THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 345 E. 47th St., New York, N.Y

Transcription

1 THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 35 E. 7th St., New York, N.Y. 117 The Society shall not be responsible for statements or opinions advanced in papers or discussion at meetings of the Society or of its Divisions or Sections, or printed in its publications. Discussion is printed only if the paper is published in an ASME Journal. Papers are available from ASME for 15 months after the meeting. Printed in U.S.A. Copyright 1993 by ASME 93-GT-378 OBSERATIONS OF WAKE-INDUCED TURBULENT SPOTS ON AN AXIAL COMPRESSOR BLADE G. J. Walker* and W. J. Solomon Department of Civil and Mechanical Engineering University of Tasmania Hobart, Tasmania, Australia J. P. Gostelow* School of Mechanical Engineering University of Technology, Sydney Sydney, New South Wales, Australia ABSTRACT Measurements of transitional flow in regions of strong adverse pressure gradient on an axial compressor stator are reported. The range of observations covers separating laminar flow at transition onset, and reattachment of intermittently turbulent periodically separated shear layers. Transition was characterised by the regular appearance of turbulent spots in association with the rotor blade wake disturbances. However, the initial breakdown did not coincide with the wake passage as has usually been observed by other workers. The spots rather evolved from the growth of instability wave packets which lagged the wake passage. Data presented from the compressor blade measurements include : mean and ensemble-average velocities and associated integral parameters; distributions of total, periodic and random disturbance components; typical individual velocity fluctuation records; contours of ensemble-average random disturbance level; and boundary layer intermittency distributions. Measurements of turbulent intermittency showed a significant fall in this quantity near the wall in the reattaching flow. This has significant implications for the interpretation of transition data from surface film gage observations. NOMENCLATURE c Blade chord i Blade incidence t Time u Longitudinal velocity component x Chordwise distance y Distance normal to surface Cf = z/pu Skin friction coefficient H '6 / Boundary layer shape factor Re,, = Uc/v Chord Reynolds number Rercf = Umbc/v Reference Reynolds number Ree = U/v Momentum thickness Reynolds number T Rotor blade passing period Tu Random disturbance level (turbulence) TuD Total disturbance level Tu Local freestream value of total disturbance level Tu Periodic disturbance level (unsteadiness) U Local free stream velocity Uinv Hypothetical inviscid velocity Umh Rotor mid-span velocity Ud alue of U at standard atmospheric conditions U mean velocity for cascade a Mean axial velocity y Intermittency of turbulence 6 Boundary layer thickness (u =.99U) S Displacement thickness = jumb Flow coefficient 8 Momentum thickness p Density zw Wall shear stress v Kinematic viscosity Superscripts, etc. < > Ensemble (phase-lock) average value Time-mean value Instantaneous fluctuation from time-mean Instantaneous fluctuation from ensemble mean INTRODUCTION The importance of boundary layer transition phenomena in Presented at the International Gas Turbine and Aeroengine Congress and Exposition Cincinnati, Ohio May -7, 1993

2 turbomachinery flows is widely recognised, and a comprehensive review of work in this field can be found in recent survey articles by Mayle (1991) and Squire (1989). Transition on turbomachine blades is strongly influenced by periodic disturbances from the turbulent wakes of neighboring upstream blade rows. Blades in a multistage environment are also subjected to high levels of random freestream disturbance from the dispersed wakes of blade rows further upstream. The pressure gradient in the neighborhood of transition may vary widely depending on the machine geometry and flow coefficient: on turbine blades, turbulent breakdown often occurs in accelerating flow, on compressor blades it most commonly occurs in decelerating flow regions, often in association with laminar separation. Much useful progress in understanding the physics of transition under these conditions has been made by two-dimensional experiments in which boundary layers on a plate have been subjected to both periodic and random disturbances under controlled conditions, using both stationary and moving grids to simulate the turbomachine disturbance environment. The majority of these experiments have been carried out in zero pressure gradient or accelerating flow, both for reasons of simplicity and the relevance of this situation to the turbine blade heat transfer problem. Relatively little work has been done on transition in decelerating flow. Two general approaches have been used for two-dimensional experiments. The test boundary layer may be generated on a flat plate with an adjacent fairing to produce the required pressure distribution, such as in the experiments of Gostelow et al. (199). Alternatively, the boundary layer behavior may be studied in cascade tests, such as those of Doug and Curnpsty (199a,b). Observations of transition in actual turbomachines are very rare. An early, largely qualitative, study of wake-induced unsteady transition on an axial compressor stator blade was reported by Walker (197). More recent quantitative studies have been carried out on a turbine rotor by Addison and Hodson (199), while Li (199) has examined the transition behavior on an axial compressor rotor. This paper describes measurements of transitional flow on an axial compressor stator blade subjected to rotor wake disturbances. The range of observations covers separating laminar flow at transition onset, and reattachment of intermittently turbulent periodically separated shear layers. This work represents a preliminary study of limited scope carried out at a fixed chordwise position. The results are being used to design a more comprehensive experimental program in which the spatial development of transitional flow on the compressor blade will be examined for a range of conditions. The present study complements research at the University of Technology, Sydney involving wind tunnel studies of artificially generated turbulent spots on a flat plate subjected to an adverse pressure gradient, as discussed by Gostelow et al.(1993). The general aim of this joint project is to gain a better understanding of transition phenomena in decelerating flow, and to apply this knowledge to improve performance predictions for axial turbomachine blades. Some more specific long-term goals are the comparison of turbulent spot behavior under two- and threedimensional flow conditions, and the resolution of controversy (Walker (199)) regarding the turbulent breakdown mechanism in wake-induced transition. EXPERIMENTAL DETAIL Research Compressor Air enters the compressor radially through a cylindrical screened inlet.13m diameter by.61 m wide. A flared bend with a 6.5 to I contraction ratio then turns the flow through 9 into a concentric cylindrical duct with 1. 1m outside diameter and.69m inside diameter which contains the compressor blade rows. Downstream of the compressor there is an annular diffuser, and a cylindrical sliding throttle at the outlet is used to control the throughflow. The compressor is a single-stage axial flow machine with three blade rows: inlet guide vanes (IG), rotor and stator. There are 38 blades in each of the stationary rows and 37 blades in the rotor, giving space/chord ratios at mid-blade height of.99 and respectively. The axial row spacings at mid-blade height in these tests were 9mm IG-rotor and 8mm rotor-stator. The blades all have a constant chord of 76mm and an aspect ratio of 3.. The blade sections were designed for free vortex flow with 5% reaction at mid-blade height at a flow coefficient ( = JU..b) of.76. The inlet and outlet blade angles from axial at mid-blade height are, respectively, as follows: IG 7.8 ; rotor and stator 5.,. Instrument slots in the outer casing of the compressor allow radial and axial traversing of measuring probes at a fixed circumferential position. The IG and stator rows are each mounted on rotatable supporting rings to permit circumferential traversing of these blades relative to a stationary probe. Further details of the research compressor can be found in Oliver (1961). Rance of Investigation Observations were made at a fixed position of 6% chord at mid-blade height on the suction surface of a stator blade. Boundary layers at five different stages of transition were observed at this fixed position by varying the compressor flow coefficient () so as to alter blade incidence (i) and move the transition region relative to the probe. The compressor speed was continuously controlled to operate at a constant reference Reynolds number (Relcf) of 9,. The stator chord Reynolds number (Re c) varied between 7, and 9, depending on flow coefficient. The rotor speed was typically 385 rpm, giving a rotor blade passing frequency of around Hz. A mid-chord to rearward laminar separation bubble develops on the stator suction surface at negative incidence. The 6% chord measurement location was chosen because it lay close to the separation position for this flow regime (Walker (1971, 1975)). The five test cases corresponded to early, mid and late transition (Cases 1, and 3), incipient transition or possibly subtransition (Case ), and an unstable laminar layer (Case 5).alues of leading parameters for these cases are given in Table 1.

3 IG ROTOR STATOR ice'_ Axial -^ Sink c o Flow c d Rotor \ Q Woke ^Umb ^^!^`^ \Ho1-Wire Wake Flow y Traverse Relative Time- I Plane To Stream Meon Flow Fig. I Cross-section of compressor blading, showing typical instantaneous wake dispersion Fig. Stator blade boundary layer - hot wire traverse detail For these tests, the IG row was positioned circumferentially relative to the stator row as shown in Fig. 1, so that the stator blade section at the measuring station lay clear of the IG wake street. This meant that the test boundary layers were subjected to a freestream disturbance field comprising periodic disturbances from rotor blade wakes interspersed with intervals of relatively low turbulence flow. The long-term average total disturbance level in the free-stream varied from.6 to 5.% over the range investigated, depending on flow coefficient and rotor blade loading. These values are closely similar to those reported by Evans (1975). Measurement Techniques The boundary layer and immediate freestream region were surveyed with a DISA 55P5 hot wire probe operated by a DISA 55MI Constant Temperature Anemometer unit. As shown in Fig., neither the hot wire traverse plane nor the passing rotor wake segments were oriented normal to the blade surface. This meant that the phase of the wake arrival varied with y, but the effects are quite small (maximum A,& 5). No correction has been applied for these phase shifts, as they do not significantly Table I Test Parameters (time.rnean values) Case i( ) UJ (ms') TuD, (mm) (nun) Ree H Cf influence the interpretation of experimental results. Effects of traverse orientation on the boundary layer mean velocity profiles are similarly negligible. The hot wire sensor was aligned parallel to the surface and the traverse position read to.5µm by a dial gage mounted outside the compressor casing. The position zero was read to 5µm, but the likely y-position accuracy was around 5mm due to uncertainty about the sensor location on the probe tip. The probe supports tapered to about.mm diameter, with the sensor located centrally on the tip. This meant that the sensor could not be traversed closer than about.1 mm to the blade surface. The anemometer output was backed with a DC offset voltage and filtered at 5kHz before analog-to-digital conversion at 1kHz and data storage. Ensemble-average values of measured quantities were obtained from 1 records, with sampling triggered at the same point on each rotor revolution so that the wakes of the same rotor blades were observed in each record. Time-mean flow data were determined from separate sets of observations with continuous sampling at random phase relative to the rotor motion and an averaging time of about minutes. In situ calibration checks were made by returning to a common reference position in the free-stream after each traverse, and any observed calibration drift was corrected accordingly. elocity values were evaluated digitally for each sample point from the full dimensionless heat transfer relation for the probe. Corrections for wall-proximity effects were applied by the method of Wills (196). The measured mean velocity profiles agreed well with the earlier observations of Walker (1971). alues of skin friction coefficient were determined from a parabolic fit to the two data points nearest the blade surface and a zero velocity value at the wall. The accuracy of this quantity is reduced by the lack of velocity data within.1 mm from the wall. Disturbance and Turbulence Level Analysis The reduction of turbulence and unsteadiness data follows the procedure of Evans(1975) with some changes in notation. The instantaneous velocity may be expressed as u= u+u' =(u)+u" (1)

4 where IT is the long-term time-mean of a continuous record, and (u)(t ; ) is the ensemble average of N samples at time t, relative to the rotor phase reference, defined by N f (u)(t) = N)lk () The periodic disturbance level or "unsteadiness" is defined by Tu = ((u) u) / U (3) alues of Tu must be evaluated from averages over an integral number of blade-passing periods; in the present case an averaging time of 3T was used. The random disturbance level or "turbulence" associated with fluctuations about the ensemble average value is given by Tu = u / U () and the total disturbance level associated with fluctuations about the long-term mean is given by TuD = u;ins / U (5) Assuming ((u)(t) u) and u" to be statistically independent, the three disturbance levels are related by Tu 1) = Tu + Tu (6) > A 1.5 in o. T =.791 (Case 5) Y/d 5(<u> u)/u+y/d 1b =.76 (Case 1) y/d 5(<u> u) /U+y/b OBSERATIONS AND DISCUSSION Free-Stream Disturbance Field As mentioned earlier, the outlet stator blades were positioned circumferentially so they were clear of the dispersed inlet guide vane wakes. The test boundary layer was thus subjected to periodic disturbances from the passing rotor wakes, interspersed with regions of relatively low turbulence flow. The free-stream total disturbance levels near the blade surface, Tu, )., were around 3-5% (see Table 1). Phase-lock averaged values of random disturbance component Tu reached peak values of about 6% within the rotor wakes; values in the inter-wake regions fell to a base level which varied between I and 3% depending on flow coefficient. It should be noted, however, that the latter figure is inflated by long-period unsteadiness arising from largescale eddies in the inlet flow which would not be expected to have a direct influence on the boundary layer. As shown in Fig., the stator suction surface boundary layer was subjected to a "negative jet" by the passing rotor wake segments. This creates a sink flow near the blade surface and also has the effect of convecting rotor wake turbulence away from the boundary layer. Positive jet and source flow effects will occur on the pressure surface at the other side of the compressor blade passage. In a turbine row these effects will be of opposite sign, with wake fluid being discharged onto the suction surface. The rotor wake velocity signature observed near the stator blade surface in the absolute frame is determined by a number of r? [i>m.y rte, Fig. 3 Ensemble-average wake velocity perturbations - stator suction surface, xlc =.6 factors including: (a) the velocity defect in the relative flow, (b) flow angle variations within the wake (over- and underturning); (c) sink or source flow effects; (d) viscous-inviscid interactions arising from fluctuations in boundary layer displacement thickness; and (e) leading edge potential flow interactions. The resultant effect may vary considerably depending on the machine geometry and operating conditions. Typical ensemble-averaged wake velocity perturbations in the present experiment are shown in Fig. 3. Leading edge effects were insignificant at the 6% chord position. There is no

5 1.5 -v^^ `^ti \Q 5 E ( Re x/c (%) Fig. Stator suction surface velocity distributions at midblade height (Compressor speed 5 rpm, Reref" 1) evidence of an increase in velocity followed by a decrease as the wake passes, as observed in moving cylinder wakes by Dong and Cumpsty (199b) and ascribed to flow over- and undertu ring; nor is such behavior observed upstream of the stator leading edge, as indicated by the work of Lockhart (1973). There is, however, definite evidence of an initial velocity decrease followed by an increase with the wake passage, which is consistent with a sink effect; this can be seen to increase in magnitude as the blade surface is approached. iscous-inviscid interactions were negligible for Case 5, which corresponds to an unstable laminar layer with little turbulent activity; they are markedly evident for Case 1, which corresponds to a transitional layer in a strongly decelerating flow with alternating regions of attached turbulent and separated laminar flow. The magnitude of the viscous-inviscid interactions increases towards the stator blade surface, as expected. Similar interaction effects associated with changes in the boundary layer state have been reported by Dong and Cumpsty (I 99b) and Liu and Rodi (1991). In conclusion, attention is drawn to the variability in signatures of different wakes. This arises from the small differences in blade geometry and setting angle of individual rotor blades which must be expected in a practical turbomachine. b u/u Fig. 5 Mean velocity variation near stator suction surface x/c =.6 ir ) =.791 (Case 5) ) =.75 (Case ) ( =.76 (Case 1) =.7 (Case ) ) =.69 (Case 3) u/u Inv J Time-mean Flow Data Surface Pressure Distributions. Representative stator suction surface velocity distributions over the range of interest are shown in Fig.. This data has been taken from Walker (1971) for closely similar conditions. The suction peak location varies from about 5 to % chord for the incidence range covered, and the surface velocity distribution is either linear or slightly convex up to the measuring station at 6% chord. Fig. 6 Boundary layer mean velocity profiles - stator suction surface, x/c =.6 Mean elocity Profiles. The mean velocity variation near the stator suction surface at 6% chord is shown in Fig.5. Measurements were continued over a normal distance of several boundary layer thicknesses and the velocity gradient across the blade passage is clearly evident. The boundary layer thickness S given in Table I was defined as the y-value at which the mean 5

6 velocity fell 1% below a straight line fit to the velocity distribution outside the boundary layer. The local free-stream velocity was defined as U = u (y, _ 8 /.99. The boundary layer mean velocity profiles presented in Fig. 6 assume the velocity defect to be the difference between the measured velocity and a hypothetical inviscid flow U(y) obtained by extrapolating the linear fit from the free-stream region. ' he values of integral properties presented in Table 1 have been calculated accordingly. This data, together with extensive additional information published by Walker(1971, 1975), indicates that the profiles for Cases and 5 correspond to laminar layers close to separation. As flow coefficient is reduced and incidence increases, the flow at 6% chord becomes transitional and the separated shear layer in Cases I and is starting to reattach. This process is well advanced for Case 3, which is taking the appearance of an attached turbulent boundary layer profile. Mean Disturbance Profiles. Profiles of time-mean disturbance levels through the boundary layer are presented in Fig. 7. The total disturbance level Tu,, exhibits a peak value of about 7 around y/6 =.5 for the unstable laminar boundary layer on the point of separation (Case 5). The peak increases in magnitude and shifts towards the wall through Cases, 1 and in which transition is progressively more advanced and the flow is reattaching. A peak value of Tu D -_.16 is observed for the most advanced stage of transition (Case 3, with y..7). The periodic disturbance level Tu decays almost monotonically from the free-stream value (determined by the rotor wake signature) to zero at the wall for the unstable laminar layer prior to transition (Case 5). As transition progresses, the distribution of Tu develops a marked bimodal form with a minimum around y/6 =. to.5 and local maxima on either side. This behavior is due to switching between laminar and turbulent velocity profiles with the passage of turbulent spots. The values of Tu in Fig. 7 underestimate the periodic disturbance component in an individual velocity trace due to phase jitter in the arrival times of wakes and associated turbulent spots on successive wake passages. Distributions of the random disturbance component Tu were obtained by subtracting the unsteadiness component from the total disturbance level according to Eqn. 6. They differ little from the profiles of TuD with the exception of Case 3, where the large peak in TuD around y/8 =.3 is effectively eliminated by subtracting the unsteadiness component. Ensemble Average Data elocity ariation. Fig. 8 shows ensemble-average velocity data obtained at different levels through the boundary layer and in the free-stream for Cases I - 5. This series of records indicates, in order of increasing turbulent intermittency, the behavior of boundary layers at different stages of transition. The records cover three complete rotor blade passages, with the wake centers passing at dimensionless times UT of about.1, 1.1 and.1 in all five cases. b 'o b a it =.791 (Case 5) v =.75 (Case ) ( O =.76 (Case 1) - m =.7 (Case ) v A =.69 (Case 3) v o Tuo t::.:: ^\ - v v-v, =.791 (Case 5) v =.75 (Case ) O co =.76 (Case 1) =.7 (Case ) v =.69 (Case 3) Tu =.791 (Case 5) 1 =.75 (Case ) =.76 (Case 1) v =.7 (Case ) v =.69 (Case 3) d^ ^a\ O O ^ Tu a e'n r Fig. 7 Boundary layer time-mean disturbance level profiles - stator suction surface, xlc =.6 (Total disturbance Tu D, Unsteadiness Tu, Turbulence Tu) 6

7 ().8 =.791 (Case 5) Y/6.66 () m =.7 (Case ) Y/ A () Y/ 6 () (b =.75 (Case ) For the unstable laminar boundary laver (Case 5) there is no significant turbulent activity and the wake velocity perturbation decays towards the wall. There is, however, some evidence of a perturbation within the boundary layer which lags the wake passage by about.t. The regular appearance of turbulent spots from this region is evident in Case. For the early stage / ).69 ) () Fig. 8 Ensemble-average velocity variation with time - stator suction surface, xlc =.6 Y/ transition layer (Case 1) the turbulent patch has become quite significant, with the usual increase in velocity near the wall, a decrease in velocity in the outer part of the boundary layer, and the appearance of potential flow perturbations in the free-stream. These effects gradually increase in magnitude for Cases and 3, in which the layers are at a more advanced stage of transition. Integral Parameters. alues of boundary layer integral parameters obtained from the ensemble-average velocity traces for Case, a layer with a maximum intermittency of about %, are shown in Fig. 9. There is considerable noisiness in the data, and significant variation between different wake passages is again evident, but the following broad features can be discerned: (a) there is an increase in boundary layer thickness, corresponding to the turbulent flow patch, which slightly lags the wake; (b) the wake passage corresponds to a maximum in shape factor H and a minimum in skin friction coefficient C r Evidently the laminar shear layer is separated prior to the arrival of the wake; (c) there is a greater lag of about.3 in before the skin friction starts to increase. 7

8 i _ I 1 I 1 i 1 imean WAKE i PO ITION 1 <O>/C *1 3 1 y/s 1 Contours of<um-.791 Intervals of %. Case <a>/c * > <C f > Fig. 9 Time variation of ensemble-average boundary layer parameters (Case ) - stator suction surface, xlc =.6 1 y/s 1 y/s =.75 Case 8 a ^ Oho J6 Case 1 Turbulence Level Contours. Contour plots of ensembleaverage random velocity fluctuation level (Tu)_(u,;,,s^/U (7) in the y t plane for Cases I - 5 are shown in Fig. 1, in order of increasing intermittency. The figures cover a period of only T, with the phase reference, t =, the same as in Figs. 8 and 9. Contours are drawn at % intervals only for simplicity, and a small amount of smoothing has been applied. It should be noted that values of (Tu) around the edge of a turbulent flow patch will be inflated by the effects of phase jitter in the spot arrival time, and may therefore overestimate the true turbulent fluctuation levels. For the unstable laminar boundary layer, Case 5, the % contour does not suggest any significant temporal variation in (Tu). However, the 6% contour (not shown here) clearly indicates a reduction in (Tu) within the boundary layer for a period of about.t immediately prior to the wake passage. An increase in (Tu) is observed for a similar period after the wake passes. This behavior is suggestive of changes in boundary layer stability characteristics associated with perturbations imposed by the wake passage. These effects are clearly seen in the Case results, where the 8% contour indicates the establishment of a turbulent spot lagging the wake passage. A calmer period is evident between these turbulent patches (e.g. from about tai' =.8 to 1.). The individual records of velocity fluctuations shown in Figs. 11 and 1 indicate that these lower disturbance levels correspond to a general absence of laminar instability waves. y/s I C 1 C y/s o' g =.7 Case 8 ^^ 1 l =.69 J Case 3 1 Fig. 1 Contours of ensemble-average random disturbance level in the y-t plane (Cases 1-6)

9 For Cases 1 and, in which the layers are at progressively more advanced stages of transition, the turbulent flow patch extends over an increasing proportion of the rotor blade passing period. This is clearly evident from the expansion of the 8% contour through Cases, 1 and. It is interesting to note, through this series, that the leading edge of the turbulent patch is occurring at progressively earlier times relative to the wake passage. A region of extremely low disturbance level adjacent to the wall is observed for Cases 5,, 1 and. This is believed to indicate a stagnant region under a separated laminar shear layer, which reaches its maximum thickness as the wake passes. It is only with the later appearance of the turbulent flow patch that reattachment occurs, and some lag in this process is evident in conformity with the lag in skin friction rise noted in Fig. 9 above. For Case 3, in which transition is most advanced, the separation region is relatively much thinner. elocity-time Records The physics of the transition process on the stator suction surface can be appreciated more readily from an inspection of individual velocity-time records obtained within the boundary layer. Fig. 11 shows some typical velocity fluctuation records for Cases I to 5. They were obtained near the critical layer (u/u.3) where the amplitude of instability wave activity is greatest. These records indicate the regular appearance of a packet of instability waves from a longer period perturbation which lags the wake passage by about =.. The instability waves are barely evident in the Case 5 trace, but their amplitude increases markedly through Cases and 1. Turbulent spots are eventually formed by the growth and breakdown of large amplitude waves within these packets. As indicated by the trace for Case 1, these breakdowns do not occur regularly on each cycle of the rotor wake passage. For the layer midway through transition (Case ) the turbulent spots are evident on each wake passage. The width of the spots has increased, resulting in a higher level of intermittency. A calming period can usually be seen after the spot passage. The recovery in velocity towards the laminar state following the end of the turbulent spot takes up to half the blade passing period. The leading edges of the turbulent spots lag the wake passage by smaller amounts than in Case 1, indicating that the spot is growing through further breakdowns of instability waves at its leading edge. For the layer in the most advanced stage of transition (Case 3) the inter-spot period has shortened to the extent that recovery to the steady laminar state is barely completed when the next spot arrives. The mean streamwise pressure gradient is greatest in this case, and the velocity in the laminar region falls almost to zero indicating the presence of a separated flow region with probable flow reversal closer to the wall. Several individual velocity-time traces taken at different positions through the same layer for Case are shown in Fig. 1. The records for different y were not obtained simultaneously. The turbulent flow patches, indicated by downward velocity spikes in the outer part of the layer and upward spikes in the inner part, clearly do not coincide with the wake passage. Laminar instability waves are again strongly evident.the U =.781 MEAN WAKE POSITION y =.75 t.76 =.7 o = Fig. 11 Typical u(t) records observed near the critical layer (Cases 1-5) - stator suction surface, xlc.6.8 T +.6 ui...7 u/u Fig. 1 Set of typical individual u(t) records for different y (Case ) - stator suction surface, x/c.6 (Observations for different y have same phase reference but are not obtained simultaneously) duration of turbulent spots in the near-wall trace is significantly lower, and there is a greater time lag between the wake passage and their appearance. 9

10 T perturbations of the boundary layer velocity profile which created =.791 (Case 5) favorable conditions for wave packet development at a particular v =.75 (Case ) phase relative to the wake passage. The existence of this ) =.76 (Case 1) alternative mechanism must be recognised in modelling wakeinduced transition. 1.5 ) =.7 (Case )- v =.69 (Case 3) There are significant similarities between the transition behavior observed on the compressor blade and the development of artificially induced turbulent spots in an adverse pressure gradient reported by Gostelow et al.(1993). This encourages the belief that such basic studies can make a useful contribution to the improvement of transition modelling for turbomachinery flows. Intermittency values in reattaching transitional shear layers on.5 the compressor blade dropped significantly near the wall, and there was a significant time lag between the appearance of turbulent spots and subsequent increases in skin friction. These factors must be taken into consideration when interpreting v--- " transition data from surface film gage observations.!1! Fig. 13 Intermittency distributions for boundary layers at different stages of transition (Cases 1-6) - stator suction surface, a/c =.6 Fi ACKNOWLEDGEMENTS The authors wish to acknowledge the support of the Australian Research Council and Rolls-Royce plc. Turbulent Intermittency Fig. 13 shows the turbulent intermittency profiles for Cases 1 to 5, taken from Walker and Solomon (199). The peak value of intermittency occurs near y/s =.5 for the incipient or early transition layers (Cases 5, and 1), which is consistent with the expected behaviour in a separating flow. In Cases and 3, where transition is further advanced, the peak intermittency occurs closer to the wall. However, there is still a marked fall in intermittency towards the wall which is confirmed by a visual inspection of individual velocity fluctuation records (see Fig. 1). This is probably due to lower turbulent shear stress and higher viscous damping levels close to the wall, combined with edge intermittency effects in a periodically separated shear layer. Similar effects of smaller magnitude were observed by Gostelow and Walker (199) in transitional layers subjected to milder deceleration. The lower near-wall intermittency values have significant implications for interpreting the progress of transition from surface film gage observations. CONCLUSIONS Quantitative observations of periodic-unsteady transition on a compressor stator have been reported. The appearance of tubulent flow in the suction surface boundary layer did not coincide with the rotor wake passage in the free-stream, as has been observed by other workers. Thus there was no direct effect of wake turbulence on transition in this particular case. Rather, the turbulent spots resulted from the growth and breakdown of packets of instability waves which significantly lagged the wake. The passing wakes influenced transition indirectly through REFERENCES Addison, J S and Hodson, H P, 199, "Unsteady Transition in an Axial Flow Turbine. Part 1 - Measurements on the Turbine Rotor", Trans. ASME, J. Turbomachinery, ol. 11, pp Aral, D, Juillen, J C et Michel, R, 1979, "Analyse Experimentale de le Transition de la Couche Limite avec Gradient de Pression Nul ou Positif", ONERA TP Dong, Y and Cumpsty, N A, 199a, "Compressor Blade Boundary Layers: Part 1 - Test Facility and Measurement with No Incident Wakes", Trans. ASME, J. Turbomachinery, ol. 11, pp. -3. Dong, Y and Cumpsty, N A, 199b, "Compressor Blade Boundary Layers: Part - Measurements with Incident Wakes", Trans. ASME, J. Turbomachinery, ol. 11, pp Evans, R L, 1975, "Turbulence and Unsteadiness Measurements Downstream of a Moving Blade Row", Trans. ASME, J. Engineering for Power, ol. 97, pp Gostelow, J P, Hong, G, Melwani, N and Walker, G J, 1993, "Turbulent Spot Development Under a Moderate Adverse Pressure Gradient", ASME Paper 93-GT-, Cincinnati. Gostelow, J P, Blunden, A Rand Walker, G J, 199, "Effects of Free-Stream Turbulence and Adverse Pressure Gradients on Boundary Layer Transition", ASME Paper 9-GT-38, Cologne. Gostelow, J P and Walker, G J, 199, "Similarity Behavior in Transitional Boundary Layers Over a Range of Adverse Pressure Gradients and Turbulence Levels", Trans. ASME, J. Turbomachinery, ol. 113, pp Li, Y S, 199, "Mixing in Axial Compressors", PhD Thesis, University of Cambridge, United Kingdom. Liu, X and Rodi, W, 1991, "Experiments on transitional boundary layers with wake-induced unsteadiness", J. Fluid 1

11 Mechanics, ol. 31, pp Lockhart, R C, 1973, "Some Unsteady Flow Phenomena Downstream of an Axial Compressor Stage", M. Eng. Sci. Thesis, University of Tasmania, Hobart. Mayle, R E, 1991, "The Role of Laminar-Turbulent Transition in Gas Turbine Engines", Trans. ASME, J. Turbomachinery, ol. 113, pp Oliver, A R, 1961, "Comparison Between Sand-Cast and Machined Blades in the ortex Wind Tunnel", Aero Res Labs, Dept of Supply, Australia, Rep ME 13. Squire, L C, 1989, "Interactions Between Wakes and Boundary Layers", Prog. Aerospace Sci., ol. 6, pp Walker, G J, 1971, "An Investigation of the Boundary Layer Behaviour on the Blading of a Single-Stage Axial-Flow Compressor", PhD Thesis, University of Tasmania, Hobart. Walker, G J, 197, "The Unsteady Nature of Boundary Layer Transition on an Axial Compressor Blade", ASME Paper 7-GT- 135, Zurich. Walker, G J, 1975, "Observations of Separated Laminar Flow on Axial Compressor Blading", ASME Paper 75-GT-63, Houston. Walker, G J, 199, "'fie Role of Laminar-Turbulent Transition in Gas Turbine Engines - A Discussion", ASME Paper 9-GT- 31, Cologne. Walker, G J and Solomon, W J, 199, "Turbulent Intermittency Measurement on an Axial Compressor Blade", 11th Australasian Fluid Mechanics Conference, Hobart, ol. 11, pp Wills, J A B, 196, "Correction of Hot Wire Readings for Proximity to a Solid Boundary", J. Fluid Mechanics, ol. 1, pp