Flow Lab

Syllabus / Module 11: Fluid Dynamics: Navier–Stokes and Beyond

Lesson 11.5

Non-dimensionalization & the Reynolds Number

Scaling, dimensionless groups, Reynolds, Mach, Froude and Strouhal numbers, dynamic similarity and wind-tunnel testing.

Written byClaude Opus

🌊
Why this matters for simulation: Matching Reynolds number is the central problem of wind-tunnel testing.

By the end of this lesson you can…

  • Rescale the incompressible Navier–Stokes equations by $U$, $L$, $L/U$ and $\rho U^2$ until only $1/\mathrm{Re}$ is left, and read $\mathrm{Re} = UL/\nu$ as inertia over viscosity.
  • Measure the two terms on an exact flow, the Taylor–Green cell, and show their ratio is $\mathrm{Re}/(2\sqrt2)$: $0.035$ at $\mathrm{Re} = 0.1$, $35\,355$ at $10^5$.
  • Derive the creeping-flow and high-Re limits, including why the viscous term survives in a layer of thickness $L\,\mathrm{Re}^{-1/2}$.
  • Compute Reynolds, Mach, Froude and Strouhal numbers for real aircraft: $5.1\times10^6$ for a light-aircraft wing, $3.2\times10^7$ at Mach $0.85$ for an airliner.
  • Find the dimensionless groups of a problem as the kernel of its dimension matrix, and check them by multiplying.
  • Design a scale-model test: the tunnel speed that matches Re, why it breaks Mach, and how pressure, cold or water rescue it.
On this page
·
  1. Units are a choice; physics is ratios
  2. Scaling Navier–Stokes
  3. Re as a ratio of forces, measured
  4. The two limits
  5. Creeping flow, Re ≪ 1
  6. High Re: viscosity hides in thin layers
  7. Mach, Froude, Strouhal, and real aircraft
  8. Buckingham π: the groups are a kernel
  9. Dynamic similarity in the wind tunnel
  10. Computing it
  11. Concept checks
  12. Problems
  13. Key Takeaways

Last lesson’s 11 mm film of air, sheared at 5050 m/s, had Reynolds number 34033403. It carried a viscosity, a speed and a gap, three numbers with units, and the claim was that only one combination of them matters. That claim is what makes wind tunnels possible. An aircraft maker does not build a new wing full size to learn how it stalls; it puts a model a few percent of full size in a tunnel and trusts that the air around the model is doing the same thing. This lesson says exactly when that trust is justified.

The tools are ones you already own: Navier–Stokes from 11.4 Viscosity & the Navier–Stokes Equations, Bernoulli and dynamic pressure from 11.3 Momentum: Pressure & the Euler Equations, and, for the cleverest step, the kernel of a matrix from 1.3 Linear Systems, Kernel & Column Space.

▶ Video Rescaling Navier–Stokes until one number is left, the term balance as Re sweeps seven decades, the dimension matrix and its kernel, and a scale model that needs a supersonic tunnelcaptions only
Transcript
  1. A one millimetre film of air, sheared at fifty metres per second, has Reynolds number three thousand four hundred twenty-three. In water, three point four two three metres per second gives the same number. The stresses differ: zero point eight nine five pascals against three point four two three. But divide each by rho U squared and both give two point nine two one four times ten to the minus four: exactly one over Re. In dimensionless clothes, the two films are one flow.
  2. Why? Measure lengths in L, speeds in U, time in L over U, pressure in rho U squared. The time derivative, the advection and the pressure all become U squared over L; the viscous term becomes nu U over L squared. Divide through by U squared over L and the only leftover is nu over U L: one over Re in front of the viscous term. Same shape, same boundary data, same Re: the same equation, the same flow. That is dynamic similarity.
  3. Is Re really a ratio of forces? Measure it on an exact solution, the Taylor-Green cell, a checkerboard of counter-rotating vortices. Pressure alone balances the inertia, and the viscous term just drains the vortices. Their root-mean-square ratio is Re over two root two: zero point zero three five four at Re zero point one, thirty-five point four at one hundred, thirty-five thousand three hundred fifty-five at ten to the fifth. The inertia line stays put; the viscous line drops a decade per decade.
  4. Real flows span enormous Re. A ten micron cloud droplet, falling at three millimetres per second, is two point one times ten to the minus three: creeping, reversible Stokes flow. A honeybee wing is one point zero three times ten to the third; a light-aircraft wing five point one three five million; an airliner thirty-two million. There, viscosity survives only in a layer of thickness L over root Re: zero point six six two millimetres on the light wing.
  5. Other physics, other ratios. Mach is speed over sound speed: density changes by about M squared over two, four point five percent at Mach point three. Froude, U over root g L, brings gravity in as one over Froude squared; for the light wing Froude is thirteen point oh three, so gravity is under one percent of inertia. Strouhal, f L over U, is about zero point two for a cylinder, so a five millimetre wire at thirty metres per second sings at twelve hundred hertz.
  6. What if you only know which variables matter? Drag on a sphere depends on F, rho, U, D and mu. Write each as a vector of mass, length and time exponents: a product is dimensionless when A times n is zero. The groups are the kernel of the dimension matrix; rank three, five variables, so two groups. The kernel is spanned by one, minus one, minus two, minus two, zero and zero, minus one, minus one, minus one, one. Multiplying checks both. They are F over rho U squared D squared, and one over Re: one curve.
  7. Now a wind tunnel. A one to ten model of the light wing needs ten times the speed to match Re: five hundred metres per second, Mach one point four seven, shock waves, a different flow. Every escape lowers the tunnel's viscosity: water needs only thirty-four point two three, air at ten atmospheres fifty. The one to twenty airliner must match Mach and Re together. In sea-level air, Mach point eight four seven means two hundred eighty-eight point three metres per second but only four point nine three five million Re, short by six point four eight. At one hundred ten kelvin, Mach needs just one hundred seventy-eight point one four, and one point seven oh nine atmospheres gives the full Re.
  8. So Re is the one dial. When it is large, one over Re says viscosity vanishes, yet no-slip keeps it alive in a thin layer. What does it spread there? Next lesson: take the curl of Navier-Stokes and pressure disappears, leaving one quantity carried by the flow and spread by viscosity: vorticity. Boundary layers are sheets of it; wakes are where it goes.

Units are a choice; physics is ratios

Metres and seconds are human conventions the air does not know, so a true law must survive a change of units. Only dimensionless combinations do.

Try it on the film. With this lesson’s air properties (the ISA model, formulas from memory, giving ν=1.4607×10−5\nu = 1.4607\times10^{-5} m2^2/s at sea level against 11.4’s rounded 1.4694×10−51.4694\times10^{-5}),

Re=Uhν=50⋅10−31.4607×10−5=3423.\mathrm{Re} = \frac{Uh}{\nu} = \frac{50\cdot10^{-3}}{1.4607\times10^{-5}} = 3423 .

The third digit moved with the property table, not the physics. Now replace the air by water, ν=1.0×10−6\nu = 1.0\times10^{-6}. The same Re needs U=3423⋅10−6/10−3=3.423U = 3423\cdot10^{-6}/10^{-3} = 3.423 m/s. The stresses differ: 0.8950.895 Pa in air, 3.4233.423 Pa in water. Divide each by ρU2\rho U^2, though, and both give 2.9214×10−42.9214\times10^{-4}, which is exactly 1/Re1/\mathrm{Re}. In dimensionless clothes the two films are one flow.

Scaling Navier–Stokes

Pick a length LL (a chord, a gap), a speed UU (the free stream), and define starred variables, each of order one in the region that matters:

x=L x∗,u=U u∗,t=LU t∗,p=ρU2 p∗.\mathbf x = L\,\mathbf x^*, \qquad \mathbf u = U\,\mathbf u^*, \qquad t = \frac{L}{U}\,t^*, \qquad p = \rho U^2\,p^* .

L/UL/U is the time a parcel takes to pass the body; ρU2\rho U^2 is twice 11.3’s dynamic pressure, the pressure change inertia can produce. Now convert each term of

∂u∂t+(u⋅∇)u=−1ρ∇p+ν∇2u.\frac{\partial\mathbf u}{\partial t} + (\mathbf u\cdot\nabla)\mathbf u = -\frac1\rho\nabla p + \nu\nabla^2\mathbf u .

A derivative in xx is ∂/∂x=(1/L) ∂/∂x∗\partial/\partial x = (1/L)\,\partial/\partial x^*, so ∇=∇∗/L\nabla = \nabla^*/L and ∇2=∇∗2/L2\nabla^2 = \nabla^{*2}/L^2. Term by term:

  • ∂u/∂t=UL/U ∂u∗∂t∗=U2L ∂u∗∂t∗\partial\mathbf u/\partial t = \dfrac{U}{L/U}\,\dfrac{\partial\mathbf u^*}{\partial t^*} = \dfrac{U^2}{L}\,\dfrac{\partial\mathbf u^*}{\partial t^*}
  • (u⋅∇)u=U⋅1L⋅U (u∗⋅∇∗)u∗=U2L(u∗⋅∇∗)u∗(\mathbf u\cdot\nabla)\mathbf u = U\cdot\dfrac1L\cdot U\,(\mathbf u^*\cdot\nabla^*)\mathbf u^* = \dfrac{U^2}{L}(\mathbf u^*\cdot\nabla^*)\mathbf u^*
  • 1ρ∇p=1ρ⋅ρU2L∇∗p∗=U2L∇∗p∗\dfrac1\rho\nabla p = \dfrac1\rho\cdot\dfrac{\rho U^2}{L}\nabla^* p^* = \dfrac{U^2}{L}\nabla^* p^*
  • ν∇2u=νUL2∇∗2u∗\nu\nabla^2\mathbf u = \dfrac{\nu U}{L^2}\nabla^{*2}\mathbf u^*

Divide everything by U2/LU^2/L. Three terms come out clean; the last keeps the factor νU/L2÷U2/L=ν/(UL)\nu U/L^2 \div U^2/L = \nu/(UL):

∂u∗∂t∗+(u∗⋅∇∗)u∗=−∇∗p∗+1Re ∇∗2u∗,∇∗⋅u∗=0,Re=ULν.\frac{\partial\mathbf u^*}{\partial t^*} + (\mathbf u^*\cdot\nabla^*)\mathbf u^* = -\nabla^* p^* + \frac{1}{\mathrm{Re}}\,\nabla^{*2}\mathbf u^*, \qquad \nabla^*\cdot\mathbf u^* = 0, \qquad \mathrm{Re} = \frac{UL}{\nu} .

Continuity, (U/L)∇∗⋅u∗=0(U/L)\nabla^*\cdot\mathbf u^* = 0, loses its constant altogether.

Re as a ratio of forces, measured

The estimate behind the name: inertia per unit mass is about U2/LU^2/L, viscous force about νU/L2\nu U/L^2, and their ratio is UL/νUL/\nu. “About” hides a shape-dependent factor, so let us measure it on a real flow.

The Taylor–Green cell is an exact Navier–Stokes solution, a checkerboard of counter-rotating vortices (verifying it is the first problem):

u∗=(sin⁡x∗cos⁡y∗,  −cos⁡x∗sin⁡y∗) e−2t∗/Re,p∗=14(cos⁡2x∗+cos⁡2y∗) e−4t∗/Re.\mathbf u^* = \big(\sin x^*\cos y^*,\; -\cos x^*\sin y^*\big)\,e^{-2t^*/\mathrm{Re}}, \qquad p^* = \tfrac14\big(\cos 2x^* + \cos 2y^*\big)\,e^{-4t^*/\mathrm{Re}} .

At t∗=0t^* = 0, direct differentiation gives two clean facts. The inertia is (u∗⋅∇∗)u∗=12(sin⁡2x∗,sin⁡2y∗)(\mathbf u^*\cdot\nabla^*)\mathbf u^* = \tfrac12(\sin 2x^*, \sin 2y^*), which is exactly −∇∗p∗-\nabla^* p^*: here pressure alone holds the inertia, as in Euler’s equations. And ∇∗2u∗=−2u∗\nabla^{*2}\mathbf u^* = -2\mathbf u^*, so the viscous term just drains the vortices, ∂u∗/∂t∗=−2u∗/Re\partial\mathbf u^*/\partial t^* = -2\mathbf u^*/\mathrm{Re}.

Average the squares over a cell: sin⁡2\sin^2 averages to 12\tfrac12, so the inertia has RMS 12\tfrac12, and the viscous term (2/Re)⋅(1/2)=2/Re(2/\mathrm{Re})\cdot(1/\sqrt2) = \sqrt2/\mathrm{Re}. The ratio is

RMS inertiaRMS viscous=Re22=0.0354, 35.4, 35 355atRe=0.1, 100, 105.\frac{\text{RMS inertia}}{\text{RMS viscous}} = \frac{\mathrm{Re}}{2\sqrt2} = 0.0354,\ 35.4,\ 35\,355 \quad\text{at}\quad \mathrm{Re} = 0.1,\ 100,\ 10^5 .

Re is the ratio, up to a factor belonging to this shape. Central differences on a 64×6464\times64 grid measure 0.49920.4992 and 1.41311.4131 instead of 0.50.5 and 1.41421.4142, ratio 0.03530.0353 at Re=0.1\mathrm{Re} = 0.1: second-order errors of a tenth of a percent.

◆ Interactive Term balance in the Taylor–Green cell: where inertia or viscosity wins, and both RMS terms against Re

Loading simulation…

Slide Re: the arrows never change; only the viscous line drops, a decade per decade, crossing the inertia line at Re=22\mathrm{Re} = 2\sqrt2. The amplitude falls by ee at t∗=Re/2t^* = \mathrm{Re}/2. At Re=0.1\mathrm{Re} = 0.1 that is t∗=0.05t^* = 0.05: the vortices die before a parcel has moved a twentieth of LL. At 10510^5 the cell survives 50 00050\,000 passage times.

The two limits

Creeping flow, Re ≪ 1

Small Re makes the viscous term huge, so the pressure scale ρU2\rho U^2 is now wrong: pressure must balance viscosity, not inertia. Multiply the starred equation by Re and define p~=Re p∗=pL/(μU)\tilde p = \mathrm{Re}\,p^* = pL/(\mu U):

Re(∂u∗∂t∗+(u∗⋅∇∗)u∗)=−∇∗p~+∇∗2u∗  → Re→0   ∇∗p~=∇∗2u∗.\mathrm{Re}\Big(\frac{\partial\mathbf u^*}{\partial t^*} + (\mathbf u^*\cdot\nabla^*)\mathbf u^*\Big) = -\nabla^*\tilde p + \nabla^{*2}\mathbf u^* \;\xrightarrow{\ \mathrm{Re}\to0\ }\; \nabla^*\tilde p = \nabla^{*2}\mathbf u^* .

These are the Stokes equations: linear, with no time derivative, so the flow at each instant is fixed by the boundary velocities at that instant. Flip the sign of every boundary velocity and (−u,−p~)(-\mathbf u, -\tilde p) solves the new problem, so a parcel retraces its path exactly when the walls retrace theirs: creeping flow is reversible (G. I. Taylor’s glycerine demonstration, from memory).

The classic result (from memory, derived in many texts) is Stokes drag on a sphere, F=3πμUDF = 3\pi\mu UD. A 1010 micron cloud droplet falls when that drag equals its weight, π6D3ρwg=3πμUD\tfrac\pi6 D^3\rho_w g = 3\pi\mu UD, so U=ρwgD2/(18μ)=3.046U = \rho_w gD^2/(18\mu) = 3.046 mm/s, and its Re is 2.1×10−32.1\times10^{-3}.

High Re: viscosity hides in thin layers

At Re=5×106\mathrm{Re} = 5\times10^6 the factor 1/Re1/\mathrm{Re} looks like permission to drop viscosity, leaving Euler. But Euler cannot meet no-slip, and 11.4 showed that dropping the highest derivative loses a boundary condition. The way out: near the wall the velocity changes over a thickness δ\delta much smaller than LL, so there ∇∗2u∗\nabla^{*2}\mathbf u^* is not of order one but of order (L/δ)2(L/\delta)^2. The viscous term becomes (1/Re)(L/δ)2(1/\mathrm{Re})(L/\delta)^2, and it rejoins the balance when that is of order one:

δL∼Re−1/2.\frac{\delta}{L} \sim \mathrm{Re}^{-1/2} .

For a light-aircraft wing, chord 1.51.5 m at 5050 m/s, Re=5.135×106\mathrm{Re} = 5.135\times10^6 and L Re−1/2=0.662L\,\mathrm{Re}^{-1/2} = 0.662 mm. 11.4’s impulsive-plate estimate, 4νx/U4\sqrt{\nu x/U} at the trailing edge, gives 2.6482.648 mm: the same law, νL/U\sqrt{\nu L/U}, with a constant 44 in front. Small coefficient, huge derivative: that product is where skin friction, separation and stall live (11.8).

Mach, Froude, Strouhal, and real aircraft

Other physics brings other ratios, each a comparison of two terms.

Mach, M=U/aM = U/a. Bernoulli changes the pressure by about 12ρU2\tfrac12\rho U^2. A gas answers with a density change Δρ≈Δp/a2\Delta\rho \approx \Delta p/a^2 (a2=dp/dρa^2 = dp/d\rho, the sound speed squared; a preview of 11.12). So Δρ/ρ≈12ρU2/(ρa2)=M2/2\Delta\rho/\rho \approx \tfrac12\rho U^2/(\rho a^2) = M^2/2. At Mach 0.30.3 that is 4.5%4.5\%: the conventional edge of “incompressible”.

Froude, Fr=U/gL\mathrm{Fr} = U/\sqrt{gL}. Add g\mathbf g to Navier–Stokes, divide by U2/LU^2/L, and it arrives as (gL/U2) g^=g^/Fr2(gL/U^2)\,\hat{\mathbf g} = \hat{\mathbf g}/\mathrm{Fr}^2. It matters at a free surface: ship waves, a seaplane float. For the light wing Fr=13.03\mathrm{Fr} = 13.03: gravity’s term is under 1%1\% of inertia.

Strouhal, St=fL/U\mathrm{St} = fL/U. When the flow has its own frequency ff, the time scale is 1/f1/f, and the unsteady term carries St instead of 11. A cylinder sheds vortices alternately at St≈0.2\mathrm{St} \approx 0.2 (from memory, for Re from a few hundred to about 2×1052\times10^5; the wind-tunnel capstone will measure it). A 55 mm antenna wire at 3030 m/s, Re 1.03×1041.03\times10^4, sings at f=0.2⋅30/0.005=1200f = 0.2\cdot30/0.005 = 1200 Hz.

FlyerLL, UUReMach
Light-aircraft wing, sea level1.51.5 m, 5050 m/s5.13×1065.13\times10^60.1470.147
Airliner wing, 1111 km (ν=3.906×10−5\nu = 3.906\times10^{-5})55 m, 250250 m/s3.20×1073.20\times10^70.8470.847
Honeybee wing (rough, from memory)33 mm, 55 m/s1.03×1031.03\times10^30.0150.015
Golf ball (from memory)42.742.7 mm, 7070 m/s2.05×1052.05\times10^50.2060.206

Bee and airliner differ by four decades of Re: a bee wing is not a small airliner wing.

◆ Interactive Same Re, different fluid, size and speed: Re and Mach strips with named regimes, and the equivalent speed in other fluids

Loading simulation…

The Re bands are for a flat-plate boundary layer with Re built on distance from the leading edge; their edges (5×1055\times10^5, 3×1063\times10^6) are from memory and move with roughness and free-stream turbulence. With them, the light wing’s layer should start transition about 0.1460.146 m behind the leading edge.

Buckingham π: the groups are a kernel

What if you know only the list of variables, not the equations? The drag FF on a sphere can depend on ρ\rho, UU, DD and μ\mu, and nothing else (no gravity, no compressibility). Write a candidate group FaρbUcDdμeF^{a}\rho^{b}U^{c}D^{d}\mu^{e} and demand it be dimensionless.

Each variable has a dimension vector in (mass, length, time): FF is kg m s−2^{-2}, so (1,1,−2)(1, 1, -2); ρ\rho is (1,−3,0)(1, -3, 0); UU is (0,1,−1)(0, 1, -1); DD is (0,1,0)(0, 1, 0); μ\mu, Pa s == kg m−1^{-1} s−1^{-1}, is (1,−1,−1)(1, -1, -1). Exponents add when you multiply, so the group is dimensionless exactly when

An=0,A=(110011−311−1−20−10−1),n=(a,b,c,d,e).A\mathbf n = \mathbf 0, \qquad A = \begin{pmatrix} 1 & 1 & 0 & 0 & 1 \\ 1 & -3 & 1 & 1 & -1 \\ -2 & 0 & -1 & 0 & -1 \end{pmatrix}, \qquad \mathbf n = (a, b, c, d, e) .

Columns are F,ρ,U,D,μF, \rho, U, D, \mu; rows are M, L, T. The dimensionless groups are the kernel of the dimension matrix. AA has rank 33, so the kernel has dimension 5−3=25 - 3 = 2: two independent groups. That count is the Buckingham π theorem.

Solve with aa and ee free. Row M: b=−a−eb = -a - e. Row T: c=−2a−ec = -2a - e. Row L: d=−a+3b−c+e=−a−3a−3e+2a+e+e=−2a−ed = -a + 3b - c + e = -a - 3a - 3e + 2a + e + e = -2a - e. So

n=a (1,−1,−2,−2,0)+e (0,−1,−1,−1,1).\mathbf n = a\,(1, -1, -2, -2, 0) + e\,(0, -1, -1, -1, 1) .

Check by multiplying. A(1,−1,−2,−2,0)A(1, -1, -2, -2, 0): row M 1−1=01 - 1 = 0; row L 1+3−2−2=01 + 3 - 2 - 2 = 0; row T −2+2=0-2 + 2 = 0. And A(0,−1,−1,−1,1)A(0, -1, -1, -1, 1): M −1+1=0-1 + 1 = 0; L 3−1−1−1=03 - 1 - 1 - 1 = 0; T 1−1=01 - 1 = 0. Both squash to zero.

The first is π1=F/(ρU2D2)\pi_1 = F/(\rho U^2D^2), a force coefficient; the second is π2=μ/(ρUD)=1/Re\pi_2 = \mu/(\rho UD) = 1/\mathrm{Re}. Every physical law relating the five variables can be written π1=Φ(π2)\pi_1 = \Phi(\pi_2). With the conventional factors,

CD=F12ρU2⋅π4D2=f(Re).C_D = \frac{F}{\tfrac12\rho U^2\cdot\tfrac\pi4D^2} = f(\mathrm{Re}) .

Five variables became one curve. In creeping flow Stokes drag fills it in: CD=3πμUD⋅8/(πρU2D2)=24/ReC_D = 3\pi\mu UD\cdot8/(\pi\rho U^2D^2) = 24/\mathrm{Re}. At high Re the curve flattens to a value of order one, then drops sharply in the drag crisis, at a Re of a few 10510^5 for a smooth sphere (from memory). The golf ball sits at 2.05×1052.05\times10^5, just below it; its dimples exist to pull the crisis down to where it flies (11.8).

Dynamic similarity in the wind tunnel

Test the light wing at 1:101{:}10, chord 0.150.15 m, in sea-level air. Matching Re needs UmLm=ULU_m L_m = U L, so

Um=U⋅LLm⋅νmν=50⋅10⋅1=500 m/s=Mach 1.47.U_m = U\cdot\frac{L}{L_m}\cdot\frac{\nu_m}{\nu} = 50\cdot10\cdot1 = 500 \text{ m/s} = \text{Mach } 1.47 .

Shock waves: a different flow. This is the central problem of wind-tunnel testing, and every escape lowers νm\nu_m:

  • Pressurize. μ\mu barely depends on pressure (from memory), and ρ∝p\rho \propto p, so ν∝1/p\nu \propto 1/p. At 1010 atm, νm\nu_m falls tenfold and 5050 m/s matches.
  • Use water. Um=500⋅1.0×10−6/1.4607×10−5=34.23U_m = 500\cdot1.0\times10^{-6}/1.4607\times10^{-5} = 34.23 m/s, and Mach is irrelevant (sound in water is about 14801480 m/s, from memory). But no compressible effect of the aircraft can be reproduced at all.
  • Cool the gas. Cold gas is denser and less viscous (Sutherland’s law).

The airliner shows why cold wins. At 1:201{:}20 its chord is 0.250.25 m and it must match Mach 0.8470.847 as well as Re=3.2×107\mathrm{Re} = 3.2\times10^7. In sea-level air, matching Mach means 288.3288.3 m/s, which gives only 4.935×1064.935\times10^6, short by a factor 6.486.48. At 110110 K the sound speed drops to 210.25210.25 m/s, Mach 0.8470.847 needs only 178.14178.14 m/s, and the gas reaches the full-scale Re at 1.7091.709 atm. Cryogenic pressurized tunnels exist for precisely this (NASA’s National Transonic Facility and the European Transonic Windtunnel use cold nitrogen, from memory; the sim uses air properties).

◆ Interactive Scale model: tunnel speed needed to match Re against model scale, with the tunnel's Mach limits and the Mach-matched speed

Loading simulation…

Sometimes no fluid can match everything. A seaplane float at 1:101{:}10, 2020 m/s full scale, in a towing tank of water: Froude matching (waves) demands 20/10=6.3220/\sqrt{10} = 6.32 m/s, Reynolds matching 200200 m/s. Naval architects match Froude and correct for the missing factor 0.03160.0316 in Re with friction formulas (from memory), accepting a known error over an impossible test.

Computing it

🐍 PythonLab: air properties (A), the film in two fluids (B), Taylor–Green term balance (C), creeping and layer scales (D), tunnels (E)
⌘/Ctrl + Enter
🧪 Coding exerciseExercise: dimensionless groups as a kernel, and the tunnel speed that matches Re
⌘/Ctrl + Enter

Exponents are graded to 10−910^{-9}, the water speed to 0.010.01 m/s; a pendulum matrix (period, length, gg, mass) catches hard-coded answers.

Concept checks

✅ Concept check

A light-aircraft wing has chord 1.5 m and flies at 50 m/s in sea-level air ($\nu = 1.4607\times10^{-5}$ m$^2$/s). What is its Reynolds number, in millions?

✅ Concept check

In creeping flow, $\mathrm{Re} \ll 1$, which statement is right?

✅ Concept check

A 1:10 model of a wing that flies at 50 m/s in sea-level air ($\nu = 1.4607\times10^{-5}$) is tested in water ($\nu = 1.0\times10^{-6}$). What speed, in m/s, matches Re?

✅ Concept check

A problem has 6 variables built from mass, length and time, and its dimension matrix has rank 3. How many independent dimensionless groups does it have?

Problems

✏️ Pen & paperTaylor–Green is exact●●○

Check that the Taylor–Green cell satisfies the starred Navier–Stokes equations and continuity.

Show worked solution

Write E=e−2t∗/ReE = e^{-2t^*/\mathrm{Re}}, u=sin⁡xcos⁡y Eu = \sin x\cos y\,E, v=−cos⁡xsin⁡y Ev = -\cos x\sin y\,E (stars dropped). Continuity: ux+vy=(cos⁡xcos⁡y−cos⁡xcos⁡y)E=0u_x + v_y = (\cos x\cos y - \cos x\cos y)E = 0.

Inertia, xx-component: uux+vuy=E2(sin⁡xcos⁡y⋅cos⁡xcos⁡y+cos⁡xsin⁡y⋅sin⁡xsin⁡y)=E2sin⁡xcos⁡x=12E2sin⁡2xuu_x + vu_y = E^2(\sin x\cos y\cdot\cos x\cos y + \cos x\sin y\cdot\sin x\sin y) = E^2\sin x\cos x = \tfrac12E^2\sin 2x. Pressure, p=14(cos⁡2x+cos⁡2y)E2p = \tfrac14(\cos 2x + \cos 2y)E^2: −px=12sin⁡2x E2-p_x = \tfrac12\sin 2x\,E^2, equal to the inertia, as the balance (u⋅∇)u=−∇p(\mathbf u\cdot\nabla)\mathbf u = -\nabla p requires. Viscous: ∇2u=−2u\nabla^2 u = -2u, and ut=−(2/Re)u=(1/Re)∇2uu_t = -(2/\mathrm{Re})u = (1/\mathrm{Re})\nabla^2u. The yy-component works the same way. Every term balances, at every Re (the lab’s block C measures the sizes).

✏️ Pen & paperStokes drag from dimensions alone●●●

In creeping flow inertia is gone, so density cannot appear. (a) Write the dimension matrix for F,U,D,μF, U, D, \mu, find its kernel and show F=c μUDF = c\,\mu UD. (b) With c=3πc = 3\pi, show CD=24/ReC_D = 24/\mathrm{Re}.

Show worked solution

(a) Columns F(1,1,−2)F(1,1,-2), U(0,1,−1)U(0,1,-1), D(0,1,0)D(0,1,0), μ(1,−1,−1)\mu(1,-1,-1). Rank 33, four columns: one group. For FaUcDdμeF^aU^cD^d\mu^e: M gives a+e=0a + e = 0; T gives −2a−c−e=0-2a - c - e = 0; L gives a+c+d−e=0a + c + d - e = 0. With a=1a = 1: e=−1e = -1, c=−1c = -1, d=−1d = -1. Check: M 1−1=01 - 1 = 0, L 1−1−1+1=01 - 1 - 1 + 1 = 0, T −2+1+1=0-2 + 1 + 1 = 0. So F/(μUD)F/(\mu UD) is a constant. Note what happened: losing ρ\rho removed a column, the kernel shrank to one vector, and the unknown function Φ\Phi became a constant.

(b) CD=3πμUD/(12ρU2π4D2)=24μ/(ρUD)=24/ReC_D = 3\pi\mu UD / (\tfrac12\rho U^2\tfrac\pi4D^2) = 24\mu/(\rho UD) = 24/\mathrm{Re}.

✏️ Pen & paperThe airliner in a cold tunnel●●○

A 1:201{:}20 model of a 55 m chord flying at Mach 0.8470.847 and Re=3.2×107\mathrm{Re} = 3.2\times10^7 is tested in air at 110110 K, where a=210.25a = 210.25 m/s and, at 11 atm, ν=2.378×10−6\nu = 2.378\times10^{-6} m2^2/s. Find the speed that matches Mach and the pressure that then matches Re.

Show worked solution

Um=0.847⋅210.25=178.1U_m = 0.847\cdot210.25 = 178.1 m/s. Required νm=UmLm/Re=178.1⋅0.25/3.2×107=1.39×10−6\nu_m = U_mL_m/\mathrm{Re} = 178.1\cdot0.25/3.2\times10^7 = 1.39\times10^{-6}. Since ν∝1/p\nu \propto 1/p at fixed TT, p=2.378/1.39=1.71p = 2.378/1.39 = 1.71 atm. The model panel, airliner at 1:201{:}20, T=110T = 110 K, prints 1.7091.709 atm, and at that pressure the Re line crosses the Mach line at k=20k = 20.

Key Takeaways

  • Physical laws relate dimensionless numbers. The 11 mm film is Re=3423\mathrm{Re} = 3423 in air at 5050 m/s and in water at 3.4233.423 m/s, with the same τ/(ρU2)=1/Re\tau/(\rho U^2) = 1/\mathrm{Re}.
  • Scaling by UU, LL, L/UL/U, ρU2\rho U^2 leaves Du∗/Dt∗=−∇∗p∗+Re−1∇∗2u∗D\mathbf u^*/Dt^* = -\nabla^*p^* + \mathrm{Re}^{-1}\nabla^{*2}\mathbf u^*: same shape and Re, same flow.
  • Re is inertia over viscosity: in the Taylor–Green cell the RMS ratio is Re/(22)\mathrm{Re}/(2\sqrt2), from 0.0350.035 to 35 35535\,355.
  • Re ≪ 1: Stokes equations, linear and reversible, CD=24/ReC_D = 24/\mathrm{Re}; a 1010 micron droplet falls at 33 mm/s.
  • Re ≫ 1: viscosity survives in layers δ/L∼Re−1/2\delta/L \sim \mathrm{Re}^{-1/2}, 0.660.66 mm on a light wing.
  • Mach (Δρ/ρ≈M2/2\Delta\rho/\rho \approx M^2/2), Froude and Strouhal compare other pairs of terms; the light wing is 5.1×1065.1\times10^6, the airliner 3.2×1073.2\times10^7 at Mach 0.850.85.
  • Buckingham π: the groups are the kernel of the dimension matrix; for sphere drag, CD=f(Re)C_D = f(\mathrm{Re}).
  • Matching Re at 1:101{:}10 in air needs 500500 m/s; pressure, water or cold gas rescue it, and a cold tunnel at 1.711.71 atm matches the airliner’s Re and Mach together.
  • Next, 11.6: at high Re the viscous term lives only in thin layers, but what does it diffuse there? Take the curl of Navier–Stokes and pressure disappears, leaving one quantity that is carried by the flow and spread by viscosity: vorticity. Boundary layers are sheets of it, and wakes and wingtip vortices are where it goes.
Finished this lesson?

Mark it complete to track your progress (saved in this browser). Reading every section does this for you.

Where this leads

These later lessons build directly on this one: