The application of a new and robust weighting scheme for establishing theta projection parameters to 1CrMoV rotor steel

Abstract

Reservations about the need to weight the estimated theta parameters within the theta projection technique have often been expressed. However, the fact that the theta values are estimated values and are therefore not certain quantities necessitates the need to weight each theta value by some measure of its reliability. Traditionally, the weights have been calculated under the assumption that the random errors around an experimental creep curve are first-order autocorrelated. Using short-term data collected at the Interdisciplinary Research Laboratories in Swansea for 1CrMoV steels, this paper shows that such an assumption is generally incorrect and presents an alternative estimator for these weights. This estimator is robust in that it requires no assumptions to be made about the nature of the autocorrelation. When applied to these short-term creep data these robust weights are well behaved and alter the theta interpolation/extrapolation functions in such a way that excellent predictions are made for the longer-term rupture times published by the National Research Institute of Metals in Japan. Failure to use a weighting scheme within the theta projection technique is shown to have a detrimental impact on such long-term predictions.