Method research of fault feature extraction based on AGA-RS
LI Hua-yin · 2014
In fault diagnosis, compressing high dimension feature space to low dimension feature space can simplifiy the design of fault classifier and improve computational efficiency. Adaptive genetic algorithm and rough sets theory are researched for feature selection and reduction, the brief features are abstracted according to the faults of diesel fuel injection system, and the neural network model is built. Experimental results indicate the method can extract effective and brief features, but also improve the capacity of neural network.