UC Berkeley Research Reports Title The Aerodynamic Performance Of Platoons: A Final Report Permalink https://escholarship.org/uc/item/8ph187fw Authors Zabat, Michael Stabile, Nick Farascaroli, Stefano et al. Publication Date 1995 eScholarship.org Powered by the California Digital Library University of California CALIFORNIA PATH PROGRAM INSTITUTE OF TRANSPORTATION STUDIES UNIVERSITY OF CALIFORNIA, BERKELEY The Aerodynamic Performance of Platoons: Final Report Michael Zabat Nick Stabile Stefano Frascaroli Frederick Browand California PATH Research Report UCB-ITS-PRR-95-35 This work was performed as part of the California PATH Program of the University of California, in cooperation with the State of California Business, Transportation, and Housing Agency, Department of Transportation; and the United States Department of Transportation, Federal Highway Administration. The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the data presented herein. The contents do not necessarily reflect the official views or policies of the State of California. This report does not constitute a standard, specification, or regulation. October 1995 ISSN 1055-1425 The Aerodynamic Performance of Platoons A Final Report Presented to the California Department of Transportation Partners for Advanced Transit and Highways (PATH) Program by, Michael Zabat Nick Stabile Stefano Frascaroli and Frederick Browand 18 May 1995 CD-ROM edition, 10 October 1995 The Aerodynamic Performance of Platoons A Final Report Presented to the California Department of Transportation Partners for Advanced Transit and Highways(PATH) Program by Michael Zabat Nick Stabile Stefano Frascaroli Frederick Browand Department of Aerospace Engineering University of Southern California Los Angeles, California 90089-1191 9 October, 1995 ABSTRACT This report details the aerodynamic performance of individual members of 2, 3 and 4-vehicle platoons. The primary purpose of the tests described here is to quantify the behavior of vehicle drag as a function of vehicle spacing. One-eighth scale models of the 1991 GM Lumina APV are used as the prototype vehicle. Each model is fitted with a force balance capable of measuring drag, side force and yawing moment. A porous ground plane equipped with suction is used to simulate the road surface. Results show a reduction in average drag for all platoon members as a function of both inter-vehicle spacing and the number of vehicles in the platoon. The average drag coefficient for a platoon of four vehicles is measured to be about 55% of the drag coefficient in isolation at spacings in the range 0.1-0.3. Extrapolating the experimental results to platoons of greater length suggests even lower average drag coefficients are possible. These low average drag coefficients for platoon operation translate directly to increased fuel savings and to less pollution per mile traveled. Keywords: Platoon, Drag, Close-following I. OVERVIEW A. Establishing Confidence in Wind Tunnel Test Results When bodies under aerodynamic study are modeled in a wind tunnel test, all possible care must be taken to establish conditions consistent with full-scale operation. In principle, the model test should reflect perfect dynamic similarity as well as perfect geometric similarity. Such perfection is never realized in practice. Even if the model tested is a full-scale vehicle, complete in every detail, the presence of the wind tunnel walls destroys perfect geometric similarity. In most wind tunnel tests, dynamic and geometric similarity are only approximately established for a variety of reasons. Wind tunnel tests still give valuable information about aerodynamic performance, provided the limitations of the test are kept in mind. Dynamic similarity first requires the definition of force coefficients for the body. Since we are primarily interested in drag force, a drag coefficient is defined, D C = (1) D 1r U2A 2 where D = measured drag r = air density U = vehicle forward speed, wind tunnel speed A = Maximum cross-sectional area of the body perpendicular to the flow Couching discussion in terms of drag coefficient rather than in terms of drag automatically accounts for differences in vehicle size and vehicle relative velocity. It allows one to make predictions for the performance of the full-scale vehicle from wind tunnel model measurements. In addition, perfect dynamic similarity requires that the Reynolds number of the model flow be made to match the Reynolds number in the full scale operation. Ud Re= (2) n where d = characteristic length scale n = kinematic viscosity of air The magnitude of the Reynolds number expresses the ratio of fluid inertia force to fluid viscous effect, or shear stress. When the Reynolds number is large, the fluid moves in response to differences in pressure 1 within the flow field. Only in very thin boundary layer regions adjacent to surfaces are viscous shearing stresses important. The characteristic length scale of our vehicle model is 1/8th that of the full-scale vehicle. Since the test is conducted with air as the fluid, the viscosity n is the same for both cases. To keep Re constant, the freestream velocity in the test case must be 8 times higher than the full-scale vehicle velocity. This, of course, is not possible(cid:190) model Reynolds numbers are lower than Reynolds numbers for full-scale operation. An automobile traveling at highway speed may correspond to Re = O(106-107), while our wind tunnel tests are conducted at Reynolds numbers about one order of magnitude lower. To account for these differences, and to judge the sensitivity of the force measurements to changes in Reynolds number, we will frequently present measurements at several Reynolds numbers. Another simulation technique is to utilize a series of small roughness elements to artificially "trip" the boundary layer on the model. Tripping insures a turbulent boundary layer more nearly in character with a boundary layer at higher Reynolds number. The model and full-scale vehicle are geometrically similar in the sense that all body dimensions have been scaled by the same ratio. The mold used to cast the models is fashioned by a stereo-lithographic process using computer-generated coordinates. For all practical purposes, the model is an exact replica. However, all minor external features such as window frames, side mirrors, door handles, windshield wipers, etc., have been removed. Also, the details of the undercarriage have been left out, and the wheels are stationary. B. The Unwanted Presence of the Ground Plane Boundary Layer In the case of ground vehicles, an additional effect is introduced by the presence of the road surface. If one uses a flat plate to model the road surface in a wind-tunnel, a layer of slow moving fluid (the ground plane boundary layer) grows continuously in the downstream direction. This layer does not exist on the actual road surface since it is the vehicle that is in motion and not the medium (air). It is necessary then to minimize the thickness of the ground plane boundary layer, d , in wind-tunnel tests. The growth of a turbulent boundary layer on a flat plate can be approximated by 0.37x d= (3) Re0.2 x where x is the distance from the leading edge. Based upon the above expression, the thickness of the boundary layer at 4 model lengths (1.22m) from the start of the ground plane at 25m/s would be about 2.46 cm. Four model lengths would correspond to a case where 2 models are in the tunnel at 1 length spacing, 2 the first model being placed 1 length from the leading edge. Further down the plate, say at x=5 m, d would be about 7.71 cm. Hucho and Sovran (1993) suggest that for a passenger car model affixed to a ground plane, the displacement thickness, d *, should be less than 10% of the model ground clearance, H, (in our case H is about 37 mm). The displacement thickness can be thought of as the amount by which an outer flow streamline is displaced due to the presence of the boundary layer. For turbulent boundary layers, d * is approximately (1/8)d . The range of d */H for our tests would generally be greater than the accepted value of 0.10, unless some means for controlling boundary layer growth is adopted. Control of boundary layer thickness can be achieved by several methods. Conceptually, the simplest way to eliminate the unwanted boundary layer is to move the ground plane at the wind tunnel speed, U. Unfortunately, this scheme is mechanically complex and impractical for the long ground plane length required to test a platoon of vehicles. Our choice is to siphon off the boundary layer through a uniformly porous ground plane surface. This is a much simpler method to implement. Other possibilities, such as tangential blowing or single slot suction, are not viable because their effectiveness would not extend over the entire length of the ground plane. C. Organization of the Manuscript This report, and the earlier PATH report Zabat et al. (1994), are repositories for all the experimental data we presently have available. In our concern for completeness, we have sacrificed some of the natural flow of a well-balanced, shorter manuscript. What we believe to be the most important portions of this work to date have been summarized in three published papers Zabat et al. (1994b), Zabat et al. (1995), and Zabat et al. (1995b). These papers, although they contain the data which is included in Zabat et al. (1994), as well as in this report, should be considered an important part of the written record for this project. Copies of these papers will be on file with CALTRANS and with PATH, or they can be obtained directly from conventional SAE and ITS sources. This report serves as the final report for the three-year project. It is more complete than the earlier PATH report in that it contains a considerable addition to the experimental results. For the sake of a single source for all the drag data, we have included in the appendix of this report, the results contained in the appendix B of Zabat et al. (1994). Please note however, that much of the discussion has not been repeated, 3 so this earlier report remains an important source of information—particularly with regard to wind-tunnel flow uniformity and the design and testing of the force balances and ground plane. The various sections of this report are arranged as follows. Section II contains a discussion of the improvements made to the force balance, and the general procedure followed when taking measurements. Section III presents the results for all the single vehicle measurements made and discusses the effectiveness of the blockage correction used, the effects of the ground plane surface, Reynolds number and boundary layer trips on the model. Section IV contains the drag measurements on 2, 3 & 4 vehicle platoons. The data from the 2, 3, and 4-vehicle platoon cases are used to predict the drag performance of an n-vehicle platoon through extrapolation. Section V presents the drag measurements for a 2-vehicle platoon with variations in geometry and ground plane surface condition. The results of the various geometries are used to define an average C ratio that takes account of geometry. In section VI, the fuel-savings benefits that D precipitate from the drag reduction are examined. Section VII presents model surface flow visualizations and measurements of the floor pressure distributions which will ultimately help explain the drag behavior as a function of spacing. In section VIII, results from tests on misaligned platoons are presented and the effects of misalignment on platoon operation are discussed. Section IX outlines future studies that will further the understanding of platoon aerodynamics. The raw data discussed in sections III, IV, V, VII and VIII are tabulated in the appendices. All the data- including the surface streamline photographs- are stored in digital format and can be retrieved and transmitted upon request to PATH. 4 II. EXPERIMENTAL METHOD Tests are conducted at the USC Dryden Wind Tunnel. This facility has a test section that is hexagonal in cross-section, 1.37 m wide and 7.4 m long. Ground simulation is accomplished using a porous plate and suction system to control the boundary layer growth on the ground plane. The majority of the tests are run at a nominal velocity of 23 m/s. Further details of the wind tunnel and ground plane are provided in Zabat et al. (1994) and Zabat et al. (1994b). A. Improving Force Balance Performance The internally mounted, 3-component force sensors used previously have been designed for nominal loads of 9.8N with an accuracy estimated at – 0.035N(cid:190) 0.35% of the full scale load, and well above the analog- to-digital converter least count resolution of – 0.006N. This kind of sensitivity requires that the flexures be very thin (0.6mm). Several significant modes of oscillation have been observed for the sensor alone, particularly at 45 and 80 Hz. When models are attached to the sensors, the additional weight reduces the lowest resonant frequency to about 20 Hz. For a bluff body, the major unsteady loads will probably be due to vortex shedding in the wake. The frequency of this shedding can be approximated by fd @ 0.2 (4) U where f = the shedding frequency d = the effective diameter ( A,the cross-sectional area ) U = the wind speed Equation (4) results in a shedding frequency of about 19 Hz for a nominal velocity of 23 m/s. This value is very close to the natural frequency exhibited by the force sensor. The original balance had very little natural damping which prevented measurements at close spacings(cid:190) where significant unsteady fluctuations are present. It was previously documented that the mean values of force measured by the sensor were not affected by the fluctuations experienced by the model at spacings greater than 0.5 vehicle lengths. At smaller spacings, the oscillations became much larger in amplitude, presumably because of larger pressure fluctuations in the region between the models. Since we felt the mean measurements might be unreliable in the presence of significant oscillations, no measurements at close spacing were presented earlier. 5 To address the limitations of the force sensor, two piston-in-pot type liquid mass dampers have been designed and installed in each sensor as shown in figure 1. The damper is, in essence, a small cylinder containing a central piston. A narrow gap of approximately 1- 2mm exists between the cylinder and the sides and bottom of the piston as shown in figure 1. The piston is attached to the model via the upper balance plate, while the cylinder is attached to the support structure through the lower balance plate. Silicon oil, which is available in a large range of viscosities, is used in the cylinder as the damping fluid. The viscosity of the oil used can be varied to adjust the amount of damping applied. For the tests shown here, each pot is filled with silicon oil with viscosity of approximately 25,000 centistokes. A A Top View Section A-A Figure 1: Schematic of force sensor dampers The response of the sensor to a drag impulse, with model attached, before and after the dampers are installed is presented in figure 2. It shows that the system is still underdamped but no longer oscillates for an extended time. The addition of the dampers has increased the damping coefficient by about a factor of 4. The useful dynamic range of the sensors can be further improved by increasing the viscosity of the silicon oil until critical damping for this lowest mode is achieved. 6