A feature extraction method based on moving multi-scale reconstruction and interactive energy entropy for gear fault diagnosis

Abstract

In order to accurately extract the sensitive features representing the type and severity of gear faults through the vibration signal, a gear fault diagnosis method using moving multi-scale reconstruction-based interactive energy entropy (MMS-IEE) is proposed. The gear vibration signal is reconstructed using a multi-scale mean at different scales and adjacent data points are used to form a sliding window, which makes the information extraction from the vibration signals sufficient. The energy distributions of the original signal and the reconstructed signal under different scale channels are calculated. Compared with the traditional energy entropy (EE) method, the feature vector obtained by the interactive superposition method can more accurately represent the energy mutation of the time-series caused by the fault. Experimental results show that the proposed MMS-IEE method has a strong fault feature extraction ability and high gear fault diagnostic accuracy under different speeds and working conditions.