Moving Mesh with Streamline Upwind Petrov-Galerkin (MM-SUPG) Method for Time-dependent Convection-Dominated Convection-Diffusion Problems

Xianping Li,Matthew McCoy

2024-05-21

Abstract:Time-dependent convection-diffusion problems is considered, particularly when the diffusivity is very small and sharp layers exist in the solutions. Nonphysical oscillations may occur in the numerical solutions when using regular mesh with standard computational methods. In this work, we develop a moving mesh SUPG (MM-SUPG) method, which integrates the streamline upwind Petrov-Galerkin (SUPG) method with the moving mesh partial differential equation (MMPDE) approach. The proposed method is designed to handle both isotropic and anisotropic diffusivity tensors. For the isotropic case, we focus on improving the stability of the numerical solution by utilizing both artificial diffusion from SUPG and mesh adaptation from MMPDE. And for the anisotropic case, we focus on the positivity of the numerical solution. We introduce a weighted diffusion tensor and develop a new metric tensor to control the mesh movement. We also develop conditions for time step size so that the numerical solution satisfies the discrete maximum principle (DMP). Numerical results demonstrate that the proposed MM-SUPG method provides results better than SUPG with fixed mesh or moving mesh without SUPG.

Numerical Analysis

What problem does this paper attempt to address?

The paper mainly addresses the issue of non-physical oscillations in the numerical solution of time-dependent convection-diffusion equations when the diffusion coefficient is very small and sharp layers exist. These problems are common in fields such as pollutant diffusion in rivers, atmospheric pollution, Stefan problems, and turbulent transport. When the diffusion tensor is anisotropic, i.e. the diffusion rates are different in different directions, the occurrence of non-physical solutions violates the discrete maximum principle (DMP).

To tackle this issue, the paper proposes a hybrid method combining the Moving Mesh (MMPDE) method and the Streamline Upwind Petrov-Galerkin (SUPG) method, called the Moving Mesh SUPG (MM-SUPG) method. This method aims to enhance stability by adapting the mesh and adding artificial diffusion, while ensuring that the numerical solution satisfies the DMP under anisotropic diffusion.

Both the MM-SUPG method can handle flows with initial sharp layers in both time-independent and time-dependent cases. For isotropic diffusion, it improves the stability of the mesh adaptation method and reduces oscillations. For anisotropic diffusion, the paper introduces a weighted diffusion tensor and a new metric tensor to control the mesh movement, and formulates time step conditions to ensure that the numerical solution satisfies the DMP.

The paper demonstrates the effectiveness of the MM-SUPG method in reducing oscillations and improving the quality of the solution compared to the fixed-grid SUPG method or the moving mesh method without SUPG, through numerical experiments for both isotropic and anisotropic diffusion problems.