The proposed project is directly aligned with the objective of the NASA transformational tools and technologies, (TTT1) project, subtopic 2.1.3 extreme scale geometry and mesh generation…

This is a $749,876.82 cooperative agreement awarded by NASA to the Old Dominion University Research Foundation (ODURF) to establish an extreme-scale parallel mesh generation and adaptation framework for computational fluid dynamics (CFD) applications. The project aligns with NASA's Transformational Tools and Technologies (TTT) program, specifically the Extreme Scale Geometry and Mesh Generation Environments subtopic. ODURF, a non-profit research organization affiliated with Old Dominion University, will develop novel parallel mesh generation methods and power-aware runtime software systems to support state-of-the-art CFD capabilities. This work aims to eliminate artifacts from domain decomposition and optimize power consumption, while also improving error estimation for anisotropic mesh adaptation. The award does not include any set-aside designations, and ODURF may collaborate with other institutions and contractors as part of this effort to advance NASA's 2030 vision for extreme-scale parallel mesh generation.

The proposed project is directly aligned with the objective of the NASA transformational tools and technologies, (TTT1) project, subtopic 2.1.3 extreme scale geometry and mesh generation environments to: establish an extreme scale geometry and mesh environment enabling large scale mesh generation and automated optimal adaptive meshing techniques that leverage the evolving high performance computing paradigm. Specifically, we propose to: design a multi layered algorithmic and software framework for 3D tetrahedral anisotropic parallel mesh generation and adaptation to support state of the art functionality develop error estimates to drive an anisotropic adaptive process design a power aware parallel runtime software system for extreme scale adaptive CFD computations including, i. Mesh generation & adaptation and, II. Consistent error based metrics for CFD adaptation. In addition we will develop and implement prototypes for selective components of the framework in order to calibrate our design using current parallel platforms. The innovative design and tools that, if successful, will transform our computational CFD capabilities include: novel parallel mesh generation methods and HPC runtime software systems that, as opposed to traditional methods: a. Adapt to the network and memory hierarchies in current and emerging parallel platforms using a nested hierarchy of data and domain decomposition approaches. This approach will eliminate the creation of artifacts, small angles, edges and faces, due to decomposition; b. Separate concerns for mesh generation and runtime software systems; and c. Optimize power; and novel error estimates to drive an anisotropic adaptive process as opposed to, a. Feature-based adaptive indicators which lack rigor in estimation of the error and have been shown to lack robustness and generality, and b. Adaptive methods using embedded refinement approaches that can only adapt within a fixed mesh topology. A team of leading high performance computing, mesh generation and CFD experts from multiple institutions will collaborate to achieve the objectives of this project. This work will deliver a framework for extreme scale parallel mesh generation and adaptation for arbitrary anisotropy and orientation based on control of estimated error, in a manner that is independent from the details of the underlying discretization.