New method of bearing fault diagnosis based on mmemd and DE_ELM

Abstract

A new method of bearing fault diagnosis based on multi-masking empirical mode decomposition (MMEMD) and extreme machine learning optimised by differential evolution algorithm (DE_ELM) is proposed in this study. MMEMD is an improvement of empirical mode decomposition (EMD). By adding masking signals to the signals to be decomposed in different levels, MMEMD can restrain low-frequency components from mixing in high-frequency components effectively in the sifting process and then suppress the mode mixing. Differential evolution algorithm is applied to determine the parameters of ELM for improving the classification accuracy. The four parameters are determined at one time by uniformly coded as the individuals of the differential evolution algorithm. To achieve the bearing fault diagnosis, the fault signals are first decomposed into different intrinsic mode functions (IMFs) and the sample entropy of each IMF was calculated as the fault feature. Then the fault feature was divided into training set and testing set. Input the training set to the DE_ELM to obtain the fault classification model. Finally, the testing set was put into the model for fault diagnosis. The experiment and wind turbine bearing fault diagnosis results show that the method could identify the different bearing faults with high reliability and accuracy.

Keywords

• machine bearings
• hilbert transforms
• wind turbines
• mechanical engineering computing
• fault diagnosis
• learning (artificial intelligence)
• entropy
• evolutionary computation
• feedforward neural nets
• signal processing
• pattern classification
• mmemd
• differential evolution algorithm
• de_elm
• masking signals
• low-frequency components
• high-frequency components
• fault signals
• fault feature
• fault classification model
• intrinsic mode functions
• multimasking empirical mode decomposition
• extreme machine learning
• sample entropy
• wind turbine bearing fault diagnosis