Research On Bearing Fault Diagnosis Method Based On One-Dimensional Convolution Neural Network

Nowadays,with the continuous development of science and technology and the continuous refinement and integration of electrical equipment,the electrical system has strict requirements for the operation quality of each equipment.As the necessary equipment in the drive system,motors are widely used in all levels of society.Once problems occur,they may cause the shutdown of the whole system,or even cause casualties.Therefore,it is of great significance to detect and diagnose motor faults,which can not only reduce economic losses,but also prevent personal injury.As the most widely used motor equipment,the bearing of asynchronous motor is also a part prone to failure.Therefore,fault detection based on motor bearing is of great significance to reduce economic losses and ensure personal safety.With the arrival of the era of electrical big data and the continuous development of artificial intelligence,intelligence has become the global development goal.The traditional signal analysis methods have gradually become inadequate,and the analysis methods based on artificial intelligence have gradually extended to the field of motor fault diagnosis.In thesis paper,the one-dimensional convolutional neural network model in depth learning is used to solve the problem of accurate diagnosis of motor bearing faults under the interference of noise and load change.Both theoretical and practical aspects are studied in depth.1)The parameters of one-dimensional convolutional neural network model and structure are analyzed.The one-dimensional convolution neural network model is applied to motor bearing fault diagnosis,and the structural parameters of the model are optimized.The one-dimensional convolution neural network(DCNN)model with the first parallel convolution kernel is designed.The research results show that the onedimensional convolutional neural network based on DCNN can analyze the motor bearing fault signals end-to-end,which proves the feasibility of this method.The parameters in the DCNN model are debugged and verified with CWRU data set,which illustrates the good effect of this method and enriches the existing motor bearing fault diagnosis methods.2)Research on bearing fault diagnosis under noise and variable load disturbance.A DCNN(DACNN)model diagnosis method with attention mechanism is proposed to solve the problem of the influence of noise and variable load on the accuracy of bearing fault diagnosis.The results show that the diagnostic accuracy of the DACNN model reaches 97.3% on average under the condition of variable load;In the case of noise interference based on different SNR under variable load,the average diagnostic accuracy of the model is 93.8%,which shows that the model has good anti noise performance and adaptive ability under variable load.3)Research on bearing fault diagnosis method based on the fusion of signal analysis and deep learning.Combining the excellent de-noising ability of signal analysis method with the powerful data processing ability of deep learning,two bearing fault diagnosis methods are proposed,which integrate wavelet analysis into DACNN and empirical mode decomposition into DACNN.The experimental results show that the method of combining empirical mode decomposition with DACNN can obtain better diagnostic effect,and the diagnostic accuracy of this method is about 5% higher than that of the DACNN model under the condition of strong noise and variable load,which proves the superiority of this method.