Evaluating the Validity of a Probabilistic Physics-of-Failure Model with Finite Element Integration for Nuclear Power Plant Piping Reliability

The authors’ previous research advanced the reliability assessment for nuclear power plant piping by integrating a probabilistic physics-of-failure (PPoF) model with a finite element (FE) analysis. This approach provides higher resolution by explicitly analyzing location-specific and time-dependent conditions. The objective of this paper is to evaluate the credibility of this FE-based PPoF model.The empirical validation of piping reliability estimates is challenging due to limited failure data for specific mechanisms and pipe diameters. As an alternative, this study implements probabilistic validation (PV), in which model validity is assessed by comparing the total uncertainty in predictions with a regulatory risk threshold of 1.0E-6 per reactor year. A distinctive feature of the FE-based PPoF model is the presence of strong correlations among the FE input parameters representing the initial and boundary conditions.This study demonstrated a unique integration of the Spearman rank correlation and the Iman-Conover transformation into PV. In addition, a global sensitivity analysis in the extended PV methodology identified dominant contributors to input uncertainty, supporting the targeted model and parameter refinements to enhance the credibility of the FE-based PPoF model. While the extended PV methodology is demonstrated for evaluating the credibility of the FE-based PPoF model, the methodology can be generally applicable for any type of the model with correlated inputs.