基于贝叶斯网络的非正常情况下高铁行车调度人因可靠性分析

Human reliability analysis based on the Bayesian network model under the abnormal circumstances in the case of high-speed rail dispatching

行车调度子系统在整个高铁运输系统中处于关键位置,虽然行车调度大多已经采用了信息化、自动化的作业程序,但发生设备故障、交通事故、自然灾害等非正常情况时,很多具体的作业组织和管理仍需行车调度作业人员的参与。提出了一种基于贝叶斯网络考虑情境环境因子交互关系的人因可靠性分析方法,以某铁路局的人因失误数据统计表为学习样本,推理出行车调度作业人员工作中存在的主要问题。结果表明,"完成任务的可用时间"、"培训"及"人员配置"的节点状态变化会对作业人员的可靠性产生很大影响。最后,以高铁行车调度人员在某种特定情境状态下的实际作业为例进行了实证研究。研究表明,所建立的方法正确可靠,能够准确、高效地辨识人因可靠性。

This paper aims to propose a new method to analyze the reliability of the high-speed rail dispatching operators. For this purpose,applying cause and effect analysis together with Bayesian network model,we have developed a method for analyzing the human factor reliability based on the Bayesian network while taking the interrelations between man- and-environment factors into account. To achieve our goal,on the one hand,we have interviewed a lot of the dispatching operators in the trade and summarized the factors that tend to affect the reliability of the operators under abnormal situations,thus gaining a Performance Shaping Factor（ PSF） in accordance with the screening criteria and the characteristic particularities of the automated control systems. And,as a result,we have worked out our own Bayesian network model for analyzing the reliability of dispatching operators under abnormal circumstances. What is more,we have also prepared a well-informed statistical diagram reflecting the human errors in the administration of a railway,and sorted out the data relevant to human factors reliability that can determine the priori probability and conditional probability of the variable node along with the Bayesian cause-effect inference and diagnostic inference respectively. Such causal-and-effect inference can contribute to identifying the probability of human error in different situations,which can also help the researchers to make further diagnostic inferences in a confident manner. In other words,the researchers can deduce the major problems in the working process of dispatching controllers. Thus,it can be concluded that the change of the node status in-situ on ＂the available time for fulfilling the task＂,＂the training＂and＂the personnel allocation＂can potentially affect the reliability of the dispatching operators to a great extent.Actually much stress has to be attached to such actual process.And,finally,we have conducted an experiment on the actual procedure of high-speed rail transportation dispatching operation in a certain situation,made a comparison between the data reflecting the actual operators＇ reliability with a predicted one in accordance with the network model offered. As a result,we have found that the relative error is 0. 3% and 6. 8% respectively within the permissible range,which proves that the modeling is reliable,and could identify the human factors reliability quite precisely and efficiently.