Efficient reliability-based design optimization for hydraulic pipeline with adaptive sampling region

In engineering practice, the active learning Kriging (AK) model is always embedded in the computation of reliability-based design optimization (RBDO) problem. Traditionally, with the iteration of the design points, the candidate sample pool for establishing the Kriging model must be constructed repeatedly. However, there are overlaps in the sample space, which affect the efficiency of the RBDO. In this study, a novel method combining the rejection sampling (RS) with AK is proposed to further ensure the accuracy of the Kriging model. The proposed method aims to reduce the number of candidate sample pools and improve the efficacy of sample utilization. In the optimization process, RS can capture the variation of the sampling region using the rejection algorithm, which avoids the need to update the Kriging model in the entire overlapping sample space, and focuses on improving the model accuracy in the additional sample space with the new design point. Thus, the Kriging model continuously updates with the iteration of the design variable in the optimization process, which significantly reduces the computational cost. Several examples were employed to validate the performance of the proposed method. Lastly, the proposed method was applied to the RBDO of a hydraulic pipeline system.