An adaptive hybrid approach for reliability-based design optimization

Abstract

Reliability-based design optimization (RBDO) is a powerful tool for design optimization with consideration of uncertainty. It can be solved by double loop approaches or single loop approaches, while double loop approaches are robust but their implementation is computationally costly. On the other hand, single loop approaches are highly efficient but may have convergence problem for highly nonlinear performance measure functions. To mend their respective drawbacks, we resort to a transition between them and propose the so-called adaptive hybrid approach (AHA) to take advantage of these two approaches. Based on a function type criterion, AHA adaptively selects the single loop or double loop approaches during the iteration. When single loop strategy is selected, the advanced mean value (AMV) method is used. When double loop strategy is selected, an improved adaptive chaos control (ACC) method is proposed to searches for the most probable target point (MPTP) of black-box function robustly and efficiently. Four illustrative examples, including two nonlinear analytical problems and two engineering applications, demonstrate the superior efficiency and robustness of the AHA over other prevalent approaches.