Research On Fault Diagnosis And Remaining Life Prediction Method Of Rolling Bearing Based On Deep Learning

Rolling bearings are the basic equipment of rotating machinery,and their running state will affect the operation of the mechanical system.The management of the health status of the rolling bearing,the realization of the fault diagnosis of the rolling bearing and the prediction of the remaining life,are the guarantee for the healthy operation of the entire mechanical system.For the prediction of the remaining life of rolling bearings,the traditionally used prediction methods process the characteristics of a single domain of the signal,and the modeling combined with a single machine learning algorithm has great limitations.For the fault diagnosis of rolling bearings,traditional machine learning methods are difficult to achieve a stable and high accuracy rate.In view of the limitations of rolling bearing life prediction and fault diagnosis modeling,the main research work of this paper is as follows:1.In view of the limitations of traditional machine learning algorithm modeling for bearing life prediction,this article proposes two solutions: Option 1,the entire degenerate failure process of the bearing is divided into three stages to study,respectively,the stable degradation period and rapid degradation period and fast failure period,through the division of stages,the characteristics of the bearing in each stage can be more prominent;the second plan is to improve the particle swarm algorithm from three aspects,and establish an improved particle swarm optimization support vector machine bearing life prediction model.In this study,the vibration signal was denoised by wavelet packet,and then the time domain and frequency domain features were extracted and the principal component analysis method was used to reduce the dimension.Then,through the traditional particle swarm and improved particle swarm optimization support vector machine bearing life prediction model,The output comparison analysis in the three divided stages can show the effectiveness of the solution proposed in this paper for bearing life prediction.2.Aiming at the limitation of traditional machine learning algorithm modeling for bearing fault diagnosis,this paper proposes a bearing fault diagnosis model based on Bayesian optimization convolutional neural network.The vibration signal is converted into a time-frequency graph,and the accuracy of the three models is through the over-fitting model of the convolutional neural network,the slow-fitting model of the convolutional neural network,and the establishment of the hyperparameter model of the Bayesian optimization convolutional neural network.The output comparison analysis of loss function,etc.can show the superiority of the bearing fault diagnosis model based on Bayesian optimization convolutional neural network proposed in this paper,and the fault diagnosis accuracy rate is obviously improved,the accuracy rate reaches 99.84%.