Statistical nature of breakdown sequences in a liquid insulant

Publisher Website
Google Scholar
Abstract
Liquid dielectrics have the advantage that, given effective discharge suppression, long sequences of measurements can be made on a single sample with minimal variation arising from spark damage. This means that the considerable variations that occur between samples do not have to be considered. Study of such sequences suggests that the statistical scatter is fundamental to the breakdown process, rather than just arising from random errors. As breakdown is a catastrophic, irreversible process, the distribution of results cannot be treated as a direct approximation to an a priori probability function, but the relationship is fairly simple. This relationship has been used to develop an analysis of breakdown under constant-stress and ramp-stress conditions. Statistical procedures allow the formative time to be separated from the total breakdown time. A statistical model has been developed for the ramp test, and this form of test is put forward as the only one which can cover the whole range of time and stress. Although a small rate-of-rise effect has to be taken into consideration, the smallest-value distribution is shown to apply to slow ramp-breakdown results. The fact that small changes in the test method produce considerable changes in the form of the distribution emphasises the importance of rigorous experimental control.

This publication has 1 reference indexed in Scilit: