The reliability and the efficiency of the convex linearization method are greatly enhanced for structural optimization problems. The reliability is improved by automatically controlling and updating the parameters needed in the optimization process. The efficiency is enhanced by controlling the number of ε-active constraints.
A computer program is developed by modifying the STROD (structural optimal design) program and the ADS (automated design synthesis) program. Ten examples are selected from the typical and practical problems of structural optimization and the results are compared with those of the recursive quadratic programming methohd and those of the sequential linear programming method. Convergence to local optima is obtained in about ten iterations, in all cases and regardless of initial designs. The improved method gives local optima in a considerably shorter CPU time than the other methods.
Future researches will be needed to reduce further the number of iterations and to improve the efficiency of solving sub-problems obtained by convex approximation.
