On-Site Mill Balancing: Fine Grinding, Product Purity, and Weights Without Welding

A fine-grinding mill punishes imbalance with more than just worn bearings. The clearance between the hammers and the liner here is a fraction of a millimetre, speeds start at three thousand rpm and go up from there, and not a gram of stray metal can be allowed to fall into the product. We balance rotors like this right in their own bearing housings, without dismantling anything, and we do it with methods that need no welding and add nothing extra to the chamber. Below is the detail for rotor and hammer mills, pigment grinders, and paint-production rotors.

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

Symptoms: what mill imbalance looks like in your plant

A fine-grinding mill rarely gives you a clean, obvious hum that tells the whole story. More often you notice the imbalance not by ear but through the product: particle size has drifted, the residue on the control sieve has grown, or metal particles have shown up in a sample. On machines like this, the clearance between the working elements and the liner is measured in fractions of a millimetre, and an unbalanced mass eats into that clearance first.

A second scenario is familiar to anyone working to a maintenance schedule. You stripped the rotor down for washing, put it back together, started it up, and the machine runs differently. Or you fitted a new set of hammers. Or you ran a shift on wet pigment and the level crept up. A mill almost always has an event the step change is tied to. Remember what it was, and half the diagnosis is already done.

• Vibration increased the same day the hammers, pins or paddles on the rotor were changed.
• After stripping the rotor down for washing and reassembling it, the level changed even though nothing else was replaced.
• Particle size has drifted, the residue on the sieve has grown, and the product is running hotter than usual.
• Metal particles have shown up in the product, or signs of liner wear.
• The machine shakes just as much at idle as it does under load. That means the cause is not the material feed.
• The level creeps up over the course of a shift and drops partway after the rotor is cleaned.
• Coast-down has become noisy: vibration spikes sharply in a narrow speed range, then fades.
• Bearing housings run hot, and the shaft seals have started leaking and drawing in product.

A useful figure to have before you call: overall vibration in mm/s RMS (root-mean-square value — what any vibration meter shows) at each bearing housing, taken at idle in the radial direction. If you can, take it twice: right before washing and right after. The difference between the dirty and clean readings tells us more than any description in words.

Mill rotor: three designs, and where the imbalance comes from

Rotors in this family fall into three layouts, and they are balanced differently. First: a stack of discs on a shaft with hinge-mounted hammers or rigidly fixed paddles, the rotor running between two bearings. Second: a single disc with pins or knives, mounted overhung, short and almost disc-shaped. Third: a long overhung shaft with the working element on the end, as on a dissolver or a vertical bead mill for paints.

What they have in common is different: speeds are higher than on crushers — fine grinding needs peripheral speed. Centrifugal force grows with the square of speed, so a mass nobody would notice on a crusher produces both vibration and contact on a mill.

And a third point. A mill rotor is almost always designed to come apart, because it gets washed. On a paint-mixing line, the rotor is removed and washed regularly for colour changes, sometimes several times a week. Every disassembly changes the angular position of the parts relative to the shaft, and the imbalance changes along with it.

• A spread of masses across the set of hammers, pins or paddles. At high speed, a difference of tens of grams already reads clearly on the instrument.
• One-sided wear on the working edges. Abrasive fillers and mineral pigments wear the elements unevenly, because the material inside the chamber is not distributed evenly.
• Product building up on the discs, hammers, and inside the rotor cavities. This is the most common problem with mills, and there is a separate section on it below.
• Reassembly after washing in a different angular position: the rotor has gone back onto its taper or key differently than it sat before disassembly.
• A loosened fit on the shaft: a worn keyway, a clamping sleeve that has lost its grip, an undertorqued taper nut. The rotor turns a fraction of a degree and shifts the phase — the angle by which the instrument locates the heavy spot on the rotor.
• Shaft bending after a contact event or after overheating. This is a common story on a long overhung shaft.
• Replacing one part instead of a pair. A new hammer is tens of grams heavier than its worn neighbor, and you get a step change in vibration out of nowhere.
• Leftover old product in blind cavities of the rotor that washing cannot reach.

Imbalance on a mill is almost always self-inflicted, and it comes back with every service. That is why we keep your rotor's influence coefficients on file — your particular machine's response to a known weight: the next balancing job runs off them, with no trial runs needed. More detail is in our article on the influence coefficient method.

What we check before the first weight, and what figure we're aiming for

Some mill jobs get closed out without fitting a single weight. Several causes contribute to vibration here at once, and balancing only fixes one of them: unbalanced rotor mass. So the first hour of the visit goes into measurement and inspection.

We compare overall vibration (the total level across all frequencies) with the 1x running-speed component — the part of the vibration exactly at the rotational frequency, which is what imbalance produces — look at the spectrum (vibration broken down by frequency), and always check the time waveform. Grinding produces a broadband impact component that smears across the spectrum and raises the noise floor. In the time waveform everything separates out: imbalance is a clean sine wave, one cycle per revolution, while contact, looseness and a crack produce regular spikes on top of it. Then we listen to the machine during coast-down — free deceleration after the drive is switched off. A long coast-down on a heavy rotor is convenient for finding resonance: one deceleration gives you almost the whole picture of amplitude and phase against speed.

And a word on the target right away. Permissible residual imbalance is inversely proportional to speed: at the same balance quality grade, a machine at 6000 rpm is allowed exactly half the specific imbalance of a machine at 3000. So a mill needs a stricter grade than a crusher of the same size, the correction has to be fitted more precisely, and the mounting radius has to be measured, not eyeballed.

• The rotor is clean. Measuring a caked-up rotor is pointless, and we don't start.
• The set of hammers, pins or paddles: every element present, intact, weighed, and matched by mass.
• The rotor's fit on the shaft: taper nut, clamping sleeve, key, signs of slipping, and whether the angular position is marked.
• Shaft and working-element runout, checked with a dial indicator while turning the rotor slowly. Mandatory on a long overhung shaft.
• Clearance between the working elements and the liner or screen, and the condition of the liner and its fastening.
• Bearing housings: play, noise, temperature, and the condition of the grease and shaft seals.
• Fastening to the frame and foundation: anchors, foot flatness, vibration isolators, cracks in the concrete.
• Drive: belt tension and condition, pulley imbalance of its own, and shaft alignment on a coupled drive.
• Resonance and proximity to critical speed: we check the 1x amplitude and phase during coast-down, and at neighboring VFD setpoints.

Misalignment and imbalance give similar numbers on a single channel, but are fixed in opposite ways. Balance a misaligned machine, and the weight compensates for a force that was never really there, so after shaft alignment the vibration ends up higher than it started. The order is: foot flatness, shaft alignment or belt tension, then weights. More detail is in our article on telling imbalance apart from misalignment. We take the balance quality grade and zone boundaries from the applicable parts of ISO 21940-11 and ISO 20816, recording the edition, measurement points, frequency band and operating mode in the report.

Sources: ISO 21940-11:2016 · ISO 13373-3:2015 · ISO 281:2007

How the work proceeds on site

A visit to one machine takes a shift. The measurements themselves are quick; the time gets eaten up by stopping the machine and preparing the chamber to be opened.

1. Step 1

   We agree on the zone, safety, and permitted operations

   Before the visit, we establish the key facts: whether the area is classified for dust explosion hazard, and whether hot work is permitted. Pigments, organic dyes, starch, feed mixes and powder coatings produce combustible dust clouds, and welding in areas like this is usually prohibited, or only allowed under a permit with full cleaning and purging of the chamber. Our instrument and laptop are not intrinsically safe, so we work on a stopped, cleaned and ventilated machine, and if needed keep the laptop outside the zone boundary. If the zone is permanently classified, we agree the work procedure with your industrial-safety department in advance, not on site.

2. Step 2

   We mount the sensors and the mark

   Two accelerometers (vibration sensors) go on the rotor's bearing housings, right up against the bearings, on a magnet on a cleaned-down pad or on a stud. Direction is radial, usually horizontal. On vertical mills and dissolvers, we take two mutually perpendicular radial directions, choose the one with higher vibration, and hold it fixed from then on. We stick the reflective mark on an exposed section of the shaft between the seal and the bearing, or on the coupling half, and aim the laser phase sensor through a hatch or a removed guard. In a dusty environment we check the mark before every run: a layer of product on it throws off the speed count and shifts the phase.

3. Step 3

   Baseline measurement, on a clean rotor

   Run at operating speed, always at idle, right after washing or cleaning. We record overall vibration in mm/s RMS, the 1x amplitude and phase, speed, spectrum, and time waveform on both channels. This is also where we decide whether fitting weights makes sense. If there is a VFD, we measure at the operating setpoint and at a couple of neighboring ones: on a high-speed machine this is a cheap way to spot resonance before you pay for it in extra runs.

4. Step 4

   Trial weight

   We mechanically fasten a temporary mass of known size at a known radius in the first plane, run the machine, then move it to the second. The instrument calculates the influence coefficients of your system: rotor, bearings, frame, foundation. A valid trial run changes the 1x amplitude by at least 20–30% or the phase by 20–30°. We do not use modelling clay, putty or magnets as a trial mass on a mill. At these speeds and inside this chamber, the trial mass has to hold just as securely as the permanent one, and be made of a material approved for your product.

5. Step 5

   Correction

   The software outputs a mass and an angle for each plane, or a position number if we are working from fixed positions: by hammer or pin count, or by the disc's factory holes. On a mill we work by position almost every time. Marking out with a protractor inside the chamber is awkward, and getting the angle-reckoning direction wrong — with or against rotation — costs you an extra stop with a full coast-down. We fit mass with mechanical fasteners or remove it by drilling; the next section covers this.

6. Step 6

   Check run, measurement under product, report

   Same speed, same points, same sensors. If we don't hit the tolerance right away, the software calculates an addition to the masses already fitted. Then we clear away swarf and burrs, recount the fasteners, close up the machine, and take a measurement in the working mode for comparison. The report gets the before-and-after figures, masses, radii, positions, material and fastening method, plus a list of everything introduced into the chamber.

Every iteration on a mill is expensive in time. Coast-down on a heavy rotor with good bearings can last minutes, the chamber has to be opened and closed, and on many production lines it has to be cleaned before opening too. So we plan for the minimum number of stops and keep the influence coefficients on file: at the next set change or after the next wash, we do a fine trim-balancing pass off them, with no trial runs.

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

One plane or two, and where they sit on a mill rotor

The number of planes is set by geometry: the ratio of the working length L to the diameter D at the weight-mounting zone. A short disc with L/D less than roughly 0.5 is usually brought into spec with a single mass. Anything elongated needs two planes, otherwise a moment (couple) imbalance is left behind, and you get the familiar picture: one bearing goes quiet, the other one shows no change at all.

Mills show both extremes, often in the same plant. A pin mill is close to a pure disc. A fine-grinding hammer mill with a long stack is an elongated rotor. We take the planes as far apart along the length as possible: the longer the moment arm, the smaller the required mass, and the less you're introducing into the machine.

A second plane costs one additional trial run — one extra stop with a full coast-down and reopening the chamber. On a mill that is a noticeable chunk of a shift, and we flag it in advance. We have a separate article on the L/D rule and choosing the number of planes.

Sources: ISO 21940-11:2016 · ISO 21940-12:2016

How to fit weights where welding is out and the product cannot be contaminated

This is the main question about mills, and it comes up almost every time. On a crusher, a weight gets welded on without a second thought. Here there are two constraints at once: combustible dust rules out sparks, and product requirements rule out introducing anything extra into the machine. That does not stop the work — it just means choosing a different method.

The rule we work from is simple. Nothing goes into the grinding chamber that could come loose, flake off, or discolour the batch. Everything introduced is counted and logged.

• Removing metal by drilling. The best option wherever disc thickness allows it. A weight that doesn't exist can never end up in the product. The software calculates the hole diameter and depth for the required mass, and we deburr and clean up the edge. Drilling also produces sparks and local heating, so in a dust-hazard area we do it on a stopped, cleaned and ventilated machine under an agreed procedure.
• Factory holes and factory fasteners. A weight washer under a hammer or pin nut, a longer bolt, an extra washer of the same material. Nothing new is introduced into the design; the mass of an existing part simply changes.
• The weight material matches the rotor material. On food, pharmaceutical and pigment lines that means stainless steel of the same grade. We do not use galvanized steel, brass, lead or aluminum: they produce colour contamination and wear particles that have no business being in the batch.
• Thread-locking is mandatory. Thread-lock compound, a lock washer, a lock nut, or a split pin, torqued to the manufacturer's documentation. A mill vibrates at high frequency and will unscrew anything that isn't locked.
• No adhesive, epoxy or putty weights anywhere the product makes contact. They flake off under heat and contaminate the batch, and what you get at the other end is a customer complaint, not a quiet machine.
• Factory-welded bosses. If the manufacturer built them into the rotor, they are ready-made, legitimate mounting points, and we use them first.
• Rearranging and matching the working elements. Weighing the hammers and arranging them so that the sums at diametrically opposite positions match often removes most of the imbalance without a single weight. We check this first, because it costs nothing.
• Accounting for what's introduced. Before closing the chamber, we recount the fasteners, check them against our log, and record the mass, radius, position, material and fastening method of every weight in the report. An error in the radius is a direct error in the mass.

Welding on a mill is possible where there is no combustible dust and no product-purity requirement — grinding mineral feedstock, for example. We confirm the rotor's steel grade and whether welding is acceptable from the manufacturer's documentation before the visit, not on site. Fastening methods are covered in more detail in our article on fastening correction weights.

If the rotor cakes up with product every shift

Buildup does more damage to balancing than anything else. Wet pigment, a resinous or greasy fraction, a heat-sensitive polymer settles unevenly on the hammers and inside the rotor cavities, gains mass over hours, and then breaks off in a lump at some random moment. You balance one state of the rotor, and by the next shift a different one is running.

Hence a hard rule: we measure and balance only on a clean rotor. If the rotor is caked up, cleaning comes first, measurement after. Otherwise you're paying for figures that won't survive to the end of the shift.

But a clean rotor is not the whole answer if it cakes up regularly. Then the task changes. What's needed is not to "fit weights" but to turn vibration into a manageable metric that drives your maintenance decisions.

• Balance the rotor while it's clean, and record that level as your baseline: a figure in mm/s RMS, the point, direction, speed, and mode.
• Take the same figure before and right after every cleaning. The difference shows how much mass the rotor picks up per cycle, and translates into a simple answer: how many hours until it's time to wash again.
• Set a vibration threshold based on that figure. Washing on-condition rather than on a calendar saves you both on vibration and on downtime.
• Don't chase the buildup with weights. Mass fitted to compensate for caked-on product turns into imbalance the moment the machine is washed.
• Get to the bottom of what's causing the buildup: chamber temperature, feedstock moisture, the condition of the liner and clearance, feed rate. It's often cheaper to remove the cause than to live with it every shift.
• Mark the rotor's angular position relative to the shaft before stripping it down for washing, and always reassemble it the same way. Otherwise every reassembly gives you a new imbalance, and the saved influence coefficients lose their meaning.

The same technique works with hammer wear. Vibration rises as the set runs, and that is normal. The value is not in any single figure but in the trend: it gives you a predictable date for the next service instead of an emergency shutdown. We have a separate article on vibration monitoring covering baselines, thresholds, and measurement routes.

When on-site work will not work

Let's be honest: on mills, the share of cases where weights don't settle the question is higher than on fans and pumps. There are two reasons. Fine grinding does not forgive small errors, and access to the rotor is often sealed shut.

• No access to the correction planes without fully dismantling the chamber. On bead mills and sealed grinding units, opening it up takes about as much work as removing the rotor entirely, and then it's more honest to take the rotor to a shop machine.
• The machine will not hold a stable speed. Calculating influence coefficients requires repeatability, and speed variation from run to run breaks the phase.
• The running frequency is close to the shaft's first critical speed — the speed at which the shaft starts to flex elastically. A long overhung shaft on a dissolver or a vertical mill behaves like a flexible rotor, and a correction made at one speed does not hold at another. We have a separate article on critical speed and flexible rotors.
• A bent shaft, a worn rotor fit, a clamping sleeve that has lost its grip, a sheared key. The 1x phase stops repeating from run to run, and the correction calculation becomes unreliable.
• Contact between the working elements and the liner or screen. In the time waveform this shows up as regular spikes on top of the sine wave. What's needed here is clearance, not a weight.
• Worn bearings and damaged bearing housings. Overall vibration is many times the 1x level, and non-synchronous components show up in the spectrum.
• Resonance in the frame, the service platform, or the housing. This is a common finding on a high-speed machine mounted on structural steel, especially after a VFD setpoint change.
• A dust-hazard zone that cannot be taken out of classification for the duration of the work. In that case the scope and order of operations are agreed with your safety department, and some correction methods are no longer available.
• The rotor is caked with product and cannot be cleaned before the measurement. We will not balance over a buildup.

The worst-case outcome for you sounds like this: the vibration is coming from the overall condition of the machine, not from rotor imbalance. In that case you get a measurement, spectra, time waveforms, and a prioritized diagnosis instead of weights. We have a separate article on cases where balancing doesn't help.

Sources: ISO 20816-1:2016 · ISO 13373-3:2015

What you get, and how to book a visit

The result of the work is not the phrase "it got better" but figures you can check again six months from now, and a list of everything introduced into the machine.

We are the engineers who design and manufacture the Balanset instruments, and we do the on-site balancing ourselves. We are based in Vila Nova de Gaia, near Porto, and we travel throughout Portugal. Vibration diagnostics with a report cost EUR 300 per unit, balancing adds from EUR 250, the minimum invoice for a visit is EUR 500. The calculator on the website gives you an exact figure for your machine. If you have several mills on site, count them all at once: one visit.

• Mill type and what you are grinding, including the dust's combustibility hazard.
• Rotor operating speed, whether there is a VFD, drive power, and whether the drive is belt or coupled.
• Approximate rotor mass, working-section length, diameter, number of hammers or pins, and the mass of one element.
• Bearing arrangement: rotor between bearings or overhung, horizontal or vertical configuration.
• Requirements for materials in contact with the product, and the rotor's steel grade.
• Whether the area is classified for dust explosion hazard, and whether hot work is permitted.
• Access to the rotor when stopped: hatches, a removable chamber cover, a service platform.
• What was done to the machine before the vibration appeared: hammer replacement, washing with the rotor stripped down, bearing repair, a speed setpoint change.
• Measurement figures in mm/s and the points, if you have measured, and photos of the rotor, bearings, and the machine's nameplate.

• Measurement before work: overall vibration in mm/s RMS and the 1x running-speed component at each bearing housing, speed, points and directions.
• Measurement after, in the same mode and at the same points, so the drop in 1x is directly visible.
• Before-and-after spectra and time waveforms, showing what remains in the vibration beyond imbalance.
• The mass, radius, position, material and fastening method of every mass fitted, and hole diameter and depth wherever metal was removed.
• A list of parts introduced into the machine, with a fastener count checked before opening and after closing the chamber.
• A baseline vibration level on the clean rotor, so you can track buildup and wear by trend going forward.
• An assessment of machine condition by overall-vibration zones, noting the applicable part and edition of the standard.
• A residual-imbalance and G-grade tolerance calculation, if needed for acceptance, plus a list of mechanical findings.

"Within tolerance" in the software means one thing: residual 1x below the target value you set. It is not an assessment of the machine's overall condition by overall vibration, and not confirmation of a G balance quality grade. We keep these three tolerances separate in the report. If you want to do this yourselves, we sell the same instrument we use. Balanset-1A: two accelerometers, a laser phase sensor working off a reflective mark, a two-channel USB module, software on a laptop, one- and two-plane balancing by the influence coefficient method, overall vibration and 1x, phase, speed, FFT spectrum and time waveform, fixed positions and drilling calculations, saved influence coefficients, trim balancing, G-grade tolerance calculation, an archive and reports.

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

Frequently asked questions

Related content