MULTILEVEL OPTIMIZATION OF THE DYNAMIC BEHAVIOUR OF A LINEAR MECHANICAL SYSTEM WITH MULTIPOINT APPROXIMATION

Abstract

The problem of optimizing the ride characteristics of a truck subjected to a random excitation due to road irregularities is considered. The weighted acceleration (ride index) at the driver position is to be minimized. The design variables comprise geometry as well as spring and damper properties. As a general optimization method an iterative multipoint explicit approximation technique is used. In each iteration step, the objective and constraint functions are approximated by simpler ones on the basis of the values of the original functions at several points. Derivatives may be used, if available. Because of the relatively large number of design variables (20), the straightforward optimization is computationally expensive, though feasible. To reduce the total computational cost, a two-level optimization approach is used. At the first level a succession of problems with fewer design variables (sizing) and constraints related to the subsystems are solved. At the second level, the optimization of the geometry with the full set of constraints is considered. The procedure is iterated until the optimum is reached. It is demonstrated that the procedure is efficient for the optimization of the dynamic behaviour of complex structures and that it can be used for large scale problems.