Computational efficiency of non‐linear programming methods on a class of structural problems

Results are presented of an investigation as to the computational efficiency of several non‐linear programming algorithms on the constrained structural weight optimization of plane trusses and plane stress plates. The optimization algorithms considered are the methods of sequential linear programming, of feasible directions based on Zoutendijk's procedure P1, and of sequential unconstrained minimizations using the methods of Powell, Stewart, Fletcher–Powell, and Newton. Results are presented on computational efficiency as a function of number of design parameters and as a function of the function evaluation effort to derivative evaluation effort ratio.