrecognized. Mayle (1991) and Walker (1992) emphasized the limitations in predicting separated-flow transition and the necessity of more work of a fundamental nature in this area. Many existing separated-flow transition models assume "instantaneous" transition at an empirically determined "transition location" within the shear layer, which gives good results for a laminar bubble in the short mode, but completely fails to predict the transition onset of the transitional mode and the length of transition of a laminar bubble in the long mode. The present work intends to address these issues with the support of experimental results. Methodology for Establishing the Separated-Flow Transition Parameters In the present study, the separation location, the separation bubble extent and elevation were primarily judged by analyzing the mean velocity profiles, the rms velocity fluctuations and the isovelocity contours. The extent of separation was also cross-checked with a tuft mounted on a thin pin and moved along the surface. 6 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/abou 1.7 1.8 1.9 11.1 11.2 11.3 11.4 11.5 665 2.71 0.69 -101 313 12 10 8 2.43 0.18 -2186 92 3.69 0.38 -2009 291 4.84 1.17 -91 954 1.74 0.47 -4108 -215 1.56 0.36 -1737 75 2.66 0.28 -2826 104 E S 2.78 734 835 4.98 1.88 181 710 3.67 0.97 -629 546 2.28 0.14 -1985 160 1.59 0.42 -2079 -135 E E So $ o o x a x al 0 li x li x ivID R 02 4 60 24 6 0 24 6 U [m/s e E o a x E o u x 6 u x E E 6 II x at o n x R S 0 01 02 0 01 02 0 OA 02 0 CO 02 0 01 02 0 OA 02 0 01 02 (b) 12 10 E , 8 8 6 II E 6 x E S 6 II x 4 . 6.42 0246024 (a) x = 0.66M 646 11 1.45 0.001S01 h‘. 5.11 2 0 CT 1318 m 1405 E CO cii o i 2 0 12 10 8 1 E6 ;4 2 0 NOTE: The negative Resin. and Reams will be discussed in Part 2. 5 E 6 II x E Zii a il x MD 1---'-'s 0204T02 0.4 0024 (c) E 1g c; o II a x Ic- E 0 Ill 00 II a x L. 0 0.5 1 0 0.5 1 (d) S d u x R S 0 02t4 E CO CR 6° a II x t- 0 02 0.4 J 02 0.4 E •- ea? au 06 a u x L. R 0 0.5 1 0 0.5 1 0 0.5 1 F (ky - For bubbles generated at low Reynolds numbers, which were characterized by large displacement and relatively strong reverse flow vortex, the tuft method proved to be effective in determining the separation bubble extent. However, this method gave ambiguous results for the transitional separation bubbles, which were relatively thin. Complementary information that supported the decision-making was drawn from the pressure gradient and integral parameters distributions. Fig. 7 shows the near-wall region of the mean and normalized rms velocity profiles of a separated boundary layer in laminar separation - short bubble mode of transition (CASE 1.4). The separation parameters (xs, itmD, xR) can be directly determined from the mean velocity profiles by determining the elevation of the interface between the shear layer and the recirculation region, yD, at each station. By plotting y0(x) versus x, as in Fig. 8 (a), despite the uncertainty in determining the wall-position in the separated region, a monotonic increase of yD can be obtained for the laminar part of the bubble. This plot presents a good outline of the separation bubble geometry and provides a close estimation for the separation point, xs, maximum displacement location, amp, and the reattachment point, xR. At y = yo, the velocity fluctuations u values are relatively high, and the mean streamwise velocity 13 reaches a local minimum. Consequently, the interface y o between the separated region and the shear layer is clearly indicated by a local maximum in the rms streamwise velocity normalized by the local velocity, u'/U profiles. E t E ..4. r.: o o x x =0.96 m r = 098 1.6 4.55 0 It x ei II I-: o u x r= o.o4 1.5 6.94 E 64 E Nr z x = 0.74 m 1.4 871 x10-5 E o wl x= 0.66 m 1.3 885 Sear • x= 0.58m 1.2 10.73 Rem. RestLT r= o 1. 1 Rex t x1 05 5.05 to Base Separated-Flow Transition Mode natural transition transitional separation mode transitional separation mode dominant transitional mode laminar separation short bubble mode laminar separation short bubble mode dominant laminar mode laminar separation long bubble mode laminar separation long bubble mode laminar separation long bubble mode transitional separation mode transitional separation mode laminar separation short bubble mode laminar separation forced short bubble laminar separation forced short bubble o , CASE 12 10 E 8 0 Table 2 Separated-flow transition parameters zero velocity line, yo Fig.? Separated-flow transition parameters for the short bubble mode (Case 14) The location of maximum displacement, xm D, is easy to determine from the mean velocity profiles, as the location where the near-wall low velocity region is the thickest (see Fig. 7 (a)). The extent of the near-wall low velocity region is obvious even for the thinnest separation bubbles. The separation point is indicated by the occurrence of the first sign of near-zero velocity gradient in the near-wall region (e.g. Fig. 7 (a), at x = 0.5 m). If the actual separation point is situated between two measuring locations, the point of the boundary layer detachment, x s, can be determined by extrapolating the line that connects the yD values of the first two stations in the separation region until it crosses the y = 0 line, as shown in Fig. 8 (a). The contour plot of U/Uoi, can also provide a visual aid for identifying the separation location. The contour plots of U/1.1. and u'/U. are presented in Part 2 of this paper. For a boundary layer measurement very close to the separation point, the (u7U) profiles may also show a local near-wall peak. The final decision regarding xs was made by cross-checking all the above features. The location of the reattachment "point" can not be precisely determined due to the unsteady nature of the reattachment process. Defining a "reattachment region" seems to be more appropriate since a relatively extended region (5% to 15% of the separation 7 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/abo bubble length) presents characteristics which can be associated with the reattachment process. The reattachment location, xR, can be identified by examining the experimental data for three distinctive features. First; the reattachment process is accompanied by a drastic drop in the mean free- stream velocity; consequently, the local pressure gradient parameter Ktocai reaches a minimum value at reattachment. Second, the mean velocity profiles in the reattachment region have a particular shape that reassembles Horton ' s "universal reattachment profile " and is characterized by (aU/ay)y, = 0 and linear mean velocity profile across the boundary layer from zero to the free stream value. To illustrate this feature, the velocity profiles at x = 0.74 m and x = 0.81 m are replotted in Fig. 9 along with Horton 's reattachment profile. The velocity profile at x = 0.74 in is situated upstream of reattachment region while the profile at x = 0.81 m resembles Horton 's reattachment profile and can be considered at the reattachment location. The analysis of the shape of the velocity profiles acquired in the reattachment region for all the present experimental cases has shown that they are similar, although they deviated from Horton 's theoretical velocity profile in the near -wall region where the condition: (aU/ay)y-0 = 0 cannot be strictly satisfied due to the flow unsteadiness. The present study found that the reattachment location can be better determined from the mean shear (au/ay) contour plots which clearly show a near-wall minimum at reattachment location (See Part 2). Finally, the reattachment location can be also determined from the distribution of the rms streamwise velocity normalized by the local mean velocity, (u '/U). The near-wall (u '/U) peak always reaches the maximum value at xR. For the laminar separation - short bubble mode, the absolute maximum of u ' also occurred at xR. The integral parameters streamwise evolution can also give supplementary indications to confirm the chosen locations for xs, amp and xR, as shown in Fig. 8 (b). The boundary layer thickness 6, the boundary layer displacement thickness 81 and the shape factor H12 start increasing immediately downstream of the separation point. The maximum values of 81 or H12 are clear indicators for amp. The decrease in the displacement thickness 81 downstream of Xp,m, coupled with a steeper reduction in velocity between amp and xR, lead to displacement thickness Reynolds number values at reattachment much smaller than the displacement thickness Reynolds number values at maximum displacement location. As a rule, the displacement thickness Reynolds number reaches a peak value at xMD. For laminas boundary layer separation the start of transition x, coincides with xmD, therefore, for both short and long modes, the transition length expressed in terms of displacement thickness Reynolds number (Rents - Reathic, < 0) has negative values (see Table 2). The onset and end of transition were primarily judged by analyzing the velocity traces, the spectral behavior and the intermittency distribution. The onset of transition, xt, was ascribed to the first measurement traverse for which the intermittency factor was unambiguously I" > 0. It was observed that, for the laminar separation - short bubble mode, the xt location always coincides with that of the maximum displacement xmp. The mid-transition point xis imax is, by definition, the x -location where the ruts streamwise velocity fluctuation reaches a maximum in the boundary layer, and it can be easily identified in Fig. 7 (b) with the (u 'max /11...) profile at x = 81 m. 3.2 2.8 2.4 "E" 2 E 1.6 >.? 1.2 0.8 0.4 0 0.3 0.4 0.5 (a) —0-- — 0.6 0.7 0.8 x Imi 0.9 1 8 82 s H • MD R 12 5 „ 0.4 (b) 0.6 0.8 x [m] Fig. 8 Characteristic features of separated - flow transition: (a) shear layer - recirculation region interface; (b) streamwise distribution of integral parameters. 10 8 0.5 1 0 0.5 10 0.5 1 U/U Fig. 9 Comparison between Horton's mean reattachment velocity profile and experimental data For the laminar separation short bubble mode of transition, the itu: max location always coincides with that of the reattachment point xR and (01.10,),sias reaches values of about 0.18. The first indication for the end of transition is given by the streamwise tins velocity fluctuation evolution. As x -r is approached, the (u/U.) profiles converge asymptotically to the characteristic shape of the turbulent boundary layer profile, with a pointed, near-wall peak, which preserves a constant value of approximately 0.12 downstream of x.r (see Fig. 7b). The decisive information regarding the end of transition is provided by the spectral analysis, especially by the streamwise evolution of the PSD, as shown in Fig.10 (a). 8 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/abo 12 10 —8 E 6 10 7.4 2 a FIRS '2 4 In (a) 02 30 fl 0 talaill 4J 4 4 U (m/s] -50 (a) 400 600 Frequency (Hz) 800 1000 '.4 2 0 0 0.15 0 0.150 0.15 0 0.150 015 0 015 00,150 0150 0. 5 0 015 IXE (b) Taylor 1-D Spectrum U E u/(3.(u)2] 12 E 10 I 86 4 2 0 x=0.74m S E d u x o u x E E 3 co d, n x Nu d u x 8 0 II x 2 0 u x R1 t;, i " 1..:'i Trenernmrttimuntryternmr. (c) 101 (b) o II x E 8 x= 0.74M 200 E 42 o u x 0 u/U 12 10 1Cf Normalized Frequency : 8 E6 2 rt10 i 2 0 IL—I &-SIL Stin-1 1 00.5 00.5 0 0.5 0 OS 00$ 00.5 00.5 00.5 00.5 0 05 0 05 (d) - 10 4 (c) zero velocity line, yo Fig.1. 1 Separated-flow transition parameters for the long bubble mode (Case 1.7). 10 1 10 0 Normalized Frequency : Fig. 11 presents the data for CASE 1.7, which represents the laminar separation - long bubble mode of transition. The behavior up to the maximum displacement location is similar to that of the separated boundary layer in the short bubble mode. The onset of transition xt is coincidental with the location of the maximum displacement xmD. Despite the fact that the separated shear layer extends far downstream, the profiles at x = 0.70 m show the characteristics of reattachment, i.e., Kt ocat reaches a minimum (not shown here), the mean velocity profile matches the Horton's universal reattachment velocity profile, and the near-wall peak in the (u'/U) profiles reaches a maximum value. This reattachment-like behavior is identified as the first reattachment, 'cm. For the laminar separation long bubble mode of transition, the location always coincides with that of the first reattaclunent 4(1 where (u'/U...,) reaches a maximum value of about 0.15. The second reattachment is typically situated in the turbulent region, and it can be detected mainly from the mean velocity profiles. The boundary layer data for case 1.7 do not show a second reattachment location. Fig.10 Spectral features of separated - flow transition: (a) streamwise evolution of power spectral density; (b) Taylor 1-D normalized power , spectrum at x = 0.74 m (c) Taylor 1-0 normalized power spectrum at x = 0.96 m. The end of transition is identified with the location downstream of which the power spectrum no longer changes. Additionally, for a turbulent signal, the normalized power spectrum, U.. Ett/IX (t0 2] vs. ), where A is the integral length scale, shows an established cascade dissipation at a - 5/3 slope, as shown in Fig. 10 (b) for x = 0.96 m station. The end of transition can be further verified by examining the turbulent intermittency distribution which approaches r= I as x -> xi.. as shown in Fig. 7 (d). The above methodology for determining x s, xmD, AR, yD, xt, xu.ma, • and xi- applies to all separated-flow transition modes. In addition to the aforementioned common features, some aspects specific to each individual mode must be addressed. 9 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/a • - 5- 2 E o 0:1 6 s x E o r,. O s x 0 10 0 10 a) E a E o 6 II X II X a 0 10 Three primary separated-flow transition modes were identified: the transitional separation mode, for which the transition starts upstream of the separation point and develops mostly as natural transition; the laminar separation -short bubble mode, for which the transition starts downstream of the separation point and rapidly reaches the turbulent state, and the laminar separation -long bubble mode, for which the transition also starts downstream of the separation point, however, for this mode, the transition completion is delayed. Passing from one fundamental mode to another takes place continuously through a succession of intermediate stages. This paper presented a detailed discussion of the methodology implemented in determining the separated-flow transition main parameters (x,, a m p, XR, 3R , x,zu•nu,x , and xT) from experimental data. Specific aspects for each of the primary separated-flow transition mode were identified. u x NI o a li X 0 X E E 0 cc? o II X o N: o II x a X Ft 0.15 • 0 0.15 0 0.15 a cc' o x X ruD 0 0.3 0.6 0 0.3 0.6 0 0.3 0.6 0 0.3 0.6 • 0.3 0.6 E i e 0. 1 ti °Ii t X 10 u'/U a n X tk (c) 6 5 —.4 al 0 NCI 0.15 0 u'/U 1 0 n x = 0.70 m C 0 0 a c0 X 0 0.15 0 0 el E 0 10 0 6 a E O 0 0.15 0 0.15 (b) o cl 10 o 3 S 1 0 i-0 0.3 0.6 1 0 0 10 0 U [m/s] E o O .0 ° x = 0.50 m G.1f) 6 5 'I 4 E 3 •;.: 2 1 0 E 3 o u x IVI) o u x S 1 0 6 5 4 E3 E o ul o u x E o ro o u x 5 4 E3 2 6 MD xA MaX S 0 0.5 1 0 0.5 1 0 0.5 1 (d) 0 0.5 1 0 0.5 1 0 0.5 1 - zero velocity I ne, yo Fig.12 Separated - flow transition parameters for transitional separation mode (Case 1.1). Fig. 12 presents a transitional separation mode case (CASE 1.1). For this case, the onset of transition, xt, takes place upstream of the separation point. For the transitional separation mode of transition, the ke n= location coincides with that of the maximum displacement xiviD, where (u'/U.),,,a, reaches values of about 0.16. In this case, the data for the actual reattachment point were not acquired since xR was situated between two measuring locations. By analyzing the evolutions of the mean velocity profiles and ul/U profiles, it can be concluded that the reattachment location must be situated between x = 0.9 m and x = 1.0 m. CONCLUSIONS This paper presented a summary of the most important aspects associated with the experimental set-up and procedures utilized in the experimental investigation of the laminar to turbulent transition in two-dimensional separated boundary layers on a flat plate at zero incidence. A description of the new approach in analyzing separated-flow transition was introduced, the key parameters were specified, and the primary separated-flow transition modes were briefly discussed. 10 Downloaded From: https://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/82028/ on 06/16/2017 Terms of Use: http://www.asme.org/abou
© Copyright 2025 Paperzz