Mesh Generation
Mesh generation partitions a domain into elements or cells; mesh quality strongly controls the accuracy and stability of every discretization on it.
Discretizing the geometry
Before any finite-element, finite-volume, or finite-difference computation, the physical domain must be divided into a mesh of cells or elements. The mesh defines where unknowns live and how they connect. Its structure and quality directly determine solution accuracy, solver conditioning, and computational cost.
Structured versus unstructured
Structured meshes have a regular grid topology, so neighbors are found by index arithmetic; they are memory-efficient and fast but hard to fit to complex shapes. Unstructured meshes use arbitrary connectivity of triangles or tetrahedra, fitting any geometry at the cost of storing explicit connectivity. Hybrid and block-structured meshes combine both.
Quality metrics
- Aspect ratio: highly stretched cells hurt accuracy and conditioning.
- Skewness and minimum angle: sliver elements with tiny angles degrade finite-element accuracy.
- Smoothness: abrupt jumps in cell size introduce error.
- Orthogonality: matters especially for finite-volume flux accuracy.
Generation techniques and refinement
Common algorithms include Delaunay triangulation, advancing-front methods, and octree-based generation. Adaptive mesh refinement concentrates cells where the solution varies rapidly, guided by error estimators, and coarsens where it is smooth. This delivers accuracy where it is needed without a uniformly fine, expensive mesh.
Boundary-layer regions and sharp features often need anisotropic elements aligned with the flow or field. Generating high-quality meshes for the intricate geometry of breeder Hyperion components is a substantial part of setting up an accurate simulation.