Nuclear power plant (NPP) simulators are proliferating in academic research institutions and national laboratories in response to the availability of affordable, digital simulator platforms. Accompanying the new research facilities is a renewed interest in using data collected in NPP simulators for Human Reliability Analysis (HRA) research.
An experiment conducted in The Ohio State University (OSU) NPP Simulator Facility develops data collection methods and analytical tools to improve use of simulator data in HRA. In the pilot experiment, student operators respond to design basis accidents in the OSU NPP Simulator Facility. Thirty-three undergraduate and graduate engineering students participated in the research. Following each accident scenario, student operators completed a survey about perceived simulator biases and watched a video of the scenario. During the video, they periodically recorded their perceived strength of significant Performance Shaping Factors (PSFs) such as Stress.
This dissertation reviews three aspects of simulator-based research using the data collected in the OSU NPP Simulator Facility:
First, a qualitative comparison of student operator performance to computer simulations of expected operator performance generated by the Information Decision Action Crew (IDAC) HRA method. Areas of comparison include procedure steps, timing of operator actions, and PSFs.
Second, development of a quantitative model of the simulator bias introduced by the simulator environment. Two types of bias are defined: Environmental Bias and Motivational Bias. This research examines Motivational Bias— that is, the effect of the simulator environment on an operator’s motivations, goals, and priorities. A bias causal map is introduced to model motivational bias interactions in the OSU experiment. Data collected in the OSU NPP Simulator Facility are analyzed using Structural Equation Modeling (SEM). Data include crew characteristics, operator surveys, and time to recognize and diagnose the accident in the scenario. These models estimate how the effects of the scenario conditions are mediated by simulator bias, and demonstrate how to quantify the strength of the simulator bias.
Third, development of a quantitative model of subjective PSFs based on objective data (plant parameters, alarms, etc.) and PSF values reported by student operators. The objective PSF model is based on the PSF network in the IDAC HRA method. The final model is a mixed effects Bayesian hierarchical linear regression model. The subjective PSF model includes three factors: The Environmental PSF, the simulator Bias, and the Context. The Environmental Bias is mediated by an operator sensitivity coefficient that captures the variation in operator reactions to plant conditions.
The data collected in the pilot experiments are not expected to reflect professional NPP operator performance, because the students are still novice operators. However, the models used in this research and the methods developed to analyze them demonstrate how to consider simulator bias in experiment design and how to use simulator data to enhance the technical basis of a complex HRA method.
The contributions of the research include a framework for discussing simulator bias, a quantitative method for estimating simulator bias, a method for obtaining operator-reported PSF values, and a quantitative method for incorporating the variability in operator perception into PSF models. The research demonstrates applications of Structural Equation Modeling and hierarchical Bayesian linear regression models in HRA. Finally, the research demonstrates the benefits of using student operators as a test platform for HRA research.
Keywords: Human Reliability Analysis, Bayesian Analysis, Simulator Bias, Performance Shaping Factor, Nuclear Power, Structural Equation Modeling, Student Operators, Digital Nuclear Power Plant Simulator, Risk Assessment