The PD6605 Creep Rupture Data Assessment Procedure-An Appraisal of its Application 10 Years On

By far the greatest testing effort, and therefore expense within the creep community is devoted to characterising and understanding the high temperature properties of steels and nickel base alloys for power engineering and aerospace applications. The most common test procedure - stress rupture testing - has remained essentially unchanged for many years, yet there is still no international consensus on the procedure for assessing results. Potentially, different strength values could be reported in the product standards from different standards bodies. The BS PD6605 assessment procedure issued over 10 years ago, adopts a statistical framework to select a model, but allows an element of user intervention to permit metallurgical judgment. It is an "open" procedure, allowing user-defined models (several have been developed since its issue); but importantly it adopts the ECCC Post Assessment Tests to ensure physical realism, goodness of fit, and repeatability and stability of extrapolations. Moreover, in its inclusion of unfailed test data, and statistical treatment of variations in data scatter, it offers several advantages over previous least squares fitting methods. Many of the concepts involved, however, are unfamiliar to potential users, which limits its exploitation. Addressing this point, we examine some of the consequences of using the Maximum Likelihood Estimation method, coupled with the Weibull and log-logistic probability distributions, to fit representative data sets; contrasting the results with those obtained from linear regression methods and the normal probability distribution. We consider how the shape of the Weibull probability distribution dictates that the median, rather than the mean, is appropriate for the model equation, and the effect this has on rupture strength values. Furthermore, we demonstrate the consequences of unfailed test data on the model equation, and its extrapolation. Finally, we consider the changing requirements for data assessment since PD6605 was originally issued. There is now a much greater interest in the creep properties for design and mechanical integrity purposes, both creep strength data and constitutive models. Principles for adapting the PD6605 procedure for such requirements are suggested.