Data Driven Multi-Modal Operation Monitoring And Fault Diagnosis Of Train Bearings

In 2018,the passenger and freight volume of China’s railway transportation reached 3.375 billion persons and 4.026 billion tons respectively.The trains has many advantages in transportation,such as efficiency and fastness,large transportation quantity,high comfort level and low energy consumption.These advantages have led to the rapid development of high-speed trains in recent years.At the same time,the safety problems of train operation have also been widely concerned.Bearings and axleboxes are key components in train operation,and their operation condition monitoring and fault diagnosis become even more important.Due to the harsh environment in which the bearing are located in the train,long-term feiction can easily cause wear,cracks and peeling and can lead to shaft breakage when it is serious.These faults cause great hidden danger to the safe operation of the train.Therefor,it is necessary to monitor and diagnose the bearing and axle box of the train,and the bearing temperature is the direct index of the performance change of the bearing and axle box.Therefore,the research on axle temperature monitoring and fault diagnosis is of great significance.This research project relies on the maj or project of NSFC,real-time diagnosis theory and prediction method for micro and complex fault diagnosis of high-speed train information control system,and cooperates with Tsinghua University,Central South University and Zhuzhou Institute of Hunan Province.On the basis of summarizing the existing research methods of bearing fault diagnosis and the research status at home and abroad,it is difficult to adopt the model-based fault diagnosis method in view of the complicated bearing temperature mechanism,and the existing false diagnosis method based on single mode data driven fault diagnosis is relatively high.Combined with the space-time characteristics of bearing temperature data in train operation,the data driven multi-mode operation monitoring and control of train bearings is studied.The main innovations of fault diagnosis methods are as follows:1.In view of the lack of data and outliers in the real-time collected bearing temperature data,a preprocessing method of train axle temperature data combining linear interpolation and dynamic principal component search is proposed.First of all,in view of the fact that the axle temperature data collected in the axle temperature monitoring system has the phenomenon of data missing,a preprocessing method of axle temperature data based on linear interpolation is proposed;secondly,in view of the inaccuracy of modeling caused by the outlier data of bearing temperature data,a preprocessing method of train axle temperature data based on dynamic principal component search is proposed to process the missing data.2.The traditional single mode modeling and monitoring methods do not take into account the dynamic characteristics of bearings in different operating intervals,and propose a multimodal operation monitoring method for train bearings.According to the clustering results of the similarity of bearing temperature changes in different operation regions,different operation modes of the same train are identified,and combined with the correlation between operation regions extracted by fault alarm geographic information,the operation mode identification method based on train operation track is proposed.According to this method,the bearing variables of the train are divided into multiple modes and the divided modes are respectively modeled by the multi-modal dynamic inner principle component analysis(m-dipca).On this basis,the bearing multi-modal modeling method based on m-dipca and the abnormal bearing variable location method based on contribution graph are proposed.3.According to the different fault types of train bearings and the waveform similarity of the same type of fault,a fault diagnosis method based on dynamic time wrapping(DTW)is proposed.Firstly,the axle temperature data of abnormal bearing variables are extracted as test samples,which are compared with the reference samples in the failure mode database to find out the dynamic time warping distance;secondly,the bearing failure type is determined by comparing the dynamic time warping distance between the test samples and the reference samples.The simulation data is used to study the method,and the actual train operating axle temperature data is used to verify the experiment.The experimental results show that the proposed method can monitor the running state of the train and diagnose the causes of bearing faults,reduce the false alarm rate,and the proposed method is effective,which can ensure the safe and efficient operation of the train.