A data mining framework within the Chinese NPPs operating experience feedback system for identifying intrinsic correlations among human factors

With the continuous increase in the number of operating nuclear power plants (NPPs) in China, the amount of operating experience feedback (OEF) increases significantly. On the other hand, the safe operation of NPPs has become an urgent problem that the National Nuclear Safety Administration (NNSA) must solve. To this end, NNSA established a nationalwide OEF system to improve the safety level of NPPs and strengthen the exchange of operating experience. Analyzing the human factors events (HFEs) is an important part of OEF and it is significant to improve human performance and prevent human error. Data mining has been recognized as an effective way to analyze data. With the continuous increase in operating event reports, data mining related to nuclear safety becomes a new domain of study. In this paper, we propose a data mining framework in support of the OEF system. The framework combines three statistical approaches (i.e., correlation analysis, cluster analysis and association rule mining) for identifying intrinsic correlations among human factors: correlation analysis measures the strength of linear relationship between human factors; cluster analysis classifies human factors into relevant groups; association rule mining identifies associations and causalities among human factors. For illustration, we apply the proposed framework to 162 human factors events (screened out from 313 events collected from the OEF system), and the results reflect the feasibility and effectiveness of the framework in identifying the intrinsic correlations among human factors. Besides, further suggestions for improving human performance and preventing human errors in NPPs are also discussed.