Thumbnail Image


Publication or External Link






Many accidents are attributed to human errors. Abundant evidences could be found in major accidents in petro-chemical, nuclear, aviation, and other industries. In the nuclear power industry, safe operation heavily relies on the operators' interaction with plant systems. For example, Three Mile Island accident was exacerbated by the operators' misdiagnosis of the situation, which led to the termination of the plant's automatic protection system that could have prevented meltdown of the reactor core. Hence, human Reliability Analysis (HRA) is an important ingredient of Probabilistic Risk Assessment (PRA), particularly in the nuclear industry. HRA aims to predict possible human errors, identify "error forcing contexts", and assess error probabilities. Despite advances in HRA discipline over the past two decades, virtually all existing methods lack a causal model and few leverage the theoretical and empirical insights for error prediction in a systematic and formal way. One approach that has attempted to address this major shortcoming is IDAC crew simulation model of ADS-IDAC dynamic PRA platform. Through the interactions between an IDAC crew model and a pressurizer water reactor plant model, ADS-IDAC dynamically simulates the operators' cognitive activities and actions in an accident condition. The goal of proposed research is to introduce an advanced reasoning capability and structured knowledge representation to enhance the realism and predictive power of in the IDAC model for situations where crew behaviors are governed by both the Emergency Operating Procedure (EOP) and their knowledge of the plant. This is achieved by: 1) Developing and implementing a cognitive architecture to simulate operators' understanding of accident conditions and plant response, their reasoning processes and knowledge utilization to make a diagnosis. A reasoning module has been added to the individual operator model within IDAC model to mimic operators knowledge-based reasoning processes; 2) Developing and applying a comprehensive set of Performance Shaping Factors (PSF) to model the impacts of situational and cognitive factors on operators' behaviors. The effects and interdependencies of PSFs are incorporated the reasoning module; and 3) Performing a calibration and validation of the model predictions by comparing the simulation results with results of a number of plant-crew simulator exercises.