Fault Feature Extraction of Compound Planetary Gear Based on VMD and DE

Abstract

Vibration signal of compound planetary gears is complex, so it is very hard to extract fault feature and diagnosis fault. This paper proposes new fault characteristic parameters based on VMD (Variational Mode Decomposition) and DE (Dispersion Entropy). Firstly, VMD is adopted to decompose the vibration signal and obtain a set of IMF (intrinsic modal function). Second, the signals is reconstructed by some IMFs according to the mutual information criterion. Third, dispersion entropy of the reconstructed signal is calculated. Finally, DE is input as a eigenvalue to the PSO-SVM (particle swarm optimization and support vector machine) classifier to implement fault pattern recognition. The experimental results show that the features proposed in this paper can distinguish the three states of normal gear, sun gear spall and planetary gear spall with 100% accuracy.