针对重载组合列车ECP制动系统——这一典型复杂、具有邻接耦合特性的互联分布式系统，建立基于邻接节点信息的具有自适应学习机制的分布式故障诊断体系结构，实现对重载列车空气管路漏泄的故障诊断。首先，提出一种不确定性建模方法，利用互联参数来描述邻接耦合特性，将参数不确定性转化为系统的结构化未知扰动，结合ECP制动系统的编组特性，建立列车管路充排风过程数学模型；其次，针对系统模型参数和互联参数不确定性的变化特点，构建一类基于自适应神经网络的参数估计器，设计自适应学习律对邻接节点信息进行状态估计，同时实现未知参数的在线辨识；最后，在故障函数估计器中引入基于极值搜索的故障函数在线辨识算法，通过研究参数可调的闭环搜索框架和反馈激励函数，保证其过程稳定性、收敛域和收敛速度，实现故障的检测、隔离和辨识。通过在ECP制动系统的应用，为重载列车安全运行提供一种有效方法，对故障诊断技术的发展提供一定的理论基础。

This research aim at the ECP (Electronically Controlled Pneumatic) brake system of heavy-haul freight train, which is a typically complex distributed system with adjacent coupling interconnection. A distributed fault diagnostic architecture which is based on the adjacent information and adaptive learning mechanisms is established to diagnose the leakage fault in brake pipes of freight train. Firstly, a modeling method with uncertainty parameters is presented and the adjacent coupling characteristics are described by interconnection parameters, uncertain parameters are converted to structured unknown disturbances, and with assemble features of ECP brake system, the charging and exhausting processes for freight train brake pipe are built then. Secondly, a class of adaptive neural parameter estimators is constructed based on parameters uncertainty characteristics of system model and interconnection; an adaptive learning law is designed to realize adjacent node estimation and to identify unknown system parameters. Thirdly, an online parameter approximation algorithm based on extremum seeking is lead into the fault function estimator to realize the FDII (fault detection, isolation and identification), and the process stability, convergence domain, convergence speed are ensured by feedback activation function and closed-loop searching framework with adjustable parameters. Finally, application and experimental verification will be discussed on interconnetive ECP brake system, and the security level of heavy-haul freight train will be proved then. This research will also provide theoretical and practical foundation to promote the development of distributed fault diagnostic techniques.