A Reliability-Based Design Optimization Method with Probabilistic-Interval Hybrid Reliability Model and Adaptive Kriging Model

Interval variables are more appropriate for situations where the number of samples in engineering is small and the probability distribution cannot be obtained. Most variables in engineering have upper and lower bounds and are better described using truncated tail probability variables. A truncated probability-interval hybrid reliability index is more appropriate for characterizing uncertainty in engineering. To achieve lightweight structures in engineering or determine the optimal objective value, traditional deterministic optimization methods are insufficient as they do not consider variable uncertainty. Therefore, it is more reasonable to use uncertainty optimization. The two-level reliability optimization algorithm, traditionally used, suffers from iterative non-convergence and low robustness. A more efficient approach is to use Sequence Optimization and Reliability Assessment (SORA). This paper combines sequence optimization and reliability assessment methods with the hybrid reliability index to obtain optimized results that meet the index's requirements. For the problem of implicit functions, the Kriging model and sequence optimization are combined with methods for reliability assessment. Optimized Latin hypercube sampling is used to meet the global sampling requirement. The method of adding samples in the vicinity of the point of maximum possible failure using the Expected Improvement (EI) criterion is contrasted with the method of initially taking a large number of samples. The results indicate that utilizing the initial small sample plus locally increasing samples approach requires fewer final samples and yields more precise results.