Abstract for 1998 SIAM Annual Meeting
Design Optimization Using Surrogates


We consider problems where the number of design variables is relatively small (less than 100) but where the cost of a single simulation to evaluate the objective is computationally expensive. We make use of surrogates of the objective to drive the optimization procedure since such surrogates are constructed, in part, to be inexpensive to evaluate. We present a general approach that is both computationally tractable and globally convergent and present results from a helicopter rotor blade design problem.