Abstract: Aiming at the nonstationary characteristics of the fault vibration signal of high pressure diaphragm pump check valve, a fault diagnosis method based on local meandecomposition (LMD) and grey correlation theory is proposed. Firstly, LMD is used to decompose the vibration signals of the check valve of high pressure diaphragm pump under different operating conditions into multiple product functions. Then the cross-correlation coefficients of all product functions are calculated, and the product functions with high cross-correlation degree are selected to extract the corresponding feature vectors. Finally, the grey correlation theory is introduced to construct the fault diagnosis and identification model of one-way valve, and the results are compared with those obtained by empirical mode decomposition (EMD), mixed feature modeling and single feature modeling. The experimental results show that the proposed method can decompose the fault signal of one-way valve well, and effectively identify the fault signal of one-way valve under the condition of small samples.