Damage localization in hydraulic turbine blades using kernel-independent component analysis and support vector machines

A hydraulic turbine runner has a complex structure, and traditional source location methods do not have the higher accuracy to meet engineering requirements. The source location of crack acoustic emission (AE) signals in hydraulic turbine blades has been researched by combining it with kernel-independent component analysis (KICA) as feature extraction, with support vector machines (SVMs) as position recognition. This method is compared with those applied SVMs with feature extraction using kernel principal components analysis without feature extraction. The results show that the recognition rate in the crack region is 100 per cent by using both original AE parameters and feature parameters. Support vector regression by feature extraction using KICA can perform better than the other methods. As a result, it is a better method for source location of complex big size structures to combine KICA with SVM. It decreases the dimensionality of input signals and also improves the accuracy of location.