BP neural networks technology was applied to the fault diagnosis of railway vehicle gearbox, and BP neural networksdiagnose model was created. An improved BP algorithm was put forward, and using the typical fault signal extracted from the routinediagnostic methods as input data of the neural network, simulation and testing in terms of the improved BP neural network were carried out.The results showed that the improved BP neural network system had the better performance of fault diagnosis and fault identification, andcould achieve the anticipated diagnosis result.