Application of improved DBD algorithm based bp neural network on fault diagnosis for fuel supply system in a certain diesel engine

Abstract

In order to overcome the drawbacks of a neural network based on back propagation (BP) algorithm, such as too slow to converge and easy to be trapped into a local minimum, a new modified algorithm is proposed in this paper, in which the grads information of the network are exchanged dynamically in each iteration step, and the increment factor of learning rate and interaction function in delta-bar-delta (DBD) algorithm are improved based on the idea of cross and mutation in Genetic algorithm (GA). The new algorithm has been applied in the fault diagnosis of a fuel supply system in a certain diesel engine successfully.