Data assimilation based on the ensemble Kalman filter for dislocation motion using dislocation dynamics simulation

Abstract Dislocation structures and their dynamic behavior have been widely observed by transmission electron microscopy (TEM), and effectively reproducing these phenomena in computer simulations is a powerful method for understanding their mechanisms from a multiscale perspective. Discrete dislocation dynamics (DDD) is a computational method of simulating the three-dimensional behavior of dislocations. However, it is difficult to reproduce dislocation motions observed in experiments through DDD simulations because of the presence of uncertain parameters. Data assimilation is a method of combining experimental observation and numerical simulation for a more accurate analysis of physical phenomena, providing insights into the mechanical properties altered during deformation and not accessible through experiments alone. In this study, we developed a data assimilation method for dislocation motion by DDD simulations. We conducted numerical tests using synthetic observation data instead of using actual experimental results. By numerical tests focused on the mobility of screw dislocations in DDD simulations, we validated the proposed method to be sufficient for data assimilation. The results of the numerical experiments indicate that the proposed method enables data assimilation using actual experimental observations such as TEM video data.