Accuracy-improved bearing fault diagnosis method based on AVMD theory and AWPSO-ELM model

Abstract

In order to further strengthen the performance of vibration signal decomposition and fault diagnosis accuracy, an accuracy-improved bearing fault diagnosis method based on adaptive parameter optimized variational mode decomposition (AVMD) theory and extreme learning machine optimized by adaptive weight particle swarm optimization (AWPSO-ELM) model is put forward. First, spectrum degree of cross-correlation (SPC) is introduced to help select the optimal penalty factor in light of the modal aliasing and information integrity. Meanwhile, reconstruction precision (RCP) is introduced to warrant the precision between reconstructed signal and the original. Its validity is verified on a simulated vibration signal. Then oscillation energy is extracted as the fault feature and fed into the AWPSO-ELM model with fault type as its output. The experiment shows that the fault feature extracted via AVMD method are more prominent than those extracted via other methods and AWPSO-ELM model reaches an accuracy of 100% on the testing dataset.