A Research On Intelligent Fault Diagnosis Algorithm Of Rolling Bearing Based On Deep Learning Theory

As one of the core parts of mechanical equipment,the failure of Rolling bearing will bring irreparable loss to the whole mechanical system.Therefore,it is particularly necessary to establish a fault detection system.With the vigorous development of sensors and various detection equipment,more and more data is collected from sensors,the traditional method of signal processing and shallow machine learning can not meet the needs of fault data’s analysis and processing under the background of "big data" due to its limitations.As a key role in the wave of artificial intelligence,deep learning technology is sweeping though many fields with its powerful data mining and nonlinear feature extraction ability,including fault diagnosis.Therefore,the fault diagnosis method based on deep learning theory provides a new idea for modern fault diagnosis methods.Taking rolling bearing as the research object,based on SDAE network and improved CNN network,a variety of fault diagnosis methods are proposed in this paper,which can automatically complete the feature extraction and recognition of rolling bearing fault.The main tasks are as follows:(1)The fault form and fault characteristic frequency of rolling bearing are analyzed.Combined with the characteristics of fault signal,a FVMD signal denoising algorithm based on VMD and wavelet algorithm is proposed and applied to practical fault data.The simulation results show that the method can effectively filter out the noise interference in the original signal,which lays a foundation for the follow-up research.(2)By focusing on the study of structure principle on SDAE network,its internal training mechanism and data flow can be described in detail.At the same time,the performance of SDAE network under different super-parameter combinations is carried out.The optimal superparameter combination is determined by using rolling bearing fault data set.On this basis,a rolling bearing fault diagnosis model based on SDAE network is built.The simulation results show that the SDAE network under the optimal super-parameter combination can effectively identify the fault type,and has higher accuracy than other network models.(3)The application of image feature extraction method in rolling bearing fault diagnosis is studied,and a fault diagnosis method based on GLCM-SDAE model is proposed.The gray level co-occurrence matrix which can effectively analyze the texture features of fault timefrequency images is selected as the feature extraction tool,and the extracted fault features are input into the improved SDAE network model.Compared with the feature extraction ability by different models,it is found that the fault diagnosis method based on image feature extraction plays an obvious role in fault diagnosis under different types and different working conditions.(4)Aiming at the identification of fault location and fault size of rolling bearing,a hierarchical intelligent fault diagnosis model based on multi-channel convolution neural network and SDAE network is proposed.In the first layer,the concept of coarse-grained feature extraction is introduced to increase the robustness and richness of the extracted fault features.In the second layer,a SDAE model is established for each fault location to identify the fault size of rolling bearings.The simulation results show that the model has high recognition accuracy in fault location recognition and fault size recognition.The hierarchical intelligent fault diagnosis method designed in this paper can effectively identify the fault location and fault size of rolling bearings,which provides an idea for guiding the replacement of mechanical parts and the prediction of the remaining life of equipment.