Recently, numerical simulation has attracted attentions as a promising tool that enables systematic and quantitative evaluations of material microstructures. In particular, phase-field (PF) model is becoming mainstream for predicting microstructural evolution processes in continuum scale. To achieve the accurate prediction of microstructure formations based on the PF model, it is essential to account for the anisotropies in grain boundary energy and mobility. However, their inclination-dependent variations are still not clear. As a numerical method to evaluate grain boundary properties, molecular dynamics (MD)-based atomistic calculations are widely performed. However, evaluating the inclination-dependent boundary properties using MD is extremely laborious because it requires a large number of calculations for grain boundaries with different inclinations.

To resolve the above problem, this study focuses on data assimilation (DA), a methodology to estimate unobservable states and/or parameters involved in numerical models by integrating observation data into the models. we propose a DA system that can estimate grain boundary properties simultaneously for various inclinations via the assimilation of MD and PF simulations. Furthermore, the usefulness of proposed DA system is demonstrated by testing the accuracy of the estimated grain boundary properties. By assimilating MD grain boundary migration simulation into a PF model, we succeeded in an accurate PF simulation that reproduces MD-simulated grain boundary shapes. Through this DA calculation, inclination-dependent grain boundary properties are also estimated.