AXILINEAcceptance Testing After Repair: Vibration, Baseline and the Report
A contractor finishes repairing an induced-draft fan. The machine turns, there are no odd noises, you sign an acceptance report saying “no issues found.” A week later a support runs hot, a month later it's reading 6 mm/s, and there's nothing to argue with: the report has no numbers, no regime, no measurement points. Below we go through how to accept a machine so that, a month later, you have a measured baseline to compare against, not a memory of how things felt.
Acceptance is a measurement, not a signature
A hands-and-ears walk-round answers the question “is there nothing obviously alarming.” Acceptance testing answers a different question: does the machine meet a number named in advance, and exactly what condition is it in as it goes into service. The difference between these two questions becomes visible at the exact moment vibration has risen and someone has to decide who fixes it at their own expense.
Keep in mind that acceptance testing gives you two different results. The first is one-off: an “accepted” or “not accepted” decision at every point. The second works for years: a set of numbers and spectra from the sound machine that you'll come back to at the first hint of suspicion. The second result is almost always the more valuable one, and writing it into the report takes ten minutes.
Let's be clear about where the method's boundary sits. A vibration reading doesn't check assembly quality as such: it won't tell you whether the bearing was fitted correctly, or whether the right grease went in. It checks the outcome of the assembly, in aggregate, through how the machine behaves at its operating regime. That's why vibration acceptance follows the mechanical checks, and doesn't replace them.
And one more thing that causes especially frequent disputes. The words “in tolerance” in a balancing instrument's software are not machine acceptance. They mean only that the residual running-speed component dropped below a target value that a person entered into that software. We have a separate article on the three different tolerances that all get called by the same word.
What exactly gets measured at acceptance
An acceptance reading is a set of quantities, not a single number. Each one closes a specific gap in the evidence: drop any one of them, and you lose a specific argument in a future dispute.
Take points at every bearing support on the unit, not just wherever's convenient to reach. For a motor-plus-driven-machine pair, that's usually four supports. Three directions: horizontal-radial, vertical-radial and axial. Axial is especially needed where you expect misalignment after setting the machine in place, and on overhung rotors.
Measure support temperature with a pyrometer on the housing, at the same point every time. This is a second, independent channel of information that costs little and pays off regularly: a bearing tightened down too hard will show a temperature rise before a clear rise in vibration velocity does.
| What we measure | Where | Why it belongs in the report |
|---|---|---|
| Overall vibration velocity, mm/s RMS | every bearing support, directions H, V, A | This is the number that gets compared against the acceptance criterion and the ISO 20816 zones |
| 1x running-speed component, mm/s | same points | Shows what share of the overall vibration comes from unbalance. If 1x is small and the overall level is high, the machine was accepted with something else going on inside |
| 1x phase, degrees | same points | Comparing phases at two supports tells static unbalance apart from a couple, and hints at misalignment. On a repeat diagnosis months later, phase tells you more than amplitude does |
| FFT spectrum | at least one point per support, in the most informative direction | The vibration picture broken down by frequency, which you'll compare against later. Six months on, the question isn't “how much has it gone up,” it's “what exactly has risen” |
| Time waveform | supports that raised doubts | Individual impacts, clipped peaks, modulation. The spectrum averages these out and hides them |
| Rotational speed, rpm | from the tachometer, actual | Without the actual speed, the reading can't be compared to future ones. Unbalanced force grows with the square of rotational frequency |
| Load and regime | from the unit's own instrumentation: output, pressure, motor current, damper position | Half of future “vibration increases” are explained right here, with no need to look further |
| Bearing support temperature, °C | pyrometer on the housing, at a marked point | A second, independent indicator. Catches over-tightening and grease shortage before vibration velocity does |
One point per unit doesn't count as acceptance testing. Unbalance, misalignment and looseness show up at different supports and in different directions, and a single reading at the most accessible support can lead you to accept a machine with a problem at the opposite end of the shaft.
Acceptance conditions: what state the machine must be in at the time of the reading
A reading with no recorded regime can't be compared against the criterion, or against a future reading. This is the cheapest place for acceptance testing to fall apart, and the most frustrating one, because it's fixed with discipline, not money.
Warm-up is mandatory. A cold machine shows different clearances, different grease viscosity and different bore geometry. The working sign of warm-up: bearing support temperature has stopped rising, with a change of no more than two to three degrees over ten minutes. On most industrial units that's thirty to sixty minutes of running.
A word on machines with a variable-frequency drive. Acceptance at a single frequency doesn't cover the whole operating range: at a different rotational frequency the unit can hit resonance, and there won't be a word about it in the report. Sweep the operating range and take readings at at least three frequencies, including both endpoints.
Now for the honest part. On handover day, the normal load is often unavailable: no product feed, the line isn't assembled yet, the ductwork isn't ready. Don't pretend idle running will do. Split acceptance into two stages: a preliminary one at whatever regime is available, with a direct caveat in the report, and a final one at the normal regime by an agreed date. Until that date is named in the report as an actual figure, final acceptance will never happen.
- The machine is warmed up: support temperature is stable to within two to three degrees over ten minutes.
- Speed is at the normal value, recorded from the tachometer, not copied off the nameplate.
- Load is at the normal value, and the figure is recorded: output, pressure, motor current.
- Dampers, gates and valves are in their normal operating position, and the position is recorded in the report.
- The support type and installation method match the design: rigid or compliant foundation, the condition of the vibration isolators, all anchor bolts tightened.
- Soft foot (one foot not sitting flush: tightening the bolt distorts the housing) has been checked for and fixed before start-up, not after the first bad reading.
- Shaft alignment has been performed, and its result is attached as a separate report.
- Sensors sit on marked, cleaned pads, the mounting is the same at every point, and it's recorded in the report.
- The quantity, units and frequency band are set before the reading, and match what the criterion is written in.
- On a variable-speed drive, the entire operating speed range has been swept.
Baseline: the main result that goes into the archive
An “accepted” decision lives for one day. A baseline lives until the next major overhaul, and it's exactly what turns future readings into a meaningful picture.
The moment of acceptance is unique in that the machine is in the best condition it will ever be in: fasteners tightened, shafts aligned, bearings new, the rotor balanced if needed. There won't be another moment like it before the next repair. Record this condition as your zero point.
Don't take the baseline with a single reading. Take two or three in a row, removing and remounting the sensor each time. The spread between them shows your own procedure's inherent noise, and tells you which future change already means something, and which one sits inside the mounting error. A spread bigger than ten to fifteen percent means the measurement procedure needs fixing, not the machine.
The baseline is specific to the point, not to the machine. The motor support and the fan support on the same unit give different numbers, sometimes several times over. Two identical pumps off the same shelf will give different baselines too, because their foundations and pipework differ. That's why the baseline values in the report are listed point by point, not as a single line saying “vibration normal.”
Put more than just a figure into the archive. The spectrum, the 1x phase, the speed, the regime, the temperature, and a photograph of every point with the sensor mounted. Six months later, that photograph will settle an argument over exactly where the sensor sat faster than anyone's memory will. How these numbers later feed into setpoints and a trend is covered in our article on vibration monitoring and the baseline level.
A baseline benefits both sides equally. For the client, it's proof of deterioration. For the contractor, it's proof that the machine was handed over sound, and whatever rose did so afterward. Without a baseline, arguing over the quality of a repair is a game both sides lose.
The acceptance criterion: how to write it so it can be checked
The phrase “vibration within normal limits” isn't a criterion. A checkable criterion has four parts: the quantity and units, the frequency band, the numerical value, and a reference to the applicable part and edition of the standard. Drop any one of the four, and a month later you'll have two engineers with two equally plausible opinions.
For overall machine condition you assess overall vibration velocity on non-rotating parts, usually as mm/s RMS over the 10–1000 Hz band. The ISO 20816 zones read as follows: zone A for a new or repaired machine, zone B for extended operation with no restrictions, zone C for restricted operation while the cause is addressed, zone D unacceptable. That gives you a natural reference point for post-repair acceptance: zone A, with the top of zone B as an agreed compromise if the machine has never historically fallen inside zone A.
The zone boundaries depend on the machine group, power, rotational speed, and whether the foundation is rigid or compliant. It's easy to get this wrong in the contractor's favor, or against them. Check exactly which part and edition of the standard applies to your machine: parts have limits on power and speed, and exclusions by machine type. Check small fans, belt-driven units and mulchers on mobile equipment separately, because they often fall outside the scope. There's a detailed breakdown of zones and groups in our article on vibration limits.
If the repair scope included balancing the rotor, the residual-unbalance criterion goes on a separate line, in different units: g·mm or g·mm/kg, balance quality grade G per ISO 21940-11. This is a property of the rotor, not the machine, and it doesn't substitute for assessing overall vibration. The reverse holds too: a machine in zone A doesn't prove the rotor was balanced to the stated G grade.
The measurement band has to cover the band the criterion is written in. An instrument that computes RMS vibration velocity over the 5–200 Hz band is more than adequate for running-speed defects and balancing, but that's not the 10–1000 Hz band. If contractual acceptance specifically requires 10–1000 Hz, agree on the instrument in advance, not on handover day. While you're at it, check the detection type and representation too: 4.5 mm/s RMS is 6.4 mm/s peak, and swapping one for the other manufactures an “exceedance” where nothing actually changed.
Sources: ISO 20816-1:2016 · ISO 21940-11:2016
The acceptance procedure: eight steps
- Step 1
Agree the criterion and the points before the repair starts
Quantity, units, frequency band, numerical value, the applicable part and edition of the standard, the numbering scheme for supports and directions, the measurement regime. All of this on paper, and before the machine is taken apart. Once it's disassembled, any one of these lines becomes negotiable.
- Step 2
Close out the mechanics before start-up
Anchor and foundation bolt tightness, soft foot, the condition of the frame and its welds, clearances, the condition of the vibration isolators. Perform shaft alignment and attach its report separately. Misalignment is the single most common cause of vibration appearing right after a repair, and on a vibration reading it shows up as a 2x component (at twice rotational frequency) and axial vibration, but you'll still fix it with alignment either way.
- Step 3
Mark and label the points
Clean the pads down to bare metal for the magnet's footprint, or fit studs. Label the machine number, point number and direction, and mark axial with an arrow. Photograph every point with the sensor mounted: you'll need that photograph again.
- Step 4
Start the machine and bring it up to its normal regime
Warm-up to a stable support temperature, normal speed, normal load, normal damper position. Record the actual regime figures at the time of the reading, not the design ones.
- Step 5
Take the full set at every point
Overall vibration velocity, 1x amplitude and phase, speed, spectrum, support temperature. A two-channel instrument takes both supports in one run, in one regime, so you never have to wonder whether the machine was running the same way twenty minutes ago.
- Step 6
Read the picture, not just the numbers
Work out the 1x share of the overall vibration at each support. Look for a noticeable 2x and axial component, a harmonic comb (a series of peaks at frequencies that are multiples of running speed), a raised noise floor, peaks above a kilohertz. Acceptance in zone A with an unexplained spectral peak is an accepted machine with a deferred problem.
- Step 7
Didn't pass: name the cause and the deadline, not “issues noted”
The phrase “accepted with issues noted,” with no figure and no date, doesn't work. Write down which point and which direction is over, what the actual reading was, what the criterion is, what's suspected as the cause, and by what date the repeat reading will happen.
- Step 8
Complete the report and file the baseline in the archive
The report follows the structure in the next section, plus attachments: spectra, the polar diagram and balancing report if the rotor was balanced, the alignment report, photographs of the points. File the point-by-point baseline values wherever you keep your equipment history, the same day.
Acceptance report structure template
Below is a structure you can carry over into your own form as-is. The point is that every line answers a question someone is bound to ask a month from now.
| Report section | What to write | Example wording |
|---|---|---|
| Site and machine | Designation per the site diagram, type, power, rated speed, foundation type | Induced-draft fan D-1, tag 12-V-01, 55 kW, 1480 rpm, rigid foundation |
| Unit and scope of work | What exactly was repaired, and what was done to the rotor | Bearings replaced at supports 1 and 2, impeller balanced in two planes, shafts aligned |
| Date and participants | Date, start and end time of the readings, who was present from the client and the contractor | 27.07.2026, 10:40–11:25, attended by the chief engineer and a contractor representative |
| Regime and conditions | Actual speed from the tachometer, load, damper or valve position, support temperature, warm-up duration | 1478 rpm, damper 100% open, motor current 92 A, support temperature 48 and 51 °C, 45 min warm-up |
| Points and directions | Numbering scheme and the full list | Supp. 1 and supp. 2 on the machine, directions H, V, A; supp. 1 and supp. 2 on the motor, directions H, V |
| Measured quantity and units | What exactly was measured, and in what | RMS vibration velocity, mm/s |
| Frequency band | The measurement band, not “broadband” | 10–1000 Hz |
| Values by point | One line per point and direction: overall vibration velocity, 1x amplitude, 1x phase, support temperature | Supp. 1 H: overall 1.3 mm/s, 1x 0.9 mm/s, phase 214°, 48 °C |
| Acceptance criterion | Numerical value, zone, reference to the applicable part and edition of the standard | No more than 1.4 mm/s RMS over the 10–1000 Hz band at every support, zone A under the applicable part of ISO 20816 for this machine group |
| Conclusion by point | Separately for each point, with the actual value, not a blanket phrase | All points in tolerance, maximum value 1.3 mm/s at supp. 1 H |
| Residual unbalance | If the rotor was balanced: the masses and radii of the weights fitted per plane, the residual value in g·mm or g·mm/kg, the stated G grade | Plane 1: 34 g at R 310 mm; plane 2: 21 g at R 310 mm; residual unbalance meets grade G 6.3 per ISO 21940-11 |
| Baseline | The same point-by-point values, marked as baseline, with the repair date | Take the table's values as the baseline dated 27.07.2026 |
| Who measured, and with what | Name and job title, instrument type and serial number, sensor type, mounting method, measurement settings | Balanset-1A instrument, serial no. …, two accelerometers on magnets, laser phase sensor against a reflective marker |
| Attachments | Spectra by point, time waveforms, the polar diagram and balancing report, the shaft alignment report, photographs of the points with sensors fitted | Attachments 1–5, 7 pages |
| Acceptance limitations | What regime the reading was taken at, what couldn't be reproduced, the date for final acceptance | Reading taken at 100% load; no limitations |
| Signatures | Signatures of both parties and the date | Client and contractor signatures |
A report with no frequency band, no regime and no point labels isn't evidence of anything. These are exactly the three lines that get dropped most often, and exactly the ones disputes fall apart on a month later.
Sources: ISO 20816-1:2016 · ISO 21940-11:2016
Six disputes that start after the signature
Accepted at idle
The most common and most expensive case. A fan with no load runs at a different point on its performance curve, with different aerodynamics and a different temperature. A pump with no pressure shows neither cavitation nor hydraulic forces. A motor disconnected from its coupling physically cannot show misalignment. The right move: two acceptance stages, with a direct caveat in the report and a named date for the final reading.
Measured on the cowling or a guard
A cowling, a mesh guard, a cladding panel and piping are all separate oscillating systems with their own natural frequencies. You can easily get 12 mm/s on a fan cowling where the bearing support reads 2.5 mm/s, because the panel has hit resonance at running speed. The reverse case is no better: soft cladding damps the signal, and you accept a machine with a real defect. The point goes on the bearing support only, as close to the bearing as possible.
No baseline taken
A month later the client says “it's gotten worse,” the contractor says “it's always been like that.” Both are being honest, and neither can prove anything. What follows is haggling over an absolute figure that, on its own, says very little: a 3.4 mm/s level that's held steady for three years is normal operation, while that same 3.4 following 1.2 a month earlier is a warning sign.
Compared against the wrong machine group's limit
A reading of 4.5 mm/s is already zone C for a small fan, and normal operation for a large crusher on a rigid foundation. Check the foundation row separately: a compliant foundation allows more vibration than a rigid one, because it damps the oscillation. In plenty of tables floating around online, these two rows are swapped, and acceptance testing against a table like that hits one side or the other at random.
The two sides used different quantities and bands
The client measured true RMS over the 10–1000 Hz band, the contractor took a peak reading over the 5–200 Hz band with a different instrument and different sensor mounting. Both instruments are working fine, and the discrepancy is a multiple. Add to that an instrument that defaults to ips, inches per second: a modest-looking 0.28 ips on the display is 7.1 mm/s. Quantity, representation, band and mounting method all get fixed before the reading, not after.
A balancing report passed off as an acceptance report
The instrument's software said “in tolerance,” the contractor attached the printout, and considers the matter closed. These are different documents about different things. A balancing report speaks to the residual running-speed component against a target value a person entered. An acceptance report speaks to the machine's overall vibration at every support, in its normal regime, against a named standard.
If the machine fails acceptance
Don't sign the report until you've worked out exactly what's risen. The first step takes a minute: work out the running-speed component's share of overall vibration at each support. If 1x accounts for most of it, the unbalance hypothesis stands and balancing is appropriate. If overall vibration is high and 1x is small, weights won't help, and sometimes will make things worse.
The typical set of causes behind a machine failing acceptance specifically after a repair is short and predictable. Misalignment after setting the unit in place gives a 2x and a noticeable axial component, and is cured by shaft alignment, not weights. Soft foot and under-tightened anchor bolts give looseness, a harmonic comb, and a raised noise floor. Resonance that's appeared after reworking the frame or replacing the vibration isolators gives itself away with a narrow peak at running speed and a roughly 180° swing in 1x phase as you pass through it. An over-tightened or incorrectly fitted bearing gives a temperature rise at the support before a clear rise in vibration velocity. An impeller assembled with different geometry, or with leftover build-up, gives pure unbalance, which is removed by on-site balancing.
On-site balancing covers a specific share of the cases on this list, and we won't pretend it covers all of them. What it does do is cover them without taking the unit apart, in a single site visit, on the real supports and the real foundation — that is, in the exact system the machine will actually run in afterward.
AXILINE's engineers are the same people who design and manufacture the Balanset instruments and balance with them on site themselves. We come out for the acceptance testing, mark and label the points, take the full set of quantities on two channels at once, read the spectrum, and tell you plainly whether it's unbalance, misalignment, looseness, a bearing, or resonance. If it's unbalance, we balance in one or two planes on site and document the result with a report including the polar diagram, the masses and angles per plane, a residual-unbalance calculation by G grade, and baseline values by point.
If you want to run acceptance testing yourself, you can get the instrument. The Balanset-1A is a portable two-channel vibration analyzer and balancer: two accelerometers, a laser phase sensor against a reflective marker, a two-channel USB module with preamplifiers, integrators and an ADC, Windows software. It displays overall vibration and the running-speed component side by side, along with phase, speed, an FFT spectrum and the time waveform, calculates the correction by the influence-coefficient method, supports fixed positions and the drilling calculation, saved influence coefficients, trim balancing, a tolerance calculation by G grade, and keeps an archive of reports. For building into machines and rigs, there's a Balanset-1A OEM version without the case. Advisory support is included.
Sources: ISO 20816-1:2016 · ISO 21940-11:2016
Frequently asked questions
Can you accept a machine at idle if the normal load isn't available on handover day?
Not as final acceptance. A fan with no load runs at a different point on its performance curve, a pump with no pressure shows no hydraulic forces, and a motor with no coupling physically cannot show misalignment. Run two stages: a preliminary reading at whatever regime is available, with a direct caveat in the report, and a final one at the normal regime by a named date. The key is that the date for the final reading has to be an actual figure in the report, or it will never happen.
What value counts as normal at post-repair acceptance?
The reference point for a repaired machine is zone A under ISO 20816, with the top of zone B as an agreed compromise if the machine has never historically fallen inside zone A. The specific boundaries depend on the machine group, power, speed, and whether the foundation is rigid or compliant, so there's no single universal number to name. Check which part and edition of the standard applies to your machine: parts have limits on power and speed, and exclusions by machine type. Check small fans and belt-driven units separately.
Do you have to measure in three directions at every support?
The working minimum for acceptance is horizontal-radial and vertical-radial at every bearing support, plus axial wherever you expect misalignment or the rotor is overhung. The full H, V and A set at every support is better, and costs a few extra minutes. Cutting corners here is risky because misalignment often shows up specifically in the axial direction, and it's one of the main causes of vibration right after a repair.
There's no baseline from before the repair. Does acceptance lose its point?
No, the opposite: acceptance is the best moment for a baseline to finally exist. The machine is now in the best condition it will ever be in, and that's exactly what you record as your zero point. If you want a “before and after” comparison and there were no pre-repair readings, say so in the report, and compare against the absolute zones instead, then against the slope of the line from future readings.
How long should you warm up the machine before an acceptance reading?
Until the bearing support temperature stabilizes. The working sign: a change of no more than two to three degrees over ten minutes. On most industrial units that's thirty to sixty minutes of running at the normal regime. Record the warm-up time and the support temperatures at the time of the reading in the report: without them, the next reading won't be comparable, because a cold machine and a warmed-up one show different clearances and different geometry.
Is a balancing report enough instead of an acceptance report?
No, these are documents about different things. A balancing report speaks to the residual running-speed component against a target value a person entered into the software, and to the rotor's residual unbalance by G grade under ISO 21940-11. An acceptance report speaks to the machine's overall vibration at every support in its normal regime, against the applicable part of ISO 20816. A balancing report becomes an attachment to the acceptance report, not a substitute for it.
Related content
Vibration came back after balancing: why, and what to check
Balancing fixes the rotor's mass distribution at the moment of the visit. If the machine came within tolerance and then, weeks later, the vibration climbed back up, look not for an error in the calculation but for a change in the machine: product build-up on the impeller, blade erosion, a correction weight that came loose, a developing bearing defect, loosened fasteners, or a different operating mode. You tell them apart by the shape of the rise, by how the amplitude and phase of the running-speed component 1x — the vibration at the rotor's rotation frequency — behave, and by whether it's only that component that rose, or the overall vibration too. The procedure is the same every time: a repeat reading compared against the report from the previous balancing job, and only then a decision on whether weights are needed.
Vibration Measurement Units: mm/s, g, µm, and What RMS Means
Vibration is described by three quantities: displacement in µm (usually peak-to-peak), velocity in mm/s (usually RMS — root-mean-square value), and acceleration in m/s² or g. Displacement works at low frequencies and on shafts, velocity gives a universal condition assessment for housings over the 10–1000 Hz band, and acceleration shows up high frequencies, bearings and impacts. Converting between quantities is only possible for a single component at a known frequency, and the 1.41 factor between RMS and peak only holds for a sine wave. That's why a figure with no stated quantity, amplitude type, frequency band and measurement point simply has nothing to be compared against.
Rotor Balancing After Rewinding and Repair: Assembled Trim Balancing On Site
Yes, we balance rotors on site, assembled and at running speed, after rewinding, repair and bearing replacement. Two conditions apply: the running-speed component 1x — vibration at the rotor's rotational frequency, which is what imbalance produces — must dominate the spectrum, and the fits and bearings must be sound after the repair. If the rotor isn't assembled into the machine yet, it's more honest to balance it on a lathe in the shop first, then take a confirmation reading after assembly and do an assembled trim balance if needed. Lathe balancing of the bare rotor doesn't account for the cooling fan, the coupling half, the key, or the stiffness of the actual bearing supports, so "balanced in the shop" and "runs quietly in the machine" are not the same thing.
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.