Artificial intelligence algorithms for the recognition of PD-generating defects in rotating machines

This paper presents an innovative technique for partial discharge source identification in rotating machines. Source identification is based on artificial intelligence tools that rely upon fuzzy inference algorithms and operate through a tree-like structure. The third level of the tree-like structure, specific for rotating machine defects, is described in this paper with emphasis on bar-to-bar discharges. Its effectiveness in the identification of defects generating PD, even in the presence of noise or overlapped PD phenomena, is discussed on the basis of lab tests performed on defective generator bars.