The research approach of engine fault diagnosis based on neural network and rough sets

Neural network is useful in fault diagnosis, but its training speed is slow. Based on the rules generated through Rough Sets, a new forward neural network with three layers is maintained. Considering the importance of the different fault attributes, the neural network, trained with BP algorithms, has directly Links between input layers and output layers. The application of the neural network to automobile engine fault diagnosis is also described. The result shows that the training speed of the new neural network has been increased and it is useful in iault diagnosis.