Fault Diagnosis Method For Rolling Bearing Based On Double Sparse Dictionary Joint With DBN

Rolling bearing intelligent fault diagnosis technology is a research hotspot in the field of rotating machinery condition monitoring.Among them,feature extraction is the key technology of rolling bearing intelligent fault diagnosis method under the background of data driving.The fault features extracted based on the time-frequency domain are highly subjective,the work is complicated and the acquired features sometimes cannot well represent the fault feature information contained in the signal itself,and the discrimination is low.In view of this,the signal sparse representation theory is applied to the field of fault diagnosis,and the fault diagnosis method for rolling bearings based on deep belief network is researched.The research results of this article are as follows:(1)Aiming at the traditional intelligent fault diagnosis method,the feature parameters in the time-frequency domain are sometimes unable to fully describe the characteristics of the signal fault category when feature extraction is performed,and a fault feature extraction method of rolling bearing based on the sparse representation of the double sparse dictionary learning algorithm is proposed.Firstly,the sub-dictionaries were trained according to the double sparse dictionary learning algorithm.Then,in order to reduce the feature dimension,the redundant atoms in the double sparse sub-dictionary obtained from the training of various fault signals were removed according to the selected sparse decomposition algorithm.And then,the remaining atoms are arranged in order to obtain the final double sparse comprehensive dictionary to extract fault features.The results of simulation experiments show that this method not only reduces the feature dimension of the original signal,but also the data points of the sparse representation signal are sparse.In addition,the distribution position of the sparse points of the characteristic signal of each type of fault is obviously different from signals of other types.(2)In traditional rolling bearing intelligent fault diagnosis methods,the choice of feature extraction method depends on the subjective judgment of the researchers when using timefrequency domain features for fault diagnosis.The performance of the extracted features depends heavily on the knowledge of the relevant practitioners at the level of experience,the fault features extracted by some models are sometimes not clear and the discrimination is low,which leads to misjudgment easily in the pattern recognition process.Aiming at this,a fault diagnosis method for rolling bearings based on double sparse and deep belief networks(DBN)is proposed.Firstly,a sparse comprehensive dictionary is used to sparsely decompose each type of fault signal and fault labels are established correspondingly.Then a DBN is constructed.The final fault diagnosis model is trained by using sparse representation characteristic signals.Finally,the classification accuracy of the entire model is tested.The results of simulation experiments show that the method has the advantages of fast speed,high accuracy and stability in identifying the fault types of rolling bearings.(3)Among the traditional intelligent fault diagnosis methods,commonly used intelligent fault diagnosis methods include a method based on time-frequency domain characteristics combined with support vector machine(SVM),directly using original signals combined with deep neural networks(such as multiple hidden layer BPNN)and shallow neural network(such as single hidden layer BPNN)for fault diagnosis.The simulation results of the rolling bearing fault diagnosis method compared with the commonly used intelligent fault diagnosis method are compared and analyzed by simulation experiments,which further confirms the feasibility of the proposed method,and analyzes the underlying reasons of the final results obtained by different methods.Finally,by comparing the fault diagnosis effect of using the original signal directly combined with DBN and the method proposed in this paper,the role of dual sparse comprehensive dictionary in fault diagnosis is summarized.