Why Balancing Doesn't Reduce Vibration: 9 Causes and How to Fix Each One

The Core Problem: Balancing Fixes Exactly One Thing

Balancing corrects mass asymmetry in a rotating part. That's it. The rotor's center of mass doesn't coincide with its rotation axis, so every revolution generates a centrifugal force that shakes the machine. Correction weights shift the mass center back to the axis. Vibration drops.

But vibration in rotating machinery has many sources, and unbalance is only one of them. This guide covers the nine that most often make a balancing job fail: resonance, mechanical looseness, misalignment, a bent shaft, bearing defects, a dirty rotor, an undersized trial weight, an angle error, and thermal distortion. Most of them produce vibration that looks like unbalance: it is synchronous or near-synchronous, it's periodic, and it shakes the machine in the radial direction. The frustrating part is that adding correction weights to a machine suffering from looseness or resonance doesn't just fail — it can make things worse.

The Balanset-1A is a balancer, but it's also a vibration analyzer with FFT spectrum analysis (F8 – Charts) and a vibration meter (F5 – Vibration Meter). These diagnostic tools are the key to identifying which of the nine causes you're actually dealing with — before you waste time on trial weights.

The "Fake Unbalance" — 5 Faults That Mimic It

Resonance: the trap that catches everyone at least once

Near resonance, the phase angle between the unbalance force and the vibration response shifts rapidly with tiny speed changes. If the machine runs at 1,480 RPM and the structural natural frequency is at 1,500 RPM, a 1% speed drift can swing the phase by roughly 10–35°, depending on how lightly the structure is damped (the lighter the damping, the sharper the swing). The balancing software sees a different angle every run and computes a different correction every time.

The first test is simple: open the Balanset-1A's Vibration Meter (F5 from the main screen), hold a constant speed and watch the phase — on screen, V1s= is the broadband RMS, V1o= the 1× component, and F1= is the 1× phase in degrees (not a function key). If the phase wanders more than 10–20° while the RPM readout is stable, you're near resonance. The Polar tab makes the drift visible: add each new reading as a point on the polar plot and watch the cluster smear. The fix is not more trial weights — it's either changing the operating speed (run at a different RPM) or modifying the structure's stiffness or mass to shift the natural frequency away from running speed.

The definitive test is a bump test or a coast-down, and the Balanset-1A has a built-in Bump Test mode: strike the housing with a soft hammer while the machine is stopped, and the software shows the structure's natural frequency along with the ring-down decay. Compare it with your running speed — if they are within a few percent, no amount of trial weights will help. On a coast-down, a resonance shows a clear amplitude peak with a ~90° phase shift at that speed.

Looseness: the one that breaks the math

Balancing math is linear algebra. It assumes that doubling the unbalance force doubles the vibration response. Looseness violates this assumption. A loose bearing pedestal may be stiff in one direction but floppy in another. A soft foot lifts the machine off one mount at a certain vibration amplitude, changing the effective stiffness mid-cycle.

Before balancing any machine, check: all anchor bolts torqued, no soft foot (feeler gauge under each foot), no cracks in the baseplate, no play in bearing pedestals. If the Balanset-1A shows a "forest" of harmonics instead of a clean 1× peak — the F4-Harmonics (1x..Nx) tab in F8 – Charts displays the 1× to N× amplitudes directly — fix the structure first.

Misalignment: the 2× signature

Coupling misalignment produces forces primarily at 2× RPM (and sometimes 3×). If the Balanset-1A FFT shows a strong 2× component — especially combined with high axial vibration — alignment is the problem, not balance. Laser-align the shafts first. Then check if balancing is still needed. Often it isn't.

Rotor Condition: Dirty Impellers and Bent Shafts

The dirty rotor problem

Dust, product buildup, calcium deposits, corrosion — any of these on fan blades, pump impellers, or centrifuge rotors create uneven mass distribution. The machine vibrates. The temptation is to balance it "as is" and get back to production.

Don't. The Balanset-1A will produce a correction solution for a dirty rotor. It doesn't know the rotor is dirty — it just measures vibration and calculates. But those deposits flake off during operation. In a fan processing hot gas, a chunk of scale drops at 2 AM on a Saturday. Now the rotor is instantly out of balance — except worse, because your correction weights were compensating for the dirt that just fell off. The weights are now the unbalance source.

Bent shafts: why heavy weights at one speed don't help

A bent shaft creates eccentricity — the geometric center doesn't match the rotation center. This looks like unbalance at 1× RPM. The critical difference: a bent shaft produces vibration that's speed-dependent in a way that simple unbalance isn't. You can sometimes reduce vibration at one specific speed with a large correction weight, but at any other speed the vibration is worse. And the shaft stress increases, shortening bearing and coupling life.

The verification is mechanical: measure runout with a dial indicator while turning the shaft slowly by hand. If total indicated runout (TIR) exceeds the machine's tolerance — typically 0.02–0.05 mm for precision rotors, up to 0.1 mm for heavy industrial — the shaft must be straightened or replaced. Balancing cannot fix geometry.

Procedural Errors: Trial Weight, Angle, and Temperature

Sometimes the machine is healthy and the fault is in the procedure. These are the errors that make technicians think "the instrument is broken" when actually the input data is wrong.

Trial weight too small

The Balanset-1A learns the system by measuring how it responds to a known trial weight. If the trial weight is too small, the change in amplitude and phase is buried in measurement noise. The software computes influence coefficients from noise, and the resulting correction is essentially random.

Target: the trial weight should change amplitude by at least 20–30% or phase by at least 20–30°. If you add 10 g and the reading barely moves, try 20 g or 30 g. Start conservatively, but don't be afraid to go bigger if needed. The math needs a clear signal. You don't have to guess the starting mass: after Run 0 the Balanset-1A software offers a First Trial Weight Estimator that computes both the trial mass and the installation angle.

Angle measurement errors

Balancing is vector math. A 10 g weight at the right angle cancels the unbalance. The same 10 g at 180° from the right angle doubles the unbalance. How bad an angle error is, exactly: the residual after a correction placed θ off target is 2·U·sin(θ/2) of the original unbalance U. 20° off leaves 35%, 45° off leaves 77% — still an improvement. At 60° you are back where you started, and beyond 60° the machine is worse than before you touched it.

Two common errors easily cost more than 60°: measuring the angle in the wrong direction — with the Balanset-1A the correction angle is always measured from the trial weight position in the direction of rotation, and for rotors that turn counter-clockwise as seen from the sensor there is a mirror option on the polar diagram — and moving the tachometer or reflective mark between runs, which shifts the zero reference.

Both are silent killers — the software shows a confident correction, you install it, and vibration jumps. If vibration increased after installing the calculated correction, the first thing to check is whether the angle was measured in the correct direction.

Thermal distortion: the "it was fine this morning" problem

A motor balanced at 20°C winding temperature may vibrate badly at 80°C. Hot-gas fans that handle 200–400°C process gas develop thermal bow — the shaft or impeller warps slightly as temperature rises, shifting the mass distribution. The balance you achieved cold is gone when hot.

The fix: run the machine to thermal steady state (full operating temperature, stable conditions) before the final trim balance run. Balance "hot" for machines that run hot. If the machine has significant cold-to-hot vibration change, document both conditions — some customers accept higher cold-start vibration knowing it drops once the machine warms up.

Decision Table: What Does the Spectrum Tell You?

Open F8 – Charts in the Balanset-1A software and switch to the F5-Spectrum (Hz) tab. Look at the peaks. Match the pattern to the fault.

Field Report: The Fan That Kept Coming Back

A grain processing plant called about a large induced-draft fan, 45 kW, running at 1,470 RPM. They'd balanced it three times in six months. Each time: vibration dropped to about 2 mm/s, and within 3–4 weeks it climbed back above 8 mm/s. The previous technician had welded correction weights after each balance — three sets from three separate visits, all still on the impeller.

First thing I did was open the Balanset-1A spectrum (F8 – Charts). The FFT showed a clean 1× peak at 24.5 Hz (shaft speed) — so it looked like unbalance. Phase was stable. No looseness. No misalignment signature. That part checked out.

Then I looked at the impeller. Heavy grain dust coating, 3–5 mm thick, unevenly distributed. The previous technician had balanced against the dust each time. Dust accumulated, shifted, partially fell off — and the vibration returned. The correction weights from three visits were now fighting each other.

We removed all previous correction weights (three sets, 11 weights total). Cleaned the impeller to bare metal. Balanced from scratch. Single 2-plane correction: 22 g front, 15 g rear.

The plant installed a monthly cleaning schedule for the impeller. Vibration was 0.9 mm/s right after the job and 1.1 mm/s six months later — no rebalance needed. The three previous visits — removal of old weights, welding, measurement — cost more in total than a single correct diagnosis would have.

For context: a 45 kW machine falls into ISO 10816-1 Class II (medium machines, 15–75 kW), where the vibration severity zone boundaries are 1.12 / 2.80 / 7.10 mm/s RMS (A/B, B/C, C/D). The original 8.4 mm/s was deep in zone D — vibration severe enough to cause damage — while both 0.9 and 1.1 mm/s sit in zone A, the as-new range. See our vibration severity chart for other machine classes.

Pre-Balance Checklist

Before you place a trial weight on any machine, verify every item on this list. If any check fails, fix it first. Balancing a machine that fails one of these checks is wasted time.

1. 1
   Rotor clean?

   Bare metal. No dust, no deposits, no product buildup. If you can't clean it, document the risk and tell the customer the balance may not hold.
2. 2
   Shaft straight?

   Dial indicator check. TIR within machine tolerance (0.02–0.05 mm for precision, 0.1 mm for heavy industrial). If out, straighten or replace.
3. 3
   No looseness?

   All bolts torqued. Feeler gauge under every foot — no soft foot. No cracks in baseplate. Bearing pedestals solid. Spectrum: no "forest" of harmonics.
4. 4
   Alignment acceptable?

   Axial vibration less than 50% of radial. The Balanset-1A's two channels normally mount radially — turn one sensor to the axial direction for a separate run to check this. No strong 2× in spectrum. If suspect, laser-align first.
5. 5
   Not near resonance?

   Phase stable — wandering less than ±10–20° while the RPM readout looks steady. If it wanders more, run the built-in Bump Test; change speed or modify structure before balancing.
6. 6
   At operating temperature?

   For hot-running machines: balance at thermal steady state, not cold. If cold/hot difference is significant, document both.
7. 7
   Tachometer and reference fixed?

   Reflective mark in place. Tachometer secured. Angle direction verified (with or against rotation). Do not move any reference after the first run.
8. 8
   Acceptance criterion agreed?

   Know the ISO 10816 zone limit for the machine before you start, and the residual-unbalance grade G (ISO 1940-1, now ISO 21940-11) you are aiming at. The Balanset-1A software has both built in: an ISO 1940 tolerance calculator on the plane panel and ISO 10816-1 machine classes in the route/trend module.

Frequently Asked Questions