Uncertainty-based Validation of Maintenance-Physical Degradation Coupled Models in the Absence of Maintenance Performance Data: Application for Nuclear Power Plant Piping Reliability Analysis

Components experience physical degradation and maintenance activities that impact their reliability. Simulation techniques are used to couple Human Reliability Analysis (HRA)-based maintenance models with physical degradation models, but their empirical validation is difficult due to the unavailability of relevant data. This research advances the Probabilistic Validation methodology (Base PV), an uncertainty-based approach to validation. The advancement of the Base PV methodology is necessary for this study, as it currently cannot account for bi-directional interactions within the same hierarchical level, nor does it fully address the unique characteristics of HRA-based models. This necessitates an advanced approach to uncertainty propagation and aggregation across the coupled models. Another advancement develops the capability to handle situations where no validation data is available at the output level. In the Base PV methodology, it was assumed that limited output-level validation data would be accessible. To address this, an advanced methodology is needed to quantify model-form uncertainty. A literature review of existing uncertainty analysis approaches in maintenance models revealed a lack of systematic approach for identifying uncertainties and a limited exploration of human performance and its associated uncertainties. This research addresses these gaps and offers (i) a methodology for characterizing and quantifying uncertainty in HRA-based maintenance models using a theoretical causal model and Bayesian Belief Network quantification, enabling comprehensive coverage of epistemic uncertainties and (ii) a computational platform for propagating uncertainties from HRA-based models to system reliability. The advancements are applied to a nuclear power plant piping reliability analysis.