Research On Sparse Representation Theory Method For Rolling Bearing Fault Diagnosis

As one of the most widely used part of mechanical equipment,the rolling bearing failure will affect the performance of the mechanical equipment immediately.Consequently,it’s very crucial for predicting and diagnosing the rolling bearing fault.The vibration and fault-related information generated by the mechanical equipment during operation are the primary data reflecting the change of the mechanical equipment and its operation state.The vibration signal data is obtained by the testing instrument and the vibration signal-based fault diagnosis method is the main way to carry out the status monitoring and fault diagnosis of the mechanical equipment.In the early failure period,the fault feature from the vibration signal is very weak and often submerged in the strong background noise due to being influenced by the equipment itself,the environmental noise and the complex transmission path.This situation will make the fault feature of the equipment signal difficult to identify,hence the condition monitoring and fault diagnosis of the rolling bearing become more difficult.Therefore,it’s extremely important to study how to extract the impulse feature from the vibration signal for fault diagnosis.Aiming at the difficulty in feature extraction of early weak faults,the research about the rolling bearing fault diagnosis is carried out based on the sparse representation theory in this paper.Firstly,an adaptive sparse representation method is proposed for fault diagnosis based on the time-domain feature of the vibration signal.Meanwhile,a fault feature extraction method is proposed for retaining the existing fault information of the vibration signal via combining the sparse model and its corresponding solution algorithm.Moreover,the typical fault feature enhancement method is optimized by the sparse representation method to strengthen the weak fault feature of the vibration signal further.Finally,the fault feature of the compound fault signal is extracted combining with the advantages of the singular spectral decomposition method and the sparse representation theory method.The main research is listed as follows:(1)When the rolling bearing failure occurs,the measured vibration signal has the characteristic of the periodic impulse.Based on the time-domain characteristics of the fault vibration signal,an adjacent difference sparse representation model is proposed to improve the anti-noise property and sparsity of the sparse representation model.Meanwhile,a new sparse representation model is proposed based on the l1-norm regularization term and the adjacent difference sparse representation model.Moreover,a weighted function method is developed for setting the regularization parameters of the sparse representation model adaptively.Finally,the simulation signal data and laboratory rolling bearing data are taken as the experimental object for verifying the performance of the proposed method.The experimental results show that this proposed method can remove the noise interference effectively and extract the fault feature from the bearing vibration signal.(2)In the previous regularization parameter setup method,the regularization parameters are often not set based on the characteristics of sparse model and the solving algorithm,which results in the partial loss of fault information.To solve this problem,a sparse representation method is proposed based on the integration of the sparse representation model and its corresponding solution algorithm,which can set the regularization parameters without loss of fault information.Because the weighted Lasso model has good sparse property,the weighted Lasso model is used as the sparse optimization model,and the forward backward splitting algorithm is adopted to solve the sparse model.On basis of the characteristics of the corresponding solution algorithm,the sparse quantization index is introduced to adjust the number of nonzero signal points during the iteration process,so as to adjust the regularization parameter dynamically and achieve the purpose of adaptive regularization parameter setup.Compared with other sparse representation algorithms and typical filtering algorithms,the proposed method has the better filtering effect and fault feature extraction ability.(3)In the early failure period of rolling bearing,the fault feature of the vibration signal is very weak.For strengthening the fault feature of the vibration signal,a new minimum entropy deconvolution method is proposed based on an adaptive sparse representation method.Based on the fact that the sparse representation method has better ability for extracting the impulse feature,the adjacent difference sparse representation is applied to improve the performance of minimum entropy deconvolution.The experimental results of the numerical simulation signal experiment and real bearing fault signal show that the proposed method has the low sensitivity to the filter length parameter and can enhance the impulsive feature of the fault signal effectively.(4)Aiming at the problem that there are often many types of fault in the mechanical equipment and its components,the compound fault diagnosis method is proposed based on the integration of singular spectrum decomposition and sparse representation method.Firstly,the singular spectrum decomposition method is used to decompose the compound fault signal,and the scale containing the fault information is preserved via calculating the kurtosis index in this paper.Moreover,the proposed adaptive sparse representation method is used to remove the noise of the fault signal at each scale,suppress the noise interference on the fault feature component,and extract the fault feature from the noisy signal accurately.Therefore,the compound fault analysis can be performed on the fault signal.Due to the difficulty in extracting the early weak fault feature,the fault feature extraction method is proposed based on sparse representation with the vibration signal in this paper.Moreover,the single fault and compound fault can be extracted for rolling bearing by the proposed method in this paper effectively.In addition,the method combining sparse representation and the previous fault feature enhancement method is used to strengthen the fault feature of the vibration signal.In this paper,the feature extraction and fault diagnosis have been realized for the bearing fault signal effectively in this paper.