Optimization of Nonlinear Dynamical Problems Using Successive Linear Approximations in LS-OPT

This paper focuses on a successive response surface method for the optimiza- tion of problems in nonlinear dynamics. The response surfaces are built using linear mid-range approximations. To assure convergence, the method employs two dynamic parameters to adjust the move limits. These are determined by the proximity of successive optimal points and the degree of oscillation, respective- ly. Three diverse examples namely in impact design, sheet metal process design and system identification are used to demonstrate the method. The methodolo- gy has been incorporated as a parallel solver in the commercial software code LS-OPT.