Towards reliability evaluation involving correlated multivariates under incomplete probability information: A reconstructed joint probability distribution for isoprobabilistic transformation

Reliability evaluation under incomplete probability information (prescribed marginal distributions and correlation coefficients) is a challenging task. The widely used Nataf transformation inherently assumes a normal copula for dependence modeling, which can be inappropriate in some cases. This paper aims to provide a more general isoprobabilistic transformation method for reliability evaluations under incomplete probability information. To this end, the joint probability distribution is represented using the pair-copula decomposition approach, which is highly flexible in dependence modeling. The desired pair-copula parameters are retrieved from the incomplete probability information by a simulation based method. Finally, based on the reconstructed joint probability distribution, the Rosenblatt's transformation is adopted for the subsequent reliability evaluation. The proposed method is illustrated in a tunnel excavation reliability problem. Several dependence structures characterized by different pair copulas are investigated to provide insights into the effect of copula selection on reliability results. (C) 2017 Elsevier Ltd. All rights reserved.