A Universal Methodology for Comparing Computer Models to Experimental Data

摘要

This paper describes the procedure and rationale used to quantify and compare the performance of a multi-wheeled combat vehicle computer model, developed on the basis of design information only, to that of a real vehicle tested on the U.S. Army Aberdeen Test Center's test courses. The real vehicle testing comprised 60 separate runs consisting of different combinations of speed, tire pressure and course terrain. Fifty-four channels of data were selectively recorded during each run in the broad categories of component strain, vehicle and strut acceleration, strut pressure and displacement, vehicle roll, pitch and yaw rates, and steering angle. Five test courses and events (Belgian Block, 6-inch Washboard, 10-inch Half Round, Top lane-change, and right and left J-turn) were numerically characterized and incorporated into the dynamic model executed in the DADS software. Model run print intervals were at 279.018 Hz and were low-pass filtered at 50 Hz to match filtering done on the test data. The simulation model data were compared with the experimental test data on the basis of parameters which as a combination were unique to an event. As an example of the latter, the vehicle roll, pitch and yaw rates were used to characterize the vehicle performance for the J-turns, whereas statistical moments were selected to characterize the vehicle performance on the 6-inch Washboard. A "figure of merit" was computed for each event on the basis of these unique parameters. To obtain an overall result for the vehicle model, a weighted summation of the individual events was performed. The proposed methodology when fully implemented is expected to serve as a measure of the utility of a multi-body dynamics model in the developmental phase of a vehicle design, prior to the availability of a test item.