How to Tell If a CFD Report Is Actually Trustworthy
Five practical checks from ASME V&V 20 to judge a CFD report: verification versus validation, mesh convergence, boundary conditions and reporting.
CFD
8/13/20262 min read
A CFD report can look perfect — smooth contours, neat conclusions — and still be wrong. Most readers can’t re‑run the simulation, but you can still spot weak reports with five practical checks, grounded in the ASME V&V 20 standard for verification and validation.
Verification vs validation
Verification: Was the simulation solved correctly? (mesh, convergence, numerical errors)
Validation: Is it modelling the right physics? (comparison to real data or accepted benchmarks)
A report can be numerically clean and physically meaningless if it skips validation.
Check 1: Mesh independence study
A trustworthy report runs at least three meshes (coarse, medium, fine) and shows key results (pressure drop, velocity, temperature) stabilising as the mesh is refined. Ideally, changes between refinements are small (<1–2%) and discretisation uncertainty is quantified (e.g. GCI). One mesh only = no proof the answer isn’t a mesh artefact.
Check 2: Boundary conditions stated and justified
Results depend heavily on inputs: inlet velocity/mass flow, outlet pressure, temperatures, turbulence intensity, etc. A good report lists all boundary conditions and explains their source (design standard, measured data, code requirement, or explicit assumption). No boundary conditions = you don’t even know what question was solved.
Check 3: Convergence on quantities that matter
Residuals dropping to 1e‑4 or 1e‑5 show numerical balance, not physical convergence. A solid report also tracks monitor points (e.g. velocity at a plane, pressure drop, temperature in a zone) and shows they stop changing. Residuals only = incomplete convergence story.
Check 4: Turbulence model named and appropriate
Turbulence is modelled, not solved exactly. k‑ε, k‑ω SST, Spalart–Allmaras, LES — each suits different flows (internal ducts, external wind, separation, etc.). A credible report names the model and briefly justifies it, including near‑wall treatment (y⁺, wall functions). No model named = missing core methodology.
Check 5: Comparison to something real
Validation means comparing results to wind tunnel tests, field measurements, published correlations, or code‑based expected ranges, with uncertainties discussed. A report with zero comparison to reality may be numerically correct and physically wrong, with no way to tell from the document alone.
What to do?
You don’t need to run CFD to judge a report. Ask for:
Mesh independence (3+ meshes, stable results)
Explicit, justified boundary conditions
Convergence on key quantities, not just residuals
Named turbulence model with rationale
Some comparison to real or published data
No coloured plot replaces this. If a report can’t show these, it’s asking you to trust it on faith.
AcouBIM Engineering is an acoustic consultancy and BIM coordination practice based in Ajman Free Zone, working across the UAE and Saudi Arabia. If you’re reviewing engineering analysis on a project and want a second set of eyes on the methodology, get in touch:
info@acoubim.com · +971 58 563 0037 · www.acoubim.com