The auto-regressive(AR) model of exhaust noise in diesel engine’s seven operating modes about the valve train was established. The auto-regressive parameters were regarded as the characteristic vectors to build the fault diagnosis model which was based on extreme learning machine. The proposed method was also compared with that of BP neural network, RBF neural network and support vector machine. Experimental results show that exhaust noise can be used for fault diagnosis in diesel engine’s value train. The classification accuracies of above four different intelligent methods are all higher than 95%. But the method of extreme learning machine has advantage apparently in training rate.