Determination of kinetic parameters of a phase-field model by a neural-network method
Korobeynikov S. A.
1, Korobeynikov D. M.
1,2, Lebedev V.G.
1, Lad’yanov V. I.
11Udmurt Federal Research Center, Ural Branch Russian Academy of Sciences, Izhevsk, Russia
2Udmurt State University, Izhevsk, Russia
Email: sa.korobeynikov@yandex.ru, dimakor-12@mail.ru, lvg@udman.ru, vilad@udman.ru
The inverse problem of restoring the kinetic parameters of the phase-field model of pure iron crystallization in an isothermal setting is considered herein. A neural network method is proposed that estimates the mobility and relaxation time parameters of the phase field based on the dependence of the front velocity on supercooling. The training is executed through an iterative cycle that aligns with the underlying physics of the process. The demonstration is made of the reproduction of target velocity curves for reference data from molecular dynamics. The proposed method has the potential to reduce the labor intensity of calibrating models that are transferable to related materials and other physical problem formulations. Keywords: phase-field method, neural network, pure element, parameter optimization.
- W.J. Boettinger, J.A. Warren, C. Beckermann, A. Karma, Annu. Rev. Mater. Res., 32 (1), 163 (2002). DOI: 10.1146/annurev.matsci.32.101901.155803
- V.G. Lebedev, A.A. Lebedeva, P.K. Galenko, JETP Lett., 101 (2), 136 (2015). DOI: 10.1134/S0021364015020101
- S.-L. Wang, R. Sekerka, A. Wheeler, B. Murray, S. Coriell, R. Braun, G. McFadden, Physica D, 69 (1-2), 189 (1993). DOI: 10.1016/0167-2789(93)90189-8
- S.R. de Groot, P. Mazur, Non-equilibrium thermodynamics (Dover Publ., N.Y., 1984)
- A.T. Dinsdale, Calphad, 15 (4), 317 (1991). DOI: 10.1016/0364-5916(91)90030-N
- V.G. Lebedev, A.A. Obukhov, V.P. Bovin, V.I. Ladyanov, Khimicheskaya fizika i mezoskopiya, 3 (23), 312 (2021) (in Russian). DOI: 10.15350/17270529.2021.3.28
- Y. Ashkenazy, R.S. Averback, Europhys. Lett., 79 (2), 26005 (2007). DOI: 10.1209/0295-5075/79/26005
Подсчитывается количество просмотров абстрактов ("html" на диаграммах) и полных версий статей ("pdf"). Просмотры с одинаковых IP-адресов засчитываются, если происходят с интервалом не менее 2-х часов.
Дата начала обработки статистических данных - 27 января 2016 г.