Rotating Machinery Fault Diagnosis Based on EEMD Time-Frequency Energy and SOM Neural Network

Abstract

This paper proposes a method of fault diagnosis for non-stationary fault signals of rotating machinery based on ensemble empirical mode decomposition (EEMD) time-frequency energy and a self-organizing map (SOM) neural network. The method uses EEMD to decompose the fault signal, obtaining an Hilbert–Huang transform time-frequency spectrum based on all the intrinsic mode functions. The time-frequency plane is then segmented into several equal blocks, where the fault feature vector is composed of the energy of each block. All of the feature vectors of the training samples are then put into the SOM neural network to train the network. The output layer is clustered into several regions, with each region corresponding to a fault. Finally, new samples are added to the trained SOM network so faults are recognized according to regions based on the location of the output neuron. Experimental results indicate that this method can eliminate the mode-mixing problem and low-frequency false components that exist with EMDresults. Diagnosis accuracy with the proposed method is higher than what can be achieved using EMD, and the diagnostic results also have high visibility.