Research On Fault Diagnosis Of Motor Bearing Based On Convolutional Neural Network

Bearing fault is one of the common faults of motor,so the research on its diagnosis method is of great significance.With the advent of the era of big data,traditional diagnosis methods that require expert experience and manual feature extraction are not conducive to real-time fault monitoring and intelligent operation and maintenance of motor units.Aiming at this problem,this paper proposes a fault diagnosis method for motor bearings based on convolutional neural network,which can automatically extract fault features and has high fault identification performance.(1)A fault diagnosis process for motor bearings based on one-dimensional time-domain vibration signals is proposed,and a one-dimensional convolutional neural network is constructed to automatically extract fault features through training to achieve the purpose of fault identification.The model demonstrates its performance on the Case Western Reserve University Bearing Failure dataset.(2)Propose a data processing method that converts one-dimensional time-domain signals into two-dimensional grayscale images,constructs a two-dimensional convolutional neural network,effectively utilizes the image processing advantages of convolutional neural networks in the visual field,and improves the diagnosis speed and diagnosis.precision.Introduce data enhancement technology to improve effective samples,add regular terms to the model,effectively suppress overfitting and improve generalization ability.(3)Combined with the time-frequency domain analysis method,the vibration signal is wavelet transformed to obtain a color time-frequency map,a two-dimensional convolutional neural network based on the time-frequency map is constructed,and the optimal hyperparameters of the model are determined by the control variable method.In addition,transfer learning is introduced into the model to solve the problem of low diagnostic accuracy when the effective samples are insufficient under variable working conditions,and to improve the fault identification ability of the model under variable working conditions in practical application.(4)The visualization of the two two-dimensional neural networks proposed in this paper is carried out.First,the intermediate outputs of the two models are visualized,showing the feature maps learned by each layer.Secondly,using the Tensor Board visualization framework,the framework structure visualization,indicator monitoring visualization and activation parameter visualization of the model proposed in this paper are realized,and the visualization results are analyzed.Through visualization,the internal mechanism and operation principle of convolutional neural network in fault diagnosis process are studied.The models proposed in this paper have high fault identification perform ance and strong generalization,and provide a new idea for the research of motor bearing fault diagnosis.