Human Reliability Analysis Of Subway Train Dispatching System Based On The Bayesian Networks

With the improvement of the automation systems technology,human-caused accidents have become more prominent in human-machine systems.Meanwhile,the subway train dispatching system is the core of subway operation.Consequently,it is crucial to analyze the reliability of human factors of the subway train dispatching system.The reliability of human factors has developed maturely in the fields of nuclear power and aviation.Nevertheless,it is still at the initial stage of research in the subway train dispatching system.Based on the preceding analysis,the following researches are carried out:Firstly,this paper introduce the function and structure of the subway train dispatching system in detail,analyze the characteristics of workers in the system as well.Based on the result of relevant papers and reports,investigating,and interviewing,the factors influencing the reliability of subway train dispatching system are preliminarily screened out.After that,the author would design a questionnaire named Questionnaire on factors influencing the human reliability of subway train dispatching system.The influencing factors were further determined by pre-investigation and analysis,and the questionnaire was optimized.After the formal investigation,the reliability and validity of the questionnaire are excellent and the sample size is reasonable.By using the exploratory factor analysis(EFA),the factors influencing the reliability of subway train scheduling system are determined by 4 main factors(individual,organization,equipment and environmental factors)and 20 observable sub factors.Secondly,this paper will establish a structural equation model of the human reliability of subway train dispatching system,which is constructed by 20 sub factors,4 main factors and the human reliability of subway train dispatching system as variables.Then,the t-rule recognition shows that the confirmatory factor analysis of the structural equation model can be analyzed.And the AMOS 23.0 software is used to complete the model construction,model validation factor analysis and evaluation,and the unweighted least square method is used to estimate the parameters.According to the standardization coefficient,we get the relative influence of the organization factor on the human reliability of the subway train dispatching system is the most,and the relative importance of the sub influence factors in themain influencing factors is also obtained.According to the fitting index,we get the causal relationship between the variables of the model.Eventually,the causal relationship between various influence factors of the human reliability in the subway train dispatching system is determined.Finally,based on the causality between the various factors,a Bayesian network of human reliability is constructed for the subway train dispatching system.The triangular fuzzy number is used to transform the semantics of the questionnaire data into probability values,so as to determine the marginal probability of the root nodes of the Bayesian network.In this paper,the Noisy-OR model is used to compute conditional probability of intermediate nodes and leaf nodes in Bayesian network.At the same time,the reliability quantification value P(A)=0.9899 is obtained for the subway train dispatching system under the emergency of subway operation.The posterior probability of each node under the condition of the failure of the subway train dispatching system is obtained by using Bayesian network diagnostic reasoning.According to the percentage of the change of probability,the sensitive factors that affect the reliability of the subway train dispatching system are acquired.Among them,the sensitivity of “team cooperation”,“emotional state”,and “education and training adequacy” rank the top three,which have great influence on the human reliability of the subway train dispatching system.The corresponding suggestions and measures should be put forward to solve the problem.