Abstract
Considering the difficulties in constructing physical models to simulate fuel systems of gas turbines, we introduce an improved approach based on SOM neural network for the fault diagnosis of fuel systems. In this model the competitive layer structure is decided by an objective function, the parameter functions are selected with genetic algorithm, and the weight vectors are initialized with a pre-segmentation method. Meanwhile, before the data is inputted, PCA dimensional reduction is used to decrease the training consumption. Eventually, practical dataset verification suggests that this improved SOM neural network performs better in recognition rate than other classification algorithms and original SOM network in fault diagnosis of gas turbine fuel system.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call