AMA: A High-Performance Anisotropic Mesh Adaptation Software for US3D

The main objective of this D2P2 project is to develop AMA, a module for anisotropic mesh adaptation whose interface will seamlessly interconnect with the well-established hypersonic flow solver US3D. Because US3D is a finite volume, cell-centered flow solver and supports general unstructured meshes, AMA will support finite volume, cell-centered semi-discretizations on hybrid unstructured meshes composed of tetrahedra, prisms, pyramids, and hexahedra: in this primary context, hanging nodes do not raise any issues. To maximize the potential for commercialization, AMA will also support finite volume, vertex-based flow solvers that operate on unstructured tetrahedral meshes, as such solvers are very popular for many other CFD applications. AMA will also be interoperable with finite element based computational fluid dynamics and computational structural dynamics codes that support tetrahedral elements. In all cases, AMA will be grounded in mesh adaptation criteria and technologies that are robust with respect to curved geometries as well as changes in geometry and flight conditions across an entire hypersonic vehicle trajectory or envelope resolving aeroheating and high-temperature wakes. To achieve parallel scalability on massively parallel high performance computing systems, AMA will feature a predictive, mesh repartitioning-based, dynamic load balancing scheme with a proven track record and where potential large memory spikes are anticipated and avoided. To achieve seamless integration with US3D, the interface of AMA will be designed to be compatible with the restart/checkpoint and post-processing workflows of US3D, and sufficiently versatile to be potentially compatible with the counterpart workflows of other code bases employed by the DoD and its prime contractors developing next-generation hypersonic vehicles.