Micaiah Smith-Pierce

(Advisor: Prof. Ruffin)

will propose a doctoral thesis entitled,

An Automated Approach to CFD with Cartesian-Grid-Based Meshing and a Discontinuous Galerkin Solver

On

Thursday, May 2 at 10:00 a.m.

Montgomery Knight Building 317

Abstract

This work concerns automating Computational Fluid Dynamics (CFD) simulations, a key challenge in aerospace engineering. CFD is currently the highest-fidelity means of predicting many quantities that are critical to aerospace design, such as drag on aircraft and atmospheric entry heating on spacecraft. The importance and complexity of CFD simulations in aerospace engineering is growing, increasing the pressure to provide rapid and reliable results. However, CFD continues to require a substantial amount of time from a human user, especially in the meshing and post-processing phases, and the results are strongly dependent on the user's skill. This work aims to reduce user workload in both of these phases and improve repeatability by coupling a robustly automated Cartesian-grid-based meshing scheme with a reliably accurate high-order Discontinuous Galerkin (DG) solver. The Cartesian-grid-based approach has previously been developed for automated all-hex meshing. This work extends it to achieve robustness to faults in the input geometry and direct support for high-order curved meshes. The high-order DG solver complements this by providing very high accuracy and reduced mesh sensitivity, mitigating quality issues associated with Cartesian-grid-based meshing. To make DG viable in practice, this work proposes more efficient explicit time integration schemes as well as a novel shock capturing method. This method, referred to as discretization-independent artificial viscosity, eliminates nonphysical oscillations and achieves mesh convergence, which popular existing schemes fail to do. Meshing tests involving two- and three-dimensional geometry demonstrate good resolution distribution and element quality comparable to industry-standard unstructured mesh generators. Accurate flow solutions are demonstrated across a range of flight conditions, from subsonic to hypersonic, including smooth aerodynamic heating results. Furthermore, preliminary performance benchmarking shows promise that the solver can provide speed competitive with industry-standard solvers.

Committee

• Prof. Stephen M. Ruffin – School of Aerospace Engineering (advisor)
• Prof. Lakshmi N. Sankar – School of Aerospace Engineering
• Prof. Spencer Bryngelson – School of Computational Science and Engineering