Gearbox Fault Diagnosis Based on Multifractal Detrended Fluctuation Analysis and Improved K Means Clustering

This paper addresses a new method of fault diagnosis for parallel shaft gearbox. Aiming at the nonlinear and non-stationary characteristics of gearbox vibration signals, the Multifractal Detrended Fluctuation Analysis (MFDFA) is introduced to calculate the multifractal spectrum parameters as fault features, and combined with the improved K means clustering to detect the failure of gearbox. The above methods are verified through using the fault data of gearbox preposition failure experiment, and the result shows that the methods have good effect.