Abstract:Aiming at the difficulty of fault feature extraction and fault identification of rolling bearing,a fault diagnosis method for rolling bearings based on local mean decomposition (LMD) approximate entropy and improved particle swarm optimization based extreme learning machine (PSO-ELM) was proposed.The signals under different working conditions were decomposed into a series of multiplicative components by LMD.The approximate entropy values of the signal under different working conditions changed in different frequency bands.Combined with the correlation coefficient,the first three components were selected,and the approximate entropy value was calculated as the input eigenvector.Aiming at the shortcomings of premature convergence of PSO,the adaptive weight method and DE algorithm were introduced to improve the PSO,and the eigenvalues were input into the improved PSO-ELM network model,the fault identification and classification was performed for different working conditions of rolling bearings.The results show that the ELM based on LMD approximate entropy and improved particle swarm optimization can not only identify the fault types of rolling bearings,but also have a higher classification accuracy,which verifies the feasibility of the method.