The economics of vibration, and building the business case

What vibration costs a plant: how to calculate the cost and build the case for the work

You show up with a reading of 7.2 mm/s on an induced-draft fan's bearing housing and ask for a four-hour window. The answer you get back is 'but it's running fine,' and the conversation is over: you're speaking in millimetres per second, and the decision gets made in euros per hour. A translation between the two is possible, but not through percentages pulled from industry surveys - through your own data from the last two years. Below is a calculation framework with blank fields, where to pull each figure from, and what to put on the table for management.

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

In short: Calculate not the vibration, but the events it has already caused on your line: hours of unplanned shutdowns, the extra cost of emergency repair over planned maintenance, premature consumption of bearings and seals, secondary damage, scrap. Multiply your downtime hours by an hourly rate agreed with finance, and add up the remaining lines over 12-24 months from your failure log, spare-parts stores, and shift reports. Calculate the difference, not the absolute amount: you'd have replaced the bearing eventually anyway - the damage from vibration is only the difference between a planned replacement and an emergency one. Keep the risk of injury and a regulator-ordered shutdown on its own line, and don't add it into the actual costs.

'It's shaking' isn't an argument. An hour of downtime is.

The conversation about vibration almost always runs into the same wall. You show a number on a display, and you're told the machine has been running for three years and hasn't bothered anyone. Both sides are right, and both are speaking different languages. Vibration isn't a cost that arrives as a line on a monthly invoice. It's a rate of wear and a probability of stopping at an inconvenient moment, and things like that don't get their own line in a budget.

So what goes into the invoice isn't the vibration, but its consequences - the ones that have already happened on your line and already been paid for. The good news is they're documented. Your key figure doesn't live in the instrument - it lives in the failure log, in the stores' delivery notes, and in shift reports from the last two years. The instrument is for something else: it tells you which machine is next.

The second rule matters more than the first. Calculate the difference, not the absolute figure. You'd have replaced the bearing anyway - the only question is when, and at what price. The unit would have stopped once a year regardless - the question is who picks the date. The moment you start calculating the difference, the number gets smaller, but it becomes defensible, and that's exactly what you need going into a meeting.

Third: don't mix actual costs with risk. Actual costs can be checked against documents. Risk is an estimate, and it lives on its own line. If you add the two into one total, the first person in the room with a finance background will strike out the whole calculation - and they'll be right to.

Don't treat the full cost of a repair as the damage. The damage from vibration is only the mark-up: the difference between a planned replacement and an emergency one, between a stop inside a maintenance window and a stop on a Friday night.

Eight categories of cost that elevated vibration creates

Below is the full list of what you're paying for. Not every line needs filling in at once: some can be worked out in half a day from delivery notes, some need measurements, and some may simply not apply to your process at all. Fill in the ones where you have documents, and honestly write 'not calculated' where you don't. A blank line with a note holds up better under scrutiny than a plausible-looking number pulled from thin air.

1. Unplanned shutdown and lost output

Almost always the most expensive line. Count the actual hours from the 'stop' command to the process being back at steady state, not just the time repair crews had tools in hand. Warm-up, purging, ramping up to setpoint, and the first hours of off-spec product are downtime too. In shift reports, this cause is most often logged simply as 'mechanical,' and your first job is to break that down by component.

2. Emergency repair versus planned maintenance

The same job costs differently depending on who picked the timing. In an emergency, you're paying for overtime and night shifts, for courier delivery of a part instead of pulling it from stores, for a contractor at a rush rate, for the absence of prepared tooling, and for inspection done on the spot instead of knowing the scope in advance. Only the difference between the two versions goes into the tally.

3. Accelerated consumption of bearings and seals

Imbalance produces a centrifugal force - a constant dynamic load on the support. That force is part of the bearing's equivalent dynamic load. Life calculations under ISO 281 depend on the ratio of the dynamic load rating to that load, raised to the power of 3 for ball bearings and 10/3 for roller bearings: doubling the load drops the calculated life by roughly 8-10 times. In a real machine the effect is gentler, because the load also includes the rotor's own weight and belt tension. Calculate from fact: how many bearings of this size have left stores for this unit over 24 months, against what the maintenance schedule called for.

4. Secondary damage

The line that makes the total big. Shafts and keyways, loosened fits on wheels and half-couplings, elastomeric elements and gear-coupling teeth, frame weld seams, anchor bolts and foundation grout, connecting pipework and pipelines, covers and guards. Restoring a frame or a foundation costs an order of magnitude more than a bearing, and it comes with an extended shutdown, often with a contractor brought in.

5. Higher energy consumption and lost efficiency

This one needs care, or the calculation is easy to challenge. The energy that goes into shaking the structure itself is small. What's expensive is whatever is causing the imbalance in the first place: a worn or product-caked impeller, rubbing contact, widened seal clearances, a slipping belt. Only put this line into the tally if you've measured motor current and actual output before and after the work. Percentages borrowed from someone else's article won't survive scrutiny.

6. Scrap and reduced product quality

This applies wherever vibration feeds directly into the process. On a machine-tool spindle, that's surface waviness and faceting, dimensional drift, and faster wear on tooling and dressing. On a crusher or mulcher, it's a wider spread in particle size. On a mixer or centrifuge, it's unevenness and repeat cycles. Calculate it from your own control charts: the share of nonconformances and rework at this process step against comparable steps, tied to specific dates and shifts.

7. Working conditions: workplace noise and vibration

In the EU, a worker's exposure to vibration and noise is regulated separately from the machine's condition, and it's a cost item independent of the equipment itself. Directive 2002/44/EC sets a normalized A(8) value for vibration - the level referenced to an eight-hour shift. For hand-arm vibration, the action value is 2.5 m/s² and the limit value is 5 m/s²; for whole-body vibration, 0.5 and 1.15 m/s². Directive 2003/10/EC on noise works with levels of 80, 85, and 87 dB(A). Exceeding them costs money: measurements, PPE, restricted working time, health surveillance, organizational measures. Check separately how these requirements have been transposed into national law: the rules and the paperwork involved differ from country to country.

8. Risk of injury and a regulator-ordered shutdown

An impeller failure, a fastener letting go, a weight or a fragment flying off - that isn't a repair line item. This covers an investigation, a suspended production area, repairs mandated by an enforcement notice, insurance consequences, and loss of certification with a customer. Assess this line as probability multiplied by the cost of the consequence, and keep it separate from actual costs. Don't fold it into the total.

Sources: ISO 281:2007

Where your numbers live inside the company, and which fields you need from them

You don't need to collect data from scratch - it already exists. The problem is elsewhere: it lives across four or five systems that don't line up with each other, because the same induced-draft fan is named differently in each one. The first thing you do is establish a single unit identifier. Without it, no amount of exporting data will let it come together.

Use a 24-month window, 12 as an absolute minimum. Twenty-four months matters for two reasons: it captures seasonal variation in load, and it stops a single large failure from dominating the whole result. If you have less than a year of history, do the calculation anyway, but say so: data covers a partial period, the estimate is preliminary.

And don't try to start with perfect data across the entire fleet. Take one group of identical machines - say, all the induced-draft fans, or all the pumps on one section - and carry the calculation through to completion for them. One finished calculation covering ten units is more persuasive than an unfinished one covering three hundred.

The maintenance-management system or failure log. You need these fields: date, unit, component, failure description, a 'planned or unplanned' flag, downtime hours, labour in person-hours.
The control-room log and production shift reports. Start and end of the stoppage, cause, lost output in physical units. This is your most expensive field, and at the same time the most carelessly filled in.
Stores and parts requisitions. Item, quantity, date issued, unit, price. Consumption of one bearing size broken down by unit points to the problem machines before you've even picked up an instrument. Nobody dresses up this database, because it's about money and stock levels.
Work orders and contractor completion certificates. Amount, date, an urgency flag, overtime, the cost of lifting equipment and rigging.
Quality department. The share of nonconformances and rework by process step, tied to date, shift, and equipment.
Health and safety department. Workplace noise and vibration measurement records, enforcement notices, PPE issued, work-time restrictions.
Energy metering and the control system. Per-unit meters, if you have them, and motor current logs. Usually the weakest point in the whole dataset, which is why the energy line has to be confirmed with your own measurements instead.

A single unit identifier, the same across every system.
Date, start and end time, and actual hours the process was stopped.
A 'planned or unplanned' flag. One field that solves half the problem.
Component and nature of the failure: bearing, shaft, fit, coupling, fasteners, frame, seal, impeller.
Direct cost per event: parts plus labour plus contractor.
Lost output in physical units, not money. You convert it to money yourself, at an agreed rate.

If the log has no 'planned or unplanned' field, add it today, and don't try to reconstruct the past from memory. Checkboxes filled in retroactively always show, and they kill trust in the entire calculation.

The cost of one hour of downtime: the number worth sitting through a conversation with finance for

Without this figure, not a single line turns into money. With it, all of them do at once - including the ones you'll calculate a year from now. So it's worth getting it right, and getting it once: not a back-of-the-envelope guess, but a calculation someone has signed off on.

It's built by adding up five components. Below is a table with blank fields, and a note on who in the company holds each figure.

Then comes an honest fork in the logic that usually goes unmentioned. If the line isn't running at full capacity and the output can be made up on the next shift, an hour of downtime doesn't cost the lost margin. It costs overtime and a share of overheads. If the line is a bottleneck and the product is already sold, an hour of downtime costs the full marginal profit. The difference between the two cases is a multiple, not a fraction. Anyone who blurs the two in their own favour loses credibility once, and for good.

It follows that you don't have one rate - you have three: an hour of downtime at a bottleneck, an hour of downtime on a unit with a fast-switching standby, and an hour of stoppage inside a planned maintenance window, where lost output is close to zero. That third rate is exactly the economic point of planned intervention.

Agree the figures with finance, write them down together with the date and the assumptions, and use only those going forward. After that, the discussion stops being an argument about how much it costs and becomes a conversation about what to do.

Component of the hourly cost	How to calculate it	Where the figure comes from
Output shortfall, if it can't be made up	output rate, units/h × marginal contribution per unit	Planning and economics department, production report
Making up the output, if it can be caught up	overtime hours needed to catch up × fully loaded rate	Timesheets, payroll department
Costs that don't stop	wages of the idle shift, rent, depreciation, heating, standby energy use	Accounting, the line's overhead figures
Ramp-up back to steady state after start-up	warm-up and purge hours × the same rate + the cost of off-spec product	Process engineers, shift reports
Other direct losses	spoiled feedstock, draining and flushing, disposal, re-setup	Process engineers, raw-materials stores
Total cost per hour of downtime	sum of the lines above	Agreed with finance, dated

Don't take your hourly downtime cost from industry surveys. Published figures there vary by orders of magnitude, and any finance director knows it. Your own five-line calculation carries more weight than any outside reference.

The damage-calculation table for the period

Now we bring it all together. You fill it in once, then update it quarterly: that takes an hour, and you always have a current figure ready for any conversation. We deliberately don't plug in rates or percentages here, because they depend on your equipment, your tariffs, and your product.

The main trap when filling this in is double counting. An emergency repair is often the same event as the stoppage itself, and the same hours can easily end up in both lines 1 and 2. Settle this upfront: line 1 gets only the hours and the lost output, line 2 gets only the mark-up on the work itself. The same applies between lines 3 and 4: the bearing is counted once, and the shaft, the fit, and the frame are counted separately.

Pay attention to the last two lines. The total is made up of lines 1-7 - actual costs, each backed by a document you can check. Line 8 sits below the total and isn't part of it. Against that total, you set the cost of the program, and that one needs to be calculated just as honestly and completely, including your own people's time.

No.	Cost category	Formula	Where to get the data
1	Downtime from unplanned shutdowns for vibration-related causes	Σ stoppage hours × cost per hour of downtime	Control-room log, shift reports, agreed rate
2	Extra cost of emergency repair over planned maintenance	(average cost of an emergency event - average cost of a planned one) × number of emergency events	Work orders, contractor completion certificates, overtime timesheets
3	Premature replacement of bearings and seals	(actual number of replacements - number scheduled) × (component price + labour)	Stores, requisitions, maintenance schedule
4	Secondary damage	Σ cost of restoring shafts, fits, couplings, the frame, the foundation, pipework	Work orders, completion certificates, invoices for parts manufacture
5	Excess energy consumption	ΔP, kW × operating hours × tariff	Power and current measurements before and after, tariff
6	Scrap and downgraded product	Σ (volume of nonconformances + rework) × cost per unit	Control charts, quality department
7	Workplace noise and vibration measures	measurements + PPE + health surveillance + organizational measures	Health and safety department, enforcement notices, invoices
	TOTAL actual costs	lines 1 + 2 + 3 + 4 + 5 + 6 + 7	Every line backed by a document
8	Risk. A separate line, not part of the total	probability of the event × the cost of the consequence	Repair estimate, failure history for your fleet
	Cost of the vibration-monitoring program	visits per year + instrument + your own people's time + planned windows × window rate	Vendor quotes, timesheets

Leave lines blank where you have no documents, and label them 'not calculated.' That keeps the whole calculation checkable. A table filled in with guesses falls apart at the very first 'where did this number come from.'

Planned intervention versus emergency shutdown: exactly where the difference comes from

This table is a convenient thing to show instead of a long explanation. It doesn't contain a single number, and it still works, because every point is instantly recognizable.

Fill in the middle and right columns with your own hours and figures, from the last real event on your line. The row-by-row difference is exactly what justifies the planned work. Not percentages, not 'industry practice' - one specific event that everyone in the room remembers.

Calculate the cost of the planned intervention in full, and don't lowball it: the site visit, a few hours of shutdown window, fasteners and weights, your own people, a follow-up verification measurement. An understated figure will work against you the second time around, when the job costs more than promised.

What's being compared	Planned intervention	Emergency shutdown
Who picks the date	You, together with production	The machine
Length of the stoppage	Known in advance, work combined with other tasks	Grows as it goes, because inspection happens on the spot
Part	From stores, or ordered through the normal process	Courier, rush rate, sometimes made to order
People	Your own shift, during normal working hours	Overtime, nights, weekends, a contractor at a rush rate
Scope of work	Only what's needed	Plus whatever secondary damage has already happened
Output	Stoppage inside a planned window, losses close to zero	Lost output at the full hourly cost
Diagnostics	The cause is known before the stoppage: imbalance (vibration at the rotation frequency, 1x) or something else	Worked out after the fact, from the failure itself
People and safety	Work follows a prepared procedure	Decisions made in a hurry, risk of injury and of a repeat failure

If you have no access to the correction plane - the location where the correction weights get fitted - on the running machine, the gap between these two columns narrows, and the cost of the planned option rises. We cover when to balance on site versus when to send the rotor to a machine in a separate article.

How to honestly calculate the effect of a vibration-monitoring program

The temptation is obvious: promise a specific percentage reduction in failure rates. Don't. A promise like that works exactly once, and when the number doesn't add up a year later, the whole topic closes with it for years to come. What you lose isn't the project - it's trust in the approach itself.

What works instead is this. Declare your metric in advance, before the program starts, along with the comparison period and which units are included. Then a year later you're bringing a measurement, not an excuse. Even if the result turns out modest, it'll be yours, and it'll be checkable.

The key technical detail: normalize against operating hours. A unit that ran 4000 hours in a year and one that ran 1200 aren't comparable, and in seasonal production, hours can swing by a factor of two. Calculate failures per 1000 operating hours, or you'll mistake a change in loading for the effect of your own work.

Small sample size

Six failures before and three after isn't a result - it's noise. On a fleet of a dozen machines, any 'halved' figure means nothing statistically. Compare total downtime hours and the makeup of the stoppages, not a ratio built from a handful of events.

Simultaneous changes

You started your rounds, and in the same quarter you also switched bearing suppliers, repaired a foundation, and lowered the speed on a VFD. You can't credit the program with the effect, and someone will rightly point that out. Keep a log of every change with its date, and the conversation becomes something you can actually work with.

The attention effect

Machines added to the route start getting inspected, lubricated, and tightened more often, simply because someone now visits them regularly. That's a real benefit too. But it isn't down to the measurements themselves, and it's worth saying so plainly.

Regression to the mean

The worst-performing units are usually the first ones added to the program, and their figures would partly improve on their own regardless. There's one way to control for this: keep a control group of comparable machines outside the program, and compare both groups over the same period.

Number of unplanned shutdowns for mechanical causes per 1000 operating hours, over periods of equal length.
Mean time between failures (MTBF) for a group of identical machines.
The share of unplanned stoppages within the total number of stoppages. A vibration-monitoring program doesn't eliminate stoppages - it converts them from unplanned to planned, and that ratio is what shows the result earliest.
Average length of a single stoppage. This falls before the number of stoppages does, because you're coming into a prepared job, with a known scope and the part already on hand.
Number of bearing and seal replacements per unit per year, from stores data.
Number of machines in zones C and D (elevated and unacceptable vibration) under the applicable part of ISO 20816, at the start and end of the period. A simple, visible, and checkable metric for the fleet's condition.
Total downtime hours from mechanical causes over the period. On a small fleet, this is the best metric, because it doesn't break down when the number of events is small.

Sources: ISO 20816-1:2016

What to put on the table for management: one page

The eight-line calculation is for you, not for management. What goes into the meeting is one page, and its job isn't to convince anyone that vibration matters - it's to get a specific decision: a window of so many hours, on such a date, for such a unit.

The order on this page is the reverse of the order you worked in. First the two condition numbers, then the trend line, then the money, and the request at the end. Keep the spectrum (vibration broken down by frequency), the polar plot, and the phase readings on a second page, for anyone who asks.

And one observation from actually defending calculations like this. A table with blank fields and labelled sources is more convincing than one with tidy totals and no references. What matters to the person making the decision isn't that the total is large - it's that they can spot-check any line at random and have it hold up.

Baseline and current level for critical machines: mm/s RMS (root-mean-square) over the 10-1000 Hz band, the measurement point and direction, and the zone under the applicable part of ISO 20816. Two numbers side by side, not one.
Trend: a 12-month line for at least one machine where the slope is visible. A slope is more convincing than any absolute value, because it shows how much time you have left.
The split between overall vibration and the 1x running-speed component. This is the proof that you're asking for balancing specifically, not 'something done to the machine.'
This line's hourly downtime cost, agreed with finance, with the date it was agreed and the name of who signed off on it.
Cost of the planned intervention: window hours, the site visit, materials, your own people, the verification measurement.
Actual costs over 12 months from your table, with the source stated for every line.
Risk on its own line, explicitly labelled as an estimate, not a fact.
A specific request with a date. Not 'we need a monitoring program' - a decision that can actually be made in this meeting.

One page, two numbers, one line, one request. Everything else goes in the back half of the folder. We cover how to mark measurement points, establish a baseline, and set alarm thresholds in a separate article on vibration monitoring.

How AXILINE's engineers can help

You do the economics, because all the figures live inside your company, and nobody hands those out. We cover the measurement side that your calculation's line items are pulled from: baseline, trend, the split between overall vibration and the running-speed component, and a report from before and after the work.

The Balanset-1A measures, simultaneously on two channels, overall RMS vibration velocity, the running-speed component's amplitude and phase, and rotor RPM. Alongside that, it shows the time waveform and the FFT spectrum. The range for 1x is 0.02 to 80 mm/s, phase measurement accuracy is ±1°, RPM range is 100 to 100,000 per minute, and the whole kit in its case weighs under 5 kg. Two channels at once means you capture both bearings in a single run, in the same mode, at the same RPM - and afterward you're not arguing over whether the machine was running the same way twenty minutes ago.

One more thing matters for your calculation: every result is saved to an archive and turned into a report. A measurement taken after a repair, and one taken six months later, stop being a memory and become a dated document - one that goes into the machine's file, and into that same cost table. Reports like these are exactly what keep your calculation checkable.

AXILINE's engineers are the same people who design and manufacture the Balanset instruments and use them to balance rotors on site themselves. We come out, measure, balance the rotor in its own operating position, help you mark and label your measurement points, establish baselines after a repair, and work through an alarm that's tripped. Consulting support is also available: send us your own measurements and spectra, and we'll look together at whether it's imbalance or something else.

Let's be direct about the limits. We don't calculate your hourly downtime cost for you, and we don't replace your maintenance department. Vibration is one cause of failures among several, and an honest calculation shows its share, not your entire repair budget. If, after going through the data, it turns out your stoppages are mostly coming from shaft misalignment, loose fasteners, or the operating regime, we'll say so before the visit, not after.

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

Frequently asked questions

Management says 'but it's running, why touch it.' What do you say?

Don't argue about the vibration itself - you'll lose that argument, because the machine really is running. Bring three things instead: the baseline and current level for this machine, a trend line over the last few months, and the line's hourly downtime cost, agreed with finance. Then frame the request as a choice between two dates: a four-hour window at the next planned opportunity, or an unknown number of hours whenever the machine decides. That's a conversation where you have the arguments on your side.

We don't have a failure log or decent data. Where do we start?

Start with stores. Bearing and seal consumption by unit is the most honest database in the company, because it's kept for stock and money, not for reporting, and nobody dresses it up. Pull 24 months of it, break it down by unit, and the problem machines will show themselves. In parallel, add one 'planned or unplanned' field to the repair log and start filling it in from today. A year from now you'll have a proper dataset.

Can we just use ready-made savings figures from industry surveys?

Not for a real conversation. Published estimates of downtime cost vary by orders of magnitude, because they lump together incomparable operations, and whoever is making the decision usually knows that. Use an outside figure only to get a sense of scale, and to sanity-check your own calculation for a gross error. What goes into the document is your own figure, signed off by finance.

How do you calculate the damage if the failure hasn't happened yet?

Through the difference, and through risk, but without mixing the two. The difference is calculated from documents: the cost of a planned intervention today against the cost of an emergency version, based on the last real event on a similar machine. Risk goes on its own line, as probability multiplied by the cost of the consequence, and is explicitly labelled as an estimate. And the main argument in this case isn't a sum - it's the trend: a rising line shows how much time you have left to prepare.

Is it worth including excess energy consumption in the calculation?

Only if you've actually measured it. Motor current and actual output before and after the work, at the same mode and the same load. Imbalance by itself doesn't consume much; what's expensive is whatever is causing it - a product-caked or worn wheel, rubbing contact, widened clearances. If you don't have measurements, leave the line blank and mark it 'not calculated.' That's more defensible than putting in a percentage you'll be asked to justify.

What if the number of failures hasn't dropped after starting the program?

First, look not at the number of failures, but at their makeup and at total downtime hours. Converting stoppages from unplanned to planned, even at the same count, is a result in its own right, and it's visible in the money. If the hours haven't changed either, admit that honestly and test the hypothesis: maybe on your fleet, the main causes of failure aren't vibration-related at all, but lubrication, operating mode, parts quality, or installation errors. A conclusion like that is worth money too, because it stops effort being spent in the wrong direction.

Describe your equipment and the problem

We'll answer your questions, clarify the details, and let you know what's needed for an estimate and a visit.

Submit a Request WhatsApp Pricing