Base Flaps and Oscillatory Perturbations to Decrease Base Drag Tsun-Ya Hsu, Mustapha Hammache, and Fred Browand Aerospace & Mechanical Engineering, University of Southern California, Los Angeles, CA 90089-1191 Abstract The objective of this investigation is to study possible means for reducing the base drag of a tractor-trailer. The experiments are conducted in the Dryden wind tunnel at the USC Ground Vehicle Aerodynamics Laboratory. A roughly 1/15 scale model resembling a trailer is utilized for the study. The model is fitted with a shaped nose-piece to ensure attached flow over the forward portion of the model. The model is equipped with a force balance to measure drag. In addition base pressures are measured, and hot-wire wake surveys are conducted downstream from the model base. The Reynolds numbers (based on the square-root of the model cross-sectional area), range from 0.1 x 106 to 0.4 x 106. Drag reduction is effected by means of flaps attached along the edges of the model base, and inclined inward to decrease the size of the downstream wake. In addition, an oscillatory perturbation is applied at the flap origin in an attempt to maintain attached flow for larger angles of flap inclination. The present study has found that a simple, passive base-flap deflection—no forcing whatsoever—produces significant drag saving. The maximum drag reduction is 0.06 – 0.08 at an angle of 9-10 degrees. The magnitude of the saving is in accord with both early and recent measurements in other laboratories. The present results also show that oscillatory momentum addition has little effect on drag reduction unless the net oscillatory momentum flux coefficient is equal or greater than 0.1%. Increasing the oscillatory momentum perturbation to a coefficient value of 0.3% produces drag savings at angles greater than 9-10 degrees, but has very little effect upon the maximum saving at 9-10 degrees. To further study this anomalous behavior, follow-on experiments are planned to investigate a larger range of forcing amplitudes, and a variety forcing-function duty cycles. In addition, Digital Particle Image Velocimetry will be used to capture the detailed flow-field in the vicinity of the flap. 304 T.-Y. Hsu, M. Hammache, and F. Browand Introduction The objective of this research is to provide guidance for the reduction of base drag of a typical tractor-trailer. Roughly speaking, the total drag of a tractor-trailer can be broken down into a contribution from the tractor and undercarriage, a contribution from the trailer base and a much smaller contribution resulting from skin friction along the (long) trailer sides. As truck tractors become more streamlined, and other sources of drag are minimized, the base drag of the trailer will remain and assume even more importance. As a hypothetical example, assume a tractor-trailer with a drag coefficient CD ª D , where D is the measured 0.7. The drag coefficient is defined as C D = 1 2 rU A 2 drag; r is the air density; U is the wind tunnel speed, and A is the area of the cross-sectional body perpendicular to the flow. In this case, the contribution from the trailer base might be Cbased ª 0.14, which is 20 percent of the total drag. Reducing trailer base drag by 50 percent would reduce total drag by approximately 10 percent. Now imagine continued improvements to the tractor that decrease the total drag to CD ª 0.5. The base drag coefficient might actually increase somewhat because the thinner boundary layer separating from the end of the trailer is more efficient in entraining air from the base region. More efficient entrainment may increase the base drag to a value Cbased ª 0.20. In this circumstance, trailer base drag becomes approximately 40 percent of the total drag, and reducing trailer base drag by 50 percent now represents a total drag reduction of 20 percent. The observation that base drag and fore-body drag are inversely related was put forth by Hoerner [5] based upon observations on aircraft shapes having blunt bases. The argument has also been suggested to hold for heavy trucks (see Diebler & Smith in this proceeding, and Saltzman & Meyer [10]). There are two general approaches for the reduction of base drag. One approach, we would term passive control, is to alter the geometry of the base region is some way. The boat-tail-plate attachment studied by Lanser et al., [8], and illustrated in Fig. 1, represents one possible alteration of base geometry. Measurements on a full-scale tractor-trailer in the 80x120 Foot Wind Tunnel at NASA Ames indicate total drag reductions of the order of 10 percent, and good performance at angles of yaw (Lanser et al., [8]). Other more recent investigations of the boat-tail-plate attachment include Khalighi et al., [6], and Storms et al., [11]. Storms et al., show that total drag decreases from 0.263 without the boat-tail-plates to 0.215 with the boat-tail-plates for a clean tractor-trailer geometry without wheels—the GTS model. The total drag reduction is 18 percent. The second approach is to attempt flow control by means of an active forcing (such as an oscillating flap, or a blowing slot), meant to alter the boundary layer properties—usually to avoid an unwanted separation. The blowing device discussed by Englar in this proceeding and in [3], falls within this second category. Our approach encompasses both a modification of the base geometry by means of the addition of flat-panel flaps, and an additional active control by means of an oscillatory mass flow perturbation within the boundary layer meant to delay flow separation over the surface of the flaps. The (flat) flaps are attached to the trailer Base Flaps and Oscillatory Perturbations to Decrease Base Drag 305 base along the trailer base edges, and are inclined to the free stream to close the wake more efficiently, as in Fig. 2. They are attractive, because they can be folded flat against the base when not in use. Fig. 1 Boat-tail-plate attachment at a truck base (from NASA Ames [11]) Fig. 2 Base flaps attached to model truck base, recent test at NASA Ames The addition of passive flaps to the base of a realistic tractor-trailer model was first studied by Cooper [2]. Cooper shows that the drag coefficient for a typical straight-sided truck decreases from 0.78 without flaps to 0.72 with flaps. The optimum flap angle is within the range 10-15 degrees. Further, Cooper demonstrates that the effectiveness of the drag reduction increases with nondimensional flap length, l, defined as l = Lf/sqrt(A), where Lf is the flap length and A is the cross-sectional area of a truck. He points out that most of the drag reduction is accomplished for flap length less than a value of approximately, l = 0.18. With regard to active flow control, Nishri and Wygnanski [9] reveal that an oscillatory jet introducing momentum flux with a net zero mass flow is more effective than a steady blowing jet in delaying flow separation over a flapped 306 T.-Y. Hsu, M. Hammache, and F. Browand airfoil. They are concerned primarily with increasing lift by means of oscillatory suction/blowing. Their oscillatory jet, generated by a loudspeaker, is located at the origin of the flap as illustrated in Fig. 3 from Nishri and Wygnanski. They show that the effectiveness in delaying flow separation is determined by the location of the jet, the frequency of the induced oscillation, the net momentum flux coefficient, and the shape size of the slot. The frequency of the induced oscillation, f, is nondimensionalized by the flap length and free-stream velocity, U∞. It is defined as f* Lf F+ = . The net oscillatory momentum flux coefficient, Cμ, is defined as U• Uj 2 g *( ) , where g is the slot height, i.e., the gap between the flap and Lf U• the side wall, and Uj is the amplitude of the oscillatory jet fluctuation. Favorable delays in separation are obtained for slot heights of the order of 1/3-1/2 of the incoming boundary layer displacement thickness, and for Cμ values less than 0.1%. The most effective non-dimensional frequencies appear to lie in the range F+ = 0.3—1.5. We wish to carry over these previous insights and technologies, and to apply them for the purpose of base drag reduction. Specifically, we wish to address the following questions. • For the flat-panel base flap geometry applied to our model, what is the optimum flap angle to achieve drag reduction? • Will the oscillatory jet delay the separation of the flow, resulting in additional drag reduction? • If the oscillatory jet has the positive effect of drag reduction, what will be the magnitude of the drag reduction? • What are the critical parameters for the application of active flow control? Cm = 2 * Fig. 3 Sketch of the flap (from Nishri & Wygnanski [9]) Base Flaps and Oscillatory Perturbations to Decrease Base Drag 307 Experimental Apparatus Wind Tunnel The USC Dryden closed-circuit wind tunnel facility consists of a settling chamber, contraction, test section, diffuser and axial fan section. The contraction represents an area change of 10:1. The test section—6.1 meters in length—has an octagonal cross-section. The dimension between any two parallel sides of the octagon is 1.37 meters. For the ground vehicle studies, a porous ground-plane 1.37 m in width and 5.8 m in length is installed in the test section. A small amount of suction is applied over the surface of the ground-plane to maintain a thin boundary layer. The maximum free stream velocity in the wind tunnel is approximately 30 m/s. Truck Model A bluff-body model consisting of a parallelepiped having a rounded nose is used to study base drag reduction, Fig. 4(a). At this point, no attempt is made to model an actual tractor-trailer. The width of the model is 0.197 m. The ratio of width: height: length for the parallelepiped is [1:1.4:4]. It is mounted above the ground plane a distance 50.8 mm. To this parallelepiped is attached a rounded nose of length 0.934 m, as shown in blue in Fig. 4(a). The nose is carefully shaped to avoid flow separation. The cross-sectional area of the model, A = 0.0535 m2, is used as the reference area. Model drag, side force and yawing moment are measured by means of a strain gage sandwich designed specifically for this purpose. The model is mounted directly to the upper surface of the sandwich. The lower surface of the sandwich is anchored below the surface of the ground plane by means of the two protruding pins seen in Fig. 4(a). Also seen within the model is a Scanivalve pressure sequencer. There are fourteen pressure ports on each side of the model, and sixteen pressure ports on the model base. Each of these ports is connected to the Scanivalve and sequentially sent to a Baratron pressure transducer located outside the wind tunnel. The model incorporates flat-panel flaps along each edge of the base. In Fig. 4(a), the flaps are shown fabricated from acrylic plastic flat stock. To provide a cleaner design and more reliable flap-slot geometry, we now construct the flaps from 0.89 mm brass flat stock, as in Fig. 4(b). The flaps are soldered along the four common edges to form a box-like structure attached to a plate mounted directly to the base of the model. A total of eight independent sets of flaps at flap angles 0, 4, 8, 12, 16.5, 20.5, 25 and 29 degrees are tested. Flaps can be doubled in length by taping on the appropriate flap extension, as the second photograph in Fig. 4(b) illustrates. A detailed picture of the model base is shown in Fig. 5. A schematic drawing looking from the top is shown to the left. A commercial hi-fi loudspeaker made by Vifa (frequency range from 35 to 4000 Hz) is used to produce the oscillatory forcing. Motion of the speaker diaphragm in an otherwise sealed enclosure forces air out through the slot gaps, g = 1mm, along all four edges of the base. Two flap lengths, Lf = 5.4 and 10.48 cm, are studied. When normalized with the 308 T.-Y. Hsu, M. Hammache, and F. Browand characteristic scale, ÷A = 0.231 m, the non-dimensional flap lengths become, respectively, l = 0.233 and 0.453). Hot Film Anemometer In order to measure the amplitude of the oscillatory jet, a hot-film anemometer is adopted (see Blackwelder 1981 [1] for details of anemometers). A TSI Model 1210-10 general-purpose probe is used for single component velocity measurements. The probe is calibrated by varying the wind tunnel speed as measured by a Pitot-static tube connected to a second Baratron. (a) (b) Fig. 4 USC model equipped with rounded nose and base flaps Base Flaps and Oscillatory Perturbations to Decrease Base Drag 309 Fig. 5 Details at the base Experimental Method Experimental Conditions All data sets are recorded at wind tunnel speeds between 13 and 25 m/s. Reynolds numbers, based on the square-root of the model cross-sectional area are, respectively 0.11 x 106 and 0.36 x 106. A sine wave is chosen as the forcing function. The forcing frequency is zero or within the range 40—600 Hz. The corresponding non-dimensional frequencies are, F+ = 0, 0.17—3.93. The oscillatory momentum flux coefficient, Cμ, ranges from 0 (no forcing) to 0.3%. Data Acquisition and Data Processing Signals from the force balance, the Baratron, and the hot-wire probe are digitized using Labview. The nominal sampling rate for the force measurements in the wind tunnel is 100-1000 Hz for a total sampling time of 5-8 seconds for each data point. The sampling rate for the pressure measurements is similar. The averaging time of 5-8 seconds is sufficient to obtain accurate mean values. Previous observations of wind tunnel performance (Zabat et al., [12]) have shown that virtually all of the wind tunnel turbulence (including low frequency unsteadiness) lies above 0.2 Hz. Typically, 5-second averages of the free stream velocity lie within .0015*U• of the long-time mean. Hot-wire signals are normally sampled at about 20 samples per cycle for a suitable number of cycles. The rms amplitude of the unsteady velocity in the vicinity of the slot gap is measured outside the wind tunnel by placing the hot-wire a distance of about one gap width downstream from the slot. This measurement is needed to determine the magnitude of the applied forcing (the value of Cμ). Observations at several positions along the length of the slot verify that the influence of the corner is restricted about two slot widths. The slot flow is two-dimensional within ± 8%. 310 T.-Y. Hsu, M. Hammache, and F. Browand The rms fluctuation amplitude that determines the value of Cμ is measured at ambient pressure on a bench test, as is common practice [4], [7], [9]. The particular level of ambient pressure has no influence on the measurement of rms fluctuation since all surfaces are exposed to the same uniform pressure. In the wind tunnel, however, the situation is slightly different. One side of the driving speaker is open to the slot gap while the other side is exposed to a static pressure equivalent to a pressure along the underside of the body. The speaker must be driven against this pressure difference. In our case these pressures are both close to the wind tunnel static pressure, and no correction to Cμ has been made. A more thorough in situ calibration of forcing is planned. Fig. 6 shows the large variation in rms amplitudes that are a consequence of the dynamic response characteristic of the speaker plus enclosure at the different applied frequencies. The largest rms amplitudes occur in two periods—one at about 50 Hz and the other centered at about 140 Hz. The hot-wire is also used to determine boundary layer thickness at the trailing edge of the model. The boundary layer there is turbulent with a displacement thickness of approximately 2 mm at a wind tunnel speed of 16 m/s. The hot-wire is used to measure the downstream wake of the model truck for a variety of flap angles. For this purpose, the hot-wire is positioned at the mid-height of the model in a cross stream plane 76.2 mm behind the flaps. Amplitude (V) 10 9 rms vel (m/s) 8 5.5 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 7 6 5 4 3 0 100 Frequency (Hz) 200 300 Fig. 6 Strength of the oscillatory jet for 1 mm gap with sine wave forcing Base Flaps and Oscillatory Perturbations to Decrease Base Drag 311 Results and Discussion Passive Control We first examine the addition of flat-panel base flaps in the absence of the oscillatory jet forcing. The effect of flap angle upon drag reduction is shown in Fig. 7. The horizontal axis represents the flap angle in degrees, and the vertical axis represents the change in drag coefficient, DCD = CDnoflaps - CDflaps,. A positive value of DCD corresponds to a reduction in drag. Data is shown here from three widely different experiments. The two curves in red are from our wind tunnel experiment on a model having an aerodynamically clean fore-body at Reynolds numbers of 0.23 and 0.36x106. The experimental reliability is indicated by the scatter of data points representing multiple trials. The blue curve is an earlier mentioned result from Cooper [2]. Although the flaps are similar, the model truck geometries are different. His flaps are attached to the base of a truck model along the vertical edges and the top edge. The bottom edge has an attached flap, but this flap is not deflected. The result shown represents flaps on a model truck having straight sides, wheels and a trailer with rounded nose tested at a Reynolds number of Re÷A ≈ 1.5x106. The data shown in black is from a recent experiment performed in the NASA Ames 12-foot pressurized wind tunnel at a Reynolds number of 6x106, which is comparable to a Reynolds number for a full-scale truck at highway speeds. In this case, the model is a generic truck shape having a gap between tractor and trailer, cab extenders and simple axles with wheels. Multiple data points again give an indication of the measurement reliability. In spite of the different model geometries, all of the DCD curves have a roughly similar shape. Drag savings first increase and then decrease with increasing flap angle. Thus there is an optimum angle for maximum drag saving. The maximum saving is in the range DCD ª .06 - .08 for all three experiments. Overall, the three data sets suggest a robust drag saving that is neither particularly dependent upon the details of truck shape nor upon the Reynolds number of operation. However, the angle for maximum saving does seem to depend upon Reynolds number. The optimum angle seems to increase steadily from 9-10 degrees at 0.3x106 to perhaps 20 degrees at 6x106. The drag reduction resulting from base flaps can also be observed in the wake profile as a reduction in the wake momentum deficit. Wake velocity profiles are shown in Fig. 8 for flap angles of 0 and 12 degrees. The horizontal axis represents the span direction, non-dimensionalized by the square-root of the model crosssectional area. The model centerline is at the zero span position. The vertical axis represents the non-dimensional velocities. The blue curve indicates the wake profile for flap angle equal to zero. It can be seen that the wake width is the order of the model width. The red curve shows the wake profile for flap angle equal to 12 degrees. The wake width is now considerably narrowed. The small velocity increase at mid-span may actually represent a recirculation (backflow) along the center-plane. The single hot-wire used here is not capable of determining flow direction and may thus “demodulate” the recirculation. 312 T.-Y. Hsu, M. Hammache, and F. Browand Active Control Now consider the addition of oscillatory blowing/suction along the edges of the model base, coincident with the origin of the flap. The purpose of the oscillatory blowing/suction is to attempt to maintain attached flow over the flap for larger flap deflection angles, and thereby to further reduce drag. Fig. 9 summarizes the best of our results. The drag coefficient difference, DC D, is now shown for two flap lengths—with and without oscillatory blowing. The two blue curves represent flaps of length 0.233 and 0.453 in the absence of forcing. The drag saving peaks at about 9 degrees as in Fig.7, and doubling the flap length has no significant effect on the result. The results in red, for a momentum coefficients of Cμ = 0.15% and 0.3%, demonstrate the effectiveness of oscillatory blowing/suction. There is a modest additional drag saving at the higher flap angles, but the application of forcing has very little effect in the vicinity of peak saving. Thus the peak saving is almost unchanged. The cause of this unexpected result is still under investigating. Examination of the forced and unforced results in the vicinity of 10 degrees suggests that the momentum addition does delay the onset of separation in some manner. However, we had anticipated that in the presence of forcing and delayed separation, the drag saving would continue to increase beyond the unforced peak at 9-10 degrees. Rather, the demonstrated effect of forcing is to broaden the peak region by extension to larger angles but not to increase the maximum value of drag saving. For the shorter flap length, the optimum forcing frequency definitely lies in the range F+ ª 0.6 – 0.7. For the longer flap, the forcing frequency may be slightly below optimum. Here the non-dimensional value F+ ª 0.33 is dictated by the need to stay in the frequency band near 40 Hz to achieve the highest momentum coefficient possible. Our experiments suggest that there is a rather broad range of non-dimensional forcing frequencies F+ ª 0.3 – 1.0 that will effect flow modification, but that forcing amplitude is critical. We begin to see noticeable change only at momentum coefficients of the order of 10-3 or greater (Cμ ≥ 0.1%). Nishri & Wygnanski [9] show observable increases in lift over deflected flaps at momentum coefficients as small as 0.01%. Evidently drag reduction for a three-dimensional body such as ours is a much more subtle proposition. Base Flaps and Oscillatory Perturbations to Decrease Base Drag Fig. 7 Effect of base flap angle on total drag Fig. 8 Effect of base flap angle on the momentum defect in the wake 313 314 T.-Y. Hsu, M. Hammache, and F. Browand Fig. 9 Effect of oscillatory forcing on drag decrease Conclusions Present Study The present study describes the addition of flat-panel flaps to all four sides of the downstream end of a bluff-body model resembling a tractor-trailer. The nose of the model is rounded to prevent unwanted upstream separation (and boundary layer thickening). It is possible to form partial answers to the questions posed at the end of the introduction. • The flaps—when deflected to flap angles of 9-10 degrees—produce a total drag reduction of DCD ª 0.075, as measured by a force balance within the model. The drag reduction is broadly consistent with reductions measured earlier by Cooper [2], and with more recent measurements acquired in the 12-foot wind tunnel at NASA Ames. • The amplitude of the oscillatory forcing applied at the origin of the flaps is most important. Forcing has little effect at small values of the oscillatory momentum coefficient, Cμ = 0.1%. Forcing at amplitudes of Cμ = 0.3%, does produce modest additional drag reduction at larger flap angles, but does not alter the peak drag reduction at 9-10 degrees. • The forcing does appear to delay separation over the flap. • Non-dimensional forcing frequencies in the broad range, F+=f*Lf/U∞ ª 0.3 – 1.0, do appear to provide the greatest effect on the flow, although this is a tentative conclusion. Base Flaps and Oscillatory Perturbations to Decrease Base Drag 315 Modifications to Present Experiment To date, the results from oscillatory forcing leave several questions unanswered. We believe oscillatory forcing should lead to greater drag savings. This view is supported by the more recent work of Kjellgren et al., [7], who investigated drag changes for a two-dimensional V-22 model wing in hover mode. In this case, the bluff-body is the wing itself with down-flow normal to the upper surface of the wing, as would be achieved in hover mode (with tilt rotors above the fixed wing providing the down-flow). Forcing at the leading and trailing edges near flap junctures does delay separation to larger flap deflection angles, and appears to provide DCD of as much as 40-50 percent. What is evident from this work is that drag improvements may require larger forcing amplitudes—of the order of Cμ ≥0.01. The present experiment will be modified in several important respects to allow an expansion of the range of forcing parameters—particularly the forcing amplitude. The present loudspeaker arrangement cannot achieve Cμ values beyond 0.3%, when all four sides are forced. A means for forcing at higher amplitudes is currently being investigated, and will be incorporated into the redesign of the model. We plan to investigate forcing-signals other than sinusoidal. For example, a step in speaker voltage produces a blowing impulse response or a suction impulse response as sketched in Fig. 10(a). If either the suction or the blowing portion of the cycle should provide more effective control (there is anecdotal support for the importance of suction), one or the other can be emphasized by means of saw-tooth signals as in case (b). Finally, we plan DPIV studies (Digital Particle Image Velocimetry) of the flow adjacent to the flap with and without forcing. The intent is to answer the question regarding the state of the boundary layer and the degree of attached/unattached flow present. Fig. 10 New designs: (a) Example of equal blowing and suction response, (b) Saw-tooth emphasizing either blowing response or suction response Acknowledgement: Support from Dr. Sidney Diamond, Chief of Technology Development, Heavy Vehicle Systems, Office of FreedomCAR and Vehicle Technologies, is gratefully acknowledged. Additional thanks to Dale Satran and to NASA Ames for providing the recent experimental results for base flaps operating at full-scale Reynolds numbers. 316 T.-Y. Hsu, M. Hammache, and F. Browand References 1. R. Blackwelder, “Hot-Wire and Hot-Film Anemometers,” Methods of Experimental Physics: Fluid Dynamics, Vol. 18 Part A, 1981. 2. K. Cooper, “The Effect of Front-Edge Rounding and Rear-Edge Shaping on the Aerodynamic Drag of Bluff Vehicles in Ground Proximity,” SAE, 850288, 1985. 3. R. Englar, “Advanced Aerodynamic Devices to Improve the Performance, Economics, Handling and Safety of Heavy Vehicles,” SAE, 2001-01-2072, 2001. 4. R. Grife, A. Darabi & I. Wygnanski, “Download Reduction on a Three Dimensional V-22 Model Using Active Flow Control,” AIAA, 2002-3071, 2002. 5. S. Hoerner, Fluid Dynamic Drag, self-published, Library of Congress Card Number 64-19666, Washington, D.C., 1965. 6. B. Khalighi, S. Zhang, C. Koromilas, S. R. Balkanyi, L. Bernal, G. Laccarino & P. Moin, “Experimental and Computational Study of Unsteady Wake Flow behind a Bluff with a Drag Reduction Device,” SAE, 2001-01-1042, 2001. 7. P. Kjellgren, D. Cerchie, L. Cullen & I. Wygnanski, “Active Flow Control on Bluff Bodies with Distinct Separation Locations,” AIAA, 2002-3069, 2002. 8. W. Lanser, J. Ross & A. Kaufman, “Aerodynamic Performance of a Drag Reduction Device on a Full-Scale Tractor/Trailer,” SAE, 912125, 1991. 9. B. Nishri & I. Wygnanski, “Effects of Periodic Excitation on Turbulent Flow Separation from a Flap,” AIAA Journal, Vol. 36, No. 4, April 1998. 10.E. Saltzman & R. Meyer, Jr., “A Reassessment of Heavy-Duty Truck Aerodynamic Design Features and Priorities,” NASA TP-1999-206574, 1999. 11.B. Storms, J. Ross, J. Heineck, S. Walker, D. Driver, & G. Zilliac, “A n Experimental Study of the Ground Transportation System (GTS) Model in the NASA Ames 7- by 10-Ft Wind Tunnel,” NASA TM, 2000. 12.M. Zabat, S. Frascaroli, F. Browand, “Drag Measurements on a Platoon of Vehicles,” UCB-ITS-PRR-93-27, California PATH, University of California, Berkeley, 1993.