自动控制文献翻译——摘要

中国列车控制系统(CTCS-3)，是一个规模大、结构复杂、有冗余配置的安全性计算机控制系统。CTCS-3以及相关的设备和技术，能确保的高速运行情况下列车的安全性和可靠性，所以对其可靠性的评估变得非常重要。一般情况下，在大型复杂系统的可靠性时评估时，常见因素的原因(CCF)和故障模式的多态性这两个元素不能被评估。应用传统的故障树分析(FTA)来对上面提到的两个因素分析时会有局限性，考虑到贝叶斯网络(BN)具有双向推理和不确定性知识解决问题的能力，因此引进BN来改变缺陷的。该方法首先建立了系统从上到下的故障树分析模型，然后故障树分析模式以分层形式从下到上转换成BN模型。最终，可靠性指标是使用关于CCF和多态系数的BN计算的。通过整合和评估子模型使用故障树分析的做法在CTCS-3系统中与BN的结合，收获了一些有趣的结果。相对结果表明，所提出的方法是相当有效的，并且还为提高系统的可靠性提供了理论依据。

关键词:CTCS-3；故障树；贝叶斯网络；可靠性评估

1 引言

高速列车运行的可靠性直接关系到乘客的生命和财产的安全。发生事故时必然会有非常严重的后果。列车控制系统是由核心设备和技术保证高速列车的安全性、可靠性和高效运行，因此它的可靠性和安全性评估具有非常重要的意义。在应用了大量的复杂电子相关元件和计算机系统的高速列车信号系统中，中国列车控制系统(CTCS-3)是安全性控制的电脑系统。CTCS-3有大型复杂结构和关键设备的冗余配置。因此，CTCS-3的可靠性和安全性评估要求越来越高。

在CTCS-3相关性是其失败的共同特征。共因故障(CCF)是一种依赖性的故障，也是一个造成系统内部故障的重要的依赖性因素。如果共因故障被忽略，容易产生较大的误差。目前，在CCF分析中常常应用β因子模型。然而，它只用于在2单元冗余系统。在传统的可靠性分析，常常假定部件或系统只有两种状态，如工作或失败，同时也忽视了相关组件和其他的故障状态对系统性能影响。既定的可靠性分析模型和实际情况存在较大差异。因此，多态系统(MSS)的可靠性分析已受到广泛关注。基于与CCF多态系统可靠性分析被报告。然而，CTCS-3缺乏应用程序。

在系统可靠性分析中，故障树分析(FTA)是常见的方法，已广泛应用于铁路可靠性分析。通过使用数字推导和建立逻辑图，FTA可以计算系统的可靠性和组件的重要性。FTA假设事件是双值的且独立的，所以它不能够解决复杂系统的建模问题。随着末事件和逻辑门电路个数的增加，计算精度会变低，并且该过程很耗时。对于像CTCS-3这样的复杂系统，随着结构和功能的复杂性的增加，系统故障过程呈现出复杂动态特性。同时，还有一些现有的如CCF和系统多态性现象等等。因此，在列车控制可靠性评估中，故障树分析的分析能力有限。

在可靠性领域，贝叶斯网络(BN)的不断演变和发展弥补了传统可靠性研究方法的局限性。BN表示节点之间的依赖关系图，简单易懂、容易双向推理。BN不仅降低了非根节点间的概率维度，同时也通过利用变量之间条件独立关系大大降低计算推理过程的复杂度。因此，它是适用于表达和分析不确定知识。从双向推理机制和状态描述来看，它不仅有FTA的优点，而且有处理CCF以及系统的多态性和不确定逻辑关系的能力。

本文中BN与传统的故障树分析相结合，......

Abstract

China Train Control System (CTCS-3) is a safety critical computer system that features large-scale, complex structure, and redundant configuration. CTCS-3 and as well as the equipment and technologies related to it can ensure the safety and reliability running of high- speed trains, and so the assessment on its reliability becomes very important. Generally, the two elements such as common cause factor (CCF) and failure mode polymorphism can not be ignored when assessing the reliability on a large complicated system. There are some limitations existing when applying the traditional fault tree analysis (FTA) is used to deal with the two factors mentioned above, and considering that Bayesian network (BN) possesses the abilities of bidirectional reasoning and the uncertain knowledge solving, and therefore BN is introduced to change the flaws. The method firstly establishes the FTA model of system from top to bottom, and then converts the FTA model into BN model from the lower to the upper, hierarchically. Eventually, the reliability indices are calculated using the BN with regard to the CCF and multi-state factors. Through integrating and evaluating sub-models using the approach with FTA combined with BN in the CTCS-3 system, some interesting results are acquired. The relative results show that the proposed approach is quite effective, and also provides a theoretical basis to improve the system reliability.

Key Words: CTCS-3, Fault tree, Bayesian networks, Reliability evaluation

1 Introduction

The reliability of high-speed train operation directly relates to the security of passenger lives and properties. There shall be very serious consequences when an accident occurs. Train control system is the core equipment and technologies to guarantee the safety, reliability and efficient operation of the high-speed train, and so its reliability and security assessment possesses very importance significance. China Train Control System (CTCS-3) is a safety- critical computer system in high speed train signaling system where a large number of complex electronic associated components and computer systems are applied. CTCS-3 possesses large complicated structure and redundancy configuration for the key equipment. Therefore, the requirements for reliability and security assessment on CTCS-3 is becoming higher and higher. 

In CTCS-3, the correlation is the common feature of its failures. Common cause failure (CCF) is a kind of dependent failure, and also is an important factor causing system internal failure, dependently. It is apt to generate greater error if common cause failure is ignored. Presently, the β factor model is often applied in CCF analysis. However it just is used in the 2-unit redundant system. In traditional reliability analysis, it is always assumed that the components or systems have only two states, such as work or failure, while it overlooks the system performance affected by dependent components and other failure states. There exists great difference between the established reliability analysis model and real situation [5]. Thus, Multi-state System (MSS) reliability analysis has received extensive attention.......