# On-Site Balancing of a Grinding Wheel Assembled With Its Arbor

> You read a wheel's imbalance not off a vibrometer but off the workpiece: regular waviness starts showing up, burn marks appear, size stops holding. We come to your site, mount sensors on the grinding-head housing, and balance the wheel assembled with its flanges and arbor at running speed. And we tell you right away if the vibration is coming from runout, the fit, or the bearings, because weights won't fix that.

**In short:** Yes, we balance the grinding wheel on site, without removing the arbor from the machine. The rotor here isn't the wheel — it's the whole assembly: the rear flange, the blotters, the wheel, the front flange, the nut, and the arbor. We mount two accelerometers on the grinding-head housing, radially, at the front and rear spindle bearings, and aim the laser phase sensor at reflective tape on the flange or arbor end. We fit mass only at the factory-provided locations: the flange's ring groove with slugs (movable weights) or threaded holes. Four conditions apply: the wheel has no cracks, dressing has been done, the spindle bearings are sound, and speed is up to roughly 12,000 rpm. And keep the main point in mind: wheel balancing isn't a one-off job — it's a scheduled operation repeated after every substantial dressing pass.

Source: https://axiline.pt/en/equipment/on-site-balancing-grinding-wheel-assembled/  
Publisher: AXILINE · Vila Nova de Gaia, Portugal · +351 931 831 229 · axilinegeral@gmail.com

## Symptoms: the workpiece suffers first, the machine gets loud after

An unbalanced wheel writes its own signature onto the workpiece. The first sign is regular waviness with a pitch equal to the feed per wheel revolution. The second is burn marks: contact pressure pulsates, the wheel presses harder then backs off, and at the moment of contact the grinding zone overheats. The third is that size drifts out of tolerance, and the wheel's profile wears unevenly, so dressing has to be done more often.

- [x] Waviness with a pitch equal to the feed per wheel revolution
- [x] Burn marks and temper colours at settings that used to work fine
- [x] Vibration increased right after fitting a new wheel, flanges, or blotters
- [x] The wheel ran smoothly on Friday; on Monday, after standing idle with coolant, it started humming
- [x] Dressing has to be done more often, and the picture is different after every pass
- [x] Diamond consumption has increased, the wheel loses its profile unevenly
- [x] The guard shakes on spin-up and coast-down; on a tool grinder the wheel can be heard knocking

> A check worth doing before we arrive. Measure vibration on the grinding-head housing with the spindle bare, then with the arbor fitted but no wheel, then with the whole assembly on. If vibration jumps up a step at the last stage, it's a balancing job and gets closed out in one visit. If it's already high with the bare spindle, the conversation will be about bearings and the fit.

## The wheel, flanges and arbor are one rotor

It's the assembly that needs balancing, not the wheel on its own: take it apart and put it back together, and the angular position of the parts changes, so the result has to be obtained all over again. That's why we mark the relative position of every part and the arbor in the spindle taper.

### Non-uniformity of the abrasive mass

The wheel is pressed and fired. Bond density, grain distribution and porosity are uneven throughout the volume, and so is the mass. The manufacturer balances the wheel to its own unbalance class, but that tolerance is set for safety and general use. It's usually not tight enough for precision grinding.

### Clearance fit

The wheel is fitted onto the arbor with clearance, otherwise it would crack when it heats up. That means it centres with an error of tens, sometimes hundreds, of microns. Add uneven compression of the elastic blotters and uneven tightening around the circumference, and you get a mass eccentricity that neither the wheel nor the arbor had on its own.

### Coolant inside the wheel

A vitrified-bond abrasive is porous and absorbs fluid. A wheel stopped after wet grinding drains coolant into its lower sector, and overnight a balanced wheel becomes unbalanced. There's no point balancing a wheel that's been standing with its lower part in the coolant bath until it's been run up and has shed the fluid.

### Dressing redistributes mass

The diamond removes a layer unevenly around the circumference and exposes fresh abrasive of a different density. After dressing the wheel is geometrically round, but the mass sits differently. Hence the strict order: dressing first, balancing after, never the other way round.

### Flanges, nut, arbor

The key, the locking screws, an unevenly walled nut, and a worn arbor taper each contribute their share. Most often, the set of factory balance slugs goes missing: the groove is there in the flange, but there are no weights in it.

> The practical consequence: the wheel's contribution and the arbor-with-flanges' contribution are separated by measurement. If one arbor is used with several wheels, it's worth trimming it once, then balancing only the wheel from then on. How this is done is covered in the section on on-site work.

## Dressing removes runout, weights remove imbalance

Two different causes live on a wheel, and both produce vibration exactly at the rotational frequency — the 1x component in the spectrum. Runout is geometry: the outer surface isn't concentric with the rotation axis. Imbalance is mass. A dial indicator shows only the first; weights fix only the second. Balancing a wheel that runs out badly is pointless: bearing vibration will drop, but contact with the workpiece will stay intermittent.

- [x] Wheel data plate: maximum operating speed, spindle speed, shelf life and storage conditions
- [x] Cracks and chips at the bore, ring-testing a dry, unmounted wheel
- [x] Radial and face runout of the wheel with a dial indicator after mounting and after dressing
- [x] Runout and cleanliness of the arbor taper and spindle seat, abrasive dust under the flange
- [x] Flanges of matching diameter, intact blotters of the same thickness, tightening torque per the instructions
- [x] Completeness of the factory balance weights and the condition of their locking screws
- [x] Spindle bearings: heat, noise, vibration-acceleration spectrum, signs of lost preload
- [x] A 15-30 minute warm-up at running speed, otherwise the measurement won't repeat

| What you see on the machine and the workpiece | What the instrument shows | Cause and what helps |
| --- | --- | --- |
| Waviness with a pitch equal to the feed per revolution, vibration grows with speed | 1x dominates, phase repeats from run to run | imbalance in the assembly — on-site balancing works |
| The wheel runs out, the dial indicator on the outer surface shows a noticeable spread | 1x is present, but dressing changes it more than any weights would | runout and fit eccentricity — fixed by dressing and cleaning the flanges |
| Vibration appeared after standing overnight, and dropped on its own by midday | 1x is present, phase shifts by tens of degrees between runs | coolant in the wheel's pores — there's nothing to balance |
| A periodic herringbone mark, squealing, noise depends on the settings | the peak isn't a multiple of the running speed, a broad band | chatter (self-excited vibration during cutting): fixed with cutting parameters, mounting stiffness, and dressing condition |
| Humming at all speeds, the front bearing runs hot | peaks at high frequencies, not multiples of running speed | spindle bearings — the unit needs repair |
| Vibration rises sharply in a narrow speed range | a narrow peak, phase swings by about 180° | resonance of the guard, column, or grinding head |

> We don't intervene in the spindle assembly itself: preload adjustment and bearing replacement are service-department work. Residual unbalance adds a rotating load to the supports, and an ISO 281 life calculation won't account for it unless you feed it in. Reading the spectrum and telling apart imbalance, misalignment, and bearing defects are covered in our separate articles.

Sources: [ISO 281:2007](https://www.iso.org/standard/38102.html) · [ISO 13373-3:2015](https://www.iso.org/standard/40840.html)

## How the on-site work proceeds

1. **Wheel, data plate, speed** — We check the wheel's maximum operating speed against the spindle speed and calculate the peripheral speed. We check for cracks, check the flanges, blotters, and the presence of the factory weights. We agree the setup and lockout procedure with the person responsible for the machine.
2. **Accelerometers on the grinding head** — Two accelerometers (vibration sensors), rigidly mounted on the metal housing, radially, as close as possible to the front and rear spindle bearings. Not on the wheel guard: the guard would show its own resonance instead of the rotor's vibration. We keep the direction the same across all measurements.
3. **Laser phase sensor** — We stick a single reflective tape onto the front flange, the arbor end, or the protruding end of the spindle. We aim the laser through a viewing window or a service hatch in the guard. On a belt-driven machine we take the mark only from the grinding spindle — the motor shaft would give the wrong phase.
4. **Baseline measurement** — The wheel at running speed, out of contact with the workpiece; we fix the coolant regime and keep it the same. We record overall vibration, the 1x amplitude and phase at both bearings, and the spectrum. Two consecutive runs show whether the phase repeats.
5. **Index check: whose imbalance is it** — We measure the assembly, stop the machine, rotate the wheel 180 degrees relative to the flanges using the marks, and measure again. If the vector rotated along with the wheel, the imbalance is in the wheel. If it stayed put, the arbor, flanges and fit are contributing it, and those are what need trimming.
6. **Trial weight with a factory slug** — We fit the weight only in the flange's ring groove or a factory-provided threaded hole, weighed on precision scales with the actual radius measured. A valid response is a change in 1x of at least 20-30 percent in amplitude or 20-30 degrees in phase. On a rigid grinding spindle, a few grams produce that.
7. **Calculation and fitting of masses** — The software gives mass and angle, or a fixed-position number. We translate this into the position of the factory weights in the groove and tighten the locking screws to the torque specified in the instructions. The angle's zero reference is where the trial weight sat.
8. **Confirmation run, dressing, marking** — The same regime and the same speed, with one or two iterations of weight adjustment if needed. We do a finishing dressing pass and repeat the measurement, so you can see how much dressing changes the imbalance. We mark the angular position of every part in the assembly.

> We work with the portable two-channel vibration analyser and balancer Balanset-1A: two accelerometers, a laser phase sensor reading the reflective tape, a two-channel USB module and software on a laptop. Single- and two-plane balancing by the influence-coefficient method: the instrument remembers how a trial weight changes the vibration and uses that relationship to calculate the correction. The software provides overall vibration and 1x, phase, speed, spectrum and time waveform, fixed positions, tolerance calculation by grade G, an archive and reports. We save the assembly's influence coefficients: after the next dressing, a repeat balancing job goes ahead without trial runs.

Sources: [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/) · [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/)

## One plane or two, and what figures we bring it down to

A single wheel assembly behaves like a disc, with a width-to-diameter ratio well below 0.5, and a single correction plane is enough. The manufacturer set the plane itself: the flange's ring groove. Exceptions below.

- The target residual 1x in the instrument's software. You set this number in mm/s, and it remains the procedure's internal criterion
- Overall vibration in mm/s RMS (root-mean-square value) on the grinding-head housing, for assessing machine condition. Record the applicable part and edition of ISO 20816, the measurement points and the operating regime in the work order — there are caveats there for machine tools
- Residual unbalance in g·mm/kg by grade G from ISO 21940-11. As a reference: G 1 for a wheel assembly, G 0.4 for precision grinding spindles. The unbalance class the manufacturer assigns a new wheel is a different scale with a different tolerance

| What's on the spindle | Planes | Why |
| --- | --- | --- |
| A single wheel on an arbor, width much less than diameter | One | A disc-shaped rotor, correction in one flange's groove |
| A wide wheel, grooves present in both flanges | Two | At large width a couple component remains |
| A set of wheels on one arbor | Two | Otherwise vibration at one of the bearings won't drop |
| A tool grinder with wheels on both ends of the spindle | Two, one wheel per plane | Two discs, spaced out along the shaft's length |
| A cup wheel, a face-grinding plate | One | One face accessible, mass into factory holes |
| A grinding head with a wheel on a long overhang | One; two at a large overhang | The farther the mass from the front bearing, the greater the moment |

> What grade G means in grams. The allowable specific unbalance equals G multiplied by 9549 and divided by the speed. For G 1 at 1,450 rpm that's 6.6 g·mm/kg: on a 20 kg assembly that works out to about 130 g·mm, roughly 1.3 g at a 100 mm groove radius. On a tool grinder at 2,850 rpm with an 8 kg assembly, the tolerance drops to 27 g·mm, about 0.3 g at an 80 mm radius. That's the source of the whole practice: weights are weighed on precision scales, and the radius is measured, not taken from a drawing. And a consequence for the future: as the wheel wears down, its diameter falls, speed is raised to keep peripheral speed up, and the gram tolerance on the same rotor gets tighter.

Sources: [ISO 21940-11:2016](https://www.iso.org/standard/54074.html) · [ISO 20816-1:2016](https://www.iso.org/standard/63180.html)

## Factory weights in the flange groove, and what's forbidden

There are exactly two places for mass here, and both are built into the design: the flange's ring groove with slugs, and threaded holes on the face of the flange or arbor. Everything else is forbidden. The convenience of the factory slugs is that their mass is known and identical, and the needed vector is built up with a pair of them.

- Starting position: all weights spread symmetrically around the groove, their vectors cancelling each other out. That's the zero point everything works from
- A pair brought together gives the maximum, spread 180 degrees apart gives zero. Intermediate values are set by the angle between them: the correction magnitude varies as the cosine of half that angle
- The correction direction is set by the bisector of the angle between the pair of weights, the magnitude by the angle's spread
- If the magnitude isn't enough, an additional slug of the same profile and the same manufacturer's series is added — never a homemade one
- The threaded holes work as fixed positions. The instrument gives a position number and a mass, ruling out a mirror-image error in angle direction
- The actual mounting radius is measured and entered into the software, otherwise the mass calculation would be wrong. The locking screws are tightened per the instructions and checked after a series of runs

- [x] Don't drill the wheel: the abrasive is brittle, and any hole is a stress concentrator and a crack
- [x] Don't drill or mill the flanges and arbor — the wheel's clamping depends on their geometry
- [x] Don't weld anything to the flanges or arbor
- [x] Don't glue weights onto the wheel, don't wind on wire, don't stick on putty: above 30 m/s peripheral speed it becomes a projectile
- [x] Don't fit magnetic weights, not even as trial weights
- [x] Don't exceed the maximum operating speed marked on the wheel. Speed is raised as the wheel wears down only within the limits of the data plate
- [x] A wheel with a crack, a chipped bore, or signs of impact isn't allowed to be balanced — it's taken off
- [x] Every run with the guard closed, the laser aimed through the window, and nobody standing in the wheel's plane of rotation
- [x] A new wheel is run idle behind the guard before the first measurement
- [x] Sensor cables are routed clear of the rotating zone, moving parts, and the coolant supply

> Peripheral speed is calculated like this: diameter in metres multiplied by 3.14 and by the speed in rpm, divided by 60. General-purpose vitrified-bond wheels are usually rated up to about 35 m/s; high-speed ones for a higher speed, and that's marked on the wheel itself. Balancing doesn't give you licence to push speed above the rated value. It gives you licence to run at the rated speed without defects. Methods for attaching correction weights and safety rules for on-site work are covered in our separate articles.

## Why a wheel is balanced regularly, not just once

On a fan, balancing is a repair event that happens once every few years. On a grinding wheel it's a routine maintenance operation, just like dressing. Imbalance doesn't come back from years of wear — it comes back from the normal actions you carry out every shift.

### After every substantial dressing pass

The diamond removes a layer unevenly and exposes fresh abrasive of a different density. The more precise the operation, the sooner it shows on the workpiece. With saved influence coefficients, a repeat balancing job goes ahead in a single run and takes minutes.

### After standing idle with coolant

Fluid in the pores drains downward. This isn't fixed by weights but by a schedule: don't leave the wheel sitting in the coolant, let it dry off idling after wet grinding, and rotate it periodically during a long shutdown.

### After every reinstallation

Removing and refitting changes the angular position of the parts and the nature of the fit. Marking the assembly and arbor recovers most of the result, but a confirmation reading is still needed.

### As the wheel wears down

Diameter decreases, speed increases, the gram tolerance tightens, and the abrasive's distribution changes. The same residual unbalance that was fine early on ends up exceeding the class by the end of the wheel's service life.

### After switching to a new batch of wheels

Wheels from different batches and different manufacturers behave differently. A new wheel's imbalance can't be predicted — it has to be measured.

- [x] Balance after mounting on the arbor, and again after the first dressing pass
- [x] Repeat after every substantial dressing pass on precision operations
- [x] Keep the assembly's saved influence coefficients: then a repeat cycle goes ahead in a single run
- [x] Keep the angular-position marking on the wheel, flanges, and arbor
- [x] Keep a log: speed, overall vibration, and residual 1x after every balancing job. A trend will reveal a bearing problem before it becomes a failure

## When on-site work won't succeed

### Speed above roughly 12,000 rpm

The instrument calculates the RMS vibration velocity (root-mean-square value) in the 5-200 Hz band, and 200 Hz is 12,000 rpm. Internal-grinding heads and electric spindles at 18,000 and 30,000 rpm fall outside that band; their tooling is balanced on a specialized arbor-balancing machine. We can still do bearing diagnostics and a vibration assessment on site.

### No groove and no threaded holes

No correction plane is provided, there's nowhere to put mass, and drilling the wheel, flange or arbor isn't allowed. What's left is balancing the arbor at the permitted locations, or ordering a flange with a balance ring.

### The wheel runs out, dressing hasn't been done

As long as the outer surface isn't concentric with the rotation axis, weights will give only a partial result, and contact with the workpiece will stay intermittent. Dressing first, balancing after.

### A wet wheel and a drifting phase

Coolant in the pores, play in the fit, a loose nut. The 1x phase wanders by tens of degrees from run to run — there's nothing to balance until the cause is eliminated.

### A worn bearing or lost preload

The vibration isn't from mass, weights won't remove it, and balancing would mask the one external sign of the defect. Repair the unit first, with balancing as the finishing operation.

### Chatter and resonance

Self-excited vibration in the machine-wheel-workpiece system and resonance of the guard or column are fixed with cutting parameters, dressing, mounting stiffness, and machine set-up. Masses on the rotor are useless here.

- A wheel with a crack, or past its shelf life, is taken off — not balanced
- Misalignment of the grinding-head drive is fixed by shaft alignment, not weights
- Grade-G certification when accepting a new assembly is done on a balancing machine with its own measurement setup

> Seven typical cases where balancing doesn't help are covered in a separate article of ours. On a visit, we go through this list first and tell you the outcome before you've paid for any weights.

## What you get, what it costs, and how to order

The result isn't just one number that dropped — it's two states of the assembly, measured under identical conditions, plus a method your own grinder operator can use to repeat the work.

As a price reference: vibration diagnostics with a report costs 300 EUR per unit, balancing adds from 250 EUR, minimum invoice for a visit 500 EUR. The calculator on the website works out an exact amount for your machine: it depends on the number of rotors and correction planes, access, and how far the site is. We're based in Vila Nova de Gaia, near Porto, and travel throughout Portugal.

- Measurement before and after at the same regime: speed, overall vibration in mm/s RMS, 1x amplitude and phase at the front and rear bearings
- Spectrum and time waveform, showing what dropped and what remains
- Masses and positions of the weights fitted in the groove, the actual radius, and a layout diagram marked right on the flange
- An assessment against grade G, where assembly mass, speed and correction radius are known
- A mechanical finding: runout of the wheel, taper and arbor, condition of the bearings, flanges and guard, and the machine's mounting
- Saved influence coefficients for the assembly and position markings, so the next job after dressing fits into a single run

> The engineers who design and manufacture Balanset instruments come out and do the balancing themselves. The same instrument can be bought: on a shop floor with many wheels and dressing every shift, it's usually more cost-effective to set up the method once and balance in-house, calling us in only for disputed cases. In your request, state the machine type and operation, spindle speed, wheel diameter, width and bond, assembly mass, whether the flange has a balance groove with factory weights, and what defect you're seeing on the workpiece.

Sources: [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/)

## Frequently asked questions

**Is it true the wheel has to be balanced after every dressing pass?**

On precision operations, yes. The diamond removes a layer unevenly around the circumference and exposes fresh abrasive of a different density, so the imbalance after dressing is different every time. You'll see it as waviness and burn marks on the workpiece, and as accelerated loss of the wheel's profile. A repeat cycle using the saved influence coefficients goes ahead without trial runs, in a single run, so in the schedule it's a matter of minutes, not half a day.

**The wheel is new, and the manufacturer already balanced it. Why balance it again?**

The manufacturer balances the wheel to its own unbalance class, and that tolerance is set primarily for safety during spin-up. It's looser than precision grinding needs. Plus your rotor isn't the wheel — it's the assembly: the clearance fit, the elastic blotters, the tightening, the arbor and flanges all introduce their own mass eccentricity. That's why the assembled unit is balanced at its own running speed.

**There isn't enough weight capacity in the groove. Can the flange or the wheel itself be drilled?**

No. The wheel can't be drilled: the abrasive is brittle, and a hole becomes a stress concentrator and a crack site. The flange and arbor can't be drilled or welded either — the wheel's clamping depends on their geometry and stiffness. The right approach: add a factory slug of the same profile from the manufacturer, use the threaded holes in the flange, or a second balance ring if the design has one.

**The wheel runs out. Is that fixed by balancing?**

No, runout and imbalance are different things, even though both produce vibration at the rotational frequency. Runout is geometry — it's visible with a dial indicator and is removed by dressing and cleaning the fit surfaces. Imbalance is mass — a dial indicator can't see it, and it's removed by weights. The order of work is strict: mounting, break-in, dressing, then balancing. Otherwise contact between the wheel and the workpiece stays intermittent no matter how many weights you fit.

**How can you tell whether the imbalance is in the wheel or the arbor?**

With an index check. We measure the assembly, stop the machine, rotate the wheel 180 degrees relative to the flanges using the marks, and measure again. If the vector rotated along with the wheel, the imbalance is in the wheel. If it stayed put, the arbor, flanges and fit are contributing it. The point is straightforward: the arbor gets trimmed once, and after that every new wheel is balanced with a short cycle.

**We have an internal-grinding head at 30,000 rpm. Can you do it?**

Not on site, with our equipment. The instrument calculates RMS vibration velocity in the 5-200 Hz band, i.e. up to roughly 12,000 rpm, and 30,000 rpm is 500 Hz. Tooling for such heads is balanced on a specialized arbor-balancing machine. Cylindrical, surface, and tool grinders fall well within our range. We can do bearing diagnostics and a vibration assessment of the high-speed head on site.