Improvements in AEPD location identification by removing outliers and post processing

Abstract

The mathematical model of an Acoustic Emission Partial Discharge (AEPD) system is solved in the literature using Newton's method with redundant number of sensors (more than 4; eight in this case). The system for numerical experiments consists of eight sensors. The algorithm is implemented using three different initial guesses. For the calculated PD source coordinates, histograms are plotted. After finding the mean and standard deviation, coordinate values which are lying outside different fractions of sigma are removed. The average of remaining set is calculated and it is found that, the accuracy of location identification can be greatly improved.