Abstract:
In order to further improve the fault prediction accuracy and maintenance efficiency of switch machine, a fault prediction method of switch machine based on wavelet energy spectrum entropy and improved Extreme Learning Machine (ELM) is proposed. Firstly, the collected power data of switch machine is preprocessed with Complementary Ensemble Empirical Mode Decomposition (CEEMD) method. Then the wavelet energy spectrum entropy of each IMF component is calculated, the multidimensional characteristic data is reduced to one dimension by KPCA method, and the degradation performance index of switch machine is constructed, and the failure threshold is obtained. Finally, the weight and threshold of ELM prediction model are calculated by Adaptive Whale Optimization Algorithm (AWOA) in order to obtain the optimal prediction model, the fault trend of switch machine can be predicted. Matlab software is used to analyze the method, and compared with Support Vector Machine (SVM) and ELM model. The simulation results show that, compared with the traditional prediction model, the improved ELM model has higher prediction accuracy and better performance, which proves the feasibility of applying this method to the fault prediction of switch machine, and provides theoretical support for the realization of intelligent fault prediction of switch machine and on-site maintenance.