Guide for the Statistical Analysis of Electrical Insulation Breakdown Data
|Publication Date:||23 September 2004|
Electrical insulation systems and materials may be tested using constant stress tests in which times to breakdown are measured for a number of test specimens, and progressive stress tests in which breakdown voltages may be measured. In either case, it will be found that a different result is obtained for each specimen and that, for given test conditions, the data obtained may be represented by a statistical distribution. This guide describes, with examples, statistical methods to analyze such data.
The purpose of this guide is to define statistical methods to analyze times to breakdown and breakdown voltage data obtained from electrical testing of solid insulating materials, for purposes including characterization of the system, comparison with another insulator system, and prediction of the probability of breakdown at given times or voltages.
Methods are given for analyzing complete data sets and also censored data sets in which not all the specimens broke down. The guide includes methods, with examples, for determining whether the data is a good fit to the distribution, graphical and computer-based techniques for estimating the most likely parameters of the distribution, computer-based techniques for estimating statistical confidence intervals, and techniques for comparing data sets and some case studies. The methods of analysis are fully described for the Weibull distribution. Some methods are also presented for the Gumbel and lognormal distributions. All the examples of computer-based techniques used in this guide may be downloaded from the following web site "http:// grouper.ieee.org/gro