A kind of railway vehicle fault detection method

Abstract

The invention discloses a railway vehicle fault detection method. According to the needs of driving safety, determine one or more components as the object to be detected for monitoring abnormalities. Construct a lightweight gradient lifting machine model to establish the importance of various driving parameters of the vehicle for the prediction of the temperature or deformation of a single object to be detected; determine the input time step of the features required by the predictor variable; combine the spatial position distribution of the components to be detected Analyze the relationship between the same vehicle parts with the Pearson coefficient, and predict the variables of multiple objects to be detected based on the concept of the same type of parts to be detected; build a prediction model, use the convolutional network to predict the time series signal, and obtain the real value and the predicted value The residual value; the obtained residual value is described in two-dimensional space according to the concept of similar components, and the abnormality detection of the residual is carried out by using the isolated forest method, so as to give an early warning of the abnormality of the components detected during the running of the rail vehicle.

Description

technical field [0001] The invention belongs to the field of rail vehicle abnormality detection and relates to a rail vehicle fault detection method. Background technique [0002] In recent years, my country's rail vehicles have continued to develop, and the safety and reliability of railways have continued to receive the attention of the entire industry. In the field of railways, Beijing Jiaotong University, Central South University, Southwest Jiaotong University, and CRRC Qingdao Sifang Research Institute have all conducted research on the safety of rail vehicles. [0003] The safety of the running part of the train is particularly important to the safe operation of the train, and its safety issues have attracted the attention of engineers. During the running of the train, the main failure forms of the running part are: the failure of the axle, bearing, and brake disc caused by the high temperature of the train components, the polygonal failure of the wheel caused by the ...

AI Technical Summary

Problems solved by technology

[0008] In view of the problems mentioned in the background technology, the low prediction accuracy caused by the weak data processing ability of the current model, the small application range, and the slow positioning and training caused by the prediction of a single target, the present invention proposes a lightweight The gradient hoisting machine conducts auxiliary analysis on the data collected by the sensor, selects the input time step of the vehicle history data required for the prediction data, and combines the spatial position distribution of each component to be detected and the Pearson coefficient to analyze the relationship between the corresponding components. A method for simultaneously predicting routine failures of a key component

Images