On-Site Balancing of Commutator-Machine Armatures and Generator Rotors

An armature comes back from repair, and the crane starts humming again. Or a generator is nearly silent at idle but starts shaking the frame under load. These are two different machines and two different causes, but we're called out for the same question either way: balance it, or look elsewhere. We record the running-speed component's amplitude and phase at both bearing housings and answer that question before the first weight ever gets picked up.

Updated 27 August 2026 · by AXILINE · Vila Nova de Gaia

Symptoms: what brings people to call about armatures and generators

There's almost always a triggering event. A repair with a banding replacement, commutator turning, a new cooling fan on the shaft, a freshly pressed-on coupling half, moving the generator onto a different frame. Recall that event before you call: it cuts an hour off the diagnosis and immediately rules out half the possibilities.

• The armature came back from repair with new banding or after impregnation, and the vibration is higher than it was before the repair
• The brushes spark all along the rocker, the commutator is darkening in bands, and by hand the vibration feels stronger at the brush gear than at the bearing end shield
• The generator's vibration rises with the load and drops when the load is shed, while the speed stays unchanged
• The vibration rises together with the field current and keeps climbing as the rotor warms up
• Two runs in a row give different 1x amplitude and phase, even though the operating conditions are identical
• It all started after replacing the cooling fan on the shaft or pressing on a new coupling half
• The bearing on the commutator side moves noticeably less than the one on the drive side
• A diesel generator set shakes the frame, while the running-speed component on the generator itself is small
• A routine check measurement put the machine in zone C or D under the applicable part of ISO 20816 (elevated or unacceptable vibration)

The costly symptom here isn't the noise. Residual unbalance creates a rotating radial force that adds to the rotor's weight and the axial load. It isn't in the ISO 281 life calculation unless it was deliberately included, so the bearing on the side of the heavy plane fails ahead of its rated life, and it looks like a factory defect.

Sources: ISO 281:2007 · ISO 20816-1:2016

Design: where the unbalance in an armature or generator rotor comes from

On a fan, imbalance is almost always in one place; on an armature or a generator rotor it builds up from four or five small contributions at a large radius. Below is what we look for first.

• Armatures of crane, traction, and machine-tool DC machines: the rotor is long, with a length-to-diameter ratio noticeably above 0.5, so practically always two correction planes
• Generator rotors on diesel and gas sets: they run as part of a driven assembly, so the assembly's vibration and the rotor's vibration have to be separated before any calculation
• Generator rotors running at 1500 and 3000 rpm: the residual-unbalance requirements are tighter, because centrifugal force grows as the square of the rotation speed
• Rotors with a brushless exciter: there's an extra core with a diode bridge hanging on the shaft, an additional mass with its own imbalance and its own mounting

Commutator and slip rings: what balancing doesn't fix

This is the first fork in the road on any commutator machine. Commutator runout and armature imbalance produce similar complaints but are fixed by opposite means. Balancing doesn't remove runout, and, conversely, a perfectly turned commutator won't save you from unbalance.

The mechanism is simple. A brush has to keep contact with the running surface. Radial commutator runout, bars sitting at uneven heights, protruding mica insulation, or an oval shape after overheating make the brush bounce. Contact breaks, sparking appears, brush and bar wear increases, and the brush gear itself starts to shake. In the spectrum this lives high up: at the bar-passing frequency, meaning the number of bars times the rotation speed, plus its harmonics. At the bearing housing that peak can look modest, so we add an extra sensor closer to the brush rocker.

All of this is checked without an instrument. A dial indicator on the running surface while turning slowly gives the radial runout, the same indicator taken bar-by-bar shows the height variation, and a visual inspection shows the condition of the mica undercutting and the risers. The fix is mechanical: turning, undercutting, grinding, and seating the brushes to the commutator.

On generators with slip rings the picture is the same but simpler. There are two rings, they're smooth, and ring eccentricity or a worn groove produces sparking and scorching, not rotor imbalance. On a brushless machine this fork doesn't exist at all, and the diagnosis is shorter.

There's only one correct order of work on an armature: commutator mechanics first, balancing second. Fit the weights and then turn the commutator, and the metal removed will come off asymmetrically, forcing the whole correction to be recalculated. The general logic of tracking down a vibration source by spectrum is covered in the article on how to determine the cause of vibration; here we apply it to a commutator machine.

The generator within the assembly: the drive's contribution and electrical causes

A generator almost never runs alone. On one side there's a diesel, a gas engine, a turbine, or a power take-off shaft; on the other, an exciter and a load. The vibration you hear is a sum, and balancing removes only one term of it.

The prime mover produces its own frequencies, and they sit below the running speed. On a four-stroke diesel that's half the running-speed frequency and the firing frequency, equal to the speed multiplied by the number of cylinders and divided by two. Uneven cylinder operation, a bad injector, or unequal compression raise these low-order components. No weight on the generator rotor will touch them. A fatigued flexible coupling between the diesel and the generator, and sagged frame vibration isolators, belong in the same category.

On an electric motor, electrical causes are ruled out by cutting the power. A generator has no power supply to cut, but there are three conditions that can be run through in a single visit: rotating with the rotor unexcited, idling with excitation applied, and running under load in steps. Imbalance doesn't depend on excitation or load current at all, and gives the same 1x with a repeatable phase in all three conditions. Anything that changes together with the current points toward electrical causes and the rotor's thermal state.

We agree the mode checks and the excitation cut-off on coast-down with your staff, and only carry them out where it's safe for the machine and for the loads it feeds. At sites where the generator can't be unloaded, we limit ourselves to whatever mode you can give us, and state that plainly in the report.

Another feature of skid-mounted assemblies: a diesel generator set most often sits on rubber vibration isolators, meaning a compliant base, and the allowable vibration for it under the applicable part of ISO 20816 is higher than for the same machine on a rigid foundation. A sagged or cracked mount acts like a soft foot and rocks the whole frame. It needs to be checked before balancing, not after.

Sources: ISO 13373-3:2015 · ISO 20816-1:2016

What we check before we bring out the weights

The walk-around takes anywhere from half an hour to an hour, part of it done by hand on the stopped and locked-out machine, part of it instrumented on the running one. On armatures and generators it runs longer than usual, because there are more off-limits zones here than permitted ones.

• Banding: cracks in the varnish, signs of slippage, shifted alignment marks, a dull sound on tapping instead of a ringing one
• Commutator: radial runout and bar-height variation by dial indicator, condition of the mica undercutting, risers and signs of resoldering, the color and pattern of the surface glaze
• Slip rings: runout, worn grooves, scorching, brush wear and seating, and free play of the brushes in their holders
• Salient-pole rotor: tightness of the pole-mounting bolts, condition of the wedges and interpole spacers, and the fit of the field coils
• Cooling fan: cracks at the blade roots, the fit on the shaft after a repair, and carbon dust caked between the blades
• Coupling half: radial and axial runout by dial indicator, both keys in place, the condition of the flexible elements, and the fit on the overhung end of the shaft
• Shaft runout by dial indicator: residual bow after straightening, after overheating, or after long storage
• Bearings and fits: interference on the shaft and in the end shield, signs of a spun ring, and fluting on the races from current passing through the bearing
• The mounting to the frame and foundation, the condition of vibration isolators on skid-mounted assemblies, and shaft alignment between the generator and the prime mover
• Temperature trend: a reading on the cold machine, repeated after 30 or 60 minutes running under load

Loose banding and a loose pole give themselves away the same way: an unstable phase. That's why, before any trial weight, we always run two consecutive runs in the same condition. If the 1x phase drifts by more than about ten degrees between them, balancing is pointless: the influence coefficient — the relationship of "weight added, vibration changed like this," which the whole calculation rests on — would be obtained for one state of the rotor while it would end up working in a different one.

How the on-site work proceeds

1. Step 1

   Access, inspection, lockout

   We establish who operates the machine and who locks out the start circuit, where the key is kept, and how the fan cover comes off to give access to the commutator, the brushes, and the coupling half. For a generator we separately find out whether it can be unloaded and de-excited, and which loads would feel it. We work through the walk-around using the checklist above.

2. Step 2

   Sensors

   Two accelerometers on the bearing housings, as close to the bearings themselves as possible: on a boss or rib of the end shield, on a cleaned, flat spot, by magnet or on a stud. The direction is radial, usually horizontal, kept the same on every run. On a commutator machine we add a check reading near the brush rocker, to catch brush bounce. A sensor on the cover, the terminal box, or a guard produces large numbers that have nothing to do with the rotor.

3. Step 3

   Speed and the reference phase

   We stick the reflective marker on a degreased section of the overhung shaft end or on the coupling half, and aim the laser sensor at it. There must be exactly one marker; a second stripe would double the indicated speed. We make sure the laser doesn't catch a stray reflection from the keyway, a chamfer, or the polished surface of a slip ring.

4. Run 0

   Baseline measurement

   Operating speed and operating conditions. We record speed, overall vibration in mm/s RMS (root-mean-square value), 1x amplitude and phase at each bearing, and the spectrum and time waveform. Axial vibration at the coupling is measured separately, once.

5. Modes

   Unexcited, idling, under load

   For a generator this is the key run. We repeat the same measurement in three states and compare the running-speed component. If it doesn't change, the cause is mechanical. If it rises with field current or with load, diagnostics follows next, not weights.

6. Separation

   Imbalance, commutator, electrical, or the prime mover

   We compare 1x against overall vibration, look at the low-order components from the prime mover, the commutator bar-passing frequency, and twice the line frequency, and check phase repeatability with two runs. Only after that do we say balancing is appropriate.

7. Run 1

   Trial weight in the first plane

   We weigh the trial weight, fit it at the marked point just as securely as we would a permanent one, and enter the actual mass and actual radius. We treat a trial run as valid if the 1x amplitude changed by at least 20-30 percent or the phase by at least 20-30 degrees. If the change is small, we increase the weight and repeat.

8. Run 2

   Trial weight in the second plane

   For a two-plane calculation we move the weight as directed by the software. In total: one plane needs two runs, two planes need three. On an armature it's almost always the two-plane scenario.

9. Correction

   Masses, positions, fastening

   The software outputs a mass and a position for each plane, with the zero reference set where the trial weight was. On the fan blades, the coupling-half bolts, and the pressure plate's threaded holes we work in fixed-position mode: the instrument states a position number and a mass, and an error in the direction of reference is ruled out.

10. Check

    Verification run and trim

    Same speed and same conditions as the baseline run. The running-speed component should drop by a large factor. If the target isn't hit right away, the software calculates trim masses to add to the ones already fitted, and one or two iterations are usually enough. On a generator we also repeat the verification measurement under load.

11. Report

    Recording the result

    Before-and-after numbers for each bearing and direction, speed, operating mode, field current and load at the time of each measurement, temperature, and the masses, radii, and positions of the weights. The influence coefficients stay in the instrument's archive, and a repeat correction on this machine will go ahead without trial runs.

We work with the Balanset-1A: two accelerometers, a laser phase sensor reading a reflective marker, a two-channel USB module, and software on a laptop. Calculation in one and two planes by the influence-coefficient method, overall vibration and 1x, phase, speed, spectrum and time waveform, fixed-position mode, drilling calculations for metal removal, tolerance assessment against grades G (balance quality grades), and an archive with reports.

Sources: Balanset-1A operation manual · Balanset-1A manufacturer specification

Correction planes on an armature or generator rotor: where they're permitted, and what to do if there are none

An armature is long, with a length-to-diameter ratio noticeably above 0.5, so the practical answer is almost always two planes. A single mass on a rotor like this can't cut vibration at both bearings at once; a couple component will remain. The rule for choosing the number of planes itself is covered in a separate article; what matters more here is that on this rotor the planes are dictated not by calculation but by what's actually allowed to be touched at all.

Permitted spots are few, and they're always substantial, standard parts. Anything belonging to the machine's active parts, to the winding, or to its insulation, drops off the list immediately.

Sometimes there's not a single accessible plane. A cast, non-removable cover, the commutator on one side and a solid end shield on the other, a coupling guard that only comes off by taking the assembly apart. There are then four options, and we work through them in this order. First: look for standard threaded holes in places they're not expected — on the pressure plate, on the fan rim, on the end face of the pole-system rim. Second: work in a single plane on the accessible side and say up front that a residual will remain at the far bearing. Third, on skid-mounted assemblies: use the prime mover's flywheel, which has bolts around its circumference, but that's a plane belonging to the assembly, not the generator rotor, and it doesn't solve every case. Fourth: stop at diagnostics and send the rotor to a workshop, where standard correction spots will be machined and a grade-G document issued.

• Pressure plates and winding retainers on the armature: solid metal, often with standard threaded holes, allowing both bolted weights and metal removal by drilling
• Balancing grooves and rings, if the manufacturer provided them: a sector weight is dropped into the groove and locked in place, with no drilling needed
• The end faces of the pole shoes and the pole-system rim on a salient-pole rotor: only through standard holes and only with the manufacturer's agreement
• The retaining rings on a non-salient-pole (cylindrical) rotor, used as a mounting spot for a sector weight, without touching the banding itself
• The rim and blades of the shaft-mounted cooling fan: a bolted weight through a standard hole, the most accessible plane on the side opposite the drive
• Coupling-half bolts: calibrated washers under the heads, the fastest method on the drive side
• The prime mover's flywheel on a skid-mounted assembly: an accessible plane, but one that belongs to the prime mover, not the generator rotor

• Permitted: bolted fastening through the standard holes of the fan, the pressure plate, and the coupling half; washers under the coupling-half bolts; sector weights in balancing grooves
• Permitted with a caveat: welding on a plate only onto substantial parts, well clear of the winding, without heating the core, and with the bearings protected from welding current
• Permitted with a caveat: metal removal by drilling only in the pressure plates, the fan rim, and other substantial parts, with the software calculating the depth and diameter
• Prohibited: drilling the rotor core — the magnetic symmetry would be disrupted, producing vibration at twice the line frequency that nothing can remove afterward
• Prohibited: drilling, heating, or trimming the commutator, the risers, the slip rings, and the damper-winding bars
• Prohibited: touching the banding, the winding end turns, the field coils, and their insulation — any intervention here voids the repair warranty
• Magnetic weights are only good as trial weights, for the duration of a measurement, and only where the stray magnetic field won't shift them

A separate note on high speeds. If the operating speed is close to the first critical frequency — the speed at which the shaft goes into resonance — the rotor starts to bow, and two planes may not be enough: a correction that removes vibration at the operating speed will worsen behavior during run-up. We check whether the ordinary low-speed procedure applies to rotors like this against the current edition of ISO 21940-12, and say plainly when the machine belongs on a balancing machine instead.

Sources: ISO 21940-12:2016

When on-site work won't succeed

When it's more honest to pull the rotor. If the machine is going for a teardown anyway, if a grade-G balance-quality document is needed, or if the rotor is flexible or has no accessible planes, a workshop balancing machine is more accurate and, in the end, cheaper. A comparison of the two approaches is in a separate article on on-site versus workshop balancing, and after reassembly we take a check measurement and a trim correction if needed.

What you get, and how to book

What comes out isn't just a quiet machine but numbers you can keep working with: compare against the next measurement, show to a repair contractor, or build into a maintenance plan.

Vibration diagnostics with a report costs 300 EUR per unit, balancing adds from 250 EUR, and the minimum invoice per visit is 500 EUR. The calculator on our website gives an exact estimate for your machine: it accounts for the number of rotors on site, the number of correction planes, access to them, the scope of diagnostics, and distance. We're based in Vila Nova de Gaia near Porto and travel throughout Portugal. Diagnostics without balancing is also a full result of the visit: sometimes the whole outcome is that you don't spend money on weights, and instead take the commutator off for turning.

• A report with the initial and residual vibration for each bearing and direction, with the overall level in mm/s RMS and the running-speed component listed separately
• Spectra before and after, showing exactly what dropped and what remained
• Speed, operating mode, field current, and load at the time of each measurement, plus a diagram of the sensor mounting points
• Masses, radii, and positions of the fitted weights, with a description of the fastening method
• Commutator or slip-ring runout readings, if taken, and a conclusion on the brush gear
• An assessment of overall vibration by zone under the applicable part of ISO 20816, stating the edition, frequency band, support type, and measurement points
• A residual-unbalance tolerance calculation against grades G from ISO 21940-11, if the rotor mass, speed, and correction radius are known
• A separate list of what balancing doesn't remove, with a recommended order of work
• This machine's saved influence coefficients: the next correction will proceed without trial runs

• Machine type: a DC machine armature, or a generator rotor — synchronous salient-pole, non-salient-pole, with slip rings, or brushless
• Power, rated speed, voltage, and drive type; for a skid-mounted assembly, the diesel type and number of cylinders
• What was done before the vibration appeared: a repair, banding replacement, commutator turning, bearing replacement, a new fan, or a new coupling half
• Photos from both sides: bearing end shields, the commutator or slip rings with the cover removed, the fan, the coupling half, the frame, and the vibration isolators
• Any existing vibration readings, a workshop rotor-balancing report, and walk-around logs
• Whether the machine can be unloaded and stopped, what time windows are available, and whether a repair contractor's warranty is in effect

AXILINE is engineers who design and manufacture the Balanset instruments and do the on-site balancing with them ourselves. That's why we're equally at ease coming out to a site or selling the instrument to those who want to do the work themselves. A standard-procedure armature balance is within reach of an in-house maintenance team, and the difficult cases on these machines are almost always diagnostic ones: the commutator, the banding, excitation, the prime mover. Don't conflate three separate assessments: the tolerance line in the software means only that residual 1x is below the entered target; the machine's condition is assessed by overall vibration and ISO 20816 zones; and rotor balance quality is assessed by residual unbalance in g·mm/kg against grades G from ISO 21940-11.

Sources: ISO 20816-1:2016 · ISO 21940-11:2016 · Balanset-1A manufacturer specification

Frequently asked questions

Related content