What is the actual difference between verification and validation?

Verification asks "did we solve the equations right" — it is a purely mathematical and numerical question about whether the discretized model (the mesh, the solver, the numerical scheme) accurately solves the governing equations it was set up to solve, checked through mesh convergence studies and code/solution verification against known analytical or benchmark solutions, with no reference to the real world at all. Validation asks "did we solve the right equations" — it is an empirical question about whether the chosen physics, material models, and boundary conditions actually represent reality, checked by comparing simulation output against physical test data. A simulation can be perfectly verified (numerically exact for the equations it solves) and still completely invalid if those equations do not represent the real physics, and vice versa a solution can be qualitatively realistic while still containing large, unquantified discretization error.

How fine does a mesh convergence study actually need to be before I trust the result?

There is no universal element count; the target is a demonstrated trend, not a specific density. Run the same model at three or more progressively finer mesh levels, track a key output (peak stress, drag coefficient, natural frequency), and confirm the value is converging toward a stable asymptotic result rather than continuing to change significantly with each refinement. A common practical threshold is that the change between the two finest meshes falls under roughly 2-5 percent, but the acceptable threshold should scale with how the result will be used — a rough trade study can tolerate more discretization error than a result feeding a fatigue-life or certification calculation.

What is the Grid Convergence Index and why do some reports require it?

The Grid Convergence Index (GCI), based on Richardson extrapolation, is a standardized way to estimate the numerical uncertainty in a mesh-convergence result rather than just eyeballing whether the curve looks flat. It compares results from at least three systematically refined meshes, estimates the apparent order of convergence, and extrapolates toward the theoretical mesh-independent (zero cell size) answer, reporting a percentage uncertainty band around the finest-mesh result. Some certification-adjacent industries (aerospace, nuclear, and increasingly automotive CFD reports) require a GCI or equivalent because it replaces a qualitative "the curve flattened out" judgment call with a defensible, quantitative discretization-error estimate.

What are the biggest red flags that a simulation report should not be trusted?

The most common is the complete absence of any mesh convergence evidence — a single mesh, a single run, and a colorful stress or velocity plot presented as if it were self-evidently correct. Other red flags include boundary conditions and loads that are not clearly documented or justified, a nonlinear or CFD solution that "converged" only because residual or force-balance criteria were loosened, a peak stress or other output pulled from a single element or node at a singularity (a sharp reentrant corner or point load) without acknowledging that such values are mathematically unbounded and mesh-dependent, and no comparison at all against hand calculations, test data, or a simpler validated case. A credible report shows its convergence work, states its assumptions, and includes at least a sanity check against something independent of the simulation itself.

Engineering·10 min read·July 12, 2026

🔍 Simulation Verification & Validation (V&V): Mesh Convergence & Model Credibility

Why running FEA or CFD software correctly is not the same as producing a trustworthy result — mesh convergence studies, h-refinement vs p-refinement, the verification-vs-validation distinction, uncertainty quantification basics, and the red flags that separate a credible simulation report from a pretty picture.

📖Engineering Software Handbook (Full Access)

View →

Running the Software Correctly Is Not the Same as Being Right

A finite element or CFD solver will almost always return an answer. It will color a part blue and red, plot a velocity field, and report a peak stress value to four significant figures, regardless of whether the mesh was adequate, the boundary conditions were realistic, or the material model was appropriate for the loading regime. The software has no concept of whether its own answer is trustworthy — that judgment is entirely the engineer's responsibility, and the discipline for making that judgment rigorously and defensibly is called Verification and Validation (V&V). This is a different skill from operating ANSYS, Abaqus, or COMSOL competently, and it is the skill that separates a simulation result an organization can stake a design decision, a certification submission, or a safety case on from a simulation result that merely looks plausible.

Topics covered

verification and validationV&V simulationmesh convergence studyh-refinement vs p-refinementsimulation credibilityuncertainty quantificationASME V&V 10grid convergence indexRichardson extrapolationmodel validation FEACFD verificationsimulation red flagsdiscretization errornumerical error FEA CFD