On-site balancing of flywheels, drive discs, and brake discs, in place

A flywheel forgives a lot while it's standing still, and forgives nothing once it's turning. A hundred kilograms of cast iron with its centre of mass off by a few tenths of a millimetre can shake a press hard enough to loosen the foundation bolts. We come out and balance flywheels, drive discs, and brake discs right on the machine, in its own bearing housings. Base in Vila Nova de Gaia near Porto, service visits across Portugal.

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

What brings us out: symptoms and events

For a flywheel, imbalance is almost always tied to an event. It was resurfaced after a clutch repair. The ring gear was replaced. It was removed and refitted while changing a seal. A disc that was balanced at the factory years ago ends up with a new mass distribution after any of these jobs, and the machine reports it on the very first start.

The second group of call-outs comes with no event at all: vibration builds up over months. Here the culprit is more often the fit, not the mass. The flange wears loose, the bolts stretch, and the flywheel starts sitting eccentric. In both cases the instrument shows a peak at the running frequency — called 1x, once per revolution — so we tell the causes apart with a dial indicator and measurement, not by eye.

• Noise and shaking appeared after the flywheel was resurfaced, the ring gear was replaced, or the clutch was repaired
• Vibration at the main-shaft or crankshaft bearing supports grows roughly with the square of speed
• The brake disc runs out, the pads wear in a wedge pattern, the disc heats up in patches
• Rising noise during run-up and coast-down, the whole machine housing shakes at operating speed
• The flywheel mounting bolts keep loosening, and rust-coloured fretting dust — wear debris from a loose joint — comes out of the joint
• You're replacing the bearing supports on the flywheel shaft more often than the others
• After reassembling the flywheel-plus-half-coupling-plus-pulley unit, vibration is higher than before it was taken apart

Absolute vibration limits depend on the machine and the stiffness of its supports; check them against the applicable part and edition of ISO 20816. For deciding whether a visit is needed, the trend is enough: if the level at the supports has risen one-and-a-half to two times after work on the flywheel, the cause is worth looking for in whatever hands touched.

Sources: ISO 20816-1:2016

Why a heavy disc is sensitive to grams

Centrifugal force grows linearly with mass and with the square of speed. One gram at a radius of 250 mm pulls with a force of about 6 N at 1500 rpm, and almost 25 N at 3000 rpm. Twenty extra grams on the rim at 3000 rpm is half a thousand newtons of rotating load that changes direction every revolution and passes entirely through the bearing supports. Bearing life calculated to ISO 281 won't see this force unless it was built into the calculation.

Now about tolerance. Flywheels usually use balance quality grade G6.3; some machines require G2.5 — check applicability against the current part and edition of ISO 21940. At 1500 rpm, grade G6.3 means a residual centre-of-mass offset of about 40 μm; at 3000 rpm, about 20 μm. For a 150 kg flywheel that's only 20–25 g at the rim radius. A heavy disc is balanced with small weights, and a step of a few grams already moves the vector noticeably.

There's a flip side too. A flywheel effectively has just one correction plane — the rim — and its diameter is small relative to the disc's mass. Those same grams need to go on at exactly the right angle: being off by one position sends part of the correction astray. That's why we work with precise weighing, fixed positions, and a verification run after every correction. The Balanset-1A calculates the weight using the influence coefficient method — from the machine's response to a trial weight — and checks the tolerance against G grades right inside the software, so the report shows not just "it got quieter" but a value landing inside the tolerance field, in g·mm.

Sources: ISO 21940-11:2016 · ISO 281:2007 · Balanset-1A manufacturer specification

Fit and runout: a separate cause that weights can't fix

A flywheel sits either on a crankshaft flange with a centring spigot and bolts on a bolt circle, or on the shaft end with a key or a taper, as on presses. Both arrangements have their own routes to vibration that have nothing to do with mass distribution. We check them before the first run, with the machine stopped and locked out.

The sign of a loose fit on the measurement: the 1x phase — the angular position of the vibration relative to the mark on the shaft — drifts by tens of degrees between runs at the same speed. When that happens, we stop the balancing and show what needs repair. How to tell imbalance from shaft misalignment by phase at the supports is covered in a separate article; where there's a coupling to the drive, we separately assess shaft alignment.

The assembly: flywheel, half-coupling, pulley

On a real machine, a flywheel rarely lives alone. The same shaft carries a half-coupling, a drive disc, sometimes a pulley. The imbalances of the individual parts add up as vectors, and an assembly built from parts that were each balanced separately can easily end up outside tolerance. So we balance the whole assembly, in its own bearing supports, at operating speed.

When we need to work out exactly which part is contributing the residual, index balancing helps: the part is repositioned relative to the shaft by a known angle, the measurement is repeated, and from the shift in the vector the instrument separates the part's contribution from the shaft's. The Balanset-1A runs this procedure as a standard feature. It also helps with assemblies that get taken apart regularly for maintenance: each part's contribution is known, and after reassembly a short trim balancing pass using the saved coefficients is enough, instead of a full cycle with trial runs.

• Before disassembly, we mark the relative position of the flywheel, half-coupling, and shaft with a punch mark or paint
• We reassemble in the same angular position: a different position means a new imbalance
• After resurfacing, a ring gear replacement, or a clutch job, we balance the assembly again; the old result no longer applies
• A repeat visit to the same machine is shorter: saved influence coefficients save on trial runs

Sources: Balanset-1A operation manual

How the work goes on site

1. Inspection

   Access and geometry

   The machine is stopped and locked out. We remove the cover or open the housing window, and decide where the weight will go and where drilling is allowed. We take radial and face runout readings with a dial indicator, and check bolt torque and the fit.

2. Sensors

   Supports and mark

   Two accelerometers on the bearing supports of the flywheel shaft, radial direction, on bare metal near the bearing. We attach a reflective mark to the rim or shaft, within sight of the laser phase sensor. We route the cables clear of rotating parts.

3. Run 0

   Baseline measurement

   Operating speed, steady-state conditions. We record overall vibration in mm/s RMS (root mean square), the 1x amplitude and phase at both supports, the spectrum — the breakdown of vibration by frequency — and the time waveform. If harmonics or bearing frequencies dominate rather than the running-speed component, we say so right away.

4. Marking

   Fixed positions

   We number the standard holes, bolts, or spokes in the direction of rotation. We enter the number of positions and the actual radius into the software. The instrument responds with a position number and a mass, so it's impossible to mix up the angle or the reference direction.

5. Trial run

   Small weight, one run per plane

   We fit a weighed trial mass at the first position. On a heavy disc, the response to grams is noticeable, so we keep the trial weight small. A valid run changes the 1x amplitude by 20–30 percent or the phase by 20–30 degrees.

6. Correction

   Weight or drill, verification, report

   The software outputs the mass and position, or, for metal removal, the hole diameter and depth. We fit the weight or drill, run a verification pass, and if needed add a small weight to what's already fitted. The report includes speed, before-and-after vectors, masses, positions, and the assessment against the G grade.

A flywheel with normal access to the rim in a single plane is usually wrapped up in two to four hours, including stops. What eats the time is covers and machine starts, not the calculation. Thick press flywheels and assemblies with widely spaced planes sometimes need a second plane, which is one more run.

Sources: Balanset-1A manufacturer specification

Mounting weight on a heavy disc: why we drill instead of welding

Requirements for weight mounting on a flywheel are tighter than on a light pulley. The mount has to hold the centrifugal force with a large safety margin: those same 20 g on the rim at 3000 rpm are loaded with fifty-odd newtons, and every start and stop adds another loading cycle. Adhesive isn't a mount at all here; magnetic weights are used only as trial weights and are removed before the machine is handed back.

We don't weld on flywheels, and here's why. Most flywheels are cast iron, and cast iron doesn't take well to welding: the weld and the heat-affected zone develop cracks, often with a time delay. The arc heats the disc locally and warps its geometry, and runout drifts along with it. Weld-deposited metal can't be dosed precisely: put on five grams too many and you have a new imbalance. And most importantly, a welded weight on a rim moving at 50–90 m/s is an object that cannot be given even the smallest chance of coming loose.

Removing metal by drilling is free of all that. Mass is dosed by hole diameter and depth, and the Balanset-1A calculates these right in the software. There's no thermal effect, and the geometry doesn't drift. There's nothing to come loose: a hole isn't a separate part. We drill the rim from the face or around the periphery, wherever the design allows, clear of the ring gear, friction tracks, and stiffening ribs. Weight-mounting methods in general are compared in a separate article; the conclusion that matters here is that for heavy discs, drilling is almost always the better option.

• Bolt-on weight only into the standard holes or onto threaded studs, with locking: a lock nut, a lock washer, or thread locker
• We weigh the weight before fitting it, and enter the actual mass and radius into the software
• We fit it at a radius that can be reproduced at a repeat balancing
• No weights on the friction surfaces of brake discs or on clutch friction tracks

Sources: Balanset-1A operation manual

Brake discs: heat matters as much as mass

An industrial machine's brake disc lives through thermal cycles: seconds of braking heat it up by hundreds of degrees, then it cools back down. Uneven heating warps the disc, and runout grows along with temperature. So vibration that wasn't there in the morning but is there by the end of the shift more often points to warping than to imbalance.

We separate these causes with measurement: a cold machine and a machine after a working cycle. A 1x vector that stays stable across both measurements means imbalance; one that grows with warm-up means geometry. We correct only outside the friction surfaces: the hub, the mounting flange, the standard holes. Drilling or weld-building the friction track is not allowed. After resurfacing, we balance the disc again: the machine tool removed metal relative to its own mounting axis, not the machine's axis of rotation.

Adhesive and magnetic weights don't survive on brake discs; the thermal cycle removes them. Only mechanical mounting outside the friction zone works. The general method for tracking down the causes of vibration from the spectrum is covered in separate diagnostic articles.

Sources: ISO 13373-3:2015

Safety: energy stored in the rim

A flywheel stores energy — that's its job. A 200 kg disc, 600 mm in diameter, at 1500 rpm carries around 100 kJ, about as much as a passenger car at just over fifty kilometres an hour. The rim's peripheral speed at that point is around 50 m/s, and twice that on high-speed machines. Any object that leaves the rim flies off tangentially at that speed.

Hence the strict procedure. Anything done by hand is done with the machine stopped and locked out, with a check for a full coast-down: a heavy flywheel keeps turning for a long time after it's switched off. Trial and correction weights are locked in place. A start after fitting any weight happens only with the cover closed, people are cleared out of the plane of rotation, and the tangential zone off an open rim is kept empty during any start. A detailed breakdown of lockout and the throw zone is in a separate article on safety in on-site balancing.

When on-site won't work, pricing, and visits

On price: packages from 550 EUR per unit (diagnostics 300 EUR + balancing from 250 EUR), minimum invoice per visit 500 EUR, and the calculator on the site works out the exact figure for your machine. We are the engineers who design and manufacture Balanset instruments, and we do the on-site balancing ourselves. Base in Vila Nova de Gaia near Porto, we work across all of Portugal. Send us photos of the flywheel, the cover, and the machine's nameplate: we'll tell you in advance whether a weight will fit, whether drilling is needed, and what we'll need from you on site.

• The fit is worn out, the flange or keyway is worn: repair comes first, weights won't hold the result
• Cracks in the rim, hub, or ring gear: balancing isn't possible, the part gets replaced
• There's no access to the rim even through a cover window, and nowhere to put the weight: disassembly is needed, sometimes with machine balancing in the shop
• The machine doesn't reach a stable speed, or runout grows with warm-up: the cause is fixed first, then balancing
• The dominant vibration isn't 1x but bearings, shaft misalignment, or resonance: we'll show you the spectrum and tell you what to fix instead of adding weights

Frequently asked questions

Related content