12.1 Airfoils, Circulation & Lift ended on a problem: the Joukowski map makes only its airfoils, with a cusp at the back and two knobs to turn. Real sections, like the NACA 2412 on a light trainer, have a finite trailing-edge angle and no neat map to a circle. So we stop mapping and replace the airfoil by a sheet of vortices lying on it, with strengths chosen so that no air passes through the skin. By hand, on a thin wing, that gives thin airfoil theory and a clean . On a computer, with a few hundred straight panels, it gives a linear system , solved by the Gaussian elimination you already know: the panel method, the core of XFOIL and most airfoil-design tools (from memory).
Tools: 11.6 Vorticity Dynamics & Circulation, 11.7 Potential Flow & Stream Functions, 1.3 Linear Systems, Kernel & Column Space (free directions) and 4.1 Local Error, Global Error & Order (convergence orders).
Transcript
- Last lesson ended on a problem: a real section, like the NACA 2412, has no neat map to a circle. So lay a sheet of vortices on the skin, with strengths chosen so that no air passes through. Across a sheet the tangential velocity jumps by gamma. With still air inside, the outside speed is gamma, so C p equals one minus gamma over U, squared.
- On a thin wing, put the sheet on the camber line and demand a streamline: an integral equation for gamma. Substitute x equals c over two times one minus cosine theta, and the camber slope becomes a cosine series with Fourier coefficients A n. Then C L is two pi times alpha minus alpha L zero. For the 2412: alpha L zero is minus 2.0772 degrees, A one 0.08150, A two 0.01386, and the quarter-chord moment minus 0.05312.
- For real thickness, cut the skin into N straight panels, with an unknown strength at each of the N plus one nodes. At each panel's midpoint demand no flow through it: N equations for N plus one unknowns, because a closed body is a streamline for every circulation. The last row is the Kutta condition: g zero plus g N equals zero.
- Stack the rows into a matrix. It is bright along the diagonal, neighbouring panels, and marked along the anti-diagonal, panel i facing panel N minus 1 minus i across the airfoil. The outlined last row is Kutta. It is nearly singular, since no fluid enters a closed body: condition number 1.30 times 10 to the 4 at 160 panels, about 4 of 16 digits lost. Harmless.
- Does it converge? NACA 0012 at 5 degrees: thin theory says 0.5483. The panels give 0.5982 with 20, then 0.6017, 0.6027, 0.6029, and 0.6030 with 300: second order. It settles at 1.0996 times 2 pi alpha, because thickness raises the slope.
- Back to the 2412. Lift is the area between the pressure curves, with a suction peak of minus 1.399 at 0.014 of the chord. The panels give C L of 0.2595, 0.7415 and 1.2199 at 0, 4 and 8 degrees, against thin theory's 0.2278, 0.6664 and 1.1051. The zero-lift angle is minus 2.149 degrees: the same line, steeper by 1.10.
- One honest warning. At the trailing edge the node strengths are unreliable. For the 0012 at zero degrees, g zero reads 0.42, 0.74, 0.91 and 1.02 at 40, 80, 160 and 320 panels: wrong sign, not converging. It is a plus-minus pair that cancels in the lift and the midpoint C p.
- Now the light aircraft. The 2412 reaches C L 0.5 at 1.993 degrees. But the wing is finite: trailing vortices push the air down. With aspect ratio 7.333 the induced angle is 1.2435 degrees, so it needs 3.237. The downwash is 1.0852 metres per second, and the induced drag 274.2 newtons, by C L squared over pi AR and by lift times downwash over speed.
- Now the honest reconciliation. 11.6's shortcut gave 0.3454 and 87.3 newtons: pi times too small. That pair carried the mean circulation, not the root: a factor 4 over pi. And a sink rate belongs to the far wake, rolled up into cores 8.639 metres apart, while the wing sits in the flat sheet: a factor pi squared over 4. The far wake keeps the momentum: 12 633 newtons.
- Lift, and drag, from a section. But once the aircraft makes lift, how does it tumble in three dimensions? Next, 12.3: rotation matrices, quaternions, and the inertia tensor, whose principal axes are eigenvectors once again.
The vortex sheet
A point vortex of circulation (clockwise positive, as in 12.1) swirls the air around it at speed : the 2D Biot–Savart law of 11.6. Spread vortices along a curve with strength per unit length, so a short piece is a point vortex : a vortex sheet.
Right at the sheet, draw a thin rectangle straddling it, length . The clockwise circulation round it equals the vorticity inside, (Stokes); the short sides vanish as the rectangle flattens, so
The tangential velocity jumps by ; the normal velocity is continuous. Put a sheet on the skin and arrange for the air inside to sit still: then the outside speed is , Bernoulli gives , and the lift is with (12.1). Everything hangs on finding .
Thin airfoil theory
Take a thin airfoil with camber line , , small slope and small . Put the sheet on the chord line (the camber is small) and demand that the camber line be a streamline.
The boundary condition. The free stream’s component across the camber line, upward, is about : the stream climbs at , the line at . The sheet must cancel it. A clockwise vortex at induces at a point behind it a downward velocity . Adding them all:
One equation for every : an integral equation for . The panel method will turn it into one row per panel.
The substitution. Write and , so runs from (leading edge) to (trailing edge), and . The integral that keeps appearing is Glauert’s integral (a standard table entry, from memory):
The flat plate. Try . The cancels against , and the left side becomes . Now look at : the sheet induces no normal velocity anywhere on the plate, so any amount of it can be added. That is 12.1’s circulation freedom again, a free direction like the kernel of a matrix. The Kutta condition removes it: our is at the trailing edge, , while blows up there.
Camber: a Fourier series. For a cambered line try
With and Glauert, each term gives exactly , so the equation becomes
The camber slope, written in , is a cosine series, and the are its Fourier coefficients. Cosines are orthogonal on , so each is one integral (Module 8 treats Fourier series properly):
Lift. , so with :
The slope is for every thin shape; camber only shifts the line by the zero-lift angle. The moment follows from pressure times lever arm; the standard result (from memory) is , and adding to move to the quarter chord,
independent of : the quarter chord is the aerodynamic centre 12.1 promised.
Worked example: the circular arc and the NACA 2412
Circular arc of camber : , so . Then , the value 12.1 quoted. Only survives, so . For : rad and .
NACA 2412: camber at chord, thick. Its camber line is two parabolas, ahead of and behind (). The lab cell below does the integrals by the midpoint rule in : (the textbook , from memory), , , , nose-down. Thin theory then puts at .
From sheet to panels
To keep thickness, put the sheet on the actual skin and discretise it:
- Panels. straight panels between nodes, numbered clockwise from the trailing edge along the lower surface and back along the upper (first and last nodes coincide), bunched toward both edges by .
- Unknowns. The sheet strength at each node, linear along each panel.
- Rows. Zero velocity across each panel at its midpoint, the control point.
What one row says. The free stream’s flow across panel plus the flow every panel’s sheet induces across it adds to zero: no air goes through panel . In symbols , with the outward normal. Each influence coefficient is the velocity across panel caused by a unit strength at node alone.
The influence coefficients. In panel ‘s own frame the panel runs from to , the control point sits at , and two geometric quantities appear: the angle the panel subtends at the point, , and the log of the ratio of the distances to its ends, . Integrating the point-vortex velocity along the panel, a unit strength gives velocity (along, across), and a strength rising from to gives . The panel’s second node gets the rising part, the first node the remainder. Rotate back and take the component along (the exercise’s influence function).
One row short. rows for unknowns, and not by accident: a closed body is a streamline for every circulation (12.1), so no-flow rows cannot decide the circulation. As in thin theory, the solutions form a line. The Kutta condition supplies the last row, : the sheet strength vanishes at the trailing edge, so both surfaces leave it at the same speed. Now is .
Conditioning. The matrix is still nearly singular, in two ways the sim measures. Weight the no-flow rows by panel length and add them: that is the total flow into the body, zero because a closed body cannot swallow fluid. So the rows are almost dependent; for the 2412 at 160 panels the weighted sum is of a typical row. (Codes with one unknown per panel overwrite one nearly redundant row with Kutta instead; Anderson’s textbook does, from memory.) And the combination of unknowns the rows barely see is a pair at the two trailing-edge nodes, whose panels almost lie on top of each other; that pair satisfies Kutta too. The 1-norm condition number is at 160 panels and at 300, costing about 4 of double precision’s 16 digits: harmless. Dense elimination costs about multiply-adds, million at 300 panels.
Superposition. The system is linear in the free stream, so one elimination with right-hand sides for and gives every : , an exact sine curve with no stall.
The matrix image is bright along the diagonal (neighbouring panels), with a dark sign-flip seam on the anti-diagonal (panel faces panel across the thin airfoil), with the Kutta row outlined at the bottom.
What the solver finds
NACA 0012 at , against thin theory’s (the exercise reproduces this):
| panels | ||
|---|---|---|
| 20 | 0.5982 | 1.0909 |
| 40 | 0.6017 | 1.0974 |
| 80 | 0.6027 | 1.0992 |
| 160 | 0.6029 | 1.0996 |
| 300 | 0.6030 | 1.0997 |
The gap to the 300-panel value shrinks about fourfold per doubling (, , ): second-order convergence. At , to round-off, as symmetry demands.
The thickness bonus. The slope at is per radian, not : thickness speeds the air over the whole body, and the circulation needed for smooth departure grows with it (12.1’s Joukowski section: ). Real wings measure slightly below , the boundary layer decambering the section (12.1, from memory).
Peak suction. At the 0012’s lowest is at ( at 300 panels), matching about from the NACA reports (from memory). At the peak jumps onto the nose, at , and still moving ( at 300 panels).
NACA 2412, 160 panels, against thin theory (, ):
| panel | thin | panel | peak at | |
|---|---|---|---|---|
| 0.2595 | 0.2278 | at | ||
| 0.7415 | 0.6664 | at | ||
| 1.2199 | 1.1051 | at |
The panel zero-lift angle is , and the line is steeper by the same . The moment starts near thin theory but drifts, : this thick inviscid section’s aerodynamic centre sits about behind the quarter chord (the 0012 agrees: at ). Exactly is a thin-wing result. The 4412 doubles thin theory’s and (, ).
Tick loading vs thin theory to compare with thin theory’s : similar shapes, thin theory lower over the front three quarters (2412, , : against ), slightly higher near the trailing edge, and infinite at the leading edge.
Closing the light aircraft
The trainer of 11.6 and 12.1 (chord m, span m, m, m/s) lifts N with N: . Press alpha for C_L = 0.5: a NACA 2412 section needs (thin theory , the flat plate ).
What a section does not tell you. A 2D section is an infinite wing. A real wing’s bound vortex turns at the tips into 11.6’s trailing vortices, which push the air at the wing down: a downwash . The section meets air at , its lift tilts back by the induced angle , and the backward part is induced drag, .
Lifting line (a glimpse). Prandtl modelled the wing as a line of horseshoe vortices with circulation along the span; each change sheds a trailing vortex. Those start at the wing, so there they are half-infinite lines inducing , half of 11.6’s . For the elliptic loading the downwash is uniform, and the standard results (from memory) are
with aspect ratio and for the elliptic wing (real planforms: a little below 1). The third follows from the first: the section makes ; insert and solve.
The numbers (sim readout):
- Root circulation, from : m/s.
- rad , so m/s. Summing Biot–Savart over 400 half-infinite trailing vortices, the sim finds m/s at every station.
- The wing needs ; its slope drops from the section’s to per radian.
- , so N; and N. The routes agree.
And 11.6’s sink rate? The shortcut m/s gives N, times too small. 11.6’s pair carried the mean circulation , not the root : a factor . And a sink rate belongs to the far wake, rolled up into cores m apart, m/s, while the wing sits in the flat sheet’s m/s: a factor . What the far wake keeps is the momentum (Problem 3).
Computing it
The coefficients fall fast (): few cosines needed.
Graded: zero lift at , the Kutta row, no flow through any control point (to ), and at , 160 panels.
Concept checks
Problems
For the flat plate, . (a) Find and . (b) Show that adding changes but not the boundary condition, and say what Kutta does to . (c) Where is infinite, and what does that mean?
Show worked solution
(a) , so .
(b) times leaves , whose normal velocity is proportional to (Glauert, ). It adds , but is infinite at unless ; Kutta sets .
(c) At the leading edge, : the air whips round the sharp edge (12.1’s leading-edge suction). A round nose spreads it into the finite peak the panels find.
The elliptic wing’s sheet rolls up into two vortices of strength , apart. (a) Show that for the light aircraft. (b) 11.6 used m/s with the same and found of the lift. Why? (c) Why does the pair’s sink rate not give the induced drag?
Show worked solution
(a) A vortex pair of strength , spacing , carries impulse per metre of wake; laying down metres per second gives downward momentum flux , the lift. With it is exact: N.
(b) is the mean circulation ; the elliptic root value is .
(c) Induced drag is set by the downwash at the wing, m/s; the rolled-up pair far behind sinks at m/s. Rolling up keeps the impulse, not the speed.
Key Takeaways
- A vortex sheet makes the tangential velocity jump by ; with still air inside, the surface speed is .
- Thin airfoil theory: one tangency equation per chord point; with the camber slope becomes a Fourier cosine series, giving and . NACA 2412: and ; circular arc: and .
- Panel method: one no-flow row per panel plus the Kutta row, for node strengths. No-flow rows cannot fix circulation; Kutta does. The rows are nearly dependent (no fluid enters the body): condition number about , harmless.
- NACA 0012 at : () at 160 panels, second-order convergence; thickness adds to the slope.
- The light aircraft on a NACA 2412: in 2D, as an elliptic wing (): m/s downwash, N induced drag by and alike.
- Limits: inviscid, no stall, edge artefacts, 2D plus a lifting-line glimpse.
- Next, 12.3: once the aircraft can make lift, how does it rotate and tumble in three dimensions? That takes rotation matrices, quaternions and the inertia tensor, whose principal axes are eigenvectors once again.