Use of a mixed-Weibull distribution for the identification of PD phenomena [discussion and reply]

摘要

The authors of the original paper [see ibid., vol. 2, no. 4, 1995] propose to use a Weibull statistical analysis to identify when two separate PD mechanisms are active simultaneously. Five parameters are needed to be estimated using a complex iterative technique. Their paper raises the following question. In the many thousands of pulse height analysis (PHA) plots I have collected, it has normally been very easy to determine that two (or more) mechanisms are occurring by simply looking at the plots. The author shows such straightforward examples. In fact observation shows it is easier to see the two mechanisms and their demarcations with the PHA plots than the Weibull plots. This makes sense to me since the Weibull transformation fundamentally involves a double logarithm transformation which tends to reduce differences. Furthermore, straightforward regression analysis techniques are readily available to separate 'objectively' the two mechanisms from the PHA. In addition, there are now many pattern recognition methods available. Thus why do the authors propose that we should go through the complicated process of transforming the data to Weibull format and using more complicated analytical procedures? A reply is given.