Abstract
To explore the deficiency of the traditional neural network in fault diagnosis, a combination of fuzzy theory and neural network based on improved BP algorithm was proposed, and used in the fault diagnosis of automatic transmissions. Through the estabLishment of the common failure knowledge base, fuzzy theory was used to process the fault information and to obtain the neural network training samples. With the simulation by Matlab software, the result shows the method can effectively overcome the deficiency of standard BP algorithm, and provides efficient way for the fault diagnosis of automatic transmissions.