Bearing Fault Diagnosis Based on Extreme Machine Learning Optimized by Differential Evolution

In order to improve the accuracy of bearing fault diagnosis, a novel method based on multi-masking empirical mode decomposition (MMEMD) and extreme machine learning optimized by differential evolution algorithm (DE_ELM) is proposed. MMEMD is a new improved empirical mode decomposition (EMD), which can overcome the defect of mode mixing and improve the feature effectiveness. Differential evolution algorithm is used for determining the parameters of extreme machine learning (ELM) to improve the classification accuracy. In implementation of the proposed method, firstly, bearing signals are decomposed into different intrinsic mode functions (IMF) and sample entropy of each IMF is calculated as the fault feature. Then, the training set is input to the DE_ELM and the fault classification model is obtained. Finally, the testing set is input to the model for fault diagnosis. The proposed method examined by the bearing fault diagnosis experiment. The results show that the method can reliably identify the different faults and has a high fault diagnosis accuracy.