Study On Human Error And Reliability In Digital Control System Of Nuclear Power Plant

The instrumentation and control (I&C) system of nuclear power plants is transformed from traditional analog control to digital control, which makes the contextural factors influencing human reliability changed, such as technology systems, human-machine interface, procedure et al, operators’cognitive and action modes also changed, then new human error mechanism emerges. The traditional human reliability analysis models pose problem for digital control systems. Technology of HRA in digital control system turns to be hotspot at present. Aiming at the limitations mentioned above, the paper focus on the HRA in digital control system and the main research work can be conducted as follows:(1) The causal model of human error of digital control system for nuclear power plants.The operators’cognitive model is expanded on the basis of Rasmussen’s SRK cognitive framework, the cognitive activities and human errors are analyzed in digital control system for NPPs. The typical performance shaping factors (PSFs) of digital control system for NPPs are identified by the analysis of contextual features of digital control system. Based on the analytical results of human factor events, the prevalence of organizational defects and correlation relationships of PSFs are futher identified. Through the studies related above, the causal conceptual model of human error is built from organizational factors, situational factors, individual factors and human errors based on system theory, and the influencing relationships and degree (types) between human errors and individual factors, situational factors and individual factors are analyzed, which provides a theoretical model and guiding framework for human error analysis and HRA.(2) The HRA method considering the contextual causal relationships.Based on the established human error causal conceptual model, the HRA method considering the contextual causal relationships is built to simulate causal relationships between PSFs. It overcomes the shortcomings that double counting of the effects of PSFs on human reliability because of the dependent and nonorthogonal of PSFs in the traditional HRA methods. The prior probability and conditional probability of node variables in HRA model are obtained by fuzzy method in order to reduce uncertainties due to the expert’s subjective judgments, which makes the quantification of human error probability (HEP) more realistic. Furthermore considering the effects of organizational and management factors on human reliability, the fuzzy Bayesian network method of HRA integrating organizational and management factors is established, the quality of HRA is improved. By the diagnostic reasoning of Bayesian network, the main contributing factors causing human error are identified, which provides decision-making support for the prevention of human errors.(3) The method of dependency of operators’activities considering error recovery.Compared with the traditional analog control system, the dependency of operators’activities is more prevalent in digital control system. Based on the analysis of the characteristics of human factors in digital control system, the influencing factors on the dependency of operators’activities are identified, and the work model for analysing the dependency of operators’activities is established. A fuzzy logic-based approach for analysing the dependency of operators’activities is proposed to assess the dependency level of operators’activities. Namely, it considers the effects of error recovery factors on human reliability to futher improve HRA model and revise the HEP, which makes the results more practical and improves the quality of HRA.(4) The technique of human error risk assessment integrating error effects.The traditional technique of human error risk which only considering HEP to quantify human error risk is modified. Human Error Probability (HEP), Error-Effect Probability (EEP) and Error Consequence Severity (ECS) are integrated to develop an adaptive network-fuzzy inference system (ANFIS)-based approach for assessing human error risk. Furthermore considering the relative weight to obtain data used to model, which makes the evaluation more reasonable. It is a simple, practical and reliable risk prediction tool of human errors in short.