Determination of voltage level from audible DC corona noise by using neural network

In this study, a signal recognition application is presented for classification of positive DC corona voltage level using recorded sound data of corona (electrical discharge) and utilizing probabilistic neural network (NN). The recorded positive DC corona sound data are acquired experimentally from a test set-up intentionally producing corona sound by applying different levels of positive DC high-voltage. Recordings have been used in training and test sets of the neural network. The main objective of this study is developing a model to determine source voltage level by only analyzing the recorded corona sound data. During the application of algorithmic method, linear prediction coefficients are used to pre-process the sound data for feature extraction. It is shown that reasonable results can be obtained by following the proposed method.