Mesh Density, Refinement & Transition Strategy

Technical provenance

Applicable standards / specifications

• ASME V&V 10 (2019) — Standard for Verification and Validation in Computational Solid Mechanics
• ASME VVUQ 10.2 (2021) — The Role of Uncertainty Quantification in Verification and Validation of Computational Solid Mechanics Models

References

• ASME Verification, Validation and Uncertainty Quantification (VVUQ) — Authoritative standards family for computational-model verification and validation.
• Bathe, K.-J. — Finite Element Procedures — Background reference for finite-element formulation, discretisation, solution and verification.

What Is It?

Mesh density, refinement and transition strategy address how the finite element mesh is distributed over the structure — where the elements should be small (refined) and where they can be large (coarse), and how the mesh transitions between fine and coarse regions. The mesh density determines the numerical accuracy of the solution — a finer mesh resolves the displacement and stress fields more accurately. But a uniformly fine mesh is impractical for large structures — the model size and the computational cost would be excessive. The strategy is to refine the mesh where the structural response changes rapidly (stress concentrations, contact, load introduction) and to keep the mesh coarse where the response changes slowly (uniform stress regions). The transition between fine and coarse regions must be smooth to avoid artificial stress concentrations from abrupt mesh changes.

Why It Matters

The mesh is the numerical discretisation — it determines how accurately the finite element model represents the continuous displacement and stress fields. A mesh that is too coarse in a critical region produces non-converged stress — the stress may be significantly under-predicted, leading to an unsafe design. A mesh that is uniformly fine produces a huge model that is expensive to solve and difficult to manage. The art of meshing is to refine where needed and to keep the rest coarse — to concentrate the mesh resolution where the structural response changes most strongly and to transition smoothly to the coarser regions. The right mesh is not the finest mesh — it is the coarsest mesh that resolves the physics and the quantity of interest adequately.

THE RIGHT MESH IS THE COARSEST MESH THAT RESOLVES THE PHYSICS AND THE QUANTITY OF INTEREST ADEQUATELY. A converged mesh is one where the quantity of interest (stress, displacement, frequency) does not change significantly with further refinement. Refining beyond convergence wastes computational effort. Refining short of convergence produces inaccurate results.

Global vs Local Mesh

A global mesh is a mesh that is uniformly refined over the entire structure. A local mesh is a mesh that is refined in specific regions and coarse elsewhere. The global mesh is simple to generate but expensive — it refines everywhere, including regions where the response is smooth and does not need refinement. The local mesh is more efficient — it refines where needed and keeps the rest coarse. The local mesh requires engineering judgement — the analyst must identify the regions that need refinement and must create smooth transitions. For most real structures, the local mesh is the practical approach — the global mesh is too expensive for anything but the smallest structures. The submodelling approach (discussed in a later article) is an extension of the local mesh concept — a coarse global model provides boundary conditions for a fine local model.

Where to Refine

The mesh should be refined where the structural response changes rapidly — where the stress gradient is steep, where the geometry changes sharply, or where the load is introduced. The specific regions that typically need refinement are:

• Fillets and radii — stress concentrations; the stress varies rapidly around the radius
• Holes and notches — stress concentrations; the stress peaks at the hole edge
• Joints and connections — load transfer; the stress changes at the interface
• Contact regions — the contact pressure varies rapidly at the contact edge
• Load introduction — the load is applied at a point or region; the stress spreads from the introduction
• High curvature — the stress varies with the curvature; sharp curves need refinement
• Stress gradients — any region where the stress changes rapidly over a short distance

Transition Zones

The transition between fine and coarse mesh regions must be smooth — the element size should change gradually, not abruptly. An abrupt transition — a small element directly adjacent to a large element — creates an artificial stress concentration at the transition. The small element has a different stiffness than the large element, and the mismatch creates a local disturbance. The transition should be gradual — the element size should increase by a factor of no more than approximately 1.5–2 per transition layer. This means several layers of progressively larger elements between the fine and coarse regions. The transition zone adds some elements but avoids the artificial stress concentration. The transition should also be located away from the region of interest — the disturbance from the transition, though small, should not be in the critical region.

Abrupt Size Change

An abrupt mesh size change — a fine mesh directly adjacent to a coarse mesh with a large size ratio — is a common meshing error. The abrupt change creates a discontinuity in the element stiffness at the interface, which produces an artificial stress concentration. The stress at the interface is not reliable — it is a numerical artifact, not a physical stress. The remedy is a gradual transition (several layers of progressively larger elements) or a transition element (if the solver supports it). The abrupt change is particularly problematic in automatic meshers, which may create a fine mesh in one region and a coarse mesh in an adjacent region without a smooth transition. The analyst should check the mesh transition and ensure that the size change is gradual.

Output-Driven Refinement

The mesh should be refined based on the quantity of interest — the output that the analysis is trying to predict. If the quantity is a local stress at a fillet, the mesh should be refined at the fillet. If the quantity is a global displacement, the mesh may be relatively coarse everywhere (displacement converges faster than stress). If the quantity is a natural frequency, the mesh should be refined to capture the mode shape — a coarse mesh may miss a high-order mode. If the quantity is a buckling load, the mesh should be refined in the buckling region (where the buckling mode has a short wavelength). The output-driven approach ensures that the refinement is targeted — the mesh is fine where it needs to be and coarse where it does not, optimising the computational effort for the specific quantity of interest.

Convergence

Convergence is the process of refining the mesh and checking that the quantity of interest has stabilised — that further refinement does not significantly change the result. A converged result is one where the discretisation error is small enough that the result is reliable. The convergence study runs the analysis at progressively finer mesh densities and plots the quantity of interest against the mesh density (or the number of elements). When the quantity stops changing significantly, the mesh has converged. The convergence study is the primary method for demonstrating that the mesh is adequate. Without a convergence study, the analyst cannot claim that the result is mesh-independent. The convergence study should be performed for each critical quantity — a mesh that is converged for displacement may not be converged for stress.

MODEL CHECK: Perform a convergence study for the quantity of interest. Run the analysis at two or more mesh densities and compare the results. If the quantity changes significantly, the mesh is not converged — refine and re-run. If the quantity is stable, the mesh is adequate. Document the convergence study as evidence of mesh adequacy.

Key Takeaways

• Refine the mesh where the structural response changes rapidly; keep it coarse elsewhere
• Typical refinement regions: fillets, holes, joints, contact, load introduction, stress gradients
• Transition zones should be gradual — abrupt size changes create artificial stress concentrations
• Output-driven refinement targets the mesh at the quantity of interest
• Convergence study — refine and check that the quantity of interest has stabilised