Issue link: https://resources.system-analysis.cadence.com/i/1472078
Preparation of Geometry Models for Mesh Generation and CFD 2 www.cadence.com [Geometry model suitability is not just a process-oriented efficiency issue. Geometry should be considered a first-order effect on the CFD solution's accuracy. If the shape of the object is not modeled accurately, the usefulness of the simulation result is unclear at best and completely misleading at worst.] A goal of "providing better geometry models" seems to imply that current geometry modeling tools and practices are inade- quate or lax. While it certainly is true that programming and application practices can negatively influence suitability, the suitability of a geometry model depends mainly on inherent aspects of the methods, purpose, and timing of its creation. f Geometry models are created using specific mathematical methods that implicitly imbue them with attributes unique to those methods, attributes that may or may not be suitable for simulation. f Geometry models are created for many purposes, only one of which is CFD simulation. Models are created for product definition, bill of materials generation, manufacturing, visualization and presentation, simulation other than computational, computational simulation other than CFD, and more. f The phase of the design process during which a geometry model is created can also influence its suitability for CFD. In conceptual design, models are generated frequently using tools more attuned to design exploration. During the detailed design phase, the tools and methods are more formalized and carry more data. During the validation stage of the design process, the geometry model will be fully realized to a high level of detail. Taken in combination, method, purpose, and design phase are the major influences on whether a geometry model is suitable for simulation, if preparation prior to use will be necessary, and the level of effort involved in preparation. For this discussion, downstream benefits of the geometry model are generating a mesh and computing a CFD solution on that mesh. While the former is obvious, the latter can use a geometry model for surface mesh adaptation or mesh curving and degree elevation for high-order methods. [When a flow solver adapts the mesh to the flowfield solution, it will need to insert or move mesh points on the geometry model. Therefore, the CFD flow solver will need access to the geometry model and software for evaluating the geometry. Unless the original mesh was generated by the flow solver, it will need additional information like the mapping of mesh points to geometry model entities. Knowing the specific geometry model entity – surface or curve - to which a mesh point should be projected is essential, not just for efficiency in the computations but in terms of accuracy. The accuracy issue primarily involves mesh points in the vicinity of intersecting surfaces (e.g., a wing-fuselage juncture) which are only computed to within a tolerance. The mesh generation software will place mesh points precisely along the intersection curve. Without a mapping to guide it, adaptation in the flow solver may blur the intersection by creating mesh cells with poor geometric shape metrics that span across the intersection. Knowing the mapping of mesh to geometry model will also inform decisions related to domain decomposition and distribution of the geometry model in a high-performance computing (HPC) environment. https://info. pointwise.com/improve-cfd-efficiency-solution-based-mesh-adaptation] The burden of geometry model preparation for simulation is not a new phenomenon. Thompson et al. in 1988 [2] noted the challenges of meshing complex geometry models and proposed the development of a "CAD system oriented to CFD applica- tions," seeming to prioritize delivering better models rather than improving post-delivery model preparation. Geometry Modeling Fundamentals A computational geometry model is an idealized mathematical representation of an object's shape. It is a model used to generate a mesh in precisely the same reason as a turbulence model is used to solve the Reynolds-Averaged Navier-Stokes equations. Both models are useful approximations of reality. Not all geometry models are the same, even those purportedly of the same object. For example, a spline model of an aircraft differs significantly from a faceted model of the same aircraft, not just in representation but how the actual shape is approximated. All geometry models contain approximations and simplifica- tions of varying degrees. Simplifications and approximations can occur during model creation (e.g., omitting manufacturing details such as seams between adjacent surfaces) and during preparation for simulation (e.g., removing details that complicate the simulation, such as protruding fasteners). There are two categories of solid object representation: boundary representation and volumetric representation. The descrip- tions provided herein are necessarily brief. See Taylor & Haimes [3] for a more in-depth treatment of geometry modeling for computational simulation.