Every field balancer eventually meets the session from hell: the procedure was followed, the weights went exactly where the software said, and the vibration did not drop. Sometimes it got worse. The reflex is to blame the instrument and run the procedure again — and again. The actual fix is almost never another run; it is finding which assumption of the balancing method your machine is violating.
This guide is the diagnostic tree for that moment. It assumes you know the basic procedure (if not, start with the fan or mulcher walk-throughs).
First: is the problem even imbalance?
Balancing corrects exactly one defect — uneven mass distribution. Its signature is specific: vibration dominated by the 1× component (once per revolution), with a stable phase, driven by a centrifugal force that grows with the square of speed. Before any weights, take a minute in F5 Vibration Meter and in the spectrum view (F8 Charts → F5-Spectrum (Hz)):
| What you see | What it suggests |
|---|---|
| Dominant 1×, stable phase | Imbalance — balance it |
| Strong 2× (and axial vibration) | Shaft misalignment — align the coupling, then re-measure |
| Harmonics series (1×, 2×, 3×…), clipped waveform | Mechanical looseness — find and torque the joint |
| Non-synchronous peaks (not whole multiples of 1×), or a raised broadband floor | Bearing wear — replace, then balance if needed |
| Peak at blade-pass frequency — or at gear mesh, when the mesh frequency falls inside the band | Aerodynamic/gear sources — not a balancing problem |
A machine can carry two defects at once — a worn bearing and imbalance. Fix the structural defect first: weights cannot compensate a bearing, and the bearing’s noise corrupts the balancing measurement. Keep the instrument’s scope in mind while you read the table: it measures vibration velocity in mm/s RMS over a 5–1000 Hz band and has no envelope or demodulation mode, so it will show you an advanced bearing fault but not an early one. That band also caps the last row — blade-pass on a fan or pump normally lands inside it, but gear mesh (shaft speed × tooth count) is often above 1000 Hz and simply will not appear.
Unstable readings: the session-killer
The influence-coefficient method has one hard requirement: repeatability. The same rotor state must produce the same amplitude and phase. If "Run 0 again" gives a different answer than Run 0, no calculation can succeed. When readings float, walk this list in order:
- Speed. Is every run at the same rpm? A drifting drive (PTO, hydraulics, a loaded motor) changes both amplitude and phase. Hold the speed; let readings settle before recording.
- Tachometer. Did the stand get nudged? Is the reflective mark clean and unique (one mark only)? Is direct sunlight or a work lamp hitting the optics? Any of these scrambles the phase reference.
- Sensors. Flat, clean, paint-free mounting spot; magnet seated firmly; cable not slapping against the machine; sensor not touching a moving part.
- Loose mass. Liquid or debris inside the rotor (sand in a mulcher tube, fluid in a hollow shaft) relocates on every start — the phase wanders run to run. Repeat Run 0 two or three times: a phase that won’t repeat with nothing else changed is the classic symptom.
- Resonance. If the working speed sits near a structural resonance, tiny speed differences produce huge reading swings. Shift the test speed 10–15% and watch: if amplitude collapses or phase jumps, you are balancing on a resonance — move the test speed off it, or stiffen/isolate the structure. A bump test on the stopped machine, or a rundown record, shows where the natural frequency actually sits.
- Thermal drift. Some machines need to reach working temperature before behaving repeatably. Balance warm if the machine runs warm.
The weights went on, vibration went up
Three causes account for nearly all of these:
-
Angle counted the wrong way. The correction angle is measured from the trial-weight position, in the direction of rotation. Counting against rotation mirrors the position: the weight lands 2θ away from where the math put it, so part of the correction is thrown away — and when the angle is near a quarter turn, the mirrored weight sits roughly opposite the correct spot and makes the vibration worse instead of better. This is the single most common field error.
-
The machine changed between runs. A support placed under the frame, a removed guard, loosened then retightened mounts, a knife replaced mid-session — any change of mass or stiffness invalidates the influence coefficients measured two runs ago. Change something structural → start the session over.
-
Trial weight too small. A trial run counts as a valid response only if it changes the amplitude by at least 20–30% or shifts the phase by at least 20–30°. If Run 1 did neither, the influence coefficient is computed from noise and the correction it produces is fiction. Go heavier — but size the next weight, don’t guess it. The rule is a trial mass whose centrifugal force is a fraction k of the rotor’s weight: m = k·M·g/(R·ω²), with M the rotor mass in kilograms, R the mounting radius in metres, ω the rotational speed in radians per second (ω = 2π·n/60 for n in rpm) and g = 9.81 m/s². Fed those SI units the formula returns kilograms — multiply by 1000 to get grams. Worked through: a 100 kg rotor at 1,500 rpm with the weight at R = 0.2 m and k = 0.07 gives ω = 157 rad/s and m = 0.0139 kg, i.e. about 14 g. Usable k runs from 0.05 to 0.10; if your first weight sat at the low end of that band, step up toward 0.10 and stop there. The upper end is a safety limit, not a formality — "just double it" walks straight past it, and the trial weight’s own centrifugal force is what tears fixings off. Then redo the run.
If you are working in the modern unified balancing form, the First trial weight estimate window does this arithmetic for you: type in the rotor mass and mounting radius, record Run 0, and it returns a mass in grams plus an installation angle. Mind the datum: that angle is counted from the tacho mark, not from the trial-weight position the correction angle above uses, and it only means anything if the sensors are mounted coaxially with the tachometer. Two further caveats. It is not in the classic balancing form that a fresh install and any older firmware open, so on most machines you do the arithmetic yourself. And k = 0.07 is what it applies on rigid supports — on the default soft-support setting it estimates from the measured amplitude instead and uses the k = 0.07 result only as a ceiling.
Convergence that stalls
Vibration drops, but each trim gains less, and the level plateaus above target:
- You are balancing the residual of another defect. A bent shaft or eccentric seat produces genuine 1× vibration that mimics imbalance but cannot be fully weighted away. The tell: corrections converge to a stubborn floor. Check runout with an indicator.
- The weights are not where the math thinks. Welded weight placed "about there", or the correction radius entered wrong. Angle and radius precision matter — a few degrees off costs real percentage points.
- Weight attachment is moving. A clip-on weight that shifted, a weld that cracked. Re-verify mass positions before the next trim.
- One plane was enough — until it wasn’t. Wide rotors corrected in a single plane can hit a floor where the remaining vibration is a moment imbalance. Switch to F3 Two-plane and start the session over.
A useful rule from field practice: one trim run (Run T) is routine; three or more means stop and re-diagnose.
When it is genuinely not balanceable in place
Some situations are out of scope for field balancing, and recognizing them early saves a day: cracked rotor bodies (balancing a cracked rotor is dangerous, full stop), bearing seats hammered beyond repair, rotors operating through a resonance with no speed flexibility, and flexible rotors that must be balanced at several speeds on equipment built for it. In those cases the honest output of your session is a diagnosis, not a balance — and what you hand over is the before/after record the instrument saves for you (F6 Reports on the Balanset-1A), documenting the defect you found.
Two field write-ups from hard machines, read as job records rather than as step-by-step method references: a crusher drive shaft and a vibrating-screen motor. And if you are still choosing the instrument that does both the measuring and the explaining, start with Choosing a Field Balancer.