The bearing fault diagnosis method of VMD-KELM based on parameter adaptive optimization

In order to solve the problems of the difficulties in feature extraction and low recognition accuracy for rolling bearing fault signals, a bearing fault diagnosis method based on Subtraction-Average-Based Optimize (SABO) optimizing Variational Mode Decomposition (VMD) parameters and using Kernel Limit Learning Machine (KELM) for fault classification is proposed. Firstly, a mathematical behavior based subtractive average search strategy is adopted and using minimum envelope entropy as fitness value to adaptively optimize the parameters of variational modal decomposition VMD, obtaining the number of modal components Κ and penalty factor α the best combination. Then, take the optimized values [Κ, α] of the parameters and the index values of the minimum envelope entropy fitness back into VMD to get the most suitable IMF. Meanwhile extract the 9 time-domain indicator features from the IMF to construct fault feature vectors. Finally, the KELM rolling bearing fault diagnosis classifier is be established. The effectiveness of the algorithm was verified using vibration data of the rolling bearing.