Investigating the EBVC Discretization Method for Pressure-based Solvers over Tetrahedral Computational Domains
Swiss partners
-
Hochschule Luzern HSLU: Luca Mangani (main applicant), Mhamad Mahdi Alloush
Partners in the MENA region
- American University of Beirut, Liban: Fadl Moukalled (main applicant), Adam Fares
Presentation of the projet
This project focuses on the investigation of a pressure-based, fully-coupled computational fluid dynamics (CFD) solver. This solver utilizes an edge-based vertex-centered finite volume method (EBVC-FVM). The motivation behind studying the EBVC-FVM approach stems from its superiority over the conventional cell-centered finite volume method (CC-FVM) when dealing with tetrahedral meshes. In the CC-FVM approach on a tetrahedral mesh, there is a limitation in computing gradients within the control volume. Gradients are computed from face values, typically representing averages from the control volume and its neighboring cells. However, tetrahedra have only four faces adjacent to their control volumes, potentially resulting in gradients with lower resolution. On the contrary, the EBVC-FVM approach considers a polyhedral dual mesh corresponding to the tetrahedral mesh. The polyhedral dual control volume exhibits a significantly larger number of faces. Consequently, gradient computation is more accurate as it incorporates information from a greater number of neighboring cells. Tetrahedral meshes are also known to produce numerical diffusion, causing a detrimental effect on the accuracy of the solution particularly in simulations where high-order discretization is crucial, as is the case with Large Eddy Simulation (LES).
Nonetheless, it is noteworthy that highly efficient tetrahedral mesh generators have recently emerged in the industry, drastically reducing mesh generation time. This development encourages the community to increasingly rely on tetrahedral meshes, especially when addressing more complex fluid flow problems involving complex flows and/or geometries that demand denser meshes. Given the extensive adoption of these tetrahedral mesh generators, selecting a discretization approach like EBVC is highly suitable for a sophisticated solver, as it effectively handles tetrahedral meshes. In the existing literature, there is no evidence of solvers that incorporate a dual mesh for their pressure-based approach. Typically, those who do employ a dual mesh tend to use density-based methods. In our project, our goal is to investigate a pressure-based algorithm using the EBVC method, specifically applied to a dualised tetrahedral mesh, and designed to employ an effective vectorization approach, ensuring the creation of a scalable parallel performance. This entails offering methods for GPU acceleration in addition to distributing the computational load across CPUs.