Neural networks prediction of ionic mobilities in SF6-N2 mixture

Abstract

The present work outlines the application of neural networks in the modelling and the Prediction of ionic mobility (μ) in SF6-N2 gas mixture using experimental data. At higher pressures, the mobility ? measured with conventional models is inversely proportional to the gas density (N-1). Experimental data of ionic mobilities for N2+SF6 have been obtained previously by the use indirect method, which, consists of measuring the voltage-current characteristics of corona discharges. The results obtained by prediction are significantly consistent with those measured experimentally. The average relative errors on predicted ionic mobility are found to be less than 10% for training as well as for testing. Since the average errors are less than 10%, the proposed ANNs technique is highly recommended for the prediction of ionic mobilities of corona discharges in N2+SF6 gas mixtures.