An efficient method for estimating failure probability bound functions of composite structure under the random-interval mixed uncertainties

For composite structure under random-interval mixed uncertainties (RIMU), the failure probability bound functions (FPBFs), which are defined as the function of the failure probability upper and lower bounds varying with the random input distribution parameters, are significant for decoupling the reliability-based design optimization. Although FPBFs can be estimated by repeatedly executing the FPB analysis at each distribution parameter realization, this strategy is unaffordable for engineering composite structure. To address this issue, an efficient method is proposed based on the line sample (LS) extended to RIMU (LS-RIMU), which requires only one execution of LS-RIMU at a given distribution parameter, instead of repeated executions of LS-RIMU at each distribution parameter. The efficiency improvement results from that the limit state surface (LSS) in the original random input space (ORIS) is determined by the failure physics, which is independent of the distribution parameter, thus it can be employed as an information exchanging station. By use of the LSS in ORIS, the LSS samples at one distribution parameter can be transformed to those at other distribution parameters, on which FPBFs can be estimated without additional computational cost. The efficiency and accuracy of the proposed method are verified by the presented numerical and composite structure examples.