基于流数据的列车在线状态监测方法研究与应用

Research and Application of an Online Train Condition Monitoring Method Based on Streaming Data

为实时地监测列车在线状态,传统的阈值检测方法只能从数值表现上对设备或系统状态做出判断,而忽略了数据趋势性变化所反映出的列车健康状态变化信息。文章针对列车在线运行过程中产生的流数据提出了一种基于一致性表现的流数据分析方法,其在离群异常点提供报警的同时,能够从特征变化趋势中挖掘信息,同时开发了一套应用于列车多功能车辆总线MVB网络的智能维护设备来搭载和验证算法;基于列车实测数据和故障注入技术在实验室模拟了列车总线网络测试环境,以列车气制动系统为分析对象对所提方法和所开发的设备进行了测试。结果表明,其能够有效地对突发性异常进行报警,在健康状态早期、中期和后期的报警识别率分别为85.7%, 71.4%和57.1%,且能够较好地识别和监测设备性能的衰退。

In order to detect the online status of trains in real time,traditional threshold detection methods can only make judgments on the status of equipment or systems based on numerical performance,while ignoring the train health status changes reflected in the trend of data.This paper proposes a consistent performance-based streaming data analysis method for the streaming data generated in the online train operation process.It provides alarms for outliers and abnormal points,and can extract information from the characteristic change trends.At the same time,an intelligent maintenance equipment which applied to the train multifunction vehicle bus MVB network is designed to carry and verify the algorithm.After that,the train bus network test environment is simulated in the laboratory based on the train measured data and fault injection technology,and the train air brake system is used as the analysis object.Tested with the equipment,the results show that it can effectively alarm sudden abnormalities.In the early,middle and late stages of the health state,the alarm recognition rate is 85.7%,71.4%and 57.1%,and it can better identify and detect equipment performance degradation.