A rolling bearing fault diagnosis approach was proposed based on the adaptive multiscale fuzzy entropy

ILS (iterative Laplacian score) and PSO-SVM (particle swarm algorithm optimization support vector machine). In the proposed method

the variational mode decomposition was used for the decomposition and reconstruction. Then composite multiscale entropy fuzzy of the reconstructed signals was calculated. Besides

the iterative Laplacian score algorithm was used for the sensitive fault feature selection and the selected features were input to the PSO-SVM based classifier for training and recognition. Finally

the proposed method was applied to experiment data of rolling bearing. Results showed that the identifying rate of proposed method was 100%. Also

the ILS based feature selection was compared with the SFS (sequential forward selection) method; and the result indicated that the highest identifying fault rate of SFS based method was 92.86% while the identifying fault rate of the ILS based fault diagnosis method reached to 100%.