Initial CFD analysis of the gnuA7 wing with OpenFOAM

A deeper look at the wing with OpenFOAM: a wind-tunnel check of the codes, the 2D checks and the open air intake, the 3D mesh and the billow valleys, and a first use of the 3D pressures in the Playground

A report for the LEparagliding open-source community  |  Francois de Villiers  |  16 November 2026, updated 17 November 2026  |  Initial version

How to read this report

Every result carries one of three labels. Measured means a run or a file exists. Inferred means the result is a conclusion drawn from measurements and was not tested directly. Not done means what it says. Open points are listed as plain facts in section 10, with no recommendations, and the checks planned next are in section 11. Each figure has a caption that says what is drawn and what it shows. Terms such as calage, K, plumb point, x/c and FSI are explained in section 2.3.

The wing geometry, the design file format and the design numbers come from LEparagliding (Laboratori d'Envol) and from LEP Studio 0.4.2 (Michael Holm) but with big modifications over the last two months. OpenFOAM, Flow5 (with its XFoil), gmsh and ParaView were used as tools, and none of them was modified.

1. Summary

2. Introduction and set-up

2.1 Why OpenFOAM, and what this report covers

The Playground and the aerodynamic loop of the project need pressures on the canopy. Flow5's panel method delivers them quickly. Its pressures are those of potential flow: its viscous option adds skin friction, but in a test on the gnuC2 wing at 8 degrees it left the lift unchanged (1033.9 N without, 1032.9 N with the viscous option). The pressures therefore show no separation and no loss of lift from the boundary layer. A wind-tunnel test of a rigid arched wing (section 3) indicated that the other codes over-predict the lift slope, so a viscous solver was needed as a second opinion. OpenFOAM was chosen because it is free, runs on the PC used for the rest of the work, and can write its surface pressures in the file format that the Playground already reads.

This report follows the work in the order it happened. It starts with a wind-tunnel referee for the codes (section 3), because the 3D CFD differed greatly from the panel method. Then comes the 2D study of rib 1, where a fully resolved mesh is affordable: the boundary-layer and orthogonal-stacking analyses, the closed and the open air intake, and the internal pressure over the angle sweep (sections 4 to 6). After that the work goes back to 3D, with what was learned about stacking and about the billow valleys (section 7). The last part uses the 3D billow pressures in the Playground (section 8). It flies the wing at 8, 9 and 10 degrees by the calage and compares the result with the Flow5-based flights of the earlier report. The billow mesh is not yet satisfactory and the best method for the valleys is not settled, and the report says where.

2.2 The wing, the conditions and the reference results

2.3 Terms used

2.4 Set-up of the runs

ItemSetting
Solver and modelOpenFOAM ESI v2512 in WSL2 (Ubuntu 22.04); steady simpleFoam; kOmegaSST; wall functions (nutUSpalding, omegaWallFunction, kqRWallFunction); no transition model unless stated; linearUpwind for velocity, first-order upwind for k and omega unless stated
3D meshsnappyHexMesh in two passes (castellate and snap, then layers only); 2 to 3.4 million cells (4.15 million for the finest mesh); refinement of the nose and of a band along the billow creases in the later meshes; stack 8 to 48 mm
2D meshgmsh quad layers on a cubic spline through the LEP foil points, converted with gmshToFoam; 43 thousand to 460 thousand cells; every mesh is a named preset of one script
StoppingForce-based: stop when CL, CD and Cm each vary less than a small band over 100 iterations; residuals alone do not reach 1e-6 on these meshes
HardwareOne laptop-class PC (6 cores, 32 GB; WSL sees 15 GB); 6 MPI ranks; 25 to 110 minutes per 3D angle when restarted from the neighbouring angle

3. Is the modelling correct? A wind-tunnel referee

The 3D OpenFOAM lift at 9 degrees is about 16 % below the Flow5 panel method (section 7.3). The question was whether viscous RANS or the panel method is closer to a real wing. A published wind-tunnel test of a rigid arched wing served as the referee (Belloc, J. Aircraft 52(2), 703-708, 2015). That wing is a rigid 1/8-scale model with a NACA 23015 section and 12 line segments, with a flat area of 28.56 m2 at full scale, measured in an open elliptical 3 m by 2 m test section at a Reynolds number of 0.92 million (central chord) and a turbulence intensity of 0.5 %. It is not the gnuA7. It is a benchmark for the codes and not for the gnuA7 numbers. The paper applies no wall corrections and says nothing about transition. The same STEP model that was used in October for the ParaCFD (a GPU cut-cell code), VSPAERO and Flow5 runs was run in OpenFOAM with the same dictionaries and layer settings as the billow work (2.45 million cells).

Reading (inferred): viscous RANS is much closer to the tunnel than any of the other codes. This supports the view that the panel method over-predicts the lift of a real arched wing. It does not show that the OpenFOAM numbers are correct to within a few per cent.

Figure 1. Lift slope of the Belloc wing: tunnel against codes. Left: horizontal bars of dCL/dalpha per radian (flat-area reference). Black: the wind tunnel (3.01), also marked by the dashed vertical l

Figure 1. Lift slope of the Belloc wing: tunnel against codes. Left: horizontal bars of dCL/dalpha per radian (flat-area reference). Black: the wind tunnel (3.01), also marked by the dashed vertical line; grey: two lifting-line results quoted in the paper; red-orange: VSPAERO; blue: Flow5 (inviscid); green: ParaCFD with the wall law off (slip) and on; orange and brown: OpenFOAM at 10 m/s and at the tunnel's Reynolds number. Right: the OpenFOAM CL against angle, from which the OpenFOAM slopes are taken; the two Reynolds numbers lie on top of each other.

Figure 2. The Belloc wing: tunnel (digitised from the paper) against OpenFOAM. Top left: lift curve; black diamonds are the tunnel data, grey dashed and dotted lines show the paper's flat-wing correct

Figure 2. The Belloc wing: tunnel (digitised from the paper) against OpenFOAM. Top left: lift curve; black diamonds are the tunnel data, grey dashed and dotted lines show the paper's flat-wing correction (illustrative), orange and brown are OpenFOAM at 10 m/s and at 4.93 m/s (tunnel Reynolds number). Top right: polar, CL against CD. Bottom left: pitching moment about the central quarter chord. Bottom right: pitching moment about the centre of gravity, 1 m below the quarter chord on the model. The OpenFOAM lift curve is parallel to the tunnel's but shifted up; the moment about the quarter chord is offset by about 0.02.

Why the work went back to 2D. The tunnel test shows which code is closer to a real wing, but not how accurate a 3D mesh is. A 3D wall-function mesh cannot be checked against a fully resolved 3D mesh, because that is impossible on this PC. A 2D section can be checked. Rib 1 can be meshed with a first cell 0.04 mm high and tens of layers at a small fraction of the 3D cost, compared with XFoil (the program inside Flow5), and then degraded to the kind of mesh that snappyHexMesh builds in 3D, to see what that costs. This is the 2D study of section 4. In 2D the 3D recipe was therefore tested before anything more was built in 3D.

4. The 2D study of rib 1

4.1 Foil, mesh presets and the reference

The foil is the LEP rib-1 section (gnua.txt scaled by 0.9; chord 2.91336 m; the same points that LEP writes for Flow5/XFLR5), at a Reynolds number of 1.77 million, with kOmegaSST and the 3D settings (inlet turbulence 0.5 %, eddy-viscosity ratio 10). Every mesh is one named preset of a single script. The preset sets the surface cell size, the nose and trailing-edge sizes, the wake size, the first-layer height, the growth ratio, the stack thickness and the trailing-edge treatment. One script run makes the mesh, and a second one converts it, builds the case and checks it. The reference is the fully resolved mesh with a flat trailing-edge base ("st3rkx": 0.04 mm first layer, growth 1.09, one uniform 202 mm stack, 457 thousand cells). At 9 degrees it gives CL 1.0634 and CD 0.0191.

First 2D results (9 November, measured). A mesh that copies the 3D recipe (42,651 cells, four layers of 1.5 mm, y+ 17 to 20) gave CL 0.959 and CD 0.0232 at 9 degrees. XFoil with a trip at 2 % chord gave 1.033 and 0.0162, and XFoil with free transition gave 1.157. Early resolved meshes gave 1.11. The wall-function mesh separated at 78 % chord. Both references were revised later, as sections 4.3 and 4.5 show.

4.2 Analysis A: how thick is the boundary layer compared with the mesh?

The edge of the boundary layer was extracted from the resolved solution at every angle and drawn on the meshes (Figures 3 and 4). At 9 degrees on the upper skin it is 6, 14, 24.5, 37, 49, 62, 75, 98, 125 and 148 mm thick at x/c 0.1, 0.2 and so on to 0.95 (the 1 % total-pressure edge, measured). The "resolved" mesh used first has a 71 mm stack, so it contained the boundary layer only to about 65 % of the chord and was not resolved in the aft third. A stack of 12 layers and 59 mm covers the boundary layer only forward of x/c 0.6. It has 8 to 22 cells inside the layer where the reference has 33 to 53, and it ends inside the layer. The lower skin is thinner (10 to 39 mm). At higher angles the layer grows quickly: on the upper skin at x/c 0.9 it is 86 mm at 4 degrees and 369 mm at 15 degrees, and it leaves the 71 mm structured band at x/c 0.7 at 14 degrees and at x/c 0.45 at 15 degrees.

The cell count alone does not decide the result. The 3D-style mesh (7 to 20 cells in the layer) and the 59 mm stack (8 to 22 cells) have similar counts, yet their lift differed by 13 % (0.959 against 1.088 at 9 degrees; measured, before the trailing-edge artefact of section 4.3 was found). A stack that follows the local boundary-layer thickness (59, 108 and 159 mm over three parts of the chord) appeared to recover nearly all of the gap to the 165 mm mesh (CL 1.1407 against 1.1593). Most of that gain turned out to be the trailing-edge artefact described next.

Figure 3. The real boundary-layer edge drawn on the reference 2D mesh (angles 4 to 15 degrees). Six windows of the mesh (upper skin above, lower skin below) at x/c 0.30 to 0.42, 0.58 to 0.72 and 0.84

Figure 3. The real boundary-layer edge drawn on the reference 2D mesh (angles 4 to 15 degrees). Six windows of the mesh (upper skin above, lower skin below) at x/c 0.30 to 0.42, 0.58 to 0.72 and 0.84 to 0.98; the thin grey lines are mesh cells (the dense band next to the wall is the prism stack, the triangles outside it are the free mesh). Each coloured line is the edge of the boundary layer at one angle (dark blue 4 degrees to orange 15 degrees); the text boxes give the number of cells inside the layer and its thickness in mm (for example 42 cells in 86 mm at 4 degrees and 64 cells in 369 mm at 15 degrees near the trailing edge). The edge climbs out of the structured band on the upper skin for the higher angles; on the lower skin the layer stays thin.

Figure 4. The boundary layer against a 59 mm wall-function stack. The same windows for the 12-layer, 59 mm stack (mesh "th12"). The red line is the real boundary-layer edge of the 9-degree resolved so

Figure 4. The boundary layer against a 59 mm wall-function stack. The same windows for the 12-layer, 59 mm stack (mesh "th12"). The red line is the real boundary-layer edge of the 9-degree resolved solution, the blue lines are wall normals and the labels count the cells between the wall and the red line (for example 8 cells in 25 mm at x/c 0.30, 18 cells in 126 mm at x/c 0.90 on the upper skin; 5 to 10 cells on the lower skin). Upper skin: the red line leaves the structured stack at about 60 % chord.

4.3 Analysis B: orthogonal stacking and the trailing edge

A good stack has its cells perpendicular to the wall all along the foil. At the sharp trailing edge this is impossible. gmsh closes the stack with a fan of slanted cells, a semicircular cap whose radius equals the stack thickness, and all the highly skewed faces of the mesh sit at the apex of that fan. Because the fan grows with the stack, it looked like a gain from a thicker stack. The trailing edge was therefore replaced by a flat base: the foil is cut 2 mm short of the tip (0.07 % of the chord), which leaves a 0.70 mm flat base. The layers end in a clean cut and the gap behind is filled with triangles (Figure 5).

At 9 degrees (measured) the flat-base meshes with different stacks give CL 1.0538 to 1.0640, a spread of 1.0 %. The same stacks with the fan give 1.0876, 1.1407 and 1.1622, a spread of 6.9 %, and a fan limited to four elements gives 1.0631. The earlier statements that a thick or stepped stack adds 4 to 7 % of lift, and the 165 mm "fully resolved" reference of 1.159, were therefore artefacts of the fan. Every earlier percentage of "gap closed" that used them is withdrawn. With a flat base the stack thickness matters by less than 1 % in the thick-stack regime. The boundary-layer thickness itself is the same with and without the fan, so the fan changed the circulation and not the boundary layer. The flat base adds a small base drag that XFoil does not have. The layer cells are perpendicular to the wall along the whole foil by construction. Only a few faces exceed the non-orthogonality limit of 70 degrees (2 to 29, depending on the mesh), and they sit at the trailing edge and, in the open mesh, at the cavity end. The flat-base meshes still report two failed checkMesh categories (small-determinant cells and low-weight faces). With a flat base the forces of the different presets agree within 1 %.

Figure 5. The flat trailing-edge base. Top: the last 30 mm of the rib-1 foil drawn to true scale (1 mm along is 1 mm up). The grey shape is the foil, the black dot is the sharp tip; the red bar is the

Figure 5. The flat trailing-edge base. Top: the last 30 mm of the rib-1 foil drawn to true scale (1 mm along is 1 mm up). The grey shape is the foil, the black dot is the sharp tip; the red bar is the flat base made by a 2 mm cut (0.70 mm high), the blue bar the larger base of a 6 mm cut (2.11 mm high), and the coloured areas are what is cut off. Bottom: a zoom on the last 9 mm with the same scale in both directions. Even the 6 mm cut removes only 0.2 % of the 2913 mm chord.

What the real trailing edge is, and what the model leaves out (described by the author, not measured in this study). On the wing the cloth of the upper and lower skin is folded three or four times at the trailing edge into a flat seam about 1 cm long and 1 to 1.5 mm thick (Figure 5b). The thickness is an estimate, not yet resolved, and is left for fine tuning. The internal pressure pushes the two skins apart, so they bulge into an almost round profile that runs into this seam. The 0.70 mm flat base of the CFD is of the same order as the seam thickness, but it is a cut foil with a flat base and not a rounded bulge feeding a seam. None of this has been modelled correctly in any run of this report (not done). The trailing edge was checked thoroughly on a single open cell of the gnuC2 wing in earlier work. The same check for the gnuA7, on a single cell in the Playground, has not been made yet (not done). The author wants the following analysed at every angle of attack (not done): the radius of the rounded end that the internal pressure produces, the seam, the back (the base behind the seam) and the mesh around them.

Figure 5b. The real trailing edge, in the author's hand sketch (photograph of the sketch, not to scale). The upper skin and the lower skin come together at the trailing edge. The small arrows are the

Figure 5b. The real trailing edge, in the author's hand sketch (photograph of the sketch, not to scale). The upper skin and the lower skin come together at the trailing edge. The small arrows are the internal pressure pushing both skins outwards, which rounds the end. A flat seam leaves the round end towards the back: about 1 cm long ("1 cm") and "1 - 1.5 thick" (the unit is not written on the sketch; millimetres are meant, as in the text). The CFD of this report replaces all of this by a foil with a 0.70 mm flat base (Figure 5).

4.4 Wall-function stacks against the resolved mesh

With the flat base, the stack was reduced step by step on a surface mesh of the 3D kind, to find out what a wall-function mesh costs. All runs are at 9 degrees, against the resolved reference of CL 1.0634 and CD 0.0191 (measured):

StackCL against the referenceCD against the reference
59 mm (12 layers)-0.6 to -0.8 %+6 to +7 %
48 mm (2 mm x 8, growth 1.3)-1.3 %+5 %
34 to 40 mm-2.6 to -3.6 %
17 to 19 mm-4.6 to -5.6 %+12 to +13 %
8 mm-7.6 % (CL 0.982)+17 % (CD 0.0224)

Fewer layers with faster growth do not help, and y+ of 20 against 38 makes no difference. The thickness of the stack decides, not the first-cell height. Wall functions also fail earlier at high angle: at 12 degrees the lift is 2 to 4 % low even with the 59 mm stack, and at 15 degrees the wall-function runs break down (not converged). On the closed foil the usable range of the wall-function approach is therefore up to about 9 to 11 degrees.

Figure 6. Wall-function meshes against the resolved mesh on rib 1. Top row: Cp (suction upwards) on both skins at 0, 3, 9, 12 and 15 degrees; bottom row: y+ of the first cell on a logarithmic scale. B

Figure 6. Wall-function meshes against the resolved mesh on rib 1. Top row: Cp (suction upwards) on both skins at 0, 3, 9, 12 and 15 degrees; bottom row: y+ of the first cell on a logarithmic scale. Black: resolved mesh (first cell 0.04 mm, y+ below 1); blue dashed: 1.5 mm first layer (y+ about 20); red dotted: 3 mm first layer (y+ about 40). The pressure distributions are the same up to 9 degrees; at 12 degrees the wall-function suction peak is slightly lower, and at 15 degrees the wall-function runs have lost the suction peak (-2.4 and -3.0 against -4.2) and the flow is separated.

4.5 Polar, turbulence and transition

The resolved flat-base mesh was run at 4 to 15 degrees (measured, Figure 7). CL is 0.574, 0.778 and 1.064 at 4, 6 and 9 degrees, then 1.153, 1.236, 1.312, 1.378, 1.422 and 1.420 at 10 to 15 degrees. CD rises from 0.0210 at 10 degrees to 0.0437 at 15 degrees. The maximum lift of about 1.42 is reached at 14 to 15 degrees with a soft stall, and the separation point moves from 96 % chord at 11 degrees to 62 % at 15 degrees.

The comparison with XFoil showed that the OpenFOAM runs are not "fully turbulent" (a wrong label that was used earlier and is withdrawn). kOmegaSST has no transition model, so the layer turns turbulent gradually over x/c 0.1 to 0.4 (eddy viscosity ratio 0 at 0.005, 6 at 0.05, 13 at 0.1, 45 at 0.2 and 118 at 0.4 on the upper skin). XFoil's own lift at 9 degrees depends on where the transition is placed: 1.023 with a trip at 0.1 %, 1.033 at 2 %, 1.054 at 5 %, 1.089 at 10 %, 1.110 at 15 % and 1.157 free. The flat-base OpenFOAM lift (1.054 to 1.064) lies on the 5 % trip, but the drag is higher than in every XFoil case (0.0192 against 0.0109 free and 0.0153 at a 5 % trip). The earlier statement that kOmegaSST "trips" at 4.5 to 5 % chord was an over-reading of the shape factor and is withdrawn. Changing the inlet turbulence from 0.1 to 2 % and the viscosity ratio from 1 to 10 changed nothing (CL 1.0639 to 1.0644). The turbulence is produced near the nose and is not carried in from the inlet.

A transition model (kOmegaSST with the gamma-Re-theta transition model) was run at 9 degrees only (measured, Figure 8). It gives CL 1.117 to 1.129 and CD 0.0140 to 0.0142, against 1.064 and 0.0192 for the baseline. The upper skin is laminar to x/c 0.14, has a laminar bubble between 0.14 and 0.17 and is turbulent after 0.2. This agrees with XFoil's natural transition and with the first seam of the real wing at 400 mm (x/c 0.137 on rib 1). The lower skin stays laminar to the trailing edge, as in XFoil. The 2D model has no air intake, so this is an upper bound on the laminar run. XFoil's natural transition on the upper skin lies at x/c 0.124 to 0.241 over critical amplification factors of 2 to 12 and angles of 0 to 10 degrees. The transition model was not used for the 3D work (not done in 3D), so the 3D drag includes a layer that turns turbulent gradually rather than at the seam (inferred from the 2D behaviour).

Figure 7. The 2D polar set of rib 1 (Reynolds number 1.77 million). Top row: lift curve, drag against angle, polar (CL against CD) and pitching moment about the quarter chord (nose-up positive). Botto

Figure 7. The 2D polar set of rib 1 (Reynolds number 1.77 million). Top row: lift curve, drag against angle, polar (CL against CD) and pitching moment about the quarter chord (nose-up positive). Bottom row: lift-to-drag ratio, and skin friction coefficient Cf along the upper and the lower skin at 9 degrees. Orange diamonds with a solid line: OpenFOAM kOmegaSST (4 to 15 degrees); blue circle and square: OpenFOAM with the transition model at 9 degrees (two inlet settings); blue dashed and dotted: XFoil with free transition; orange dashed: XFoil with both skins turbulent from the nose (trip at 0.1 % on the upper and 2 % on the lower skin); brown dash-dot: XFoil with both skins tripped at 5 % chord. A table at the lower right lists the 9-degree values (for example kOmegaSST CL 1.064, CD 0.0192; transition model 1.129, 0.0142; XFoil free at critical amplification 4.3: 1.105, 0.0109). Free transition gives much less drag and a laminar run on the upper skin to 14 % chord; the kOmegaSST drag lies above all of the XFoil cases.

Figure 8. The transition model against kOmegaSST (9 degrees, first 60 % of the chord). Top row: upper skin; bottom row: lower skin. Left: skin friction Cf; middle: boundary-layer shape factor H (dotte

Figure 8. The transition model against kOmegaSST (9 degrees, first 60 % of the chord). Top row: upper skin; bottom row: lower skin. Left: skin friction Cf; middle: boundary-layer shape factor H (dotted line H = 2); right: the intermittency gamma of the transition model. Grey: kOmegaSST baseline; red and orange: the two transition-model runs; dash-dot lines: XFoil free transition and a 5 % trip. With the transition model, Cf on the upper skin follows XFoil: low and laminar up to x/c 0.14, a short bubble (Cf touches zero at x/c 0.14 to 0.17) and the jump to the turbulent value; the intermittency jumps from 0.4 to 1.0 at x/c 0.15. On the lower skin the layer stays laminar (gamma 0.2).

4.6 Towards an automatic mesh

The aim of the mesh work is that a new foil or a new angle needs no hand work. This is the state on 16 November (measured where a script exists):

5. Closed and open 2D: the air intake

The real wing breathes through an air intake on the lower surface near the nose. In LEP's input (Section 2, the columns "intake in" and "intake out") the intake of the gnuA7 is the lower-surface part from 1.6 to 6.3 % of the chord on every rib. Ribs 1 to 14 are open and ribs 15 to 17 (the tip) are closed. On rib 1 this is a mouth 145 mm long, from 46.6 to 183.5 mm behind the nose, 72 to 121 mm below the chord line and tilted -19.5 degrees to the chord. In the terms of Belloc et al. (Int. J. Aerodynamics 5(2), 83-104, 2016, the air-inlet study of an open paraglider airfoil) the width is 4.9 % and the position 5.5 % of the lower-surface length. That is wider than his widest case (4 %) and slightly further back than his furthest (5.3 %); his best case was 1 % wide at 2 %.

The model (measured). The skin is a 1 mm thick solid with the inner face offset along the normal. The cavity reaches to 94 % chord and has a rounded end, because a flat end wall produced skewed cells. The lips are 1 mm thick. The reinforced edge of the real intake is at least that thick, so the lip end faces are physical and not an artefact. Separate patches for the outer faces, the inner faces, the lips and the cavity end make it possible to read the force on each. There are two meshes: op1 (261,584 cells, thick stack on the whole inner skin) and op2 (255,729 cells, thick stack only over the first 20 % of the chord inside and a thin 16-layer stack behind). Their forces agree within 0.1 % in lift and 0.4 % in drag (Figure 9).

Result at 9 degrees (measured, Figures 10 to 12). The total CL is 1.0666 and CD 0.01976 on op1, against 1.0635 and 0.01916 for the closed foil: +0.3 % lift and +3.1 % drag. With op2 and the finer closed reference of the later sweep the figures are +0.15 % and +3.4 %. On op2 the outer faces carry CL 1.0258 and the inner faces the rest (CL 0.0396, CD 0.0216), which is the pressure inside times the mouth area. The pressure inside is constant, Cp 0.913, equal to the outer pressure at the forward lip (0.915). The stagnation point sits at x/c 0.0097, just ahead of the mouth. On the closed foil the mean Cp across the gap would have been 0.779 and the maximum 0.965. The air moves slowly in vortices inside the cavity (below 0.2 of the free-stream speed). It enters over the front part of the mouth at a total pressure of about 0.99 and leaves near the aft lip. Behind the aft lip a separation bubble sits on the outside skin from 6.4 to 8.3 % chord. The Cp there dips to 0.11 and recovers to 0.55 at x/c 0.086. The bubble differs in size between op1 and op2 (the lip Cp is 0.27 against 0.12), so its detail depends on the mesh. The edge thickness of the lip, which is a parameter of the real material, probably sets its size (inferred, not tested). The external flow of the open foil is almost that of the closed foil. A check with Belloc's equation 5, which builds the open-foil force from the closed foil, shows the external flow changing little (CL -0.003, CD -0.002).

The pressure difference across the skin (inside minus outside, positive pushing the skin outwards) is 0.8 to 4.05 on the upper skin and 0.4 to 0.9 on the lower skin. The exception is a small inward push (-0.09) on the lower skin at x/c 0.006 to 0.016, just before the mouth (measured, Figure 12).

Figure 9. The mesh of the open foil (op2). Six windows. Top row, left to right: the whole foil (orange upper, blue lower skin, with the wake box at the right); the nose and the mouth; the mouth enlarg

Figure 9. The mesh of the open foil (op2). Six windows. Top row, left to right: the whole foil (orange upper, blue lower skin, with the wake box at the right); the nose and the mouth; the mouth enlarged. Bottom row: the forward lip, the aft lip and the rounded cavity end near the trailing edge, drawn to scale. The skin is a 1 mm solid (the white strip in the lip windows) with its own boundary-layer stack on the outer face and on the inner face. The mouth is the gap between the two lips; the large round region inside is the cavity, meshed with fine triangles.

Figure 10. Pressure on the outer and inner faces of the skin (open foil, 9 degrees). Three views of -Cp against x/c (suction upwards): the whole chord, the front 20 % and the lower skin from 0 to 10 %

Figure 10. Pressure on the outer and inner faces of the skin (open foil, 9 degrees). Three views of -Cp against x/c (suction upwards): the whole chord, the front 20 % and the lower skin from 0 to 10 %. The yellow band is the intake (x/c 0.016 to 0.063). Solid lines are the outer faces (red upper, blue lower), dashed lines the inner faces: both are flat at Cp 0.913 (-Cp of -0.913). The thin black line is the closed foil. The green bar is the mean pressure the closed foil has across the gap (0.779); the orange line is the pressure of the flow on the mouth line (0.91 at the front, 0.83 at the back); black squares are the lip end faces. After the aft lip the lower skin shows the dip and recovery of the separation bubble. The annotations give the stagnation point (x/c 0.0097, Cp 0.999).

Figure 11. The flow in and around the mouth (open foil, 9 degrees). Left: speed relative to the free stream (colour, 0 to 1.4) with streamlines, over the mouth and the cavity. The forward lip, the aft

Figure 11. The flow in and around the mouth (open foil, 9 degrees). Left: speed relative to the free stream (colour, 0 to 1.4) with streamlines, over the mouth and the cavity. The forward lip, the aft lip and the 145 mm mouth (red dashed) are marked; the pale yellow region is the slow air in the cavity with two vortices, and the blue region below the aft lip is the bubble outside the skin. Right: an enlargement behind the aft lip showing the horizontal velocity relative to the free stream (blue towards the trailing edge, red towards the nose); the magenta line encloses the reversed-flow bubble.

Figure 12. Pressure on both sides of the skin and the difference across it (open foil, 9 degrees). Left: upper skin, -Cp on the outer face (red) and the inner face (dashed black); middle: lower skin,

Figure 12. Pressure on both sides of the skin and the difference across it (open foil, 9 degrees). Left: upper skin, -Cp on the outer face (red) and the inner face (dashed black); middle: lower skin, same; right: the pressure-coefficient difference inside minus outside (red upper skin, blue lower skin; positive means outwards). The inner faces are at -Cp -0.9 everywhere; the difference is about 4 at the suction peak and 0.8 to 0.9 over most of the lower skin. The panel lists the forces: total CL 1.0654, outer faces 1.0258, inner faces 0.0396; internal Cp 0.912.

6. The internal pressure over the whole angle sweep

The open and the closed foil were run at 0, 3, 4, 6, 9, 10, 11, 12, 13, 14 and 15 degrees (measured). The 0, 3, 4, 6 and 15 degree runs stopped at the iteration cap of 1500. There the inner forces are flat to 1 to 2 %, and the total drag scatters by 4 to 10 % at 0 to 4 degrees.

Angle (deg)Cp inside the open foilClosed foil: Cp at the front lip (x/c 0.016)Inside / front lipInside / highest closed Cp across the gapStagnation point, open (x/c)CD open against closed
0-0.293-0.33188 %86 %-0.001+30 %
30.2090.32864 %68 %0.000+21 %
40.3170.49864 %66 %0.001+17 %
60.6010.76479 %81 %0.003+11 %
90.9120.97494 %94 %0.010+3.4 %
100.9640.99597 %97 %0.012+2.6 %
110.9800.99798 %98 %0.015+1.5 %
120.9950.981101 %99.6 %0.016+0.5 %
130.9990.941106 %99.9 %0.016+0.1 %
141.0000.902111 %99.9 %0.089+0.5 %
151.0000.861116 %100 %0.086+3.5 %

What drives the pressure (measured, Figures 13 and 14):

Limits (not done or open): the runs are steady 2D RANS for rib 1 only. There is no flow along the span between the cells of the real wing, no refinement study of the mouth mesh, and the low-angle mouth lines are wavy (9 degrees is smooth). The intake geometry was not changed in the CFD apart from the one LEP geometry. Only kOmegaSST was used; a Spalart-Allmaras run as in Belloc's study was not made. The seam height is not modelled: the cloth at the intake edge and at the seams is folded three or four times, its thickness is estimated by the author at about 1 mm and is not resolved, and the lip of the CFD is 1 mm.

Figure 13. The internal pressure of the open foil against the angle of attack (open against closed). Top left: Cp against angle; the thick orange line is the pressure inside the cavity (the pale band

Figure 13. The internal pressure of the open foil against the angle of attack (open against closed). Top left: Cp against angle; the thick orange line is the pressure inside the cavity (the pale band is the spread over the inner walls), the grey lines are the closed-foil pressure at the front lip, the highest and the mean across the gap and at the aft lip. Top right: the internal pressure as a percentage of three closed-foil values (front lip orange, mean green, highest grey), showing how the shortcut changes with angle (100 % = equal). Bottom left: position of the stagnation point on the lower surface for the open (orange) and closed (grey) foil, with the yellow band marking the intake; at 14 degrees the open stagnation point jumps behind the mouth. Bottom right: lift (solid) and drag (dashed) of the open (orange) and closed (grey) foil; lift lies on top, drag is higher with the intake open at low angle.

Figure 14. The flow through the mouth at 0, 3 and 9 degrees. Three columns, one angle each; the horizontal axis runs from the front lip (0) to the aft lip (1). Top row: speed normal to the mouth line,

Figure 14. The flow through the mouth at 0, 3 and 9 degrees. Three columns, one angle each; the horizontal axis runs from the front lip (0) to the aft lip (1). Top row: speed normal to the mouth line, green into the cavity and red out of it (the share of the mouth going out is given above: 57 %, 54 % and 35 %). Middle row: static Cp on the mouth line (blue) against the cavity Cp (dashed black); where the mouth-line pressure is above the cavity pressure air flows in. Bottom row: total pressure on the mouth line (purple). At 9 degrees the mouth-line pressure is near the cavity pressure over the front half, at 0 and 3 degrees it varies strongly along the mouth and the in- and outflow regions swap places.

Where the nose is pushed in (measured, Figure 14b). Pressure arrows show it more plainly than Cp. At 0 degrees the air stagnates on the nose and pushes the skin inwards, while the cavity (Cp -0.29) pulls the skin outwards. The net pressure pushes the skin inwards from the nose tip to x/c 0.038 on the upper skin and to x/c 0.008 on the lower skin, by up to 1.3 q (about 66 Pa) at the tip. At 3 degrees the inward zone is shorter and weaker: up to 0.8 q (about 40 Pa), within x/c 0.010 on the upper skin and 0.015 on the lower skin. At 9 degrees the upper nose is pushed outwards everywhere (suction outside, Cp 0.91 inside), and the lower skin is pushed outwards except for a small inward push (0.09 q, about 5 Pa) at x/c 0.010 just before the mouth. The computation is steady and the skin is rigid. How far the nose of the real, soft canopy deforms is not computed (not done).

Figure 14b. Pressure arrows on the nose of the open foil at 0, 3 and 9 degrees (rib 1, mesh op2). One column per angle; the front of the profile (x from the nose to 0.66 m behind it, x/c 0 to 0.23, x

Figure 14b. Pressure arrows on the nose of the open foil at 0, 3 and 9 degrees (rib 1, mesh op2). One column per angle; the front of the profile (x from the nose to 0.66 m behind it, x/c 0 to 0.23, x and z in metres,same scale in both directions) with the skin as a black line and the mouth (x/c 0.016 to 0.063, the gap in the lower skin) in yellow. Top row: the pressure on both faces of the skin, one arrow every 12 mm along the skin, 30 mm long per unit of Cp; the arrows outside the skin belong to the outer face, the arrows inside the skin to the inner face (the cavity). A red arrow points at the skin (positive Cp: the air pushes on the skin), a blue arrow points away from it (negative Cp: suction pulls on the skin). Bottom row: the net pressure on the skin (inner face minus outer face, same length scale): an orange arrow points at the skin (net push inwards, the nose caves in), a green arrow points away (net push outwards, the skin is inflated). The three grey arrows in front of the nose are the wind: the foil is drawn with its chord horizontal at every angle, so the wind comes from the front and from below, tilted by the angle of attack from the dotted line (the chord line); their length has no meaning. At 0 degrees the stagnation pressure on the nose (red, outside) is larger than the cavity pressure and pushes the nose in; at 3 degrees the push is weaker; at 9 degrees the suction on the upper nose and the cavity pressure push the skin outwards. The arrows get longer with angle, not shorter: the lowest Cp on the upper skin is about -1.3 at 0 degrees, -1.9 at 3, -3.1 at 9 and -4.2 at 15 degrees. (A first version of this figure and of the animation drew these long suction arrows in a window that was too small, so they were cut off at the higher angles, and it stopped the arrows at 0.47 m behind the nose on the upper and the lower skin while the skin line went on; the window now contains every arrow and the arrows run to its right edge.) An animation of all the angles that were run (0 to 4, 6 and 9 to 15 degrees) is Figure 14c below; each frame also carries the close-up of the Cp curve on the lower skin (as in Figure 9, without the closed foil) with the orange mouth-line curve, the pressure inside the cavity and the stagnation point.

Angle of attack9°
degrees

Pressure arrows on rib 1 at 9 degrees angle of attack

Figure 14c (slider). The same pressure arrows for every angle of attack that was run (rib 1, open air intake, mesh op2): 0 to 4, 6 and 9 to 15 degrees. Move the slider, press an angle or press Play. Left: arrows on the outer face (outside the skin) and the inner face (inside the cavity); red points at the skin (the air pushes), blue points away (suction); 30 mm of arrow per unit of Cp. Middle: the net pressure on the skin, inner face minus outer face; orange = pushed inwards (the nose caves in), green = pushed outwards. The circle is the stagnation point (the highest Cp on the outer face of the lower skin). The three grey arrows in front of the nose are the wind: the foil is drawn with its chord horizontal at every angle, so the wind comes from the front and from below, tilted by the angle of attack from the dotted line (the chord line); their length has no meaning. Right: the close-up of the Cp curve on the lower skin, x/c 0 to 0.10: blue = outer face, blue dashed = the inner face, orange = Cp from the flow field on the line between the two lips (black squares), shaded = the mouth; the numbers give the pressure inside the cavity, the stagnation point and the pressure at the lips. The arrows run along the skin to 0.66 m behind the nose. At 14 and 15 degrees the highest pressure on the lower skin lies behind the mouth, at x/c 0.086 to 0.089, below the pressure inside. The animated version of the same frames (0.9 s per frame) is the file images/skin-arrows-op2-sweep.gif in the package.

7. 3D: stacking and the billow valleys

7.1 What the 3D mesher can build

The 2D result calls for a stack thick enough to contain the boundary layer, or at least 40 to 60 mm. In 3D, snappyHexMesh with its default layer settings builds only 17 to 19 mm stacks reliably (measured, 13 November):

3D stack attempt (no billow)StackShare of the skin that got the layers (upper / lower)Comment
Band mesh8 mm98 % / 98 %3.40 million cells
wf1A17 mm (3 mm x 4, growth 1.25)96 % / 93 %2.01 million cells, y+ 38
wf1B19 mm (1.5 mm x 6, growth 1.3)about 95 %2.30 million cells, y+ 21
Default settings, thicker34 to 48 mm10 to 38 %
Default settings59 mm0.4 %
Finer surface (7.8 mm) with 48 mm48 mm0.1 %
Relaxed quality limits (T1, T2)47.7 mm20 % / 37 % and 18 % / 32 %mesh poor (non-orthogonality 79 to 84 degrees)
U1: four outer iterations and no layer growth limit47.7 mm79.7 % / 76.7 %2.47 million cells, non-orthogonality up to 75 degrees, 5 failed checkMesh checks; the layer pass took 2.4 hours
U2: displacement shrinker, U3: first-and-relative final layer47.7 mm6 / 13 % and 1.7 %failed
cfMesh (probed once)59 mm (62.5 mm wall cells)1,018 negative-volume cells, 10 failed checks: unusable

In the U1 mesh the missing layers sit outboard of 1 m from the centre (inboard they are nearly complete). The shape of the 3D wing, with its swept and arched outer part, blocks the thick stack that the same mesher builds at 99.6 % on a 2.5D slab of a single rib (all stacks from 7.7 to 59.3 mm). Figures 15 and 16 show the difference.

Figure 15. 3D meshes cut perpendicular to the span: 17 mm stack against 47.7 mm stack. Four rows, three columns each (whole section, nose, trailing edge). Rows 1 and 2: the 17 mm stack (coverage 95.7

Figure 15. 3D meshes cut perpendicular to the span: 17 mm stack against 47.7 mm stack. Four rows, three columns each (whole section, nose, trailing edge). Rows 1 and 2: the 17 mm stack (coverage 95.7 / 92.9 %) at 0.5 m and 3.0 m from the centre. Rows 3 and 4: the 47.7 mm stack (coverage 79.7 / 76.7 %) at the same two places. At 0.5 m the layers form a clean band around the nose in both meshes. At 3.0 m the 47.7 mm stack is broken up at the nose (row 4, middle) and the layers are missing along part of the lower skin.

Figure 16. The 2.5D slab of rib 1 with three stacks (all 99.6 % covered). Rows: 17.3 mm (4 layers), 47.7 mm (8 layers) and 59.3 mm (8 layers); columns: the whole foil, the nose, the trailing edge. The

Figure 16. The 2.5D slab of rib 1 with three stacks (all 99.6 % covered). Rows: 17.3 mm (4 layers), 47.7 mm (8 layers) and 59.3 mm (8 layers); columns: the whole foil, the nose, the trailing edge. The mesher builds all three stacks around the nose and along both skins to the trailing edge, which indicates that the failures of Figure 15 are due to the shape of the 3D wing and not to the stack thickness itself (inferred).

7.2 The billow valleys

The billow adds creases (about 130 degrees included angle, a skin kink of about 50 degrees) that run along the flow between the ribs. They cause two problems for the mesh and one for the physics (measured).

What the billow does to the section lift (measured): the billowed slab with the 8 mm stack oscillates by +-1.6 % and was stopped by hand at CL 0.870 and CD 0.0380, against 1.0710 and 0.0247 for the smooth twin. That is -18.8 % in lift. The cross-flow angle on the skin of the 3D billow wing is 3.4 degrees (median) inboard and 11 to 14 degrees outboard, the same as without billow (the arch), so the slab assumption of flow along the grooves was kept (inferred as good enough). A fully resolved billow valley, with a stack thick enough for the 36 to 59 mm layer, does not exist: this is not solved. The billow mesh is not yet satisfactory, and which method is best for the valleys is not decided. What snappyHexMesh and OpenFOAM can build and solve in a crease is limited (Figures 17, 18 and 19).

Figure 17. Where the first billow mesh has no layers. The wing seen from above (x aft, y along the span; nose on the left) with the skin faces that have no prism layers painted red: left the upper ski

Figure 17. Where the first billow mesh has no layers. The wing seen from above (x aft, y along the span; nose on the left) with the skin faces that have no prism layers painted red: left the upper skin (34.9 % of the faces without layers), right the lower skin (30.7 %). The red stripes run along the rib creases from the nose to the trailing edge; the grey patches near the nose are layered. This is the mesh with the refined band and the 3 mm x 4-layer stack, before the thin stack was introduced.

Figure 18. The boundary layer across three grooves (billow, band mesh, upper skin, x = 1.0 m). Top left: the skin unrolled along the arc length (horizontal, 0 to 1 m) and the height above it (vertical

Figure 18. The boundary layer across three grooves (billow, band mesh, upper skin, x = 1.0 m). Top left: the skin unrolled along the arc length (horizontal, 0 to 1 m) and the height above it (vertical, 0 to 60 mm); colour is the speed along the flow relative to the free stream (blue is slow, red is fast), the white line is the boundary-layer thickness, the black dots are mesh cell centres (rows of dots near the wall are the prism layers). Top right: speed profiles at five stations (the groove in red, the crowns and the flanks in other colours); the groove profile is much fuller of slow air. Bottom left: the boundary-layer thickness along the arc length: 15 to 23 mm on the crowns and peaks of 51 to 59 mm at the three grooves. Bottom right: the shape of the skin in the plane x = 1.0 m with the groove marked.

Figure 19. Layers in the billow slab (four variants). Rows: 8 mm stack with the refinement band (98.9 / 98.3 %); 8 mm stack without the band (99.7 %); 17 mm stack with the band (65 %); 47.7 mm stack w

Figure 19. Layers in the billow slab (four variants). Rows: 8 mm stack with the refinement band (98.9 / 98.3 %); 8 mm stack without the band (99.7 %); 17 mm stack with the band (65 %); 47.7 mm stack with the band (1 %). Columns: a cut across the flow at x = 1.0 m with the crease in the middle; a cut along the flow through the crease; a cut along the flow through the crown (the middle of a panel). The 8 mm stack follows the V-shaped crease; the 17 mm stack fills the crease with distorted cells; the 47.7 mm stack has almost no layers there, while the crown (right column) keeps a clean layer stack in every case.

7.3 What the 3D sweep shows, and what it does not

The no-billow wing was run from 3 to 15 degrees with three stacks: 8 mm at 3 to 15 degrees, and 17 mm and 47.7 mm at 3, 5, 7, 9, 11, 13 and 15 degrees (measured, Figure 20). The three stacks agree within 0.7 to 0.9 % in CL and 1 to 2.7 % in CD at every angle (CL at 15 degrees 0.899, 0.899 and 0.893). The lift slope between 3 and 11 degrees is 3.10, 3.12 and 3.08 per radian. The 17 mm run at 15 degrees did not converge: the plateau of iterations 100 to 300 is plotted (CL 0.899, CD 0.0945), and later the drag creeps up and unsteady separation sets in. The 8 and 48 mm runs at 15 degrees converged.

Against the other results at 9 degrees:

9 degrees, no billowCLCDCL/CDCm about the common point
3D OpenFOAM, 8 / 17 / 47.7 mm0.600 / 0.601 / 0.5950.0481 / 0.0483 / 0.047612.5 / 12.4 / 12.5+0.002 / +0.002 / +0.004
3D Flow50.7110.056512.6-0.028
Lifting line from the resolved 2D polar0.6280.051212.3
Lifting line from the 2D polar with 8 / 17 / 47.7 mm stack0.589 / 0.600 / 0.6140.0508 / 0.0513 / 0.052111.6 / 11.7 / 11.8
2D rib 1 closed, resolved1.0640.019055.9+0.015
2D rib 1 open intake (op2)1.0650.019754.1+0.014
2D closed, 8 / 17 / 47.7 mm stack0.982 / 1.014 / 1.0500.0224 / 0.0213 / 0.020043.9 / 47.6 / 52.5
Flow5 profile rib 1 (XFoil, trip 2 % / free)1.030 / 1.1460.0162 / 0.011063.6 / 104.2+0.029 / +0.009

All pitching moments are about one point: a quarter of the rib-1 chord, 0.728 m behind the nose of the root rib, nose-up positive. CFD and Flow5 both use the nose of the root rib as the origin. The surface integrals reproduce the CFD moment exactly and the Flow5 moment to within 2 %. The 3D centre of pressure of the CFD stays at about the quarter chord of the root rib at every angle (Cm 0.000 to 0.006), while Flow5 moves from -0.014 to -0.033. The sign of the Flow5 moment is not an error, since both the wing and the profile are negative about the leading edge and the profiles are positive only about the quarter chord. Flow5's sections carry almost no nose-up moment (inferred: potential-flow panel pressures).

The unexplained point (measured, cause not established). In 2D the thin stacks lose 7.6, 4.6 and 1.3 % of the lift (8, 17, 47.7 mm). A lifting line built from those 2D polars predicts that the 3D lift should rise by 4 % from the 8 mm to the 47.7 mm stack. The 3D CFD shows -0.8 %. The 47.7 mm stack is only 77 to 80 % built and is missing outboard, so it may not be a thick stack in the sense of the 2D test. The 3D wing may also respond differently. None of this was tested. By the 2D analogy the 3D CL could be 2 to 5 % low and the CD 5 to 12 % high with the 17 mm stack (inferred, not tested in 3D). The 3D numbers are therefore repeatable (three meshes within 0.9 %) but not confirmed as accurate. The lifting line from the resolved 2D polar (no twist) follows the CFD curves within a few per cent from 3 to 14 degrees (ratio to the CFD 1.09 at 3 degrees to 1.02), and the Flow5 lift slope is 20 % above all of them.

Figure 20. The 3D sweep against the 2D profile, Flow5 and the lifting-line estimate. Top row: lift curve, drag, polar and lift-to-drag ratio with everything on one scale. The upper group of curves (CL

Figure 20. The 3D sweep against the 2D profile, Flow5 and the lifting-line estimate. Top row: lift curve, drag, polar and lift-to-drag ratio with everything on one scale. The upper group of curves (CL up to 1.4) is 2D per unit span: teal closed foil, dark brown open intake, pink XFoil profile (Flow5's polar) with trip at 2 % and free transition, plus the three 2D stack points at 9 degrees (blue circle 8 mm, orange square 17 mm, red triangle 47.7 mm). The lower group is the 3D wing: blue and orange and red markers for the three stacks (they lie on top of each other), black dashed for Flow5 3D, purple dotted for the lifting-line estimate. Bottom row: the same 3D curves enlarged (lift, drag, polar, lift to drag) and, at the right, the pitching moment for all curves about the common point. The two tables below the plots give the 9-degree values. The 3D stack curves coincide, while the three 2D stack points are spread apart.

8. The 3D pressures in the Playground, compared with the earlier report

The Playground is Studio's model of the whole wing under load (dynamics, loading, trim). It takes the pressure on every cell from a file. This section writes such a file from the 3D OpenFOAM billow sweep (8.1), shows the Flow5 flights of the earlier report (Lep Studio test on gnuA7 (24 m2), 29 October 2026) as the reference (8.2), and flies the wing in the Playground with the OpenFOAM pressures at 8, 9 and 10 degrees by changing the calage (8.3). Sections 8.4 and 8.5 explain why the OpenFOAM wing flies faster than the Flow5 wing and give the forces of the flights and the position of the plumb point against the centre of pressure. Section 8.6 lists what still differs between the OpenFOAM wing, the Playground wing and the real wing.

8.1 The pressure file and the loads the Playground adds

The billow sweep (3 to 14 degrees, section 7.2) was converted into the Playground's pressure file. For every cell of the wing, for the upper and the lower skin and for 40 chordwise bins, it holds the area-weighted mean Cp of the CFD surface faces that fall in the bin (measured). The Playground reads the file at the chordwise position (x/c) of each of its own triangles, linearly between the bin centres, and applies the force -q Cp times the area vector of the triangle.

The OpenFOAM pressure file is a closed wing. The Playground adds the viscous shear drag of the CFD (CDv times q times S, 14 N at 9 degrees), the drag of the open intake, the drag of the lines and the drag of the pilot (0.5 times air density times 0.35 m2 times V squared). The intake drag is a separate term from the 2D study of rib 1 (section 5): the difference in drag coefficient between the open and the closed foil at the same angle, kOmegaSST, from 3 to 15 degrees (+0.0034 at 0 degrees, +0.00065 at 9, +0.00004 at 13, +0.0015 at 15). It is read for every cell at the angle of the section of that cell in the flown wing, multiplied by the dynamic pressure and the planform area of the cell, and put on the mouth faces of the cell along the wind. Cells without a mouth get nothing. In the flights the term is 0.9 to 2.0 N for the whole wing (1 to 3 % of the wing drag). The real trailing edge (section 10) is in neither the OpenFOAM file nor the Playground. An allowance of 1 to 2 % of the wing drag is expected to be missing. The Playground has an adjustable term for it, which is set to zero in all flights.

8.2 The Flow5 flights of the earlier report (the reference)

The earlier report flew two designs of the gnuA7 in the Playground with a pressure file written from Flow5. They are the reference for the comparison in section 8.3. B2 has the design values glide ratio 6.499, plumb point 34.24 % and calage 33.23 % (Section 8 of the design file). B3 is the same design with the calage lowered by 4.13 percentage points of the central chord (plumb point 30.11 %, calage 29.10 %). The values are the settled states of the Flow5 flights of that report (mean of the last 2 s of 15 s flights; lift and drag from its history files), once without and once with FSI (3 cycles). K is the calage the wing settles at. The angle is the angle of the wing; the angle to the central chord line is 0.48 degrees higher.

Settled flight stateK (%)Speed (m/s)Angle (deg)Lift (N)Drag (N)L/D
B2: design calage 33.23 % (glide ratio 6.5, plumb point 34.24 %)
Flow5 pressure file b2, no FSI (earlier report)37.159.119.87817.5136.16.01
Flow5, with FSI, 3 cycles (earlier report)37.1810.277.86820.7113.97.20
B3: design calage 29.10 % (the same design lowered by 4.13 points)
Flow5 pressure file b2, no FSI (earlier report)33.049.638.60817.7137.35.96
Flow5, with FSI, 3 cycles (earlier report)33.2110.896.79820.1118.46.93

8.3 Flying the gnuA7 at 8, 9 and 10 degrees by the design calage

With the OpenFOAM pressure file and the separate terms of section 8.1, the wing was flown at 8, 9 and 10 degrees by changing the calage of the design (Section 8 of the design file). The plumb point is moved with it and the lines are laid out again; the risers are not adjusted. Every flight starts from the design shape and runs 15 s of simulated time (900 frames, damping 10, the engine export of the design, closed mouth, no riser adjustment, cde 1.0 so that the intake is only the term of section 8.1). The values are the means of the last second. The wing is still creeping by 0.01 to 0.04 degrees per second at the end, so the settled angles are good to about 0.1 degree. The wing angle is the mean angle of all ribs. The central-chord angle is the angle of the chord of the central cell, which is the angle of the CFD, and the OpenFOAM values are read at the central-chord angle. K is the calage the wing settles at. The Cm = 0 point of the closed wing is the centre of pressure of the OpenFOAM force alone (pressure and friction, from the solver totals at the flown angle, measured about the nose of the central chord, positive nose-up: x = -Cm times the reference length divided by the force normal to the chord). The plumb point is where the vertical through the pilot meets the central chord (from the flown state).

FlightDesign calage (%)Wing angle (deg)Central-chord angle (deg)K (%)Speed (m/s)Lift / drag (N)L/DPlumb point (% chord)Cm = 0 point of the closed wing (% chord)
C13-r4138.007.9126.511.59817 / 142.25.7534.420.4
C20-r4209.119.0631.310.96817 / 136.16.0032.920.8
C24p8-r524.89.999.9834.910.54816 / 133.06.1432.121.1
C26-r42610.2810.2835.910.42816 / 132.56.1631.821.1

Figure 21. Settled angle against the design calage (left) and airspeed against angle (right). Left: the wing angle after 15 s for the flights of the table above; the dotted lines are 8, 9 and 10 degre

Figure 21. Settled angle against the design calage (left) and airspeed against angle (right). Left: the wing angle after 15 s for the flights of the table above; the dotted lines are 8, 9 and 10 degrees. Right: airspeed against wing angle; the green line is the Flow5 flights of the earlier report (B3 and B2, calage 29.10 and 33.23 %).

The final flights (measured). The design calage that gives 8, 9 and 10 degrees is 13, 20 and 24.8 % of the central chord (flights C13-r4, C20-r4 and C24p8-r5). The wing flies at 8.00, 9.11 and 9.99 degrees (central chord 7.91, 9.06 and 9.98), at 11.59, 10.96 and 10.54 m/s, with a glide ratio of 5.75, 6.00 and 6.14 (lift about 817 N, the weight of the wing and the pilot). The calage that gives a wing angle of 9 degrees is about 20 %, against about 30.4 % for the Flow5 flights of the earlier report (interpolated between B3 at 8.60 degrees with calage 29.10 % and B2 at 9.87 degrees with 33.23 %). That is about 10 percentage points lower. The speed at 9 degrees is 10.96 m/s against about 9.5 m/s for Flow5 (interpolated), which is 1.5 m/s higher, and the glide ratio is 6.00 against 5.96 to 6.01 (B3 and B2).

8.4 Why the OpenFOAM wing flies faster than the Flow5 wing

Why the wing flies faster than with Flow5, and where the centre of pressure is (inferred from the two pressure files). The lift coefficient at 9 degrees is 0.534 in the OpenFOAM pressure file (billow wing) and 0.743 in the Flow5 pressure file used in the earlier flights, which is 28 % less (the reference areas are 20.73 and 20.94 m2). For the same lift, the weight of 817 N, the dynamic pressure has to be about 1.4 times higher and the speed about 1.18 times: q S CL of the OpenFOAM file at 10.96 m/s and 9 degrees is 816 N. The flown speeds follow this: 9.5 m/s for the Flow5 flights (interpolated) and 10.96 m/s for the OpenFOAM wing at 9 degrees. The Cm = 0 point of the closed wing, the centre of pressure of the wing alone, is at 28.4 % of the chord at 9 degrees in the Flow5 pressure file and at 20.8 % in the OpenFOAM flight. That is 7.6 percentage points of the chord further forward (measured from the totals of the two files). With the centre of pressure further forward, the wing needs a lower calage to fly at the same angle. The glide ratio at 9 degrees (6.00) is close to that of the Flow5 flights (5.96 to 6.01).

8.5 Forces, plumb point and Cm = 0 point

FlightWing angle (deg)OpenFOAM lift at this angle (N)Playground wing lift (N)OpenFOAM pressure drag incl. induced (N)OpenFOAM friction (N)Playground wing drag (N)of which intake (N)Lines (N)Pilot (N)
C13-r48.0081782552.416.071.12.043.728.8
C20-r49.1182082456.414.272.31.339.325.8
C24p8-r59.9982282359.913.173.90.936.423.8

Forces (measured). In the final flights the lift of the wing in the Playground is 825, 824 and 823 N against 817, 820 and 822 N of the OpenFOAM file read at the central-chord angle (within 1 %; the Playground reads the file at the angle of each face). The wing drag is the pressure drag of the file (induced drag included) plus the friction, plus the intake term of 0.9 to 2.0 N. The Playground adds the drag of the lines (36 to 44 N) and of the pilot (24 to 29 N).

Plumb point and Cm = 0 point (measured). The plumb point lies 14.0, 12.2 and 11.0 % of the chord behind the Cm = 0 point of the closed wing at 8, 9 and 10 degrees (plumb 34.4, 32.9, 32.1 %; Cm = 0 point 20.4, 20.8, 21.1 %). The gap is of the order that the drag of lines and pilot gives (inferred: 65 N acting several metres below the wing against a lift of 817 N is a shift of about 12 % of the chord at 9 degrees). The Cm = 0 point of the closed wing moves by less than 1 % of the chord between 8 and 10 degrees, whereas the plumb point moves by 2.3 percentage points. The wing is trimmed by the calage, not by a movement of the centre of pressure.

8.6 What still differs between the OpenFOAM wing and the Playground wing

The following differences remain between the CFD and the Playground, or between the model and the real wing. They are listed as causes of differences in the results, and none is quantified unless a number is given.

9. Corrections made along the way

Earlier statementWhat was found
The OpenFOAM runs are fully turbulentThere is no transition model; the layer turns turbulent over x/c 0.1 to 0.4; inlet turbulence from 0.1 to 2 % changes nothing.
kOmegaSST trips at 4.5 to 5 % chordWithdrawn: it has no trip position; that figure came from a shape-factor criterion and from the lift matching a 5 % XFoil trip.
A stack that follows the boundary layer adds 4 to 7 % lift; the 165 mm mesh (CL 1.159) is the referenceArtefact of the gmsh fan at the sharp trailing edge; flat-base reference is 1.0634 and every earlier "gap closed" percentage is withdrawn.
The Belloc slope agreement is +7 %At equal angle the lift is 0.024 to 0.078 above the tunnel and the zero-lift angle differs by 0.7 degrees; the paper applies no wall corrections; the wing is not the gnuA7.
The tunnel slope 3.01 per radian is a reference for the gnuA7It belongs to the Belloc wing; it is a referee for the codes only.
A Flow5 pressure file (a1) had a defect at the outer nose166 triangles (the worst-shaped ones, minimum angle median 9.9 degrees) carry Cp from 0 to -0.4 beside neighbours at -1.3 to -2.0; the effect on the total lift is +0.10 % (0.4 m2 of 20.9 m2); file b2 has none; Flow5 was not rerun.
The 17 mm stack at 15 degrees is a converged pointIt is not: the plateau value is plotted and footnoted.

10. What is not solved or not done

11. Future work

Checks and developments intended next. None of them has been done, and the expected effects are inferences, not results.

12. Conclusion