Automatic evaluation of move-limits in structural optimization

The two-point exponential approximation method was introduced by Fadelet al. (1990) and tested on structural optimization problems with stress and displacement constraints. It was subsequently tested on problems with frequency constraints (Sareenet al. 1991). The results reported in earlier papers showed reductions in the number of iterations needed to reach an optimum, and a smoother convergence towards that optimum. The method, which consists in correcting Taylor series approximations using previous design history, is used in the present paper to automatically determine move-limits. Move-limits are the allowable changes in design variables during the optimization of the approximate problem. The exponents, computed in the two-point exponential approximation by matching slopes between two design iterations, are used as a measure of non-linearity of the objective function and constraints with respect to each of the design variables. The relationships between the move-limits and the exponents are established and individual move-limits are computed and applied to each design variable. The method is applied to two classical structural optimization examples. It provides the engineer with more flexibility when choosing the move-limits and typically converges in less iterations.