A machine that vibrates strongly at one speed but calms down at others usually has a resonance problem, not an unbalance problem. Adding correction weights to a resonant structure wastes time and rarely helps. This article shows you how to find natural frequencies on site with a bump test and how to confirm them with a coast-down measurement, so you can decide between stiffening, damping, detuning, or balancing.
Why resonance mimics unbalance
Unbalance creates a rotating force:
F = m · r · ω²
where m is the unbalance mass in kg, r is the distance from the shaft axis in m, and ω is the angular speed in rad/s. This force grows with the square of speed. A resonance amplifies whatever force excites it, so when the 1× running speed passes near a natural frequency, vibration rises sharply and falls again. The symptom looks like unbalance, but the cause is a structural natural frequency close to the operating speed.
The decision rule is simple: if 1× vibration is high across a narrow speed band and low outside it, suspect resonance. If 1× vibration climbs steadily with speed and phase stays stable, suspect unbalance. The Vibration Fault Spectrum Simulator shows how each fault signature differs.
ISO 21940-14 covers balancing machines and how they are proved; ISO 21940-11 sets balance tolerances for rotors. Both assume the rotor runs away from its critical speeds. Confirm the natural frequencies first, then use the ISO 21940-11 Balance Tolerance Calculator if balancing is the right fix.
The bump test: procedure and setup
A bump test (impact test) measures the natural frequencies of a structure at standstill. You strike the machine with an instrumented hammer or a heavy mallet and record the response with an accelerometer. The structure rings down at its natural frequencies, which appear as peaks in the response spectrum.
Follow this checklist on site:
- Stop and lock out the machine. The rotor must be at rest.
- Mount a stud-mounted or magnetically attached accelerometer on the bearing housing or the point of highest vibration. Keep the cable fixed so it does not flap.
- Set the analyzer to a frequency span at least 2 to 3 times the highest expected natural frequency. Use a low-frequency range (for example 0 to 500 Hz) for large slow machines.
- Strike the structure firmly and briefly with a soft-tip hammer. The impact must be short compared with the period of the highest frequency of interest.
- Record 3 to 5 averages. Check that the impact force stays consistent between hits; a double hit or a glancing blow corrupts the spectrum.
- Identify peaks in the response spectrum. Peaks with high amplitude and low damping (sharp, narrow peaks) matter most.
- Repeat at several measurement points and in horizontal, vertical, and axial directions. A structure can have different natural frequencies in each direction.
Compare the measured natural frequencies with the 1× running speed. A common rule: keep the separation margin between any natural frequency and the operating speed above roughly 15 to 20 percent. The exact margin depends on the machine class and damping; ISO 21940-21 discusses acceptable separation margins for rotors with flexible supports.
Reading the bump test result
A sharp peak marks a lightly damped natural frequency. Estimate the damping from the half-power bandwidth:
ζ = Δf / (2 · fn)
where ζ is the damping ratio, Δf is the frequency width in Hz between the two points where the peak amplitude falls to 1/√2 (about 70.7 percent) of its maximum, and fn is the natural frequency in Hz. A ζ below about 0.05 means the resonance amplifies vibration strongly and will be sensitive to speed changes.
Watch for these pitfalls:
- The hammer tip controls the excitation band. A soft tip excites low frequencies only; a hard tip reaches higher frequencies. Match the tip to the range you care about.
- A running machine nearby can inject its own frequencies into the spectrum. Bump-test results taken during nearby operation need careful reading.
- Nonlinear supports (loose feet, cracked foundations) shift natural frequencies with impact force. If the peak moves between light and hard hits, find and fix the looseness first.
The Free Vibration Simulator: Mass–Spring–Damper lets you see how damping ratio changes peak width and amplification.
Coast-down: confirming resonance at speed
A coast-down measurement records vibration amplitude and phase while the machine coasts from operating speed to a stop after the power is cut. It needs a phase reference: a tachometer or optical pickup on the shaft. As speed falls, a plot of 1× amplitude against speed shows a peak wherever the speed crosses a natural frequency. This peak is the critical speed.
Phase behavior confirms the diagnosis. Below a critical speed, the 1× phase response lags the unbalance force by a small angle. Passing through the critical speed, phase shifts by roughly 180 degrees. If your coast-down plot shows an amplitude peak together with a phase shift of about 180 degrees, you have found a true resonance, not just a speed-dependent unbalance response.
Coast-down also distinguishes resonance from other faults:
- Resonance: amplitude peaks at one speed, phase shifts through that speed, amplitude drops on both sides.
- Rub: amplitude stays high to low speed, spectrum shows harmonics and subharmonics.
- Looseness: amplitude may drop with speed but the spectrum shows many harmonics of 1×.
Record coast-down data at a sample rate high enough to resolve the speed sweep, and log speed from the tachometer simultaneously. The Time Waveform and FFT Spectrum Simulator helps you practice reading spectra at different speeds before you collect real data.
Run-up and coast-down with limited instrumentation
If you lack a phase reference, a coast-down amplitude plot alone still shows where natural frequencies lie: any sharp amplitude peak versus speed marks a resonance crossing. You lose the phase confirmation, so cross-check with a bump test at the same frequency.
If you cannot shut the machine down, a run-up (starting from rest and recording during acceleration) gives the same information in reverse. Accelerate slowly through the suspected band. A fast sweep can miss a narrow resonance because the structure has no time to build up amplitude; the slower the speed change, the closer the measured peak comes to the true natural frequency.
For machines with variable speed drives, step the speed in small increments, hold each step for several seconds, and record the settled amplitude at each step. This produces a clean amplitude-versus-speed curve without a full coast-down.
Deciding what to do with the result
Once you know a natural frequency sits near the operating speed, you have four options:
- Stiffen. Raising the natural frequency moves it away from 1×. Tighten foundation bolts, grout the baseplate, add bracing. Check bolt torque first; loose feet alone can drop a natural frequency into the running band.
- Add damping. Damping lowers the amplification at resonance without moving the frequency. Viscoelastic pads or tuned mass dampers work where stiffening is impractical.
- Detune by speed. If the drive allows it, move the operating speed away from the critical speed by more than the separation margin.
- Balance carefully. Balancing helps only if the residual unbalance force drives the resonance. A well-balanced rotor excites a resonance far less. Use the Single-Plane Balancing Calculator or Two-Plane Balancing Calculator only after you have confirmed the resonance is driven by 1× unbalance and the phase is stable.
ISO 10816-3 (now superseded by ISO 20816-3 for the same machine groups) gives vibration severity zones for machines up to 300 kW on rigid and flexible foundations. Use it to judge whether the vibration level at the operating speed is acceptable after your fix. The ISO 10816-3 Vibration Severity Evaluator applies the zones directly. For a structured condition-monitoring program that catches resonance drift early, ISO 17359 and ISO 13373 define the measurement and diagnostic framework, and ISO 18436 covers personnel certification for the analysts doing the work.
When to call for an on-site measurement
Call in a specialist when any of these apply:
- The bump test shows a natural frequency within the separation margin and you cannot change speed or stiffness yourself.
- Coast-down phase data is ambiguous, for example with two close natural frequencies or a heavily damped peak.
- Vibration exceeds the ISO 10816-3 / ISO 20816-3 alarm zone and the cause is still unclear after bump and coast-down tests.
- The machine trips on vibration during start-up or shutdown, which points to a critical speed crossing that needs measured confirmation before any modification.
Our field balancing team brings multi-channel 24-bit DAQ equipment for bump tests, coast-downs, and operating deflection shape measurements, and works to ISO 21940 and ISO 20816 procedures. If you want to screen the fault first, start with the Diagnosis Wizard and bring the results to the discussion.
