Research On Fault Diagnosis And Early Warning Method For High-speed Train Axle Box Bearing Based On Data-driven

In the past ten years,China’s high-speed railways have developed rapidly,and it is particularly important to ensure the safety and stability of trains during operation.As a key component of the running gear of high-speed trains,axle box bearings are directly related to the safety and stability of train operation.At present,high-speed trains in China mainly monitor the running status of axle box bearings through the on-board axle temperature detection system,and alarm based on a fixed axle temperature threshold.However,this system has the problem of high false alarm and underreport rate,and the large amount of data collected during the operation has not been fully utilized.In response to the above problems,this thesis analyzes the temperature rise mechanism of high-speed train axlebox bearings.Based on the data of axle temperature,ambient temperature and speed collected during the operation of high-speed trains,through data processing and statistical analysis,the axles are established using different artificial neural networks.The diagnosis model for abnormal temperature rise of box bearings provides an early warning method for train axle box bearing faults.Based on this,a secondary early warning mechanism for high-speed train axle box bearing faults is proposed.The main works are as follows:(1)The heat production and heat dissipation mechanism of the axle box bearing of the high-speed train is analyzed.Combined with the structure and heat transfer mechanism of the axle box bearing,the heat transfer model of the train axle box is simplified.Based on the simplified model,the temperature and temperature rise rate of the axle box are analyzed by analytical solution which provide a theoretical basis for the establishment of subsequent fault diagnosis models and the advancement of early warning methods.(2)In view of the low data quality and the large amount of data,a data preprocessing plan was developed.The missing values in the data were processed through interpolation and resampling and the different data variables were integrated and aligned;in view of the characteristics of trains starting and stopping at different stations,A time window segmentation algorithm suitable for the data in this paper is formulated;in order to further improve the quality of the data,the high-frequency noise in the data is removed by lowpass filtering in the frequency domain;based on the characteristics of the data in this paper as a discrete time series,the temperature rise rate of the bearing is calculated by numerical differentiation;finally,the distribution of the fault data and the normal data is compared and analyzed,and the correlation between the different data variables is analyzed quantitatively and qualitatively,and the effect of different data variables on the temperature and temperature rise rate of the axle box bearing is obtained.These works have laid the foundation for the establishment of fault diagnosis models and the advancement of early warning methods.(3)A diagnosis model for abnormal temperature rise of axle box bearings within a time window was built and a fault warning method based on the axle temperature trend in the equilibrium state was proposed.First,for the serious imbalance of positive and negative samples in the data,a generative adversarial network was built and trained to enhance the imbalanced data set;then real normal samples and generated fault samples were used as training data to build and train multivariate long-term and short-term memory fully convolution neural network is used as a fault diagnosis model to identify the faults of the axlebox bearing in different time windows;the fault diagnosis model for the axlebox bearing within the time window can only identify whether the bearing has abnormal temperature rise in different time windows,and cannot be updated.It reflects the dynamic evolution process of the fault degree of the faulty bearing over time in a long time,and a method is proposed to reflect the dynamic change of the bearing fault degree through the shaft temperature trend under the equilibrium state.(4)A two-stage early warning strategy for axle box bearing failures of high-speed trains is proposed.Based on the diagnosis model of abnormal temperature rise of axle box bearing in time window and the early warning method of axle temperature trend in equilibrium state,a secondary early warning strategy for high-speed train axle box bearings was developed and an early warning mechanism was developed.Finally,an example was used to verify the proposed secondary early warning.The strategy can identify train axle box bearing failures in advance and avoid false alarms.