Research On Fault Diagnosis Method Of Motor Based On Deep Learning

As an energy conversion device,the motor is widely used in people’s daily life and industrial production.In industrial production,due to the long-term high-speed and high-load operation of the motor,coupled with the lack of daily inspection and maintenance,the operation of the motor has left a safety hazard.Once the motor fails,it will not only affect the smooth operation of the equipment,but also cause certain losses to the enterprise,and may even threaten the personal safety of the on-site staff.Therefore,in order to ensure the smooth,efficient and reliable operation of the motor,it is very important for the motor to perform remote monitoring and fault diagnosis.The object of this thesis is asynchronous motor,and the method of deep learning is used to conduct in-depth research on the intelligent fault diagnosis method of motor.The main work of the paper is as follows:(1)Aiming at the problems of extraction of motor fault features in traditional methods,relying on expert knowledge and low fault recognition rate,this paper proposes a motor fault diagnosis method based on one-dimensional convolutional wavelet neural network(1D-CWNN)to realize end-to-end motor fault diagnosis.This method improves the traditional CNN model,adds a batch normalization layer to optimize the network structure to avoid gradient explosion,extracts the deep features of the information through the convolution layer and the pooling layer,and then outputs the wavelet layer through the global average pooling,adding wavelet layer to make the wavelet function replace the traditional activation function to improve the convergence ability of the model,and finally classify and identify through the Soft Max layer.Through experiments,it can be seen that the fault recognition accuracy of the 1DCWNN model can reach 98.2%,which can well diagnose motor faults.Compared with the 1D-CNN model,the 1D-CWNN model can effectively suppress local overfitting,and the convergence speed is faster.(2)In order to further improve the accuracy of the motor fault diagnosis model and enhance the robustness of the fault diagnosis model in the environment of strong noise interference,a motor fault diagnosis method based on MSWN-ALSTM is proposed.On the basis of 1D-CWNN,a multi-scale module and a long short-term memory(LSTM)network are introduced to improve the model,and use the attention mechanism to highlight the characteristics of feature information that are closely related to faults,and propose a motor fault diagnosis method based on MSWN-ALSTM.Through experiments,it can be seen that the model has excellent diagnostic ability,and its diagnostic ability and robustness are better than other models in the environment of strong noise interference,and it has strong anti-noise ability.(3)In this thesis,the overall design of the motor remote monitoring and fault diagnosis system is carried out.The wireless sensor is an important part of the motor operation information collection in the system.In this thesis,the hardware design of the wireless sensor node is carried out,the components of each module are selected and the schematic diagram is designed to provide a specific solution for the remote wireless collection of motor operation data.