Getting the most out of historic reliability data

Abstract

In this paper, an intensive statistical analysis on the UK transmission power transformer historic reliability data is presented. It is shown that deriving the reliability of transformer at a certain age by simply calculating the hazard rate is inadequate, as the hazard rate in each age has a statistical range in which the confidence band width is related to the amount of the reliability data involved in that age. Since the post-mortem investigations on the failed units have verified that all failures up till now follow random failure mechanism, a general hazard rate of 0.27% with standard deviation of 0.03% has been derived for the UK transmission power transformers based on all the reliability data available. In term of data sufficiency, as supported by the hazard rate's confidence band analysis, only if the number of transformer at one age is more than 200 should the calculated hazard rate be reliable.