Skip to content

Plane channel and the wall model ​

IDMilestoneTypeLegsBackendsFigure
WM#119 · T3.1–T3.6P+Cwm1–wm12CPU, FNLyes

Case and scope ​

T3 adds the channel infrastructure (mass-flow forcing, isothermal walls, T3.1), the wall-modelled boundary (T3.3) with the equilibrium model (T3.4) and the inverse-velocity-transformation model (T3.5), its input filter, the eddy viscosity below the matching height, and 2:1 faces along a modelled wall (T3.6). The methods are on the turbulence page. The legs check each piece against a laminar exact solution, a restart or a rank change. The campaign runs the compressible channel of Coleman, Kim & Moser (1995) at bulk Mach 1.5 and 3, against a wall-resolved reference (T3.2) and the DNS.

Legs ​

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

LegPhaseWhat it verifiesMeasured
wm1T3.1laminar channel, isothermal walls, Re 100: force against 3/Re; one rank against 2order 1.99, 4.8⋅10−4 at 64 across; 4⋅10−11
wm2T3.3, T3.4test_wall_model; laminar channel on modelled wallssee unit tests; order 1.84, 1.9 % at 64
wm3T3.6half the wall refined, 2:1: force between the coarse and the fine uniform runs0.03 % from their mean, mass 10−16
wm4T3.3hot wall, modelled against resolved: energy change; one rank against 2 on modelled walls7.8 % (bound 10 %); 4⋅10−11
wm5T3.4matching height 0.3, laminar channel at Re 5 against the exact slip solution 3/Re/(1+3h2/2); semi-local against classical damping6.4⋅10−3, 2.2⋅10−3, 3.0⋅10−5 at 16, 32, 64; 4⋅10−7
wm6T3.5semi-local against classical damping on a hot wall, Re 100force 0.7 % apart (min 0.01 %)
wm7T3.5input filter: Tf→0 against no filter; Tf=0.05 near the steady state; Tf=1 over a transient; one rank against 20 (bit for bit); 3.0⋅10−4 (bound 10−3); 1.4⋅10−2 (min 10−3); 4⋅10−11 (CPU), 10−13 (FNL); CPU against FNL 5⋅10−11
wm8T3.5inverse-transformation model: laminar channel against the equilibrium model; hot wall; one rank against 2 with the filter1.0⋅10−4, the heat flux of the Reynolds analogy moving the energy by 2.3⋅10−3 (bound 10−4); 9.8⋅10−3 (min 10−3); 4⋅10−11 (CPU), 0 (FNL); CPU against FNL 4⋅10−11
wm9T3.5restart with the filter averages (.wmf, one per rank): 10 steps + restart to 20 against 20 straight, inverse-transformation model, Tf=0.01; control with the .wmf removedfields and histories bit for bit (CPU and FNL); the control differs and warns
wm10T3.1 (#49 NV-5, NV-9)dimensional units on a Vreman channel with the wall model and the forcing, gas cp, cv: input bit for bit (NV-5), output against scaling.py dimensionalize (NV-9), forcing and turbulence histories includedbit for bit (CPU and FNL); bit for bit, 4 histories
wm11Mach 3 channelthe wall model's eddy viscosity below the matching height (wall_model_eddy_viscosity), perturbed channel at Re 1000, h=0.3: it acts; 2 against 4 blocks per direction; one rank against 2; the same with Vremannot recorded (bounds: force change above 10−3; blocks within channel_oracle agree's bounds, 2.4⋅10−6 against 10−9 with the ghost columns left out)
wm12#121 S4dthe pressure gradient in the equilibrium model (wall_model_pressure_gradient) on wm11's perturbed channel with the model's eddy viscosity: it acts; 2 against 4 blocks per direction (the gradient's central difference reads the ghost cells along the wall); one rank against N, CPU against GPU (--agree); its filter wall_model_pressure_gradient_filter =0.01 acts; the filtered gradient across a restart (the 6th average of the .wmf) bit for bit; a .wmf of another layout restarts the averages with a warningnot recorded (bounds: force change above 10−6; blocks and ranks within channel_oracle agree's bounds; the restart bit for bit)
wm14#126the viscosity coupling (wall_model_coupling = viscosity) on wm11's perturbed channel: it acts; an explicit flux against the default; 2 against 4 blocks per direction; one rank against 2; CPU against GPU (--agree); on wm5's laminar channel the history's τw against the driving force at CFL 0.5 and 0.25 (the forcing adds its momentum after the Runge–Kutta stages, so the stages' first cell lags the committed one by about fΔt and a stress linear in it lags too)force 0.38 % from the flux coupling's; 0 (bit for bit); 1.0⋅10−12; 2.2⋅10−11; CPU against FNL 1.3⋅10−10; offsets 3.78⋅10−3 and 1.87⋅10−3, ratio 2.02, extrapolated to Δt→0 −3.6⋅10−5 (bound 10−3)

Mach 1.5 ​

Setup
caseColeman, Kim & Moser (1995): bulk Mach 1.5, bulk Re 3000 on the half height and the wall viscosity
box4π×2×43π, periodic in x and z, mass-flow forcing
gas, wallsviscosity (T/Tw)0.7, Pr 0.7; isothermal walls
schemehybrid-kep-weno
referenceReτ about 220–222 for the DNS, as cited by Qi et al. (2022), Cao et al. (2024), Alberti et al. (2025b)
runturbulence/channel_campaign.sh with channel_cases_t35.txt, 4 A30
measuresτw: the mean force per unit volume times h; Reτ=3000τwρw with ρw=1.30 for every row, so the rows compare among themselves (the reference run's own wall density, extrapolated from its last checkpoint, is 1.25, which lowers every Reτ by 2 %); uc, Tc/Tw: the Favre plane means at the centreline of the last checkpoint
inverse modeledge state given: ue, Te/Tw=1.29, the centreline of the h=0.15 semi-local runs

The plane channel

τw against time; dashed: Reτ 220 and 222; dotted: the laminar value.

CaseCellsτw, 100<t≤150150<t≤200Reτuc, Tc/Tw
resolved walls256×192×128, Δy+≈2.33.82e-3–211 (50<t≤100: 218)1.16, 1.26
resolved walls64×32×322.35e-32.72e-3166, 1781.22, 1.26
equilibrium, 3rd cell64×32×324.37e-34.49e-3226, 2291.21, 1.29
equilibrium, 3rd cell96×48×484.36e-34.52e-3226, 2301.19, 1.30
equilibrium, 3rd cell128×64×644.69e-34.66e-3234, 2341.18, 1.31
equilibrium, 3rd cell, semi-local128×64×644.33e-34.33e-3225, 2251.18, 1.30
equilibrium, h=0.1596×48×484.71e-34.77e-3235, 2361.19, 1.30
equilibrium, h=0.15128×64×644.89e-34.90e-3239, 2391.19, 1.30
equilibrium, h=0.15, semi-local96×48×484.15e-34.09e-3220, 2191.18, 1.29
equilibrium, h=0.15, semi-local128×64×644.13e-34.21e-3220, 2221.17, 1.29
… and the filter Tf=196×48×484.15e-34.03e-3220, 2171.18, 1.29
… and the filter Tf=1128×64×644.15e-34.16e-3220, 2211.19, 1.29
inverse, h=0.15, ue=1.18128×64×644.13e-34.08e-3220, 2181.16, 1.21
… and the filter Tf=1128×64×644.04e-34.11e-3217, 2191.17, 1.21
inverse, h=0.15, ue=1.1896×48×482.03e-31.62e-3154, 138: relaminarises1.51, 1.11
inverse, 3rd cell, ue=1.1896×48×481.61e-31.40e-3137, 128: relaminarises1.44, 1.12
inverse, h=0.15, ue=1.1096×48×483.61e-34.15e-3206, 2201.18, 1.21

Every run conserves mass to 10−16.

What it shows.

  • Matching height 0.15 with semi-local damping puts the equilibrium model in the DNS band on both grids (219–222). Either change alone is worse: the 3rd-cell sample is 226–234 and rises with the grid, h=0.15 with the classical damping 235–239, the semi-local damping at the 3rd cell 225. The fixed height removes the grid trend; the semi-local damping removes the excess. This is one Mach and Reynolds number, single runs, and the band sits within the ±2% of the wall density.
  • The input filter does not move the means (within 1.5 %); on the coarser grid it slows the transient.
  • The inverse model is in the band on the finer grid with a cooler centreline (Tc/Tw 1.21 against 1.29 for the equilibrium model, 1.26 for the resolved run): its heat flux differs.
  • On 96×48×48 the inverse model relaminarises with ue=1.18, at h=0.15 and at the 3rd cell, but not with ue=1.10. The model is sensitive to the given edge velocity, which the flow does not set (the paper takes the edge from the flow; a face kernel cannot).
  • The wall-resolved reference is 4–5 % short of the DNS (211). The modelled rows in the band agree with the DNS, not with it.
  • The coarse grid with resolved walls falls 20 % short: without a model the grid cannot carry the wall layer.

Mach 3 ​

Coleman, Kim & Moser's second case: bulk Mach 3, bulk Re 4880, Reτ about 451 for the DNS; the rest as at Mach 1.5, the equilibrium model only.

Result. With the wall model's eddy viscosity below h (wall_model_eddy_viscosity) and run to a thermal steady state at t=800, h=0.15 with semi-local damping gives Reτ=443–445 against about 451 for the DNS; the classical damping overshoots by 13–21 %. Without the key the first cell drains, as the runs to t=200 show.

Without a flux path below h, to t=200 ​

Setup
runchannel_cases_t3_mach3.txt, to t=200; the first-cell history from channel_cases_t3_mach3_drain.txt (the same runs, bitwise, a checkpoint every 500 steps)
ρw, Reτthe wall density changes with the run at this Mach number, so each row takes its own: ρw is the first-cell mean pressure of the last checkpoint over Tw, Reτ=4880τwρw
T1/Twthe first cell (Favre, both walls)
Qw/Wwall heat flux over forcing work for 180<t≤200, from the total-energy budget (channel_wall_cells.py budget; Qw is the residual): 1 at a thermal steady state

The plane channel at Mach 3

Left: τw against time (dotted: the laminar value). Middle and right: the first cell of the two h=0.15 runs on 128×64×64 against time, from a checkpoint every 500 steps.

CaseCellsτw, 100<t≤150150<t≤200ρwReτuc, Tc/TwT1/TwQw/W
equilibrium, 3rd cell96×48×481.77e-31.43e-31.65264, 238: relaminarises1.43, 1.630.860.71
equilibrium, h=0.15, semi-local96×48×481.89e-31.59e-31.54263, 241: relaminarises1.50, 1.530.640.67
equilibrium, 3rd cell128×64×644.29e-34.44e-32.12465, 4731.25, 2.221.030.87
equilibrium, 3rd cell, semi-local128×64×643.20e-33.03e-32.02392, 3811.20, 2.041.280.85
equilibrium, h=0.15, semi-local128×64×643.58e-33.46e-31.91404, 3971.28, 2.010.810.83
equilibrium, h=0.15128×64×646.46e-37.49e-32.12571, 615: still rising1.48, 2.630.410.88

What it shows.

  • No run is at a thermal steady state at t=200: the total energy still rises in every run (Qw/W 0.67–0.88), so these Reτ are values on the way, not means.
  • The Mach 1.5 setting does not carry over: h=0.15 with semi-local damping is 12 % below the DNS; the 3rd-cell sample is 5 % above. The coarse grid relaminarises with either setting.
  • The first cell is drained. The wall face takes the modelled τw and qw from the sample at h; the first cell's own state never enters, and nothing below h carries those fluxes (no subgrid model): the molecular plus resolved stress at the face between cells 1 and 2 is 0.54 τw with semi-local damping and 0.21–0.39 with the classical one (after t=30).
  • With h=0.15 and the classical damping the run does not settle. The first cell falls to 0.26Tw, reverses its flow and holds 17 % of the mass at t≈80; τw follows it, down (t 80–110) and up again.
  • With semi-local damping the drain is slower and τw stays flat, but the first cell is still below Tw from t≈125. A first cell colder than an isothermal wall that takes heat out of it is not a boundary layer: the treatment below h fails, not the model's law.

With the eddy viscosity below h, to t=800 ​

Setup
keywall_model_eddy_viscosity = .true.: μt=max(μt,SGS, κyρτwD2) in the cells whose centre is below h (turbulence)
runsthe three 128×64×64 inputs above plus the key, to t=200 (channel_cases_t3_mach3_ev.txt), continued from the restart to t=800 (channel_cases_t3_mach3_ev800.txt)
measuresρw and T1/Tw at the last checkpoint as above; Qw/W over 700<t≤800

The plane channel at Mach 3 with the eddy viscosity below h

Left: Reτ against time, with the wall density of each checkpoint (dashed: the DNS). Middle: the first cell. Right: the wall heat flux over the forcing work in 50-unit windows, 1 at a thermal steady state.

Caseτw, 600<t≤700700<t≤800ρwReτuc, Tc/TwT1/TwQw/W
equilibrium, h=0.15, semi-local3.52e-33.48e-32.36445, 4431.17, 2.441.790.99
equilibrium, 3rd cell4.51e-34.35e-32.51519, 5101.21, 2.651.570.99
equilibrium, h=0.155.17e-35.15e-32.44548, 5471.21, 2.531.841.00

What it shows.

  • The drain is gone. The first cell is 1.6–1.8Tw at every time after the transient, the two walls agree within 2 %, and the two first-cell layers hold 4–5 % of the mass (3.1 % if uniform), against 16 % at 0.41Tw without the key.
  • The runs are thermally steady at t=800: Qw/W is 0.99–1.00 and dE/dt about 60 times smaller than at t=200.
  • h=0.15 with semi-local damping matches the DNS within 2 %. With the classical damping Reτ is 13 % (3rd cell) and 21 % (h=0.15) above.
  • Reτ needs the thermal steady state at this Mach number. The bulk heats until t≈600, ρw rises from about 2.1 to 2.4–2.5, and Reτ rises with it: at t=200 the 3rd-cell run gave 438, 448, which looked like a match.
  • Cost. From the same t=800 state of the semi-local run to t=810 (4 A30), the key on takes 1.69 times the wall time of the key off: 1.28 times the steps (the eddy viscosity enters the viscous time-step limit) and 1.32 times the time per step.

The flow ​

Instantaneous streamwise velocity in the two channels

The streamwise velocity over the bulk velocity of the last checkpoint: the x-y plane at mid-span (left), the wall-parallel plane at the matching height h=0.15 above the lower wall (middle, its own colour range) and the cross-flow plane (right). Top: bulk Mach 1.5, 128×64×64, h=0.15 with semi-local damping, t=200. Bottom: bulk Mach 3, the same grid and model with the eddy viscosity below h, t=800. Planes from turbulence/field_slices.py; the figure is make_doc_figures.py --only channel-fields.

  • At the matching height the flow is in streaks along x, but outer-scaled ones. At Mach 3 the first minimum of the spanwise correlation of u in that plane puts them 1.05h apart, about 470 wall units at Reτ≈445, against the 100 of the near-wall cycle; at Mach 1.5 the correlation has no clear minimum within the span. The near-wall cycle lives below the matching height (y+≈33 at Mach 1.5), in the wall model, and the grid (Δz+≈14, Δx+≈22 at Mach 1.5) does not form it above.
  • At Mach 3 the streaks look longer and more regular than at Mach 1.5. One snapshot each: a description, not a measurement.

Not done ​

  • Averaged profiles against the DNS: only τw and the centreline are compared, and the DNS profiles are not in this tree.
  • The cause of the wall-resolved reference's 4–5 % shortfall: its resolution or the hybrid scheme's dissipation.
  • The relaminarisation of the inverse model on 96×48×48 with ue=1.18; an edge state taken from the flow.
  • Why the classical damping overshoots at Mach 3; a cheaper flux path below h.
  • Runtime regridding along a modelled wall.