# How to verify a balancing result: a verification measurement that can't be disputed

> The contractor packs up the instrument, the noise is clearly lower, everyone's happy. A month later, reliability asks how much the vibration actually dropped, and there's nothing to answer with. We break down exactly what gets compared before and after, how 'within tolerance' in the software differs from the acceptable level for the machine, and which checklist to run through before you sign off.

**In short:** Compare four numbers at each point: overall vibration in mm/s RMS, the 1x amplitude, its phase, and the speed. An 'after' measurement is only valid if it's taken at the same points, in the same directions, with the same sensor mounting, in the same quantity and frequency band, at the same operating regime, and at the same temperature as the 'before' measurement. Signs of an honest result: it's specifically the running-speed component that dropped, the phase is stable, the amplitude repeats from run to run, and the rest of the spectrum hasn't grown. 'Within tolerance' in the instrument's software means exactly one thing: the residual 1x is below the target you entered yourself.

Source: https://axiline.pt/en/articles/how-verify-balancing-result/  
Publisher: AXILINE · Vila Nova de Gaia, Portugal · +351 931 831 229 · axilinegeral@gmail.com

## 'Sounds a bit quieter' is not a result

The ear is a poor instrument here. People hear high frequencies well, but balancing works at the rotational frequency: for a machine at 1450 rpm that's 24 Hz, for a slow-turning drum it's 3–5 Hz. Down there you barely hear anything, but you feel it through the housing and through a hand on the railing. So 'it's quieter' and 'the vibration dropped' are two different claims, and they don't always agree.

The result of balancing is four numbers at each point before the work and the same four numbers after: overall vibration in mm/s RMS, the amplitude of the 1x running-speed component (vibration at the rotational frequency — exactly what unbalance produces), its phase (the angle relative to the mark on the rotor), and the rotational speed. Everything else is impressions.

The word 'same' is doing all the work here. An 'after' measurement is only comparable to a 'before' measurement when everything listed below matches.

- Point. A specific bearing housing, not 'the drive side.' Mark the pads with a marker or paint and take measurements from exactly there.
- Direction. Compare horizontal-radial against horizontal-radial. On the same machine, horizontal and vertical readings can differ several times over.
- Sensor mounting. Only identical mounting can be compared: a magnet on a pad cleaned down to bare metal against a magnet on that same pad. A handheld-probe reading and a stud-mounted reading aren't comparable at all.
- Quantity. Vibration velocity, root mean square, mm/s. Don't mix RMS with a peak amplitude value, or mm/s with micrometers of displacement.
- Frequency band. Overall vibration depends entirely on where the band's limits are set. Fix it and don't change it.
- Operating regime. Speed, load, damper or inlet guide vane position, flow, pressure, current.
- Temperature. Compare a warmed-up machine against a warmed-up machine.

> One 'before' measurement and one 'after' is the bare minimum. Take two or three measurements at each point: the spread between them tells you which number you can trust and which one is noise.

## Why a measurement taken under a different regime proves nothing

The centrifugal force from an unbalanced mass grows with the square of the rotational frequency. Drop a fan's speed from 1450 to 1000 rpm, and the force from the exact same unbalance falls by roughly half. Vibration dutifully drops right along with it. The rotor, meanwhile, stays exactly as it was.

That leads to an unpleasant conclusion: an 'after' measurement taken at reduced speed will show success even where no weights were fitted at all. It also works the other way around. Raise the speed by 20%, and a machine that was honestly balanced looks worse than it was.

The bearings react even more sharply. Amplitude is force divided by the dynamic stiffness of the rotor-bearings-base system, and stiffness depends on frequency. Near the structure's natural frequency (the frequency at which it's inclined to sway on its own), a 5% difference in speed can change the amplitude several times over and swing the phase by tens of degrees. We covered how to recognize this in a separate article on resonance and its signs.

Operating regime works its way into the number through more than just speed. A damper in a different position changes the forces on the impeller, a pump running outside its duty point adds hydraulic effects and cavitation, and a motor under load has its current and slip change.

| What changed between 'before' and 'after' | What happens to the reading | What this turns into in the report |
| --- | --- | --- |
| Speed is 30% lower | 1x drops by roughly half from this alone | A false success |
| Speed moved closer to the natural frequency | Amplitude grows several times over at the same unbalance | A false failure |
| Damper or inlet guide vane in a different position | Forces on the impeller and overall vibration change | The numbers aren't comparable |
| Machine cold versus warmed up | Clearances, shaft-to-shaft position, and preload change | A 20–50% spread with no visible cause |
| Sensor moved from the housing to the casing | Panel vibration at its own frequencies gets added in | A meaningless number |

> The correct order is: fix the operating regime and the points first, then balance. If the regime had to change partway through the job, capture a new 'before' baseline at the new regime — otherwise there's nothing to compare against.

## 'Within tolerance' in the software and 'within normal range' for the machine are different things

When the Balanset-1A software reports after a verification run that you're within tolerance, it's comparing one quantity against one number: the residual 1x against the target value you entered yourself into the tolerance field when you created the archive record. If you also entered the rotor mass, speed, and correction radius, it will also compare the achieved residual unbalance in g·mm against the tolerance for the selected G grade (the balance quality grade).

The check is honest and useful. But it answers exactly one question: whether you reached the target you set for yourself. Set the target at 4.5 mm/s, and 'within tolerance' shows up at 4.4.

What this message doesn't confirm:

- It doesn't confirm zone A or B for overall vibration. Overall vibration includes everything else: shaft misalignment, looseness, bearings, aerodynamics.
- It doesn't confirm the G balance quality grade if you didn't enter the rotor mass, speed, and the weight-fitting radius.
- It doesn't confirm that the machine is in good condition. A cracked blade and a wrecked bearing can happily coexist with a residual 1x of 0.8 mm/s.

> In the report, write three separate lines: residual 1x against the target, overall vibration against the norm for this machine, and residual unbalance against the G grade. We cover this in detail in a separate article on the three tolerances in balancing.

Sources: [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) · [ISO 21940-11:2016](https://www.iso.org/standard/54074.html) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/)

## Four signs of an honest result

This is checked with one extra stop. Do a verification run, record three numbers at each point, stop the machine, restart it, and record again. Two matching lines in the report are worth more than any amount of reassurance.

A polar plot shows the same thing visually — a chart where the 1x amplitude and phase are drawn as a single vector arrow. The 1x vector should move from its starting point toward the center and stay close to it. If the vector moves sideways or grows longer after fitting the weight, look for an error in the angle, in the reference direction, or in the fitting radius — don't just increase the mass at random.

### It's specifically 1x that dropped

Look at the running-speed component at each point, not at the overall impression. A normal picture is a drop of several times over: from 8–12 mm/s down to 1–2 mm/s. A 15% drop in 1x isn't a result, it's measurement scatter.

### 1x phase is stable

On a verification run, the phase holds steady and drifts within a few degrees. If it swings by 30–60° between repeat runs, you're not measuring unbalance: that's how resonance, play in a fit, and loosened fasteners behave.

### The numbers repeat from run to run

Stop it, start it, measure again: the 1x amplitude should repeat within roughly 10%. Unbalance doesn't disappear or appear between runs. A twofold spread means something in the machine isn't fastened down.

### The rest of the spectrum hasn't grown

Compare the before and after spectra: 2x, harmonics, the noise floor, the bearing-frequency range. A weight shouldn't add anything. A 2x that grows after correction is a signal that you've overloaded a bearing housing or loosened a fit.

## 1x dropped, but overall vibration barely did: what it means

Let's work through a typical case in numbers. Before the work, at one point: overall 9.0 mm/s, 1x 4.0 mm/s. After correction: 1x 0.8 mm/s, overall 8.1 mm/s. You removed the unbalance almost completely, and the overall figure only moved by 10%.

There's no error here, that's just how the arithmetic works. Vibration components don't add up like a column of numbers — they add like the legs of a right triangle: with an overall of 9.0 and a running-speed component of 4.0, everything else already comes to 8.1 mm/s. Nobody can subtract 4.0 from 9.0 and get 5.0.

This kind of result isn't a failure, it's a diagnosis. It's telling you: unbalance's contribution to this machine's vibration was small, and a second cause is producing most of it. From here on, you're no longer working with weights.

1. **Look at where the remainder sits** — Take a spectrum at the same points. A pronounced 2x — check shaft alignment and looseness. Many harmonics and a raised noise floor — mechanical looseness, play, or rubbing. Peaks at high frequencies that aren't multiples of the speed — bearings. A peak at blade-pass frequency, meaning the number of blades multiplied by the speed — aerodynamics or hydraulics.
2. **Compare the points and directions against each other** — Axial vibration comparable to the radial vibration, and a phase difference of about 180° across a coupling between bearings, point to shaft misalignment, not rotor mass.
3. **Check the mechanics by hand** — Anchor bolt and foot tightness, soft foot, the condition of shims and the frame, the rigidity of the pedestals under the bearing housings, cracks in welds. This is done with a wrench and a visual check, no instrument needed.
4. **Check for resonance on coastdown** — Record the 1x amplitude and phase during coastdown — the free decline in speed after the drive is switched off. A narrow amplitude peak with a phase swing near the operating speed means the number is being driven by the structure's dynamics, not by unbalance.
5. **Decide from the data, not from impressions** — Repair, shaft alignment, bearing replacement, remounting the bearing housings. Only after that comes a repeat vibration assessment: once the main cause is removed, you'll sometimes find that one more balancing correction is needed.

> Wording for the report: 'balancing completed, residual 1x reduced from 4.0 to 0.8 mm/s; overall vibration of 8.1 mm/s is caused by the 2x component, probable cause is shaft misalignment; shaft alignment with a repeat measurement is recommended.' That's an honest result, not an excuse. More detail in our articles on the ratio between overall vibration and 1x, and on cases where balancing won't help.

Sources: [ISO 13373-3:2015](https://www.iso.org/standard/40840.html) · [ISO 13373-5:2020](https://www.iso.org/standard/62202.html)

## A second verification measurement a few days later

A measurement taken right after correction proves one thing: you calculated correctly and fitted the weight correctly. It doesn't prove the result holds up. That's what a second measurement answers, taken after 3–7 days of running or after the machine's first full process cycle.

Four questions only this second measurement can settle:

- Is the weight still in place? A two-point tack weld cracks from vibration and heat, a bolt without locking works itself loose, an adhesive or magnetic weight slides on a hot surface. A sharp rise in 1x with a phase shift of tens of degrees is the recognizable signature of a lost weight.
- Is the rotor building up deposits again? On induced-draft fans, fans handling wet dust, mulchers, and screw conveyors, buildup comes back within days, and a balancing job done on a dirty impeller only lasts until the next cleaning.
- Has the machine reached its real thermal state? Clearances and shaft-to-shaft position are different on a warmed-up machine, which means the vibration is different too.
- Is there a baseline level? Two or three numbers on a warmed-up machine in good condition after a repair form the reference point you'll later use to track the trend and set thresholds from. Without it, absolute norms operate blind.

> Record the repeat measurement on the same form and at the same points: overall, 1x, phase, speed, regime, date. A rise in 1x of more than 20–30% from the verification value within a week isn't scatter, it's a process. We cover what's behind it in a separate article on vibration that came back after balancing.

Sources: [ISO 20816-1:2016](https://www.iso.org/standard/63180.html)

## Checklist for accepting a contractor's work on site

- [x] The contractor showed the 'before' numbers for each point: overall, 1x, phase, speed. From the instrument's archive, not from memory.
- [x] The measurement points are marked on the machine and match the ones used for the original vibration reading.
- [x] The speed in the 'before' and 'after' measurements differs by no more than a couple of percent, and the regime and damper position are the same.
- [x] The verification run was done on a warmed-up machine and under working load, not at idle.
- [x] You saw at least one verification run and the numbers on the screen with your own eyes.
- [x] There were two or three verification runs, with a stop between them, and the 1x amplitude repeated.
- [x] A spectrum after balancing was shown, and no new peaks appeared in it and 2x hasn't grown.
- [x] The mass of each correction weight in grams, the radius, and the exact fitting location were stated; the weights can be seen and touched.
- [x] The weight's mounting method suits the machine: a continuous weld instead of two tack welds, a bolt with locking, nothing adhesive on a hot surface.
- [x] The result is stated as three separate lines: residual 1x against the target, overall vibration against the norm, residual unbalance against the G grade if it's specified in the contract.
- [x] If overall vibration didn't come within the norm, the cause of the remainder is stated and a specific recommendation is given.
- [x] A date for the repeat measurement after a few days of running has been agreed.
- [x] It's been agreed when and in what form you'll receive the report.

> One phrase saves half of all future disputes: 'please show me the before-and-after measurement from the instrument's archive on one screen.'

## What the report needs so the check can be repeated

The report has one practical job. Six months from now, a different person with a different instrument should be able to come to this machine, repeat the measurement from your description, and get the same numbers. Everything needed for that goes into the report. Everything else is optional.

The Balanset-1A keeps an archive for each rotor: a folder of diagrams, a time-stamped history file of every run, saved influence coefficients, and a report that opens in the built-in editor, can be edited, and printed. That's where the rows in the table below come from: almost everything is already recorded, and all that's left is to add the measurement conditions and conclusions.

| Report section | What's in it | What you can't repeat the measurement without |
| --- | --- | --- |
| Machine and rotor | Name, position on the diagram, power, rotor mass, operating speed, support type: rigid or flexible | Without mass, speed, and radius you can't recalculate the G grade; without the support type you can't select the assessment zone |
| Point diagram | A sketch or photo with the points marked, measurement directions, sensor mounting method | Otherwise the next person will place the sensor somewhere else and get a different number |
| Measurement conditions | Speed at each run, load and regime, temperature, the measured quantity and frequency band | This is the main condition for comparability |
| 'Before' measurement | Overall, 1x, and phase at each point | The reference point — without it the result can't be proven |
| Balancing process | Number of planes, the trial weight's mass and radius, its position, the influence coefficients obtained | Lets you get by with one run instead of three next time |
| Correction | Mass of each weight, radius, angle or fixed-position number, mounting method, metal added or drilled out | A year from now you'll find your own weights and know what can be removed |
| 'After' measurement and verification runs | The same three numbers at each point for every run, spectrum, polar plot | The actual proof of the result |
| Assessment | Residual 1x against the target, overall against the norm citing the applicable part and edition of the standard, residual unbalance in g·mm and the G grade | Three tolerances instead of one blanket statement |
| Conclusions | What's left in the vibration, what's causing it, what to do next, and when to repeat the measurement | Continuity of work for whoever handles it next |

> The frequency band is worth stating outright. The Balanset-1A measures RMS (root mean square) vibration velocity over a 5–200 Hz range, and that isn't the same as the wide 10–1000 Hz band used in overall-level criteria. The running-speed component and the lower harmonics of a machine running at 300–3000 rpm fall inside that range, but a wideband number taken over a different band stays a different number. State the band, and you've headed off a future dispute.

Sources: [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/) · [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) · [ISO 21940-11:2016](https://www.iso.org/standard/54074.html)

## If you'd rather hand the check to engineers

AXILINE's engineers design and manufacture the Balanset instruments and do the balancing themselves on site visits. That's why a verification measurement is part of the work for us, not a favor done at the customer's request.

Here's what that looks like on site. We mark out points on the bearing housings and record the directions. We take the starting overall vibration, 1x, phase, and speed. We balance the rotor in place, in its own bearings, in one or two planes. We do verification runs with stops in between, and show you the spectrum and polar plot on the screen. We hand over a report that records the measurement conditions, not just the final numbers. If overall vibration doesn't come within the norm, we tell you plainly what's causing the remainder and what to do about it.

We also do stand-alone verification: an independent verification measurement after someone else's work, following the same rules, with a report. Sometimes that's enough to settle a disagreement between reliability and a contractor.

You can also get the instrument yourself and check machines on your own. The Balanset-1A kit includes two accelerometers, a laser phase sensor using reflective tape, a two-channel USB module with preamplifiers, integrators, and an ADC, and Windows software: overall vibration and 1x with phase, speed, FFT spectrum and time waveform, single- and two-plane balancing, fixed positions and drilling calculations, tolerance calculation by G grade, trim balancing and saved influence coefficients, a polar plot, and a per-rotor archive with run history and report printing. For building into machine tools and test stands, there's a Balanset-1A OEM version without the case. Methodology consulting support is included.

Write to us with what kind of machine it is, what speed it runs at, and the numbers you got before and after. Three figures are usually enough to tell whether a result was proven or just declared.

Sources: [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/)

## Frequently asked questions

**Is one verification run enough to accept the work?**

No. One run shows that the weight was calculated and fitted correctly, but it doesn't show repeatability. Do two or three runs with a stop between them and compare the 1x amplitude and phase. A discrepancy of up to 10% in amplitude and a few degrees in phase is normal. A spread of several times over means something in the machine isn't fastened down, and it's too early to accept the work.

**How much should vibration drop for balancing to count as successful?**

The right question isn't 'by how many percent' but 'down to what number.' Success is a residual 1x below the agreed target and overall vibration in the zone you committed to bringing the machine into. A benchmark for 1x: a drop of several times over, usually from single or double-digit mm/s down to 0.5–2 mm/s. If 1x only dropped by 15–20%, either unbalance wasn't the main cause, or the correction was done imprecisely.

**We balanced at 900 rpm, but the machine runs at 1450. Is the measurement valid?**

As proof of the result at the operating regime — no. The force from unbalance grows with the square of the speed, and the bearings respond differently at a different frequency. Take a verification measurement at the operating speed and under working load. If you can't reach that regime for process reasons, record in the report what regime the numbers were taken at, and don't extrapolate them to the operating regime.

**1x dropped fivefold, but overall vibration only by 8%. Did the contractor do a bad job?**

More likely the opposite: they did what weights can do, and the numbers show it. The components add quadratically, so you can't subtract 4 mm/s of running-speed component from 9 mm/s of overall and get 5. Ask a different question instead: did they show a spectrum, name the cause of the remainder, and give a recommendation? If yes, the work is done, and what's needed next is a different kind of repair.

**Why take a repeat measurement after a few days if everything was already within tolerance?**

It checks what the first one can't: whether the weight is still in place, whether the impeller has built up deposits again, and how the machine behaves at its real thermal state. Plus you get a baseline level for a machine in good condition, which you'll later use to track the trend and set thresholds. A practical interval is 3 to 7 days of running, or one full process cycle.

**The contractor didn't leave a report, just said 'within tolerance.' What do we do?**

Without recorded measurement conditions, a statement like that can't be verified, and in a dispute it's worth nothing. Request an export from the instrument's archive: the before-and-after numbers for each point, speed, weight masses and fitting locations, spectrum. If that doesn't exist, the only approach that works is an independent verification measurement at the marked points, to get your own reference point for the future.