To identify different crack depth of the fault gear,taking the spur gears of the gear transmission system as the research object,three gears already set different crack depth and a gear of no crack in structures were installed in the signal acquisition test platform.The collected vibration signals were analyzed in timefrequency domain and EMD decomposition,and then the energy fault feature vectors with different dimensions were extracted.The SVM model was established based on the radial basis kernel function algorithm,and the recognition rate of gears with different crack depths was compared.The results show that the proper dimensionality of energy fault eigenvectors,combined with EMD signal decomposition and SVM classification method,can accurately identify the types of gears with different crack depths.It provides a reference for the early diagnosis of gear crack faults.