コンテンツにスキップ

powerlaw-poiseuille2d

英語 | 日本語

« Home · « 層流・解析解による検証

D3Q19 power-law (Ostwald-de Waele, non-Newtonian) plane Poiseuille flow -- a body-force-driven laminar channel whose LOCAL viscosity depends on the local shear rate, nu_app = K_c * gammadot^(n-1). This is the classic non-Newtonian LBM verification case (OpenLB gallery menu: laminar/powerLaw2d / nonNewtonianPoiseuille). CLEAN-ROOM reimplementation: no OpenLB (GPLv2) source read or ported; only the choice of case was borrowed. Thin periodic streamwise (x) and spanwise (y), halfway bounce-back no-slip walls at the two z faces, uniform body force G in +x, BGK collision with Guo (2002) forcing. The relaxation time tau(x) is set per cell every step from the locally measured shear rate.

概要: 冪乗則(Ostwald-de Waele)非ニュートン流体のチャネル流。局所せん断速度から見かけ粘性を毎ステップ決める。n<1 でプラグ状、n>1 で尖った分布になることを解析解と照合する。

Geometry

powerlaw-poiseuille2d 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。

powerlaw-poiseuille2d umag

powerlaw-poiseuille2d ux

Build & run

cmake -S native -B native/build && cmake --build native/build -j
./native/build/orheo-powerlaw-poiseuille2d

powerlaw_poiseuille2d [N [Re [steps [csv]]]]

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)): 8.3 s wall。

Validation

Tier T1 — 自動ゲートあり
QoI velocity-L2-error(abs)
Reference analytic Ostwald-de Waele power-law channel profile
Gate 🧩 モジュールゲート (closures)

Source & run details

View source

Item Value
Target orheo-powerlaw-poiseuille2d
Source native/benchmarks/powerlaw_poiseuille2d.cpp
Lattice D3Q19
GPU twin no
Modules lbm_d3q19.hpp
Gallery run 2026-07-27 @ 7628c0cf
Args ``
Wall time 8.28 s
CPU time 590.22 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)。

=== power-law plane Poiseuille (D3Q19, local shear-dependent tau) ===
tau clamp band [0.51, 2.00]; body force set for u_max(target)=0.100

powerlaw n=0.50  N=64 Re=100.0 steps=200000  Kc=2.5298e-03 Gx=7.6547e-06 u_max(exact)=0.1000
    step      0: u_centre=0.000011  d=1.00e+09
    step  20000: u_centre=0.066616  d=6.66e-02
    step  40000: u_centre=0.086962  d=2.03e-02
    step  60000: u_centre=0.094736  d=7.77e-03
    step  80000: u_centre=0.097851  d=3.11e-03
    step 100000: u_centre=0.099119  d=1.27e-03
    step 120000: u_centre=0.099639  d=5.20e-04
    step 140000: u_centre=0.099853  d=2.14e-04
    step 160000: u_centre=0.099941  d=8.79e-05
    step 180000: u_centre=0.099977  d=3.62e-05
...

    step 180000: u_centre=0.099866  d=9.88e-15
    step 199999: u_centre=0.099866  d=3.25e-15
  n=1.50  u_max: lbm=0.09987 exact=0.10000 (-0.13%)  L2=1.602e-04  tau in [0.5876, 0.8487]

=== summary (L2 profile error vs Bird-Stewart-Lightfoot) ===
  n=0.50  L2=6.749e-04  tau in [0.5796, 2.0000]  shear-thinning (blunt/plug centre)
  n=1.00  L2=2.686e-04  tau in [0.6920, 0.6920]  Newtonian (parabola)
  n=1.50  L2=1.602e-04  tau in [0.5876, 0.8487]  shear-thickening (sharp centre)
PASS (all profiles match the analytic power-law solution)