A procedure is described for the application of fuzzy logic systems to the
classification of partial discharge pulse patterns in terms of cavity or vo
id size. The features employed in the pattern recognition task are those re
lated to the form or shape of the partial discharge pulses and their associ
ated apparent charge transfer Preliminary results, obtained with the fuzzy
logic system on simple partial discharge sources, indicate a performance ap
proaching that attainable with artificial neural networks.