Reliability sensitivity based on profust model: An application to aircraft icing analysis

Local and global reliability sensitivity analysis (RSA) based on the Profust model, a reliability model based on probability inputs and fuzzy-state assumption, can provide useful information in the Profust reliability-based design optimization. It is still a challenge to efficiently estimate two RSAs simultaneously. This paper mainly focuses on the estimation of two RSAs. First, by employing the general performance function (GPF) theory and Bayes theory, a single-loop Monte Carlo simulation (S-MCS) method is proposed to efficiently estimate two RSAs simultaneously. Based on the existing derivation of the GPF, the Profust failure probability is transformed into the traditional failure probability, and then the Bayes theory is combined with the GPF so that two RSAs can be simultaneously completed as the by-product of estimating the Profust failure probability. Second, the key of the proposed S-MCS method is screening out the failure samples from the MCS sample pool, and the adaptive kriging model is embedded in the MCS sample pool to identify the failure samples, which further improves the efficiency of the S-MCS method while ensuring the accuracy. Four examples are given to demonstrate the rationality and efficiency of the proposed algorithm. Meanwhile, two RSAs of an aircraft icing severity analysis model also verify its wide engineering applicability.