b'details of the microstructure thereby facilitating decision making (e.g., material acceptance,This project will allow for a rigorous assessment of the design of new experiments). performance of components under complex loading scenarios as a function of microstructure of metals. It could Accomplishments: pave the way for optimal design of experiments.The starting point was a proof-of-concept creep surrogate model (Los Alamos Reduced Order Model for Advanced Non-linear Constitutiveeach tile separately. Furthermore,of the future work and develop-Equations [LAROMance]) for HT9.a weighted least squares regressionment of LAROMance in general This model needed significantmethod has been implementedand shall transfer to a possible improvements in terms of stability,to allow for emphasizing certainnext version of the HT9 surrogate range of utilization (i.e., stress,parts of the model. In this work, themodel. Indeed, some regions of the temperatures) and overall accuracy.transient and steady-state regimesmodel seem to overestimate the In FY23, significant changes wereare autonomously detected by thecreep response, which may yield to made to the surrogate model toalgorithm, which allows to adjustconservatism in decision making address the aforementioned needs.the weight of either regime duringrelated to safety criteria when this Specifically, a new version of thethe calibration process. Further, themodel is used for lifetime predic-creep surrogate model (LARO- metrics for model quality in termstions of engineering structures.Mance) for HT9 has been calibratedhave also been revisited and distin-from an extended synthetic databaseguish between the transient and the of the materials response. Thesteady state regime of the simulated database comprises long-termcreep response. This enables a more mean-field polycrystal simula- thorough analysis of the surrogate tions. As an improvement to themodel, which is useful for further first surrogate model, the databasedevelopments.has been extended in the stressThese additions to the framework domain, temperature domain, asallow for a more tuneable surrogate well as the temporal domain (longermodel to be easily adjusted to the simulations). A total of ~100 millionspecific problem at hand. Overall, observations are now includedthe updated surrogate model is in the calibration process of thebetter-behaved, in particular for surrogate model. longer-term simulations, both in The HT9 surrogate model formterms of continuity/smoothness and has been revised, so that onlyaccuracy, than its predecessor. The Legendre polynomials up to thesurrogate model shall be made avail-linear degree have been includedable in the framework of MOOSE in the calibration process, andas part of the LAROMance family the input/output transformationof creep surrogate models. Further mappings have been optimized forimprovements to the model are part 2023|AFC ACCOMPLISHMENTS 155'