Evaluation of Automation Trustworthiness in Nuclear Power Plants: A Risk-Informed Approach using Probabilistic Validation and Integrated Probabilistic Risk Assessment

Hammad Khalid, Istiaque Ahmed, Samrendra Roy, Ha Bui, Seyed A. Reihani, Ahmad Al Rashdan, Zahra Mohaghegh

Research output: Chapter in Book/Report/Conference proceedingConference contribution


In the pursuit of enhanced efficiency and operational safety, the U.S. nuclear industry is progressively integrating automation technologies into Nuclear Power Plants (NPPs). To make informed decisions about large-scale investments in these technologies, stakeholders require robust evidence of their transparency, trustworthiness, and operational acceptability. This study presents a risk-informed approach to evaluate automation trustworthiness, combining an advanced Integrated Probabilistic Risk Assessment (I-PRA) methodological framework with a Probabilistic Validation (PV) methodology. This paper is part of an ongoing project, funded by the Department of Energy (DOE) under the Nuclear Energy University Program (NEUP), which aims to develop a probabilistic validation and risk importance ranking methodology for automation trustworthiness and transparency in NPPs. The I-PRA framework, developed by the PI's team in previous studies, connects simulation models of underlying physical and social phenomena with the existing plant PRA model through a probabilistic interface. I-PRA adds realism into the risk estimation by explicitly incorporating time and space into the underlying models while avoiding significant changes to the plant PRA model and its associated costs (e.g., peer review). In this project, I-PRA is advanced by adding or advancing automation-related modules to enable a clear tracing and capture of relationships between the plant risk metrics and input parameters associated with the underlying physics, automation, and human performance. A Global Importance Measure (GIM) method is added to I-PRA to rank significant sources of uncertainty at the level of underlying physics, automation, and human performance concerning their contribution to the plant risk uncertainty. The PV methodology lies at the core of our approach for evaluating automation trustworthiness. It characterizes and propagates the uncertainties associated with the human-automation-physics coupling, with the degree of automation trustworthiness measured by the magnitude of epistemic uncertainty associated with the automation output. Execution of the PV methodology for evaluation of automation trustworthiness in NPPs requires that the uncertainties from the underlying automation model level be propagated up to the level of plant risk metrics, which necessitates a computational platform that is achievable by leveraging the advanced I-PRA framework. The acceptability of the degree of automaton trustworthiness for a specific automation application is evaluated against predefined acceptance criteria that can be found available at one or more levels of the system hierarchy (e.g., automation output level or plant risk metric level). The PV methodology, in conjunction with advanced risk importance ranking results, provides an efficient algorithm to enhance automation trustworthiness. To demonstrate the feasibility and practicality of the advanced I-PRA framework and the PV methodology, this paper presents a case study centered on an AI-based automated firewatch system. By integrating the advanced I-PRA framework with the PV methodology, the proposed approach offers a holistic evaluation of automation technologies' trustworthiness, considering complex interactions among human operators, automation technologies, and the evolution of physical phenomena in NPPs. This comprehensive evaluation will assist in making informed decisions about the deployment and operation of automation technologies in NPPs, ultimately contributing to the enhancement of plant safety and efficiency.
Original languageEnglish (US)
Title of host publicationProbabilistic Safety Assessment and Management (PSAM) International Topical Meeting on Artificial Intelligence (AI) and Risk Analysis
StatePublished - Oct 25 2023


  • Integrated probabilistic risk assessment
  • Nuclear Power Plants
  • Uncertainty analysis
  • AI-based firewatch
  • Probabilistic validation (PV)
  • Automation trustworthiness
  • Risk informed


Dive into the research topics of 'Evaluation of Automation Trustworthiness in Nuclear Power Plants: A Risk-Informed Approach using Probabilistic Validation and Integrated Probabilistic Risk Assessment'. Together they form a unique fingerprint.

Cite this