An estimation method (LOESS) for dealing with the imperfections of existing parametric creep models for time to failure: Illustrated using 2.25Cr–1Mo steel

It is important to be able to predict the creep life of materials used in power plants. This paper illustrates the inadequacies of existing parametric creep models in achieving this aim for 2.25Cr–1Mo steel. This breakdown occurs due to different creep mechanisms being present at different test conditions. This paper puts forward a locally weighted estimation technique (LOESS) to tackle this problem that importantly maintains the parametric model's structure by using local (to the test condition) weighted regression procedures. It was observed that this technique not only produces more sensible values for key parameters such as the activation energy, but it also produces much more accurate predictions of lifetimes close to operating conditions when using only short-term data (less than 10,000 h) to quantify the creep model's unknown parameters. The variation in the model's creep parameters with stress were also consistent with previous studies on the changing creep mechanisms for this material.