Sequential Approximate Reliability-Based Design Optimization Using Window Management

In this research, a new sequential approximate reliability-based design optimization (SARBDO) algorithm is proposed. Existing RBDO algorithms require an iterative numerical analyses due to the reliability analysis (RA) and a double-loop structure of the design optimization. Because of the numerical burden of RBDO, it is hard to apply existing RBDO algorithms to practical design problems required numerically expensive simulation. In order to solve these drawbacks, the proposed approximate RBDO algorithm adopts a concept of the sequential approximate optimization(SAO) which is efficiently superior. The proposed algorithm doesn't need additional simulation except for the generation of the metamodel and gradient evaluations. The proposed algorithm assigns a design window according to the target reliability index, and sets the size of an approximation window larger than the design window for avoiding extrapolation result from sampling method which is one of RA. The proposed algorithm adaptively updates the location and size of the design and approximation windows considering the convergence history of the sequential approximate RBDO