Comparison of the Weibull and the Crow-AMSAA Model in Prediction of Early Cable Joint Failures

Abstract

This paper compares the application of the Weibull distribution and the Crow-AMSAA (C-A) model to the analysis of cable joint failures. The procedures on how to use the two models to analyze failure data and to predict the future number of failures have been described before the models are applied to a set of early failure data. The data, which include 16 failures and 1126 suspensions, were collected from a regional power-supply company in China. This paper proves that the Weibull distribution provides more reliable results in the analysis of early-failure data since it considers the time to failure of each event. Whilst the method is more straightforward and requires less information, applying the C-A model can yield confusing results if not handled carefully. When analyzing the events, using time as x-axis and using the cable joint population as an axis can complement each other. Separating those data into subsections and analyzing them independently can yield useful information. Recent failure data can better reflect the current state of cable joints. The results of this paper should help utility asset managers better analyze their past failure events.