Research On Fault Diagnosis Of Motorized Spindle Of Machining Center Based On Deep Belief Net-work

In industries such as automobiles,molds,and aviation,machining centers are important equipment for processing precision and complex mechanical parts.If the electric spindle of its key components fails,it will affect the safety and reliability of the overall operation of the ma-chining center.Therefore,it is of great significance to carry out the research on the fault diag-nosis of the electro-spindle.Due to the complex internal structure of the electro-spindle,in the process of monitoring its different operating states,the signal collection of multi-sensor fusion and multi-frequency sampling settings causes the fault data to have the characteristics of high dimensionality,heterogeneity and nonlinearity.Since traditional fault diagnosis technology cannot extract deep features from complex data,Deep Belief Network can adaptively extract rich and effective features from complex data,and then more accurately map the signal and The complex relationship between the fault characteristics,that is,this article uses DBN to re-place the traditional diagnosis technology.Therefore,this paper uses the effective time-frequency domain characteristics of the fault vibration signal of the electric spindle as the train-ing data of DBN to complete the intelligent fault diagnosis of the electric spindle of the ma-chining center.The main research contents of this paper are as follows:(1)Explain the internal structure and working principle of the electro-spindle,and then the statistical data shows that the common faults of the electro-spindle include bearings,rotat-ing parts,housings,etc.At the same time,based on the above analysis,further study the cause of the failure and clarify the common faults Vibration characteristics of is the main feature,and the effective value,root mean square value,variance and kurtosis and other characteristic indi-cators that can characterize the state of the electro-spindle system are selected.(2)Starting from the principle and structure of DBN,the pre-training process and back-propagation process of the network are explained.Among them,the Tanh activation function used in the back-propagation process has problems such as the disappearance of gradients,which is based on Tanh.Activation function(IM-Tanh),and theoretically deduced this function not only improves the training efficiency of DBN,but also improves the accuracy of classifica-tion,and is verified with the public MINST handwritten data set.(3)Through the single fault experiment of the electric spindle bearing and the compound fault experiment of the electric spindle system,the effectiveness of the DBN method based on IM-Tanh proposed in this paper is proved.The method is compared and analyzed with the DBN method using traditional activation functions such as Sigmoid,Tanh,Relu,etc.and tradi-tional fault diagnosis methods such as empirical mode decomposition(EMD),ensemble empir-ical mode decomposition(EEMD),etc.The results show that the fault diagnosis proposed in this paper The method is highly efficient.(4)Develop a fault diagnosis system with a good human-computer interaction interface based on Python Web,use uploaded real-time data and online fault diagnosis DBN model to realize real-time monitoring of equipment operating status and visual presentation,laying the foundation for subsequent use in actual scenarios.