A Case Study on the Response Surface Method Applied to the Optimization of the Dynamical Behavior of Vehicles

This paper describes the application of statistical techniques related to the condensation of computational models so that gradient based optimization procedures can be used more effectively. The adoption of these techniques is encouraged by the possibility of an important reduction in time and cost associated to the vehicle development process. A sophisticated computational model of a Mini-baja vehicle is defined in the virtual environment by means of CAD/CAE software, intending to provide the major information related to the study of its dynamic behaviour and to define the statistical surrogates (approximate models). The creation of the computational model deals with the determination of physical and geometric properties, and is fed by stiffness and damping parameters obtained through experimental procedures. The response surface methodology (RSM) uses values obtained from simulations of the original multi-body model, aiming at establishing polynomial models which approximate the response of interest. These responses are represented by algebraic functions which are able to evaluate some of the main aspects of vehicle dynamics, such as: comfort, stability and safety. Finally, optimization results obtained from both the original (multi-body) and the approximate (obtained by means of the response surface method) models are compared, taking into account the difference between the number of objective function evaluations necessary in the first approach, and the number of computational experiments performed to obtain the response surface model.