A Novel Intelligent Fault Detection Scheme for Rolling Bearing Based on Morphological Multiscale Dispersion Entropy

Abstract

This paper develops a novel fault detection method for rolling bearing based on morphological multiscale dispersion entropy (MMDE). MMDE is mainly made up of two parts (i.e. morphological filtering and multiscale dispersion entropy (MDE)). Concretely, the original vibration data under different fault status is first preprocessed by morphology-hat product operation (MHPO). Afterwards, MDE of the filtered signal is calculated for the purpose of feature extraction under multi-fault conditions. Finally, the constructed fault feature matrix is fed into particle swarm optimization-based support vector machine (PSO-SVM) for realizing the identification of different bearing fault conditions. The efficacy of the presented method is validated by applying the experimental examples. Results demonstrate that MMDE can work more effectively in recognizing bearing fault type than several popular entropies (e.g. MDE, MPE and MSE).