A Novel Correlation‐Based Reliability and Nonparametric Sensitivity Evaluation Method for Correlated Systems

ABSTRACT In structural reliability analysis, input variable correlation affects modeling accuracy, especially with non‐normal, dependent data. Traditional methods often fail to preserve dependence, causing errors in reliability estimates. We propose a novel framework based on fourth‐order L‐moments. This method uses a polynomial transformation to map correlated non‐normal variables into an independent standard normal space, maintaining distribution and correlation structure. Coefficients are optimized for accurate moment matching. In the transformed space, the most probable failure point is identified, and failure probability is corrected for improved efficiency in high‐dimensional problems. A non‐parametric sensitivity analysis evaluates input variable impacts. The method is validated with numerical examples, showing strong agreement with Monte Carlo results. Relative error is reduced by 87.6%, with errors under 1.0% in the cam mechanism case.