# On-site balancing of electric motor and generator rotors

> The motor started humming after a repair, the bearing is running hot, and in a month it'll need to come apart again. Vibration in an electric machine almost never has a single cause: part of it comes from rotor imbalance, part from electromagnetic forces in the air gap, part from shaft misalignment with the driven machine. We come out with a two-channel vibration analyser, break the vibration down by frequency, and balance the rotor in its own bearing supports. If the cause isn't imbalance, you'll know within the first hour, not after the weights have been fitted.

**In short:** Yes, we balance electric motor and generator rotors on site, in their own supports, at operating speed. Three conditions apply. First: the vibration has to be dominated by the 1x running-speed component — vibration at the rotor's rotational frequency, the main sign of imbalance — rather than by line frequency and its second harmonic. Second: at least one correction plane needs to be accessible, meaning a spot where a weight can be fitted: usually the cooling fan, the coupling, a balancing ring, or the rotor face. Third: the fits, bearings, and fasteners have to be sound, because weights don't cure play. If the rotor is coming off anyway (rewinding, bearing replacement, restoring the mounting surfaces), it makes more sense to balance it on a machine in the workshop, and we'll say so plainly rather than fit weights through a hatch.

Source: https://axiline.pt/en/equipment/on-site-balancing-electric-motor-generator/  
Publisher: AXILINE · Vila Nova de Gaia, Portugal · +351 931 831 229 · axilinegeral@gmail.com

## Symptoms: what brings people to us, and when to call

An electric machine rarely starts vibrating for no reason. There's almost always an event behind it: a repair, a bearing replacement, a new fan, a different load, a move onto a different bedplate. Recall that event before you call — it saves an hour of diagnosis.

- [x] Humming and vibration increased right after a repair, rewinding, bearing replacement, or unit reassembly
- [x] Vibration rises as the machine warms up and returns to its previous level once it cools down
- [x] It all started after fitting a new cooling fan, a coupling, or a repair sleeve on the shaft
- [x] Bearings last noticeably less than their rated life, grease darkens and gets pushed out
- [x] The vibration level changes with speed on a VFD: at one setpoint frequency the machine hums, at another it's quiet
- [x] Foot bolts have loosened or sheared, there's a crack in a foot or the mounting base, the bearing fit in the end shield is worn
- [x] A round-check measurement put the machine in zone C or D of the applicable part of ISO 20816 (levels at which continued operation is not recommended or not permitted)
- [x] The generator is noisy under load and noticeably quieter at no load
- [x] Axial vibration at the coupling has increased, while radial vibration has barely changed

> The costly symptom here isn't the noise, it's bearings failing early. Residual imbalance creates a rotating radial force that adds to rotor weight, belt tension, and axial load. That force won't appear in a service-life calculation to ISO 281 unless someone built it in, which is why a bearing "unexpectedly" lasts half as long as expected.

Sources: [ISO 281:2007](https://www.iso.org/standard/38102.html)

## Which rotors in this family we balance

The family is broad but shares common mechanics: a rotor in two bearing supports, a cooling fan at one end, a coupling or pulley at the other. Below are the subgroups and exactly what creates imbalance in each of them.

### Induction motors

Rotors of general-purpose industrial induction motors: drives for pumps, fans, compressors, conveyors, crushers, mixers. Squirrel-cage or wound-rotor with slip rings. The core's own imbalance is usually small, but the cooling fan, the coupling, and the traces of a repair can produce plenty of it.

### Synchronous motors and generators

Rotors of salient-pole synchronous motors and generator rotors, including generators on diesel and gas sets. Here a second measurement is essential: at no load and under load. The electromagnetic component rises together with the current, while imbalance doesn't depend on load at all.

### Commutator-machine armatures

Armatures of DC motors: crane, traction, and drives for machine-tool and rolling-mill equipment. The rotor is long, almost always two planes. We correct via the banding rings and pressure plates, and we don't touch the commutator or the winding end-turns.

### Drive motors in units

Rotors of drive motors that are part of a machine: a pump, fan, or compressor unit. Often both the motor rotor and the driven machine's impeller need balancing, with shaft alignment checked between them as well. We decide the order of work on site from the spectrum, not by guesswork.

### High-speed machines

Rotors of high-speed electric motors: two-pole at 3000 rpm, spindle drives, machines driven above rated speed by a VFD. The requirements on residual imbalance are tighter, because centrifugal force rises as the square of speed, and the first critical speed can already be close by.

- Rotors after a repair or rewinding: a new winding lay-up, a banding wound on site, wedges, and excess impregnation that has set on one side all change the mass distribution around the circumference
- Rotors after a bearing replacement: the rotor's position in the air gap and the support stiffness have changed, and with them the whole vibration picture, phase included
- Rotors after restoring the mounting surfaces: hardfacing, spraying, chrome plating, or a repair sleeve add metal asymmetrically and almost always require a repeat balancing
- Rotors after fitting a cooling fan: a new or simply repositioned fan brings its own imbalance, especially a cast-aluminium one with uneven blade thickness
- Rotors after fitting or replacing a coupling: it sits on the shaft overhang, so its eccentricity and own imbalance act on a long lever arm and load the near support

> It's worth being precise about rewinding. A wound rotor, a DC-machine armature, and the poles of a synchronous machine carry a winding on the rotor, and a repair there directly changes the mass distribution. On a squirrel-cage induction motor it's the stator that gets rewound, and that affects rotor imbalance only indirectly: through disassembly, bearing replacement, removing and refitting the fan, straightening the shaft. Vibration rises after a repair like that too, it's just that the cause is assembly-related rather than electrical.

## Imbalance or electrical: how we tell them apart

This is the main way an electric machine differs from a fan or a pump. A motor has a source of vibration that no other rotor has: electromagnetic forces in the air gap. They live at their own frequencies and can't be removed by weights, as a matter of principle.

The markers are simple. Imbalance sits at the rotational frequency, that is, at 1x. Electromagnetic forces sit at line frequency and its second harmonic: 50 and 100 Hz on a 50 Hz supply. An uneven air gap, eccentricity of the rotor or the stator bore, loose winding in the slots, and a turn-to-turn short show up mainly at 100 Hz.

A broken or cracked squirrel-cage bar looks different. Sidebands appear around the running frequency and around 100 Hz, spaced at the slip frequency (the small difference between field speed and rotor speed) multiplied by the number of poles. That spacing is small, so we set a high enough number of spectral lines that neighbouring peaks actually separate instead of merging into one hump.

There's a trap that catches people often. On a four-pole motor at 1500 rpm, the second harmonic of the running frequency (2x) equals exactly 50 Hz and coincides with line frequency. On a two-pole motor at 2900 rpm, the first harmonic sits around 48 Hz, almost right up against 50 Hz. Peaks like that are hard to tell apart from the spectrum alone. From the machine's behaviour, it's easy.

The power-off check takes one run. The motor reaches its operating mode, we record 1x amplitude and phase and the overall level, then power is removed. The electromagnetic force disappears together with the current: amplitude drops in a step within a fraction of a second, while the rotor is still turning at nearly the same speed. Imbalance doesn't behave that way: it fades smoothly with coast-down, that is, as the rotor slows on its own, because centrifugal force is proportional to the square of the rotational frequency. We agree this check with your staff and only do it where a free coast-down is safe for the machine and the process.

| What we see | What it most likely is | What we do |
| --- | --- | --- |
| A 1x peak dominates, phase repeats from run to run within a few degrees | Rotor imbalance | We balance on site, in one or two planes |
| Amplitude drops in a step the moment power is removed | Electromagnetic component | Weights won't help: measure the air gap, check the windings and the stator fit |
| A noticeable peak at twice line frequency (100 Hz on 50 Hz) | An uneven air gap, rotor or stator eccentricity, loose winding in the slots | Electrical-side work; balancing is postponed |
| Sidebands around 1x and around 100 Hz spaced at "slip × pole count" | A broken or cracked squirrel-cage bar | Rotor flaw detection and cage repair; imbalance here is secondary |
| Strong 2x plus high axial vibration at the coupling | Shaft misalignment | Shaft alignment, and only then balancing if still needed |
| A comb of harmonics at 1x, 2x, 3x and above, unstable phase | Mechanical looseness: foot bolts, a soft foot, a worn fit | Fasteners and fit repair; there's nothing to balance in this state |
| Peaks in the high-frequency region, not a multiple of speed | Bearing defects | Diagnosis and replacement; weights won't fix it |

> On a generator we add one more cross-section. We measure at no load and under load, ideally at two current levels. Imbalance doesn't depend on load at all. Vibration that rises together with current or with winding temperature points toward electrical causes and the rotor's thermal state. The general method for finding the cause is covered in a separate article; here we apply it to an electric machine.

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

## What we check before we bring out the weights

The walk-round takes fifteen minutes to an hour and decides whether it's worth starting the procedure at all. Some checks are done by hand on a stopped, locked-out machine, others with instruments on the running one.

- [x] Mounting to the frame and foundation: foot-bolt tightness, a soft foot, cracks in the feet and mounting base, the stack of shims under a foot, condition of the anchor bolts
- [x] Bearing fits: interference on the shaft and in the end shield, signs of a spun outer race, running-in marks and heat-tint colours
- [x] Bearing condition from the high-frequency components of the spectrum, noise and temperature, fluting on the raceways from bearing currents
- [x] Cooling fan: cracks at the blade roots, missing or broken blades, dust buildup between the blades, the fit on the shaft after a repair
- [x] Coupling or pulley: radial and face runout by dial indicator, the fit on the shaft overhang, condition of the flexible elements, presence of both keys
- [x] Shaft runout by dial indicator: residual bow after long storage resting on one side, after heating, or after straightening
- [x] Key and keyway: an empty keyway under the coupling creates its own imbalance, and a second key changes the calculation picture
- [x] Rotor after rewinding: protruding wedges, an uneven banding, impregnation runs, asymmetrically laid end-turns
- [x] Process mode: speed stability, VFD setpoint, generator load, valve position on a driven pump
- [x] Thermal state: measurement on the cold machine, repeated after 30 or 60 minutes of running under load

> Thermal bow shows up on electric machines more often than you'd think. The rotor heats unevenly, the shaft bends, and imbalance becomes a function of temperature. The tell-tale sign: 1x rises slowly as the machine warms up, and phase drifts by tens of degrees and doesn't come back until the machine cools down. A machine like that can be balanced, but the result will only hold for one thermal state. We look for the cause separately: a turn-to-turn short, uneven impregnation, blocked cooling passages, rotor rubbing.

## How the work proceeds on site

1. **Access, inspection, lockout** — We establish who operates the machine and who applies the lockout, where the key is, how the fan cover opens, and whether the coupling is accessible. We go through the walk-round from the checklist above. This is also where we decide whether a free coast-down is safe: it's needed for the electrical and resonance checks.
2. **Sensors and the phase mark** — We fit two accelerometers on the motor's bearing supports, as close to the bearing as possible: on a rib or boss of the end shield, on a cleaned flat spot, magnet-mounted or on a stud. Direction is radial, usually horizontal, and the same on every run. A sensor on the fan cover, the terminal box, or a guard cover gives large numbers that have nothing to do with the rotor.
3. **Speed and the phase reference** — We aim the laser sensor at a reflective tape mark. We stick it on a degreased section of the shaft overhang, on the coupling, or on a visible section of the shaft on the fan side. Exactly one mark: a second strip would give a speed reading twice too high. We make sure the laser doesn't pick up a stray reflection from the keyway or a chamfer.
4. **Baseline measurement** — The machine at operating speed and mode. We record speed, overall vibration in mm/s RMS, 1x amplitude and phase at each support, spectrum, and time-domain signal. We measure axial vibration at the coupling separately, once.
5. **Imbalance, electrical, or mechanical** — We compare 1x against overall vibration, look at 50 and 100 Hz and the sidebands, do the power-off check and watch the coast-down, and check phase repeatability with two consecutive runs. Only after that do we say whether balancing is worth doing or the cause is something else. If it's something else, diagnostics follow, not weights.
6. **Trial weight in the first plane** — We weigh the trial weight, fix it at the marked point just as securely as the permanent one will be, and enter the actual mass and actual radius. We count the trial run as valid if 1x amplitude has changed by at least 20–30% or phase has turned by at least 20–30°. If the change is smaller, the weight is too small: we increase it and repeat.
7. **Trial weight in the second plane** — Only for a two-plane calculation. We move the weight as directed by the software and repeat the measurement. In total: one plane needs two runs, two planes need three.
8. **Masses, angles, and fixing** — The software outputs a mass and position for the weight in each plane. The zero reference is wherever the trial weight sat. On the fan blades and the coupling bolts, fixed-position mode is more convenient: the instrument returns a position number and a mass, and errors in the direction of reading disappear. We fix the weight with a bolt through a factory hole, with washers under the coupling bolts, by welding on a plate where welding is allowed and won't heat the winding, or by removing metal by drilling in a solid part. Drilling the laminated core, the cage bars, or the banding is not allowed.
9. **Verification run and trim** — Same speed and same mode as the baseline run. The running-speed component should drop several times over. If we don't hit the target right away, the software will suggest adding small masses to what's already fitted. One or two iterations are usually enough.
10. **Recording the result** — We record the before-and-after numbers for each support and direction, the speed, mode, temperature, and the masses, radii, and positions of the weights. The influence coefficients — this specific machine's measured response to a trial weight — stay in the instrument's archive, so a repeat balancing of the same machine will run without trial weights.

> We work with the Balanset-1A: two accelerometers, a laser phase sensor, a two-channel USB module with preamplifiers and an ADC, software on a laptop. The instrument calculates the correction in one and two planes using the influence coefficient method, displays overall vibration, 1x, phase, speed, spectrum, and time-domain signal, splits the mass across fixed positions, calculates drilling, assesses tolerance by G class, and keeps an archive for the report.

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

## One plane or two for electric-machine rotors

Rotor shape and speed determine the number of correction planes. We work out the L/D ratio, where L is the distance between possible correction planes and D is the diameter at the weight-mounting zone. The rule itself is covered in a separate article; below is the practical picture for electric machines.

The difference between a motor and a fan is that the available planes are fixed by the design, not by what you'd prefer. As a rule that means a cooling fan on one side and a coupling or pulley on the other. They sit at opposite ends of the shaft, which is a convenient geometry for a two-plane calculation.

| Rotor type | Planes | Why |
| --- | --- | --- |
| Short rotor on a small motor, L/D under 0.5, up to 1500 rpm | One | Behaves like a disc, static imbalance dominates, one trial run is enough |
| Typical rotor, medium to large power, L/D over 0.5 | Two | Otherwise couple imbalance remains: one mass can't reduce vibration at both supports at once |
| Two-pole motor at 3000 rpm | Two | High speed amplifies the couple component, requirements on the residual are tighter |
| DC-machine armature | Two | The rotor is long, correction via the banding rings and pressure plates at both ends |
| Only the cooling fan is accessible | One, with a caveat | We'll reduce vibration at the near support; a residual may remain at the far one. We'll say so before starting work |
| Both the fan and the coupling are accessible | Two | The planes sit at opposite ends of the shaft, the calculation converges reliably |
| The rotor runs close to its first critical speed | Limited on-site | A flexible rotor behaves differently, two planes may not be enough. We check applicability against the current edition of ISO 21940-12 |

> A special case: correcting at the coupling. Mass at the coupling sits on the shaft overhang, beyond the bearing, so its effect on the far support can be opposite in sign and non-obvious. That's exactly why we don't guess, but obtain the influence coefficients with a trial run on this specific machine.

Sources: [ISO 21940-12:2016](https://www.iso.org/standard/50429.html)

## What gets in the way, and when on-site won't work

### No access to the correction plane

An enclosed machine, a cast fan cover, a fan cast into the housing, a coupling covered by a guard that can't be removed without dismantling the unit. If not a single plane can be opened up, on-site balancing is physically impossible.

### The cause is electrical

A broken cage bar, eccentricity of the rotor or the stator bore, a turn-to-turn short, loose winding in the slots. Weights are useless here: you'd balance the rotor, and the 100 Hz vibration would stay right where it was. We'll show that with numbers and won't charge for balancing that isn't needed.

### Thermal bow

Vibration rises as the machine warms up, phase drifts by tens of degrees. Balancing will only give a result for one thermal state, and on a cold or overheated machine the vibration will come back. The cause of the warming comes first, weights come second.

### Play and worn fits

A spun bearing race, a worn end shield, a loosened fan or coupling fit. The rotor's mass isn't fixed, while the calculation assumes it is. Restoring the mounting surfaces comes first, balancing comes second.

### VFD and drifting speed

Balancing requires a constant speed. On a VFD we fix the setpoint, disable PID control and pressure- or flow-based auto-adjustment, and wait for a stable frequency. If speed drifts by more than a couple of percent, the influence coefficients drift too, and the correction comes out essentially random. We balance at the frequency where the machine actually runs most of the time, and if there are two operating modes, we check the result at both.

### Resonance in the supports and frame

During coast-down, amplitude shows a narrow peak and phase swings through about 180 degrees. Balancing is unstable inside resonance: the result scatters from run to run. We address the stiffness of the frame and foundation, or move away from that speed.

### The machine can't be stopped

Fitting weights requires a shutdown and lockout. One plane means at least two stops, two planes mean three. If the process doesn't allow stops, we plan the work into a window or push it to a scheduled shutdown.

> When it's more honest to pull the rotor. If the machine is going for disassembly anyway, if a G-class quality document is needed, if the rotor is flexible, or if not one plane is accessible on site, workshop balancing on a machine is more accurate and cheaper in the end. We'll say so at the start, not after three runs. A comparison of the two approaches is in a separate article on on-site versus workshop balancing.

## What you get at the end

The result of the work isn't just a quiet machine, but numbers you can carry forward: compare against a previous measurement, show to the manufacturer, build into the maintenance plan.

- A report with baseline and residual vibration for each support and each direction, the overall level in mm/s RMS stated separately from the 1x running-speed component
- Speed, mode, load, and temperature at the time of each measurement, plus a diagram of the sensor mounting points
- Spectra before and after, so you can see exactly what dropped and what remains
- Masses, radii, and positions of the fitted weights, with a description of the fixing method
- An assessment of overall vibration by zone under the applicable part of ISO 20816, citing the edition, frequency band, support type, and measurement points
- A residual-imbalance tolerance calculation by G class under ISO 21940-11, if rotor mass, speed, and correction radius are known
- A separate list of what balancing doesn't fix: shaft misalignment, bearings, electrical issues, resonance, loose fasteners, with recommendations on the order of work
- Saved influence coefficients for this machine: the next correction will run without trial weights

> It matters not to mix up three different assessments. The "within tolerance" line in the software means one thing: residual 1x is below the entered target value. The machine's overall condition is assessed by overall vibration and the ISO 20816 zones, while balancing quality is assessed by residual imbalance in g·mm/kg under the G classes of ISO 21940-11. Which part and edition of the standard applies to your machine we check separately: small units, belt-driven machines, and special designs carry exceptions. A breakdown of the three tolerances and how to choose the accuracy grade is covered in separate articles.

Sources: [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) · [ISO 21940-11:2016](https://www.iso.org/standard/54074.html)

## How much it costs and how to book

vibration diagnostics with a report costs EUR 300 per unit, balancing adds from EUR 250, and the minimum invoice for a visit is EUR 500. The calculator on our website gives an exact figure for your machine: it accounts for the number of rotors on site, the number of correction planes, access to the planes, the scope of diagnostics, and how far the site is. We work in Portugal, based in Vila Nova de Gaia near Porto, with visits across the country.

Diagnostics without balancing is a piece of work in its own right too. Sometimes that's the whole visit: we show that the cause is shaft misalignment, a bearing, or something electrical, and you save on weights that wouldn't have changed anything.

- Machine type, power, rated speed, and number of poles
- Line supply or VFD, operating setpoint frequencies
- What was done to the machine before the vibration appeared: a repair, rewinding, bearing replacement, a new fan or coupling
- Photos of the machine from both ends: end shields, fan cover, coupling, frame, and foundation mounting
- Any existing vibration readings, round-check reports, monitoring records
- Possible shutdown windows and site access requirements

> AXILINE is the engineers who design and manufacture the Balanset instruments and use them on-site themselves. That's why we're equally comfortable coming out to your site and selling the instrument to those who want to do it in-house, with consulting support for the specific machine. Routine balancing of a motor rotor is well within reach of an in-house maintenance team, and the hard cases usually come down to diagnostics, not calculation.

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

## Frequently asked questions

**Can a motor rotor be balanced without dismantling the machine?**

Yes, that's the primary method. The rotor is balanced in its own bearing supports at operating speed: sensors on the end shields, a laser phase sensor on the shaft mark, weights fitted through the fan cover or on the coupling. Dismantling is needed when not one correction plane can be opened up, when the rotor is going for repair anyway, or when G-class acceptance on a balancing machine is required.

**The motor started humming after rewinding. Is that imbalance?**

It could be, but it's worth checking more broadly. If the rotor was rewound (a wound rotor, an armature, the poles of a synchronous machine), the mass around the circumference changed directly, and balancing is almost mandatory. If the stator was rewound, the rotor was disassembled in the process, bearings were changed, and the fan was removed and refitted, so the source could just as well be assembly, a fit, or new air-gap geometry. We run one test with a power-off check and give you a definite answer.

**How do you tell that vibration is electrical, not from imbalance?**

Three signs. First: a noticeable peak at twice line frequency, that is, around 100 Hz on a 50 Hz supply. Second: sidebands around 1x and around 100 Hz spaced at the slip frequency multiplied by the pole count — that's the signature of a broken cage bar. Third, and most convincing: when power is removed, amplitude drops in a step within a fraction of a second, even though the rotor is still turning. Imbalance fades smoothly, in step with the coast-down.

**Where do correction weights go on a motor rotor?**

In descending order of convenience: the factory holes and blades of the cooling fan, washers under the coupling bolts, balancing rings and grooves if the manufacturer provided them, pressure plates and banding rings on armatures, welding on a plate where welding is allowed and won't heat the winding, and metal removal by drilling in a solid part. The laminated core, the squirrel-cage bars, the commutator, and the banding are never drilled or welded. Magnetic weights are only good as trial weights.

**Vibration rises as the motor warms up. Will balancing help?**

Most likely not, and that's an important sign. A 1x that rises with temperature, together with drifting phase, points to thermal bow of the shaft or rotor: uneven heating, a turn-to-turn short, blocked cooling passages, signs of rubbing. Balancing will make the machine quiet in one thermal state, and the vibration will come back in another. We sort out the cause of the heating first, then fit weights.

**The motor runs off a VFD at different speeds. Which one do you balance at?**

At the frequency where the machine runs most of the time, with the setpoint fixed and auto-adjustment disabled. Drifting speed corrupts the influence coefficients, and the calculation becomes essentially random. If there genuinely are two operating modes, we balance at the primary one and check the result at the second. While we're at it, we check whether either mode falls into a frame resonance — a common finding on VFD-driven machines.