Skip to content

mf-tee2d

English | Japanese

« Home · « Multi-fidelity (1D ⇄ FVM ⇄ LBM)

MF-5: a 2D dividing tee, solved three ways -- the BRANCHING axis of the multi-fidelity ladder. --------------------------------------------------------------------------- Every multi-fidelity case so far has been a SERIES line: one flow, one path, one coefficient at a time. A tee adds the axis those cannot exercise -- the flow SPLIT is itself unknown, and it is what the junction coefficients decide. So the same geometry is solved:

Build & run

cmake -S native -B native/build && cmake --build native/build -j
./native/build/orheo-mf-tee2d

mf_tee2d <config-file> (KEY = VALUE lines; NOT KEY=VALUE argv)

Cost of the computation that made the figures on this page (Miyabi-G Grace CPU (72 cores, -stdpar=multicore)): 35.9 s wall.

Validation

Tier T1 — automatic gate
Gate ✅ direct gate

Source & run details

View source

Item Value
Target orheo-mf-tee2d
Source native/benchmarks/mf_tee2d.cpp
Lattice D2Q9
GPU twin no
Modules channel2d.hpp, stations.hpp
Gallery run 2026-07-28 @ c9b0572c
Args h=12 L1=32 Ls=50 Lb=34 steps=25000 dn_start=20 up_lo=30 up_hi=2 bias=0.004
Wall time 35.87 s
CPU time 2571.84 s
Device Miyabi-G Grace CPU (72 cores, -stdpar=multicore)

Computation log

The actual output of this computation (head and tail).

=== mf_tee2d: h=12  trunk 32 h / straight 50 h / branch 34 h  (996 x 420 = 0.42 M cells)
    Re_h=100  ubulk=0.0400  tau=0.5144  nu=0.004800  steps=25000
    window rule: legs measured from 20 h past the box over 8 h
    K convention: measured and injected on the COMBINED head, BULK column; energy column reported for judgement

[measure] biased runs (the K(beta) table)
  bias=+0.0040  beta=0.4778  V_c=0.04060  K_cb=+0.6872/+1.1099  K_cs=-0.1188/+0.2802  (bulk/energy)
        alpha c/s/b = 1.536/1.537/1.536 (54/35=1.543)   mass=0.092%   [I] 1.0e-02/1.0e-02
        [I] decomposition -- measured (p_tot_e-p_tot)/q_c vs (alpha-1)vsq:  trunk 0.5472/0.5362  straight 0.1482/0.1474  branch 0.1244/0.1240
        per-leg relative: trunk 2.04%  straight 0.52%  branch 0.39%  (trust <= 3%)
        developed grad/alpha  trunk 1.031/0.996  straight 1.024/0.996  branch 1.030/0.996   TRUSTED
  bias=-0.0040  beta=0.6320  V_c=0.04061  K_cb=+0.6290/+0.9570  K_cs=+0.0345/+0.5061  (bulk/energy)
        alpha c/s/b = 1.536/1.537/1.536 (54/35=1.543)   mass=0.063%   [I] 9.7e-03/8.3e-03
        [I] decomposition -- measured (p_tot_e-p_tot)/q_c vs (alpha-1)vsq:  trunk 0.5449/0.5362  straight 0.0733/0.0729  branch 0.2169/0.2179
...


[iii] whole LBM (reference)  beta=0.5548  dp=4.9867e-03  TRUSTED

[P3] split      handbook -0.0140 abs  (reported, NEVER gated -- the spread IS the result)
     injected   -0.0003 abs
     dp: handbook +5.88%   injected -2.54%

wrote ./mf_tee2d.csv
RESULT h=12 re=100 beta_lbm=0.554784 beta_handbook=0.540779 beta_injected=0.554465 dp_lbm=4.986659201e-03 dp_handbook=5.279811801e-03 dp_injected=4.860059773e-03 npts=3 ref_trusted=1