Research And Application Of Life Prediction Model Of Railway Freight Car Brake Shoe Based On Transfer Learning

With the rapid development of China’s railway,information technology has become an important support to ensure the safe and efficient operation of railway transportation.In order to ensure the safe operation of railway freight cars,it is necessary to repair the key parts of railway freight cars in time.As the tread braking based on brake shoe is the main braking mode of railway freight cars,it is very important to replace the failed brake shoe in time to ensure the safe operation of railway freight cars.At present,the railway department mainly carries out daily inspection and regular maintenance for railway freight cars,but this maintenance method has some problems,such as short maintenance cycle,high maintenance cost,excess maintenance,etc.With the development of sensor technology,a large number of monitoring data can be obtained to provide the basis for data-driven life prediction.According to the predicted remaining life of the brake shoe,the condition based maintenance of the brake shoe can be realized,which greatly improves the maintenance efficiency and reduces the maintenance cost.Due to the complex driving environment of railway vehicles,the brake shoe data obtained from them are generally contaminated by noise,and the working conditions are complex and changeable.The current machine learning model is difficult to be directly applied to the actual industrial production environment under variable working conditions.In this thesis,a deep convolution neural network residual life prediction model based on transfer learning is proposed.On this basis,according to the characteristics and prediction requirements of railway freight car brake shoes,a freight car brake shoe residual life prediction model is proposed.Finally,the proposed prediction model is applied to the railway condition based maintenance system to provide support for the maintenance decision of the key parts of railway freight cars.The main research contents of this thesis include:(1)Aiming at the problem that the traditional residual life prediction method can only predict the life of parts under one working condition,and the data feature extraction effect is not good under complex working conditions,a depth migration convolution neural network(DTCNN)prediction model is proposed.By using the multi-layer superposition convolution layer in the deep convolution neural network model to better learn the high-level representation of each original feature,the transfer learning is introduced to apply the data under different working conditions from the target domain data set to the life prediction modeling of the target data set,so as to realize the life prediction of parts under the condition of few samples.At the same time,the open data set C-MAPSS is used to verify the effectiveness of the model.(2)Aiming at the problems of missing data and complex working conditions in the key parts of railway freight car brake shoe,a prediction method for the remaining service life of railway freight car brake shoe is proposed based on the proposed DTCNN model.Firstly,the missing brake shoe data is completed by using the generated counter interpolation network(GAIN).Then,combined with the characteristics of the brake shoe and the prediction requirements,the service life of the brake shoe is predicted by using the depth migration convolution neural network(DTCNN).The relevant experiments are completed by using the brake shoe data under actual working conditions,which verifies the accuracy of this method in the service life prediction under variable working conditions.(3)The design and implementation of brake shoe remaining service life prediction system.Using the proposed method for predicting the remaining service life of railway freight car brake shoe,a prediction system for the remaining service life of brake shoe is designed and developed,and put into test use,which provides technical support for the maintenance decision of railway freight car brake shoe,so as to improve the operation efficiency of freight car.