Based on the wavelet neural network algorithm,the vibration signals of the automobile gearbox is analyzed from the time domain and the frequency domain,and the values with time domain and frequency domain characterization are extracted,and a method of BP neural network algorithm for the fault diagnosis of the automobile gearbox is designed. The fault of the gearbox is analyzed and extracted by empirical mode decomposition method in of the time-frequency domain. In addition,the method of BP neural network algorithm and the diagnosis models is erected. The JZQ-250 gearbox is used to train the algorithm of neural network algorithm to meet the needs of fault diagnose. The experimental results show that the BP neural network algorithm can reflect the correction of diagnosis under complicated fault of the automobile gearbox,which is of great practicality for fault diagnosis of the automobile gearbox. An idea and some experience for vehicle gearbox fault diagnosis in practice are provided.