Safety during on-site balancing: start-up, lockout, weights, and the throw zone

You start a machine with a guard removed, add mass to a spinning rotor, and approach the bearing housings while someone stands nearby with a hand on the controls. No correction calculation is worth an injury, which is why how the fieldwork is organized matters more than the instrument. Below we cover who's in charge of what, how a start-up gets locked out, what a weight should be mounted with, and what to do when vibration goes wrong during run-up. At the end: a checklist for before the first run, and a list of conditions under which work stops.

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

Your own person runs the controls, not the technician

The first thing we ask for on site is two names: the person responsible for work safety, and the person responsible for start-up and shutdown. The second person needs to be physically present for the whole shift, because balancing isn't really made up of measurements - it's made up of starts and stops. There will be anywhere from three to six of them, and each one needs a decision from someone who actually has the authority to make it.

The technician doesn't touch the controls. That's not a formality - it's because they don't know your machine as a process asset. They can't see whether the damper is closed, whether cooling water is flowing, whether the pump is primed, or whether work is under way on a neighbouring section of the same line. They don't know your interlocks, and they don't know that a setpoint-triggered auto-start in your control system can issue a command without their involvement. Starting the machine is a technical decision inside your own production process, and it stays yours.

There's a second, more practical reason. A visiting engineer doesn't hold your site authorization and isn't in your logbook. If something goes wrong during a start-up they performed themselves, you're no longer dealing with a normal operational incident with a clear line of investigation - you're dealing with a violation that everyone ends up answering for at once. Splitting the roles only looks bureaucratic until the first incident.

Agree on a communication protocol before the first run, not at the moment someone is already standing at the machine. Direct line of sight is more reliable than a radio. If there's no line of sight, test the radio link beforehand and adopt a hard rule: no answer means no start. The command is repeated aloud both ways, the way it's done with a crane: 'zone clear, go ahead' and back 'starting.'

• The engineer is responsible for the instrument, sensors, calculation, and weights. Your operator is responsible for the machine's condition, start-up, shutdown, and operating mode.
• One command, one person giving it. If two people start directing the start-up, sooner or later one of them will say 'go' before the other has cleared the zone.
• Before the first run, ask to be shown the nearest emergency-stop button and have it explained exactly what it does: cut power to the drive, or just switch it to 'stop' mode.
• Ask about the stopping time. On a large induced-draft fan or centrifuge, minutes can pass between pressing the button and the rotor actually coming to rest, and that changes the whole logic of working near the machine.
• If the only person authorized to start the machine sits in a different room and goes to lunch, you don't actually have an agreed balancing window, whatever the work order calls it.

We covered the organizational side of a site visit - access, the window for runs, and the machine data needed - in more detail in a separate article on preparing equipment for field balancing. Here the focus is safety only.

Work permit, lockout, and tagging: two separate phases of the job

Split the work into two machine states and never mix them. The 'setup' state is fitting the sensors, marking the zero reference, attaching the reflective marker, and installing or removing weights. The rotor is stationary throughout, and the drive is de-energized and locked out. The 'run' state is the measurement itself. Nobody in the zone, the guard back in place, tools cleared away.

You'll switch between these two states five or six times over a shift. Every switch is handled the same way, with no exception for the last weight that you just want to 'nudge for a second.' It's precisely on that last weight that people get caught by a start-up, because by then the procedure has worn thin.

Have the work permit or work authorization issued before the engineer arrives. It should state more than just 'vibration measurement' - it needs the number of runs planned, the fact that a guard will be removed, and the names of both people in charge. If correction weights will be welded, clear the hot work separately, on its own permit. Missing this is the single most common reason for losing half a shift on a site visit.

1. 1

   Shutdown and full coast-down

   The machine is stopped through its normal procedure, and you wait for rotation to fully stop. The clock starts not from pressing the button, but from the shaft mark actually holding still.

2. 2

   Cutting power, not just switching to 'stop'

   A panel button, a command from the control system, or a stopped VFD all leave the drive energized and under remote control. Power is cut at the source: disconnector, circuit breaker, or starter set to 'off.'

3. 3

   Lock and tag

   A lock and a tag with the person's name and the time go on the isolating device. If several people are working in the zone, one lock per person goes on the hasp, and each person removes only their own.

4. 4

   Verifying zero energy and a test start attempt

   A qualified electrician verifies there's no voltage present. After that, a test press of 'start' is made: the machine must not move at all. This step catches a mixed-up isolating device more often than people assume.

5. 5

   Residual energy

   Let the VFD's capacitors discharge, waiting the time stated in its manual. Release any system pressure, slacken the drive belt tension, and lock a vertical rotor against turning under its own weight.

6. 6

   Securing against self-rotation

   A fan wheel on an induced-draft fan will turn under natural draft, a pump impeller under reverse flow, a drum under its own imbalance. Close the damper and fit a mechanical lock or a wedge before bringing a hand anywhere near the correction plane - the location on the rotor where weights are fitted and removed.

7. 7

   Removing the lockout before a run

   Only the person who applied a lock removes it, and only after they've left the zone themselves and confirmed everyone else has too. By that point the guard is back in place, tools are cleared away, the weight is tightened down, and the rotor has been turned a full revolution by hand.

The most dangerous phrase you'll hear on site is: 'I've got my finger on the button, no one's starting it.' A finger is not a lockout. A lockout is an isolated device, a lock, and a tag, with the key in the pocket of the person standing inside the zone.

Guards and the throw zone: where people stand during a run

A guard is removed and refitted only on a rotor that's stopped and locked out. For every run, it goes back in place. That's how most machines work, and the matter ends there.

Some rotors can't be balanced any other way: an open impeller, a removed inlet duct, a fan with its screen off, a machine where the correction plane is only accessible with the housing disassembled. In that case, you set up a zone instead of a guard. Mark the boundary physically - with tape or a barrier - not with words. During runs, nobody is inside that boundary.

Understanding the geometry of the hazard matters more than remembering a rule. A weight that breaks loose, or a chunk of blade, flies off tangentially to the circle of rotation - meaning it stays within the rotor's plane of rotation. The danger isn't a circle around the machine, it's a sector within that plane. You need to move sideways, clear of the wheel's plane, not just 'further away along the conveyor's axis.'

Last but not least: during a run, watch the screen, not the machine. That's exactly what the instrument is for. The amplitude and phase of the running-speed component (vibration at the rotation frequency - which is exactly what imbalance produces) tell you far more about what's happening than a view from two metres away, and two metres is precisely the distance at which injuries happen.

• Move out of the rotor's plane of rotation, not just 'further away.' Distance along the shaft axis helps; distance within the wheel's plane barely helps at all.
• Thin sheet metal, an inspection hatch, a flexible connector, and glass stop nothing. Don't stand facing them, even when the housing is technically closed.
• Move the laptop and the person running the instrument out of the sector. Sensor cables are made long precisely for this, not for convenience of routing.
• Block off neighbouring walkways, stairs, and platforms too. Someone walking through the shop on their own business has no idea a run is about to start.
• Nobody stays in the zone during a run. Not 'someone will just hold the sensor,' not 'I'll quickly get it on video,' not 'I'll hold the tripod, it's about to fall.'
• Route and secure sensor cables so that no loop can reach a coupling, a belt, or a cooling fan. A cable that gets pulled in drags the sensor, the tripod, and whatever hand is holding it along with it.
• Nobody approaches the machine until it has fully stopped, even if a coast-down measurement is still running. The instrument keeps logging on its own - it doesn't need anyone standing next to it.

If the tripod holding the laser phase sensor ends up inside the hazard sector, move it. A reflective-tape marker is visible from several metres away, and a spot for the tripod outside the plane of rotation can almost always be found.

Mounting the weight: the highest-stakes operation of the visit

Fitting the trial and correction weights is the one balancing operation that directly creates a new hazard. Everything else is measurement - here you're adding mass to a spinning rotor and relying on it to stay put. Treat a trial weight with the same seriousness as a permanent one: centrifugal force doesn't know or care whether it's temporary.

The trial weight's mass isn't chosen by 'bigger just to be safe' - it's chosen from the machine's response. The change you're after is 20 to 30 percent in the running-speed amplitude, or 20 to 30 degrees in phase; we cover the logic behind that in a separate article on trial weights. On top of that sits a safety limit that matters more than convenience: the weight must not push the machine into an unacceptable vibration zone, overload the bearings, or bring anything into contact with a stationary part.

The practical limit we use in the field is this: the centrifugal force from a trial weight shouldn't exceed roughly a tenth of the rotor's own weight. That calculation takes a minute, and it saves you from a situation where the 'trial' weight ends up stronger than the imbalance itself, and the machine starts tearing at itself during run-up.

The mounting radius is chosen just as carefully. The further from the axis, the stronger the effect for the same mass - but the higher the separation speed, and usually the weaker the structure at that point. A blade edge, a protective screen, thin housing cladding are not places for a weight, no matter how convenient the access looks. Mount the weight on a solid element: a disc, a hub, a balance ring, a back plate, or existing bolt holes.

If the only available correction plane won't let you mount the weight securely, the problem gets solved with calculation, not improvisation. The software can recalculate the correction mass for a different set of available planes. That costs more time, and it's always cheaper than repairing a housing.

Sources: Balanset-1A operation manual

If a weight breaks loose: what happens in the first second

It's worth running the numbers once, so the rule about the lock nut stops feeling like a formality. Centrifugal force rises as the square of rotation speed, and you simply can't hear that: a machine at 3000 RPM sounds louder than at 1500 RPM, but nowhere near four times more dangerous to the ear. Yet the force and the energy of a separation really do grow fourfold.

The table below is a calculated example for a 50-gram weight at a 300-millimetre radius. There's no actual measurement behind it - it's pure arithmetic from the centrifugal-force formula, included only to give a sense of scale.

Two hundred and twenty joules is the energy of a two-kilogram block falling from eleven metres up. An object like that doesn't leave the machine vertically - it travels horizontally, in the plane of rotation, and in its path sits a guard made of three-millimetre sheet metal, and a person who thought they were standing 'off to the side.'

There's a second part to a weight breaking loose that people tend to forget. The weight is gone - which means the imbalance has jumped by the same amount, just in the opposite direction. Within a single revolution, the machine swings into a high vibration level, clearances start working against you, seals and the wheel's fit on the shaft suffer, and you end up with damage that wasn't there before the balancing job. The sign on screen is unmistakable: the running-speed amplitude jumps, and the phase flips. See that, and you stop the machine and go find out where your weight is.

There's a simple practical consequence here. At 750 RPM, a mounting mistake usually ends in a scare and a search for the weight under the machine. At 3000 RPM, it ends with a hole punched through the housing. Mounting requirements on a high-speed machine aren't stricter on paper - they're stricter in physics.

Hot equipment, dust, and explosive atmospheres

An induced-draft fan, a hot-air fan, a pump handling a hot product: the bearing housing on machines like these can easily be hot enough to cause an instant burn. Plan for it in advance, because the fix always costs time.

An accelerometer has a housing temperature limit, and it's stated on that specific sensor's datasheet. Don't estimate it by eye. A magnetic mount on a hot housing fails you twice over: a permanent magnet's holding force drops noticeably as it heats up, and the magnet itself conducts heat straight into the sensor. A stud, or an adapter pad that doubles as a heat sink, is more reliable on a hot housing. Route the cable so it doesn't lie against hot metal.

Dust is more dangerous than it looks, and not because of the dust itself. Hot work to fit a correction weight in a room with an accumulated layer of combustible dust - flour, wood shavings, coal, sugar, whatever it is - throws sparks, and sparks start a smouldering fire. Dust is cleared by wet methods. Blowing it off with compressed air makes things worse: you raise a cloud, and it's exactly a cloud of combustible dust that can explode.

Let's be direct about explosive atmospheres. The Balanset-1A is a general industrial instrument. It has no explosion-proof rating or certification for use in classified zones, and it must not be brought into one. There are exactly three options: run the measuring part outside the boundary of the zone, if the sensor cables are long enough; set up hot-work-style temporary permit work under your own rules, with gas monitoring and purging, if your safety department allows it; or forgo on-site balancing and pull the rotor instead. We won't offer you a fourth option, because there isn't one.

• Heat-resistant gloves for the whole time you're working near a hot housing, including removing the sensor after a run. The housing cools more slowly than it looks like it will.
• Check the sensor's rated temperature limit on its datasheet. If it's exceeded, change the mounting method or the measurement point - don't just count on the measurement being quick.
• The instrument itself has its own operating limits: +5 to +50°C, and humidity up to 85% with no condensation. Freezing temperatures on an outdoor site and condensation forming are reasons to postpone the work, not to 'just work fast.'
• Wet-clean the dust before hot work, keep a fire extinguisher and a fire watch nearby, and monitor the weld area after the work is finished. A smouldering fire doesn't always show itself right away.
• Welding over paint, an oil film, or a layer of product gives you neither a sound weld nor safety. Clean down to bare metal.
• Steam, hot exhaust, and relief lines near a bearing housing are more dangerous than the housing itself, because you can't see the jet. Find out where they are before you set up the tripod.

Sources: Balanset-1A operation manual

Instrument power, working at height, confined spaces, and PPE

Ear muffs and earplugs make it harder to hear a 'stop' call. Agree on three hand signals before the first run: start, immediate stop, zone clear. Confirm both people read them the same way, and don't rely on shouting over a running machine.

RPM, emergency stop, and abnormal vibration during a run

Balance at the operating speed, and only at that speed. The temptation to push the frequency up on a VFD 'to see the vibration more clearly' comes up regularly, and it's a bad idea. The wheel's strength is rated for the nominal speed, and the centrifugal load on the blades - and on your weight - rises as the square of RPM. Fifteen percent over nominal adds roughly a third more load.

Treat the first run of a machine with a large imbalance with extra care. A simple check beforehand: turn a horizontal-axis rotor 90 degrees and let go. If it swings back to the same position on its own every time, static unbalance is significant. With a variable-speed drive, bring the speed up in steps and watch the level as you go; with a direct-on-line start, you don't have that option, which makes it all the more important that nobody is in the zone for that run.

Resonance is a hazard in its own right, not just a nuisance for the calculation. In a resonance band, amplitude climbs several-fold, the running-speed phase becomes unstable, and the load on the supports goes well beyond what they're rated for. Pass through that band quickly during run-up, and don't linger there to take a measurement. We cover the signs of resonance and how to check for it in a separate article.

The single most important rule in this article is short. An abnormal sign during a run means an immediate stop. Not 'let's just finish the measurement,' not 'ten more seconds and the instrument will average it out.' A measurement takes seconds and can be repeated for free. Every extra revolution with a weight that has come off, a crack growing in a blade, or contact just starting costs you a wheel, a bearing, or a person.

• A sharp rise in vibration level during run-up, or a jump at steady speed. Stop immediately, without waiting for the measurement to finish.
• A new sound: an impact, a grinding noise, a rhythmic knock, the whistle of something rubbing. The ear is faster than the instrument here.
• A jump in the running-speed amplitude together with a phase flip. Read this as 'the weight has come off.'
• Vibration level above the agreed stop threshold. Set that threshold in actual numbers before the run - for example, at the boundary of the unacceptable zone in the applicable part of ISO 20816 for your machine group - and don't renegotiate it on the fly.
• Amplitude and phase readings won't settle at steady speed, drifting by more than 10-15 percent over the course of the measurement. There's no point continuing, and it isn't safe to.
• Smoke, a burning smell, a bearing housing hotter than usual, grease being thrown out, a leak through a seal.
• Any movement of the machine itself: a shim shifting, a foot loosening, the housing rocking, the tripod starting to move.
• Anyone at all entering the hazard zone. Stop first, work out who and why afterward.

Every person on the job, including the visiting engineer, must have stop authority, and that isn't up for debate on site. A stop on a false alarm costs you one extra run. A run that didn't get stopped in time costs a great deal more.

Sources: ISO 20816-1:2016

Checklist before the first run, and when work stops

• The start-up can't be locked out. No de-energizing scheme, no lock, the only safeguard is someone's promise not to press the button. Work does not begin.
• A guard is being removed or refitted on a rotor that's still turning. Work stops until the situation is reviewed.
• The weight can't be mounted securely. Only thin cladding, a screen, or a blade edge is accessible, there's nowhere for a bolt, and welding to that material isn't allowed. Find a different correction plane, or pull the rotor instead.
• There's no one responsible for start-up, or they've been pulled onto another task. Work pauses until they're back - nobody starts the machine 'carefully, on their own.'
• The throw zone isn't marked, and people are walking through it. Work stops until the walkway is blocked off.
• A measurement point is only reachable from a stepladder, or from a spot you can't get out of quickly. Skip that point and record the reason in the report.
• A classified explosive zone without properly organized permit work. The instrument does not go into that zone.
• Vibration during a run is abnormal: a jump, a new sound, drifting readings. Stop, inspect, and only then decide whether to continue.
• Site conditions fall outside the operating range: freezing cold, rain on an outdoor site, condensation. The work is postponed.
• The process won't allow the number of stops needed. Find a maintenance window instead of cutting corners on safety.

• The work permit or work authorization is issued, and it states that a guard will be removed and how many runs are planned.
• The person responsible for work safety has been named and is on site.
• The person responsible for start-up and shutdown has been named, is on site, and communication with them has been tested.
• Hand signals are agreed: start, immediate stop, zone clear.
• The lockout scheme is clear: what isolates the drive, who applies the lock and tag, and who holds the key.
• The nearest emergency-stop button has been pointed out, and what it actually does is understood.
• The guard is back in place, or the hazard zone is marked physically, not just described verbally.
• The rotor's plane of rotation has been identified, and everyone knows which way to move clear: sideways, not along the axis.
• The weight is tightened down: washers in place, a lock nut or self-locking nut fitted, two or three thread turns above the nut.
• The rotor has been turned a full revolution by hand, with no contact anywhere in that turn.
• The weight's mass and radius have been measured and entered into the software as actual values.
• Tools, rags, fasteners, and bags have been cleared out of the machine and out of the zone.
• Sensor cables are secured and routed clear of the coupling, the belt, and the cooling fan.
• The laptop and the person running the instrument are out of the hazard sector.
• The vibration-level stop threshold has been agreed in actual numbers before the run.
• PPE is on everyone: hard hat, glasses, protective footwear, hearing protection near a loud machine, nothing loose on clothing or hair.

Safety on a site visit isn't a separate document, or a signature in a logbook before you start. It's a sequence of decisions that repeats as many times as you have runs in a shift. Cut a corner just once, on that last weight, and everything that came before it stops mattering.

Sources: Balanset-1A manufacturer specification

Frequently asked questions

Related content