In order to solve the fault recognition problems caused by a complex signal transfer path,severe noise pollution and the weak fault features,a method for gear based on variational mode decomposition（ VMD）and adaptive neuro-fuzzy inference system（ ANFIS） is proposed. Firstly,VMD is used to decompose a fault signal into several intrinsic mode functions（ IMFs）,and introduced the permutation entropy to construct the feature vectors characterizing the modal component information. Then the extracted feature vectors are input into the adaptive neuro-fuzzy inference system to establish the fault diagnosis model. Finally,the model is validated by the vibration signal data of the gears and compared with the support vector machine（ SVM） method. The results show that the proposed method has a strong learning ability and it can effectively diagnose the gear fault,improve the accuracy of fault identification. The recognition effect is obviously superior to SVM.