Physics-Informed Neural Networks for Boltzmann-BGK Equations with Absorbing Boundary Layers

Aygün A., Karakuş A.

SIAM Conference on Computational Science and Engineering 2023, Amsterdam, Hollanda, 27 Şubat - 03 Mart 2023

  • Yayın Türü: Bildiri / Yayınlanmadı
  • Basıldığı Şehir: Amsterdam
  • Basıldığı Ülke: Hollanda
  • Orta Doğu Teknik Üniversitesi Adresli: Evet

Özet

We introduce the application of physics-informed neural networks (PINN) for Boltzmann-BGK equations for nearly incompressible flows. The governing equations are discretized in the velocity space using Hermite polynomials, resulting in a first-order conservation law. A perfectly matching layer (PML) surrounding the computational domain is used to dampen the waves leaving the domain. A multi-domain approach is used with PINN to ensure the continuity of the solution. The damping profile of the PML is an output of the PINN offering a suitable constant for absorbing the waves.