Reconstruction Of Missing Measurement Data Of High-speed Train Based On Generative Adversarial Network

The data-based operation and maintenance system is an important guarantee for the safe operation of high-speed trains.Only by ensuring the authenticity and integrity of the data and enabling the operation and maintenance system to achieve real-time and accurate monitoring of the train,can the safe operation of the train be guaranteed.However,when high-speed trains run in special conditions such as tunnels and mountainous areas,it often has a certain impact on the high-speed train data collection system,resulting in problems such as difficulty in obtaining some measurement data or missing key data.Aiming at the situation that the train data collection system is faced with small samples and missing measurement data under special working conditions,and the existing models are difficult to effectively solve the problem of such small sample data reconstruction,this study develops a high-speed train based on generative adversarial network.Missing Measurement Data Reconstruction Study.The specific research contents are as follows:(1)In view of the problem of missing measurement data under special working conditions,a method for reconstruction of missing measurement data based on generative adversarial network is proposed,and appropriate parameters are set for the framework.First,the discrete measurement data is used as input to preprocess the data dimension upscaling;secondly,the convolutional neural network is used to learn the correlation between different eigenvalues of each device in the form of unsupervised training;finally,the authenticity context similarity constraint is used to improve the Model reconstruction accuracy.Experiments show that the model in this study can still maintain a high reconstruction accuracy in the case of missing measurement data in different degrees,and the reconstructed data can also well conform to the distribution law of measurement data.(2)Aiming at the situation of small samples and missing of measurement operation and maintenance data under special working conditions,a transfer learning generative adversarial network data reconstruction method for small sample data is proposed.In this method,a new variable A split-autoencoder-generative adversarial semantic fusion network(VAE-FGAN)is used to reconstruct missing data.First,the GRU module is introduced into the encoder to fuse the low-level features and high-level features of the data,so that VAE-FGAN can learn the correlation between measurement data in the form of unsupervised training;secondly,the SE-NET attention mechanism is introduced into the entire generation network.In order to improve the expression of data features by the feature extraction network;finally,parameter sharing is achieved through transfer learning and pre-training,which solves the problem that it is difficult to train the model in this study due to the small amount of operation and maintenance data in some parts.The experimental results show that in the case of missing small sample data,the reconstruction accuracy MAE and MAPE indexes can be kept below 1.5,and the reconstructed data can also well conform to the distribution law of the measurement data.