1998 年 64 巻 621 号 p. 1138-1145
Though genetic algorithms are effective for combinatorial optimization, they often need large number of function evaluations. Moreover, integration of analysis software with the genetic algorithm is sometimes cumbersome. Response surface methodology has been proposed recently for overcoming these two difficulties. In the present study, response surfaces are constructed in lamination parameters for laminated composites. The construction of the response surface in lamination parameters reduces the number of design parameters, with an attendant reduction of the design cost. Applicability of this approach is studied for optimizations of stacking sequences of simply supported laminated plate for maximization of buckling load. A two-stage approach, relying on initial continuous optimization, is used to zoom the response surface to the region of the optimum design for uncreased accuracy.