Fault Diagnosis Of Motor Bearing Based On Multi-input Modified GoogLeNet Convolutional Neural Network

Today,electricity has become an indispensable power source in today’s social development and daily activities,and the motor in the power system is the most important power source and driving device,and the bearing is one of the important components of the motor.Bearing health monitoring and fault diagnosis is of great significance.The traditional fault diagnosis methods require rich fault diagnosis knowledge,and the diagnosis process is extremely complex,time-consuming and labor-intensive,which limits the development of diagnosis technology.develop rapidly.Therefore,in view of the fault diagnosis of motor bearings,this paper proposes three fault diagnosis models for motor bearings,combining the characteristics of the two convolutional neural network models and the advantages of data processing methods.Firstly,in view of the complex and tedious problem of the traditional fault diagnosis method in the diagnosis process,combined with the structural characteristics of the GoogLeNet model and the Dense Net model,an improved GoogLeNet model was constructed.Compared with the original GoogLeNet model,the improved model has stronger feature extraction ability and higher fault diagnosis accuracy,and the improved GoogLeNet convolutional neural network is verified.Feasibility of network models in fault diagnosis of motor bearings.Then,in order to further improve the accuracy of the improved GoogLeNet model in the fault diagnosis of motor bearings,an improved GoogLeNet based on time-frequency domain features is proposed to solve the problem that the fault features in the original time-domain signal are not prominent,difficult to extract and express.The motor bearing fault diagnosis model of convolutional neural network,the test converts the original data into a twodimensional time-frequency graph,which is used as the input data to verify the improved GoogLeNet model,and the optimal parameters of the model are obtained through the control variable test,and finally the construction A model with better fault diagnosis performance under any single working condition is developed.Finally,with the in-depth research,it is found that the proposed model does not perform well under multiple working conditions.Therefore,for the problem of motor bearing fault diagnosis under multiple working conditions,an improved GoogLeNet convolutional neural network model based on multiple inputs is proposed.The model has two input layers,which can extract features from different types of data at the same time,deepen the utilization of features,and finally perform feature channel fusion and expansion for fault diagnosis.Through experiments,it is proved that the improved GoogLeNet convolutional neural network model based on multiple inputs has the advantages of high fault diagnosis accuracy,strong feature extraction ability,more effective and comprehensive feature extraction in motor bearing fault diagnosis under multiple working conditions.The variable comparison test determines the optimal parameters of the model,which has a certain use value.