Study On Fault Diagnosis Method Of Rotor System Based On Recurrent Neural Network

Rotating machinery is the main equipment in industrial production.Rolling bearings are one of the important components of rotating machinery.Fault diagnosis is a prerequisite for ensuring safe production,improving enterprise efficiency and avoiding economic and property losses.The working environment of rolling bearings is harsh and changeable,and the background noise and variable load make the fault information submerged in the interference signal,making it difficult to extract fault features.At present,the data-driven convolutional neural network(CNN)method can realize autonomous learning.The characteristics of local connection and weight sharing can reduce the amount of calculation and increase the feature extraction ability of vibration signals.It has broad prospects in the field of fault diagnosis.In this paper,aiming at the problem that the early fault characteristic signals of rolling bearings are mixed in the noise signals,and the fault features are difficult to extract,an improved variational mode decomposition combined with cross wavelet transform(IVMDXWT)noise reduction method is proposed to preprocess the vibration signals of rolling bearings;The multi-channel improved convolutional neural network(MC-ICNN)fault diagnosis model fully extracts fault diagnosis features and improves the accuracy of fault diagnosis;combines the attention mechanism to build a multi-channel improved convolutional neural network based on the attention mechanism(ATT-MC-ICNN)fault diagnosis model to solve the feature extraction ability of the model under complex working conditions.Based on this,the fault diagnosis and detection system software is developed.The main research content of the paper is as follows:(1)Aiming at the problem that the bearing fault features are weak and difficult to extract,a signal processing method of improving the joint noise reduction of variational mode decomposition and cross wavelet transform is proposed.Through the spectrum envelope method,the useless frequency components in the vibration signal are screened out,and the effective frequency components in the vibration signal are retained to determine the optimal mode number K.The original vibration signal of the rolling bearing is subjected to cross wavelet transform to obtain the original wavelet coherence spectrum;the component decomposed by IVMD and the original vibration signal with a large kurtosis value are respectively subjected to cross wavelet to obtain the component wavelet coherence spectrum,and the wavelet coherence spectrum of the original signal is combined with Comparing component wavelet coherent spectrograms,selecting the most relevant frequency bands,and removing noise frequency bands,so as to realize signal reconstruction and achieve the effect of noise reduction.(2)Aiming at the problem that the single-channel input cannot fully extract fault features and the convolutional neural network has insufficient ability to extract time-series features of vibration signals,a multi-channel improved convolutional neural network(MC-ICNN)rolling bearing fault diagnosis method is proposed.Reconstruct the signal through the IVMD-XWT method to obtain multi-channel signals,fully extract the fault feature information,introduce the Bi GRU layer to extract the time series related information of the vibration signal,and determine the optimal network structure,learning rate and the number of convolution kernels of the model according to the experiment,build the MC-ICNN model.(3)Aiming at the problems of strong noise and variable speed in actual engineering,a fault diagnosis method combining attention mechanism and improved convolutional neural network(ATT-MC-ICNN)is proposed,and the ATT-MC-ICNN model is constructed.The attention mechanism is introduced into the improved convolutional neural network to improve the feature extraction ability of the model under strong noise and variable speed.(4)An end-to-end fault diagnosis system is developed based on the integration of the fault diagnosis model proposed in this paper based on the Python programming language and the Py Qt5 framework.