The use of statistical techniques for the identification of partial discharges in electrical machines

The condition of insulating systems, the prediction of faults and problems are of major concern to machine manufacturers and machines users. They allow corrective maintenance and planning and reduce the overall maintenance costs. Up until now, manufacturers have undertaken a range of tests including partial discharge (PD) tests. Whilst these indicate that there are some faults in the insulation, they do not give details of the type of faults, and therefore fall short of satisfying the manufacturer´s needs. The introduction of computers and digital equipment has introduced a new era in PD measurement and analysis. This paper describes the application of statistical analysis of discharge signals in improving the recognition of PD sources