Abstract: Based on the test data of a series of pump speeds and working conditions, a matching method based on the general characteristics of the hydraulic torque converter was proposed.This method was used to match engine with hydraulic torque converter.Firstly, error sources of the traditional matching method were analyzed.Secondly, this paper used the back propagation (BP) neural network, which is good at fitting and generalization, to establish a general characteristics prediction model of torque converter.The number of hidden layer nodes of neural network structure is determined.The comparison between the BP model and the traditional modification model indicated pronounced improvement with regard to performance prediction.Combined with the engine performance and the general characteristics matching model, it was found that the matching approach proposed in this paper was able to reflect the influence of pump speed on the matching conditions and improve the matching accuracy.