复杂交互作用下高铁列控系统危险致因辨识及风险耦合分析方法研究

The risk coupling problem caused by complex interaction of potential danger causation among systems has strong concealment, complexity and harmfulness, which has become a urgently critical problem in the safety analysis field of high-speed train control system. Based on the viewpoint of system theory and control theory，this project synthetically utilizes Systems-Theoretic Accident Model and Process(STAMP) and multi-agent theory to research the danger causation identification and risk coupling analysis method for train control system. Firstly, aiming at the complexity, hybrid and timing characteristic of train control system, the construction method of hierarchical control formalized model for operational scenarios is researched, which can accurately represent the complex interaction behavior and control logic of train control system under different operational scenarios. Secondly, the research of automatic fault injection algorithm is by injecting fault information into hierarchical control formalized model for operational scenarios, which can achieve the automatic search of system danger causation and risk coupling relationship. Thirdly, on the basis of all above, system risk coupling intelligent learning method based on multi-agent simulation and decision tree is researched, which can deeply dig system danger causation and risk coupling relationship, and then make a supplement or correct errors in the results of automatic search. Through the research of this project, it is helpful to reveal the complex danger causation mechanism of the train control system, which has important theory significance and application values to improve the operation safety of train control system.

系统间潜在的危险致因通过复杂交互作用导致的风险耦合问题，具有极强的隐蔽性、复杂性与危害性，已成为高铁列控系统安全分析领域亟待解决的关键问题。本项目从系统论与控制论的视角，综合利用基于系统理论的事故模型及过程（STAMP）与多智能体理论，研究列控系统危险致因辨识及风险耦合分析方法。首先，针对列控系统的复杂性、混成性与时序性特点，研究运营场景分层控制形式化模型的构建方法，准确表征不同运营场景下列控系统复杂的交互行为与控制逻辑；其次，研究故障自动注入算法，将故障信息注入到运营场景分层控制形式化模型中，实现系统危险致因及风险耦合关系的自动搜索；最后，基于以上基础，研究基于多智能体仿真与决策树的系统风险耦合智能学习方法，深入挖掘系统危险致因及风险耦合关系，对自动搜索结果存在的遗漏或错误进行补充与修正。通过本项目的研究，将有助于揭示列控系统复杂的危险致因机理，对提高列控系统运营安全具有重要理论意义。

系统间潜在的危险致因通过复杂交互作用导致的风险耦合问题，具有极强的隐蔽性、复杂性与危害性，是高铁列控系统安全分析领域亟待解决的关键问题。本项目从系统论与控制论的视角，综合利用基于系统理论的事故模型及过程（STAMP）与多智能体理论，研究提出了列控系统危险致因辨识及风险耦合分析方法。本项目主要研究内容具体包括：1）针对列控系统的复杂性、混成性与时序性特点，研究提出了运营场景分层控制形式化模型的构建方法，可以准确表征不同运营场景下列控系统复杂的交互行为与控制逻辑；2）研究设计了面向运营场景分层控制形式化模型的故障注入算法和风险耦合路径搜索算法，可以将危险致因信息注入到运营场景分层控制形式化模型中，实现系统危险致因事件间风险耦合路径的自动搜索；3）研究提出了基于多智能体仿真与机器学习的系统风险耦合智能学习方法，可以深入挖掘系统危险致因间的风险耦合规则。通过本项目的研究，为揭示列控系统复杂的危险致因机理及其风险耦合关系提供了重要的方法支撑，对提高列控系统运营安全具有重要理论意义。