Equipment resonance: signs, how to check, and what to do

A fan starts humming after cleaning, the vibration meter shows 10 mm/s. You balance the impeller, the vibration drops, and two weeks later it's back. In a situation like this, the culprit is often not the rotor but resonance in the supports and frame. Below, we'll cover how to identify it from a single coast-down — the machine's free run-down after the power is cut — and what to do about it.

Updated 27 August 2026 · by AXILINE · Vila Nova de Gaia

What resonance is, and why it blows vibration out of proportion

Every structure has a natural frequency. For a simple system it's estimated as fₙ = (1/2π)·√(k/m), where k is stiffness and m is mass. The logic is simple. Weld a diagonal strut to the frame, and the stiffness goes up, so the natural frequency moves up too. Hang a heavy guard off a support, and the frequency comes down.

Resonance appears when the exciting frequency coincides with fₙ. On a rotating machine, the source of excitation is usually the residual imbalance at the rotation frequency (1x). The force stays the same, but the response grows by a factor of Q — that's how much resonance amplifies the vibration. The Q factor is calculated as Q = 1/(2ζ). For a welded steel frame, the damping ratio ζ is around 0.05, which puts Q at around 10. The same imbalance that nobody would notice at other speeds produces ten times the vibration here.

There's a practical takeaway here. High vibration at 1x doesn't prove imbalance. It proves that there's a force at the rotation frequency, and a response to it. Resonance inflates the second part, not the first.

Keep two different phenomena separate. Structural resonance is the natural frequency of the stationary parts: the bearing housing, the frame, the foundation, a platform, piping. Critical speed is resonance of the shaft itself. A rigid rotor operates below the first critical speed; a flexible one passes through it. For a fan, pump, screw conveyor, or crusher on a plant floor, you're almost always dealing with structural resonance.

No instrument prints the word “resonance” on its own line. Resonance is a conclusion you draw from how the amplitude and phase behave as speed changes.

Five signs that give resonance away

Where resonance comes from

Resonance doesn't appear on its own. It appears when someone has changed the system's stiffness or mass. Here are the usual culprits.

• Loose fasteners. Foundation bolts that aren't torqued down reduce the support's stiffness, and the natural frequency drifts down into the operating speed range.
• Soft foot. One foot isn't touching the base, or it's sitting on a warped shim. The support behaves like a spring.
• A flexible or modified frame. A welded frame with no diagonal bracing, a reinforcement that's been cut out, the machine relocated onto a new base, shims stacked under the feet.
• Cracks and corrosion. A cracked weld, deteriorated concrete around an anchor, a rusted-through embedded plate. Stiffness drops while the geometry looks unchanged from the outside.
• A change in mass. A new impeller in place of the old one, a different pulley, a guard, added piping.
• A change in speed. A variable-frequency drive with a new setpoint is a common cause. A machine ran at 1450 rpm for years, then at 1180 rpm it landed right in the frame's resonance.
• The bearing housing itself. Worn-out seating surfaces, a loosened fit of the bearing housing.
• Sagged rubber vibration mounts. Their stiffness has changed, and the designed-in frequency separation has disappeared.

If vibration went up right after a repair, cleaning, an impeller replacement, or a relocation, check for resonance first. You changed the mass or the stiffness, and the system moved to a different point.

How to check: speed, coast-down, bump test

1. 01

   Take the baseline figures

   At operating speed, measure the overall vibration (the total level across all frequencies) and 1x at each bearing housing, in three directions. Record the 1x phase. Mount the sensor rigidly on the bearing housing, with a magnet on a clean, flat surface. Keep the direction the same from reading to reading. The machine's overall condition is assessed from the overall vibration in mm/s RMS (root mean square) over the 10-1000 Hz band, by zones A/B/C/D — from A “good” to D “not acceptable.” The zone boundaries depend on the machine group and the applicable part and edition of ISO 20816, so check them separately, especially for small fans and belt-driven units.

2. 02

   Sweep the speed upward

   If there's a variable-frequency drive, raise the frequency in 5% steps, recording the 1x amplitude and phase at each step. You're not looking for the maximum — you're looking for the shape of the curve: where it spikes, and where the phase starts to shift. Hold each point until the readings settle.

3. 03

   Record the coast-down

   Without a drive, the coast-down does the job. Cut the power and record the vibration until the rotor stops. The machine sweeps through every speed from top to bottom on its own, and resonance shows up as a hump in the recording. On the Balanset-1A, switch on “Diagrams” mode, select overall vibration, and record for up to 20 seconds on both channels at once. On a large fan the coast-down runs for minutes, so 20 seconds only gives you a rough picture: make several consecutive recordings, and take the exact frequency value from a bump test.

4. 04

   Strike the stopped machine

   A bump test answers the question directly. Stop and de-energize the machine, leave the sensor where it is, strike the support or frame near the sensor with a soft mallet, and look at the response spectrum. The structure rings at its own natural frequencies, and you read them off the spectrum. Conditions: the machine is stopped, the strike is a single clean hit with no “double tap,” the tip is soft, and averaging is switched off. Averaging will smear a single impulse into noise.

5. 05

   Calculate the margin

   Convert the rotation frequency to hertz by dividing by 60. At 1450 rpm that's 24.2 Hz. A working guide for frequency separation — the gap between the exciting and natural frequencies — is |f_exc − fₙ| / fₙ greater than 15-20%. That means a natural frequency below 20 Hz or above 29 Hz is fine, while landing in the 22-26 Hz range means a problem. Check the blade-pass frequency too — speed multiplied by the number of blades: for a six-blade impeller at 1450 rpm that's 145 Hz, and it has its own set of coincidences with the natural frequencies of the guard and the ductwork.

A bump test on a running machine won't give you an answer. The response to the strike mixes with the operating vibration, and you won't be able to separate out the natural frequency.

Resonance, imbalance, or looseness: how to tell them apart

Imbalance and misalignment often produce equally high 1x. Without phase, you can't tell them apart, so measure phase at both bearings, not just amplitude.

Why you can't balance in resonance

Balancing works through the influence coefficient. The instrument measures the baseline vibration, you fit a trial weight of known mass at a known radius, the instrument measures again. The difference shows how the system responds to the added mass. From there, the software calculates the correction mass and angle.

All of that arithmetic rests on one assumption: the system is linear and stable. The response is proportional to the force, and the influence coefficient doesn't change between runs. In resonance, that assumption falls apart. Here's what you get in practice.

• The influence coefficient comes out enormous, because a small trial weight produces a large change in vibration. The software will calculate a tiny correction, and it won't work.
• Phase is steep against speed in resonance. The speed drifts by 1%, and the phase shifts by tens of degrees. The weight ends up somewhere other than where it needs to be.
• A verification run with the same weight gives different numbers. You keep repeating trim balancing over and over and never reach tolerance.
• Sometimes you do land within norm anyway. But balancing only removes the exciting force, and the Q factor hasn't gone anywhere. A little wear, dust building up on the blades, or a weight shifting, and the vibration is back within a week.

The tell that you're fighting resonance rather than imbalance: balancing “works” every time, but it doesn't hold. If a rotor needs touching up every month, the problem isn't the rotor.

What to do: detuning by stiffness, mass, and speed

Resonance is removed in three ways: change the stiffness, change the mass, or change the speed. A fourth option is adding damping — absorbing the vibration energy. That doesn't shift the frequency, it knocks down the height of the peak.

The relationship fₙ_new = fₙ·√((k_new/k_old)·(m_old/m_new)) helps estimate the shift. Stiffness and mass sit under the square root. To raise the natural frequency by 20%, you need to increase the stiffness by roughly a factor of 1.44. One decorative gusset plate won't get you there.

Where this approach doesn't work. Resonance of a flexible rotor at its critical speed can't be removed by detuning the supports. That calls for a rotordynamics calculation and, possibly, flexible-rotor balancing across several speeds. If your machine runs above its first critical speed, that's a separate undertaking, and we'll tell you plainly when it falls outside the scope of a site visit.

Where to start on your own site

The Balanset-1A gives you everything needed for this kind of check. The kit includes two accelerometers for the bearing housings, a laser phase sensor for the shaft's reflective mark, a two-channel USB module, and Windows software. The instrument shows speed, overall vibration, and 1x with amplitude and phase, builds the FFT spectrum, records the time waveform on both channels at once, and stores readings and spectra in an archive. You build the report from that archive. The same instrument calculates the trial weight, the correction, tolerance against balance quality grade G, and splits the weight across fixed positions, once it comes down to actually balancing.

AXILINE comes to the site anywhere in Portugal. We measure the vibration, separate resonance from imbalance and shaft misalignment, balance the rotor in its own bearings when balancing genuinely solves the problem, and tell you plainly when work on the frame or foundation is needed instead. The Balanset instruments are designed, manufactured, and used on site visits by the same engineers who provide consulting support on them.

Tell us what the machine is, what speed it runs at, and what figures you're seeing at the bearings. We'll tell you whether it looks like resonance or not, and what to check first. If you'd rather work on it yourself, we'll supply the instrument and help you get up to speed with it.

• Compare the overall vibration and 1x at each bearing housing. If the overall level is several times higher than 1x, imbalance isn't the main cause.
• Measure in three directions and compare the figures. Resonance singles out one direction.
• Find out what changed before the vibration appeared: a repair, a new impeller, a drive setpoint, relocating the unit.
• Torque the fasteners and check for soft foot before any diagnostics.
• Record the coast-down or sweep the speed in steps, recording the 1x amplitude and phase.
• Confirm the natural frequency with a bump test on the stopped, de-energized machine.
• Calculate the detuning margin. The guide is more than 15-20% of fₙ.
• Balance only after confirming the operating speed is outside the resonant zone.

Sources: ISO 20816-1:2016 · ISO 13373-3:2015 · ISO 13373-5:2020 · ISO 21940-11:2016 · ISO 21940-12:2016 · Balanset-1A operation manual · Balanset-1A manufacturer specification

Frequently asked questions

Related content