Reduction of discretisation-induced anisotropy in the phase-field modelling of dendritic growth by meshless approach
| Autor: | Božidar Šarler, Boštjan Mavrič, Tadej Dobravec |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Physics
General Computer Science Basis (linear algebra) Discretization Mathematical analysis Finite difference General Physics and Astronomy 02 engineering and technology General Chemistry 010402 general chemistry 021001 nanoscience & nanotechnology 01 natural sciences 0104 chemical sciences Polyharmonic spline Computational Mathematics Dendrite (crystal) Mechanics of Materials Orientation (geometry) General Materials Science Radial basis function 0210 nano-technology Anisotropy |
| Zdroj: | Computational Materials Science. 172:109166 |
| ISSN: | 0927-0256 |
| DOI: | 10.1016/j.commatsci.2019.109166 |
| Popis: | A numerical procedure is developed to assess and reduce the discretisation-induced anisotropy in the solution of the phase-field model for dendritic growth. The meshless radial basis function-generated finite difference (RBF-FD) method and the forward Euler scheme are used for the spatial and temporal discretisation of the phase-field equations, respectively. A second-order accurate RBF-FD method is ensured by the use of augmentation with monomials up to the second order, while shape-parameter-free polyharmonic splines of the fifth-order are used as the radial basis functions. The performance of the RBF-FD method is assessed on regular and scattered node distributions by observing the mean phase field, the size of the primary dendrite arm, and the growth velocity. The observables at different orientation angles are compared to assess the orientation dependence of the solution. We show for the first time that the use of the RBF-FD method on a scattered node distribution provides a robust approach for the solution of the phase-field model for dendritic solidification with respect to an arbitrary preferential growth direction. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect