Statistical scatter and variability of creep property estimates in θ projection method

Abstract

The prediction of long term creep properties from short term tests is discussed in terms of the θ projection method. The method has been shown to yield reasonable extrapolation functions and the present paper investigates the characterisation of creep scatter and the procedures required to give estimates of variability of creep predictions. The nature of errors in individual creep curves and in multiple curves obtained from various specimens under differing test conditions are discussed and formulae are developed that allow an assessment of the reliability of derived parameters. These methods are then extended to estimated creep properties (both interpolated and extrapolated) so that the variances of predicted quantities can be calculated. The analysis is carried out for general creep curve functions, but is illustrated with reference to the double exponential description of creep strain ξ with time: ξ=θ₁(1−e^−θ₂t) + θ₃(e^θ₄t − 1). Numerical examples relating to the creep properties of IN 100 are given and the significance of the procedures in relation to design for creep is discussed.

MST/988