Transverse flow over a wavy cylinder Anwar Ahmed and Byram BaysMuchmore Citation: Physics of Fluids A: Fluid Dynamics (1989-1993) 4, 1959 (1992); doi: 10.1063/1.858365 View online: http://dx.doi.org/10.1063/1.858365 View Table of Contents: http://scitation.aip.org/content/aip/journal/pofa/4/9?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Centrifugal instability of Couette flow over a wavy wall Phys. Fluids 14, 312 (2002); 10.1063/1.1416185 Noise radiation from transverse air flow over cylinders of constant and varying diameter: An empirical predictive model J. Acoust. Soc. Am. 83, S25 (1988); 10.1121/1.2025272 Noise radiation from transverse air flow over cylinders of constant and varying diameter: Experimental investigation of parameters J. Acoust. Soc. Am. 83, S25 (1988); 10.1121/1.2025271 Computation of turbulent flow over a moving wavy boundary Phys. Fluids 26, 2065 (1983); 10.1063/1.864410 Nonlinear analysis of laminar boundary layer flow over a periodic wavy surface Phys. Fluids 21, 1471 (1978); 10.1063/1.862409 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 Transverse flow over a wavy cylinder Anwar Ahmed and Byram Bays-Muchmorea) Department of Aerospace Engineering, Texas A&M University, College Station, Texas 77843-3141 (Received 15 May 1991; accepted 13 May 1992) Transverse flow over a wavy cylinder was investigated experimentally; surface-pressure distributions and flow visualizations were obtained for a set of wavy cylinders with different axial wavelengths. Significant spanwise pressure gradients were present, resulting in three-dimensional separation lines and the formation of streamwise trailing vortex structures near the geometric nodes. Despite the symmetry of the geometries, the separated Sow structures near the geometric nodes were distinctly asymmetric a large fraction of time. Integration of the pressure data revealed greater sectional drag coefficients at the geometric nodes than at the geometric saddles. 1. INTRODUCTION Unsteady flow fields associated with general threedimensional, bluff-body separation and transitional or turbulent wakes are still too complex to be determined using the available computational tools. The current work is part of a program to study the physics of such flows experimentally, with the goal of aiding in the development of engineering tools. In recent years greater attention has been focused on the three-dimensional phenomena present in the nominally two-dimensional flow over a right circular cylinder. l-3 In the current work three-dimensionality was explicitly introduced into the flow field by the wavycylinder geometry shown in Fig. 1. This geometry was identified as a natural extension of the right-circularcylinder geometry for a study of three-dimensional separation and wake development. This paper reports on some the salient characteristics of the wavy-cylinder flow field. The only information available in the literature on wavy-cylinder flows concerns the attachment-line boundary-layer behavior between nodal and saddle points of attachment. Brown,4 Cooke and Robbins, and more recently von Kerczek6 solved the laminar boundary-layer equations for flow from nodal to saddle points of attachment using a variety of solution techniques. They all found that for waviness of sufficiently small amplitude, the entire boundary layer flows from geometric node to geometric saddle without separation. For amplitudes greater than this “borderline case,” the earlier works speculated significant separation would occur, disrupting the outer flow. von Kerczek’s calculations also predicted that reversed flow would occur (directed away from the saddle and toward the node) but that it would be limited to the bottom of the boundary layer and not disrupt the outer flow. These efforts were all based on various assumed forms for the pressure distribution from nodal to saddle points of attachment, as there were no prior experimental data available for the pressure distribution over a wavy cylinder. ‘)Present address: The Boeing Company, Seattle, Washington 98124. II. DESCRIPTION OF EXPERIMENTS A. Model geometries The terminology used to describe the wavy cylinders is shown in Fig. 1. The axial locations of maximum diameter are hereafter termed “geometric nodes” and the axial locations of minimum diameter are termed “geometric saddles.” A right-circular cylinder and four wavy cylinder models were used in the wind-tunnel and water-tunnel tests. The wavy cylinders (shown in Fig. 2) are described by the equation: R ,ocal=Rmean-A cos(2rZ/j). For each of the four cylinders R,,,, was 3 1.75 mm and A was 6.35 mm; the four different wavelengths (A> were 76.2, 101.6, 127.0, and 152.4 mm. Hereafter the cylinders are referred to by the appropriate value of wavelength/ mean diameter (/z/D): either 1.2, 1.6, 2.0, or 2.4. The cylinders were each slightly over 610 mm long. Each cylinder spanned the height of the wind tunnel and extended from the floor of the water tunnel to a point above the waterline. The base geometry of each cylinder (the lowest 20 mm) was slightly different as shown in Fig. 2. The spanwise-waviness geometry abruptly changed to a constant 76.2 mm diameter cylinder for the /z/D= 1.2, 1.6, and 2.0 cylinders while the UD=2.4 cylinder base ended smoothly near a geometric node. B. Wind-tunnel tests Boundary-layer transition and surface-pressuremeasurement tests were conducted in the Texas A&M 2 ~3 ft low-speed wind tunnel at a Reynolds number of 20 000 based on mean cylinder diameter. Transition was detected using a Pitot probe connected to a set of earphones; turbulent shear layers were identified by an audible hiss reflecting rapid variations in total pressure. This method of transition detection, despite being intrusive, proved to be adequate and far easier to use than alternatives such as hot-film sensors, sublimating chemicals, and liquid crystals.7-‘0 @ 1992 American institute of Physics 1959 0899-8213/92/091959-i 0$04.00 Phys. Fluids A 4 (9), September 1992 1959 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 seven diameters upstream of the cylinder revealed a spanwise free-stream velocity variation of up to 1%. The pressure coefficient was calculated using C,= ( P-Pf,)/Qf, where Q,, was calculated for each pressure tap using the free-stream velocity measured upstream at the spanwise location of the tap (rather than using one averaged value for Qf,) . This procedure had a maximum effect of changing C, by 0.02. Using the symmetry of the wavy-cylinder geometry, all data presented are averages from measurements at positive and negative angular locations. These C, values agreed within 0.02 (or 2% of the free-stream dynamic pressure value) for 6 less than to 28” (see Fig. 1 for the definition of angle 6) and within 0.04 everywhere else with a few exceptions occurring for 6 above 60”, primarily in the post-separation regions. Further descriptions of the wind tunnel and instrumentation hardware are given elsewhere.” Geometric Node Geometric Saddle FIG. 1. The geometry and coordinate system of the wavy cylinders. Surface pressures were measured using 11 pressure taps on each wavy-cylinder model. The taps were spaced every O-15/2,extending from one geometric saddle to a geometric node 1.51 further along the span. Because of the symmetry of the geometry, the data were combined to give the pressure distribution over a typical half-wavelength with pressure measurements every 0.05d along the span. [This procedure assumes the asymmetric separation structures near the nodes (discussed later) were either symmetric in a time-averaged sense or else had negligible effect on the surface pressure distribution.] The pressure distribution over the entire surface of each wavy cylinder was obtained by rotating the model 360” in 4” increments. The only corrections applied to the data were to account for a spanwise variation in free-stream dynamic pressure. A survey of the wind-tunnel test section conducted C. Water-tunnel tests The flow-visualization data were obtained in the Texas A&M University 2 X 3 ft water tunnel. The tunnel has a test section velocity range of 61 to 610 mm/set. The empty test-section turbulence intensity is less than 1.0% for velocities up to 305 mm/set (the highest velocity used in the current work). The wavy-cylinder surface-flow behavior was observed by injecting a mixture of food coloring and water through the pressure taps described earlier. The pressure on the dye injection line to each tap was regulated separately to the minimum value required to produce a visible dye line. Dye was also painted onto the surfaces of the cylinders and allowed to dry prior to insertion in the water tunnel. The surface-flow separation behavior was then observed as the dye was washed away. The behavior of the attachment-line flow for each cylinder was observed with and without a wake splitter plate installed. The splitter plates had serpentine edges to fit the wavy cylinder model geometries and extended seven diameters downstream from the centers of the cylinders. The location of the separation line was determined by projecting a laser beam from the side of the water tunnel and measuring the distance required to move the beam from a reference location (X=0, see Fig. 1) to the separation-line location. For these measurements the separation line was identified using the injected-dye streaklines. Ill. RESULTS AND DISCUSSION A. Attachment-line FIG. 2. Wavy-cylinder models used in the wind-tunnel and water-tunnel tests. From left to right the values of /z/D are 1.2, 1.6, 2.0, and 2.4. A right-circular cylinder is also shown. 1980 Phys. Fluids A, Vol. 4, No. 9, September 1992 flow For each cylinder tested, dye injected from the surface showed the attachment-line location to oscillate at the von K&man shedding frequency: dye injected near 8=0” was carried over alternate sides of the cylinder. The oscillation was much more dramatic for Re= 10 000 and 20 000 than for Re=5 000. For the wavy cylinders this motion was observed everywhere along the attachment line, from geometric node to geometric saddle, and was in phase over at A. Ahmed and B. Bays-Muchmore 1960 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 0 AttachmentLine, 0 = 0 deg. h/D 0 = 1.2 0 = 1.6 A = 2.0 v = 2.4 -EC.2 -- = Borderline Case Saddle FIG. 3. The attachment-line pressure distribution measured on each of the cylinders. The dashed curve is the theoretical borderline pressure distribution for the onset of attachment-line boundary-layer separation by von Kerczek.6 (b) least one half-wavelength. In each case it was possible to suppress the attachment-line oscillation by inserting a splitter plate along the wake centerline. This demonstrated that the attachment-line oscillation was due to induced effects of the primary wake instability, rather than a local instability at the attachment line. These results are consistent with those previously reported for unsteadiness at the attachment line of right-circular cylinders.12-‘7 The splitter plates were not used in any of the other tests reported here. The mean attachment-line pressure distributions for all four wavy cylinders are shown in Fig. 3. The pressure variation is fairly insensitive to the spanwise wavelength, although there is a slight trend of increasing amplitude with decreasing cylinder wavelength, particularly near the geometric node. Each attachment-line pressure distribution is skewed toward the node, indicating the outer flow acceleration is more rapid near the node than the subsequent deceleration near the saddle. As discussed earlier, if the amplitude of the waviness is large enough, the attachment-line boundary layer leaving the node will not reach the saddle. The theoretical pressure distribution (taken from Ref. 6) for the onset of boundarylayer separation along the attachment line is also shown in Fig. 3. The data for all the cylinders tested fall well below this “borderline case,” both in amplitude and maximum adverse pressure gradient. The flow visualization confirmed that there were no disruptions along the attachment line. For both the oscillating and nearly steady flow, the spanwise flow of dye along the attachment line was di- FIG. 4. Dye injected from the surface of the wavy cylinders at Re = 10 OC0 at various angular locations. (a) 0=0”. (b) 0=30”. rected from geometric node to geometric saddle for each wavy cylinder. B. Boundary-layer flow The boundary-layer region on each cylinder extended from the attachment line near 8=0” to the primary separation line near 6=80“. Figure 4 shows streaklines produced by injecting dye from the surfaces of the A/D= 1.2 and 2.4 cylinders at Re= 10 000. The dye lines show that for each cylinder the flow was primarily circumferential, but did contain a spanwise velocity component directed from geometric node to geometric saddle until just before the separation line, where the flow turned back toward the node. This spanwise flow was more pronounced for the shorter wavelength cylinders. The surface-streakline patterns showed no significant variation over the Reynolds number range tested (5000 to 20 000). The entire streakline patterns appeared to be slightly unsteady due to the von K&man vortex shedding, significantly more so for Re = 10 000 and 20 000. The movement did not significantly affect the appearance of the streaklines ahead of separation. It was expected that the boundary layers in these experiments would not transition via Tollmien-Schlichting waves prior to separation because the test Reynolds number (Re = 20 000) was an order of magnitude lower than A. Ahmed and B. Bays-Muchmore 1961 Phys. Fluids A, Vol. 4, No. 9, September 1992 1961 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 0 Node no 0180 Lx&&&4 2.0 000002.4 RCC Saddle 100 75 Sepa?ation85Ang1~0 (degy FIG. 7. The separation line locations measured for each of the cylinders at Re=20 Ooo. tions characteristic of turbulent boundary layers were not present on the cylinder surfaces. FIG. 5. Boundary-layer separation line of a right-circular cylinder. C. The boundary-layer the critical Reynolds number for a “two-dimensional” cylinder. However, the spanwise flow over the wavy cylinders introduced the possibility of transition triggered by crossflow instabilities. The Pitot-probe audible tests revealed that the boundary layer did remain laminar up to separation for each cylinder. This result justifies the use of the intrusive Pitot-probe technique: the presence of the probe is expected to enhance, rather than suppress, transition upstream in the boundary layer. Distinct frequencies were detected above the cylinder surfaces near the separation lines (corresponding to the separation structures described later) but the broadband frequency total pressure fluctua- separation line The structure of the separation line was observed by painting dye onto the surface of the cylinder and allowing it to wash off. As shown in Fig. 5, the separation line for a right-circular cylinder was found to be very crisp and uniform (“two dimensional”) along the central 60% of the span, in agreement with past work at subcritical Reynolds numbers.’ As shown in Fig. 6, the wavy-cylinder separation lines were clearly three dimensional, originating at the geometric saddles at approximately 8 = 75” and terminating near the geometric nodes at approximately 8=92”. As seen from the measured separation-line locations given in Fig. 7, the distortion of the separation line was greater for the shorter wavelength cylinders. An interpretation of the topology of the flow field near the wavy-cylinder boundary-layer separation line is offered in Fig. 8 (where S and F denote saddles and foci in the : Ix 1 h/D FIG. 6. Boundary-layer 1962 ~&+I, = 1.2 separation line of a wavy cylinder. Phys. Fluids A, Vol. 4, No. 9, September 1992 I FIG. 8. Topology of the velocity field near the wavy-cylinder boundarylayer separation line. A. Ahmed and B. Bays-Muchmore 1962 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 s s F N F s s I: S S @ ill FIG. 9. Kinematically possible topologies of the surface shear-stress field near the geometric node of wavy cylinder. topology of the surface shear-stress field). That there exists a saddle point of separation at each geometric saddle, and that the boundary-layer fluid rolls up into streamwise vortical structures near each geometric node is quite clear from the earlier figures and all of the flow-visualization tests conducted. Less clear, due to the nature of the watertunnel flow-visualization techniques, are the details of the singularity (or singularities) in the surface shear-stress field at the geometric node. Several possible alternative topologies for the separating flow near the geometric node are shown in Fig. 9 (where N denotes an ordinary node in the surface shear-stress field). Dye patterns consistent with each of these topologies were observed during the experiments. In cases where there appeared to be two trailing vortices as suggested in Fig. 8, the two consistently combined into a single structure (see Fig. 6) resulting in an asymmetric flow pattern despite the symmetric geometry. D. Complete surface pressure distributions An overview of the mean static pressure distribution over each of the wavy cylinders is presented in Fig. 10, where the surface contours were determined based on linear interpolation between the data points and the mean circumference was used in scaling the lengths of the axes. These contour plots show that the pressure drops more quickly and attains more negative values over the geometric saddles than over the geometric nodes. The pressure contours indicate that the flow also separates earlier at the geometric saddles than at the geometric nodes. Beyond the primary separation region, especially in the range from 0= 1 lo” to 150”, the contour lines are roughly parallel to the axes of the cylinders. From 8= 160” to 180” each cylinder shows a region of lower pressure near the geometric node. The circumferential pressure distributions for all of the cylinders are plotted together in Fig. 11. At the geometric saddle [Fig. 11 (a)] each wavy cylinder shows a more rapid pressure drop, an earlier, higher suction peak, and an earlier separation than the right-circular cylinder. These effects become more pronounced as cylinder wavelength decreases, except for the lower suction peak visible for the A./D= 1.2 cylinder. Halfway between geometric node and saddle [Fig. 11 (b)] the circumferential pressure distribu- tion for each wavy cylinder closely matches that for the right-circular cylinder. At the geometric node [Fig. 11 (c)] the trends from Fig. 11 (a) are reversed: the wavy cylinders show a more ‘gradual pressure drop, to a ‘later -and less extreme suction peak and a later separation. In all of these figures the spanwise-averaged basepressure coefficient is slightly lower (more suction) for the /z/D= 1.6 cylinder and slightly higher (less suction) for cylinder. These effects are likely due to the the A/0=2.4 cylinder end conditions; each of the wavy cylinders intersected the wind-tunnel ceiling at a different location between node and saddle. Williamson and Roshkoi8 have shown that for a cylinder at Reynolds numbers of 105 and 302, the end conditions can have a significant effect on the oblique nature of the von K&man vortex shedding and the value of the base-pressure coefficient, especially within 20 diameters of the ends of the cylinder. The results of Fig. 11 indicate that the spanwise-mean base-pressure coefficient for a wavy cylinder of aspect ratio 9.6 is more sensitive to the geometry at the cylinder-wall interface than to the local geometry of the waviness, even at a Reynolds number as high as 20 000 where the wake is highly three dimensional. The spanwise pressure gradients are shown in Fig. 12. The attachment-line pressure distribution given in Fig. 3 is repeated in Fig. 12(a) on the same scale as the rest of the plots in Fig. 12. By 8=20” [Fig. 12(b)] the spanwise pressure gradient is favorable from node to saddle, and the increase of this gradient with decreasing cylinder spanwise wavelength is visible. The spanwise gradient grows until 8= 52” [Fig. 12(c)]. The variation of spanwise pressure gradient with wavelength is even more pronounced ‘iii physical dimensions than it appears to be in Fig. 12(c), since the abscissa in the figure is nondimensionalized’ by the spanwise wavelength. For angles between 56” and 80” Zig the point of minimum pressure moves from saddle to node. The typical case of f3= 72” is shown in Fig. 12(d) . At 8 = 80” the saddle pressure has reached its plateau level, indicating separation, and the node pressure is at its peak and is about to start recovery. By 8=96” even the /z/D = 1.2 wavy-cylinder node shows signs of having separated. For 8= loo”-156” there are no discernible spanwise pressure variations. From 8 = 160”-180” there is a pressure drop from the saddle to the node, shown for 8= 180” in Fig. 12(e). E. Pressure-drag calculations Although the flow is three dimensional, it is interesting to examine the spanwise variations in “local pressure-drag coefficient per unit length,” cd-rocal,measured at a fixed axial location and based on local diameter: as well as the “mean pressure-drag coefficient per unit length,” qmean, based on mean diameter: 1963 Phys. Fluids A, Vol. 4, No. 9, September 1992 A. Ahmed and B. Bays-Muchmore 1963 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 s ---I&----[[;--,.2; 4”5 1 *-Q---&\~ ---__-. -Gy---- ______0s --4:--z. ----. ---____-_--_ I= - b.0 d.1 d.2 Saddle Ii.5 ii.4 0-s d.3 (a) G s ‘-0 ,::- p ~;G-:l.&y~~-= ,’ ,/--;-bo.‘---__ I’ \----;-NL--- Y f& -1.2 - : g Node Saddle (cl Node Z/A : ._---_ ‘\-. -. -.-_ __;_------ ____--.-_-- .* z \c- ./ --. 0.0 Saddle 0.1 d.3 6.2 6.4 z/x Node Saddle FIG. 10. Isobars of static pressure coefficient for each of the wavy cylinders. Re=20 CEO. (a) A/D= --2 I,“==boy J(;;;o,’ (“FL@) Cd-mean- 1964 Phys. Fluids A, Vol. 4, No. 9, September Node (d) (b) 1992 1.2; (b) L/D= 1.6; (c) A/0=2.0; (d) ,X/0=2.4. This equation for cd-,, is valid for the wavy cylinders as long as the axial length to which it is applied consists of an integral number of half-wavelengths. The spanwise variacdSIOCd and the value of c&meanwere evaluated using trapezoidal integration and are shown for the A/D= 1.2 A. Ahmed and B. Bays-Muchmore 1964 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 0 0 A v -= h/D = 1.2 = 1.6 = 2.0 = 2.4 cc! h/D 0 = 1.2 0 = 1.6 A = v = e @ed 2.4 I e @cd 0 0 A v (b) 2.0 h/D = 1.2 = 1.6 = 2.0 = 2.4 FIG. 11. Circumferential pressure distributions for each of the wavy cylinders at three locations along the span, Re=20 Ooo. (a) Z/l.=O.O (the geometric saddle); (b) Z/A=O.25; (c) Z/2,=0.5 (the geometric node). 0 (deg) cylinder in Fig. 13. The value of cd-ioc~ at the geometric node is over 10% greater than the value of c&mean,and the value of c&loCalat the saddle is more than 10% less. The local drag force per unit length is approximately 30% greater at the geometric node than the mean value and 30% less at the geometric saddle, due largely to the spanwise variation in local diameter. The difference between c&&al at the geometric nodes and the geometric saddles decreases with increasing spanwise wavelength as shown in Fig. 14. The mean pressure-drag coefficient per unit span is given for each of the cylinders in Table I. Despite the large spanwise pressure variations, the integrated pressure-drag coefficient for the wavy cylinders is not significantly different from that for a right-circular cylinder. The pressure drag is dominated by the mean level of the base-pressure coefficient, which seems far more sensitive to the cylinder end conditions than to the spanwise waviness of the geometries tested. IV. CONCLUDING REMARKS In the current work we have explored the pressure field and flow patterns over the surface of the wavy-cylinder geometry. This geometry is promising as a testbed for experimental, theoretical, and numerical research in unsteady, three-dimensional, bluff-body separation and wake development because of its simplicity and symmetry. The significant spanwise pressure gradients suggest the use of 1965 Phys. Fluids A, Vol. 4, No. 9, September 1992 A. Ahmed and B. Bays-Muchmore 1965 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 2 , h/D 0 = 0 = A = v = -=CO 2 2 “: 7 -~ A/D 1.2 1.6 2.0 2.4 ‘c4 ” -i 62 6 T- F LLI‘: c.lo 2 L 0 f- 2 2 0 N a* v B A 0 v8po -r,, 0 0 LI 0 = 0 = A = v = -=CO 1.2 1.6 2.0 2.4 @ x0: 0 O 8 I, E, 0 0 zI I 9 I’ B m I 0.1 % 8 1 0.2 Saddle @ I 0.3 i? B , 0.4 @ 7 0. 5 0.0 1 Saddle Node Z/h (at 0.1 t 8 0.2 d -I 0.4 I 0.3 t 0.: blad Z/h (d) te -i 6; a t-3 0 i f “a:+ Fi 9, f, A 0 v v g 0 @ q 0 0 ” n Q : v v on 0 cl n I ” v 0 i1 % 2 0.0 # 0.1 Saddle 1 0.3 0.2 1 0.4 I, .5 0.0 Saddle (e) Node Z/X (b) 0 0.1 4 0.2 I 0.3 Z/h d.4 0.5 Node FIG. 12. Spanwise pressure distributions for each of the wavy cylinders at lOCatiOns, (a) 8=0”; (b) 6’=20”; (c) 8~52”; (d) 0~72~; (e) various 8= 180”. 2 II 0.0 Saddle I 0.1 8 0.3 d.2 1 0.4 Z/h 5 Ntde (cl 1966 Phye. Fluids A, Vol. 4, No. 9, September 1992 A. Ahmed and B. Bays-Muchmore 1966 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27 TABLE Re,=20 I. Integrated mean pressure-drag coefficients per unit span, OCO.The values given were repeatable to within 0.01. 1.30 1.25 1.20 ‘;; Sl 15 48 . 0 1.10 1.00 I 0.00 0 ( 0.10 I I 0.20 Saddle I I 0.30 a , 0.40 1 1.24 1.19 1.21 1.20 1.13 This work was conducted while the second author was supported by Rockwell International through the Rockwell Doctoral Fellowship Program. , 0.50 Node Z/h FIG. 13. Spanwise distribution of cd.iocalfor the A/D= 1.2 wavy cylinder. The dashed line indicates the value of cd-mea”for this cylinder. this geometry at higher Reynolds numbers as a testbed for cross-flow transition work. Symmetrical geometry does not ensure symmetrical flow patterns, as illustrated by some of the shear-stress field topologies sketched in Fig. 9. In the current tests the flow appeared far more stable and symmetrical about the geometric saddle than the geometric node; therefore it is suggested that if the wavy-cylinder geometry is used for numerical research into the instabilities of separation structures, the cylinder should intercept the computational boundaries at the geometric saddles. The base-pressure coefficient was found to be slightly lower (more suction) at the geometric nodes than at the geometric saddles, suggesting a spanwise variation in the wake structure. The separation structures and turbulent wake of the wavy-cylinder geometry are the subjects of work currently in progress, and will be addressed in a forthcoming paper. 0.30 1 0.25 - 0.20 \ ;0.15- 2 --~- x 0.10 0.05 i I 0.6 I 1.2 1 1.6 I 2.0 1 2.4 I 2.8 h/D FIG.14. The effect of cylinder wavelength on the spanwise variation in cd-,ocll = cd.local ( Acd-,,a, cd-mean RCC A/D= 1.2 /z/D= 1.6 A/0=2.0 k/D=24 ACKNOWLEDGMENT 1.05 0.00 Cylinder (node) - cd4,1 (saddle) ). ‘U. Dallmann and G. Schewe, “On topological changes of separating flow structures at transition Reynolds numbers,” AIAA Paper No. AIAA-87-1266, 1987. *M. Hayakawa and F. Hussain, “Three-dimensionality in a plane turbulent wake,” J. Fluid Mech. 206, 375 (1989). 3C. H. K. Williamson, “Three-dimensional aspects and transition of the wake of a circular cylinder,” Proceedings of the 7th Symposium on Turbulent Shear Flows, Stanford University, 21-23 August 1989, Paper No. 6.2. 4S. N. Brown, “Singularities associated with separating boundary layers,” Proc. R. Sot. London Ser. A 257, 409 (1965). ‘J. C. Cooke and A. J. Robins, “Boundary-layer flow between nodal and saddle points of attachment,” J. Fluid Mech. 41, 823 (1970). %. von Kerczek, “The symmetry plane boundary layer on a corrugated cylinder in cross flow,” AIAA Paper No. AIAA-88-3544-CP, 1988. ‘B. J. Holmes, C. J. Obara, and L. P. Yip, “Natural laminar flow experiments on modem airplane surfaces,” NASA Tech. Publ. TP-2256 (1984). ‘B. J. Holmes, C. C. Croom, P. D. Gall, G. S. Manuel, and D. L. Carraway, “Advanced transition measurement methods for flight applications,” AIAA Paper No. AL4A-86-9786, 1986. 9B. J. Holmes, P. D. Gall, C. C. Croom, G. S. Manuel, and W. C. Kelliher, “A new method for laminar boundary layer transition visualization in flight-color changes in liquid crystal coatings,” NASA Tech. Memo. TM-87666 (1986). “C. J. Obara and B. J. Holmes, “Flight-measured laminar boundarylayer transition phenomena including stability theory analysis,” NASA Tech. Publ. TP-2417 (1985). “B. Bays-Muchmore, “An experimental investigation of the surface flow and wake dynamics associated with transverse flow over wavy cylinders,” Ph.D. dissertation, Texas A&M University, 1991. “D. M. McGregor, “An experimental investigation of the oscillating pressures on a circular cylinder in a fluid stream,” UTIA Tech. Note TN-14 (1957). “W . D . James, S. W. Paris, and G. N. Malcolm, “Study of viscous crossflow effects on circular cylinders at high Reynolds numbers,” AIAA J. 18, 1066 (1980). 14C. K. Cheung and W. H. Melbourne, “Wind tunnel blockage effects on a circular cylinder in turbulent flows,” Proceedings of the 7th Australasian Hydraulics and Fluid Mechanics Conference, Brisbane, 18-22 August 1980, pp. 127-130. “T. Igarashi, “Effect of vortex generators on the flow around a circular cylinder normal to an airstream,” Bull. Jpn. Sot. Mech. Eng. JSME28, 274 (1985). 16M. Braza, P. Chassaing, and H. Minh, “Numerical study and physical analysis of the pressure and velocity fields in the near wake of a circular cylinder,” J. Fluid Mech. 165, 79 (1985). “S. M. Mangalam and L. R. Kubendran, “Experimental observations on the relationship between stagnation region flow oscillations and eddy shedding for a circular cylinder,” Instability and Transition, edited by M. Y. Hussaini and R. G. Voigt (Springer-Verlag, New York, 1989), Vol. 1, pp. 372-386. 18C. H. K. Williamson and A. Roshko, “Measurements of base pressure in the wake of a cylinder at low Reynolds numbers,” Z. Flugwiss. Weltraumforsch 14, 38 (1990). A. Ahmed and B. Bays-Muchmore 1967 Phys. Fluids A, Vol. 4, No. 9, September 1992 1967 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP: 137.222.24.34 On: Fri, 30 Jan 2015 05:34:27