A multi-entropy fusion approach for rolling bearing fault diagnosis integrated with variational mode decomposition

The monitoring of bearing health is difficult due to the presence of background noise in the acquired vibration signal along with non-linear and non-stationary characteristics. Targeting these factors a new bearing fault diagnosis technique is proposed based on integration of variational mode decomposition (VMD) and multi entropy fusion approach (MEFA). This methodology is proposed to increase the efficiency of the fault classification of rolling bearings. In this approach, the acquired vibration signal is initially denoised using VMD approach. The entropy values are then calculated using MEFA comprising of approximate entropy (ApEn), sample entropy (SampEn), fuzzy entropy (FuzzyEn), permutation entropy (PE) and dispersion entropy (DisEn). These values are used as feature vectors for the present study. K-Nearest Neighbors (KNN) and Random Forest (RF) are used as classifiers for classifying different types of faults. The proposed methodology is tested on the bearing dataset of an electro-mechanical system, and its accuracy results are compared with the integrated local mean decomposition (LMD) with MEFA. The results from the experimental investigation prove the efficacy of the proposed method.