Abstract: The vibration signal of hydraulic machinery has the characteristics of strong noise, non-stationary, non-linearity, etc. The effective extraction and identification of the fault characteristic of the vibration signal has an important influence on the fault diagnosis. In this paper, two common vibration faults of misalignment and loosening of hydraulic machinery are simulated by test bench. The fault signal is extracted by variational mode decomposition and Hilbert envelope spectrum analysis. First, use the instantaneous frequency average of the components to determine the number K of variational modal components, decompose the signal, and then the envelope spectrum analysis of the decomposed and reconstructed signals is performed. Studies have shown that VMD can effectively extract the feature quantity of the signal, which is beneficial to the identification of the fault. The difference of the envelope spectrum characteristics of the reconstructed signal under different fault conditions and different degrees of the same fault is compared. It is found that the higher harmonic information is more prominent in the case of misalignment faults, and the higher harmonics increase with the increase of the degree of misalignment. The vibration caused by looseness fault is directional, and the vibration signal is accompanied by inter-harmonics. These vibration fault characteristics can provide reference for intelligent fault diagnosis.