Research On Fault Diagnosis Of Spindle Bearing For CNC Machine Tools Based On Deep Learning

Spindle is the key component of CNC machine tools.Its performance directly affects the machining accuracy and product quality.It is estimated that most of the spindle failures are mainly caused by bearing faults.The traditional fault diagnosis methods extract and select fault features manually,which results in large fluctuation of fault eigenvalue under various working conditions,affecting the fault diagnosis effect.Deep learning uses the original data to learn features and characterize the internal information.It can overcome the problem of low accuracy of fault diagnosis caused by artificial feature extraction.This study has resolved bearing fault diagnosis of spindle in multiple operating conditions using the deep learning method,the main research contents are as follows.(1)Research on single fault diagnosis of bearing in multiple operating conditions.To solve the low accuracy of single fault diagnosis for bearing under multiple working conditions caused by manual extraction fault features and setting network parameters.An ant colony optimization-deep convolutional neural network fault diagnosis method based on vibration signal fusion is proposed.Firstly,different position fault information is considered.The horizontal and vertical vibration signals are fused as the deep convolutional neural network input.Then ant colony optimization-deep convolutional neural network model is established.Using ant colony algorithm optimized deep convolutional neural network key parameters to adapt to different working conditions.Finally,the powerful feature self-extraction ability of deep convolutional neural network is used to realize single fault state identification of bearings under multiple working conditions.(2)Research on compound fault diagnosis of bearing in multiple operating conditions.The compound fault signals have characteristics of complex components and coupling among faults.The fault diagnosis time is long and the fuzziness is high under multiple operating conditions.So an improved ant colony optimization-deep convolutional neural network fault diagnosis method based on wavelet transform is proposed.Firstly,wavelet transform is used to transform one-dimensional fault signals into two-dimensional images with powerful information.Secondly,the global average pooling layer is used to replace the full connection layer of the traditional deep convolutional neural network to reduce the number of parameters and improve the performance of the method.Finally,an improved ant colony optimizationdeep convolutional neural network fault diagnosis model is established to realize compound fault state identification of bearings under multiple working conditions.(3)Design and implementation of fault diagnosis system.Taking Win CC software as the development platform.Based on the above fault diagnosis methods,this system is mainly designed from the headend and the terminal.Realizing the functions of data acquisition,data monitoring,fault warning,fault diagnosis and data management,providing a simple and convenient operation platform for operators and managers.The results show that proposed fault diagnosis methods based on deep learning can effectively resolve single faults and compound faults of bearings under multiple working conditions,and the accuracy rates of fault identification are 99.07% and 97.89% respectively.Both of which have high fault diagnosis accuracy.