コンテンツにスキップ

sphere-drag

英語 | 日本語

« Home · « 層流・解析解による検証

Canonical drag validation: uniform flow past a sphere at Re = U D / nu. A single-level D3Q19 MRT solve (no AMR, no wall model, no LES) whose ONLY purpose is to pin the sign AND magnitude of the momentum-exchange (Ladd) force used by the Ahmed body case -- the identical formula F = sum over body-adjacent links c_i * (f_i + f_iopp) (post-collide) Cd = 2 F / (rho U^2 A), A = pi R^2 (sphere frontal area) At Re = 100 the sphere wake is steady and the drag is well known: Clift-Gauvin: Cd = 24/Re (1 + 0.15 Re^0.687) = 1.09 (Re=100). A correct MEM must converge to Cd ~= +1.09 (POSITIVE, streamwise). Any sign flip or O(10-100) magnitude error here is the same bug seen in ahmed_body.

概要: 一様流中の球の抗力係数。Clift-Gauvin 相関(Re=100 で Cd≈1.09)に対し 1.95% 一致。運動量交換法の符号と大きさを固定する基準ケースで、この値がずれると外部流れ全体が疑わしくなる。

Build & run

cmake -S native -B native/build && cmake --build native/build -j
./native/build/orheo-sphere-drag

Validation

Tier T2 — 単点比較
QoI Cd
Reference Clift-Gauvin 相関, Re=100 -> Cd~1.09
Gate — ゲートなし
Measured 1.1113 (rel. err +1.95%) — cross-code run (sphere_drag)

Source & run details

View source

Item Value
Target orheo-sphere-drag
Source native/benchmarks/sphere_drag.cpp
Lattice D3Q19
GPU twin yes (orheo-sphere-drag-gpu)
Modules ahmed_solver.hpp, lbm_d3q19.hpp
Gallery run まだ実行していない(図・ログなし)