基于动态时间规整和证据理论的机车轴温监测预警研究

Research on monitoring and pre-warning of locomotive bearing temperature based on dynamic time warping and evidence theory

根据和谐型电力机车轴温状态监测的实际需求,针对监测数据异常值所导致的误报,以及阈值报警发生时,故障往往处于后期,存在现场处置时间不足的问题。采用四分位算法进行预处理消除误报,在绝对-相对轴温故障判别模型的基础上,运用动态时间规整,改进走行部关联测点的温差计算方式。将证据理论应用于轴温故障诊断,提出基于模糊集合论的基本概率分配函数,得出故障信度值为现场预警决策提供依据。实例诊断表明该方法能从大量轴温数据中查找出故障隐患车轴,提前预警处理,保障行车安全。

According to the actual demand about the bearing temperature monitoring of Harmonious electric locomotives,the purpose is settling the high misdiagnosis caused by abnormal monitoring data and a problem of insufficient disposal time to solve the fault which is in the late stage on the spot,when the threshold alarm occurs.This paper used the quartile algorithm as the pre-process to eliminate misdiagnosis.On the basis of absolute-relative bearing temperature fault discrimination model,using dynamic time warping,the calculation method of temperature difference of correlational measuring points was improved in the running part structure.The evidence theory was applied to shaft temperature fault diagnosis.The basic probability assignment function based on fuzzy set theory was put forward,and the fault trust value was obtained to provide the basis for field pre-warning decision.The diagnosis example shows that this method can find out the hidden fault bearing from a large number of axle temperature data,advance the pre-warning processing,and ensure the driving safety.