A numerical code for lay-out optimization of skeletal structures with sequential linear programming

Abstract

This paper discusses a new Sequential Linear Programming (SLP) algorithm denoted as LESLP (Linearization Error Sequential Linear Programming). The new algorithm implements an advanced strategy to choose the move limits which are defined by limiting the difference between the original nonlinear problem and its linearized counterpart. Besides, LESLP includes a trust region model to increase the design freedom and to improve the overall efficiency of the optimization process. LESLP is tested in five lay-out optimization problems of skeletal structures (i.e. bar trusses and frames) where the objective is to minimize the weight of the structure. It is obviously intended that the objective function does not require structural analysis. The new algorithm is compared to other SLP based techniques and globally convergent optimization methods like Sequential Quadratic Programming (SQP). Results indicate that LESLP is competitive with recently published algorithms and commercial softwares. Also, the new algorithm is robust with respect to starting designs.