The Application Of Fractal Dimension And Permutation Entropy In Roling Bearing Fault Diagnosis

Rolling bearing is the most widely used machine in modern society production of machinery. Along with the development of the modern society and technology progress and the production system, bearing is playing a more and more important role in the production. Therefore, the requirement of pre-fault motor diagnosis is obviously necessary.70%of electrical and mechanical equipment fault is caused by the vibration, and among the vibration faults, almost30%of them are caused by bearing fault. It is because the rolling bearing’s working environment is the worst of the complete set of equipment, which, at the same time, is the weakest link. In the operation of then equipment, it plays a role of transmitting then load and subjecting the load. Its operation state directly affects the performance of whole machine. So fault diagnose methods of bearing have always been an advanced technology.This paper described the theoretical basis of wavelet de-noising. Then, an improved wavelet de-noising method was proposed based on the advantages and disadvantages of wavelet de-noising method. The effectiveness and feasibility of the improved method was verified by the comparison of spectral analysis and data analysis.After introducing permutation entropy algorithm and analyzing the characteristic of permutation entropy algorithm, permutation entropy is introduced in fault diagnosis of rolling bearings. The results which can prove that permutation entropy could represent the change of state of rolling bearings. At same time, fractal theory is introduced. The nonlinear behaviors of box-dimensional and correlation dimension of signals are analyzed. Different characteristic of changes in the vibration data with box-dimensional data and correlation dimension after the change of state of rolling bearings is compared.This paper uses the theory of support vector machine as the intelligent diagnosis method, studies the multi-class SVM classifier and the support vector regression theory. The multi-class SVM classifier and support vector regression machine was built by adopting gauss radial basis function. Permutation entropy and fractal characteristics of rolling data in different status combine with SVM theory. In terms of the link, fault of rolling bearings is being diagnosing and predicting. By analyzing data, it is verified that permutation entropy is the better rolling bearing vibration data nonlinear feature extraction method to fractal theory.