データ同化を用いた熱分析データからの溶融金属流動物性の推定

抄録

Reliable prediction of macrosegregation requires accurate thermophysical properties that govern melt flow, particularly the thermal expansion coefficient β_T and kinematic viscosity ν. However, these properties are difficult to measure with sufficient accuracy. In this study, we develop a data assimilation approach that combines the modified lattice Boltzmann method (MLBGK) with a particle filter to estimate β_T and ν using only thermal analysis data. A two-dimensional system with a horizontal temperature gradient was considered, and twin experiments were performed to evaluate the feasibility of the method. The results show that both β_T and ν can be estimated with high accuracy. In single-parameter estimation, convergence to the true value was achieved rapidly for particle numbers N ≥ 200. Simultaneous estimation was also successful. The estimation accuracy depended on the number of measurement points and the filtering interval, and estimation of β_T is generally more accurate than ν. The proposed approach provides a practical means of determining flow-related properties of molten steel from thermal analysis measurements.