Bearing Fault Diagnosis Based on the Refined Composite Generalized Multi-Scale Bubble Entropy

摘要

Entropy is an efficient method to measure the randomness of signals and the sudden change of nonlinear dynamics. A new feature as refined composite generalized multi-scale bubble entropy (RCGMBE) is proposed to represent fault signals in order to suppress the noise interference of rolling bearing fault signals and accurately identify the rolling bearing fault types. And then the bearing fault diagnosis is achieved with the combination of neighborhood preserving embedding (NPE) and the extended nearest neighbor method (ENN). Through the validation of experimental data, it is shown that RCGMBE is more stable than the generalized multi-scale bubble entropy (GMBE), and the classification accuracy of the rolling bearing with the proposed approach is high.