Research On Fault Diagnosis Method Of Motorized Spindle Based On Deep Learning

The motorized spindle is the key component of CNC machine tools.When the motorized spindle fails,it will affect the safety and reliability of the overall operation of the machine tools.Therefore,it is of great significance to study the fault diagnosis of motorized spindle.In this paper,the motorized spindle is analyzed and studied,and the traditional fault diagnosis method and the fault diagnosis method based on deep learning are used to study the vibration signals of motorized spindle of CNC hobbing machine.The main research contents are as follows:(1)An experimental platform for fault diagnosis of motorized spindle is set up,and based on this,three types of vibration signals of motorized spindle are obtained,and the vibration signals are preliminarily analyzed.(2)The original vibration signal of motorized spindle is denoised by singular value decomposition,and the denoised signal is decomposed by variational mode,and then the energy of eigenmode function after decomposition is extracted as feature quantity.Finally,the extracted feature quantity is input to BP neural network for classification.The results show that the fault diagnosis method has a good recognition effect on motorized spindle signal,and the accuracy rate reaches 93%.(3)Aiming at the uncertainty of fault diagnosis effect caused by the traditional fault diagnosis methods relying on artificial feature extraction,a fault diagnosis method of motorized spindle based on dilated convolution capsule network(DCCN)is proposed.Firstly,the original vibration signal is subjected to short-time Fourier transform to obtain the time-frequency map,and the time-frequency map is used as the input of DCCN for classification.The results show that DCCN has better performance than the traditional fault diagnosis method,and the recognition accuracy of motorized spindle signal reaches 99.55%.(4)In order to solve the problem of unbalanced classification in fault diagnosis of motorized spindle,a fault diagnosis method of motorized spindle based on CCGAN-DCCN is proposed.A convolution-based Conditional Generation Adversarial Network(CCGAN)model is used to expand the unbalanced time-frequency map data sets,and then the time-frequency map data sets before and after expansion are input to DCCN for training.Then the DCCN model trained by each other is used to identify the time-frequency map test sets,and the recognition accuracy of the two is compared.The results show that DCCN trained by the expanded data set has higher recognition accuracy for the test set.