Preparing equipment for on-site balancing: a customer checklist

The engineer arrives with an instrument, not a repair crew. Everything they can get done in a shift depends on your preparation: whether the magnet holds on the support, whether the impeller is clean, who's authorized to press 'start,' and whether the machine can be stopped five times in a row. Visits fail not on the calculation, but on not being able to reach the correction plane — the location on the rotor where the correction weights go — without disassembly. Below is a single checklist and an honest list of the conditions under which the work gets postponed.

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

The single checklist: go through it the day before the visit

• Mechanics. Bearings with no critical play, feet bolted down, a frame with no cracks or loosened joints, the rotor not rubbing anywhere and turning smoothly.
• Cleanliness. The impeller, wheel, pulley or drum cleaned of build-up, dust and product in advance, not on the day of the work.
• Sensor mounting spots. A clean, flat spot prepared at each bearing housing: the magnet needs to seat on bare metal, not on paint over dirt.
• Tachometer marker. There's a place on the shaft or hub for the reflective marker, and the laser can see it from the tripod.
• Correction planes. You can reach the weight-installation points with the rotor stopped, without disassembling the machine.
• Hatches and guards. It's clear who removes the guard, who puts it back, and what paperwork covers that.
• A window for the runs. A window for 3–6 runs is agreed, with a full stop after each one until the rotor is completely still. All runs at the same operating mode.
• An operator. Someone authorized to start and stop the machine is on site, and won't be pulled onto another task an hour in.
• Work permit and lockout. Paperwork is arranged before the visit, and the lockout scheme is clear to everyone involved.
• Power and light. A 220 V outlet within 5–10 meters of the bearings, working light at the platform and at the correction planes.
• Tools and materials. Wrenches, a drill, plates and washers for the weights, bolted fasteners or a welder if the weights have to be welded on.
• Machine data. Nameplate speed, drive power, rotor mass, type of supports and foundation, repair and prior-balancing history.
• Photos and video. Shots of the rotor, both bearings and the correction planes, plus a short video of it running, with sound.
• A draft report. The machine's designation on your own scheme, measurement points, operating mode and the target vibration value.

The list looks long, but it takes a mechanic about an hour. One item out of these fourteen left undone usually costs you anywhere from half a shift to a full repeat visit.

Mechanical soundness: balancing doesn't substitute for repair

The correction calculation relies on the 'rotor — supports — foundation' system behaving linearly: double the unbalanced mass, and vibration doubles. The entire math of the influence-coefficient method rests on this assumption: the software calculates what correction weight to install, and where, from the machine's response to a trial weight. A worn-out bearing, a loose wheel fit on the shaft, or an unbolted frame breaks that linearity, and the instrument gets contradictory answers from the machine to the same input.

The simplest way to see this is through the stability of the readings. While the machine runs at a constant speed, the amplitude and phase of the running-speed component (the vibration at the rotor's rotating frequency; phase is its angle relative to the shaft marker) shouldn't drift by more than 10–15% over the measurement window. If the numbers are jumping, going further is pointless: the influence coefficient will come out random, and the weights will land in the wrong place.

A separate case is a machine that's simply standing on the floor or on shims, with no anchors. Under noticeable vibration, the unbalance force lifts the unit, the system's stiffness changes within a single run, and the result doesn't repeat from run to run. That kind of machine needs bolting down, not balancing.

Check one more thing by hand. Turn a horizontal-axis rotor 90° and let go. If it consistently turns back to the same position, the rotor has pronounced static unbalance: the heavy spot outweighs the rest and drops downward. That's not a reason to cancel the work, but it's useful for the engineer to know in advance: the first run on that kind of machine is done cautiously, at a reduced speed.

• Play in the bearings. Noticeable radial play, rumbling, knocking or overheating mean a repair is needed, not balancing.
• Fasteners and soft foot. All foot bolts are torqued, the shims are thick enough and don't compress, there's no gap under the foot.
• Frame and foundation. No cracks, no delaminated grout, no loosened anchors, no 'temporary' welded-on reinforcements.
• Rubbing. The rotor doesn't touch the housing, seals, inlet duct or guard mesh in any position.
• Shaft fit. The wheel, pulley, auger or coupling half is seated tightly, with no play and no chipped keyway.
• Stable speed. The drive holds its frequency: the belt doesn't slip, the variable-frequency drive doesn't wander, the load is steady.
• Misalignment. On a machine joined by a coupling, shaft alignment is done before balancing, not after.

We say this plainly here because it saves you money. Balancing only removes the vibration produced by an uneven distribution of mass around the rotation axis. It won't remove bearing knock, vibration from shaft misalignment, or the 'comb' of harmonic-multiple frequencies produced by loose fasteners — not with any weight.

Sources: Balanset-1A operation manual

Rotor cleanliness: dirt comes off right during the runs

A dirty rotor balances poorly not because the dirt is heavy, but because it isn't constant. A layer of built-up dust, product, scale or wet material clings to the blades unevenly and breaks off in chunks during run-up and coast-down. Every such loss changes the unbalance by a few grams, and the instrument sees a machine that's become a different one between two runs.

In practice it looks like this: the trial weight gave a clear response, the software calculated the correction, you installed the weights, and the check run showed vibration no lower than at the start. The engineer adds a trim weight (a small refining one), and the level drifts again. That isn't a calculation error. It's a chunk of build-up that broke off inside the exhaust fan or the fan on a dusty section.

So plan cleaning before balancing, not after. Wash or scrape the impeller, remove the film from the blades all the way around, clear the buildup off the hub and the back of the disc. If you clean the wheel only partway, you've created a new unbalance yourself, so clean either everything or nothing.

Erosion behaves differently, but with a similar outcome. An abrasive stream wears the blades unevenly, and after balancing the vibration level starts creeping up on its own. Balancing here honestly removes the vibration today rather than solving the underlying problem. If an impeller is worn noticeably and unevenly, replacing it is cheaper than sending an engineer out every quarter.

Sometimes a shutdown for cleaning costs more than the balancing itself, and you decide to balance 'as is.' That's done too, and we say so plainly: the level will come down, but the result only lasts until the next noticeable build-up. Record the before-and-after numbers so you have something to compare against a month later.

Access: sensor mounting spots, the tachometer marker, correction planes

Check access physically, not from a drawing. 'There's a hatch there' and an open hatch with the bolts already backed out describe two different states of the machine. If a hatch has been welded shut and the guard bolts are seized, say so in advance, and we'll come with a different plan.

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

Runs, power and tools: how many stops to plan for

On-site balancing is made up of runs, not measurements. The measurement itself takes minutes; the stops are what eat up the time: waiting for the rotor to fully stop, installing or removing the weight, closing the guard, restarting the machine and getting back to the same operating mode.

Count it this way. Balancing in one plane: an initial run, a trial-weight run, a check run. Three runs minimum. Balancing in two planes: an initial run, two trial runs, a check run. Four minimum. Add one or two trim-balancing runs if you don't land on the target value the first time, and you get exactly the 3–6 runs you need to agree with process engineers in advance.

All measurements are taken at a constant speed and one operating mode. If you change the damper position, the load or the temperature between runs, you're comparing two different machines, and the previous data is reset. Fix the operating mode in advance and record it in the report.

The operator controls the runs, not the visiting engineer. Agree in advance who that is, where they'll be, and how you'll stay in contact with them: direct line of sight, radio or phone. Half the time lost on a visit goes into finding the person with the key to the control panel.

• 220 V power within 5–10 meters of the bearings. If the mains quality is poor, the engineer will switch to the laptop's battery, but it's better to know that in advance.
• Working light on the platform and inside the housing, so the correction plane and position markings are visible.
• Mechanic's tools: wrenches for the guard and foot bolts, a drill, if the correction is done by removing material.
• Material for the weights: steel plates, washers, bolts of the right length. A weight is fastened as securely as a permanent one, from the very first run.
• Welding and a welder, if the weights are being welded on. Arrange hot work clearance separately in advance — it's a common cause of losing half a day.
• Scales on hand. The mass of the trial and correction weights goes in as the actual measured value, not 'about ten grams.'

A trial weight is sized so the running-speed amplitude changes by at least 20–30%, or the phase by at least 20–30°. If there's no response, the weight is increased and another run is made. Build that margin into the window: an announced 'exactly three runs' sometimes turns into five.

Sources: Balanset-1A operation manual

People and safety: who's responsible for what

Balancing happens on a running machine with the guard open or partly removed, so organizing the work matters here more than the instrument does. Name two people in advance and introduce them to the engineer in the first few minutes: the person responsible for work safety, and the person responsible for starting and stopping. These can be different people, and both need to be on site the whole time.

The guard is removed and replaced only with the rotor stopped. While someone is working at the correction plane, the start control has to be physically locked out — a verbal agreement isn't enough. The key or lockout tag stays with the person working inside, and no one else.

Fastening the weights is also a safety matter, not just an accuracy one. A trial weight is held with a bolt, a clamp or a tack weld. Tape, putty, a magnet or a cable tie won't do for any run: a weight that comes loose at operating speed will go through the guard. And the weight itself must not touch stationary parts or overload its mounting point.

• A work permit or authorization is arranged before the visit, not on the day of the work.
• A person responsible for work safety is named, and knows about the removed guard.
• A person responsible for starting and stopping is named, on site and reachable the whole time.
• The lockout scheme is clear: who locks it, who unlocks it, where the key or tag is.
• PPE on everyone involved: helmet, safety glasses, gloves, hearing protection near a loud machine.
• Hot work is cleared separately, if the correction weights are being welded on.
• A communication method between the site and the control panel is agreed: direct line of sight, radio or phone.
• It's clear where people step back to during a run, and where the danger zone is marked.

The engineer has the right to stop work if the start control can't be locked out, or if the guard is being removed on a rotor that's still turning. That isn't a formality or something to negotiate: you don't balance under an open hatch on a fan that's running.

Machine data and a draft report

Machine data isn't needed for paperwork's sake. Every figure answers a specific question before the visit: what fasteners to bring, how many correction planes to plan for, what trial weight to use, and what criterion to use for calling the job done. Send it in advance, and you'll save the first hour of the visit, which would otherwise go into measuring with a tape and calling the chief mechanic's office.

We'll separately ask for photos and a short video. Photos of the bearings show whether a magnet will hold and where the cable will run. A photo of the correction plane shows how many fixed positions there are. A video with sound often reveals things no email mentions: a whining bearing, a rhythmic knock, the sound of rubbing.

• The machine's designation and position on your own scheme, so the report matches your documentation.
• Measurement points with names and directions: bearing 1 horizontal, bearing 1 vertical, and so on.
• The operating mode during the readings: speed, load, damper or valve position, temperature.
• The measured quantity and frequency band for assessment: mm/s RMS in the 10–1000 Hz band as a working benchmark.
• The target vibration value or the balance quality grade G you want to reach.
• Who signs the report on your side, and in what form you need it.

Fix the frequency band, measurement points, operating mode and the exact part and edition of the standard before work starts, if the result is going into acceptance under a contract. Use ISO 20816 and ISO 21940-11 as a working reference and check applicability to your own machine: there are exceptions by power, speed and machine type.

Sources: ISO 20816-1:2016 · ISO 21940-11:2016

What to prepare, why, and what happens if you skip it

This table is easy to forward to the section mechanic as it stands. The right-hand column usually convinces better than the left one.

When work gets postponed or stopped, and how to arrange a visit

An honest conversation before the visit is cheaper than an honest conversation on site. There are conditions under which the engineer won't start the work, or will stop it partway through, and it's better to check them against this list in advance.

• The mechanics aren't sound. Critical bearing play, a rubbing rotor, a crack in the frame, a loosened anchor. Repair first, then balance.
• The machine isn't bolted to its base. It sits on the floor or on shims with no anchors and lifts under vibration. Bolt it down, then balance.
• Speed is unstable. The belt slips, the drive doesn't hold its frequency, the load wanders. Readings at a varying speed aren't valid for the calculation.
• The operating speed falls in a resonance zone: the rotation frequency coincides with a natural frequency of the structure, and vibration spikes sharply. The phase of the running-speed component shifts from run to run, and the result doesn't repeat. Either the mode or the mounting conditions need to change, or balancing is pointless.
• No access to the correction plane. There's nowhere to put the weight without disassembly. Either disassemble on site, or remove the rotor and balance it on a machine.
• No clearance for runs. No work permit, no one responsible for starting, no way to lock out the panel. The work gets postponed.
• The machine can't be stopped the needed number of times. A continuous process with no window. Look for another window, usually a maintenance shutdown.
• Site conditions are outside the operating range. The instrument works at +5 to +50 °C and up to 85% humidity without condensation. Frost, rain on an open platform, or condensation dripping down are reasons to postpone.
• The vibration isn't running-speed in nature. Aerodynamic or electromagnetic force grows in proportion to speed, while a weight's centrifugal force grows with its square. That kind of vibration can only be compensated at one specific operating mode, and we say so before the work, not after.
• The rotor is flexible and operates above its first critical speed — the speed at which the rotor starts bowing noticeably. That calls for a different method and a different conversation.

If you've gone through the checklist and something doesn't add up, don't cancel the visit quietly. Write in and say exactly what isn't working: the plan often changes on site without losing a day. The software can recalculate an inaccessible second correction plane onto accessible ones, and instead of an angle with a protractor, you can work by blade number.

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

Frequently asked questions

Related content