基于HFACS-RAs的铁路事故致因建模及混合学习方法研究

发布时间：2018-04-27 21:41

本文选题：人因组织 + 事故学习　；参考：《北京交通大学》2017年硕士论文

【摘要】：铁路系统作为社会复杂系统,其技术的逐步完善一方面提升了设备的可靠性,但设备高集成度和高自动化程度也增加了人机交互的难度。因此,人员的错误和组织缺陷成为导致铁路事故的重要原因。此外,事故中人和组织因素的不确定性和铁路事故文本形式的初始数据给事故调查带来难度。本文设计了针对中国高速铁路系统事故的五阶段混合学习方法,对事故中涉及的人因和系统组织缺陷进行定性定量分析,并基于分析结果给出事故预防手段的建议。论文主要内容如下:(1)结合历史事故数据设计了铁路事故人因分析及分类模型(Human Factor Analysis and Classification System-Railway Accidents,HFACS-RAs)对致因进行分类建模。由不安全行为、不安全行为的前提条件、不安全监管和组织影响四类因素组成的HFACS-RAs模型的层次结构能够从一线人员的错误行为追溯到致使人犯错并间接导致事故发生的组织变量(Governing Variables)因素,有助于将无规律的人和组织因素变得有章可循。(2)结合HFACS-RAs模型的层次分类结构设计了基于网络分析法(Analytic Network Process,ANP)的量化关联性分析法。事故中人和组织致因间的影响关系被划分为内部关联性和外部关联性,并通过搭建超矩阵(Supermatrix)以计算权重。采用模糊决策试验及评价实验法(Fuzzy Decision Making Trail and Evaluation Laboratory,F-DEMATEL)改进内部关联性分析过程,弥补了 ANP方法参数化过程中大量专家判断带来的不可靠性。通过绘制致因的因果图并计算权重,可以获取事故的关键致因,从而缩小解决问题的范围,便于有效地改善现有安全缺陷。(3)以"7 ·23"甬温线特别重大铁路交通事故为研究对象,利用HFACS-RAs模型对事故中人和组织因素分类,并分别通过ANP法和DEMATEL法对归类的致因进行因果关联性分析和权重计算,从而识别导致事故的主要致因并验证方法的可行性。
[Abstract]:Railway system as a social complex system, the gradual improvement of its technology on the one hand to enhance the reliability of equipment, but the high level of integration and automation of equipment also increased the difficulty of human-computer interaction. Therefore, personnel errors and organizational defects have become an important cause of railway accidents. In addition, the uncertainty of human and organizational factors in the accident and the initial data in the form of railway accident text make it difficult to investigate the accident. This paper designs a five-stage hybrid learning method for the accidents of high-speed railway system in China, and makes a qualitative and quantitative analysis of the human factors and system organization defects involved in the accidents. Based on the analysis results, some suggestions on the accident prevention measures are given. The main contents of this paper are as follows: (1) A human Factor Analysis and Classification System-Railway accidents (HFACS-RAsS) model for human factor analysis and classification of railway accidents is designed based on the historical accident data. By unsafe behavior, the precondition of unsafe behavior, The hierarchical structure of the HFACS-RAs model, which is composed of four factors, unsafe supervision and organizational influence, can be traced from the wrong behavior of personnel on the first line to the organizational variable governing variables that cause people to make mistakes and indirectly lead to accidents. It is helpful to make irregular human and organizational factors become rule-based.) combined with the hierarchical classification structure of HFACS-RAs model, this paper designs a quantitative correlation analysis method based on network analysis method (Analytic Network process ANPs). The influence relationship between human and organizational causes in accidents is divided into internal and external correlations, and the weight is calculated by building a supermatrix supermatrix. The fuzzy Decision Making Trail and Evaluation laboratory F-DEMATELL is used to improve the internal correlation analysis process, which makes up for the unreliability caused by a large number of expert judgment in the parameterization process of the ANP method. By drawing causality diagram of cause and effect and calculating weight, the key cause of accident can be obtained, so as to reduce the scope of solving the problem and to effectively improve the existing safety defect. The HFACS-RAs model is used to classify the human and organizational factors in the accident, and the causality analysis and weight calculation are carried out by the ANP method and the DEMATEL method, respectively, so as to identify the main cause of the accident and verify the feasibility of the method.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U298.5

【参考文献】