Fault diagnosis of ship power system based on optimized BP neural network

  Objectives  In order to realize the fault diagnosis of a ship power system, this paper proposes a fault diagnosis method based on an optimized back-propagation (BP) neural network.

  Methods  First, a momentum/adaptive learning rate adjustment algorithm is used to overcome the defects of the BP neural network. The idea of a "small network cluster" is then adopted to construct a separate network for fault identification and diagnosis. Next, 450 groups of fault data generated from a ship power system simulation platform are used for neural network training. Finally, the fault diagnosis results are demonstrated via a fault case in which the speed of a feed water pump is abnormal.

  Results  Through training in fault data, the fault diagnosis accuracy level reaches more than 99%.

  Conclusions  The proposed fault diagnosis method based on an optimized BP neural network can accurately realize the fault diagnosis of ship power systems.