Abstract: Aiming at the non-stationary and non-linear characteristics of vibration signal of check valve in high pressure diaphragm pump, a fault diagnosis method based on local mean decomposition (LMD), permutation entropy and artificial bee colony (ABC) optimization kernel extreme learning machine (KELM) is proposed. The vibration signal is decomposed into multiple component signals by LMD, and then the component signal with higher correlation degree is selected by cross-correlation criterion, and the corresponding permutation entropy is calculated as the eigenvector. The fault diagnosis model is built in KELM optimized by artificial bee colony algorithm. Through the processing and analysis of different fault state signals collected under the actual working conditions, the results show that the fault diagnosis method for the operation status of one-way valve can not only better represent the signal state information, but the fault recognition accuracy also reaches 95.65%. At the same time, compared with traditional KEML and ELM, it has higher recognition accuracy.

Key words: check valve, local mean decomposition, permutation entropy, artificial bee colony algorithm, fault diagnosis