FEM-based approaches for modeling of resource-demanding magnetization problems with magnetic scalar potential

Seminars

Laboratory of Information Technologies

Seminar of the Scientific Department of Computational Physics

Date and Time: Friday, 1 March 2024, at 11:00 AM

Venue: room 310, Meshcheryakov Laboratory of Information Technologies

Seminar topic: “FEM-based approaches for modeling of the resource-demanding magnetization problems with magnetic scalar potential”

Speaker: Alexander Chervyakov

Abstract:

Despite the excellent quality of numerical calculations, 3D FEA analysis based on the magnetic vector potential is computationally expensive and therefore limited by the available hardware resources for magnetostatic problems with complicated model geometries, large nonconducting regions, nonlinear materials, and increased requirements for accuracy of calculations. To improve the computational efficiency of finite-element modeling for such problems, instead of vector potential, we propose to use the scalar potential either in the combination with vector or even separately. In the former case, both potentials are defined by Maxwell’s equations for conducting and nonconducting regions of the problem domain and coupled together on their common interfacing boundaries. Thin cuts with the potential jumps are constructed in the current-free regions to make them simply connected and ensure the consistency of the vector-scalar formulation. In the latter case, the scalar potential is only defined for nonconducting regions, while the impact of inductors on the entire problem domain is modeled either with the help of the potential jumps across thin cuts, or by using the magnetization of linear and nonlinear permanent magnets. The comparative analysis of the numerical efficiency of proposed methods is carried out by using the model of the dipole magnet as an example. Most efficiently, these methods can be applied for modeling of the magnetic systems, where a significant number of simulations with significant variation in geometric shapes is required during development of the optimal system design.