Gearbox fault diagnosis using EMD method combinated with BP neural network

Abstract

The empirical mode decomposition (EMD) is a new method for adaptive
analysis of nonlinear and nonstationary signals. In this paper, the EMD
method was combined with the BP neural network to analyze the vibration
signals and evaluate the working conditions of the gearbox. Firstly,
vibration signals are decomposed into some components intrinsic mode
functions (IMFs) by EMD method. Then, the feature vectors are extracted
from a number of IMFs that contained the most dominant fault
information. To identify gearbox fault patterns, the feature vectors could
serve as input vectors of BP neural network. The analysis results from
gearbox vibration signals by EMD and BP neural network show that the
EMD-BPN can identify gearbox fault patterns accurately and effectively
and is superior to the combination of Wavelet packet with BPN and EMD.