逐次近似最適化を用いた信頼性に基づく最適設計手法：

大域的探索のためのアルゴリズムの提案

抄録

This study proposes a new algorithm for a response-surface-assisted Reliability-Based Design Optimization (RBDO) method. RBDO is one of optimization techniques considering uncertainty and has gained attention in engineering field. Many studies have focused on response-surface-assisted RBDO methods since the number of function calls in RBDO is often larger than the deterministic optimization problem and the burden of the reliability analysis is huge. Especially, the sequential approximation optimization technique for RBDO problems has been studied actively. On the view of additional sampling strategies, these studies assumed that it is important to improve accuracy of reliability analysis. Therefore, sample distribution is often concentrated. However, it is preferable that the distribution is uniform because the global aspects of problems can be obtained. Additionally, few studies applied global search methods for optimization. In this study, an additional sampling strategy considering both improving accuracy of reliability analysis and searching the sparse region is developed. To utilize global search methods, the formulation in RBDO is converted into a deterministic formulation. Through two numerical examples, the validity of proposed method is discussed.