CFD Validation — Benchmark Cases & Published Results

See the numbers. CFD Lab runs canonical benchmark cases — cavity, step, NACA 0012, Ahmed body, Schlichting plate — and reports measured error against published references.

A CFD result without verification is an opinion. CFD Lab's validation module runs canonical cases with known published solutions and reports the actual error, so you can see what the solver does rather than take a claim on trust.

The catalogue spans the standard set: the lid-driven cavity against Ghia et al. (1982) centreline velocities at Re=100 and Re=400, cylinder wakes against Tritton and Williamson, the backward-facing step against Armaly et al. (1983), Blasius and Schlichting flat plates, Poiseuille, Couette and Taylor-Green analytical solutions, low-Reynolds NACA 0012 sections, and the Ahmed body. Each case declares its own tolerance, and a case outside it is reported as a failure or warning, not quietly dropped.

Grid convergence is handled the way it should be: the Grid Convergence Index, a Richardson-extrapolation estimate of discretisation error computed across three grid levels, so the reported number carries an uncertainty rather than a false precision. A GCI tells you whether you have actually converged, which is a different and more useful question than whether the residual has stopped falling.

The table below is generated at build time from the committed validation-report.json, the same file the in-app Validation module reads, so this page cannot quote better numbers than the solver produced. It includes the cases that currently fail (the low-Reynolds NACA 0012 sections) or carry warnings, and the higher-Reynolds cavity cases that were skipped in that run.

Latest recorded results

From the validation run recorded on 2026-11-12 at commit d8e6456e: 35 cases, 30 passed, 1 failed, 4 with warnings and 0 skipped. The error column is the relative L2 error against the published reference profile; the tolerance is the acceptance threshold for that case. Failing and skipped cases are listed on purpose, because a validation page that only shows passes tells you nothing.

CaseReferenceSolverStatusL2 errorTolerance
Ghia Cavity Re=100Ghia et al. (1982)LBMPASS3.3e-30.050
Ghia Cavity Re=400Ghia et al. (1982)LBMPASS0.0150.050
Ghia Cavity Re=1000Ghia et al. (1982)LBMPASS9.1e-30.025
Tritton Cylinder Re=20Tritton (1959)LBMPASS0.0350.100
Tritton Cylinder Re=40Tritton (1959)LBMPASS0.0200.050
Tritton Cylinder Re=100Tritton (1959)LBMPASS0.0530.070
Poiseuille FlowAnalytical SolutionFVMPASS0.0130.020
Blasius Flat PlateBlasius (1908)LBMPASS6.7e-30.010
Schlichting Flat Plate (k-ω SST)Schlichting (1979)KOMEGAPASS0.1480.300
Ghia Cavity Re=5000Ghia et al. (1982)LBMPASS0.0330.080
Tritton Cylinder Re=200Tritton (1959)LBMPASS0.1070.150
Square Cylinder Re=40 (steady drag)Breuer et al. (2000); Dhiman et al. (2005)LBMPASS0.1190.150
Square Cylinder Re=100 (vortex shedding)Sohankar, Norberg & Davidson (1998), Table II Case 3LBMPASS0.0800.150
Backward-Facing StepArmaly et al. (1983)LBMWARNING0.1640.200
Backward-Facing Step, closed channel (Re 400)Katsoudas et al. (2025) arXiv:2507.16509, Table 1LBMPASS8.4e-30.050
Taylor-Green VortexTaylor & Green (1937)LBMPASS1.7e-30.030
Couette FlowSchlichting (Boundary Layer Theory)LBMPASS4.8e-30.020
Von Kármán Vortex Street Re=100Williamson (1988)LBMPASS4.7e-30.020
NACA 0012 Airfoil Re=5000 α=0°Swanson & Langer (2016) NASA/TM-2016-219003, laminar NACA 0012 Case 1 (M=0.5, α=0°, Re=5000), CD=0.05565LBMFAIL0.1930.020
NACA 0012 Airfoil Re=5000 α=5°Gupta et al. (2023) J. Fluid Mech. 954, A26, doi:10.1017/jfm.2022.958, Fig. 18(d) Re=5000 α=5° 2-D spectral-element Cl=0.206 (digitised)LBMPASS0.0310.070
Ghia Cavity Re=10000Ghia et al. (1982)LBMPASS0.0650.180
Hagen-Poiseuille Pipe FlowHagen (1839), Poiseuille (1840)LBMPASS0.0130.030
3D Lid-Driven Cavity Re=400Albensoeder & Kuhlmann (2005)LBMPASS0.0120.050
Rayleigh-Bénard Convection Ra=3·10⁴Prasianakis & Karlin (2007) Phys. Rev. E 76, 016702 (Nu=3.605, Ra=3·10⁴, Pr=0.71, L/H=2); Clever & Busse (1974) J. Fluid Mech. 65, 625LBMPASS0.0420.140
Turek 2D-1 SteadySchäfer & Turek (1996)LBMPASS0.0130.050
Kelvin-Helmholtz InstabilityDrazin & Reid (1981)LBMPASS0.1900.250
Plane Jet Re=500Schlichting (1933)LBMPASS0.4180.550
Ahmed Body 25° Re=2.78MAhmed & Raitby (1984); Lienhart & Becker (2000)FVMWARNING10.3870.300
Natural Convection Ra=10⁵de Vahl Davis (1983) Int. J. Numer. Meth. Fluids 3:249-264, Nu_avg = 4.519 at Ra=1e5CHTPASS0.0350.120
Risø R-1280 Wind TurbineRisø National Laboratory (2001) Risø-R-1280BEMPASS0.0760.300
FDA Nozzle BenchmarkFDA (2014) "CFD in Medical Devices" benchmark standardFVMWARNING4.3e-40.300
Schlichting Flat Plate (k-ε RANS)Schlichting (1979) Boundary Layer Theory, Table 12.1FVMPASS0.0260.300
Normal Shock M1=2 (Rankine-Hugoniot)Anderson (2003) Modern Compressible Flow Ch. 3; NACA Report 1135 / Anderson App. A.2 normal-shock table (M1 = 2.00)ANALYTICALPASS——
Holtrop-Mennen Trawler ResistanceHoltrop & Mennen (1984); ITTC 1978 friction lineANALYTICALWARNING——
Womersley Pulsatile Channel Flow (α = 4, analytical)Womersley (1955) J. Physiol. 127:553 — exact oscillatory viscous flow; plane-channel form per Loudon & Tordesillas (1998) Int. J. Numer. Meth. Fluids 27:1001.LBMPASS0.0120.020

Frequently asked questions

What is a Grid Convergence Index?

A formal estimate of discretisation error derived by Richardson extrapolation across a sequence of refined grids, with an uncertainty factor reflecting the order of convergence. It answers "how much of my error is numerical?" rather than "has the residual stopped falling?"

Which benchmarks does CFD Lab run?

Lid-driven cavity, cylinder wakes, backward-facing step, Blasius and Schlichting flat plates, channel and pipe flow, Taylor-Green vortex, Rayleigh-Bénard convection, NACA 0012, Ahmed body and a normal shock, among others. The results table on this page lists every case with its measured error, tolerance and status.

Try it now

The page you are reading is a static, crawler-readable summary. The interactive tool runs in your browser — nothing is uploaded.

Open CFD Lab Back to home