Investigation of Human Factors Using HFACS Framework—A Case Study for Unintended Reactor Trip Events in NPP

Abstract

Safety remains to be the principal concern in nuclear industry, and layered protective measures are implemented to ensure safety to public and environment. Nevertheless, accidents in a nuclear plant do occur, and human factors are confirmed to be the prime cause of these accidents. Several studies have been carried out for human factor analysis, and still, there is a need for additional methods to model human errors. In this paper, human factor analysis and classification system (HFACS) is used to analyze the causal factors for an accident/incident. HFACS is a validated and reliable model for human factor analysis and provides a simple way to evaluate the causal factors in an accident. Although HFACS framework can be applied to analyze any accident events, in this paper, it is applied to review eighteen unintended reactor trip events due to human error from 2000 to 2006. In this study, experts analyze and rate the events against each hierarchal level in HFACS. Expert judgment is sought as the uncertainty and complexity involved in modeling human and organizational factors are very high. Homogeneity of experts is assessed using inter-rated reliability and Cohen’s kappa value. Expert evaluation on human factors is quantified and based on the correlation in rating, most significant factor is identified. Further, based on the information on the reported events and the analysis, statistical tests are employed to confirm the associations between the hierarchal levels in the HFACS framework for reactor trip events. The results indicate four pairs of factors, viz. resource management, planned inappropriate actions, adverse mental states, and skill-based errors have strong associations between adjacent categories. Based on these associations, adverse mental states are deemed the most potent for accident occurrences. Further analysis indicated that slips of attention, memory failures for events such a failure to open or close the valves, and perception failures for visual illumination (i.e., ‘skilled based errors’) are the most common ‘unsafe acts’ committed and appear more frequently along with nuclear control room actions. The results reveal that human errors are caused mainly due to potential human factors associated with cognition. Few active failures are also identified. Thus, the analysis highlights the importance of attention toward possible human factors among others. Overall, HFACS tool proves useful in categorizing operator errors from reported events. HFACS framework and the quantification method help to determine which factors are more dominant and influential in the accident sequence and provide additional insight for the analysts into the significance and relative importance of each of the human factors.