【作者】 张亚东；

【导师】 郭进；

【作者基本信息】 西南交通大学 ， 交通信息工程及控制， 2013， 博士

【摘要】 高速铁路的发展,极大提高了列车的运行速度,缩短了城市间的时空距离,方便了人们的出行,促进了区域经济的繁荣与文化的交流。当列车运行速度提高到一定程度以后,依靠司机瞭望和人工驾驶已难以保证行车安全。根据国际铁路联盟规定,当列车运行时速超过160km时,为保证高速铁路的行车安全,必须装备列车运行控制系统(简称列控系统)。列控系统是实时控制列车安全运行间隔、防止列车超速运行的高速铁路核心技术装备和安全关键系统,对于保障高速铁路行车安全、提高运输效率具有重大作用。列控系统综合应用了计算机、现代通信和自动控制等技术,由车载设备和地面设备组成,系统庞大,在组成结构、功能层次、功能执行过程和状态变迁等方面都极其复杂,各种随机失效和系统失效均可能导致极其严重的后果,与传统铁路信号系统相比面临更加苛刻的安全需求。由于高速铁路列控系统是我国铁路信号领域中的新技术,系统未经过现场长期应用的验证,部分技术规范仍处于不断修订和完善之中,系统中许多潜在的安全风险尚未完全掌握,因此,基于经验及技术规范的传统安全保障手段已不能满足列控系统的安全需求。本文基于系统安全风险理论和方法,利用模糊不确定理论、基于逼近理想解的排序法(TOPSIS)、贝叶斯网络、可拓学以及Petri网等建模理论,围绕高速铁路列车运行控制系统安全风险辨识及分析的关键问题展开研究,论文的主要研究内容与成果包括：1.将列控系统自上而下划分为系统层、子系统层、单元层、单元板层和模块层,在分层的基础上,分别从系统组成、功能层次、状态变迁和功能执行过程等多维视角,提出了列控系统结构参考模型、功能分层模型、基于P/T系统的状态转移模型和基于SPN的功能执行过程模型的构建与验证方法,并结合危险与可操作性分析(HAZOP)技术,提出了基于结构参考模型、功能分层模型、状态转移模型和功能执行过程模型的安全风险辨识方法,可以提高列控系统安全风险辨识的系统性和全面性。以列控中心子系统为例,分别建立了列控中心的结构参考模型、功能分层模型、状态转移模型和临时限速设置功能执行过程模型,并给出了危险源识别的示例。2.针对安全风险等级分析过程中存在的模糊不确定性,建立了基于模糊不确定性理论的列控系统危险源的安全风险等级分析模型。首先基于风险矩阵构建了危险源的安全风险等级推理规则库,分别从危险源发生的可能性和后果严重性两个方面,利用模糊群决策、模糊层次分析法和多级模糊综合评判,建立了危险源的发生频率等级分析模型和后果严重度等级分析模型,最后结合推理规则,实现了危险源安全风险等级的模糊推理。3.在安全风险等级分析的基础上,提出了相同等级的危险源的安全风险排序问题。从危险源发生的可能性和后果的严重性两个方面,建立了列控系统危险源的风险评价指标体系,基于模糊层次分析法计算了各评价指标的权值。在指标体系构建及权值计算的基础上,应用TOPSIS法和模糊集理论,建立了基于多级模糊TOPSIS法的列控系统危险源安全风险排序模型,可以从众多相同安全风险等级的危险源中分离出风险更高的危险源,为加强危险源的重点管控提供科学依据。4.基于故障树和事件树的概率安全风险分析方法,融合贝叶斯网络和模糊集理论,研究了基于故障树和事件树的贝叶斯网络模型的构造方法、基于多专家模糊评判的贝叶斯网络根节点先验概率的分析算法以及基于等效死亡的危险源定量安全风险计算方法,提出了基于贝叶斯网络的列控系统危险源的定量安全风险分析模型,统一了危险源的原因分析模型和后果分析模型,继承了故障树和事件树建模的优点,降低了贝叶斯网络模型构造以及根节点先验概率获取的难度,既可以定量计算危险源的安全风险,又可以诊断分析导致不同后果的主要因素及其后验概率。以列控中心错误驱动信号继电器为例,建立了列控中心错误驱动信号继电器的贝叶斯网络模型,利用基于多专家模糊评判的根节点先验概率分析算法,通过仿真得到了贝叶斯网络模型中各个根节点的先验概率。最后利用基于聚类的贝叶斯网络推理算法,通过因果推理,定量计算了列控中心错误驱动信号继电器及各种可能后果的发生概率,并结合等效死亡的概念,得出了列控中心错误驱动信号继电器的安全风险；通过诊断推理,分析了导致不同后果发生的主要因素及其后验概率。5.从列控系统的系统层面和运营角度,提出了列控系统运营安全风险分析的问题。基于安全系统工程的原理,从设备、操作、维修、管理、环境和更新改造等方面,建立了列控系统运营安全风险的评价指标体系,并应用模糊层次分析法计算了各评价指标的权值。在指标体系构建及权值计算的基础上,将可拓学引入列控系统运营安全风险分析,建立了列控系统运营安全风险多级可拓分析模型,既可以评判列控系统运营的整体安全风险水平,又可以评判各单项评价指标的安全风险水平,为查找列控系统运营期存在的薄弱环节提供科学依据。更多还原

【Abstract】 The development of high-speed railway leads to the great improvement of train speed, which decreases the time and distance between cities, facilitates trips, promotes the economic development and culture exchange. Manual driving and monitoring can no longer ensure traffic safety when speed is over certain level. According to international union of railways rules, trains with speed higher than160km/h must be equipped with train control system to ensure safety. Train control system is the core equipment and safety-critical system of high-speed railway which real-time controls train safe operation distance and protects train from over speed. It plays an important role in ensuring the high-speed railway safety and improving transport efficiency. Train control system integrates technology of computer, modern communication and automation, and it is an immense system that composed of on-board unit and ground equipment. It is very complex at composition, structure, functional level, function execution process and state transition aspects. So it has more oppressive safety requirements compared with traditional railway signaling systems, all kinds of random failure and system failure may leads to extremely serious results. Because high-speed railway train control system is the new technique in Chinese railway signaling area, it hasn’t experienced tests with long-term field application, part of its technical manual are still under revise and polish, many potential safety risks are not mastered yet. So, traditional safety methods can no longer fulfill safety requirements of train control system.This thesis uses system safety risk theory and method, together with modeling theories including Fuzzy uncertainty theory, TOPSIS, Bayesian Network, Extenics, Petri Net, to do research on key problems of high-speed railway train control system safety risk identification and analysis. The main research contents and achievements are listed below.1. Train control system was divided into system level, subsystem level, constitutional unit level, unit board level and module level from top to bottom. Base on such hierarchy, the thesis integrated the views of composition, functional level, state transition and function execution process to propose build and test methods of structural reference model, function hierarchical model, state transition model based on P/T system and function execution process model based on SPN of train control system. Combined with HAZOP, it also raised safety risk identification method which based on structural reference model, function hierarchical model state transition model and function execution process model. In the way, the systematicness and comprehensiveness of train control system safety risk identification was increased. Took train control centre as an example, the thesis set up structural reference model, function hierarchy model, state transition model and temporary speed restriction execution model, and also provided examples of hazard identification.2. Aiming at fuzzy uncertainty of safety risk level analysis, the thesis established the safety risk level analysis model of train control system hazard which based on Fuzzy uncertainty theory. Firstly, it set up hazard safety risk level inference rule library based risk matrix. Then, according to possibility and seriousness of hazard, it used FGDM FAHP and fuzzy comprehensive evaluation to build analytical models of hazard frequency grade and consequence seriousness grade. At last, combined with inference rule, it completed fuzzy inference of hazard safety risk level.3. Raised the question about how to reordering hazards with the same safety risk level based on safety level analysis. Set up risk assessment index system of hazard from possibility and seriousness and calculate weight of each index using FAHP. According to such evaluation index system, the thesis used TOPSIS and fuzzy set theory to build train control system hazard safety risk ordering model which based on multilevel fuzzy TOPSIS. Using this ordering model, more serious hazard can be isolated from hazards with the same safety risk level. In this way, scientific basis of enhancing hazard control was provided.4. The thesis raised train control system hazard quantitative safety risk analysis model which based on Bayesian Network by researching Bayesian network model construction method which based on fault tree and event tree, the algorithm of prior probability of Bayesian network root node based on multiple experts fuzzy evaluation, computing method of hazards quantitative safety risk which based on equivalent of death, and also combining Bayesian network, fuzzy set theory and probability safety risk analysis method which based on fault tree and event tree. This model integrated hazard’s causes and consequences analysis model, inherited the advantage of fault tree and event tree modeling, decreased the difficulty of Bayesian network model structure and obtaining prior probability. It not only quantitatively calculated hazard’s safety risk, but also diagnosed dominant factors and posterior probability for different consequences. Used error-driven signal relay in train control center as an example, the thesis established its Bayesian network model, and calculated prior probability of each root node in Bayesian network model according to algorithm based on multiple experts’fuzzy evaluation. At last, it used cluster Bayesian network inferential algorithm and causal inference to quantitatively calculate the probability of every possible consequence of error-driven signal relay, and also quantitatively analyzed major factors and posterior probability of each possible consequence through diagnostic reasoning.5. Posed the problems of train control system operational safety risk analysis from the system layer and operating aspect. Set up evaluation index system of train control system operational safety risks, calculated weight of each index using FAHP from equipments, operation, maintaining, management, environment and updating aspects. Then introduced Extenics to train control system operational safety extermination, built train control system operational safety risks multiple grade extenics evaluation model. So it can not only estimate the overall safety risks level of train control system operation, but also estimate single index of safety risks level. In this way the scientific basis for seeking vulnerable spot of train control system during the operation service period was provided.更多还原