Mesh Strategy for High-Velocity Impact
Mesh strategy for high-velocity impact FEA — local refinement, element size, through-thickness discretisation, localisation capture, mesh dependency and computational cost.
Mesh requirements for impact
The mesh for a high-velocity impact analysis must resolve: the contact zone (projectile-target interface), the stress-wave propagation (wave front spread over several elements), the localisation (shear band, crack), and the through-thickness stress state (for thick targets). These requirements conflict with the computational cost — a mesh that resolves all features may have billions of elements. The mesh strategy must balance resolution and cost through local refinement — fine mesh where needed, coarse mesh elsewhere.
Local refinement
Local refinement concentrates fine elements in the impact zone — the region where the contact, deformation and damage occur. The impact zone typically extends a few projectile diameters from the impact point. Outside the impact zone, the mesh can be coarser — the stress waves have spread, the deformation is smaller, and the damage is absent. The transition from fine to coarse mesh must be gradual — a sudden transition can cause wave reflection at the mesh interface, producing artificial stress waves. The local refinement may be achieved through graded meshing, adaptive meshing or submodelling.
Element size
The element size in the impact zone determines the resolution of the analysis. The element size should be small enough to resolve the stress-wave front (at least 3-5 elements across the wave front), the shear band width (at least 3-5 elements across the band) and the contact zone (at least 5-10 elements across the projectile diameter). For a 10 mm diameter projectile, this may require elements of 1 mm or smaller in the impact zone. The element size also determines the failure resolution — element deletion removes elements of this size, so the failure pattern is quantised at the element size.
Through-thickness discretisation
For thick targets, the through-thickness discretisation determines the stress-wave resolution and the spall prediction. The through-thickness should have at least 5-10 elements to resolve the wave propagation and the reflected tensile wave. For spall prediction, the element size should be smaller than the expected spall thickness — otherwise the spall cannot be resolved. For thin targets, the through-thickness may be represented by shell elements (if the stress state is approximately uniform) or by 3-5 solid elements (if the through-thickness gradient is significant).
Localisation and mesh dependency
Localisation — shear banding, cracking — is mesh-dependent in element-deletion-based failure models. A finer mesh produces a narrower localisation band and a different failure pattern. The mesh dependency is a fundamental limitation — the localisation width may be smaller than any practical mesh can resolve. Nonlocal or regularised failure models (gradient, viscoplastic) can reduce the mesh dependency but are more complex and less widely available. The mesh sensitivity must be assessed by running the analysis with at least two mesh densities and comparing the results — if the penetration result changes significantly, the mesh is not converged.
Computational cost
The computational cost of penetration FEA is high — the fine mesh in the impact zone requires a small time step (Courant condition), and the analysis may require millions of time steps. The cost scales with the number of elements and the number of time steps — a 2x finer mesh requires approximately 8x more elements (3D) and 2x more time steps, giving approximately 16x more computational cost. The cost must be managed through mesh strategy: use the coarsest mesh that gives an acceptable result, use local refinement, and use submodelling or adaptive meshing where available. A mesh-sensitivity study with two or three mesh densities is essential to justify the mesh choice.
Mesh-quality metrics
Local element size is only one aspect of mesh quality. Highly skewed or stretched elements can distort wave speed, contact pressure and failure localisation before they become visibly invalid. Review aspect ratio, warpage or Jacobian measures appropriate to the element type, especially in the impact zone and at mesh transitions. The starting mesh should have enough quality margin to tolerate the severe distortion expected during the event rather than relying on erosion to remove poor elements immediately.
Mesh transitions and wave contamination
A change in element size also changes the numerical dispersion characteristics of the mesh. Very abrupt transitions can reflect or scatter high-frequency content even when the material interface is physically continuous. Use gradual transitions where stress waves must pass from the refined zone into the surrounding structure, and inspect time histories on both sides of the transition for nonphysical reflections. This is especially important when support reactions or remote equipment response are outputs, because numerical high-frequency content can propagate away from the impact site.
Local solid regions and shell structures
Thin-walled components may be represented globally with shells while the impact zone requires a three-dimensional solid description. A mixed-dimensional strategy can be efficient, but the shell-to-solid coupling must transmit membrane force, bending moment and transverse shear without creating an artificial stiffness step. Validate the coupling using a simpler dynamic problem before relying on it in a severe impact model. Alternatively, use a solid local submodel driven by a validated larger structural model when two-way load redistribution is not governing.
Convergence evidence
Mesh convergence in failure problems is rarely demonstrated by a single scalar peak stress. Compare structural quantities that should stabilise: deformation history, support impulse, damage dimensions, penetration classification, residual state and energy partition. If failure is regularised by characteristic length or fracture energy, verify that the implementation uses the actual element characteristic length as intended. Report the mesh family and the change in decision-relevant outputs so that the selected production mesh is justified rather than merely described as 'fine'.