The U.S. nuclear industry is increasingly modernizing its operations by integrating automation technologies to improve efficiency, safety, and reliability, while minimizing unnecessary costs. However, successfully deploying automation technologies for operations and maintenance (O&M) in Nuclear Power Plants (NPPs) requires addressing several critical challenges, including ensuring that the automation is trustworthy, transparent, and operationally acceptable. This paper focuses on automation transparency, a characteristic that plays a vital role in enabling safe and efficient human-automation interactions and, thus, contributing effectively to operational risk management. Despite its importance, there is currently no consensus on the definition of automation transparency across various domains, and existing methodologies for measuring transparency are predominantly subjective and context specific. This paper presents a literature review to assess the existing definitions and methodologies of automation transparency and identifies key limitations in their applicability to the nuclear domain. In addressing these limitations, this study proposes a novel definition that refers to automation transparency as “ the degree to which underlying information about the inner workings of an automation system is conveyed, relevant to its intended use. ” Aligning with this proposed definition, three evaluation approaches for automation transparency, namely attribute-based, model-replication-based, and entropy-based, are developed. A hypothetical case study that involves an Artificial Intelligence (AI)-based automated anomaly detection system used in NPP monitoring is leveraged to test these three evaluation approaches to better understand their feasibility, applicability, and limitations. Future work will apply the lessons learned from this study for a practical case study to further understand the impacts of automation transparency on human performance.

The goal of this study is to propose an approach to dynamic human reliability analysis (HRA). This paper introduces a dynamic HRA methodology called Procedure-based Investigation Method of EMRALD Risk Assessment - HRA (PRIMERA-HRA). This method provides HRA analysts with specific guidelines on how to reasonably model human actions, assign human reliability data, and evaluate the output of simulations within a dynamic probabilistic risk assessment tool called Event Modeling Risk Assessment using Linked Diagrams (EMRALD). Although the EMRALD tool was not originally designed for HRA, it possesses most of the essential functions required for implementing dynamic HRA. In this study, an EMRALD model, including human actions in an extended loss of AC power (ELAP) scenario, was developed using PRIMERA-HRA to assess the feasibility of the approach.

Historically, human reliability analysis (HRA) methods require analysts to develop static models of human error within a predetermined human failure event through expert estimation. Objective quantitative models of human performance in the form of a virtual operator offer an avenue to perform dynamic HRA through Monte Carlo simulation. Idaho National Laboratory developed the Human Unimodel for Nuclear Technology to Enhance Reliability (HUNTER) as a dynamic HRA framework and has demonstrated success in modeling existing scenarios. More data is needed to create generalizable virtual human models to explore undefined human failure event space. The Rancor Microworld Simulator is a simplified nuclear process control simulator that students can easily train to use and that can be modified to support the development of concepts of operation for advanced reactors. Rancor-HUNTER fills the data collection niche by integrating the Rancor Microworld Simulator with the HUNTER software. Rancor automates human performance data collection with a digital procedure system compatible with HUNTER’s procedure database through their integration. This digital procedure system provides implicit task level goals associated with each step that affords automatic HRA coded data collection. Therefore, high-resolution sequences of operator behaviors can now be automatically recorded with error rates and time distributions, which collectively represent a virtual human. A thermal power dispatch concept of operations development use case highlights benefits of Rancor-HUNTER to augment human factors evaluation through automated human performance data collection, which can then be used to also inform HRA models.

The Human Unimodel for Nuclear Technology to Enhance Reliability (HUNTER) software is used for dynamic human reliability analysis (HRA). HUNTER creates a digital human twin (or virtual operator) that interfaces with a digital twin (or nuclear power plant simulator). HUNTER is procedure driven, a unique characteristic of safety domains in which much decision making is rule based and captured in procedures. The outputs of HUNTER extend beyond the typical outputs of an HRA estimating method and approach the level of human performance data acquired from human-in-the-loop (HITL) studies using operators and a plant simulator. An advantage of HUNTER is that it creates a virtual human in the loop (VHITL). As such, HUNTER is a unique source of synthetic data on human performance. This paper highlights the use of HUNTER for use in automated evaluations for human factors. HUNTER augments HITL studies by providing a virtual tool to screen human interactions with novel technologies in the control room.

To support data collection for dynamic human reliability analysis (HRA), this study investigates time distributions for task primitives defined in the Goals, Operators, Methods, and Selection rules (GOMS)–Human Reliability Analysis (HRA) method and Human Reliability data EXtraction (HuREX). GOMS-HRA was developed to provide cognition-based time and human error probability (HEP) information for dynamic HRA calculations within the Human Unimodel for Nuclear Technology to Enhance Reliability (HUNTER) framework, while HuREX is a comprehensive HRA data collection method developed by the Korea Atomic Energy Research Institute (KAERI). In this paper, we examine time distributions by using experimental data collected from the Simplified Human Error Experimental Program (SHEEP) study, which proposes an HRA data collection framework to complement full-scope simulator research and gather input data for dynamic HRA by using simplified simulators such as the Rancor Microworld simulator. This paper investigates whether the time required for GOMS-HRA and HuREX task primitives fits 13 statistical distributions. Additionally, we compare and discuss the time distributions obtained from both student operators and professional operators. The result was that this study identified several time distributions for five GOMS-HRA and four HuREX task primitives. In the future, the results of this study are expected to provide objective reference data on the elapsed time for task primitives and aid in realistically simulating scenarios within dynamic HRA.

Advances in technology have led to development of new reactor designs and significant changes in human-system interfaces used by operators to monitor and control commercial nuclear power plants (NPPs). These advances have led the way to novel concepts of operations (ConOps) that are very different from those used in operations for traditional NPPs, i.e., large light water reactors. Accordingly, in collaboration with Idaho National Laboratory (INL), the U.S. Nuclear Regulatory Commission (NRC) started multiple projects to support the NRC’s guidance for human factors engineering (HFE) reviews of advanced reactor applications. As a part of these efforts, INL staff members are currently working on a project to risk-inform the scope of HFE reviews when there are changes to important human actions (IHAs). Specifically, the purpose of the project is to capture specific regulatory aspects of these novel ConOps proposals and risk-inform HFE reviews related to IHAs.

Advances in technology have led to development of new reactor designs and significant changes in human-system interfaces used by operators to monitor and control commercial nuclear power plants (NPPs). These advances have led the way to novel concepts of operations (ConOps) that are very different from those used in operations for traditional NPPs, i.e., large light water reactors. Accordingly, in collaboration with Idaho National Laboratory (INL), the U.S. Nuclear Regulatory Commission (NRC) started multiple projects to support the NRC’s guidance for human factors engineering (HFE) reviews of advanced reactor applications. As a part of these efforts, INL staff members are currently working on a project to risk-inform the scope of HFE reviews when there are changes to important human actions (IHAs). Specifically, the purpose of the project is to capture specific regulatory aspects of these novel ConOps proposals and risk-inform HFE reviews related to IHAs.

In nuclear power plants (NPPs), timely identification of sensor and human errors is critical to ensure safe and efficient plant operations. Anomaly detection models can be employed for this task. However, traditional anomaly detection approaches may have high dependency on labeled data sets and struggle with adaptability in complex, dynamic environments. Reinforcement learning (RL) has demonstrated significant potential in fault diagnosis and anomaly detection; however, its application to anomaly detection in NPPs remains a relatively underexplored research direction.Hence, to address this gap, in this study, we present a novel physics-informed RL model, PIRL-AD (physics-informed reinforcement learning for anomaly detection), that integrates domain knowledge from calorimetric equations into the RL framework for enhanced sensor and human error anomaly detection. We evaluate the performance of PIRL-AD against a nonphysics-informed RL benchmark and a support vector machine (SVM) on data collected from a forced-flow loop testbed.Experimental results suggest that PIRL-AD outperforms other baselines on a range of anomalous data sets that include both sensor and human-induced anomalies across key performance metrics, statistically outperforming the RL and SVM benchmarks with respect to geometric mean (respectively, 92.96% versus 91.06% versus 83.01%) and F1 score (respectively, 89.23% versus 86.98% versus 77.01%). The findings suggest the potential of physics-integrated RL models for enhanced anomaly detection performance in NPPs.

Nuclear power plants in the United States are critical to the nation’s energy security, accounting for 20% of all electricity produced for the power grid. As energy needs grow, 100 gigawatts of additional nuclear power will be necessary by 2050, more than double the current capacity. Realizing this target requires cutting-edge technology like artificial intelligence (AI) and machine learning (ML) that can bring about significant increases in the level of automation. Human-centered AI (HCAI) is a combination of human-centered design (human factors, human-in-the-loop, etc.) with AI/ML to help produce an efficient and reliable system with full consideration for human engagement. This paper provides a comprehensive and novel discussion of HCAI considerations in nuclear power, introducing unique applications for the existing fleet as well as new advanced reactor designs. We include real-life use cases of AI applications to work management processes at nuclear power sites and highlight lessons learned for HCAI.

For safety-critical industries, human error (HE) presents continual risks to system productivity, reliability and safety. Artificial intelligence (AI) and machine learning (ML) methods have emerged as promising approaches to understand, categorize and mitigate the risk of HE in safety-critical industries. This review offers an examination of the current landscape regarding the utilization of AI/ML with regards to HE in safety-critical industries, categorizing literature into descriptive modeling, predictive modeling, prescriptive modeling, and generative modeling techniques. Additionally, the review aims to provide insights regarding themes in literature, challenges, and future research directions. Findings of the review suggest that AI/ML methods can prove useful in addressing the HE problem across safety-critical industries.

The timely and accurate identification of incidents, such as human factor error, is important to restore nuclear power plants (NPPs) to a stable state. However, the identification of abnormal operating conditions is difficult because of the existence of multiple scenarios. In addition, to implement mitigation actions rapidly after an incident occurs, operators must accurately identify an incident by monitoring the trends of many variables. The mental burden posed by this can increase human error and cause failure in identifying incidents. Failure to identify incidents directly results in erroneous mitigation measures, which are detrimental to NPPs.In this study, we leverage uncertainty-aware models to identify such errors and thereby increase the chances of mitigating them. We use the data collected from a physical test bed. The goal is to identify both certain and accurate models. For this, the two main aspects of focus in this study are explainable artificial intelligence (XAI) and uncertainty quantification (UQ). While XAI elucidates the decision pathway, UQ evaluates decision reliability. Their integration paints a comprehensive picture, signifying that understanding decisions and their confidence should be interlinked.Thus, in this study we leverage UQ measures (e.g. entropy and mutual information) along with Shapley additive explanations to gain insights into the features contributing to both accuracy and uncertainty in error identification. Our results show that uncertainty-aware models combined with XAI tools can explain the artificial intelligence-prescribed decisions, with the potential of better explaining errors for the operators.

Nuclear power plants (NPPs) depend on precise procedures for safety and efficiency. These procedures contain work instructions detailing both sequential and conditional tasks. However, executing conditional tasks-tasks performed under specific conditions-often poses difficulties for operators. The hurdles lie in recalling crucial information, spotting key cues, and reacting appropriately to significant plant events linked to these tasks. Current computer-based procedures (CBPs) also fall short in autonomously generating operator reminders when conditional tasks are activated due to these tasks' lack of computer interpretability. This paper presents an ontology: a computer-interpretable representation of conditional tasks. The ontology supports reason about potential events, suitable responses, task constraints, and potential hazards related to the procedure. A case study demonstrates the ontology's effectiveness. Future research will focus on refining ontology and investigating automatic critical information extraction methods.

In nuclear power plants (NPPs), anomalies arising from sensors or human errors (HEs) can undermine the performance and reliability of plant operations. Anomaly detection models can be employed to detect sensor errors and HEs. Additionally, physics-informed machine learning models can utilize the known physics of the system, as described by mathematical equations, to ensure that sensor values are consistent with physical laws. Hence, we propose SPIDARman: System-level Physics-Informed Detection of Anomalies in Reactor Collected Data Considering Human Errors, a holistic physics-informed anomaly detection approach based on generative adversarial networks (GANs) to detect anomalies in both automatically collected sensor data and manually collected surveillance data. We test our approach on data collected from a flow loop testbed, showcasing its potential to detect anomalies. Results demonstrate that the proposed model performs better than the baseline GAN-based models in detecting sensor and surveillance anomalies, suggesting the potential of physics-informed anomaly detection GAN models in NPPs.