Research On Real-time Fault Warning Of High-speed Train Bearing Based On Axle Temperature Characteristics

At present,the scale of China's high-speed railway network continues to expand,the number of high-speed trains continues to grow,and research on train equipment fault diagnosis and health management has become increasingly important.High-speed train bearing often works with heavy loads,poor environment,and hot-axle is one of the most commonly found problem during the running process.Since the anomaly of axle bearing temperature is closely related to the occurrence of bearing faults,it has become an important research direction for axle bearing fault diagnosis and prognostics of highspeed train by monitoring axle temperature in real time.However,the operation mode of the equipment is greatly affected by the working conditions and the fault has nonlinear characteristics.Traditional fault prediction method is difficult to meet the actual requirements,and the research on bearing fault early warning of high-speed train is not much.Therefore,this paper carried out the research on the fault prediction,performance degradation evaluation and early warning strategy of high-speed train bearing based on the temperature characteristics.The main contents include:(1)The current research states and progress are outlined,and the bottleneck problem of high-speed train bearing fault diagnosis is summarized.The author introduced the existing axle temperature monitoring system for high-speed trains,analysis the macroscopic characteristics of axle temperature monitoring data and use statistical methods to figure out 17 features that have a greater impact on the axle temperature as the data basis for subsequent research.(2)The traditional fault prediction methods are almost completed in an offline environment.On the one hand,the prediction accuracy is difficult to break through,on the other hand,the model cannot adapt to new data changes.To solve these problems,an online learning method based on OR-ELM for bearing fault prediction of high-speed trains is proposed.The onboard computing platform is used to collect streaming sensor data in real time,which has been avoiding the complicated transmission process and reducing the loss of data quality,and a streaming data preprocessing part is designed to further improve the data quality.The real-time update of model parameters is completed online with monitoring data,which ensures that the model can adapt to new data changes rapidly.Experimental results show that the method has excellent performance in prediction accuracy and dealing with streaming data mutation.(3)A high-speed train bearing fault prediction method based on performance degradation assessment is proposed.A deep neural network was trained with health equipment monitoring data,and then calculate the error matrix which is obtained by the prediction result and the actual data.The performance degradation evaluation indicators can be calculated with error matrix,and it directly reflects the degradation degree of equipment performance.Experimental results show that this method has a large degree of separation in the performance degradation evaluation indicator calculated on the data of healthy bearings and faulty bearings,and can achieve early prediction of faults.(4)According to the differences between the two fault prediction methods,the corresponding fault warning strategies are proposed.According to the fault types,reliability and severity of the prediction results,the fault warning strategies of different types and different safety levels are designed,which can provide accurate and simple warning information to train operators in time and efficiently.The experimental results show that the two strategies can trigger the hierarchical warning message accurately before the bearing failure.