A Hybrid Fault Diagnosis Method for Rolling Bearing with Improved Multiscale Dispersion Entropy-based State Pre-detection and Feature Fusion

Accurate and efficient fault diagnosis for rolling bearing is essential for ensuring the reliability and safety of operation. On the premise of this background, a hybrid fault diagnosis method based on improved multiscale dispersion entropy (IMDE), systematically integrating the stages of state pre-detection and feature fusion, is proposed for rolling bearing. Specifically, a pre-detection about the health state of bearing is firstly implemented with the analysis of the IMDE distributions. Secondly, for the bearing having failure symptom, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is employed to acquire the intrinsic mode functions (IMFs) of original signals. Subsequently, the IMDE value of IMFs, served as the fusion features pool, can be calculated. Finally, with the Softmax model, the fusion features can be used to identity the fault states. Related experiments are conducted to demonstrate the performance of the developed method, and results obviously indicate the effectiveness and superiority of the method.