patterns recognition and classification require an
understanding of the traits commonly associated with the
different source and relationship between observed PD
activity and responsible defect sources. This paper shows
the performance of SOM using different competitive
learning algorithms to classify measured PD activities into
underlaying insulation defects or source that generate
PD’s, its showed that WTA is the better algorithm with
less error and training time, but its overall performance are
not always satisfactory, being alternative in accord at the
performance FSCL or RPCL algorithms