System identification of simplified crash models using multi-objective optimization

Abstract

A multi-objective optimization (MOO) based methodology is presented to identify simplified dynamic system simulation models. The proposed methodology is used to develop a three degree-of-freedom model for an automotive crash simulation. To date, such system identification problems have only been approached as single-objective problems. We use various MOO methods to provide new insight into the problem. Furthermore, we use this problem to study the nature of the Pareto optimal hypersurface, normalization of objectives, and specification of preferences. In general, we find that the MOO-based methodology is quite useful for dynamic system identification problems.