No abstract

Current and emerging trends in High Performance Computing (HPC) are providing transformational capabilities for simulation-based research and development and simulation-based design. Numerous efforts are underway to provide exascale systems in the next decades. HPC architectures are rapidly evolving and the tools and methods need to keep pace with both the scale and the evolving hardware architecture. Emerging HPC capabilities provide potential for simulation of increasingly complex, multi-scale and multidisciplinary applications for discovery, design and evaluation of aerospace systems. The computational mesh, along with the geometry that it represents, has a considerable impact on the quality, stability, and amount of resources required to complete numerical simulations. Extreme-scale environments require increased levels of process automation and reliability not currently available in state-of-the-art mesh generation tools. These shortcomings make geometry modeling and mesh generation a pacing bottleneck for the future. The paper will summarize the panel discussion that was held at AIAA's 2015 SciTech Conference in which the path for geometry and mesh generation as a supporting element of the NASA CFD 2030 Vision was discussed.

Geometrical models output from CAD software often require modification before they may be used for analysis-quality mesh generation. This is due primarily to the inconsistencies in tolerances used by the CAD operator and the tolerances required for analysis. This paper presents a method for construction of watertight surface meshes directly on imperfect non-modified CAD models. The method is based on a hierarchical grid topology structure that defines a surface mesh by a grid and a collection of curves defining the boundary. Curve boundaries on component surfaces are iteratively split and merged according to user-set tolerances, allowing adjacent surface meshes to become computationally watertight via their shared edge curves. The collection of watertight surface meshes may then be made model-independent through interactive agglomeration of the surface meshes, followed by refinement and decimation sweeps to remove artifacts of original surface edges. Interactive procedures used for difficult cases are also explained, as are ongoing efforts for further automation.

The NASA CFD Vision 2030 Study: A Path to Revolutionary ComputationalAerosciences is the latest in a series of significant reports that illuminate the path ahead for computational fluid dynamics. Mesh generation and the use of complex geometry models were cited by the study as key aspects requiring significant improvement if the vision for CFD in the year 2030 is to be realized. In particular, the study cited meshing for being unable to reliably generate valid, high-quality meshes on the first attempt, for being inadequate in its ability to utilize complex geometry models, for not taking advantage of high performance computing resources, and for not providing robust, solution-adaptive capabilities. This paper delves into all these challenges, expands on them, contextualizes them, and identifies paths toward their resolution. For most of these challenges, in the short term, the potential resolutions tend more toward the educational and organizational than technical. This is particularly true for CFD's use of geometry models and, to a lesser extent, mesh quality. Meshing's future use of high performance computing resources and implementation of technology such as adaption are more technical and require alignment with the needs of CFD solver technology. Nomenclature API= Application Programming Interface BREP = Boundary REPresentation CAD = Computer Aided Design CAE = Computer Aided Engineering CFD = Computational Fluid Dynamics COTS = Commercial Off-the-Shelf ECAD = Electrical/Electronic CAD FEA = Finite Element Analysis HPC = High Performance Computing LES = Large Eddy Simulation MCAD = Mechanical CAD MDAO = Multi-Disciplinary Analysis and Optimization OML = Outer Mold Lines RANS = Reynolds-Averaged Navier-Stokes STEP = STandard for the Exchange of Product model data STL = STereoLithograpy / Standard Tessellation Language

No abstract