Jixie chuandong (Jan 2018)
Fault Diagnosis of Rolling Bearing based on MEEMD-DHENN
Abstract
According to the non-stationary and nonlinearity of the fault signals from rolling bearing,a recognition method based on Modified Ensemble Empirical Mode Decomposition（ MEEMD） and Double Hiddenlayer Elman Neural Network（ DHENN） is proposed. The vibration signal is decomposed by MEEMD,and using variance contribution to select out the Principal Intrinsic Mode Function（ PIMF）,one can calculate the permutation entropy of the IMF which obtained by using Complementary Ensemble Empirical Mode Decomposition（ CEEMD）,it is basic to deal with the illusive component,and the others are reconstructed and decomposed by using EMD,the variance contribution is calculated,through Hilbert transform,characteristic matrix is constructed. The DHENN network model is constructed to recognize the fault type,getting the best combine of the double hidden neural network. Finally,compare EMD-DHENN and MEEMD-ENN,the result show that the proposed MEEMD-DHENN method has a high accuracy rate of 100% and only need 26 steps.
