High Speed Train Axle Temperature Prediction Based on Support Vector Regression

Abstract

The axle temperature of the high-speed train is the most direct reflection of the train operating conditions. In the progress of train operation, the axle temperature annunciator will raise the alarm when the temperature is abnormal. Since the axle temperature annunciator may be false negative and false positive, the prediction of the axle temperature is proposed to provide theoretical support for the work of the axle temperature annunciator. The change in axle temperature is a complex process and is influenced by excessive factors, the combination of support vector regression model and mutual information technique is applied. The feature selection and dimensionality reduction of the characteristic items influencing the axle temperature change are realized by mutual information, and then the support vector regression model based on radial basis function and polynomial kernel function is established. Through the comparison and analysis of the prediction results of the axle temperature, the feasibility of the method in the prediction of the axle temperature is verified. The support vector regression model based on the radial basis function is reduced by about 0.02 compared with the support vector regression model based on the polynomial kernel function.