Variants of the Finite Element Method for the Parabolic Heat Equation: Comparative Numerical Study

Ahmed A. Hamada,Mahmoud Ayyad,Amr Guaily

DOI: https://doi.org/10.1007/978-3-030-39847-7

2020-05-25

Abstract:Different variants of the method of weighted residual finite element method are used to get a solution for the parabolic heat equation, which is considered to be the model equation for the steady state Navier-Stokes equations. Results show that the Collocation and the Least-Squares variants are more suitable for first order systems. Results also show that the Galerkin/Least-Squares method is more diffusive than other methods, and hence gives stable solutions for a wide range of Péclet numbers.

Numerical Analysis

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to study the performance of different variants of the finite element method in solving the Parabolic Heat Equation. Specifically, the author hopes to evaluate the performance of these methods in terms of numerical viscosity and stability by comparing different weighted residual finite element methods, especially in the case of convection - dominated flows.

### Research Background and Problem Description

1. **Research Object**:
- The Parabolic Heat Equation is a typical second - order partial differential equation (PDE), which is used to describe the heat diffusion process.
- This equation can be used as a model equation for the steady - state Navier - Stokes equation, which is of great significance in fluid mechanics.

2. **Research Goal**:
- Compare the numerical performance of different variants of the finite element method in solving the Parabolic Heat Equation.
- Evaluate the stability and accuracy of these methods in dealing with convection - dominated flows.

3. **Research Method**:
- Use different variants of the weighted residual finite element method, including the Standard Galerkin method (SG), the Collocation method (C), the Least - Squares method (LS), the Galerkin/Least - Squares method (GLS), the Collocation/Galerkin method (CG), the Collocation/Least - Squares method (CLS) and the Collocation/Galerkin/Least - Squares method (CGLS).
- Conduct numerical experiments on these methods, changing the number of mesh cells and the Péclet number (Pe), to observe their performance under different conditions.

4. **Key Parameter**:
- Péclet number (\( Pe=\frac{Lu}{\alpha} \)), where \( L \) is the characteristic length, \( u \) is the local flow velocity, and \( \alpha \) is the thermal diffusivity.
- The Péclet number reflects the relative importance of the convection term and the diffusion term. When the Péclet number is large, the problem becomes convection - dominated.

### Main Findings

- **Collocation and Least - Squares methods**: They are more suitable for first - order systems because they show better stability in convection - dominated cases.
- **Galerkin/Least - Squares method (GLS)**: It is more diffusive than other methods, so it can give stable solutions in a wide range of Péclet numbers.
- **Standard Galerkin method (SG) and Collocation/Galerkin method (CG)**: They become unstable as the Péclet number increases.
- **Least - Squares method (LS)**: When using linear shape functions, it degenerates into a pure diffusion method, and shape functions of the same order as the highest - order derivative need to be used to improve accuracy.

### Conclusion

- The GLS method has been proven to be the most stable and is suitable for high Péclet number cases.
- The CLS method is suitable for very high Péclet numbers or inviscid flows (flows corresponding to very high Reynolds numbers).
- For first - order systems, such as the Euler equation or the Navier - Stokes equation as a first - order system, the LS and C methods provide higher accuracy.

Through these studies, the author provides a valuable reference for choosing a finite element method suitable for a specific problem.