Skip to content

Shock/boundary-layer interaction ​

IDMilestoneTypeLegsBackendsFigure
— (campaign tbl_cases_s4.txt)#121 · S4b–S4lPnone (the pressure-gradient key: wm12)FNLyes

Case and scope ​

The interaction of the IUSTI experiment (Dupont, Haddad & Debiève 2006, JFM 559) as computed by the STREAmS DNS (Bernardini, Modesti, Salvadore & Pirozzoli 2021, CPC, Fig. 7): Mach 2.28, an oblique shock of 8∘ flow deflection on an adiabatic wall, Reθ≈5000. The layer is the T3.7 setup; the shock comes in through the top face, the shock generator of S3. Nothing is fitted to the interaction: the wall model is the one of the zero-pressure-gradient layers. The oracle is the interaction length, the separation from Cf and the wall pressure against the DNS and the experiment.

Setup ​

Setup
box, cells100×12×6δi, 1200×144×72, cubic δi/12
shockβ=32.685∘, entering the top at x=41.30 so that the inviscid shock reaches the wall at ximp=60, past the inflow's recovery; exact regular reflection: p2/p1=1.602, p3/p1=2.458, reflected at 29.77∘
gasRe=U∞δi/ν∞=35500, Sutherland's law (S1) with S/T∞=0.776 (110.4 K over the experiment's 142 K), Pr=0.71
wall, inflow, schemewall-modelled at Tw=1.918T∞ (the recovery temperature), the 3rd cell, semi-local damping, eddy viscosity below h; the M2p5 DNS profile with its temperature excess scaled by (2.28/2.5)2 (reference/tbl_m2p28_profile.txt); hybrid-kep-weno
runtbl_cases_s4.txt, to t=1200 on 4 A30, 0.74 s per step; statistics: the span and the 50 checkpoints of 310≤t≤934 (sbli_oracle.py compare --t0 300)
referencesthe DNS wall pressure and the IUSTI points of STREAmS Fig. 7a, whose x∗ is measured from the reflected shock (below); the STREAmS DNS at Reθ=6882 of Bernardini, Della Posta, Salvadore & Martelli (2022, arXiv:2204.10531) Fig. 4a; Volpiani, Bernardini & Larsson (2018, PRF 3) Fig. 12; digitised from the vector paths (reference/sbli_m2p28_*, error in the headers)

sbli_oracle.py takes δ0, δ∗ and θ of the undisturbed layer from a linear fit over [ximp−15,ximp−6] extrapolated to ximp; Cf from the wall model's τw; the reflected shock from the mid-pressure crossings of the rows 2.5<y/δ0<5, a line extrapolated to the wall: its apparent origin xref gives Lint=ximp−xref, the length STREAmS scales x with. It also reports the separation of the resolved flow, where the time-mean u of the first cell is negative, which τw misses when the reversed layer lies below the sample (below).

Results ​

The interaction: wall pressure and skin friction

Left: the mean wall pressure (pw−p∞)/(p2−p∞) against x∗=(x−ximp)/Lint, the DNS of STREAmS (line), the DNS at Reθ=6882 (dashed) and the experiment (triangles). Right: Cf; dotted the apparent origin of the reflected shock (x∗=−1), dashed the inviscid impingement. make_doc_figures.py --only sbli.

FLUME, wall modelReference
Reθ at ximp, nominal / against the edge state5247 / 4667about 5000 (DNS, experiment)
δ0, δ0∗, H at ximp−101.807, 0.531δi, 3.58
reflected shock: angle to the wall28.95∘29.77∘ inviscid
Lint/δ03.153.22, Volpiani et al. (DNS, Reθ=1050); 3.30, Bernardini et al. (2022, DNS, Reθ=6882)
separation: Lsep/δ00, Cf never negative, minimum 0.22⋅10−3 at x∗≈−0.251.62, Volpiani et al., Reθ=1050; 2.16, Bernardini et al. (2022), Reθ=6882; the experiment has a bubble
(pw−p∞)/(p2−p∞) at x∗=−1, −0.5, 0, 0.5, 1, 20.02, 1.12, 1.72, 2.09, 2.24, 2.37DNS: 0.64, 1.75, 2.07, 2.22, 2.29, 2.36; at Reθ=6882: 0.87, 1.65, 2.01, 2.18, 2.27, 2.35
Souverein L∗=Lint/(δ∗g3), separation parameter Se1.93, 1.20

The flow ​

The interaction's flow field

One checkpoint (t=934). From the top: numerical schlieren of the mid-span plane; the spanwise-averaged pressure with the sonic line (white), the u¯=0 line (red, absent) and the oracle's shock fits (orange, its crossings as dots); u in the wall-parallel plane y=0.29 (0.16δ0; dashed the reflected shock's apparent origin, dotted ximp); the cross-flow plane at ximp and spanwise-averaged profiles. make_doc_figures.py --only sbli-fields.

The Mach 2.5 boundary layer and the interaction in 3-D

A cutaway, u/U∞, y drawn stretched twice: the Mach 2.5 boundary layer over 15<x<45 (top), and this interaction over 35<x<60, the cross-flow face at the impingement x=60 (bottom). make_doc_figures.py --only tbl-3d.

What it shows ​

  • The interaction length is right. Lint/δ0=3.15 against 3.22 in the DNS of Volpiani et al. (2018, at Reθ=1050), and the reflected shock leaves at 28.95∘, 0.8∘ under the inviscid angle. The downstream pressure plateau is the DNS's within 4 % (x∗=2).
  • The mean flow does not separate. Cf dips to 0.22⋅10−3 and recovers; the DNS, the experiment and Volpiani et al. have a bubble. It comes close: in one checkpoint the spanwise-averaged u of the first cell falls to 0.003U∞ between x=55 and 65, and the wall-parallel plane at 0.16δ0 has patches of u<0 (down to −0.14U∞) between x=57.7 and 61.9. The wall model's τw follows the velocity at its matching height, the 3rd cell, whose mean stays positive.
  • The wall pressure rises late. It starts at x∗≈−0.75 and is half-way to p2 at −0.62; the DNS are there at −1.25 and −1.09, in the separation that the DNS have and this run has not. Downstream of the impingement the pressure is the DNS's within 4 %. (The first version of this page read the reference's x∗ from ximp, as the STREAmS paper's text says, and found the rise early: the figure measures it from the reflected shock, below.)
  • Two candidate causes of the missing separation, tested since. The equilibrium model has no pressure-gradient term, and it computes τw from the velocity at the 3rd cell, 0.21δi or 0.12δ0 above the wall, which the near-wall deceleration under the adverse gradient reaches late. S4c halves the matching height: the dip of Cf halves, the mean does not separate. S4d adds the pressure gradient: a bubble of 0.6δ0, the compression unchanged.

A refined wall grid under the interaction ​

Run S4c is the S4b case with one more level, 2:1, under the interaction (tbl_cases_s4c.txt). The wall model and its matching point (the 3rd cell) are unchanged, so in the refined box the matching height halves with the cells. A bubble in this run points at the near-wall grid; none points at the equilibrium model.

Setup
gridS4b's box and base grid (δi/12) with a static 2:1 box: make_tbl.py --refine-box 30,75,2.5 refines a block when its centre is inside 30≤x≤75, y≤2.5, which takes the blocks of 25≤x≤75, y<3, every z, to δi/24; 120 blocks (64 fine, 56 coarse), about 23 M cells
matching heightthe 3rd cell: 0.104δi in the box against S4b's 0.208δi (0.06 against 0.12δ0)
seamsthe box's upstream face is at x=25, 35 δi ahead of the impingement and 29 ahead of S4b's reflected-shock origin; its top at y=3, 1.7δ0, inside the layer (FV reflux at the 2:1 faces)
startfrom t=0 with the box in place, not from S4b's state: moving a developed state onto a refined grid needs runtime regridding along a modelled wall, which is not verified (a static box is what the wall-model leg wm3 verifies). The cost is S4b's ≈300 time units of transient again
runto t=550 on 2 nodes (8 A30), a checkpoint every 2000 steps (≈14 time units); statistics over t>300, where S4b's had converged (S4b changed by at most 1 % between the windows 310–934 and 310–1200)
oraclesbli_oracle.py compare --t0 300: tbl_oracle.averages maps the two levels onto the fine grid (coarse cells repeated); Cf comes from the wall model's history, 1200 bins of equal width weighted by area, so the levels need no special case. field_slices.py cuts the planes on the fine grid the same way

Cost on two nodes ​

The first FNL run of FLUME across two medooza nodes. S4b ran 64 blocks on 4 A30 (one node) at 0.74 s per step; S4c runs 120 blocks of the same size on 8 A30 at 0.95 s per step at the start and 0.98 at t≈400 (the time step falls as the layer develops; the steps cost the same). Per step the busiest GPU advances 4.1 M cells per second against S4b's 4.2 M: the inter-node halo exchange and the 2:1 seams (reflux included) cost about 2–3 %.

The step was slower because of the partition, not the network. The 120 blocks landed 20/13/19/17/13/19/19/0 on the 8 ranks: one GPU held no block, and the busiest held 20 against an even 15, which made each step 1.33 times slower than an even split. mpi_redistribute then walked the blocks greedily and kept the 8 siblings of a refined parent together; a model of the routine reproduces that split exactly (#123). A static grid is now balanced by block count ([amr] balance_static, on by default: 15 per rank on this tree), and a runtime regrid by a min-max split of the sibling sets (busiest 18). Evenly split, the step would take about 0.71 s, so the run would need about 13.5 hours instead of 18.

Results ​

S4c against S4b: wall pressure and skin friction

Left: the mean wall pressure of S4b and S4c against the DNS and the experiment. Right: Cf from the wall model's τw against x, the 2:1 box of S4c shaded. make_doc_figures.py --only sbli-refined.

The statistics are the span and the 21 checkpoints of 307.5≤t≤550 (sbli_oracle.py compare --t0 300); S4b's column repeats the table above.

S4c, 2:1 boxS4b
Reθ at ximp, nominal; δ05314; 1.8245247; 1.807
reflected shock: angle to the wall28.63∘28.95∘
Lint/δ03.043.15 (DNS of Volpiani et al.: 3.22)
separation: Lsep/δ00, Cf never negative, minimum 0.10⋅10−3 at x=58.40, minimum 0.22⋅10−3 at x=58.6
(pw−p∞)/(p2−p∞) at x∗=−0.5, 0, 0.5, 1, 20.89, 1.76, 2.14, 2.26, 2.391.12, 1.72, 2.09, 2.24, 2.37 (DNS: 1.75, 2.07, 2.22, 2.29, 2.36)
Cf upstream of the box (x≤24)S4b's within 0.4 %
Cf in the box, before the interaction−11 % at x=26–30, behind the 2:1 face; +6 % at x=35; +10 % at x=45–55
Cf downstream of the box (x=76–90)−7 %

The flow of S4c

The panels of the flow for S4c at t=550, on the fine grid of the box (the coarse cells repeated outside it, the box outlined in orange in the schlieren). make_doc_figures.py --only sbli-refined-fields.

  • The finer near-wall grid does not open a mean bubble. Halving the matching height halves the dip of Cf, at the same place, and the mean stays positive. In single checkpoints the flow at the wall reverses in patches: at t=550 the spanwise-averaged u of the first cells is negative over a short stretch near x=58.5 (the red u=0 line), and the cross-flow plane at ximp has reversed cells along the wall. The reversal is intermittent; the mean has none.
  • The interaction length shrinks slightly, 3.04 against 3.15, and the rise is shallower (0.89 at x∗=−0.5 against S4b's 1.12 and the DNS's 1.75), its onset unchanged (below): the shape of the interaction is S4b's.
  • The zero-pressure-gradient layer depends on the matching height. Outside the box Cf is S4b's; inside it, after an adjustment over about 10δi behind the 2:1 face, it is 10 % higher before the interaction. The equilibrium model's τw follows the velocity at its matching point, and that point moved from 0.12 to 0.06δ0 (the log-layer mismatch of Kawai & Larsson 2012). A matching height fixed in δi, not in cells, would keep the two levels consistent; it does, within 1.7 % (S4k).
  • What remains is the model. The equilibrium model integrates a constant total stress across its layer, so its τw turns negative only where the velocity at the matching height reverses on average. With the matching point at 0.06δ0 it does not; the pressure-gradient term, which lets τw reverse under a strong adverse gradient while the matching velocity is still positive, was tested next (below): a bubble of 0.6δ0, a quarter of the DNS's.

The matching height fixed: S4k ​

The configuration of Kawai & Larsson (2012) and of Hoy & Bermejo-Moreno (2022) on this grid: the matching height fixed in δi instead of in cells, wall_model_height =5/24δi=0.117δ0, S4b's matching point, with 5 cells below it in the box (S4c: 2.5 cells, h halved). Plain equilibrium model, eddy viscosity below h. S4c restarted from t=550 to 850 on S4c's two nodes (8 A30, 18 h 33 min, 1.19 s per step); the statistics take the span and the 20 checkpoints of 653≤t≤850 (sbli_oracle.py compare --t0 650). Criteria fixed before the run: (a) Cf over 35≤x≤48 (in the box, past the adjustment behind the 2:1 face, ahead of the interaction) within 2 % of S4b's, where S4c is +10.9 % by the same measure; (b) a mean bubble and Lint within 10 % of the DNS's.

S4k against S4b and S4c: wall pressure and skin friction

Left: the mean wall pressure of S4b, S4c and S4k against the DNS and the experiment. Right: Cf from the wall model's τw, the box and the stretch 35≤x≤48 shaded, the stretch where the time-mean u of S4k's first cell is negative hatched. make_doc_figures.py --only sbli-fixed-h.

S4k, 2:1 box, h fixedS4c, 2:1 box, h halvedS4bDNS, Reθ=6882
Reθ at ximp, nominal; δ05253; 1.7835314; 1.8245247; 1.807
Cf over 35≤x≤48 against S4b+1.7 %+10.9 %
Cf against S4b: x=15–24 (outside the box), 45–52, 64–70, 76–90+4.2, +2.2, −4.5, −8.3 %−0.3, +7.6, +1.5, −7.6 %
Lint/δ03.263.043.153.30
Lsep/δ0 from τw; Cf minimum (10−3)0; 0.14 at x=58.60; 0.10 at x=58.40; 0.22 at x=58.62.16 (x∗=−0.8 to −0.15)
time-mean u<0: first cell; second cellx=57.6–59.5, 1.10δ0 (x∗=−0.41 to −0.08); 0.68δ0nonenone
5 % / 50 % of the rise, δ0 ahead of ximp2.58 / 2.282.33 / 2.002.34 / 1.964.13 / 3.60
(pw−p∞)/(p2−p∞) at x∗=−0.5, 0, 0.5, 1, 21.02, 1.57, 2.11, 2.28, 2.390.89, 1.76, 2.14, 2.26, 2.391.12, 1.72, 2.09, 2.24, 2.371.75, 2.07, 2.22, 2.29, 2.36
reversed area per unit span, mean (max), δ020.115 (0.138)0.069 (0.084)0.068 (0.142)
first / third cells reversed in the window23.4 % / 12.6 %14.5 % / 8.7 %9.5 % / 5.1 %

The two refined runs count fine cells, S4b coarse ones: the reversed fractions compare S4k with S4c.

  • (a) passes: the matching height fixed in δi keeps the refined layer on S4b's. Over 35≤x≤48 Cf is 1.7 % above S4b's, against S4c's 10.9 %: the log-layer mismatch of the halved matching height was S4c's drift. One difference outside the box was not planned: there h=5/24δi falls on the 3rd cell's centre, the sample weight comes out 4⋅10−16 instead of 0, and the eddy viscosity below h, which skips the sample cell only when the weight is exactly 0, also enters the 3rd cell (sample_cells, adam_flume_wall_model_library.F90). Upstream of the box Cf is then 4.2 % above S4b's, and the layer entering the box carries that. sample_cells now snaps a weight below 10−9 to 0, so a matching height on a cell centre is that cell; S4k and S4l (whose coarse region has the same defect) were not run again.
  • (b) half passes: the resolved flow separates, the wall stress does not. Lint is the DNS's within 1.2 %, and the compression starts 0.25δ0 earlier than in S4b and S4c (still 1.5δ0 behind the DNS's). For the first time with the plain equilibrium model the time-mean u of the first cell is negative, over 1.10δ0 (half the DNS's bubble, in its downstream half), and of the second over 0.68δ0; a quarter of the first cells are reversed in the window. The model's τw stays positive everywhere (minimum 0.14⋅10−3): the bubble is about two fine cells tall, below the matching height, and the model, which sees only the flow at h, keeps the positive τw of a forward flow over a reversed first cell.
  • So the remaining gap is the coupling. The face takes the model's τw whatever the first cell does; a coupling through the wall viscosity with the no-slip condition kept (De Vanna et al. 2022), where the first cell sets the sign of the wall stress, is the next test (wall models in the literature): S4l (below).

The viscosity coupling: S4l ​

S4k with the model coupled through the wall viscosity, the no-slip condition kept (#126, wall_model_coupling = viscosity, turbulence): the face stress is the larger of the first cell's molecular no-slip stress and the model's, along the first cell's wall-parallel velocity, so a reversed first cell reverses the wall stress. Otherwise S4k's set-up (plain equilibrium model, h=5/24δi, eddy viscosity below h), restarted from S4k at t=850 to 1150 on two nodes (8 A30, 18 h 38 min, 1.19 s per step); the statistics take the span and the 20 checkpoints of 953≤t≤1150 (sbli_oracle.py compare --t0 950). Criteria fixed before the run: (a) Cf over 35≤x≤48 within 2 % of S4b's; (b) a mean bubble from τw and Lint within 10 % of the DNS's. Cf is the stress applied (its x component), negative where the first cell reverses; S4k's is the model's.

S4l, viscosity couplingS4k, flux couplingS4bDNS, Reθ=6882
Reθ at ximp, nominal; δ05276; 1.7575253; 1.7835247; 1.807
Cf over 35≤x≤48 against S4b+3.3 %+1.7 %
Cf against S4b: x=15–24 (outside the box), 25–35, 45–52, 64–70, 76–90+2.7, −0.9, +2.9, −8.4, −9.8 %+4.2, −2.2, +2.2, −4.5, −8.3 %
Lint/δ03.643.263.153.30
Lsep/δ0 from τw1.66, x=56.4–59.4 (x∗=−0.56 to −0.10)002.16 (x∗=−0.8 to −0.15)
Cf minimum (10−3)−0.39 at x=58.40.14 at x=58.60.22 at x=58.6
time-mean u<0: first, second, third cell1.83, 1.14, 0.64δ01.10, 0.68δ0, nonenone
5 % / 50 % of the rise, δ0 ahead of ximp3.48 / 2.932.58 / 2.282.34 / 1.964.13 / 3.60
(pw−p∞)/(p2−p∞) at x∗=−0.5, 0, 0.5, 1, 20.89, 1.56, 2.09, 2.23, 2.351.02, 1.57, 2.11, 2.28, 2.391.12, 1.72, 2.09, 2.24, 2.371.75, 2.07, 2.22, 2.29, 2.36
sonic height at ximp−5δ0, δ00.0350.0280.036
reversed area per unit span, mean (max), δ020.188 (0.217)0.115 (0.138)0.068 (0.142)
first / third cells reversed in the window33.4 % / 18.9 %23.4 % / 12.6 %9.5 % / 5.1 %

The green line of the figure above; the reversed flow per checkpoint in the reversal figure.

  • (b) half passes: the wall stress separates in the mean, for the first time in this campaign. Cf<0 over 1.66δ0, three quarters of the DNS's bubble, at its downstream end (x∗=−0.56 to −0.10 against −0.8 to −0.15). The compression starts 0.9δ0 earlier than S4k's (5 % of the rise 3.48δ0 ahead of ximp, the DNS's 4.13), and the reversed flow grows: the first cells reverse in the mean over 1.83δ0, a third of them in the window. Lint overshoots: 3.64, 10.3 % above the DNS's 3.30, just outside the bound (Volpiani et al.'s 3.22 at Reθ=1050).
  • (a) fails: the stretch 35≤x≤48 is 3.3 % above S4b's (S4k's 1.7 %). In the box the first cell spans about 21 wall units. Under the flux coupling (S4k) its no-slip stress would be 1.14 times the applied one; under the viscosity coupling the maximum takes it, the first cell slows (u1 0.409 against S4k's 0.488, the cells above less) and its no-slip stress falls to 0.94 of the applied. Two effects then pull apart: the maximum keeps picking the molecular stress whenever the first cell runs fast (with its rms of 17 % and a steady model stress, +3.8 %), and the slower sample lowers the model's stress. On the flat plate, with the sample 2.5 cells up, the second wins: −8.7 % against the flux-coupled control (boundary layer); here, 5 fine cells up, the net is +1.6 % against S4k (a hypothesis: the model's own τw is not recorded apart from the applied one). The shift is also of the size of the scatter of S4k's own bands against S4b (from −2.2 to +4.2 %). The sonic height at ximp−5δ0 is the DNS's, 0.035δ0 against 0.036, where S4b's is 0.028. The stress also jumps at the 2:1 face (x=25, +10 % over about 3δi), which the flux coupling does not: there the first cell enters the box at the coarse level's velocity, and the maximum takes its no-slip stress, as at the restart of the flat plate.
  • Behind the interaction the layer recovers more slowly (Cf 8 to 10 % below S4b's over 64≤x≤90, S4k's 4 to 8 %).

The wall model with the pressure gradient ​

The equilibrium model has its τw from a constant-stress layer, so it reverses only where the mean velocity at the matching height does. With wall_model_pressure_gradient (turbulence) the layer carries τw+ydp/dx, and an adverse gradient strong enough turns τw negative while the sample still moves downstream.

A priori, on the mean fields ​

Before any run, the model was applied to the span and time averages of S4b and S4c: per x, the mean sample at the run's matching point and the gradient of the mean wall pressure, in a Python port of the model (sbli_oracle.py means and apriori; the port reproduces the Fortran solver to 10−14).

S4b, h=0.208δiS4c, h=0.104δi in the box
the model without dp/dx on the mean sample, against the run's Cf (x=35–50)−1.7 %−2.0 %
without dp/dx: Cf<0nowhere (minimum 0.10⋅10−3)nowhere (minimum 0.03⋅10−3)
with dp/dx: Cf<0x=56.1–61.2, Lsep/δ0=2.81x=56.1–60.7, Lsep/δ0=2.54
with dp/dx: Cf minimum−1.58⋅10−3 at x=56.5−1.28⋅10−3 at x=58.9

The equilibrium model with and without the pressure gradient on the mean samples

Top: Cf of the runs (the wall model's history), of the model on the mean sample, and of the model with the mean dp/dx, the reversed stretch shaded; bottom: the gradient of the mean wall pressure. make_doc_figures.py --only sbli-apriori.

  • The term opens a bubble on both grids, 2.5–2.8 δ0 long, against 1.62 in the DNS of Volpiani et al. (at Reθ=1050). The model without it reproduces the runs' Cf to 2 %, so the port is the run's model.
  • The term dominates there: (dp/dx)h/τw reaches about 70 in the bubble, far from the equilibrium layer the model assumes. The model keeps dp/dx and drops the convective terms, which partly balance it in the wall layer: it can over-respond.
  • The mean field is the attached one. With a bubble the pressure rise spreads and dp/dx falls, so the run decides.
  • The instantaneous gradient cannot be used as it is. On a single checkpoint of S4b upstream of the interaction (x=35–50) the turbulent wall-pressure gradient has an rms of 1.1⋅10−2 against a mean of 8⋅10−4, so |dp/dx|h/τw has a median of 1.4, and the model with it raises the mean τw of the zero-pressure-gradient layer by 13 %. The run filters the gradient in time (wall_model_pressure_gradient_filter).

The run ​

Setup
startS4b's last state (t=1200), the developed layer and interaction; its block layout kept ([amr] balance_static = .false.)
modelwall_model_pressure_gradient = .true., wall_model_pressure_gradient_filter = 10 (about 5δ0/U∞, far below the period of the interaction's low-frequency motion); the rest as S4b (the 3rd cell, h=0.208δi, semi-local damping, the model's eddy viscosity below h)
runto t=1450 on one node (4 A30), 0.74 s per step, S4b's cost; statistics over the 24 checkpoints of t≥1300 (sbli_oracle.py compare --t0 1300); the input is S4b's make_tbl.py line with --restart --time-max 1450 --wm-pressure-gradient --wm-pg-filter 10 (no case file: a one-off restart)

S4d against S4b and S4c: wall pressure and skin friction

Left: the mean wall pressure of S4b, S4c and S4d against the DNS and the experiment; right: Cf from the wall model's τw, the reversed stretch of S4d shaded. make_doc_figures.py --only sbli-pg.

S4d, with dp/dxS4bS4c, 2:1 box
separation: Lsep/δ00.60, Cf<0 from x=56.6 to 57.7, minimum −1.9⋅10−300
Lint/δ0 (DNS of Volpiani et al.: 3.22)3.123.153.04
reflected shock: angle to the wall28.2∘28.95∘28.6∘
(pw−p∞)/(p2−p∞) at x∗=−0.5 (DNS 1.75; at Reθ=6882 1.65)0.951.120.89
Reθ at ximp; δ05288; 1.7945247; 1.8075314; 1.824
Cf upstream (x=35–50) against S4b−1.7 % (S4b's own windows differ by up to 1 %; the unfiltered gradient gave +13 % a priori)

The flow of S4d

The panels of the flow for S4d at t=1450. make_doc_figures.py --only sbli-pg-fields.

  • The mean flow separates, for the first time in these runs: Cf is negative over 0.60δ0 at the foot of the reflected shock's compression, where the filtered gradient is largest, and reattaches before the impingement. The a priori test on the mean fields gave 2.5–2.8 δ0 for the same pressure field; the run's is a quarter of that, the gradient being filtered and the pressure rise unchanged.
  • The bubble is a quarter of the DNS's. The DNS at Reθ=6882 separates from x∗≈−0.8 to −0.15, Lsep/δ0≈2.16, ahead of the impingement as here; this one spans x∗=−0.60 to −0.41, inside it.
  • The interaction does not change. The wall pressure still rises at x∗≈−0.7, half-way at −0.56 against the DNS's −1.09, and Lint is S4b's: the model's stress reverses, but the resolved flow above the matching height, which sets the pressure, does not.
  • What remains. The model keeps dp/dx and drops the convective terms (the known inconsistency): a non-equilibrium layer (the thin-boundary-layer equations with convection, or an integral model) is the next test of the wall model. Its target is the DNS's separation: from x∗≈−0.8, 2.2δ0 long, and the wall pressure half-way at x∗≈−1.1 (below).

Where the compression starts ​

The earlier versions of this page found the wall pressure rising too early: from x∗≈−0.7, where the DNS of STREAmS Fig. 7a was flat until −0.3. The comparison was wrong, not the run. The STREAmS paper defines x∗=(x−ximp)/L, but its figure follows the IUSTI convention of Dupont et al. (2006), where the origin is the mean position of the reflected shock (Bernardini et al. 2022, the footnote to their Lint). Shifted by one, x∗−1, the Fig. 7a curve falls on the STREAmS DNS at Reθ=6882 of Bernardini et al. (2022, Fig. 4a, x∗ from ximp). Over −1.8≤x∗≤2.5 the rms difference is 0.034p∞, and the best shift is 1.02. A reading of the curve itself agrees: at x∗=0 the printed DNS is half-way up the rise (0.64), the value at the foot of a shock, not on the plateau a separated interaction has at its impingement. The two DNS and the experiment then agree with each other.

Where the compression starts

Left: the references as printed (dotted, open triangles) and shifted (solid, filled), the DNS at Reθ=6882 (dashed), and the runs. Right: the mean Mach number of the incoming layer at ximp−5δ0 against the DNS profile the inflow is made from. sbli_oracle.py fields and compression; make_doc_figures.py --only sbli-onset.

S4bS4c, 2:1 boxS4d, with dp/dxDNS, Reθ=6882
Lint/δ03.163.053.133.30
wall pressure at 5 % of p2−p∞: x∗ (δ0 ahead of ximp)−0.74 (2.34)−0.76 (2.33)−0.68 (2.12)−1.25 (4.1)
at 50 %: x∗ (δ0 ahead of ximp)−0.62 (1.96)−0.66 (2.00)−0.56 (1.77)−1.09 (3.6)
Lsep/δ0000.602.16 (x∗=−0.8 to −0.15)
incoming layer at ximp−5δ0: mean sonic height0.028δ00.031δ0 (1.4 cells)0.041δ00.036δ (the inflow's DNS profile)

Lint here is from the time-mean field (fields, compression), 0.01 above compare's in the tables above.

  • The incoming layer is not the cause. Its mean Mach number is the DNS profile's to 0.03 above 0.3δ and to 0.15 at the 2nd cell (0.13δ, lower: the log-layer mismatch of the wall model), and its mean sonic height, 0.03–0.04δ0, is the DNS profile's 0.036δ. The first cell, below that height, runs at M≈0.95 against the DNS's 0.85: the wall model's slip, one cell thick.
  • The compression is late by about 0.5Lint, 1.6–1.8δ0, on every grid and model. Below 0.5δ0 it crosses the layer as a vertical front at x∗≈−0.7 to −0.6; above, it is the incident shock. The reflected shock's apparent origin is right (Lint within 5 %), the plateau behind the impingement too.
  • What is missing is the separation. In the DNS a bubble of 2.2δ0 starting at x∗≈−0.8 lifts the flow and carries the compression to x∗≈−1.2. The equilibrium model opens none (S4b, S4c) or a quarter of it (S4d), and the late compression follows. Halving the near-wall cells (S4c) leaves the onset where it was. The pressure-gradient term moved the onset by 0.06Lintdownstream. S4e qualifies this bullet and the first: without the model's eddy viscosity below h the near-wall layer slows and the onset moves 1.2δ0 upstream with no bubble (below).
  • For the next wall model, non-equilibrium: the target is the DNS's bubble (x∗≈−0.8 to −0.15) and its wall pressure, half-way at x∗≈−1.1. The run's wall pressure is printed by sbli_oracle.py compare (at fixed x∗, against both DNS) and compression (where it reaches 5, 10 and 50 %); the oracle reports the run's Lsep, the DNS's is quoted from Bernardini et al. (2022).

Wall models for this interaction in the literature ​

Before a new wall model, what the literature says about wall-modelled LES of separating interactions, and the test it suggests on these runs.

SourceCaseFinding
Hickel, Touber, Bodart & Larsson (2012, CTR Summer Program)this interaction (IUSTI, 8∘), a priori on the wall-resolved LES of Touber & Sandham (2009); an adverse-pressure-gradient layerAbove the viscous layer the convective term balances the pressure gradient (Bernoulli), so a model that keeps dp/dx and drops convection, as here, is inconsistent: on the adverse-gradient layer it separates much too early, and the model of Duprat et al. (2011), whose velocity scale S4d uses, over-predicts the wall shear of the interaction. Convection parametrised as −1ρdpdxmin(y/ypg,1), ypg=4(μw2/(ρw|dp/dx|))1/3, gives a skin friction close to the reference
Larsson, Kawai, Bodart & Bermejo-Moreno (2016, Mech. Eng. Rev. 3)reviewEquilibrium models can be accurate in these interactions "provided that the outer part of the boundary layer is well resolved" (a duct interaction near incipient separation; a ramp at Reθ≈50000, where the equilibrium model fell within the experimental uncertainty as the non-equilibrium one did): no conclusive demonstration that a non-equilibrium model is better yet. Outer-layer grid: Δx≈0.08δ, Δz≈0.05δ, Δy from 0.02δ at y=0.2δ to 0.05δ at δ
Fu, Karp, Bose, Moin & Urzay (2021, JFM)Mach 6, laminar inflow, transitional interactionThe equilibrium-model LES delays the separation and advances the reattachment: a shorter bubble than the DNS's (the symptom here, in a different regime)
De Vanna, Bernardini, Picano & Benini (2022, IJHFF 98)this interaction (IUSTI, 8∘, Mach 2.28, Reτ≈700 at the inflow), WALE, Δx+=140–82, first Δy+=39–34 on a stretched gridAn equilibrium model coupled through a wall viscosity added to the molecular one, the no-slip condition kept: the first cell sets the sign of the wall stress, so Cf turns negative in the bubble. The incoming Cf matches their wall-resolved LES at every resolution, but Lint/δ0 is 7.0 / 5.2 / 5.0 (coarse to fine) against 3.3 wall-resolved and 4.18 measured, attributed to the model and to the subgrid and shock-capturing dissipation

Against these runs. S4b's grid (0.046δ0, cubic) is finer than the guideline in x, equal in z and 2.3 times coarser in y at 0.2δ; S4c's box (0.023δ0) meets it everywhere, and does not separate either. The Bernoulli-consistent model a priori on the mean fields (sbli_oracle.py apriori --ypg 4, the port of the pressure-gradient model with y replaced by g(y)=y−y2/(2ypg) below ypg and ypg/2 above):

model, no dp/dxwith dp/dx (S4d's model)Bernoulli-consistent, ypg∗=4DNS, Reθ=6882
S4b: Lsep/δ002.811.01 (x=57.8–59.6)2.16
S4c: Lsep/δ002.541.53 (x=57.4–60.2)
  • The a priori numbers bound the model, not the run. S4d's run kept a quarter of its a priori bubble (0.60 against 2.81), and its strongly negative τw left the interaction where it was: the resolved flow above the matching height decides. A Bernoulli-consistent model alone is unlikely to reach 2.2δ0.
  • The suspect is the resolved near-wall flow. The first cell runs at M≈0.95–0.98 against the DNS profile's 0.85 at that height (above): extra momentum at the wall that the adverse gradient has to stop. The model's eddy viscosity below the matching height (wall_model_eddy_viscosity) shapes that layer, so S4e takes it out.

Without the eddy viscosity below the matching height ​

Setup
startS4b's last state (t=1200); its block layout kept ([amr] balance_static = .false.)
modelS4b's without wall_model_eddy_viscosity (the cells below h get the SGS model's viscosity only); no dp/dx
runto t=1450 on one node (4 A30), 3 h 24 min; statistics over the 23 checkpoints of t≥1300 (sbli_oracle.py compare --t0 1300, fields, compression); the input is S4b's line of tbl_cases_s4.txt without --wm-eddy-viscosity, with --restart --time-max 1450 (no case file: a one-off restart)

S4e against S4b and S4d

Left: the mean wall pressure of S4b, S4d and S4e against the DNS and the experiment; middle: Cf from the wall model's τw, the reversed stretch of S4d shaded; right: the mean Mach number of the incoming layer at ximp−5δ0 against the DNS profile the inflow is made from. make_doc_figures.py --only sbli-nonu.

S4e, no eddy viscosity below hS4bDNS, Reθ=6882
Lint/δ03.513.163.30
wall pressure at 5 % of p2−p∞: distance ahead of ximp3.55δ02.34δ04.13δ0 (x∗=−1.25)
the same at 50 %2.64δ01.96δ03.60δ0 (x∗=−1.09)
separation: Lsep/δ00, Cf minimum 0.27⋅10−3 at x=58.50, minimum 0.21⋅10−3 at x=58.62.16
incoming layer at ximp−5δ0: first-cell Mach number0.75 (DNS profile 0.82 there)0.98 (DNS profile 0.83)
the same: second and third cells1.08, 1.30 (DNS 1.23, 1.37)1.13, 1.26 (DNS 1.24, 1.37)
the same: sonic height0.061δ0.030δ0.036δ
incoming layer: δ0, δ∗, Reθ1.85, 0.572, 54631.81, 0.531, 5271
Cf at x=35–45+4.8 % on S4b
  • The eddy viscosity moves the onset, not the separation. Without it the first cell slows from M=0.98 to 0.75, the sonic line rises from 0.030 to 0.061δ, and the compression starts 1.2δ0 earlier: about two thirds of the way to the DNS at 5 % of the rise, 40 % of the way at 50 %. Lint overshoots the DNS by 6 % (S4b undershoots by 4 %). The mean flow still does not separate, and the minimum of Cf stays near S4b's. The pressure-gradient term (S4d) did the opposite: a bubble, the onset moved by 0.06Lint downstream.
  • The incoming layer is not the DNS's either way. S4b's first cell is too fast and its next two too slow; S4e's first three are all too slow, and the sonic height goes from 15 % below the DNS's to 70 % above. δ∗ grows by 8 % and Cf upstream by 5 %: S4e changes the layer that meets the shock, so it is a diagnostic, not a candidate setting. The run stayed stable for its 250 time units (the eddy viscosity was introduced against the first-cell drain of the Mach 3 channel, turbulence); longer runs were not made.
  • The mean is short. 150 time units against S4b's 890; the interim mean over t=1301–1363 gave Lint/δ0=3.52, the full one 3.51.
  • What it points to. The two features have different levers: the onset answers to the momentum of the resolved near-wall layer, the separation to the sign of τw. A Bernoulli-consistent model changes τw only (a priori 1.0–1.5δ0, above), so it is expected to open a bubble without moving the onset, as S4d did; the onset needs an incoming near-wall layer closer to the DNS profile, which is the matching between the model and the resolved cells, not the model's momentum balance.

The near-wall cells of the incoming layer ​

Before changing how the model meets the resolved cells: how far the cells of the incoming layer are from the DNS profile the inflow is made from, on both grids and with and without the eddy viscosity below h.

Setup
datathe span and time averages of sbli_oracle.py fields at ximp−5δ0 (S4b t=310–1200, S4c 308–550, S4d 1305–1450, S4e 1301–1450); S4c on the fine grid of its box
referencereference/tbl_m2p28_profile.txt at the cell centres, y/δ with the run's local δ, u/ue
figuremake_doc_figures.py --only sbli-cells

The near-wall cells of the incoming layer

Left: the mean velocity of the first 8 cells of S4b, S4c, S4d and S4e against the DNS profile; right: the difference of each cell from the DNS at its centre. A cell holds the mean over its height: in the first cell, where the profile bends, the DNS's cell mean is below its centre value, so the excess of the runs is a lower bound.

u/ue, cells 1 – 41234
DNS at S4b's centres (y/δ 0.025, 0.075, 0.12, 0.17)0.4780.6690.7220.757
S4b0.558 (+17 %)0.625 (−7 %)0.684 (−5 %)0.758 (0 %)
S4d, with dp/dx+14 %−9 %−7 %0 %
S4e, no eddy viscosity below h0.441 (−7 %)0.602 (−10 %)0.698 (−3 %)−1.5 %
S4c, the box (Δy/2; y/δ 0.012, 0.037, 0.062, 0.087)+70 %+2 %−2 %+2 %

The mean stress through the face between cells 1 and 2, in units of τw (the time-mean strain times the mean viscosities; the resolved part −ρ¯(uv―−u¯v¯)):

viscousresolvedthe model's eddy viscositysum
S4b0.030.200.590.82
S4e0.080.33—0.41
  • The first cell slips with the eddy viscosity and lags without it. With the model's eddy viscosity below h it is 14–17 % faster than the DNS on S4b's grid and 70 % on S4c's finer one; without it 7 % slower.
  • Cells 2 and 3 are 5–10 % slow on S4b's grid, within 2 % on S4c's. That is the log-layer mismatch of a coarse near-wall grid (Kawai & Larsson 2012), and S4c's Δy=0.025δ removes it. From the 4th cell up every run is within 2.5 % of the DNS.
  • The budget does not close on the means. No subgrid model runs (the hybrid KEP-WENO scheme is the implicit LES), so with the eddy viscosity the explicit stresses carry 0.82 of τw through the first face, without it 0.41. The rest is the scheme's dissipation, or the correlation of the eddy viscosity with the strain, which the means cannot separate: whether the resolved and the modelled stress overlap below h is not measurable from these fields.
  • A DNS-like incoming layer does not move the onset. S4c's layer is the DNS's from the second cell up, and its onset is S4b's (above). S4e moves the onset with a layer slower than the DNS's. So matching the model to the resolved cells better, by itself, is not the lever for the interaction. The DNS's onset comes with its bubble, and the separation is what the runs lack (S4d opened a quarter of it, above).

The pressure gradient without the convection, with it, and without the eddy viscosity ​

Two more restarts of S4b at t=1200, run to t=1450, statistics over t≥1300 (sbli_oracle.py compare --t0 1300, fields, compression), as S4d and S4e:

SetupS4gS4f
modelS4d's (wall_model_pressure_gradient, filter 10) with wall_model_pressure_gradient_ypg = 4: the Bernoulli-consistent variant (turbulence)S4d's, without wall_model_eddy_viscosity
questiondoes the consistent model keep a bubble? (a priori 1.0δ0 against S4d's 2.8)does S4d's bubble grow when the near-wall flow is slower (S4e's layer)?
runone node (4 A30), 0.97 s per step (S4d 0.92, S4e 0.68), 24 checkpointsone node (4 A30), 3 h 52 min, 22 checkpoints

The pressure-gradient runs against S4b and S4e

Left: the mean wall pressure of S4b, S4e, S4d, S4g, S4f, S4h (black, below) S4i (pink, below) and S4j (olive, below) against the DNS and the experiment; right: Cf from the wall model's τw, the reversed stretches shaded. The legend's Lint is compare's, 0.01 below the table's. make_doc_figures.py --only sbli-pg-runs.

S4bS4e, no νt below hS4d, dp/dxS4g, Bernoulli-consistentS4f, dp/dx, no νt below hDNS, Reθ=6882
Lint/δ0 (the time-mean field)3.163.513.133.243.373.30
Lsep/δ0000.6000.14 (x=58.4–58.6)2.16
Cf minimum (10−3)0.210.27−1.870.15−0.02
5 % of the rise, ahead of ximp2.34δ03.55δ02.12δ02.44δ03.07δ04.13δ0
50 % of the rise1.96δ02.64δ01.77δ02.04δ02.42δ03.60δ0
first-cell Mach number at ximp−5δ00.980.750.940.950.700.83
  • The consistent model does not separate. With convection balancing the pressure gradient above ypg (a few viscous lengths of the gradient, far below h), τw stays positive: S4g's Cf dips to 0.15⋅10−3, S4b's 0.21, and its interaction is S4b's (onset within 0.1δ0, Lint within 0.08). S4d's bubble was the pressure gradient acting over the whole layer, the inconsistency Hickel et al. (2012) point out; a priori the consistent model kept 1.0δ0 of it, the run none.
  • A slower near-wall flow does not grow S4d's bubble. Without the eddy viscosity (S4f) the first cell slows to M=0.70, as in S4e, and the bubble shrinks from 0.60 to 0.14δ0 and moves 0.7δ0 downstream, toward the impingement (x=58.4–58.6 against S4d's 56.6–57.7); its Cf also dips to 0.15⋅10−3 at x≈55.4, where its compression starts, without reversing; the onset moves upstream as in S4e, by a little less (3.07δ0 against 3.55 at 5 %). The two levers do not add: each moves its own feature, and together they move both less.
  • What none of the five runs does. The DNS separates over 2.16δ0 from x∗≈−0.8; the runs open at most 0.6δ0 and only with the inconsistent model. The sample sits at h≈0.12δ and ypg about ten times lower (≈0.02δi under the interaction's gradient); where the reversed layer is thinner than h the sample still moves downstream, and an equilibrium-type model, which takes τw from that sample and a consistently balanced gradient, cannot see the reversal. The DNS's bubble is taller than h in its large state, and the runs' own mean reversed layer lies below h (below). The convection the reversal needs is the flow's own, which a model fed with the resolved field (the integral model of Yang et al. 2015, the thin-boundary-layer equations with convection) carries and these do not.

The reversed flow: the DNS's bubble and the runs' below the sample ​

Whether a reversed layer can hide below the matching height: how tall the DNS's bubble is, and where the runs' own reversed flow sits.

The DNS. Bernardini et al. (2022) give the bubble's length (2.16δ0 in the mean, from x∗≈−0.8 to −0.15) and its time history, but no height. Their Fig. 10 shows two instantaneous states on a vertical slice, in y∗=y/Lint: in the large one the u=0 line reaches y∗≈0.14–0.16, about 0.5δ0, over x∗≈−0.85 to 0.1; in the other the bubble is almost absent. The area of the bubble swings between two probable values with intermittent large excursions (their Fig. 9). Our sample sits at h=0.208δi=0.115δ0=0.035Lint: the large state is about 4h thick, ten of S4b's cells.

The runs. sbli_oracle.py reversal reads each checkpoint in ximp−4.5δ0<x<ximp+δ0, y<1.5δ0:

S4bS4c, 2:1 boxS4d, dp/dxS4e, no νt below hS4f, dp/dx, no νtS4g, Bernoulli-consistent
checkpoints712124232224
area with u<0 per unit span, mean (max), δ020.068 (0.142)0.069 (0.084)0.049 (0.086)0.140 (0.200)0.113 (0.191)0.077 (0.106)
highest reversed cell of a column, 99th percentile, mean (max), δ00.31 (0.44)0.30 (0.40)0.31 (0.44)0.38 (0.52)0.38 (0.52)0.34 (0.49)
first cells reversed9.5 %14.5 %6.1 %23.3 %17.9 %9.8 %
third cells (the sample) reversed5.1 %8.7 %3.9 %8.2 %6.6 %5.9 %
time-mean u<0 in the first cellnonex=57.8–58.9, 0.66δ0nonex=56.8–58.8, 1.13δ0x=57.5–58.9, 0.77δ0none
time-mean u<0 in cells 2, 3nonecell 2 0.32δ0; cell 3 nonenonenonenonenone
Lsep from the wall model's τw (compare)000.6000.140

The reversed flow of the runs

Left: the area with u<0 per unit span per checkpoint; right: the fraction of the first and of the third cells (the wall model's sample) reversed in the window. sbli_oracle.py reversal, make_doc_figures.py --only sbli-reversal. The time-mean reversal of the cells is printed by sbli_oracle.py compression.

  • The runs reverse as high as the DNS's large state, but rarely. In every run some columns reverse up to 0.4–0.5δ0; the span-averaged area stays at 0.02–0.2δ02, within about twice its mean in each run (S4b's 71 checkpoints: 0.04–0.14 around 0.068), where the DNS's bubble area has a bimodal distribution with a long tail of large excursions (their Fig. 9; its normalisation is not given, and a single slice is not a span average, so the areas themselves are not compared). The sample sees these reversals: its cell is reversed 35–64 % as often as the first cell.
  • In the mean, three runs separate below the sample. S4c's resolved flow reverses in the mean over 0.66δ0 in its first cell and 0.32δ0 in its second; S4e's over 1.13δ0 (x∗=−0.49 to −0.19, against the DNS's −0.8 to −0.15), S4f's over 0.77δ0; the third cell, the sample, never. The wall model's τw, which takes the sample, stays positive there: compare reports no separation for S4c and S4e, and 0.14δ0 for S4f.
  • Under that layer the wall stress has the wrong sign. A positive τw takes streamwise momentum out of a first cell whose mean flow is already reversed, where the wall would push it forward. The model is consistent only while the reversed layer reaches h.
  • So the bubble is partly there, and the oracle's Lsep misses it. The mean reversal of the first cell is the separation a wall-resolved run would report; on it, S4e is at half the DNS's length and the onset of its reversal is at x∗=−0.49, against −0.8. What the equilibrium-type models lack is a wall stress consistent with a reversed layer below h: a model fed with the flow's own convection between the wall and h (thin-boundary-layer equations, an integral model) has it by construction.

A priori: the wall model fed with the resolved convection ​

The thin-boundary-layer balance between the wall and h keeps the convection with the pressure gradient (Wang & Moin 2002; Park & Moin 2014): (μ+μt)dU/dy=τw+∫0yFdy′, F=dp/dx+∂(ρuu)/∂x+∂(ρuv)/∂y. Fed with F of the resolved cells below the sample (linear in y between the cell centres, F=dp/dx at the wall, where the convection vanishes) and applied to the mean field of a run (sbli_oracle.py convective):

Cf<0 a priori, δ0equilibriumdp/dx alone (S4d's)Bernoulli-consistent (S4g's)fed with the resolved convectionmean u<0 of the first cell
S4b's mean field02.861.013.04 (x=56.5–61.9)none
S4d's02.880.742.79none
S4e's02.571.313.33 (x=56.4–62.5)1.08
S4f's03.111.312.980.72

The wall model a priori on the mean fields

Cf of the four models on the mean fields of S4b (left) and S4e (right), and of the convection-fed one with the velocity scale from τw (pink, S4i); the mean u of the first cell dotted. sbli_oracle.py convective, make_doc_figures.py --only sbli-convective.

  • Where the resolved flow obeys Bernoulli, the model follows it. At the steepest gradient (x≈56.5) the resolved convection of cells 1–6 balances 0.9–1.4 of −dp/dx, and the model fed with it loses the spike of the pressure gradient alone (Cf=−1.58⋅10−3 there for S4b), as the Bernoulli-consistent model does.
  • Where the resolved near-wall flow is slow or reversed, the convection is small (0 to 0.3 of −dp/dx in cells 1–3 over x=57.5–58.5) and the model behaves as the pressure gradient alone: Cf<0 over about 3δ0, from x∗≈−0.6, downstream into the reattachment.
  • The a priori does not tell this model from S4d's (2.8–3.3 against 2.6–3.1), and S4d's run kept a fifth of its a priori bubble. Unlike S4d's model, this one changes with the resolved flow: its τw turns negative where the resolved convection stops balancing the pressure gradient, which is where S4c, S4e and S4f reverse below h. Only a run says whether the feedback opens the DNS's bubble.

The run: S4h ​

S4b restarted at t=1200 with S4d's model fed with the resolved convection (wall_model_pressure_gradient, filter 10, wall_model_convection; turbulence), the eddy viscosity below h kept, to t=1450 on one node (4 A30, 5 h 27 min), statistics over the 24 checkpoints of t≥1300; it is the black line of the figure above.

S4h, fed with the resolved convectionS4d, dp/dxS4g, Bernoulli-consistentS4bDNS, Reθ=6882
Lint/δ0 (the time-mean field)3.363.133.243.163.30
Lsep/δ0 from τw0.18 (x=55.96–56.29, x∗=−0.66 to −0.61)0.60 (x∗≈−0.6 to −0.4)002.16 (x∗=−0.8 to −0.15)
Cf minimum (10−3)−0.30 at x=56.1−1.870.150.21
time-mean u<0 in the first cellx=58.3–59.0, 0.41δ0nonenonenone
5 % of the rise, ahead of ximp2.63δ02.12δ02.44δ02.34δ04.13δ0
50 % of the rise2.23δ01.77δ02.04δ01.96δ03.60δ0
first-cell Mach number at ximp−5δ00.910.940.950.980.83
reversed area per unit span, mean (reversal), δ020.0910.0490.0770.068
Cf upstream against S4b: x=15–25, 25–35, 35–45, 45–52+5, +6, +9.5, +12 %+1, +1, −2, −2 %0, 0, −3, −3 %
  • The interaction moves toward the DNS. Lint is the DNS's within 2 % (the closest of the runs), the onset moves upstream by 0.3δ0 at 5 % and at 50 % of the rise, the first cell slows from M=0.98 to 0.91 (the DNS profile 0.83), and the wall stress reverses at x∗≈−0.63, where the DNS's bubble begins. The bubble is short, 0.18 against 2.16δ0; the resolved first cell reverses in the mean over another 0.41δ0 further down.
  • The zero-gradient layer moves too, which it should not. Upstream of the interaction Cf rises by 5 % near the inflow and by 12 % ahead of the interaction, where S4d and S4g stay within 3 % (S4e and S4f, without the eddy viscosity, within 7 %), and the recovery behind the reattachment is higher still. The convection fed to the model is the instantaneous resolved one, and its time mean holds the gradient of the resolved shear stress, ∂⟨ρu′v′⟩/∂y, large in the cells below h (the budget above: 0.2 of τw through the first face, about 1 by the third). The model's own eddy viscosity already carries that stress: it is counted twice, and τw grows. Park & Moin (2014) meet the same double count in their thin-boundary-layer model fed with the LES and reduce the model's eddy viscosity dynamically, so that modelled and resolved stresses at h add up to the total.
  • So the convection is a lever, and this form of it is not yet a model. It moves both the onset and Lint the right way, which no equilibrium-type variant did, but a model that changes the zero-gradient layer by 10 % fails the layer before the interaction. The correction is the reduced eddy viscosity (or the convection of the mean flow only), checked on the zero-gradient boundary layer before the interaction: S4i, below.

The run: S4i ​

S4h's model with wall_model_resolved_stress (turbulence): the filtered resolved shear stress of the convection cells removed from the forcing, and the eddy viscosity's velocity scale from τw alone. It passes the zero-gradient gate (the Mach 2.5 layer's Cf within 0.3 % of the equilibrium model's, boundary layer). S4b restarted at t=1200, to t=1450 on one node (4 A30, 6 h 01 min), statistics over the 24 checkpoints of t≥1300; the pink line of the figure above.

S4i, the resolved stress removed, scale from τwS4h, fed with the resolved convectionS4bDNS, Reθ=6882
Lint/δ0 (the time-mean field)2.943.363.163.30
Cf<0 (compare)x=56.96–62.38 (3.01δ0, x∗=−0.58 to 0.45), negative over 1.99δ0 of it0.1802.16 (x∗=−0.8 to −0.15)
Cf minimum (10−3)−3.41 at x=57.3−0.300.21
time-mean u<0 in the first cellnone (its minimum u 0.068, at x=58.9)0.41δ0none
5 % / 50 % of the rise, ahead of ximp1.85 / 1.59δ02.63 / 2.232.34 / 1.964.13 / 3.60
first-cell Mach number at ximp−5δ00.980.910.980.83
reversed area per unit span, mean (max), δ020.033 (0.054)0.0910.068 (0.142)
first / third cells reversed4.8 % / 2.5 %9.5 % / 5.1 %
Cf upstream against S4b: x=15–25, 25–35, 35–45, 45–52−0.4, −6.0, −5.8, −2.5 %+5, +6, +9.5, +12 %
Cf behind the interaction against S4b: x=64–70, 70–80−36, −11 %
A priori on the mean fields (convective), Cf<0 over 55≤x≤64dp/dx aloneconvection-fed, scale (|τw|+|S|)/ρand the scale |τw|/ρ
S4b's mean field56 %54 % (min −0.73)81 % (min −2.46)
S4e's52 %66 % (min −1.11)94 % (min −3.90)
S4i's42 %44 % (min −1.34)66 % (min −3.23)
  • The wall stress reverses, the flow does not. Cf is negative over two stretches, deep at the foot of the compression (x=56.96–57.79, down to −3.4⋅10−3, S4d's inconsistent balance −1.9⋅10−3) and shallow behind the impingement (x=59.8–62.4). Between them, at x=57.9–59.7, it is positive: there the first cell is slowest (u≈0.09), while at the deep stretch it moves at 0.33. The resolved flow reverses less than in any earlier run (half S4b's area and first-cell fraction), and its first cell never reverses in the mean.
  • The interaction shrinks. Lint falls from 3.16 to 2.94δ0 and the compression starts 0.5δ0 later than S4b's (5 % of the rise 1.85δ0 ahead of ximp, against 2.34; the DNS 4.13): the opposite of S4h's shift. A negative wall stress under a forward first cell pushes the near-wall flow downstream, against the pressure rise.
  • The cause is the velocity scale, and the mean fields show it a priori. With the scale from τw, the eddy viscosity vanishes as τw approaches zero, the model's layer becomes laminar, and the integrated adverse forcing (of order (dp/dx)h/2≈5⋅10−3) outweighs μU(h)/h (about 7⋅10−5): τw jumps to the negative branch. On S4b's mean field the variant is negative over 81 % of the interaction (the scale with |S| 54 %) with a minimum three times deeper, and its sign changes 23 times along x≥54, against 2 to 4 for the other variants. The a priori was run after S4i; it would have stopped the run.
  • So the zero-gradient gate is necessary, not sufficient. The scale that removes the zero-gradient bias (the forcing's fluctuation in |S|) also removes the eddy viscosity that the adverse gradient needs. Upstream the run confirms the zero-gradient result: Cf within 6 % of S4b's (S4h's +5 to +12 %), lower behind the inflow, as in the gate's recovery stretch.

A priori: the velocity scale of the eddy viscosity ​

S4h's scale, (|τw|+|S|)/ρ, raises the zero-gradient τw (the forcing's fluctuation inside |S|); S4i's, |τw|/ρ, collapses under the adverse gradient. Scales meant to keep the forcing's mean in the eddy viscosity and its fluctuation out, each tested a priori on both flows before any code: on the interaction's mean fields of S4b, S4e and S4i (sbli_oracle.py convective), and on the filtered averages of the Mach 2.5 zero-gradient layer kept in the .wmf of two runs (boundary layer, wmf_apriori.py), over the developed stretch 45≤x<80 (30240 wall columns, each column's x from its block's Morton code), the stretch the runs' gate measured. The criteria were fixed before the results: on the layer, the mean τw within 2 % of the equilibrium model's; on each mean field, from x=54 on, at most four sign changes of Cf, and Cf<0 only under a first cell below half its incoming speed. The models keep the resolved stress removed from the forcing.

Velocity scale, u2=; forcinglayer 45≤x<80, mean τw against the equilibrium model (rs, rs2)sign changes of Cf (S4b, S4e, S4i)of them beyond ±0.2⋅10−3 (after the fact)fastest first cell under Cf<0, of its incoming speed
(|τw|+|S|)/ρ; the column's (the P3 intermediate)+9.7, +6.5 %4, 4, 64, 2, 40.76, 0.60, 0.80
|τw|/ρ (S4i)−5.0, −6.7 %23, 30, 2217, 16, 160.93, 0.89, 0.99
|τw|/ρ+up2, up=(μw|dp/dx|/ρw2)1/3 (Duprat et al. 2011)−3.7, −5.5 %17, 27, 2411, 12, 160.90, 0.89, 1.01
(|τw|+|dp/dx―|y)/ρ, the slow gradient−2.7, −5.0 % (with the code's filter: −1.2, −3.5 %)10, 7, 102, 2, 40.66, 0.89, 0.74
|τw+S¯|/ρ with the column's forcing in the balance−6.5, −4.1 %4, 4, 64, 2, 40.76, 0.60, 0.80
|τw+S|/ρ, the column's own: the local modelled stress (Park & Moin 2014's mixing length)+2.8, +1.1 %4, 4, 64, 2, 40.76, 0.60, 0.80
the same, κ matched to the resolved slope at h (Kawai & Larsson 2013's matching)−32.6, −30.6 %4, 6, 44, 4, 40.86, 0.83, 0.87
(|τw|+|S¯|)/ρ; the slow forcing S¯ in the balance too+1.7, −0.7 %4, 4, 64, 2, 40.76, 0.60, 0.80
dp/dx alone (S4d's model, for reference)+4.7, +3.7 %4, 4, 22, 2, 20.93, 0.40, 0.96

The slow forcing and gradient: on a mean field, its own (so on the mean fields the last row is the first, and the local stress the total stress of the slow forcing); on the layer, the mean over the stretch, the limit of a long time filter in a homogeneous layer. Two .wmf: the runs rs and rs2 of the boundary-layer gate.

The velocity scales a priori on the mean fields

Cf of the convection-fed model with the resolved stress removed, under six velocity scales, on the mean fields of S4b, S4e and S4i; the mean u of the first cell dotted, half its incoming speed as a thin line. sbli_oracle.py convective, make_doc_figures.py --only sbli-scales.

  • The mean forcing of the developed layer is its physical one. At the matching height the forcing with the resolved stress removed averages −0.011τw (rs; +0.11 in rs2) on 45≤x<80, against a total-stress drop of −0.018τw in the DNS at the same y/δ (0.108); the −0.15τw of all the wall columns comes from the recovery behind the inflow (−0.21, −0.25 on 20≤x<40).
  • The bias is the fluctuating forcing inside the balance, not the scale alone. With the column's forcing every τw-type scale lowers the layer's τw by 3 to 7 %, and |S| in the scale raises it by 6 to 10 %; with the slow forcing in both the balance and the scale the layer is the equilibrium model's within 2 % (the only row that passes), and on the interaction it is S4h's |τw|+|S|, which does not chatter (4, 2, 4 sign changes beyond ±0.2⋅10−3) and keeps the eddy viscosity under the adverse gradient. Park & Moin's local stress with the column's forcing is close (+2.8, +1.1 %).
  • Duprat's scale is a near-wall scale. At the foot of the compression ρwup2≈1–4⋅10−4, against a forcing carried up to h of S(h)≈(dp/dx)h≈2⋅10−2: the eddy viscosity still vanishes with τw and Cf chatters.
  • Matching κ to the resolved slope at the third cell halves it (0.42–0.50 of 0.41 on the layer, 0.54–0.57 upstream of the interaction): the resolved velocity gradient there is too steep for the log law (the log-layer mismatch of wall-modelled LES), and the layer's τw falls by a third. The .wmf holds no fourth-cell averages, so on the layer the slope is one-sided, at 2Δy.
  • The second interaction criterion is failed by nearly every model, the pressure gradient alone included (it passes on S4e's field only, where the first cell reverses): these mean fields come from runs whose near-wall flow does not slow, so a model forced by the measured pressure rise disagrees with them by construction. A priori, on another model's mean field, it cannot be met; only a run tests it.
  • The a priori's absolute level is worth a few per cent. On the layer it puts S4i's scale at −5 to −7 %, while its run (rs2) kept Cf within 0.3 % of the equilibrium run over the same stretch: the sample is rebuilt from the averages (the speed of the third cell, the free-stream pressure, a fixed temperature), so differences between the rows are firmer than each row against the 2 % criterion.

The run: S4j ​

The slow-forcing model the a priori above favoured, after its zero-gradient gate (boundary layer: the forcing filtered over 100 keeps the layer within 1.9 % of the equilibrium model's, over 50 within 2.2 %): wall_model_resolved_stress, wall_model_velocity_scale = sum and wall_model_pressure_gradient_filter = 100. S4b restarted at t=1200 to 1650 on one node (4 A30, 10 h 48 min); its forcing averages start fresh, so the statistics take the 25 checkpoints of t≥1500, where the first sample weighs e−3; the olive line of the figure above.

S4j, the slow forcingS4h, fed with the resolved convectionS4i, scale from τwS4bDNS, Reθ=6882
Lint/δ0 (the time-mean field)3.123.362.943.163.30
Lsep/δ0 from τw0.14 (x∗=−0.62 to −0.57)0.18 (x∗=−0.66 to −0.61)3.01, under a forward first cell02.16 (x∗=−0.8 to −0.15)
Cf minimum (10−3)−0.13 at x=56.7−0.30−3.410.21
time-mean u<0 in the first cellnone0.41δ0nonenone
5 % / 50 % of the rise, ahead of ximp2.09 / 1.85δ02.63 / 2.231.85 / 1.592.34 / 1.964.13 / 3.60
first-cell Mach number at ximp−5δ00.960.910.980.980.83
reversed area per unit span, mean (max), δ020.049 (0.071)0.0910.033 (0.054)0.068 (0.142)
first / third cells reversed7.1 % / 3.5 %4.8 % / 2.5 %9.5 % / 5.1 %
Cf upstream against S4b: x=15–25, 25–35, 35–45, 45–52+1.7, −1.9, −1.2, 0.0 %+5, +6, +9.5, +12 %−0.4, −6.0, −5.8, −2.5 %
Cf behind the interaction against S4b: x=64–70, 70–80−9, +1 %+30, +24 %−36, −11 %
  • The incoming layer is S4b's. Upstream Cf stays within 2 % of S4b's, as the zero-gradient gate promised, where S4h's rose by up to 12 % and S4i's fell by 6 %.
  • The interaction is S4b's too. Lint is S4b's within 1.3 %, the compression starts 0.25δ0 later than S4b's, the first cell is as fast (Mach 0.96) and reverses less (7 % of the cells against 9.5 %). The wall stress dips to zero where the DNS's bubble begins, as S4h's did, over 0.14δ0; the resolved first cell does not reverse in the mean.
  • So S4h's gains do not survive the consistent model. S4h moved the onset 0.3δ0 upstream and Lint to the DNS's, but with a forcing filtered over 10 time units, which follows the interaction's unsteadiness, and a zero-gradient layer 5 to 12 % too high; filtered over 100, about half the low-frequency period (StL≈0.03, about 180 time units), the forcing is close to the mean, the zero-gradient layer is right, and the interaction returns to the equilibrium model's. Whether S4h's gain came from its response to the moving shock foot or from the bias of its incoming layer is not separated by these runs (a hypothesis).

Where the wall-modelled interaction stands ​

The S4 wall-model line is closed (#121). Eleven runs after S4b changed one thing each: the grid near the wall, the model's balance, its eddy viscosity, its matching height, its coupling to the face. The criterion throughout: consistent on the zero-gradient layer (its Cf within about 2 % of the reference) and closer to the DNS in the interaction.

runwhat changed against S4bLint/δ0Lsep/δ0 from τw5 % of the rise ahead of ximpincoming Cf against S4b
S4bthe equilibrium model, δi/123.1502.34δ0
S4ca 2:1 box at the wall, h halved in it3.0402.33+10.9 %
S4dthe pressure gradient in the model3.120.602.12−1.7 %
S4eno eddy viscosity below h3.5103.55+4.8 %
S4fS4d without the eddy viscosity below h3.370.143.07
S4gthe Bernoulli-consistent pressure gradient3.2402.44
S4hfed with the resolved convection3.360.182.63+5 to +12 %
S4iand the resolved stress removed2.943.01, under a forward first cell1.85−0.4 to −6 %
S4jthe slow forcing3.120.142.09within 2 %
S4kS4c's box, h fixed in δi3.260 (first cell reversed over 1.10)2.58+1.7 %
S4lS4k with the viscosity coupling3.641.663.48+3.3 % (the flat plate −8.7 %)
DNS, Reθ=68823.302.164.13

The incoming Cf is measured over 35≤x≤48 (S4c, S4k, S4l), 35≤x≤50 (S4d) or 35≤x≤45 (S4e), and as the range over the upstream bands for S4h, S4i and S4j; S4f and S4g were not measured there.

  • No variant passes both. Those consistent on the zero-gradient layer (S4d, S4j, S4k) keep the compression at least 1.5δ0 behind the DNS's, and their wall stress separates over at most 0.6δ0, S4d's by an inconsistent balance (the Bernoulli-consistent S4g has none). Those that move the onset or the bubble toward the DNS (S4e, S4h, S4l) shift the zero-gradient layer by 3 to 12 % (S4l by −8.7 % on the flat plate), or reverse the stress under a forward flow (S4i).
  • The interaction length is within 10 % of the DNS's in every run but S4i (10.9 % short) and S4l (10.3 % long, the longest bubble); the separation and the onset are not reproduced by any.
  • Where the gap is. The runs' reversed flow lies below the wall model's sample, in the first two or three cells (above); a model that sees only the sample keeps a forward stress over it (S4k), and coupling the stress to the first cell (S4l) makes it separate but slows the near-wall cells of the zero-gradient layer too.
  • Not taken: a hybrid RANS/LES near the wall (#127); the viscosity coupling without the lower bound on the wall's subgrid viscosity (#126); a wall-resolved grid over the interaction.

Not done ​

  • A separation of the right length and place, and the onset: no wall model tried here gives them with the zero-gradient layer kept (where the interaction stands); the S4 wall-model line is closed. The routes left: a hybrid RANS/LES near the wall (#127), a wall-resolved grid over the interaction.
  • The late compression (x∗≈−0.6 half-way against the DNS's −1.1), unchanged by the near-wall grid and by the pressure-gradient term; moved 40 % of the way by taking out the eddy viscosity below h (S4e), which also moves the incoming layer off the DNS's: the matching between the model and the resolved near-wall cells.
  • The wall-pressure fluctuations and the low-frequency unsteadiness.
  • The recycling–rescaling inflow (S5).