Adaptive goal-oriented data sampling in Data-Driven Computational Mechanics

Data-Driven (DD) computing is an emerging field of Computational Mechanics, motivated by recent technological advances in experimental measurements, the development highly predictive computational models, data storage and processing, which enable transition from a material data-scarce to data-rich era. The capability DD simulations contingent on quality set, i.e. its ability closely sample all strain–stress states phase space given mechanical problem. In this study, we develop methodology for increasing existing set through iterative expansions. Leveraging formulation problems treated with paradigm as distance minimization problems, identify regions poor coverage, target them additional experiments or lower-scale simulations. solution informs so that they can provide better coverage application. We first illustrate convergence properties approach finite element simulation linear elastic cylinder under triaxial compression. same numerical experiment then performed specimen Hostun sand, complex history-dependent behavior. Data sampling Level-Set Discrete Element Method (LS-DEM) calculations unit cells representative granular material, subjected loading paths determined proposed method. It shown adaptive expansion tailored particular application, leads convergent accurate predictions, without cost using large databases potentially redundant low-quality data.