A method is presented for detecting and identifying
the shape and size of a contaminating metallic particle in gas
insulated system (GIS). A fixed particle–initiated negative corona
in air insulated co-axial cylindrical configuration is investigated at
a voltage slightly above the corona onset level. The characteristics
of the negative corona pulses (Trichel pulses) are calculated by
mathematical modeling the process taking place during the
negative corona discharge. An experimental set-up is built up to
measure the Trichel pulse characteristics and to check the
accuracy of the present calculation. The calculated values of
corona Trichel pulse amplitudes and repetition rates agree well
with the measured values. So, the Trichel pulse characteristics are
calculated for different particle sizes and shapes. These data of
pulses are used as a bias for designing and training the artificial
neural network technique (ANN). For earlier detection of
insulation defect, the Trichel pulse characteristics are measured
and then given as an input data to the trained ANN. Then, we
could identify the particle shape and size.