Development Of A Fault Diagnosis And Life Prediction System For Rolling Bearing Based On Graph Convolutional Networks

Rolling bearings are the core components of large-scale mechanical equipment,and the equipment paralysis caused by the damage to rolling bearing accounts for a large proportion.It is of great significance to judge the running status of the rolling bearing in time and make corresponding results analysis.Given the complex working conditions of rolling bearing and the difficulty in extracting the fault features of vibration signals,this paper adopts a feature extraction method based on visual graph signals to diagnose the faults of rolling bearing,and effectively deal with the non-stationary and nonlinear characteristics of rolling bearing signals.Using the graph convolutional neural network model to predict the life of the rolling bearing,reflecting the real-life value of the rolling bearing,and verifying the superiority of the algorithm proposed in this paper through the relevant data sets.Finally,adding graph convolutional neural network and other algorithms to the software system verifies the feasibility of the graph convolutional neural network algorithm in the system,and through experiments to verify the effectiveness of adding graph convolutional neural network and other algorithms to the system.(1)Fault diagnosis based on feature extraction of visible graph signals.The one-dimensional vibration signal of the rolling bearing is converted into a visual map signal,and various map indicators are obtained by calculating the adjacency matrix and Laplace matrix of each visual map signal,and the two-sample Z value method is used to screen out the appropriate fault characteristics as The fault feature vector is obtained,and the bearing fault diagnosis result is obtained through the support vector machine classification algorithm.The experimental analysis shows that,compared with the traditional time-domain feature extraction method,for different types of rolling bearing fault diagnosis,the correct rate of the fault feature extraction method based on the visual map signal is increased by 8.34%;for the fault diagnosis of the rolling bearing outer ring,The correct rate of fault feature extraction method based on visual map signal is increased by 16.67%.It further shows the superiority of the method based on visual map signal feature extraction.(2)Prediction of rolling bearing life based on graph convolutional neural network.Convert the time domain signal of the rolling bearing into the frequency domain signal,extract the relevant features in the frequency domain,filter out the appropriate frequency domain features as the fault feature vector of this time and input it into the graph convolutional neural network model,and predict the remaining life of the rolling bearing.Finally,by comparing the RNN algorithm and the MLP algorithm through experiments,the mean square error,mean absolute error,and root mean square error of each model are calculated respectively.The error comparison analysis shows that the graph convolution network model is used in this paper to predict the life of rolling bearings which has a good effect.(3)Development of a fault diagnosis and life prediction system for rolling bearing based on graph convolutional neural network.Aiming at the possible difficulties in the system,the system data acquisition function,the system data transmission function,and the feasibility of the graph convolutional neural network algorithm in the system are analyzed in detail.At the same time to achieve the expected effect of the system,the overall framework of the system and the cloud database are designed.Finally,the data and related algorithms used in the third and fourth chapters of this paper are added to the system to verify the feasibility of the graph convolutional network model and other related algorithms in the system,and on this basis,the relevant experimental platform is built to verify the effectiveness of data transmission and acquisition in the practical application.