Computing Library › Numerical Methods
Numerical Methods

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

Kronos motion — mesh

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

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.