Research On Intelligent Diagnosis Method For Rolling Element Bearing Based On Machine Learning

Mechanical equipment,which plays an important role in modern industry,directly affects the normal operation of industrial production in enterprises.It is necessary to repair it in time to prevent more damage happening when the mechanical equipment is in the early failure stage,so it requires the technology can diagnose early minor faults.The diagnosis process of mechanical equipment contains basically three steps: The first step is to acquire the diagnostic information;the second step is feature extraction,which extracts the fault feature information from the mechanical vibration signal;the third step is the state recognition and fault diagnosis.Feature extraction is a key step in fault diagnosis technology,which directly affects the accuracy of fault diagnosis and the reliability of early fault prediction.Therefore,how to extract the optimal low dimensional fault features to improve the performance of fault classification is a huge challenge.Status recognition stage is about identification of samples with pattern recognition and machine learning methods.Different identification methods will affect the accuracy of identification to a certain extent.This thesis studies the feature extraction and diagnosis technology of rolling bears with different machine learning methods.There is redundancy or no correlation between multiple feature parameters of complex fault equipment,which is not conducive to fault diagnosis,and the classical linear dimension reduction method can’t meet the requirements of nonlinear fault data.In order to solve this problem,the feature extraction method of low rank discriminant projection and sparse representation classifier are combined to form a fault diagnosis model.For the sake of searching for the optimal parameters of the corresponding data,to achieve the parameters adaptively,the grid search algorithm is introduced.The simulation results show that the low rank discriminant projection algorithm can accurately describe the global structure and the structure of the data.It’s feasible to apply the model to the fault feature extraction of rolling bearing.The orthogonal local preserving projection algorithm is applied to the bearing fault feature extraction because of the existing problems: there’s huge difficulty in bearing incipient fault feature extraction,however,the traditional manifold learning algorithm does not make use of the samples’ category information,and can’t quickly deal with the new samples,High dimensional feature space is composed of time domain index and wavelet band energy.The orthogonal transformation matrix is obtained by means of orthogonal local projection method,and the low dimensional vector is obtained by transforming the orthogonal transformation matrix.The effectiveness of the proposed method is evaluated by using two indexes: the divergence between the class,the divergence within the class.The simulation results of the rolling bearing fault data verify the superiority of the algorithm.We usually have to continue to extract features when time spectrum diagram of vibration signal of mechanical equipment is acquired by using the time-frequency analysis method,however,the feature extraction is also more difficult,and the important information of vibration signal may be lost.In order to solve this problem,we introduce the support tensor machine method in the field of machine learning to solve the problem of tensor sample identification.The method can recognize the different states of the three-dimensional time spectrum directly.It not only simplifies the process of fault diagnosis,but also can improve the accuracy of fault diagnosis.