applets.

Wind Tunnel

← All apps

About Wind Tunnel

A wind tunnel blows air past a model that stays still, so you can watch the air and weigh the forces on the model. This one runs in your browser. The pictures come from a real simulation of air, worked out afresh many times a second; the numbers about full-size cars, lorries and wings come from published figures and a standard formula. The two are kept apart on purpose, for the reasons under What the page simplifies.

THE TUNNEL: pick a model and press RUN. Smoke from the rake on the left shows where the air goes; STREAMLINES traces the air's path at this instant; PRESSURE colours the air by how hard it presses (red) or pulls (blue). The two dials are the tunnel's force balance: the drag (backwards) and lift (upwards) on the model in this simulation, as coefficients: numbers that hardly change with size or speed over a wide range, so different shapes can be compared. The data sheet sets the real things' drag coefficients side by side, from the round bar down to the wing; what drag costs at speed is on DRAG.

LIFT: a wing at the angle you set. PUFF SMOKE lets go of a straight line of smoke puffs in front of it, all at the same instant, and times the ones that pass just over and just under the wing to its back edge.

DRAG: a car's speed against its fuel; a lorry with and without a cab deflector (the fairing on the cab roof) and a tapered tail; two lorries driving close together (a platoon); two cyclists, one behind the other. DOWNFORCE: an upside-down wing over a moving road, and how fast a racing car must go before its downforce is more than its weight. DRAW YOUR OWN: draw any shape in the tunnel (click and dragdrag your finger) and test it. A closed outline is filled in.

THE WOWS

Speed costs more than you think. For a given shape, air drag grows with the square of the speed, near enough: twice as fast, four times the drag. The power needed is drag times speed, so it grows with the cube: twice as fast, eight times the power. Going from 70 to 80 mph adds about 31% to the drag. A car's fuel doesn't rise as fast, because rolling the tyres costs about the same per mile at any speed and the engine's own losses cost less per mile the faster you go, but the government's own guidance is that a car uses about 25% more fuel at 85 mph than at 70 mph, and about 10% more at 70 than at 60 (Department for Transport, 2011). The data sheet's fuel model is deliberately simple, set out under How it works: it matches the 85 against 70 figure, but gives about 14% for 70 against 60, more than the guidance's 10%, because real engines work more efficiently under a heavier load and the model leaves that out.

The air over a wing doesn't wait for the air underneath. A common explanation of lift says the air going over the curved top must get to the back at the same time as the air going under the flat bottom, so it has to go faster. It is wrong, and wind tunnels show it plainly: puffs that pass over the top of a lifting wing reach the back well before the ones underneath, and they never meet up again. NASA's Glenn Research Center calls this the "equal transit" theory and lists it as incorrect. What is true is that, on a wing making lift, the air over the top goes faster and its pressure is lower, and the wing turns the air downwards; both are ways of describing the same flow. Tilt the wing nose down far enough to push downwards and it all turns round: the race in LIFT shows that too.

Downforce greater than the car's weight. The wings and floor of a top single-seater racing car push it down onto the road, and at high speed that push is bigger than the car's weight. So, the story goes, it could drive upside down on the ceiling of a tunnel. It is quoted often and is true on paper, but no one has done it with a real racing car: it would have to get onto the ceiling at speed, and keep enough grip once there. Published estimates of the speed needed run from about 100 to 160 mph, depending on the car and its set-up.

THE STORY

The first wind tunnel was built in 1871 for the Aeronautical Society of Great Britain, on the proposal of Francis Wenham, who used it to measure the lift and drag of different shapes and found that long narrow wings lifted best. In late 1901 Orville and Wilbur Wright built a wooden tunnel about six feet long and 16 inches square, blown by a fan at roughly 25 mph, and over a couple of months tested more than 200 small wing shapes on balances made from hacksaw blades and wire; the tables they made shaped the 1902 glider and the 1903 Flyer. In 1909 Gustave Eiffel ran a tunnel at the foot of his tower in Paris; in 1912 he built a bigger laboratory at Auteuil, and its tunnel still works today. At Göttingen, Ludwig Prandtl's laboratory (founded in 1907) built tunnels that sent the air round a closed loop instead of out into the room, a layout still known as the Göttingen type. Prandtl's 1904 idea of the boundary layer, the thin film of slow air clinging to every surface, is the key to most of what you see here.

Étienne-Jules Marey photographed lines of smoke flowing past shapes around 1900. Car makers' tunnels now often have a rolling road, a belt under the car moving at the speed of the air, because a still floor would grow its own slow layer of air and spoil the flow under the car; the tunnels here use one for the vehicles.

HOW IT WORKS

The air is a lattice Boltzmann simulation (the D2Q9 model with the BGK collision rule): the tunnel is a grid of 320 by 128 cells, and each cell holds nine numbers, the amounts of air moving in each of nine directions (still, the four sides, the four corners). Every step, each amount moves one cell in its own direction, then the nine in each cell are nudged towards the mix that air settles into when left alone. That is all; the swirls, wakes and eddies come out of it by themselves, and they follow the same equations as real air at low speed (the Navier–Stokes equations). The air bounces off the model's cells (no-slip walls), comes in steadily on the left and leaves on the right. The force on the model is added up from the momentum of every bounce. The page's tests check the simulation against known results: the curved speed profile of flow through a channel, and the rhythm of eddies shed behind a round bar (a Strouhal number of about 0.18, as measured in real tunnels at this Reynolds number). It runs in a background worker, a dozen or so steps for each frame you see.

Smoke is thousands of tiny particles let go from the rake's forty nozzles and carried by the simulated air, joined up into lines that widen and fade as they travel, as real smoke spreads. The rake is pulsed, as in real smoke tunnels: every nozzle stops for a moment together, so each short gap starts as a straight upright row. Where the gaps pull ahead, that air is faster: over the wing, the gaps above it reach the back before those beneath. Streamlines are traced through the air's velocity at one instant. Pressure comes from the density of the simulated air.

The full-size numbers use F = ½ ρ v² Cd A: air density ρ = 1.225 kg/m³ (sea level, 15 °C), the speed v, a drag coefficient Cd and the frontal area A. Typical figures: a family car Cd 0.30 with 2.2 m²; an articulated lorry 0.65 with about 10 m² (the US Environmental Protection Agency says American tractor-trailers improved from about 0.8 to about 0.65 over two decades, largely through cab styling); a generic open-wheel racing car with drag area CdA 1.2 m² and downforce area ClA 4.5 m², 800 kg with its driver (published estimates for top single-seaters vary widely, up to about twice this downforce); a cyclist CdA 0.3 m². Power is drag times speed. The car's fuel: rolling resistance 0.8% of its 1,500 kg weight, the engine turning 30% of the fuel's energy into work, plus about 15 kW of fuel spent on its own losses every second whatever the speed (so less per mile at higher speeds), petrol at 34.2 MJ a litre. Real cars differ; the shape of the answer is the point. The wing's figures are typical of a whole full-size wing about seven times as long as it is wide, with a slightly curved section: no lift at about −2°, lift rising about 0.08 for each degree (less than the 0.1 of the section alone, because air spills round the wing tips), stalling at about 15° with a lift coefficient of about 1.3. The lamp post uses a drag coefficient of about 1, lower than the 1.2 of an endless round bar because air also passes over its top.

WHAT THE PAGE SIMPLIFIES

It is a 2D slice, a cut down the middle of the model, as if the model went on for ever sideways. Real air can go round the sides of a car; here it can only go over or under, so blunt shapes look worse and wakes last longer than in life. Wheels are drawn faintly behind the slice: on a car or lorry they sit at the sides, off the centre line, and a bicycle's spoked wheels let air through. A lorry's side skirts work on air swirling in from the sides, which a side slice can't show, so they are left out; the makers claim up to about 5% better fuel economy from them.

The Reynolds number is low: from about a hundred to a few thousand (the wing on THE TUNNEL gets the thinnest air, so its race shows plainly), where a real car at 70 mph is around ten million. In effect the simulated air is syrupy, so its slow layer on each surface is much thicker. That is why the dials read differently from the full-size figures, and why they are best used to compare shapes with each other. It shows most in DOWNFORCE: a real upside-down wing gains downforce as it gets nearer the road, down to about a tenth of its length from it (Zerihan and Zhang, 2000), and loses it when its slow layer tears away very close to the ground; in this syrupy air that loss starts much higher up, so the slice's downforce fades as you lower the wing. A wing in LIFT shows its stall as smoke tearing off the top, but the slice's lift number doesn't collapse the way a full-size wing's does. In the platoon and the cyclists, the rear one's drag falls by far more than the real figures on the data sheet, because a 2D wake is so long-lived.

The speeds are scaled. The simulated air moves 0.1 cells a step and the eddies' timing follows from that, not from the speed on the data sheet; puffs are timed in steps of the simulation. The opening pictures of THE TUNNEL and LIFT are stored, from the same simulation run for a few thousand steps, so the page doesn't have to work them out when it opens. Not a design tool.

Sources for the figures: the US National Renewable Energy Laboratory (platooning trials, 2014: lead lorry up to 5.3% less fuel, the one behind up to 9.7%); Blocken and colleagues (cyclists drafting: the rider behind 14 to 27 per cent less drag a centimetre behind, depending on riding position; the one in front about 1 to 3 per cent; worked out by computer simulation, checked against a wind-tunnel test); Department for Transport (fuel and speed, 2011); US EPA SmartWay (American lorry drag, 2016); NASA Glenn Research Center (lift).

FURTHER WATCHING AND READING

TEST SECTION · 2D SLICE · REYNOLDS NUMBER ABOUT 300
Model
View

TRY THIS