Visual walkthrough — Computational aerodynamics — panel method (intro), CFD overview

---

Step 1 — What "flow" even means as a picture

WHAT we do: replace "air is moving somehow" with "at each point there is an arrow $\vec{V}$".

WHY: a wing feels forces only because air arrows near it bend and speed up. If we know the arrows, we know everything — pressure, lift, the lot.

PICTURE: below, the uniform far-away wind is a field of identical arrows all pointing the same way at speed $V_{\infty }$ (read "$V$-infinity" = the wind speed far from the wing, where the body has no influence yet).

$\vec{V}=(u,v)$

• $u$ ::: the sideways (horizontal, $x$) part of the arrow
• $v$ ::: the up-down (vertical, $y$) part of the arrow
• $\vec{V}$ ::: the whole arrow, built from those two parts

---

Step 2 — The one rule the wing imposes

To talk about "straight into the surface" we need one more picture-tool.

Now, how do we extract "the part of $\vec{V}$ pointing along $\hat{n}$"? We use the dot product.

WHAT we do: write the solid-wall rule as one clean equation.

WHY: this single condition is the entire fingerprint of the shape. Everything downstream exists only to satisfy it.

PICTURE: the wind arrow $\vec{V}$ is split into a piece along the wall (allowed, blue) and a piece into the wall (forbidden, pink). We must kill the pink piece.

$\boxed{\vec{V}\cdot \hat{n}=0}\text{on the body}$

• $\hat{n}$ ::: outward perpendicular unit arrow at that point of the skin
• $\vec{V}\cdot \hat{n}$ ::: the amount of flow going into the wall — must vanish
• $=0$ ::: no tunnelling; this is the flow-tangency condition

But raw wind alone does not satisfy this — the pink piece is non-zero. We need to add "something" that cancels it. Enter the sources.

---

Step 3 — The spray nozzle (a source) and why it helps

WHAT we do: place these nozzles near the wall so their outward push can cancel the forbidden into-the-wall component of the wind.

WHY this tool and not another: a source's field, added to the wind's field, is still a legal flow (both obey the same governing equation — see Laplace's Equation & Potential Flow). And a source pushing outward is exactly the thing that can neutralise wind pushing inward. Vortices spin flow around; sinks pull in; a source is the natural "anti-tunnel" tool.

PICTURE: a lone source; its arrows radiate out, longer near the middle, shorter far away.

The velocity a source of strength $\lambda$ makes at distance $r$ is: $V_{\text{source}}=\frac{\lambda }{2\pi r}$

• $\lambda$ ::: strength — how hard it sprays (bigger = stronger push)
• $r$ ::: distance from the source to where we measure
• $2\pi$ ::: the full circle it spreads its flow over; dividing by it shares the flow evenly around

---

Step 4 — Chop the outline into panels

WHAT we do: replace the smooth curve by $N$ flat panels; unknowns become the $N$ numbers ${\lambda }_1,\dots ,{\lambda }_N$.

WHY: $N$ unknown numbers → a finite, solvable problem. More panels = better shape, but (foreshadowing the parent's mistake box) cost grows fast.

PICTURE: a smooth airfoil below, then the same airfoil as straight panels. On each panel: a red dot at the middle — the control point — where we will enforce the rule of Step 2. Each panel has its own outward ${\hat{n}}_i$.

• $N$ ::: how many panels we chose
• ${\lambda }_j$ ::: constant source strength on panel $j$ (one unknown per panel)
• control point ::: the panel's midpoint, where flow-tangency is tested
• ${\hat{n}}_i$ ::: the outward normal of panel $i$

---

Step 5 — Add everything up (superposition)

WHAT we do: the total flow at any point $P$ = the wind + the push from every panel's source.

WHY: we can build the exact flow we need by adding tuned ingredients, like mixing paint.

PICTURE: three stacked layers — wind arrows, source arrows, and their sum — showing how the combined arrows curve to hug the surface.

$\vec{V}(P)=\underset{\text{wind}}{\underbrace{{\vec{V}}_{\infty }}}+\underset{\text{all panels}}{\underbrace{\sum _{j=1}^N{\lambda }_j{\vec{v}}_j(P)}}$

• ${\vec{V}}_{\infty }$ ::: the uniform far-field wind (fixed, known)
• ${\sum }_{j=1}^N$ ::: "add up over every panel $j$ from $1$ to $N$"
• ${\lambda }_j$ ::: the unknown strength we are solving for
• ${\vec{v}}_j(P)$ ::: the flow that a unit-strength source on panel $j$ makes at point $P$ (pure geometry — computable once)

---

Step 6 — Turn the rule into equations (influence coefficients)

Now apply Step 2's rule — zero into-wall flow — at each control point $i$. Dot the whole sum with that panel's ${\hat{n}}_i$:

${\vec{V}}_{\infty }\cdot {\hat{n}}_i+{\sum }_{j=1}^N{\lambda }_j({\vec{v}}_j\cdot {\hat{n}}_i)=0$

WHAT we do: one flow-tangency equation per control point → $N$ equations.

WHY: $N$ equations for the $N$ unknowns ${\lambda }_j$ — now genuinely solvable.

PICTURE: an arrow from source $j$ reaching over to control point $i$, and only its component along ${\hat{n}}_i$ (the piece we keep) highlighted. That kept piece is $A_{ij}$.

Writing $b_i=-{\vec{V}}_{\infty }\cdot {\hat{n}}_i$ (the into-wall wind we must cancel), each equation is:

${\sum }_{j=1}^N{A}_{ij}{\lambda }_j=b_i$

• $A_{ij}$ ::: geometry-only influence of panel $j$ on panel $i$
• ${\lambda }_j$ ::: unknown strengths
• $b_i$ ::: the into-wall wind at panel $i$, with a minus sign (we must undo it)

---

Step 7 — Stack them into one matrix equation

WHAT we do: pack the $N$ equations into rows of a matrix.

WHY: "solve $A\mathbf{\lambda }=\mathbf{b}$" is the single most standard job in computing — the machine inverts $A$ and reads off every ${\lambda }_j$ at once.

PICTURE: the grid $A$, the tall $\mathbf{\lambda }$, and the tall $\mathbf{b}$, with the row for panel $i$ highlighted so you see it is exactly Step 6's equation.

Once solved, get the along-wall (tangential) speed $V_{t,i}$ at each panel and read pressure from Bernoulli's Equation: $C_{p,i}=1-{\left(\frac{V_{t,i}}{V_{\infty }}\right)}^2$

• $C_{p,i}$ ::: pressure coefficient (dimensionless pressure) on panel $i$
• $V_{t,i}/V_{\infty }$ ::: how much faster/slower air slides there than the free wind
• fast flow ($V_t>V_{\infty }$) ::: low pressure (lift); slow flow ::: high pressure

---

Step 8 — The degenerate case: sources alone give ZERO lift

WHAT we do: notice sources shape the body but never lift it — and see this is a special (degenerate) outcome, not a bug.

WHY: it forces the real-world fix — add vortex panels and impose the Kutta condition (flow must leave the sharp trailing edge smoothly), which sets a non-zero $\Gamma$, hence lift. This zero-drag-and-here-zero-lift ideal is d'Alembert's Paradox in action; real drag needs Boundary Layer Theory & Skin Friction Drag.

PICTURE: two flows over the same airfoil — top (source-only) is symmetric, streamlines mirror front to back, $\Gamma =0$; bottom (source + vortex + Kutta) is tilted, air leaves the trailing edge cleanly, $\Gamma \ne 0$, lift up.

• $\Gamma$ ::: circulation — net swirl of air around the body
• $L^{\prime }=\rho V_{\infty }\Gamma$ ::: lift per unit span; no swirl ⇒ no lift
• Kutta condition ::: pins $\Gamma$ so the flow leaves the trailing edge without wrapping around it

---

The one-picture summary

The full journey on one board: wind hits shape → coat outline with tunable sources → demand zero into-wall flow at each midpoint → this becomes $A\mathbf{\lambda }=\mathbf{b}$ → solve for strengths → read pressure → (add vortices for lift).

Recall Feynman retelling — say it in plain words

Imagine wind blowing at a solid shape. Air is not allowed to go through the metal, only along it. To force that, I line the outline with lots of tiny spray nozzles and let each spray at a strength I get to choose. If I aim the sprays right, their outward push exactly cancels the part of the wind that was trying to pierce the wall — so the total flow just skims the surface. "Aim them right" means: at the middle of every little segment, the sideways-into-wall flow must add up to zero. That is one equation per segment. Each nozzle's push on each midpoint is a fixed geometry number (an influence coefficient), so all the equations together are just rows of a grid: $A$ times the list of unknown strengths equals the list of winds I must cancel. The computer inverts the grid, hands me every strength, and from the along-wall speeds I read the pressure everywhere via Bernoulli. One catch: pure sprays are symmetric and cannot lift — for lift I also add little swirls (vortices) and require the air to peel off the sharp tail smoothly. Sprays make the shape; swirls make the lift.

Recall Quick self-test

• What single physical rule pins the source strengths? → Zero normal (into-wall) velocity at each control point.
• Which math tool extracts "into-wall flow" and why? → The dot product $\vec{V}\cdot \hat{n}$; it isolates the component along $\hat{n}$.
• Why can we add wind + sources freely? → Superposition: the governing equation is linear.
• What is $A_{ij}$? → Into-wall velocity a unit source on panel $j$ makes at panel $i$'s midpoint (pure geometry).
• Why zero lift with sources only, and the fix? → No circulation ($\Gamma =0$); add vortex panels + Kutta condition.