The Applied Research of System Subtractive Clustering RBF Algorithm in Eddy Current Testing on the Conductive Structure Defect

According the problem of defect type discrimination and quantitative detection of defect depth in eddy current testing (ECT) on the conductive structure defect. The text proposes a RBF optimization algorithm based on system subtractive clustering (SISCA),first of all, according to the likelihood of the data ,it uses the system clustering method to estimate the number of clustering center, and improves mathematical model of subtractive clustering to determine the clustering scheme, then takes the minimum of the largest distance variance in the cluster as evaluation index to obtain the optimal clustering information to provide the critical initial value for training RBF network. The facts proves improved RBF algorithm takes higher accuracy and more effectiveness in the experiment of ECT.