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

As the core equipment in industrial equipment,the working state of rolling bearing directly affects the running state and performance of the entire mechanical system,which in turn affects the safety and economic level of industrial production.Therefore,it is necessary to ensure the normal operation of mechanical equipment through the research on fault diagnosis and remaining useful life prediction technology of rolling bearings.However,how to accurately diagnose multi-position and multi-type faults of rolling bearings and how to directly construct HI from the original vibration signal to maintain the integrity of the signal are hot issues in the field of fault diagnosis and remaining useful life prediction of rolling bearings today.In view of the above problems,this thesis takes rolling bearings as the research object,and uses deep learning technology to study the fault diagnosis and remaining useful life prediction technology of rolling bearings.The main contents are as follows:1.A multi-fault diagnosis method for rolling bearings based on spatiotemporal feature fusion is proposed.Firstly,the vibration acceleration signal and time series data are extracted,and the long and short-term memory network in the classification model is used to extract the time series features of the bearing data set.The convolutional network extracts the vibration acceleration signal features of the rolling bearing,then fuses the spatiotemporal features extracted by the two networks,and uses the fully connected layer fusion algorithm to update the network parameters and features.identify.The experimental results show that this method has more significant feature extraction ability compared with convolutional neural network,long and short-term memory network,support vector machine and other methods,and the final classification accuracy is better than the above traditional methods,which proves the effectiveness of this method.nature and superiority.2.A method for predicting the remaining useful life of bearings based on improved D-cov AE and GAU is proposed,which constructs HI directly from the original vibration signal,first preprocesses the original bearing vibration signal through a low-pass filter,and then The vibration signal is input into the improved D-cov AE model,and the HI of the bearing is constructed.Compared with AE,DNN,KPCA and other methods,the constructed HI value has the ability to construct a better HI from the original bearing vibration signal in terms of comprehensive performance.After several comparative experiments,the results show that the HI constructed by the quadratic function-based D-cov AE model has stronger predictive ability than the traditional data-driven HI.Finally,the HI value constructed by the D-cov AE model is input into the GAU model.The experimental results show that the proposed D-cov AE model has stronger predictive ability than the HI constructed by the traditional method.