Flow Lab

Syllabus / Module 14: Applications: Physics & Engineering

Lesson 14.5

Vibrating Beams & Structures

Euler–Bernoulli beam equation, static deflection, natural frequencies of beams, and finite-difference/FEM beam solvers.

Written byClaude Sonnet

🌊
Why this matters for simulation: Wings are beams; aeroelasticity couples them to the flow.

By the end of this lesson you can…

  • Derive the Euler-Bernoulli equations $EI\,y'''' = q$ and $\rho A\,y_{tt} + EI\,y'''' = q$ from moment-curvature and slice equilibrium, and state the two conditions that clamped, pinned, free and sliding ends impose.
  • Solve the static cantilever and simply supported beams by integration and quote $PL^3/(3EI)$, $qL^4/(8EI)$ and $5qL^4/(384EI)$.
  • Separate variables to get $\phi'''' = \beta^4\phi$, solve it with $\cos$, $\sin$, $\cosh$, $\sinh$, and find the cantilever frequencies from $1 + \cos\beta L\cosh\beta L = 0$.
  • Prove that beam modes are orthogonal, and explain why $\omega_n$ grows like $n^2$ (dispersive) where a string's grows like $n$.
  • Build a finite-difference beam solver with ghost-point end rows, check its tip deflection and frequencies against the exact values, and measure second-order convergence.
  • Compute the tip deflection and first bending frequency of a cantilever wing-beam under elliptic lift for a stated $EI$, $\rho A$ and $L$.
On this page
·
  1. From bending to EI y′′′′=qEI\,y'''' = qEIy′′′′=q
  2. Four kinds of end
  3. Static deflection
  4. Free vibration: separation of variables
  5. Orthogonality, and why a beam is not a string
  6. A numerical beam solver
  7. The wing as a cantilever
  8. Two more glimpses
  9. The lab
  10. The exercise: a beam solver
  11. Concept checks
  12. Problems
  13. What comes next
  14. Key Takeaways

14.4 Engineering Heat Transfer ended with a fin whose boundary condition turned θ′′=m2θ\theta'' = m^2\theta into a problem about cosh⁡\cosh and sinh⁡\sinh. A beam is the same bookkeeping one derivative up. Press the tip of a ruler clamped to a desk and it bends; let go and it rings. Both come from one equation of fourth order in xx, with two end conditions at each end. The static problem was previewed in 5.7 Higher-Order Linear Equations, and the ringing is the standing-wave problem of 9.4 The Wave Equation & Vibrating Strings with a twist: the frequencies are not a harmonic series. We derive both, solve them exactly, build a finite-difference solver and check it, and end on a lift-loaded wing spar.

▶ Video Bending moment and curvature, the four kinds of end, a cantilever under load, mode shapes and the frequency equation, a beam against a string, a finite-difference solver converging, and a wing as a cantilevercaptions only
Transcript
  1. A beam is the fin's boundary-condition bookkeeping one derivative up. Bending moment is E I times the curvature y double prime, where E I is Young's modulus times the second moment of area. For a rectangle I is b t cubed over 12, so doubling the thickness makes the beam 8 times stiffer. Slice equilibrium gives E I y four primes equals q, and the slice's mass adds rho A y t t. Fourth order means four constants, so two conditions at each end. Clamped sets y and slope to zero, pinned sets y and moment, free sets moment and shear, sliding sets slope and shear.
  2. Take the cantilever with a tip load P. Integrating twice, the tip moves P L cubed over 3 E I: a spring. Under uniform load q it is q L to the fourth over 8 E I. Pinned at both ends, the midspan deflection is 5 q L to the fourth over 384 E I, and clamped at both ends, q L to the fourth over 384 E I. At the same scale the cantilever sags 9.6 times more than the simply supported beam. For a steel ruler, 25 millimetres wide, 3 thick and 0.30 metres long, with E and rho from memory, E I is 11.25. A 1 newton tip load deflects it 0.800 millimetres.
  3. Separate variables: space gives phi four primes equals beta to the fourth phi, solved by cos, sin, cosh and sinh. The four end conditions have a nonzero solution only when a determinant vanishes. For the cantilever that is cos beta L times cosh beta L equals minus 1. There is no closed form, but bisection gives 1.8751, 4.6941 and 7.8548. The nth mode has n minus 1 interior nodes, and omega is beta L squared times the square root of E I over rho A L to the fourth. For the ruler, the first frequency is 170.77 radians per second, 27.18 hertz.
  4. A string has omega equal to n c pi over L, a harmonic series: 1, 2, 3. A pinned beam has omega proportional to n squared: 1, 4, 9, 16. For the cantilever the ratios are 1, 6.267, 17.55, 34.39 and 56.84, so a beam's overtones are not harmonics. In plane-wave language omega is the square root of E I over rho A times k squared. The phase speed grows with k, and the group speed is twice the phase speed. The beam is dispersive, and the string is not.
  5. A grid needs no determinant. Applying the second difference twice gives the five-point stencil, 1, minus 4, 6, minus 4, 1, over h to the fourth. Near an end it reaches outside the beam, so the end conditions supply ghost points. Clamped makes row 1 equal to 7, minus 4, 1. Free gives the last two rows 1, minus 4, 5, minus 2 and 2, minus 4, 2. For the cantilever under uniform load the tip error falls from 1.00 percent to 0.0039 percent, and the first-frequency error from minus 0.837 to minus 0.0033. Each doubling divides the error by 4.00: second order.
  6. Omega max is about 4 times the square root of E I over rho A, over h squared: at N equal to 80, 25,590 against 4 N squared, 25,600. Explicit leapfrog needs omega max times delta t at most 2, so delta t is at most h squared over twice the square root of E I over rho A. A string's limit is h over c, so halving h halves the string's limit and quarters the beam's. In a pinned-pinned run with N equal to 20, the limit is 1.258 times 10 to the minus 3. At 0.95 of it the solution stays bounded; at 1.05 it grew to 10 to the 106 in 400 steps.
  7. A wing spar is a cantilever clamped at the root. The half-wing of the light aircraft carries W equal to 6316.5 newtons over s equal to 5.5 metres. The usual first model, from memory, is elliptic lift. Superposing tip loads gives a deflection of W s cubed over E I times 1 eighth minus 4 over 45 pi. With a hypothetical E I of 1.0 times 10 to the 6 and rho A of 15, the tip moves 0.10163 metres, which is 0.7736 of the uniform-load value, because the lift sits nearer the root. The root moment is 14,744 newton metres, against 17,370 for uniform lift, and the first frequency is 4.776 hertz. Flutter, from memory, needs a second mode, and a pure beam cannot show it.
  8. The beam is a wave problem with the unusual relation omega proportional to k squared. The next lesson, acoustics, covers sound. The syllabus lists the acoustic wave equation from linearised compressible flow, the speed of sound, standing waves in pipes, spherical waves, reflection and transmission, the Doppler effect and a 2D room simulation. There the equation is the ordinary wave equation, derived from the Euler equations rather than from a force balance on a slice.

From bending to EI y′′′′=qEI\,y'''' = q

A thin beam of length LL lies along xx, and y(x)y(x) is its small transverse displacement, measured in the direction of the load. Euler-Bernoulli theory assumes that cross-sections stay flat and perpendicular to the bent centreline, that slopes are small, and that the material is linear elastic. Shear deformation and rotary inertia are ignored.

A beam with curvature κ≈y′′\kappa \approx y'' stretches the outer fibres and shortens the inner ones in proportion to their distance zz from the centre: strain −zκ-z\kappa, stress −Ezκ-Ez\kappa. The resultant moment M=∫(−σ) z dAM = \int(-\sigma)\,z\,dA is

M=EI y′′,I=∫z2 dA.M = EI\,y'', \qquad I = \int z^2\,dA .

EIEI is the bending stiffness: Young’s modulus times the second moment of area. For a rectangle of width bb and thickness tt, I=bt3/12I = bt^3/12, so doubling the thickness makes the beam eight times stiffer.

Now take a slice of width dxdx. The shear forces on its faces differ by the load, V′=qV' = q per unit length, and its moment balance (the load’s own moment is second order) gives M′=VM' = V. Eliminating,

 EI y′′′′=q(x) \boxed{\,EI\,y'''' = q(x)\,}

with EI y′′EI\,y'' the bending moment and EI y′′′EI\,y''' the shear force, up to a sign that books choose differently. For motion, Newton’s law for the slice has mass ρA dx\rho A\,dx (ρ\rho density, AA cross-section area), the load qq, and the internal force −EI y′′′′-EI\,y'''' per length:

ρA ytt+EI y′′′′=q(x,t).\rho A\,y_{tt} + EI\,y'''' = q(x, t).

The homogeneous version is the one in the table of 9.9 Infinite Domains & the Fourier Transform with β=EI/ρA\beta = EI/\rho A.

Four kinds of end

A fourth-order equation has four constants, so it needs four conditions, two at each end. Each end pairs a geometric quantity with a force quantity, and exactly one of each pair is given:

EndConditionsPhysical meaning
clamped (built in)y=0, y′=0y = 0,\ y' = 0a wall: no displacement and no rotation; the wall supplies force and moment
pinned (simply supported)y=0, y′′=0y = 0,\ y'' = 0held in place but free to turn: zero moment
freey′′=0, y′′′=0y'' = 0,\ y''' = 0nothing touches it: zero moment and zero shear
slidingy′=0, y′′′=0y' = 0,\ y''' = 0in a frictionless slot: slope held at zero, no transverse force

Some of the sixteen combinations are singular: a pinned-free beam can swing about the pin and a free-free beam can float away, so a static load has no unique answer.

Static deflection

Take the cantilever (clamped at 00, free at LL) with a tip load PP. Equilibrium of the piece beyond xx gives the moment P(L−x)P(L - x), which vanishes at the free end, so EI y′′=P(L−x)EI\,y'' = P(L - x). Integrate with y′(0)=0y'(0) = 0 and y(0)=0y(0) = 0:

EI y′=P(Lx−x22),y(x)=Px2(3L−x)6EI,δ=y(L)=PL33EI.EI\,y' = P\Bigl(Lx - \frac{x^2}{2}\Bigr), \qquad y(x) = \frac{P x^2(3L - x)}{6EI}, \qquad \delta = y(L) = \frac{PL^3}{3EI}.

So the tip load acts like a spring, stiffness k=3EI/L3k = 3EI/L^3. A uniform load, integrated four times, gives qL4/(8EI)qL^4/(8EI) (5.7 Higher-Order Linear Equations). For the simply supported beam, y=qx(L3−2Lx2+x3)/(24EI)y = qx(L^3 - 2Lx^2 + x^3)/(24EI) satisfies y=y′′=0y = y'' = 0 at both ends, and for clamped-clamped y=qx2(L−x)2/(24EI)y = qx^2(L - x)^2/(24EI) works (Problem 2). The lab’s exact solver, which solves the four end conditions as a 4×44\times4 system, confirms:

Beam and loadPeak deflection
clamped-free, tip load PPPL3/(3EI)PL^3/(3EI), at the tip
clamped-free, uniform qqqL4/(8EI)qL^4/(8EI), at the tip
pinned-pinned, uniform qq5qL4/(384EI)5qL^4/(384EI), midspan
pinned-pinned, midspan PPPL3/(48EI)PL^3/(48EI)
clamped-clamped, uniform qqqL4/(384EI)qL^4/(384EI), midspan

Under the same load a cantilever sags 4848 times more than a clamped-clamped beam and 9.69.6 times more than a simply supported one.

Worked example: a steel ruler. Take b=25b = 25 mm, t=3t = 3 mm, L=0.30L = 0.30 m, and (from memory) E=200E = 200 GPa, ρ=7850\rho = 7850 kg/m3^3. Then I=5.625×10−11I = 5.625\times10^{-11} m4^4, EI=11.25EI = 11.25 N m2^2, ρA=0.58875\rho A = 0.58875 kg/m. A 11 N tip load deflects it 0.8000.800 mm (stiffness 12501250 N/m), and its own weight, 5.7765.776 N/m, sags the tip 0.5200.520 mm.

Free vibration: separation of variables

Set q=0q = 0 and look for standing waves y=ϕ(x) T(t)y = \phi(x)\,T(t). Then ρA ϕT′′+EI ϕ′′′′T=0\rho A\,\phi T'' + EI\,\phi'''' T = 0, so ϕ′′′′/ϕ=−ρA T′′/(EI T)\phi''''/\phi = -\rho A\,T''/(EI\,T) is a constant, which we call β4\beta^4:

T′′=−ω2T,ϕ′′′′=β4ϕ,ω2=EIρA β4.T'' = -\omega^2 T, \qquad \phi'''' = \beta^4\phi, \qquad \omega^2 = \frac{EI}{\rho A}\,\beta^4 .

The sign is forced: multiply ϕ′′′′=β4ϕ\phi'''' = \beta^4\phi by ϕ\phi and integrate by parts. For every end type the boundary terms vanish and β4∫ϕ2 dx=∫(ϕ′′)2 dx≥0\beta^4\int\phi^2\,dx = \int(\phi'')^2\,dx \ge 0, the bending energy. A free-free beam has two zero-frequency modes (translation and rotation), where ϕ′′\phi'' vanishes.

The spatial equation has characteristic polynomial r4=β4r^4 = \beta^4, with roots ±β\pm\beta and ±iβ\pm i\beta (as in 5.7 Higher-Order Linear Equations), so the general solution is

ϕ(x)=Acos⁡βx+Bsin⁡βx+Ccosh⁡βx+Dsinh⁡βx.\phi(x) = A\cos\beta x + B\sin\beta x + C\cosh\beta x + D\sinh\beta x .

Compared with the string’s ϕ′′=−k2ϕ\phi'' = -k^2\phi there are two extra hyperbolic pieces, like the fin profile. The four end conditions give a 4×44\times4 homogeneous system, and nonzero solutions exist only when its determinant vanishes: a frequency equation for βL\beta L. This is the Sturm-Liouville problem of 8.2 Eigenvalue Problems & Sturm–Liouville Theory at fourth order.

Pinned-pinned. At x=0x = 0, ϕ=ϕ′′=0\phi = \phi'' = 0 gives A+C=0A + C = 0 and −A+C=0-A + C = 0, so ϕ=Bsin⁡βx+Dsinh⁡βx\phi = B\sin\beta x + D\sinh\beta x. At x=Lx = L the same two conditions give Bsin⁡βL±Dsinh⁡βL=0B\sin\beta L \pm D\sinh\beta L = 0, so D=0D = 0 and sin⁡βL=0\sin\beta L = 0: the string’s answer.

βnL=nπ,ϕn=sin⁡nπxL,ωn=n2π2EIρAL4.\beta_nL = n\pi, \qquad \phi_n = \sin\frac{n\pi x}{L}, \qquad \omega_n = n^2\pi^2\sqrt{\frac{EI}{\rho AL^4}}.

Cantilever. Clamping at 00 (ϕ=ϕ′=0\phi = \phi' = 0) leaves ϕ=A(cos⁡βx−cosh⁡βx)+B(sin⁡βx−sinh⁡βx)\phi = A(\cos\beta x - \cosh\beta x) + B(\sin\beta x - \sinh\beta x). The free end needs ϕ′′(L)=ϕ′′′(L)=0\phi''(L) = \phi'''(L) = 0:

A(cos⁡βL+cosh⁡βL)+B(sin⁡βL+sinh⁡βL)=0,A(sin⁡βL−sinh⁡βL)−B(cos⁡βL+cosh⁡βL)=0.A(\cos\beta L + \cosh\beta L) + B(\sin\beta L + \sinh\beta L) = 0, \qquad A(\sin\beta L - \sinh\beta L) - B(\cos\beta L + \cosh\beta L) = 0 .

The determinant is −(cos⁡+cosh⁡)2−(sin⁡+sinh⁡)(sin⁡−sinh⁡)=−2(1+cos⁡βLcosh⁡βL)-(\cos + \cosh)^2 - (\sin + \sinh)(\sin - \sinh) = -2(1 + \cos\beta L\cosh\beta L), using cos⁡2+sin⁡2=1\cos^2 + \sin^2 = 1 and cosh⁡2−sinh⁡2=1\cosh^2 - \sinh^2 = 1. The frequency equation is

cos⁡βL cosh⁡βL=−1.\cos\beta L\,\cosh\beta L = -1 .

There is no closed form, but bisection works: cosh⁡\cosh is huge for βL>3\beta L \gt 3, so the roots hug the zeros of cos⁡\cos, βnL≈(2n−1)π/2\beta_nL \approx (2n - 1)\pi/2 (7.85407.8540 against 7.85487.8548 for n=3n = 3). The first three are 1.8751, 4.6941, 7.85481.8751,\ 4.6941,\ 7.8548. The mode shape is ϕn=(cosh⁡βx−cos⁡βx)−σn(sinh⁡βx−sin⁡βx)\phi_n = (\cosh\beta x - \cos\beta x) - \sigma_n(\sinh\beta x - \sin\beta x) with σn=(cosh⁡βL+cos⁡βL)/(sinh⁡βL+sin⁡βL)\sigma_n = (\cosh\beta L + \cos\beta L)/(\sinh\beta L + \sin\beta L) (σ1=0.7341\sigma_1 = 0.7341); the nnth mode has n−1n - 1 interior nodes, which the lab counts.

Every other pair of ends is the same game with a different determinant:

EndsβL\beta L (first three)ω2/ω1\omega_2/\omega_1
clamped-free (cantilever)1.8751, 4.6941, 7.85481.8751,\ 4.6941,\ 7.85486.2676.267
pinned-pinnedπ, 2π, 3π\pi,\ 2\pi,\ 3\pi4.0004.000
clamped-pinned3.9266, 7.0686, 10.21023.9266,\ 7.0686,\ 10.21023.2413.241
clamped-clamped, or free-free4.7300, 7.8532, 10.99564.7300,\ 7.8532,\ 10.99562.7572.757

Clamped-clamped and free-free share a frequency equation (cos⁡βLcosh⁡βL=1\cos\beta L\cosh\beta L = 1) but not shapes. The frequency is ω=(βL)2EI/(ρAL4)\omega = (\beta L)^2\sqrt{EI/(\rho AL^4)}, so the ruler’s first cantilever frequency is (1.8751)2×11.25/(0.58875×0.304)=170.77(1.8751)^2\times\sqrt{11.25/(0.58875\times0.30^4)} = 170.77 rad/s, 27.1827.18 Hz; pinned at both ends over the same span it would ring at 76.2976.29 Hz.

Orthogonality, and why a beam is not a string

Let ϕm\phi_m and ϕn\phi_n be modes with βm≠βn\beta_m \ne \beta_n. Multiply ϕn′′′′=βn4ϕn\phi_n'''' = \beta_n^4\phi_n by ϕm\phi_m, subtract the same with mm and nn swapped, and recognise a total derivative:

(βn4−βm4)∫0Lϕmϕn dx=[ϕmϕn′′′−ϕm′ϕn′′+ϕm′′ϕn′−ϕm′′′ϕn]0L=0.(\beta_n^4 - \beta_m^4)\int_0^L\phi_m\phi_n\,dx = \Bigl[\phi_m\phi_n''' - \phi_m'\phi_n'' + \phi_m''\phi_n' - \phi_m'''\phi_n\Bigr]_0^L = 0 .

The bracket vanishes because at every end each of the four products contains a factor that the end condition sets to zero (check it for all four end types). So ∫ϕmϕn dx=0\int\phi_m\phi_n\,dx = 0: the modes are orthogonal, like the normal modes of 14.1 Coupled Oscillators & Normal Modes, and a general motion is a sum of modes with an=∫y0ϕn dx/∫ϕn2 dxa_n = \int y_0\phi_n\,dx/\int\phi_n^2\,dx. The lab evaluates ∫ϕ1ϕm/∫ϕ12\int\phi_1\phi_m/\int\phi_1^2 numerically: 11 for m=1m = 1 and about 10−1410^{-14} for the others.

The big difference is the spectrum. A string has ωn=ncπ/L\omega_n = nc\pi/L, a harmonic series. A pinned beam has ωn=n2ω1\omega_n = n^2\omega_1: 1,4,9,161, 4, 9, 16. For the cantilever ωn/ω1=1, 6.267, 17.55, 34.39, 56.84\omega_n/\omega_1 = 1,\ 6.267,\ 17.55,\ 34.39,\ 56.84, so a beam’s overtones are not harmonics. In plane-wave language, ω=EI/ρA k2\omega = \sqrt{EI/\rho A}\,k^2: the phase speed vp=ω/kv_p = \omega/k grows with kk and the group speed is vg=2vpv_g = 2v_p, as in 9.9 Infinite Domains & the Fourier Transform. The beam is dispersive; the string is not.

A numerical beam solver

The exact approach needs a determinant per pair of ends and fails for variable EI(x)EI(x). A grid does not care. Use nodes xi=ihx_i = ih, h=L/Nh = L/N. Applying the second difference twice,

y′′′′(xi)≈yi−2−4yi−1+6yi−4yi+1+yi+2h4,error h26 y(6).y''''(x_i) \approx \frac{y_{i-2} - 4y_{i-1} + 6y_i - 4y_{i+1} + y_{i+2}}{h^4}, \qquad \text{error } \frac{h^2}{6}\,y^{(6)} .

That stencil reaches two nodes each way, so near an end it needs ghost points outside the beam, and the end conditions supply them. For the cantilever with unknowns y1,…,yNy_1,\dots,y_N:

  • Clamped at x=0x = 0: y0=0y_0 = 0 drops out, and y′(0)=0y'(0) = 0 with a central difference sets y−1=y1y_{-1} = y_1. Row 11 is (7,−4,1)(7, -4, 1).
  • Free at x=Lx = L: zero moment gives yN+1=2yN−yN−1y_{N+1} = 2y_N - y_{N-1}, and zero shear yN+2=4yN−4yN−1+yN−2y_{N+2} = 4y_N - 4y_{N-1} + y_{N-2}. Row N−1N - 1 becomes (1,−4,5,−2)(1, -4, 5, -2) and row NN becomes (2,−4,2)(2, -4, 2).
  • Every other row is (1,−4,6,−4,1)(1, -4, 6, -4, 1).

Call this matrix SS and set K=(EI/h4) SK = (EI/h^4)\,S. The static problem is Ky=qK\mathbf y = \mathbf q; the vibration problem is Kv=ω2MvK\mathbf v = \omega^2 M\mathbf v with M=ρA IM = \rho A\,I, the eigenproblem of M−1KM^{-1}K from 14.1 Coupled Oscillators & Normal Modes. The end rows make KK non-symmetric, yet its eigenvalues came out real. A point load PP enters as P/hP/h at its node, or 2P/h2P/h at a free end (half a cell).

For the cantilever in units L=EI=ρA=1L = EI = \rho A = 1 (checked against an independent script), the tip deflection under uniform load and the first frequency, whose exact values are 0.1250.125 and (βL)2=3.5160(\beta L)^2 = 3.5160:

NNtip deflectionerrorω1\omega_1error
10100.1262500.1262501.00%1.00\%3.48663.4866−0.837%-0.837\%
20200.1253130.1253130.250%0.250\%3.50863.5086−0.211%-0.211\%
40400.1250780.1250780.0625%0.0625\%3.51423.5142−0.0528%-0.0528\%
80800.1250200.1250200.0156%0.0156\%3.51563.5156−0.0132%-0.0132\%
1601600.1250050.1250050.0039%0.0039\%3.51593.5159−0.0033%-0.0033\%

Each doubling of NN cuts the error by 4.004.00 for the deflection and 3.973.97 to 4.004.00 for the frequency: second order, as the stencil promised. Higher modes are harder: at N=40N = 40 the errors of ω1,…,ω5\omega_1,\dots,\omega_5 are −0.053%-0.053\%, −0.27%-0.27\%, −0.56%-0.56\%, −0.96%-0.96\%, −1.46%-1.46\% (fewer points per bend).

Finite elements (10.13 Introduction to Finite Elements) would need slope-continuous shape functions, since the beam energy is ∫EI (y′′)2 dx\int EI\,(y'')^2\,dx; from memory, the standard fix is Hermite cubic elements. I did not implement them.

The wing as a cantilever

A wing spar carries the lift of its half-wing to the fuselage: a cantilever clamped at the root with a distributed lift. Use the light aircraft of 12.1 Airfoils, Circulation & Lift: 12 63312\,633 N over an 1111 m span, so one half-wing carries W=6316.5W = 6316.5 N over s=5.5s = 5.5 m. Spanwise lift is not uniform. The usual first model (from memory, lifting-line theory, not derived in this course) is elliptic: q(x)=q01−(x/s)2q(x) = q_0\sqrt{1 - (x/s)^2}, measured from the root, with area πq0s/4=W\pi q_0s/4 = W, so q0=4W/(πs)=1462.3q_0 = 4W/(\pi s) = 1462.3 N/m.

The tip deflection needs no new solver. A point load PP at ξ\xi bends the beam to Pξ3/(3EI)P\xi^3/(3EI) at ξ\xi with slope Pξ2/(2EI)P\xi^2/(2EI), and the part beyond is straight, so the tip moves Pξ2(3s−ξ)/(6EI)P\xi^2(3s - \xi)/(6EI). Superpose over the load, with ξ=ssin⁡θ\xi = s\sin\theta:

δ=∫0sq(ξ) ξ2(3s−ξ)6EI dξ=q0s46EI∫0π/2 ⁣sin⁡2θ cos⁡2θ (3−sin⁡θ) dθ=q0s46EI(3π16−215),\delta = \int_0^s\frac{q(\xi)\,\xi^2(3s - \xi)}{6EI}\,d\xi = \frac{q_0s^4}{6EI}\int_0^{\pi/2}\!\sin^2\theta\,\cos^2\theta\,(3 - \sin\theta)\,d\theta = \frac{q_0s^4}{6EI}\Bigl(\frac{3\pi}{16} - \frac{2}{15}\Bigr),

which is δ=Ws3EI(18−445π)=0.0967 Ws3EI\delta = \dfrac{Ws^3}{EI}\Bigl(\dfrac18 - \dfrac{4}{45\pi}\Bigr) = 0.0967\,\dfrac{Ws^3}{EI}. Spread uniformly the same lift gives Ws3/(8EI)Ws^3/(8EI), so the elliptic tip moves only 0.77360.7736 of that: the lift sits nearer the root. The root moment is 4Ws/(3π)=14 7444Ws/(3\pi) = 14\,744 N m, against Ws/2=17 370Ws/2 = 17\,370 N m.

Take hypothetical constant values (a real spar tapers): EI=1.0×106EI = 1.0\times10^6 N m2^2, ρA=15\rho A = 15 kg/m. At 11 g the tip deflects 0.10160.1016 m, 1.85%1.85\% of the span, and the frequencies fn=(βnL)2EI/(ρAs4)/2πf_n = (\beta_nL)^2\sqrt{EI/(\rho As^4)}/2\pi are 4.7764.776, 29.9329.93 and 83.8183.81 Hz. At 33 g the tip is at 0.30490.3049 m, 5.5%5.5\% of the span. Finite differences with N=100N = 100 give f1=4.776f_1 = 4.776 Hz; the lab’s N=40N = 40 gives 4.7744.774 Hz and a tip about 0.5%0.5\% low, since the elliptic load’s square-root end is less smooth. The model ignores torsion and taper.

Flutter, a glimpse, entirely from memory. Lift on a bending wing depends on its deflection and velocity, so in modal form Mq¨+Kq=A(U) q+B(U) q˙M\ddot{\mathbf q} + K\mathbf q = A(U)\,\mathbf q + B(U)\,\dot{\mathbf q} and the eigenvalues move with airspeed UU. Flutter is the speed at which a pair crosses into the right half-plane, the picture of 12.6 Lateral-Directional Modes. As I remember it, it needs a second mode, usually torsion; a pure beam cannot show it.

Two more glimpses

Both from memory, not derived. Timoshenko theory adds shear deformation and rotary inertia, lowering the frequencies of short thick beams; it is negligible for the ruler (L/t=100L/t = 100). Euler buckling: a cantilever under axial compression buckles at Pcr=π2EI/(4L2)P_{cr} = \pi^2EI/(4L^2), which is 308308 N for the ruler.

The lab

◆ Interactive Static deflection with exact and finite-difference solutions, vibration modes, and the wing as a cantilever

Loading simulation…

Three experiments:

  1. Static, default (clamped-free, uniform, N=40N = 40): peak 0.1250000.125000, finite differences 0.1250780.125078, largest nodal error 6.25×10−46.25\times10^{-4} of the peak (2.50×10−32.50\times10^{-3} at N=20N = 20). A tip point load gives 0.3333330.333333; pinned-pinned 0.0130210.013021; clamped-clamped 0.0026040.002604. Pinned-free and free-free are singular, and the lab says so.
  2. Modes, clamped-free, n=3n = 3: βL=7.8548\beta L = 7.8548, ω=61.6972\omega = 61.6972, two interior nodes. At N=40N = 40 the numerical ω3=61.350\omega_3 = 61.350; for N=10N = 10 to 160160 the error is −8.26%,−2.21%,−0.563%,−0.141%,−0.035%-8.26\%, -2.21\%, -0.563\%, -0.141\%, -0.035\%.
  3. Wing, defaults: tip 0.101630.10163 m (closed form and exact solver), 0.101160.10116 m at N=40N = 40, uniform comparison 0.131360.13136 m; at 33 g, 0.30490.3049 m.

The exercise: a beam solver

The playgrounds find the cantilever roots by bisection and tabulate the wing. The exercise builds the finite-difference matrix, solves the static problem, and extracts frequencies from M−1KM^{-1}K with np.linalg.eigvals.

🐍 PythonRoots of 1 + cos cosh = 0 by bisection
⌘/Ctrl + Enter
🐍 PythonThe wing: tip deflection against load factor, and frequencies
⌘/Ctrl + Enter
🧪 Coding exerciseExercise: finite-difference cantilever
⌘/Ctrl + Enter

Concept checks

✅ Concept check

Same beam and same uniform load $q$. By what factor is the clamped-free tip deflection larger than the clamped-clamped midspan deflection?

✅ Concept check

A pinned-pinned beam has $\omega_n = n^2\omega_1$. What is $\omega_2/\omega_1$, and (for comparison) what would a string give?

✅ Concept check

You halve the thickness of a rectangular cantilever, keeping its width and length. Since $I\propto t^3$ and $\rho A\propto t$, by what factor does the first natural frequency change?

✅ Concept check

You double $N$ in the finite-difference beam solver. What happens to the error in the first frequency?

Problems

✏️ Pen & paperThe ruler as a spring●○○

For the steel ruler (EI=11.25EI = 11.25 N m2^2, L=0.30L = 0.30 m) find the tip force that gives a 55 mm deflection.

Show worked solution

k=3EI/L3=1250k = 3EI/L^3 = 1250 N/m, so P=kδ=6.25P = k\delta = 6.25 N.

✏️ Pen & paperClamped at both ends●●○

Check that y=qx2(L−x)2/(24EI)y = qx^2(L - x)^2/(24EI) solves EI y′′′′=qEI\,y'''' = q with y=y′=0y = y' = 0 at both ends, and find the midspan deflection.

Show worked solution

Expand: x2(L−x)2=L2x2−2Lx3+x4x^2(L - x)^2 = L^2x^2 - 2Lx^3 + x^4. Only x4x^4 survives four derivatives, giving 2424, so EI y′′′′=qEI\,y'''' = q. The factor x2x^2 gives y(0)=y′(0)=0y(0) = y'(0) = 0, and the factor (L−x)2(L - x)^2 gives y(L)=y′(L)=0y(L) = y'(L) = 0. At x=L/2x = L/2: y=q(L/2)2(L/2)2/(24EI)=qL4/(384EI)y = q(L/2)^2(L/2)^2/(24EI) = qL^4/(384EI), one forty-eighth of the cantilever’s qL4/(8EI)qL^4/(8EI).

✏️ Pen & paperRoot moment of a wing●●○

For elliptic lift q=q01−(x/s)2q = q_0\sqrt{1 - (x/s)^2} with total W=πq0s/4W = \pi q_0s/4, find the bending moment at the root, ∫0sq(ξ) ξ dξ\int_0^s q(\xi)\,\xi\,d\xi, and compare it with a uniform load of the same WW. Evaluate for W=6316.5W = 6316.5 N and s=5.5s = 5.5 m.

Show worked solution

With ξ=ssin⁡θ\xi = s\sin\theta: ∫0sξ1−(ξ/s)2 dξ=s2∫0π/2sin⁡θcos⁡2θ dθ=s2/3\int_0^s\xi\sqrt{1 - (\xi/s)^2}\,d\xi = s^2\int_0^{\pi/2}\sin\theta\cos^2\theta\,d\theta = s^2/3. So M=q0s2/3=4Ws/(3π)M = q_0s^2/3 = 4Ws/(3\pi). Uniform: Ws/2Ws/2. The ratio is 8/(3π)=0.8498/(3\pi) = 0.849. Numerically M=14 744M = 14\,744 N m against 17 37017\,370 N m.

What comes next

The beam is a wave problem with the unusual relation ω∝k2\omega\propto k^2. The next lesson, 14.6 Acoustics & Wave Propagation, covers sound; the syllabus lists the acoustic wave equation from linearised compressible flow, the speed of sound, standing waves in pipes, spherical waves, reflection and transmission, the Doppler effect and a 2D room simulation. There the equation is the ordinary wave equation of 9.4 The Wave Equation & Vibrating Strings, derived from the Euler equations of 11.3 Momentum: Pressure & the Euler Equations rather than from a force balance on a slice.

Key Takeaways

  • Moment-curvature M=EI y′′M = EI\,y'' plus slice equilibrium gives EI y′′′′=qEI\,y'''' = q and ρA ytt+EI y′′′′=q\rho A\,y_{tt} + EI\,y'''' = q; EI∝t3EI\propto t^3 for a rectangle.
  • Two end conditions per end: clamped (y,y′y, y'), pinned (y,y′′y, y''), free (y′′,y′′′y'', y'''), sliding (y′,y′′′y', y''').
  • Static tips: PL3/(3EI)PL^3/(3EI), qL4/(8EI)qL^4/(8EI); simply supported midspan 5qL4/(384EI)5qL^4/(384EI); clamped-clamped qL4/(384EI)qL^4/(384EI).
  • Separation gives ϕ′′′′=β4ϕ\phi'''' = \beta^4\phi; the cantilever equation cos⁡βLcosh⁡βL=−1\cos\beta L\cosh\beta L = -1 has βL=1.8751,4.6941,7.8548\beta L = 1.8751, 4.6941, 7.8548, and ω=(βL)2EI/(ρAL4)\omega = (\beta L)^2\sqrt{EI/(\rho AL^4)}.
  • Modes are orthogonal, but ωn∝n2\omega_n\propto n^2: a beam is dispersive and its overtones are not harmonics.
  • The 5-point stencil with ghost-point end rows is second order, and its explicit time step must shrink like h2h^2.
  • Elliptic lift gives the half-wing tip deflection Ws3EI(18−445π)\frac{Ws^3}{EI}\bigl(\tfrac18 - \tfrac{4}{45\pi}\bigr), 0.77360.7736 of the uniform-load value.
Finished this lesson?

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