Skip to content

The ILES baseline ​

IDMilestoneTypeLegsBackendsFigure
TT, HT#119 · T0.1–T0.6P+Ctt1–tt6, ht1–ht2CPU, FNLyes

Case and scope ​

T0 measures what the scheme alone does to turbulence, with no model: the implicit LES (ILES) baseline. The cases are the Taylor-Green vortex at Mach 0.1 (T0.3) and at M0=1.25 (T0.4), the same vortex with a static 2:1 box (T0.6) and decaying compressible isotropic turbulence (T0.5). The legs check first the diagnostics they rest on (T0.1) and the initial conditions (T0.2, T0.5). The diagnostics and the keys are on the turbulence page.

Legs ​

src/tests/flume/verification/turbulence/check.sh, CPU and FNL.

LegPhaseWhat it verifiesMeasured
tt1T0.1, T0.2first history row of taylor-green against the exact integrals, M 0.1, 1.25K 10−16; Z and ε order 3.98–4.00; εd, pd 10−30
tt2T0.1one rank against 26⋅10−14
tt3T0.3energy budget to t=2, Re 1600, M 0.1: what the integrals leave of the K drop70 %, 8.8 %, 0.28 % at 323, 643, 1283
tt4T0.1spectra.py on the initial field: Parseval, the energy of the single-mode field in shell 2not recorded (bound 10−12)
tt5T0.1(--agree TAG) this executable's tt2 history against the executable TAG: backendsnot recorded (bound 10−11)
tt6T0.1pressure work, M 0.8: the budget closes only with pd8.5 % left, 203 % without, 414 % flipped
ht1T0.5isotropic-turbulence field: every shell, K, dilatationshells 7⋅10−15, K 1.5⋅10−14
ht2T0.5one rank against 2 on the random field7⋅10−15

Taylor-Green vortex ​

Setup
box[0,2π]3, periodic
flowRe 1600; M 0.1 and M0=1.25
schemesweno (flux splitting, WENO-5, 7, 9); HLLC: weno-riemann with the 6th-order correction and the WENO sensor
timerunge-kutta-ssp-54, CFL 0.5, weights = si
runturbulence/tgv_campaign.sh, 4 ranks on A30; all values per unit volume

The ILES baseline on the Taylor-Green vortex

Top: M 0.1, WENO-5 by resolution; middle: M 0.1 at 1283 by scheme and order; bottom: M0=1.25 by resolution. Left −dK/dt, right the enstrophy; dashed: the 512³ spectral DNS, at M 0.1.

CasePeak −dK/dtat tPeak Zat tNumerical share to the peakK(20)
M 0.1, WENO-5, 6430.013775.081.716.2577 %0.0245
M 0.1, WENO-5, 12830.012428.243.408.1158 %0.0219
M 0.1, HLLC, 6430.013055.421.836.5676 %0.0284
M 0.1, HLLC, 12830.012498.283.598.1155 %0.0227
M 0.1, WENO-7, 12830.012248.193.178.1460 %0.0181
M 0.1, WENO-5, 25630.012728.756.188.5627 %0.0223
M 0.1, WENO-9, 12830.012998.284.458.0343 %0.0213
M0 1.25, WENO-5, 6430.013837.121.817.6280 %0.0206
M0 1.25, WENO-5, 12830.010818.623.349.0946 %0.0243
M0 1.25, WENO-5, 25630.0110111.125.9511.9420 %0.0245
M0 1.25, HLLC, 6430.012467.222.297.7269 %0.0299
M0 1.25, HLLC, 12830.011208.563.8611.4436 %0.0265
M0 1.25, HLLC, 25630.0112211.116.3111.9914 %0.0243

The share is that of the energy lost until the peak of −dK/dt not accounted for by ε−pd.

Reference. The 512³ pseudo-spectral DNS of the High-Order CFD Workshop (case C3.5, turbulence/reference/spectral_Re1600_512.gdiag), incompressible, so compared at M 0.1 only: peak −dK/dt 0.01286 and peak Z 10.29, both at t=8.97. The L2 errors against it are in Against the DNS. Mach 1.25 has no reference here.

What it shows.

  • Numerical dissipation dominates. Up to the peak of −dK/dt it is 77 % of the energy lost at 643 and 58 % at 1283 with WENO-5 at M 0.1.
  • The resolved gradients are far from converged. At 1283 the enstrophy peak is a third of the reference, even where the total −dK/dt looks right. At 643 the peak of −dK/dt comes at t=5, against 9.
  • Resolution matters more than scheme and order at 1283. HLLC gains a few percent, WENO-9 reaches Z 4.45. WENO-7 is below WENO-5 (Z 3.17 against 3.40); the cause is not established.
  • At M0=1.25 the numerical share falls with the resolution: 80 %, 46 %, 20 %.

A static 2:1 box ​

Setup
gridbase 643; the box of side π on the centre refined to the 1283 spacing (one eighth of the volume)
flow, schemeM 0.1, WENO-5, as above
seam operatorstricubic fill with mean restriction; quintic fill with lagrange6 restriction

The models and the 2:1 box

Bottom row: the 2:1 box on the centre of the 643 grid against the uniform grids (the upper rows are the subgrid-scale models).

CasePeak −dK/dtat tPeak Zat tK(20)max|pd|
uniform 6430.013775.081.716.250.02450.00026
box, tricubic fill, mean restriction0.013126.391.876.390.02330.0070
box, quintic fill, lagrange60.013146.391.856.380.02340.0115
uniform 12830.012428.243.408.110.02190.00007

What it shows.

  • The box delays and lowers the peak of −dK/dt, from t=5.1 to 6.4.
  • The two seam operators give the same integrals: −dK/dt and Z do not tell them apart.
  • pd cannot be read on a 2:1 grid. It is 27 to 44 times the uniform value. A central difference across a 2:1 face does not telescope, so ∫pθdV picks up the mean pressure (p≈71 at M 0.1) times a nonzero discrete ∫θdV. This explanation is untested. The budget and the numerical share are therefore not reported on a 2:1 grid.

Decaying compressible isotropic turbulence ​

Setup
fieldisotropic-turbulence, Mt=0.6, Reλ=100, k0=4, one seed
gaspower-law viscosity 0.76, Pr 0.7
time scaleτ=λ0/urms=0.5
referencenone: the DNS of Johnsen et al. (2010) is not available to this tree, so only self-convergence is measured

Isotropic turbulence

Left: the enstrophy against t/τ; right: the spectrum of ρu at t/τ=4.

CasePeak Zat t/τK(t/τ=4)E(16) at t/τ=4E(30)
no model, 64334.50.660.3037.4⋅10−48.3⋅10−6
Vreman, 64332.60.580.2925.9⋅10−45.8⋅10−6
no model, 128351.01.140.3843.0⋅10−31.0⋅10−4
Vreman, 128348.91.120.3762.7⋅10−38.1⋅10−5
no model, 256360.51.360.4134.2⋅10−32.3⋅10−4

What it shows.

  • 2563 with WENO-5 is not converged. K changes by 7 % from 1283 to 2563.
  • Vreman moves away from the finer runs at each resolution, as on the Taylor-Green vortex (subgrid-scale models).

The vorticity of both flows on the hybrid scheme is in The flow.

Not done ​

  • HLLC at 2563, M 0.1.
  • Spectra against the DNS: the Workshop file holds integrals only.
  • The DNS of Johnsen et al. for the isotropic case.
  • The pressure dilatation on a 2:1 grid: the non-telescoping explanation is not tested.