The statistical basis of impulse testing

The normal procedure of impulse-testing insulation involves the average of a series of similar tests and the determination of the percentage breakdown at a given impulse ratio. The basis of this procedure involves the properties of a series of Bernoulli trials, the outcome of each trial being either breakdown or non-breakdown of the insulation tested. These properties, which are of a statistical nature, are analysed, and the results expected for various types of insulation and impulse waveshape are given and compared with existing experimental data where possible. The probable outcome of a series of N impulse trials is shown to have a binomial distribution, and estimates of the errors involved are given. The effect of sample variation which applies particularly to tests on solid insulation is also treated briefly. The paper provides a statistical background against which the results of impulse voltage experiments may be judged.