Bearing Data Model of Correlation Probability Box Based on New G-Copula Function

Abstract

Bearing failure often occurs in rotating machinery. Fault diagnosis method based on vibration signals has been studied for many years. Considering complementary information of the vibration signals from different directions, this article proposed an applied model of a correlation probability box based on G-Copula function for diagnosing bearing faults. First, to avoid constructing binary Copula function directly from the definition of binary Copula function, a new function is defined, and a construction method of binary G-Copula function is proposed based on the new function. Then, the correlation probability box model is established based on a joint cumulative distribution of the G-Copula function to increase the independent of the input data in the support vector machine (SVM) model, and the aggregated widths of the correlation probability box model can be used to monitor a development of the bearing failure. Finally, the experimental results showed that the proposed method obtain the better classification accuracy than other data processing study.