RBFネットワークによる繊維強化複合材の逐次近似最適化

抄録

Carbon fiber reinforced plastics (CFRP) has strong anisotropy in addition to high specific strength and specific rigidity. This makes optimal design essential for effective utilization of CFRPs. Also, in recent years, it has become possible to create CFRP structures with curvilinear reinforcing fibers, and the degree of freedom for optimum design is further increased. A common approach is direct application of the metaheuristic method to these design problems. However, it is difficult to apply metaheuristics to the problem requiring long calculation time to evaluate an objective function. From such a background, it is desirable to quickly obtain an approximate optimal solution with small number of evaluations. In this research, we perform the optimization with the Sequential Approximate Optimization (SAO) using a radial basis function (RBF) network and reduce the number of calculations for evaluation of the objective function.