NOTE ON THE EFFECTIVENESS OF STOCHASTIC OPTIMIZATION ALGORITHMS FOR ROBUST DESIGN

摘要: Robust design optimization (RDO) uses statistical decision theory and techniques to optimize a over range of uncertainty (introduced by the manufacturing process unintended uses). Since engineering ob jective functions tend be costly evaluate prohibitively expensive integrate (required within RDO), surrogates are introduced allow use traditional methods find solutions. This paper explores suitability radically different (deterministic stochastic) solve prototypical robust problems. The algorithms include genetic algorithm using penalty function formulation, simultaneous perturbation stochastic approximation (SPSA) method, two gradient-based constrained nonlinear optimizers (method feasible directions sequential quadratic programming). results show that fully deterministic standard consistently more accurate, likely terminate at points, considerably less than nondeterministic algorithms.