Research On Fault Diagnosis Method Of Motor Bearing Based On Deep Learning

The motor is the most widely used power supply equipment and drive unit,and many large plant areas use motors.Due to the long-term high-intensity of operation,the motor inevitably has some troubles.In the process of industrial production,the frequency of motor failure is relatively high.As a kind of integrated electrical equipment consisting of bearings,stators,rotors,the motor often suffer from various types faults.Among them,bearing fault is one of the most common faults.Therefore,it has good theoretical and practical significance for the fault diagnosis of motor bearings.Traditional bearing fault diagnosis methods rely on the accumulated experience of technicians,but there are many cases that cannot be handled and judged in time will consume a long time and manpower.With the development of artificial intelligence,the current deep learning-based fault diagnosis technology relies on powerful feature extraction capability and massive data processing capability to achieve expected results.In this thesis,the motor of industrial spot of Anshan Iron and Steel is taken as the research background and the bearing data set of Case Western Reserve University and Anshan Iron and Steel were selected as the experimental data.Firstly,the wavelet transform is used to analyze and process the vibration signal of bearing fault.Support Vector Machine is used as a traditional machine learning method for bearing fault diagnosis.Secondly,aiming at shortcomings of traditional methods,a fault diagnosis method based on Convolutional Neural Network(CNN)is proposed.In order to solve the problem that the initial value of complete randomization of neural network will lead to slow convergence and poor learning performance,ant colony algorithm is introduced to optimize model parameters.The comparative analysis shows that Convolutional Neural Network based on ant colony algorithm has better convergence and learning performance.Finally,Convolutional Neural Network is utilized to bearing fault diagnosis.The experimental results demonstrate that the introduction of ant colony algorithm improves the convergence speed and diagnostic accuracy of Convolutional Neural Network.Convolutional Neural Network with supervised learning can not solve the problem of insufficient fault data and sample labeling in actual production.A fault diagnosis method for motor bearing based on the Generative Adversarial Networks(GAN)is proposed.In order to solve the problems of training stability and unsatisfactory effects of generation and recognition,Convolutional Neural Network optimized by the ant colony algorithm is used as a discriminator model,in case of the randomness and uncertainty in the data distribution of generated samples during training,a conditional model is introduced into the generator,and the fault type is used as the conditional variable for connection,to solve the problem of sample labeling,this thesis introduces a semi-supervised learning method and optimizes the generator and discriminator loss functions.The test proves that the CSGAN obtains better sample generation performance and classification accuracy.Finally,Generative Adversarial Network is applied to bearing fault diagnosis.The experimental results illustrate that the improved model solves the problem of insufficient samples and sample labeling,and improves the quality of generated samples and diagnostic accuracy.