An lstm method for fault detection of high-speed railway steering system based on generative confrontation network

Abstract

The invention discloses a high-speed rail steering system fault detection LSTM method based on a generative confrontation network, which belongs to the technical field of fault diagnosis. Firstly, a signal sensor is mounted on the train; under multiple normal conditions and multiple fault conditions, vibrations are respectively collected After the signal is synthesized and the vector is labeled as; then all the labeled vectors are merged into the data set, and the training set and the test set are divided; in the training set, n fault vectors and m normal vectors are selected, and the fault vector is generated by using the generated confrontation network. Oversampling to get m‑n new fault state vectors. Use the new fault state vector, n fault vectors and m normal vectors, a total of 2m data to train the LSTM network fault detection model, and perform fault detection on the test set; use the test results to calculate the index G-mean for verification; this The invention introduces a generative confrontation network for oversampling, which reduces detection errors caused by data imbalance.

Description

technical field [0001] The invention relates to fault diagnosis of a high-speed rail train steering system, and belongs to the technical field of fault diagnosis, in particular to an LSTM method for fault detection of a high-speed rail steering system based on a generative confrontation network. Background technique [0002] In recent years, my country's high-speed railway construction scale and innovations in high-speed railways have made significant progress. The rapid growth of high-speed railway operating mileage and the rapid improvement of operating timing have made high-speed train safety technology face great challenges. How to improve the safety of high-speed railways and the comfort of passengers has become one of the important research directions in the field of high-speed railways. [0003] As an important part of the car body, the bogie’s abnormal state will be reflected in the abnormal vibration of the bogie and car body. The vibration signals are all time-sequ...

AI Technical Summary

Problems solved by technology

[0003] As an important part of the car body, the bogie’s abnormal state will be reflected in the abnormal vibration of the bogie and the car body. The vibration signals are all sequential signals. The traditional fault diagnosis methods are based on static signals, and the fault diagnosis based on vibration signals Most methods are not directly applicable for diagnosis

Therefore, a high-speed train bogie fault diagnosis method based on LSTM (long-term short-term memory network) came into being. Considering that when training the LSTM fault diagnosis model, the bogie fault data is much smaller than the normal data, resulting in data imbalance, which will affect the LSTM fault diagnosis. Model accuracy and fault detection effect, so it is necessary to use the generative confrontation network to generate fault data, reduce the imbalance of training data, and improve the accuracy of the fault diagnosis model

Images