Development and Application of Minimum Creep Strain Rate Metamodeling
Abstract Numerous minimum-creep-strain-rate laws exist, creating a challenge in determining which is best for a given material database. The objective of this study is to validate the applicability of a “metamodel” and its ability to model the minimum creep strain rate (MCR). A metamodel is a model that can combine and regress into different base models, in this case, seven established MCR models. The metamodel can be exploited using a calibration algorithm to rapidly calibrate the base models. The metamodel contains ten terms and eight material constants (one is a constraint, and another is stress as an input variable). Using the metamodel and calibration software, the user can determine the best MCR model for a given material database. Using the software, the metamodel is calibrated in two approaches: constrained and pseudo-constrained. The constrained approach restricts the metamodel to regress directly into one of the base models, allowing for the base models to be equally calibrated and compared alongside each other. The pseudo-constrained approach freely optimizes all eight of the metamodel material constants; however, the metamodel is modified to include 5 Heaviside function constants that turn on/off sections of the metamodel to increase the statistical-dependencies of the final model. This pseudo-constrained approach has the potential to identify novel MCR models that exist at the interface between the seven base models. Alloy data for 9Cr-1Mo-V-Nb (ASTM P91) was used with a total of 89 points which extended over a total of three isotherms: 600°C, 625°C, and 650°C. The MCR model that best fit the data was the Johnson-Henderson-Kahn model.