基于证据理论的城轨车辆走行系融合故障诊断

Approach for fault diagnosis of the metro vehicle running gears based on evidential theory

针对城轨车辆走行系故障诊断只依赖于单一证据源造成的诊断准确率低的问题,研究利用证据理论进行走行系多证据源融合的故障诊断。经过对高冲突低信任度情况下悖论产生的原因分析,采用矛盾因子对合成规则进行了改进;针对城轨交通车辆走行系,分析了轴承和轮对所有可能发生的故障,建立了走行系轴承和轮对的识别框架;选择轴箱振动诊断、轨旁振动诊断和轴温诊断等3个证据源作为走行系融合诊断的证据源;采用专家经验法确定了不同证据源的基本信任分配函数。研究结果表明,采用单一证据源的故障诊断准确率较低或诊断对象少,而采用多证据源融合的故障诊断方法(在包含轮对和轴承两个对象情况下)的诊断准确率为80%,从而有效提高了城轨车辆走行系故障诊断的可靠性。

Aiming at the problem that traditional methods for fault diagnosis of the metro vehicle running gears depend on single evidence and lead to low accuracy, a new fusion fault diagnosis approach was proposed based on evidential theory. The reason of paradox on the condition of high conflict and low confidence was analyzed, and the combination rule was then modified. All the possible faults of bearings and wheels were collected to construct the frame of discernment for running gears. The diagnosis results of axle vibration, wayside vibration and axle temperature are selected as evidences of fusion diagnosis. The belief function was identified using experts＇ experience. The results indicate that the fault diagnosis based on single evidence may lead to low precise or cover few objects, while the proposed fusion fault diagnosis approach based on multiple evidence sources can reach a high accuracy of 80% for bearings and wheels and improve the reliability of fault diagnosis effectively.