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

As the core structure of the motorized spindle,the bearing could affect the motorized spindle performance and even the entire production line.In order to ensure safe and reliable operation of the motorized spindle,it is essential to diagnose the failure of the motorized spindle bearing.With the development of deep learning,the concept about self-learning characteristics from the data itself provides a new research direction for fault diagnosis.Sparse denoising auto encoders and convolutional neural network as a method of deep learning with the strongest generalization ability,has a broad development prospect in the field of fault diagnosis.In this thesis,taking the motorized spindle bearing fault as the research object,using bearing fault data as an input sample of a SDAE and an improved CNN,the improved method was uesd to distinguish the fault type of the motorized spindle bearing.The main work were summarized as follows:(1)Combined with theoretical knowledges of the motorized spindle bearing,the simulation experiment platform was used to collect the data of the inner and outer rings faults of the motorized spindle bearing,the rotor bearing misalignment,and the rotor bearing imbalance.(2)The structure and training method of SDAE was analyzed.At the same time,the excellent classification performance of the SDAE was verified by the simulation signals of the mild and severe faults of the inner and outer rings of the motorized spindle bearing.Compared with the classification accuracy of traditional neural network and SDAE,the advantages of SDAEwere summarized.(3)In order to improve the accuracy of motorized spindle bearing fault diagnosis and reduce an over-fitting phenomenon,a motorized spindle bearing fault diagnosis method based on improved convolutional neural network CNND.In the CNN,many kinds of methods preventing over-fitting were added.After comparison,the dropout method worked best.Simulation results show that the motorized spindle bearing fault diagnosis model established by combining the dropout optimization method with the CNN was feasible,which could avoid the over-fitting phenomenon and improve the diagnostic accuracy.(4)In order to enhance the model’s ability to extract fault data features,an improved SDAE-CNN electric spindle bearing fault diagnosis method was designed.Firstly,SDAE hidden layer number and hidden layer neuron number were optimized.Secondly,the optimal value is selected for the number of convolutional layers,the number of convolution kernels,and the size of the convolution kernel.Then,training CNN with SDAE feature expression;Finally,the Adam optimization algorithm was selected to reduce the cross entropy function value.And the fault of the degree of cracking of the outer ring of the motorized spindle bearing is taken as the research object.The simulation results show that the feature extraction ability and the final fault diagnosis accuracy of this method are higher than other methods.The improved SDAE-CNN motorized spindle bearing fault diagnosis method designed in this paper can effectively identify the fault type of the motorized spindle bearing and improve the fault diagnosis accuracy.