How to build vibration monitoring and balancing into your maintenance program

As long as vibration only gets measured on a call of 'something's making a noise,' you're always dealing with consequences, not causes. You need the same instrument, the same hands, the same callout time. The only difference is whether the measurements are built into the maintenance schedule or not. Below, we break down how to turn one-off callouts into a program: which machines to include, how often to measure, what to record, when balancing gets scheduled in advance, and who does all of it.

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

Three maintenance approaches, and where vibration monitoring pays off most

A typical call sounds like this: the induced-draft fan started making noise this morning, by noon the bearing housing was hot, please come today. By the time the engineer reaches the site, they're no longer dealing with unbalance but with its consequences: a wrecked bearing, stretched studs, sometimes a crack in a weld on the frame. Balancing is still possible in this situation, but it has become the third item after repairs, not the first.

The reason isn't the people or the instrument. It's that the machine has no history. Nobody knows what vibration level this fan had six months ago, so the very first measurement has to be interpreted blind. History only comes from a program.

• Vibration monitoring pays off most on machines with a steady operating regime and accessible bearing housings: fans and induced-draft fans, pumps, electric motors, compressors, crushers, mulchers, line drives.
• It pays off most where a defect develops gradually: product buildup and impeller wear, bearing raceway spalling, loosening fasteners, foundation settling, shaft misalignment after thermal expansion.
• It gives you little on machines with infrequent starts and a widely variable regime: each measurement lands in different conditions, and no trend forms.
• It gives almost nothing where the failure is instantaneous: a snapped belt, a foreign object ingested, an electrical winding breakdown.
• Don't pick one approach for the whole fleet. A working program mixes all three and states plainly which machine belongs to which group.

One thing is worth remembering about unbalance. The calculated bearing life under ISO 281 is tied to the load acting on it, and an unbalanced mass adds a constant rotating force on top of the working load. This force never switches off for a single minute of operation, and it grows with the square of the speed.

Sources: ISO 281:2007

Which machines to include in the program, and by what criterion

The temptation is understandable: put everything that rotates on the program. A list of two hundred items turns, by the third month, into a signature in the log with no actual measurement behind it, because the route technician physically can't keep up. Select machines deliberately, by a criterion you can explain to the chief engineer.

There's one main criterion: what happens when this machine stops without warning. Everything else just refines the answer.

• Consequences of a stoppage: a line stops, scrap starts, an environmental limit gets triggered, people are put at risk. One induced-draft fan stops the whole shop; one of three supply-air fans stops nothing.
• Whether a standby unit exists and how long it takes to bring it online. A standby pump that starts up within a minute takes half the urgency out of the question.
• Failure history and repair cost. A machine that destroys bearings once a year goes into the program regardless of its power rating.
• Parts lead time. If a housed bearing for this unit takes twelve weeks to arrive, you need a quarter's warning, not a week's.
• Observability: whether the bearing housing is accessible while running and whether the regime is stable enough. If you can't reach the housing while the machine is running, you have three honest options: build a platform, extend a stud with a sensor pad to somewhere accessible, or accept that this machine won't be in the program.
• Sort what you've selected into three groups: A — critical, no standby; B — important, with standby; C — everything else. The group determines the measurement interval, the set of parameters, and who does the work.

Review group membership once a year. The fleet changes, regimes change, and a group C machine can become the only one in the chain after a change to the production layout. A program that hasn't been reviewed in three years no longer describes your plant. We covered how to mark out points, capture a baseline level, and track a trend in a separate article on vibration monitoring and routes.

Measurement interval: base it on how fast the defect develops

The calendar is secondary here. The measurement interval has to be shorter than the time your typical defect takes to go from the first noticeable sign to failure. Working rule: the interval should be no more than half that time. Otherwise you'll see the defect exactly once, and it will already be in its final stage.

Benchmarks for typical defects explain why a month turns out to be a sensible unit for most machines.

• Unbalance from product buildup or uneven impeller wear builds up over weeks, sometimes a month. Mulchers, crushers, and induced-draft fans handling dusty gas live in this regime constantly.
• A rolling-element bearing defect usually runs from a month to six months, from the first distinguishable sign to failure.
• Gear mesh wear progresses more slowly: from three months to a year.
• Loosening fasteners and foundation settling develop over months, but sometimes jump forward after a hard start or an impact.
• That gives you a working split: group A monthly, group B quarterly, group C every six months or on request only.
• Events sit apart from the calendar. Any repair that touched the rotor, the bearing housings, or the mounting requires a measurement immediately after reassembly, regardless of when the last round happened.

Measurement intervals only make sense under consistent measurement conditions. The same induced-draft fan gives different numbers at different damper positions. Write the operating regime for the measurement into the program, and require the route technician to wait for it or log the deviation.

Four documents the program rests on

A program doesn't live in the mechanic's head or in an email thread. It needs four documents, and all four are simple. Their absence always shows up at the same moment: when the person who remembered everything goes on leave or changes jobs.

• Date, time, who performed the measurement
• Machine ID and point number from the map
• Measurement direction
• Rotational speed at the time of measurement and the operating regime: load, damper position, temperature
• Overall vibration and the 1x running-speed component (vibration at the rotor's rotational frequency), with units and amplitude type stated: RMS (root mean square) or peak
• The frequency band the value was taken in
• 1x phase, if the instrument provides it: useful later for root-cause analysis
• Anomalies you noticed by hand or by eye: an oil leak, heat, knocking, a new sound
• Comparison against the threshold and a note of any exceedance

A number without metadata can't be correctly compared against either a threshold or a previous measurement. ISO 13373-1 requires full identification of the measurement conditions for exactly this reason: so the result is reproducible. For contractual acceptance, this isn't a formality — it's the only way to avoid a dispute later.

Sources: ISO 13373-3:2015

When to schedule balancing in advance, without waiting for a measurement

Some jobs cause unbalance not as a maybe, but predictably. You're going to open the machine anyway, take an acceptance measurement anyway, and spend a shutdown anyway. It makes sense to write balancing into the work order right away, so you don't call people out a second time or open the unit twice.

• After repairing or replacing a rotor or impeller. A new impeller doesn't mean a balanced one: the factory tolerance is set for its own grade, not for your machine and your operating speed.
• After replacing blades, crusher hammers, or mulcher knives or teeth. Unbalance shows up here even with careful work, because the new parts differ in mass from the worn ones. Hammers and knives should only be replaced as a matched set, selected by mass, otherwise balancing starts out fighting the assembly itself.
• After rewinding or repairing an electric motor. The rotor was disassembled, the cooling fan was pulled off and refitted, the winding was impregnated with varnish, the banding was retied. Any one of these operations redistributes mass.
• After installation in a new location or relocating the unit. The stiffness of the base changes, and with it the response of the rotor-bearings-base system to the same residual unbalance. On a new foundation, the vibration level can turn out different even though the rotor hasn't changed.
• After cleaning or washing an impeller. Deposits almost never come off evenly: some falls away, some stays in the passage between blades. A machine that ran smoothly before cleaning starts shaking afterward.
• After weld build-up, machining, shaft straightening, or replacing a hub or pulley. Metal was removed or added, almost certainly asymmetrically.
• Before a long continuous campaign or season, when the next chance to stop the unit won't come for several months.

Scheduling balancing in advance doesn't mean balancing blind. The sequence stays the same: an acceptance measurement after reassembly, comparing the running-speed component against the overall vibration, then a decision. This step often reveals that what's needed isn't balancing but shaft alignment or tightening the mounting feet. What goes into the plan is a window and the resources for the work, not a commitment to fit weights. Evaluate acceptance against the applicable part of ISO 20816 in its current edition, and if a balance quality grade is required, separately against ISO 21940-11.

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

How to link the measurement to a decision: three threshold levels

The most common way a program breaks down looks like this: measurements keep happening, numbers keep piling up, and no decisions get made. The route technician wrote down 4.8 mm/s and moved on, because the document doesn't say what to do with that number. A threshold with no defined action doesn't work.

Build three levels. Calculate the first two from the specific point's baseline level; keep the third as an absolute value and check it against the zones in the applicable part of ISO 20816 — the standard divides machine condition into zones from A to D, from normal for a new machine to unacceptable vibration. The baseline level matters more than the absolute table: a machine with a baseline of 0.8 mm/s that now reads 2.2 has already told you about a developing defect, even though it's formally still in zone B.

• An exceedance means 'investigate,' not 'balance.' The first step is always the same: compare the overall vibration to the running-speed component. If the overall level is several times higher than 1x, balancing won't help.
• Don't set one threshold for the whole fleet. A pump running at 2950 rpm and a crusher on a rigid base running at 990 rpm have different acceptable levels and different baselines.
• Don't compare a peak value against a threshold set in RMS. The error comes out several times over, and it looks like a real increase.
• Write down who reviews thresholds and when. After a major overhaul, the baseline level is captured again, and the old thresholds become invalid.

Sources: ISO 20816-1:2016

A sample program structure

Below is a typical structure that's convenient to start from. This is an illustrative example, not anyone's actual site: you work out your own groups, points, and thresholds from your own fleet and your own baseline levels. The value of the table is that each row answers five questions at once, and once it's filled in, the program is ready to sign off.

The row people forget most often: who updates the baseline level, and when. Without it, within a year the thresholds stop matching the machine, and the program starts producing either false alarms or silence.

Who does the work: in-house staff, a contractor, or a mixed setup

This isn't a question of principle — it comes down to how often you need the work done and how fast you need a response.

• Don't count the cost of a single job — count the full cost of the decision: downtime while waiting, reopening the machine a second time, the risk of a second callout.
• If balancing is needed on your fleet more often than once a quarter, your own instrument stops being a luxury.
• If you need to decide not just 'how many grams and where' but also 'is this actually unbalance or not,' plan from the start for someone who can read a spectrum.
• We covered separately where the line falls between on-site work and shop-machine work, along with the criteria for choosing the instrument itself.

Where to start, and what your person needs to know at the first stage

Don't try to launch the program across the whole fleet at once. Take five or ten group A machines and carry the full cycle through on them.

• At the first stage, the route technician needs to be able to do five things: mount the sensor at a marked point and direction the same way every time; take the overall vibration, the running-speed component, and the speed; fill in a log line completely; compare the value against the threshold and escalate on time; and balance a rigid rotor in a single plane using fixed positions.
• At this first stage, they aren't required to run investigations, work with coastdowns and orbit plots, balance flexible rotors, or diagnose from a spectrum. Those are separate competency levels, and we covered the ISO 18436-2 category structure in a separate article.
• The single most useful skill, and one that can be picked up in a single visit: comparing 1x against the overall vibration and using that comparison to work out whether it's even worth reaching for weights.

1. 1

   Select the machines and document the points

   Five or ten units, each with a point map for its bearing housings, with directions and the sensor mounting method. Wherever a housing can't be reached while running, solve the access problem right away.

2. 2

   Capture the baseline levels

   Best done right after a repair, on a machine in good condition, at an agreed operating regime. Record the conditions. This is your reference point for years ahead.

3. 3

   Set the thresholds and actions

   Three levels, and next to each one an action, a deadline, and who's responsible. Check the wording for clarity: someone on the night shift will be reading it.

4. 4

   Add the rounds to the maintenance schedule

   The round needs to be a work order with a date and an assigned person, not a task for 'whenever there's time.' Otherwise it loses out to any emergency job.

5. 5

   Review after a quarter

   What didn't get done, where the points turned out inconvenient, where a threshold produced false alarms. After that, expand the program to group B.

AXILINE engineers design and manufacture the Balanset instruments and do the balancing themselves on site visits. We provide hands-on training on your own machines and ongoing consulting support afterward, once your staff is running the route independently. A practical way to start looks like this: on the first visit, we balance whatever is getting in the way of running right now, and at the same time mark out the points and capture the baseline levels, so the program rests on your own real numbers instead of a table from the internet. The Balanset-1A fits this setup by design: two accelerometers, a laser phase sensor using reflective tape, a two-channel USB module, and a laptop, with the case kit weighing under five kilograms. The instrument computes single- and two-plane balancing, displays overall vibration and 1x, phase, speed, spectrum, and the time waveform, supports fixed positions and drilling calculations, stores influence coefficients, keeps an archive, and generates reports. For building into a machine tool or test stand, there's a Balanset-1A OEM version without the case.

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

Frequently asked questions

Related content