| In recent years, the level of science and technology makes a spurt of progress in China. Along with the process of new industrialization, a large-scale mechanical application becomes increasingly common, automation, and informatization. It also vigorously promoted the development of the society and economy. However, the unplanned downtime and partial failure will directly lead to impaired production benefit and huge economic losses, and even accidents, in the production and application in industry. As a key component of rotating machinery, the rolling bearing, its running-state does often directly affect on the performance of the whole machine. So, it is very highly significant to monitor and diagnose the running-state.The essence of fault diagnosis of rolling bearing can be summed up as the feature extraction and fault identification. Aimed at the difficulty to extract the weak fault information and the low recognition rate at early stage of bearing fault, a intelligent diagnosis method based on dual singular value decomposition (dual-SVD) and variable predictive model based class discriminate (VPMCD) was put forward. It can adaptively extract the weak fault features, according to the vibration signal of the rolling bearing, to identify the fault.The main contents of this paper are as follows:1) This paper has studied the application mechanism of SVD in signal processing, and put forward to the method of dual-SVD. It can effectively solve the difficult problem that the different number of singular values affects the accuracy of fault diagnosis, caused by SVD for different signal. Combined with the advantage of energy moment in feature extraction, the method can be successfully applied to extract the weak fault information at early stage of bearing fault, and the results accurate and valid. Moreover, the dual-SVD using the singular value contribution rate to adaptively construct Hankel matrix, according to the signal of its characteristics. It provides technical basis for on-line monitoring and diagnosis of bearing faults, by greatly reducing the calculation time of large-scale matrix. In addition, the method further weakened the experience knowledge and limited conditions, by using the singular value curvature spectrum method to automatically select the effective singular values. They are all making the processing method to become much "smarter".2) The VPMCD is a new pattern-recognition method. The fault identification model can be established by VPMCD, which used of the mutual internal relations of eigenvalue. Then, the different fault identification models, obtained by the different fault categories, were used for the prediction of the measured-samples’eigenvalues. And then, the fault recognition based on the prediction results. So, the models are well suitable for processing the pattern-recognition problems that non-linear and multi faults. Actually, its essence is the process of parameter identification, to avoid the iterative of artificial neural network (ANN) and the optimization process of support vector machine (SVM). Moreover, it greatly makes the computation and training time smaller, and improves the algorithm efficiency.3) Aimed at the difficulty to extract the weak fault information and the low recognition rate at early stage of bearing fault, the method, based on dual-SVD and VPMCD, was applied to the actual fault diagnosis of rolling bearing under the conditions of the partial failure. The result shows that the method presented here is feasible and valid. And it has achieved the full intelligent from signal processing to pattern-recognition. In addition, it shows the superiority of the method proposed in this paper by comparing with the methods of bearing fault diagnosis in recent years, from two aspects:feature extraction and pattern-recognition. Further more, it highlights the advantages of high accuracy rate and the faster algorithm in fault diagnosis of rolling bearings. So, this method provides a new way of rolling bearing for online intelligent diagnosis.
|
PDF Full Text Request