AE Seminar
"Rethinking Mesh Curving Algorithms to Realize the Full Potential of High-Order Finite-Element CFD"
featuring
Devina Sanjaya
(Assistant Professor, University of Tennessee, Knoxville)
Thursday, October 15
TBD
About the Seminar:
Adaptive, efficient, and reliable computational fluid dynamics (CFD) is essential for designing next-generation aerospace vehicles, from micro aerial vehicles to fighter jets, missiles, and hypersonic systems. Adaptive high-order CFD methods (third-order accurate or higher) have long been regarded as a promising path toward these goals, yet after three decades of research, their full potential has not been realized in large-scale applications, especially those with complex geometries and coarse meshes. A key reason is that curving meshes to conform to the boundary and geometry, as in current practice, is necessary but insufficient for effective use of high-order methods. This talk presents theory and algorithms for curved metric-based mesh adaptation in a high-order finite-element framework, with aerospace applications illustrating their benefits; these benefits are achieved solely through improved meshing algorithms, with the flow solver unchanged. The first part of the talk establishes the importance of precisely placing high-order geometry nodes within mesh elements through a warped-element refinement method. The second part reviews metric-based mesh adaptation for linear meshes, including the benefits of anisotropic mesh adaptation and Mesh Optimization via Error Sampling and Synthesis (MOESS), a rigorous method for obtaining the metric field that drives adaptation. The third part presents the speaker’s work on high-order MOESS (HOMES), a native extension of MOESS to curved meshes, and highlights the need for a robust metric-based, a posteriori error estimator and an affordable optimization process that strongly couples vertex and high-order node movements. The unifying theme of the results is that stronger theory and algorithms for curved metric-based adaptation can reduce a target error quantity by an order of magnitude, enabling meaningful high-order CFD simulations on coarser, higher-order meshes and potentially accelerating convergence.
About the Speaker:
Devina Sanjaya is an Assistant Professor of Aerospace Engineering and a Joint Assistant Professor of Computational and Applied Mathematics at the University of Tennessee, Knoxville. She received her Ph.D. in Aerospace Engineering and Scientific Computing from the University of Michigan, Ann Arbor, in 2019 and her M.S. in Aeronautics and Astronautics from Stanford University in 2014. She was a Visiting Scientist in the NASA Advanced Supercomputing (NAS) Division at NASA Ames for three summers. Her research integrates advances in aerospace engineering, mathematics, and computer science to develop the foundational theory, numerical methods, and algorithms for adaptive, efficient, and reliable computational fluid dynamics (CFD). Much of her work has focused on developing new mesh warping and mesh curving algorithms for high-order finite-element methods. She also actively conducts engineering education research and has led an interdisciplinary NSF-REU site on Advanced Air Mobility. She has held several leadership roles in the aerospace community, including serving as the elected Chair of the AIAA Meshing, Visualization, and Computational Environments (MVCE) Technical Committee. Her honors include the MGB-SIAM Early Career Fellowship, the François-Xavier Bagnoud (FXB) Fellowship, and the Zonta International Amelia Earhart Fellowship.