taylor-couette2d¶
英語 | 日本語
Taylor-Couette flow -- laminar circular Couette flow between two concentric cylinders, the inner rotating at angular velocity Omega, the outer fixed. --------------------------------------------------------------------------- The classic viscous-flow verification: below the Taylor instability the steady azimuthal velocity has the EXACT laminar solution (Landau & Lifshitz, Fluid Mechanics sec. 18, "flow between rotating cylinders"; also any text):
概要: 同心二重円筒の層流解。u_θ = Ar + B/r という解析解と直接比較でき、曲面壁の Bouzidi 補間が効いているかを見る。
Geometry¶

Animation¶
Fields¶
この計算が最後の出力ステップで書き出した場をそのまま描いたもの。3D ケースは中央断面。


Build & run¶
cmake -S native -B native/build && cmake --build native/build -j
./native/build/orheo-taylor-couette2d
taylor_couette2d [config.txt] (KEY=VALUE; or all defaults)
このページの図を作った計算のコスト(Miyabi-G Grace CPU (72 cores, -stdpar=multicore)): 3.8 s wall。
Validation¶
| Tier | T0 — 収束次数まで確認 |
| QoI | velocity-L2-error(abs) |
| Reference | analytic Couette solution u_theta = A r + B/r; Taylor (1923) instability |
| Gate | — ゲートなし |
Source & run details¶
| Item | Value |
|---|---|
| Target | orheo-taylor-couette2d |
| Source | native/benchmarks/taylor_couette2d.cpp |
| Lattice | D3Q19 |
| GPU twin | no |
| Modules | lbm_d3q19.hpp |
| Gallery run | 2026-07-27 @ 7628c0cf |
| Args | `` |
| Wall time | 3.82 s |
| CPU time | 226.36 s |
| Device | Miyabi-G Grace CPU (72 cores, -stdpar=multicore) |
Computation log¶
この計算の実際の出力(先頭と末尾)。
Taylor-Couette 2D R1=20.0 R2=40.0 (eta=0.500 gap=20.0 cells)
grid 96x96 (nz=1, z-periodic), centre (48.0,48.0)
Omega=1.000e-03 U1=Omega*R1=0.0200 (Ma=0.035) Re=20.0 nu=2.0000e-02 tau=0.5600 MRT
analytic u_theta = A r + B/r : A=-3.33333e-04 B=5.33333e-01 ; torque=1.34041e-01
step 0 umax=0.00000e+00 |torque|=2.44987e+01 (exact 1.34041e-01, +18177.0%) d(umax)=1.00e+09
step 4000 umax=1.98654e-02 |torque|=1.48923e-01 (exact 1.34041e-01, +11.1%) d(umax)=1.99e-02
step 8000 umax=1.98962e-02 |torque|=1.39923e-01 (exact 1.34041e-01, +4.4%) d(umax)=3.08e-05
step 12000 umax=1.99006e-02 |torque|=1.38617e-01 (exact 1.34041e-01, +3.4%) d(umax)=4.32e-06
step 16000 umax=1.99011e-02 |torque|=1.38497e-01 (exact 1.34041e-01, +3.3%) d(umax)=4.89e-07
step 20000 umax=1.99011e-02 |torque|=1.38483e-01 (exact 1.34041e-01, +3.3%) d(umax)=5.72e-08
step 24000 umax=1.99011e-02 |torque|=1.38484e-01 (exact 1.34041e-01, +3.3%) d(umax)=9.27e-09
step 28000 umax=1.99011e-02 |torque|=1.38487e-01 (exact 1.34041e-01, +3.3%) d(umax)=2.70e-09
step 32000 umax=1.99011e-02 |torque|=1.38490e-01 (exact 1.34041e-01, +3.3%) d(umax)=2.00e-09
step 36000 umax=1.99011e-02 |torque|=1.38493e-01 (exact 1.34041e-01, +3.3%) d(umax)=1.88e-09
...
36.0 +0.002871 +0.002815 +1.98%
38.0 +0.001331 +0.001368 -2.71%
40.0 +0.000192 -0.000000 (wall, exact~0)
inner-cylinder torque: LBM |tau|=1.385274e-01 exact=1.340413e-01 (+3.35%)
L2(u_theta profile) = 3.3197e-02 (all bins) 3.5044e-02 (interior, wall bins excluded)
wrote taylor_couette.csv , tc_field.csv , tc_convergence.csv
PASS (laminar circular-Couette profile matches A r + B/r)