Welcome to the PSAM 18 Abstract Status page.
Abstract KR220Full Paper + Presentation
Quantifying and Reducing HRA Uncertainty in Advanced Reactors: Bayesian Inference Approach
Advanced nuclear reactors, such as small modular reactors and microreactors, introduce new challenges for Human Reliability Analysis (HRA) due to differences in design characteristics, operational context, information availability, and design maturity compared to conventional light-water reactors (LWRs). One distinct characteristic of these advanced reactors is increased digitization and automation, which can affect operators’ performance. A literature review of HRA applications for microreactors shows that existing studies largely rely on qualitative categorization, lump-sum frequencies from generic databases, or simplified nominal human error probability (HEP) values.
The U.S. Nuclear Regulatory Commission (NRC) has developed the Integrated Human Event Analysis System for Event and Condition Assessment (IDHEAS-ECA) as the state-of-the-art HRA method to support risk-informed regulation. The associated database, documented in IDHEAS-DATA (RIL 2025-01), is built primarily from simulator data and other sources specific to conventional LWRs. In general, increased digitization and automation in advanced reactors can shift cognitive failure modes and performance-influencing factors toward contexts where existing human performance data are relatively sparse.
This paper addresses two research questions: (i) how much additional uncertainty can limited data introduce into HRA for an advanced reactor context, and (ii) if the uncertainty is unacceptably large, how much additional data is needed to reduce it? A Bayesian inference approach is proposed to construct probability distributions representing uncertainties in model parameters of IDHEAS-ECA, including base HEPs and performance-influencing factor (PIF) weights. Probabilistic programming is implemented using the PyMC Python package. The proposed approach is demonstrated using synthetic data designed to reflect the structure and quantity of the IDHEAS-DATA dataset. While application to the actual IDHEAS-DATA dataset is ongoing, the results from the synthetic data presented here provide an order-of-magnitude insight into the HEP uncertainty associated with IDHEAS-ECA.
← Check another abstractThe U.S. Nuclear Regulatory Commission (NRC) has developed the Integrated Human Event Analysis System for Event and Condition Assessment (IDHEAS-ECA) as the state-of-the-art HRA method to support risk-informed regulation. The associated database, documented in IDHEAS-DATA (RIL 2025-01), is built primarily from simulator data and other sources specific to conventional LWRs. In general, increased digitization and automation in advanced reactors can shift cognitive failure modes and performance-influencing factors toward contexts where existing human performance data are relatively sparse.
This paper addresses two research questions: (i) how much additional uncertainty can limited data introduce into HRA for an advanced reactor context, and (ii) if the uncertainty is unacceptably large, how much additional data is needed to reduce it? A Bayesian inference approach is proposed to construct probability distributions representing uncertainties in model parameters of IDHEAS-ECA, including base HEPs and performance-influencing factor (PIF) weights. Probabilistic programming is implemented using the PyMC Python package. The proposed approach is demonstrated using synthetic data designed to reflect the structure and quantity of the IDHEAS-DATA dataset. While application to the actual IDHEAS-DATA dataset is ongoing, the results from the synthetic data presented here provide an order-of-magnitude insight into the HEP uncertainty associated with IDHEAS-ECA.