In this paper,first,a neural network model describing the exhaust temperature of a diesel engine sample is established based on the improved BP neural network algorithm.Next,some data of engine speed,engine power,fuel consumption and exhaust temperature are obtained from beach tests,which are then used to train the established model.Finally,an error analysis is performed to verify the model.The results indicate that the established neural network model well describes the variation of exhaust temperature,and that the errors of the identification results,which are all less than 1%,meet the requirements of calculation.In addition,the intelligent temperature control of exhaust gas recirculation（EGR） is realized by combining the BP neural network model with the fuzzy inference.