Flywheel fault identification method based on LSTM deep noise reduction auto-encoder

Abstract

The invention provides a flywheel fault identification method based on an LSTM deep noise reduction auto-encoder. The method comprises the steps: firstly, collecting a flywheel motor current signal, a flywheel output torque signal and a flywheel rotating speed signal, processing an original signal by using extreme value envelope moving average or wavelet noise reduction, and respectively constructing a current signal sample, a torque signal sample and a rotating speed signal sample in combination with a sliding window method; then, respectively constructing three independent deep noise reduction auto-encoder networks on the basis of an LSTM unit, and respectively training the three networks through the current signal sample, the torque signal sample and the rotating speed signal sample; next, generating a signal residual based on the network output, and extracting a signal residual feature; and finally, introducing a grid search method to determine an SVM optimal parameter, and utilizing the SVM to complete flywheel fault identification. During network training, only signals collected in a normal state are used, after training is completed, monitoring signals can be directly input into the model, and fault identification is completed. The technology of the invention can realize accurate identification of flywheel faults.

Description

technical field [0001] The invention belongs to satellite flywheel fault diagnosis technology, in particular to a flywheel fault identification method based on LSTM deep noise reduction autoencoder. Background technique [0002] A flywheel is a rotating rigid body that is driven by a motor to achieve different rotational speeds and generate different moment of momentum. Due to its long life, high precision and multi-purpose advantages, it has become one of the most common actuators in current high-precision attitude control satellites. The flywheel uses its own angular momentum to generate reaction torque, thereby changing the attitude of the satellite, which is very important for completing the specified attitude adjustment, resisting external disturbances and ensuring the stable operation of the satellite. [0003] However, the working environment of the flywheel is very harsh. It is disturbed by many factors such as space temperature difference, cosmic rays, electromagne...

AI Technical Summary

Problems solved by technology

In actual use, support vector machines have difficulties in dealing with multi-mode fault identification problems, and are sensitive to parameter selection, so they need to be used together with other methods; backpropagation neural networks have gradient explosion or disappearance during training, and at the same time The representation ability of complex data is limited; although the convolutional neural network can extract data features better, it requires the input data to be two-dimensional or even higher-dimensional, which limits the use of one-dimensional time series signals for satellite monitoring; the long-short-term memory network can store Learning historical information in time series, but poor robustness to nonlinear noise

Images