A new sampling approach for system reliability-based design optimization under multiple simulation models

In this paper, a new system reliability-based design optimization (SRBDO) method is proposed for problems where performance function values are obtained from different simulation models. For this purpose, a new active learning function is derived according to the system type by predicting the system reliability increase after the sample point is added to the design of experiment (DOE) of performance functions in each simulation model. In the proposed SRBDO method, a Kriging model is sequentially updated by adding the optimal sample point to the DOE of performance functions included in the critical simulation model, which can be obtained by comparing the proposed active learning function. The accuracy of the Kriging model and SRBDO optimum convergence are utilized as the stop criteria. The proposed method can be applicable to SRBDO problems regardless of system type. Three numerical and two real engineering examples are investigated to demonstrate the efficiency and accuracy of the proposed method. The validation results indicate that the proposed method is accurate and efficient in finding the SRBDO optimum under multiple simulation models.