FEA Mesh Size: How to Choose a Mesh Without Guessing

One of the most common questions in finite element analysis is: “What mesh size should I use?” Unfortunately, there is no single number that works for every model.

FEA mesh size should be selected based on geometry, material behavior, load path, stress gradients, contact, and the quantity you are trying to predict.

Why a coarse mesh can be misleading

A coarse mesh may capture the overall deformation of a simple structure reasonably well, but it can miss local stress concentrations. If a fillet, hole, notch, contact region, or thin section controls the design, the local result can change significantly as the mesh becomes finer.

The danger is not that a coarse mesh always produces a wrong answer. The danger is believing a result is accurate simply because the software produced a colorful contour plot.

Start with the physics

Before opening the meshing settings, ask where the load travels. Which areas carry tension? Where does bending occur? Where does contact happen? Where does geometry change quickly?

These regions deserve attention because stress gradients are often higher there.

Use mesh refinement intelligently

You do not necessarily need the smallest elements everywhere. A useful approach is to use a reasonable global mesh and refine critical regions.

For example, a large plate may not need the same element size as the area around a small mounting hole. Local refinement can reduce computation while improving resolution where it matters.

Mesh convergence is the key check

Mesh convergence means checking how the result changes as the mesh becomes finer. Choose a quantity that matters, such as displacement, reaction force, average stress, or a stress value away from a singularity.

Run the model with one mesh, refine it, and compare the result. Continue until further refinement produces an acceptably small change for the engineering decision.

Be careful with stress singularities

Sharp re-entrant corners, point loads, and perfectly rigid constraints can create mathematical stress singularities. The peak stress may increase as the mesh is refined instead of converging.

That does not automatically mean the software is broken. It means the idealized geometry or boundary condition creates a non-physical mathematical condition.

In such cases, evaluate stress away from the singular point, model the real fillet or load distribution, or use an appropriate engineering interpretation.

Element type matters

Mesh size cannot be separated from element formulation. Shell elements may be appropriate for thin sheet structures, while solid elements may be needed for thick or complex three-dimensional geometry. Poor aspect ratios and distorted elements can also reduce solution quality.

Check the boundary conditions

A perfect mesh cannot rescue an unrealistic model. If a bracket is fixed over an entire face when the real part is attached with bolts, the stress distribution may be very different.

Always ask whether the model represents how the part is actually loaded and constrained.

Practical FEA workflow

  1. Simplify geometry without removing important load-carrying features.
  2. Define realistic materials and contacts.
  3. Apply realistic loads and boundary conditions.
  4. Create a sensible global mesh.
  5. Refine holes, fillets, contacts, and other critical regions.
  6. Check element quality.
  7. Run a convergence study.
  8. Compare results with hand calculations or known behavior.
  9. Interpret the result in the context of the actual design.

Do not chase the prettiest contour

FEA is an engineering tool, not a picture generator. A smooth contour plot does not prove that the model is correct.

For a simple beam, compare FEA displacement with the analytical beam equation. For a pressure vessel, compare with an appropriate theoretical solution. For a bracket, compare reactions and approximate bending behavior.

Final takeaway

FEA mesh size should be driven by the physics of the problem and verified through convergence. Use refinement where the result is sensitive, avoid blindly trusting peak values at singularities, and always validate the model against engineering expectations.

If your FEA result cannot survive a basic engineering sanity check, making the mesh finer is not the first thing to fix.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top