Testing for statistically significant differences in predictions obtained from competing creep models

M. Evans
DOI: https://doi.org/10.1177/02670836231223583
IF: 1.8
2024-01-08
Materials Science and Technology
Abstract:It is important to be able to predict the creep life and other creep properties of materials used in power plants and aeroengines. Whilst existing studies have compared different creep models using different measures of predictive performance, none have identified whether these differences are real from a statistical significance perspective. This paper proposes several tests based on a review of the literature. These tests were applied to two creep models using failure time data on 2.25Cr–1Mo steel to develop a recommended approach for practitioners to adopt when selecting a creep model. All such tests concluded that the Evans model produced better long-term life predictions using only short-term data, and that this difference was statistically significant at the 5% significance level.
materials science, multidisciplinary,metallurgy & metallurgical engineering
What problem does this paper attempt to address?