Title :
Classification of partial discharge signals by means of auto-correlation function evaluation
Author :
Contin, A. ; Pastore, S.
Author_Institution :
Trieste Univ.
Abstract :
A new algorithm for the separation of partial discharge (PD) signals due to multiple sources is presented in this paper. It evaluates the similarity of the signals shape by comparing their auto-correlation functions (ACFs), on the assumption that the same PD source can exhibit signals having similar ACFs. The new classification algorithm is based on a modified K-mean clustering (KMC) method. A laboratory test of the proposed algorithm is reported. The proposed classification method may constitute a step forward in the automatic signal separation
Keywords :
correlation methods; partial discharge measurement; pattern clustering; signal classification; signal sources; source separation; statistical analysis; K-mean clustering algorithm; PD source; auto-correlation function; automatic signal separation; partial discharge measurement; signal classification; Autocorrelation; Classification algorithms; Clustering algorithms; Laboratories; Noise shaping; Partial discharges; Pulse measurements; Shape; Signal processing; Testing;
Conference_Titel :
Electrical Insulation, 2006. Conference Record of the 2006 IEEE International Symposium on
Conference_Location :
Toronto, Ont.
Print_ISBN :
1-4244-0333-2
DOI :
10.1109/ELINSL.2006.1665317