Application of Particle Swarm Optimizer Neural Network in Fault Diagnosis for Marine Auxiliary Boiler

A mixed neural network diagnose model is constructed by integrating the self-organizing feature map network with a BP neural network in cascade.Meantime, the particle swarm optimizer is used to improve the initial connecting weighting and node dependent thresholding.An ALFA LAVAL D-type water tube boiler is taken as the objective machine for verification of the diagnosis model.The diagnosis model is tested with the data from the DMS VLCC marine engine room simulator.Same experiments are carried out with the sole BP neural network and the SOM-BP network for comparison.The advantage of the development is demonstrated.