Research On Bearing Temperature Anomaly Detection Of High-speed Train Based On Spatio-temporal Comparison

Bearings are the key components that affect the safety of high-speed trains.The on-board monitoring system uses a temperature sensor to perform real-time monitoring of the bearing status and alarms based on a set temperature threshold,which effectively prevents major safety accidents.However,this strategy also has a disadvantage: once the alarm is triggered,it is necessary to take measures to reduce speed or stop immediately.It cannot provide sufficient time margin for handling the accident,seriously disrupts the train operation order,and easily causes economic loss and adverse social impact.Therefore,improving the intelligent diagnosis level and early warning capability of high-speed train bearings has important research and engineering value.Due to the complicated operating conditions of high-speed train bearings,and at the same time affected by the coupling of materials,manufacturing,and maintenance,the temperature characteristics are complicated and changeable.Researchers have proposed a series of diagnostic methods based on empirical rules and cluster analysis models,but the high false alarm rate and insufficient early warning capabilities have always been two important issues that have plagued practical applications.Therefore,this paper conducts research on bearings temperature anomaly detection of high-speed train based on spatio-temporal comparison.The main research work is as follows:(1)By comparing and analyzing the temperature characteristics of similar bearings with different spatial distributions,this paper proposes a bearing anomaly temperature detection model based on AHP-entropy method decision optimization.Firstly,according to the consistency of the temperature rise trend of similar bearings and the deviation phenomenon of anomaly bearing temperature,the K-means clustering is used to locate bearing with anomaly temperature to obtain the first-level anomaly detection result;Then,by analyzing the characteristics of anomaly temperature,combining expert experience and data characteristics,a decision model based on the optimization of AHP-entropy method is proposed to re-diagnose the first-level anomaly detection results,and obtain secondary detection results.The example verification results show that the model significantly reduces the false alarm rate,but the warning time margin is less.(2)By comparing and analyzing the temperature characteristics of the same bearing with different time distribution,a high-speed train bearing temperature anomaly detection model based on Bi LSTM real-time prediction is proposed.Firstly,the sensitive parameters are extracted from a large amount of vehicle monitoring history data through the random forest algorithm,and then the Bi LSTM is used to establish the mapping relationship between the sensitive parameters and the bearing temperature to form a real-time prediction model of the high-speed train bearing temperature.Then,the residual value analysis is performed on the predicted value and the actual value,and the determination rule of the abnormal temperature of the bearing is formulated with reference to the SPC standard.The example verification results show that the model significantly improves the early warning capability,but has a certain false alarm rate.(3)A high-speed train bearing temperature anomaly detection model based on spatio-temporal fusion decision is proposed.Firstly,the D-S evidence theory is used to combine the anomaly detection model based on the AHP-entropy method decision optimization the decision and the anomaly detection model based on Bi LSTM real-time prediction.Through comprehensive diagnosis and decision-making from the two dimensions of space and time at the same time,more comprehensive and accurate anomaly detection is achieved.The example verification results show that the spatio-temporal fusion decision model successfully eliminates the misjudgment in the diagnosis of the single-dimensional model,and has the good capability of early warning.(4)A high-speed train bearing temperature anomaly detection system was developed.Based on the high-speed train bearing temperature abnormality detection model based on the fusion decision of spatio-temporal comparison,the related functional requirements are analyzed,and the tkinter compilation library in the Python GUI module is used to complete the system interface development.Application examples show that the system can meet the expected functional requirements.