Particle filter-based phase-field data assimilation for estimating material parameters in solidification

Data assimilation by integrating experimental data with phase-field simulations is a valuable approach for inferring the material properties required for dendrite growth simulations. However, the perturbation of the phase-field profile is an inherent issue when employing the ensemble Kalman filter (EnKF) for data assimilation. In this study, we developed a phase-field data assimilation system utilizing a particle filter (PF) that avoids perturbing the phase-field profile. The performance of this system was evaluated via twin experiments and compared with that of EnKF. Our results demonstrate that the PF enables highly accurate material property estimation. Additionally, compared to EnKF, the PF enables larger filtering intervals, although it necessitates a greater number of particles. Future studies will focus on enhancing the efficiency and accuracy of the data assimilation system by combining the strengths of both PF and EnKF.