Determination of statistical distribution of water tree lengths: Monte Carlo simulation

Abstract

In this paper, results of Monte Carlo simulation, employed to determine the statistical distribution type for water tree lengths, are presented. The simulation involved generating log-normal and Weibull distributed populations from which left-truncated samples were drawn in the way resembling the process of water tree measurements performed with a finite resolution of 2.5 /spl mu/m. Then, distribution plots for truncated log-normal and truncated Weibull samples were compared, all in relation to the distribution plots typically obtained from microscopic measurements. The results obtained from the Monte Carlo simulation point to the log-normal distribution as a favorable statistical model, appropriate for representing a population of water tree lengths.