AMR is most valuable when the important flow feature moves or is not known accurately before solving. Examples include free surfaces, shocks, wakes, flame fronts and overset-interface regions. The mesh can remain coarse elsewhere and refine only where the selected criterion requests it.
Why AMR can outperform a static “refine everything” mesh
A static mesh has to anticipate every location a moving interface, shock or wake may occupy. That often means refining a large volume that contains important physics only briefly. AMR can follow the feature and reduce unnecessary cell count.
Siemens describes STAR-CCM+ model-driven AMR for free surfaces and overset interfaces, as well as user-defined refinement based on field functions such as Mach-number or pressure gradients. Later releases also expanded AMR capabilities for reacting flows and prism layers.
Choose the criterion from the physics
Free surface / VOF
Refine around the gas-liquid interface to reduce numerical smearing without refining the entire tank or wave domain.
Overset
Refine background cells where needed to improve size compatibility with the moving overset grid.
Shock / compressible flow
Use a gradient-sensitive field criterion tied to Mach number or pressure when sharp waves move through the domain.
Reacting flow
Refine around flame-front indicators when combustion structures move or change significantly.
The criterion must isolate the feature you actually care about. A noisy gradient field can otherwise refine large regions and eliminate the computational advantage.
Boundary-layer adaptation needs directional thinking
Near-wall cells are strongly anisotropic by design: very small normal spacing and much larger tangential dimensions. Isotropic refinement can therefore create unnecessary cells or poor aspect-ratio transitions. Siemens introduced anisotropic prism-layer AMR options so refinement can be directed more intelligently.
Even with AMR, the wall-treatment strategy and first-layer design remain important. Adaptation does not remove the need to understand target y+, prism growth and total thickness.
Design the base near-wall mesh first
Use the wall-treatment advisor and prism-layer calculator before relying on adaptation.
How to verify an adaptive mesh
- Plot the refinement level.
Confirm that cells refine where the intended feature exists and coarsen where it leaves.
- Check total cell-count history.
Unexpected runaway refinement often means the criterion is too broad or noisy.
- Inspect the solution feature itself.
A sharper interface or shock is useful only if the engineering outputs become more reliable.
- Compare with a stricter AMR setting.
Perform a sensitivity check on maximum refinement level, threshold or target size.
- Check timestep interaction.
Smaller cells increase local Courant number at the same timestep.
AMR creates smaller cells
Re-check transient timestep and Courant number after refinement reaches its smallest scale.
Common mistakes
- Assuming AMR automatically guarantees mesh independence.
- Using a gradient criterion that refines numerical noise rather than the target physics.
- Ignoring timestep restrictions after local cell sizes shrink.
- Starting from a poor base mesh and expecting adaptation to repair geometry or bad topology.
- Allowing refinement levels to grow without monitoring total cell count and runtime.