ahmed-body¶
English | Japanese
« Home · « External flow and wind engineering
Ahmed body external aerodynamics (ERCOFTAC case082; Ahmed et al. 1984) with the single-phase D3Q19 LBM (MRT + Smagorinsky LES). openrheo's first EXTERNAL-flow case: inflow/outflow BCs, a solid body mask (bounce-back), slip tunnel walls + a no-slip floor, a smooth mixing-length wall model, synthetic inflow turbulence, and the momentum-exchange force on the body -> drag/lift. Reference (OpenFOAM SPHERIC-training AhmedBody + ERCOFTAC): U=40 m/s, nu=1.5e-5 (air), real Re_H = U h/nu = 768000, 25 deg slant, target Cd ~ 0.285.
Geometry¶

Animation¶
Fields¶
The fields this computation wrote at its last output step, plotted as they are. 3D cases show the mid-plane.


Build & run¶
Cost of the computation that made the figures on this page (Miyabi-G Hopper GPU (GH200, -stdpar=gpu)): 26.6 s wall.
Validation¶
| Tier | T2 — single-point comparison |
| QoI | Cd, Ux(z) |
| Reference | ERCOFTAC case082 (Ahmed 25°), Cd~0.285 |
| Gate | — no gate |
Source & run details¶
| Item | Value |
|---|---|
| Target | orheo-ahmed-body |
| Source | native/benchmarks/ahmed_body.cpp |
| Lattice | D3Q19 |
| GPU twin | yes (orheo-ahmed-body-gpu) |
| Modules | ahmed_solver.hpp, lbm_d3q19.hpp, synthetic_turbulence.hpp |
| Gallery run | 2026-08-30 @ 38ad1eaa |
| Args | 16 4000 |
| Wall time | 26.63 s |
| CPU time | 26.48 s |
| Device | Miyabi-G Hopper GPU (GH200, -stdpar=gpu) |
Computation log¶
The actual output of this computation (head and tail).
Ahmed body (full-domain LBM MRT+LES): grid 833x104x78 (dx=0.0180 m, nh=16), ulat=0.050 (Mach=0.087), Re_H=20000, nu_lat=4.00e-05, tau=0.5001, slant=25 deg, solid=19117 cells
domain: 13.7H up / 34.7H wake / 4.9H tall / +-3.2H lateral; frontal area 0.1120 m^2, blockage = 4.28%
wall model: Werner-Wengle (high-Re); inflow turbulence intensity TI=0.25% (LSTM tunnel; real Re_H = U h/nu = 768000)
velocity scale: U_inf = 40 m/s = ulat 0.050 -> 1 lattice unit = 800.0 m/s (u_phys = u_lat * 800.0); inflow uniform Ux=40 m/s
body-following AMR: 2 levels (coarse + 1 nested boxes), body wall = Bouzidi interpolated bounce-back. Force on the finest level (dx=0.00900 m = 2.0x coarse, effective nh=32):
L1 (2x): grid 148x74x52 = 0.57M cells (dx=0.00900 m), solid=153088
totals: coarse 6.76M + fine 0.57M = 7.33M cells
step 0/4000 t_conv=0.00 Cd=12.6227 Cl=4.7612 umax=0.050 (40.0 m/s) | 0.4s 0 step/s 0 MLUPS
step 100/4000 t_conv=0.01 Cd=0.4908 Cl=-1.7510 umax=0.074 (59.0 m/s) | 0.9s 173 step/s 797 MLUPS
step 200/4000 t_conv=0.01 Cd=1.0710 Cl=-0.9305 umax=0.071 (56.9 m/s) | 1.6s 156 step/s 953 MLUPS
step 300/4000 t_conv=0.02 Cd=1.0670 Cl=0.7135 umax=0.071 (56.4 m/s) | 2.2s 172 step/s 1047 MLUPS
step 400/4000 t_conv=0.02 Cd=0.1317 Cl=-0.2852 umax=0.070 (55.9 m/s) | 2.8s 156 step/s 1077 MLUPS
step 500/4000 t_conv=0.03 Cd=0.6449 Cl=-0.7614 umax=0.070 (56.0 m/s) | 3.4s 172 step/s 1116 MLUPS
step 600/4000 t_conv=0.04 Cd=0.9216 Cl=0.1943 umax=0.070 (56.2 m/s) | 4.0s 156 step/s 1127 MLUPS
...
step 3600/4000 t_conv=0.22 Cd=0.6363 Cl=0.3952 umax=0.086 (68.5 m/s) | 22.7s 154 step/s 1201 MLUPS
step 3700/4000 t_conv=0.22 Cd=0.6106 Cl=0.4382 umax=0.084 (67.2 m/s) | 23.3s 170 step/s 1203 MLUPS
step 3800/4000 t_conv=0.23 Cd=0.6043 Cl=0.4587 umax=0.085 (68.4 m/s) | 23.9s 155 step/s 1203 MLUPS
step 3900/4000 t_conv=0.23 Cd=0.6601 Cl=0.4456 umax=0.085 (68.4 m/s) | 24.5s 170 step/s 1205 MLUPS
checkpoint: wrote ahmed_ckpt.bin (833x104x78)
step 4000/4000 t_conv=0.24 Cd=0.5868 Cl=0.6361 umax=0.084 (66.8 m/s) | 25.2s 145 step/s 1202 MLUPS
checkpoint: wrote ahmed_ckpt.bin (833x104x78)
wrote ahmed_profiles.csv (centreline Ux(z) at 16 stations). Time-mean over 2001 samples, steps [2000,4000] = t_conv [0.12,0.24] domain-passes = 6 H/U convective times (warm-up 6 H/U).
wrote ahmed.csv + ahmed.vti + ahmed_fine.vti (finest level). Velocity in m/s (u_lat*800.0); umax_ms column added to the CSV.