Data-Based Axle Temperature Prediction of High Speed Train by Multiple Regression Analysis

Abstract

Axle is a key equipment of high speed train, and affects the safety of train operation. The fault of axle is commonly detected by comparing the current axle temperature with a fixed temperature threshold. Owing to the complex mechanism of axle temperature rising, the axle temperature has a wide variation range even working in normal condition. Therefore, the shortcomings of the method with a fixed temperature threshold are obvious: a high threshold may leads to missing alarm, existing potential safety risk, on the contrary, a low one can ensure the safety, but easily leads to failing alarm, causing unnecessary troubleshooting and maintenance. In view of the problem, a model for the calculation of dynamical temperature threshold is proposed in this paper by relational analysis of monitoring data. Specifically, after analyzing the characteristic of axle temperature changing, the temperature prediction process is divided into three stages according to its running modes, i.e. acceleration, stable running and deceleration. Then the temperature prediction model is established and validated, and the results denoted the effectiveness and practicability.