Batch sequential design of optimal experiments for improved predictive maturity in physics-based modeling

The focus of nuclear fuel design and maintenance has been shifting from a primarily empirical endeavor to that of highly advanced simulations characterized by experiment-based calibration, validation and uncertainty quantification. The experimental data available for calibration and validation, however, is limited by the availability of resources. This limitation poses difficulties especially if the model is to be executed to predict at different settings and/or regimes within a domain. To assure that the model exhibits predictive maturity throughout the domain of applicability, a sufficient number of experiments must be conducted to explore this domain. Given the limited resources, it is therefore crucial to design validation experiments to maximize the improvement in the predictive capability of the physics-based numerical models. This article contributes to the recent developments in the optimal design of validation experiments by evaluating the performance of several experiment selection criteria that specify the specific benefits desired from future experiments. Our focus is on the Batch Sequential Design methods, which for a given set of initial experiments and a selection criteria, iteratively select a batch of future experiments. The performance of various selection criteria in improving model predictiveness are compared considering not only the empirically identified model discrepancy, but also the coverage of the domain of applicability. The manuscript provides an extensive simulation-based study on a polycrystal plasticity material model, utilizing an established index that quantifies predictive maturity of numerical models.