Everything vibrates
A washing machine on its spin cycle shuffles across the floor. A car's body bobs gently after a speed bump and then settles. A phone buzzes because a tiny off-centre mass spins inside it. In June 2000 the new Millennium Footbridge in London began to sway sideways under the feet of the crowds on its opening weekend and had to be closed for nearly two years. Tall towers sway in the wind, turbine blades flutter, and every machine with a rotating part shakes a little once per revolution.
Vibration is sometimes the point: loudspeakers, ultrasonic cleaners, the vibratory feeders that line up parts on a production line, the vibrating micro-structures inside the gyroscope of a phone. More often it is the enemy, because it causes fatigue cracks, noise, wear, discomfort and, in the worst case, collapse. Either way the engineer asks the same three questions. At what frequencies does this thing want to vibrate? How quickly does a vibration die away? And how big does it get when something shakes it? This lesson answers them with the simplest model that captures the physics, a mass on a spring with a damper, and then with two masses, which is enough to see the idea of a mode and to understand the tuned mass damper that steadies a skyscraper. It builds on the dynamics of particles and rigid bodies, and on ordinary differential equations.
A mass on a spring
Pull a mass on a spring aside and let go. The spring pulls it back; it overshoots, because it has gained speed; the spring pulls the other way; and so on. Energy sloshes between the spring, as strain energy, and the mass, as kinetic energy, twice per cycle. A stiffer spring pulls harder, so the cycle is quicker; a heavier mass responds more sluggishly, so it is slower. The natural frequency goes as .
Two shortcuts are worth knowing from the start. First, the weight of the mass stretches the spring by a static deflection , so , and the natural frequency can be read off how far the system sags under its own weight:
Second, a vibrating system forced at its own natural frequency responds enormously, the way a playground swing goes higher and higher when it is pushed once per swing, in time with its motion. That is resonance, and most of this lesson is about predicting it, using it, or keeping away from it.
The single-degree-of-freedom system
Newton's second law for the mass in the figure, with measured from static equilibrium, gives the equation that underlies all of vibration engineering:
Definition— natural frequency, damping ratio
The undamped natural frequency is in rad/s, or in hertz; the period is . The critical damping is , and the damping ratio is . The system is underdamped if , critically damped if and overdamped if .
Theorem— Free vibration of a damped oscillator
With and , every solution has the form
an oscillation at the damped natural frequency inside an exponentially decaying envelope. Successive peaks, one period apart, have a constant ratio, whose logarithm is the logarithmic decrement
Proof
Try : the equation becomes , so and . The real and imaginary parts of are and , and since the equation is linear and second order, every solution is a combination of these two. One period later the bracket repeats exactly and the envelope has shrunk by , whose logarithm is .
∎
The complex roots are the heart of it: their imaginary part sets the frequency, and their real part sets the decay. With both roots are real and negative, and the mass creeps back to rest without oscillating.
Free vibration, simulated
The program models a quarter of a car: 400 kg of body on a 25 kN/m spring, which sags 157 mm under its weight. It releases the body 50 mm above equilibrium (as if after a bump) and integrates the motion with fourth-order Runge–Kutta, for four damping ratios. It then checks the integrator against the exact solution and recovers from two successive peaks, exactly as a test engineer would from a measured decay.
import numpy as np
# A quarter-car model: 400 kg of car body on a 25 kN/m spring and a damper.
m, k = 400.0, 25e3
wn = np.sqrt(k / m) # natural frequency (rad/s)
print(f"natural frequency {wn:.2f} rad/s = {wn / (2 * np.pi):.2f} Hz, period {2 * np.pi / wn:.2f} s")
def free_response(zeta, x0=0.05, t_end=12.0, dt=1e-3):
"""Integrate m x'' + c x' + k x = 0 from x = x0, v = 0 with fourth-order Runge-Kutta."""
c = 2 * zeta * np.sqrt(k * m)
f = lambda s: np.array([s[1], -(c * s[1] + k * s[0]) / m])
s, out = np.array([x0, 0.0]), [x0]
for _ in range(int(round(t_end / dt))):
k1 = f(s); k2 = f(s + dt / 2 * k1); k3 = f(s + dt / 2 * k2); k4 = f(s + dt * k3)
s = s + dt / 6 * (k1 + 2 * k2 + 2 * k3 + k4)
out.append(s[0])
return np.array(out)
t = np.arange(0, 12.0 + 5e-4, 1e-3)
runs = {z: free_response(z) for z in (0.05, 0.3, 1.0, 2.0)}
print("\n t (s) " + "".join(f" zeta={z:<5}" for z in runs) + " (x in mm)")
for ti in [0.0, 0.2, 0.4, 0.6, 0.8, 1.0, 1.5, 2.0, 3.0]:
i = int(round(ti / 1e-3))
print(f"{ti:6.1f} " + "".join(f"{1000 * runs[z][i]:11.2f}" for z in runs))
# Check the integrator against the exact underdamped solution, and estimate zeta from the peaks.
z = 0.05
wd = wn * np.sqrt(1 - z**2)
exact = 0.05 * np.exp(-z * wn * t) * (np.cos(wd * t) + z / np.sqrt(1 - z**2) * np.sin(wd * t))
print(f"\nzeta = 0.05: largest difference from the exact solution {abs(runs[z] - exact).max():.1e} m")
x = runs[z]
peaks = [i for i in range(1, len(x) - 1) if x[i - 1] < x[i] > x[i + 1]]
delta = np.log(x[peaks[0]] / x[peaks[1]]) # logarithmic decrement
print(f"first peaks {1000 * x[peaks[0]]:.2f} mm and {1000 * x[peaks[1]]:.2f} mm, "
f"decrement {delta:.4f}, so zeta = {delta / np.sqrt(4 * np.pi**2 + delta**2):.4f}")
# Settling: when does the motion stay within 1 mm of rest for good?
for z in runs:
outside = np.nonzero(abs(runs[z]) > 1e-3)[0]
print(f"zeta = {z:4}: settles within 1 mm after {t[outside[-1] + 1]:.2f} s")The body bounces at 1.26 Hz. With , like a car whose dampers have failed, it is still swinging by 23 mm after two seconds and takes 9.6 s to settle within a millimetre. With it swings through to about 19 mm below equilibrium, back to 7 mm above, and is within a millimetre of rest after 1.4 s. Critical damping returns it fastest without any overshoot, in 0.74 s, and heavy overdamping, , is sluggish again: 1.9 s. The two peaks of the lightly damped run give back , the value put in.
Car suspensions are typically set somewhere around –: enough damping to stop the bouncing quickly, little enough that the body is not jolted by every ripple in the road, which a stiffly damped suspension transmits. Door closers, the needles of analogue meters and the suspensions of measuring instruments are set nearer critical, because there an overshoot is the thing to avoid.
Forced vibration and resonance
Now drive the system with a steady harmonic force , from a rotating imbalance, a reciprocating engine, footsteps or wind. After the free vibration dies away, the response settles to a steady state at the forcing frequency.
Theorem— Steady-state response to harmonic forcing
With frequency ratio , the steady-state response is , with
The ratio is the magnification factor: the amplitude as a multiple of the deflection the same force would cause if applied slowly.
Proof
Write the force as the real part of and look for as the real part of . Substituting, . Divide by : . The modulus of the denominator gives the amplitude and its argument gives the phase lag .
∎
Three regimes are visible. Well below resonance () the spring dominates and the mass follows the force in phase, with . Well above it () inertia dominates: the mass cannot keep up, moves less and less, and lags by almost 180°. Near the spring and inertia forces cancel each other, and only the damper is left to resist the force: the amplitude reaches at , and the peak, slightly lower in frequency, is . The number is the quality factor, the same quantity that sets the sharpness of a tuned circuit in Lumenary EE's RLC circuits, because the equation is the same.
import numpy as np
def magnification(r, zeta):
"""Steady-state amplitude X divided by the static deflection F0/k, at frequency ratio r = w/wn."""
return 1 / np.sqrt((1 - r**2) ** 2 + (2 * zeta * r) ** 2)
# The frequency-response curve, as a table and a sideways bar chart for zeta = 0.1.
zetas = [0.05, 0.1, 0.3, 0.7]
print(" r " + "".join(f" zeta={z:<5}" for z in zetas) + " zeta = 0.1")
for r in [0.0, 0.5, 0.8, 0.9, 0.95, 1.0, 1.05, 1.1, 1.2, 1.5, 2.0, 3.0]:
row = "".join(f"{magnification(r, z):11.2f}" for z in zetas)
print(f"{r:5.2f}{row} " + "#" * int(round(4 * magnification(r, 0.1))))
r = np.linspace(0, 3, 300001)
for z in zetas:
M = magnification(r, z)
print(f"zeta = {z}: peak {M.max():6.2f} at r = {r[M.argmax()]:.3f}; at r = 1 it is 1/(2 zeta) = {1 / (2 * z):5.2f}")
# Switching on a force exactly at resonance: the response grows towards its steady amplitude.
m, k, zeta, F0 = 1.0, (2 * np.pi) ** 2, 0.05, 1.0 # natural frequency 1 Hz
wn = np.sqrt(k / m); c = 2 * zeta * m * wn
dt, x, v = 1e-3, 0.0, 0.0
print("\ncycle largest |x| during the cycle, as a multiple of F0/k")
for cycle in range(1, 41):
peak = 0.0
for i in range(1000): # 1000 steps of 1 ms = one period
t = (cycle - 1) + i * dt
a = (F0 * np.cos(wn * t) - c * v - k * x) / m
v += a * dt # semi-implicit Euler: velocity first,
x += v * dt # then position with the new velocity
peak = max(peak, abs(x))
if cycle in (1, 2, 5, 10, 20, 40):
print(f"{cycle:5d} {peak * k / F0:8.2f}")The table shows how narrow the danger zone is with light damping: at the response is ten times static at , but only 2.7 times at and 2.2 times at . The second part shows why a resonance takes time to build. Switched on from rest at exactly the natural frequency, the amplitude grows by about static deflections per cycle at first, reaches 95 per cent of its final value after about ten cycles and settles at ten times static. A machine that passes quickly through a resonance while it speeds up may never reach the steady-state amplitude; one that runs at resonance always will.
A common source of forcing is rotating imbalance: a mass at radius on a rotor turning at shakes its machine with a force , growing with the square of the speed. That is the force a washing machine's drum full of wet clothes generates on its spin cycle, why its suspension and its controller try to redistribute the load before spinning up, and why rotors are balanced, as the previous lesson described.
Vibration isolation
Put a running machine on springs and ask how much of its shaking force reaches the floor; or put a delicate instrument on springs and ask how much of the floor's shaking reaches it. The two questions have the same answer. The force passed through the spring and damper to the floor is , so its amplitude divided by is the transmissibility
At it equals 1 for every damping ratio, as the right-hand graph above shows. Below that, the mounts make things worse; above it, they isolate. So an isolator must make the natural frequency low, well below the forcing frequency, which means soft mounts with a large static deflection. And above , damping makes isolation worse, because the damper becomes a stiff path for high-frequency force; yet some damping is needed to limit the motion while the machine passes through resonance on starting and stopping.
import numpy as np
def transmissibility(r, zeta):
"""Force transmitted to the floor divided by the force the machine generates (or, equally,
the machine's motion divided by the floor's motion), at frequency ratio r = w/wn."""
with np.errstate(divide="ignore"):
return np.sqrt(1 + (2 * zeta * r) ** 2) / np.sqrt((1 - r**2) ** 2 + (2 * zeta * r) ** 2)
print(" r " + "".join(f"zeta={z:<6}" for z in (0.0, 0.05, 0.2, 0.5)))
for r in [0.5, 1.0, 1.2, np.sqrt(2), 2.0, 3.0, 4.0, 6.0]:
print(f"{r:5.2f} " + "".join(f"{transmissibility(r, z):10.3f}" for z in (0.0, 0.05, 0.2, 0.5)))
# Design: a 200 kg compressor running at 1480 rpm stands on four rubber mounts (zeta about 0.05).
# Aim: transmit at most 10 per cent of its shaking force to the floor.
m, rpm, zeta, target = 200.0, 1480.0, 0.05, 0.10
f = rpm / 60 # forcing frequency (Hz)
r_lo, r_hi = np.sqrt(2), 20.0 # transmissibility falls steadily beyond sqrt(2)
for _ in range(60): # bisection for transmissibility(r) = target
r_mid = (r_lo + r_hi) / 2
r_lo, r_hi = (r_mid, r_hi) if transmissibility(r_mid, zeta) > target else (r_lo, r_mid)
fn = f / r_hi # highest natural frequency allowed (Hz)
k_total = m * (2 * np.pi * fn) ** 2
print(f"\nforcing {f:.1f} Hz; need r >= {r_hi:.2f}, so fn <= {fn:.2f} Hz")
print(f"total stiffness <= {k_total / 1e3:.0f} kN/m, i.e. {k_total / 4e3:.0f} kN/m per mount")
print(f"static deflection under the machine's weight: {1000 * m * 9.81 / k_total:.1f} mm")
print(f"while starting up through resonance (r = 1), transmissibility {transmissibility(1.0, zeta):.1f}")The compressor's 24.7 Hz forcing needs a natural frequency below 7.26 Hz: four mounts of at most 104 kN/m each, which sag 4.7 mm under the machine. A manufacturer's catalogue of rubber or spring mounts is organised by exactly these two numbers, load per mount and static deflection. During start-up, though, the machine passes through 7 Hz, where the same mounts amplify its force tenfold; it must accelerate through quickly, and its pipes and cables need flexible connections so that they do not short-circuit the springs.
The same reasoning protects things from their surroundings: electron microscopes and optical tables float on air springs with natural frequencies near 1–2 Hz, and whole buildings in earthquake zones, from Japan and New Zealand to Chile and Türkiye, stand on base isolators, laminated rubber bearings or sliding pendulum bearings that lengthen the building's period beyond the strongest shaking of an earthquake.
Two degrees of freedom: modes and the tuned mass damper
Most structures need more than one coordinate. A two-storey frame has two floors, each able to sway; a car body can bounce and pitch. With masses and coordinates , Newton's law for each mass gives two coupled equations, which in matrix form read
for the frame, where is the stiffness of the ground storey and of the upper one.
Definition— natural modes
A natural mode is a pattern of motion in which every coordinate moves at the same frequency, in or out of phase, with fixed proportions . Substituting gives the generalised eigenvalue problem , which has non-trivial solutions only when . An -degree-of-freedom system has natural frequencies and mode shapes .
Because and are symmetric (and positive definite), the mode shapes are orthogonal in the sense for , a consequence of the spectral theorem of linear algebra. In the coordinates of the modes the equations decouple into independent single-degree-of-freedom oscillators, so everything above applies to each mode separately. That is modal analysis, and it is how every structural dynamics code works.
The program finds the two modes of this frame with NumPy's symmetric eigenvalue solver, after a change of variables that turns the generalised problem into an ordinary symmetric one. Its second half adds a tuned mass damper to a tower modelled by its first mode: a second, small mass on a spring and damper, tuned to almost the same frequency as the structure, so that the pair becomes a two-degree-of-freedom system.
import numpy as np
# A two-storey frame: floor masses m1, m2 (kg); storey stiffnesses k1 (ground to floor 1) and k2 (N/m).
m1, m2 = 20e3, 15e3
k1, k2 = 30e6, 20e6
M = np.diag([m1, m2])
K = np.array([[k1 + k2, -k2],
[-k2, k2]])
# Free vibration x = phi cos(w t) needs (K - w^2 M) phi = 0: a generalised eigenvalue problem.
# With L = M^(-1/2), the symmetric matrix L K L has eigenvalues w^2 and eigenvectors L^(-1) phi.
L = np.diag(1 / np.sqrt(np.diag(M)))
w2, V = np.linalg.eigh(L @ K @ L)
modes = L @ V # mode shapes, one per column
for i in range(2):
phi = modes[:, i] / modes[-1, i] # scale so the top floor moves 1
print(f"mode {i + 1}: {np.sqrt(w2[i]) / (2 * np.pi):5.2f} Hz, shape (floor 1, floor 2) = "
f"({phi[0]:+.3f}, {phi[1]:+.3f})")
print("orthogonal through the mass matrix: phi1' M phi2 =", f"{abs(modes[:, 0] @ M @ modes[:, 1]):.1e}")
# A tuned mass damper on a tower, as a one-mode model (per unit modal mass, natural frequency 1).
mu, zeta_s = 0.02, 0.01 # damper mass ratio; the tower's own damping
f_opt = 1 / (1 + mu) # Den Hartog's tuning: damper frequency ratio
zeta_d = np.sqrt(3 * mu / (8 * (1 + mu) ** 3)) # and damper damping ratio
kd, cd = mu * f_opt**2, 2 * zeta_d * mu * f_opt
Md = np.diag([1.0, mu])
Kd = np.array([[1 + kd, -kd], [-kd, kd]])
Cd = np.array([[2 * zeta_s + cd, -cd], [-cd, cd]])
w = np.linspace(0.7, 1.3, 6001)
with_tmd = [abs(np.linalg.solve(Kd - x**2 * Md + 1j * x * Cd, [1.0, 0.0])[0]) for x in w]
without = 1 / np.sqrt((1 - w**2) ** 2 + (2 * zeta_s * w) ** 2)
print(f"\ndamper: {100 * mu:.0f}% of the modal mass, tuned to {f_opt:.3f} of the tower's frequency, "
f"damping ratio {zeta_d:.3f}")
print(f"peak response of the tower: {without.max():.1f} times static without the damper, "
f"{max(with_tmd):.1f} with it")
print(" w/wn without with")
for x in [0.90, 0.94, 0.96, 0.98, 1.00, 1.02, 1.04, 1.08]:
i = int(round((x - 0.7) / 0.0001))
print(f"{x:6.2f} {without[i]:8.1f} {with_tmd[i]:7.1f}")The frame's first mode, at 3.97 Hz, sways both floors the same way, the top one nearly twice as far; the second, at 9.02 Hz, moves them in opposition. The second half is the reason tall buildings carry heavy pendulums near their tops. A tower with 1 per cent damping responds to wind gusts near its natural frequency at up to 50 times the static deflection. A damper with only 2 per cent of the tower's modal mass, tuned as Den Hartog's classical formulas prescribe, cuts that peak to 8.7, as if the tower's own damping had risen to about 6 per cent. It works by resonating against the structure: near the tuned frequency the damper's mass swings out of phase with the tower, its spring pulls back on the tower, and its damper turns the energy into heat.
The table shows how: near and the tuned damper splits the single sharp peak into two broad, low ones, with the response nearly flat across the band between them. That flatness is what Den Hartog's tuning achieves, and it makes the damper forgiving of a structure whose frequency is not known exactly or changes as the building is fitted out.
Taipei 101, completed in 2004, hangs a steel sphere of about 660 tonnes in its upper floors as a tuned mass damper against wind sway, and visitors can see it swing; many other towers use tuned sloshing water tanks or active dampers. London's Millennium Bridge reopened in 2002 after being fitted with viscous dampers and dozens of tuned mass dampers. On a smaller scale, the dumbbell-shaped Stockbridge dampers clamped near the towers of overhead power lines are tuned absorbers against wind-excited vibration of the conductors, and the torsional vibration dampers on car crankshafts apply the same idea to twisting vibration.
From strings to skyscrapers
Pythagorean musicians knew that a string's pitch depends on its length and tension, and Galileo noticed that a pendulum's period depends on its length but hardly on its swing. Robert Hooke published his law of the spring in 1678, and in the eighteenth century Daniel Bernoulli, Leonhard Euler and Joseph-Louis Lagrange solved the vibrating string and beam and found the idea of superposing modes. Lord Rayleigh's The Theory of Sound (1877) gathered the theory of vibration into its modern form, including the energy method that still estimates natural frequencies. Engineering practice caught up when fast machinery began to shake itself apart: Hermann Frahm patented the dynamic vibration absorber in 1909 to quieten ships, J. Ormondroyd and Jacob Pieter Den Hartog worked out its optimal tuning in 1928, and Den Hartog's Mechanical Vibrations (1934) taught generations of engineers. The Tacoma Narrows Bridge collapse in 1940 forced bridge engineers to treat wind and structure together, and from the 1970s fast Fourier transform analysers and finite element programs turned modal testing and modal analysis into everyday tools.
Vibration engineering in practice
Modal testing. A structure's real natural frequencies, mode shapes and damping are measured by exciting it, with an instrumented impact hammer or an electrodynamic shaker, and recording its response with accelerometers. Software fits the measured frequency-response functions with a model like the ones above. Car bodies, aircraft wings, satellites before launch and machine-tool frames are all tested this way, and the results are used to correct finite element models.
Modal analysis by finite elements. Before anything is built, its modes are computed by the finite element method, which you will meet in The finite element method: the same eigenvalue problem, , with millions of degrees of freedom. Open-source codes such as CalculiX and EDF's Code_Aster do it, as do commercial packages. Designers then keep natural frequencies away from known forcing frequencies: an engine's firing rate, a rotor's running speed and its multiples, the blade-passing frequency of a fan, the pace of walking (about 2 Hz vertically and 1 Hz sideways, which is what caught out the Millennium Bridge).
Condition monitoring. A machine's vibration signature changes when a bearing wears, a shaft cracks or a coupling loosens. Power stations, factories and wind farms monitor vibration continuously and judge it against severity criteria such as those in the ISO 20816 series published by ISO.
People. Long exposure to vibration harms people: hand-arm vibration from grinders, breakers and chainsaws damages nerves and blood vessels, and whole-body vibration in vehicles affects the spine. Workplaces limit daily exposure, choose low-vibration tools and maintain them, following guidance such as that of the UK's Health and Safety Executive and the ISO 2631 series. The most comfortable vehicles are tuned so that the body's natural frequencies sit away from 4–8 Hz, the band in which people are most sensitive to vertical vibration.
Control. Where passive springs and dampers are not enough, actuators push back against the vibration under feedback control: active engine mounts, semi-active car suspensions whose damping changes in milliseconds, and active mass dampers in some buildings. That meets Lumenary EE's feedback control and the control of mechanical systems later in Lumenary ME, in Control of mechanical systems.
Optional Christian reflection · A small helm
A reflection from the Christian tradition that inspires Lumenary Academy. It is not part of the mathematics and you are welcome to skip it.
The letter of James observes that ships, "though they be so great, and are driven of fierce winds, yet are they turned about with a very small helm" (James 3:4, King James Version). A tuned mass damper is a small helm of the same kind: two per cent of a tower's mass, placed and tuned with care, steadies the whole of it. Small things rightly ordered can hold steady what is far larger than themselves. This reflection is optional, and nothing in the lesson depends on it.
Common misconceptions
Misconception
"More damping is always better." Near resonance, yes. But for vibration isolation above , extra damping passes more force to the floor, and in a car, heavy damping transmits every small bump to the passengers. Damping is a design choice, not a quantity to maximise.
Misconception
"Making a structure stiffer always reduces its vibration." Stiffening raises the natural frequency, which helps if the forcing is below it, but can move the structure into resonance with forcing that used to be above it. The rule is to separate natural frequencies from forcing frequencies, in whichever direction is practical.
Misconception
"The Tacoma Narrows Bridge fell because the wind matched its natural frequency." The 1940 collapse was not simple forced resonance. It was aeroelastic flutter: the bridge's twisting motion changed the airflow so that the wind fed energy into the motion, a self-excited instability, in effect negative damping. Modern long-span bridges are shaped and tested in wind tunnels to prevent it.
Misconception
"A tuned mass damper works because it is heavy." It is small, typically a few per cent of the modal mass. It works because it is tuned: its frequency is set close to the structure's, so that it moves out of phase with it, and its damping is set to spread and flatten the response. Mistuned, the same mass does little.
Exercises
Use m/s² throughout. Two exercises ask for Python functions; run them in the editor before submitting.
1. Vibrations
Warm-upHumanA machine on springs vibrates at a natural frequency of 12 Hz. It is rebuilt with four times the mass on the same springs. What is its new natural frequency?
2. Vibrations
Warm-upHumanA motor on spring mounts sinks 10 mm when it is set down on them. Estimate its natural frequency of vertical vibration in cycles per minute (the unit machinery speeds are quoted in), to the nearest whole number.
3. Vibrations
CoreHumanA tap test on a steel footbridge records a decaying vibration. Two successive positive peaks, one period apart, measure 12.0 mm and 7.5 mm. Estimate the damping ratio . Give 100 × ζ, rounded to the nearest whole number.
4. Vibrations
CoreHumanA pump runs at three times the natural frequency of its mounts (), and the mounts have a damping ratio of 0.05. What percentage of the pump's shaking force reaches the floor? Round to the nearest whole per cent.
5. Vibrations
CoreHumanA fan whose imbalance shakes it at 30 Hz stands on rubber mounts (damping ratio about 0.05) that give it a natural frequency of 25 Hz. The floor vibrates noticeably. Which single change will reduce the force reaching the floor the most?
6. Vibrations
CoreComputationalWrite natural_frequencies(M, K) for an undamped system with symmetric (positive definite) and . Return (freqs, modes): the natural frequencies in hertz, ascending, and a matrix whose columns are the mode shapes, scaled so that (mass-normalised).
The starter makes a classic mistake. Run the tests to see it fail, then fix it.
7. Vibrations
StretchComputationalWrite isolator_stiffness(m, f_forcing, tr_max, zeta=0.0): the largest total stiffness (N/m) of the mounts under a machine of mass m (kg), shaking at f_forcing (Hz), such that at most the fraction tr_max (less than 1) of its shaking force reaches the floor, given the mounts' damping ratio zeta. A transmissibility(r, zeta) function is provided.
Further reading and resources
- MIT OpenCourseWare 2.003SC, Engineering Dynamics · Massachusetts Institute of TechnologyIts later lectures cover single- and multi-degree-of-freedom vibration, with recitation problems.
- NumPy: numpy.linalg.eighThe symmetric eigenvalue solver used for the mode shapes here.
- Code_Aster · EDFAn open-source finite element code from the French electricity utility, widely used for structural dynamics and modal analysis.
- ISO, the International Organization for StandardizationPublishes the ISO 20816 series on measuring and evaluating machine vibration and ISO 2631 on human exposure to whole-body vibration.
- Health and Safety Executive (UK): vibration at workReadable guidance on hand-arm and whole-body vibration from power tools and vehicles.
- Mechanical Vibrations · J. P. Den Hartog · 1934The classic engineering text, including the theory of the tuned vibration absorber; later editions are reprinted by Dover.