Neural networks for breakdown voltage estimation of various gas mixtures

In this paper, AC breakdown strengths of a mixture of 99.875% CO ₂+0.125% SF₆ and those of N2+SF6 mixtures containing 0.125, 0.5, 1 % of SF₆ in non-uniform field are studied. The relative gas pressure and the electrode gap spacing are varied within the range of 100-500 kPa and of 5-15 mm, respectively. The results are first measured experimentally and then estimated by means of Feedforward Neural Network Approach. The comparisons of measured and computed values show that there is a good agreement between two values. The breakdown voltages of the mixtures can be found correctly by the Feedforward Neural Network (FNN) approach. Therefore, the Feed-Forward Neural Network Approach can be considered an alternative tool to estimate the new values in or out of the measurement range.