Mounting correction weights: welding, bolting, riveting, and removing metal by drilling

The instrument works out the mass and angle in five minutes. What comes next is the part of the job that decides whether the balancing holds for years or comes back in a month, along with a second trip out. Here's what actually holds a correction weight in place: welding, bolting, riveting, clamping - and when it's better to remove metal by drilling instead of adding it.

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

What you're actually deciding when you choose how to mount the weight

The instrument comes back with 48 g at an angle of 63° on a 220 mm radius. What comes next is the part no algorithm calculates for you: finding where to physically put that mass, and what to hold it there with.

The mounting method settles three things at once. Whether the weight holds at operating speed and under heat. Whether the mass stays exactly where, and at the radius, the instrument calculated it for. And whether it gets in the machine's way - the flow, the clearances, servicing.

A mounting error shows up on the instrument differently from a calculation error. A calculation error produces vibration right away, on the verification run, and you fix it during the same visit. A mounting error waits. A week, a month, the first hot start-up. Then the 1x running-speed component (vibration at the rotation frequency - exactly what imbalance produces) jumps in a single run, the phase - the angle of that oscillation - shifts by tens of degrees, and you're making a second trip.

That gives you a working rule. Mount the correction weight as if it's going to stay there for the rotor's entire service life. The same goes for the trial weight - it's spinning at the same speed and the same radius.

The instrument calculates the product of mass and radius. Everything you do by hand needs to preserve exactly that product. A 20% deviation in mass or in radius produces a residual unbalance that wasn't in the calculation.

Welding: a full weld, not two tacks

On steel rotors, welding remains the primary method. It just has to actually be a weld.

A two-point tack weld isn't a mounting method - it's a way to trial-fit the weight. Vibration applies an alternating load, first one way then the other, right at the root of the tack, and that's exactly where a crack starts growing. A continuous weld around the plate's perimeter spreads the load along its full length and works in shear, rather than pulling a single point apart.

Preparing the pad matters just as much as the weld itself. Paint, primer, filler, product buildup, oil - all of it ends up sandwiched between the weight and the rotor. Clean down to bare metal, with margin beyond the plate's outline, degrease, then weld. A weight welded on top of paint eventually comes off along with the paint.

Heat works against you. Welding puts heat into one spot, and on a thin-walled impeller that causes distortion: a blade pulls, a disc warps. That distortion creates an imbalance of its own that wasn't there before you arrived. The thinner the wall, the smaller the plate and the shorter the weld need to be - and the more a bolt starts making more sense than welding.

The material limits your choices hard. Carbon and low-alloy steel weld without issue. Aluminium and cast wheels, cast iron, and stainless impellers in food production almost always mean the answer is 'don't weld.' On stainless steel, the heat-affected zone loses its corrosion resistance, and a year later you'll see rust running exactly along the outline of your weld.

A separate point on coatings. Plating, zinc, polymer, and enamel all burn away next to a weld, leaving you with a corrosion site on a rotor that had otherwise stood for fifteen years. Restoring the coating is part of the job, not an afterthought.

• A continuous weld around the plate's perimeter, not two tacks and not a single spot in the middle.
• The pad is cleaned to bare metal and degreased, wider than the weight's outline.
• The welder's return cable is clamped directly onto the rotor, right next to the weld. Current passing through the bearings burns tracks into them, and you'll end up with a new problem instead of a solved one.
• The plate is a compatible material, with no sharp edges facing the flow.
• Hot work is cleared on its own permit before the visit. Missing this is a common way to lose half a day.
• The coating is restored after cooling, and spatter and slag are cleaned off.
• The verification run happens after everything has cooled. A hot rotor is distorted, and a measurement taken right after welding won't show what it will an hour later.

A tack weld is acceptable in exactly one case: you're trial-fitting a weight to check that it brings you within tolerance, and you're certain to come back to it on the same visit. Leaving the site with a tack weld as the final fix is not acceptable.

Sources: ISO 20816-1:2016

Bolted mounting: often better than welding

A bolt beats welding more often than people assume. It's reversible, it doesn't heat the metal, it works on aluminium and stainless, and you know the assembly's mass exactly, because you weigh the whole stack: bolt, nut, washers, plate.

Look for existing holes first. Wheels, pulleys, couplings, flywheels, and drums almost always have spare holes in the back disc, the gussets, or the rim. That's the cheapest place to correct at, and the manufacturer has already rated it for load.

Drill a new hole only where the metal isn't carrying a structural load, and ideally with the manufacturer's sign-off. A blade, a blade root, a hub in the shaft-fit zone, a strip running alongside a weld seam - none of these is a place for a new hole.

A stack of washers is convenient because you build up the exact grams you need on a scale. Keep in mind that the stack extends along the axis, away from the correction plane - the rotor cross-section the instrument calculated the mass for. On a short disc that doesn't matter; on a long rotor, you're effectively putting the mass in a different plane. Keep the stack short, and, where you can, symmetrical about the disc.

Thread-locking is mandatory, no exceptions. Vibration will back out any threaded joint unless something stops it: a lock nut, a nylon-insert lock nut, wedge-lock washers, a split pin, or a medium-strength anaerobic thread locker. A single spring washer is weak insurance. Take the tightening torque from a table for the bolt's diameter and strength class, not 'by hand until it feels tight.'

A self-tapping screw, an aluminium-bodied blind rivet, and a cable tie are not fasteners for a rotating mass. Not under any circumstances.

• After the verification run, check the tightness again. The first few minutes of operation take up any slack left in an under-torqued joint.
• Check it again after the first working shift. If the joint has moved, you'll find out now, not on the next site visit.
• Paint a witness mark across the nut and bolt. If the mark no longer lines up, the joint has turned.
• Record the assembly's mass, radius, and position number in the report. Six months from now, a different person may be the one standing at this machine.

Magnets, glue, rivets, clamps: what's good for what, and for how long

The rule is simple. Anything that can be removed by hand, without a tool, cannot be left on the rotor after you leave.

Removing metal instead of adding it: drilling and milling

Sometimes there's nowhere to add mass. The rim is taken, the flow won't tolerate a protrusion, the housing sits two millimetres away, or the food-production plant won't allow a weld. In that case, you remove metal at a point 180° opposite where the weight would have gone. Same vector, same physics.

The volume calculation is done on the spot. The mass you remove equals the material's density multiplied by the volume removed. For a cylindrical hole, the volume is π·d²/4·h. Use 7.8 g/cm³ for steel, 7.2 for cast iron, 2.7 for aluminium.

An example. A 12 mm drill at a 20 mm depth in steel removes about 2.3 cm³, roughly 18 g. To remove 60 g, you'd need either four such holes, one larger-diameter hole, or a milled pocket. Work this out before you pick up the drill: it often turns out the part simply doesn't have that much depth to give.

Space multiple holes symmetrically around the calculated angle. Four holes placed symmetrically around the correction angle give you the same vector. Four holes drilled in a row to one side of it shift the centre of the removed mass off-axis, and you end up with a different angle than the one you calculated.

A real hole isn't a cylinder - it has a conical tip from the drill's geometry. At depths under three diameters, the discrepancy from the formula becomes noticeable. It's simpler to remove material in stages and check the result by measurement rather than by arithmetic alone.

The key thing about drilling: you can't weld the metal back on. So remove less than the full amount at first. Take out 60-70% of the calculated volume, run a verification run, then remove the rest. The instrument's software has trim balancing for exactly this step - a fine correction based on the machine's already-measured response, with no new trial runs needed.

There are places where metal is never removed, whether by drilling or milling:

• Thin walls. Monitor the remaining wall thickness - don't judge it 'by eye, from the chips.'
• Weld seams, and the strip of metal alongside them.
• Blades, blade roots, gussets, any load-bearing ribs.
• The hub in the shaft-fit zone, and the shaft itself.
• Areas with cracks, corrosion, or signs of prior repair or weld build-up.
• Surfaces responsible for sealing, corrosion resistance, or hygiene.
• Anything the manufacturer's documentation for the machine specifically prohibits touching.

In the Balanset-1A software, you enter the drill diameter and radius, and it gives you the depth and number of holes instead of an abstract number of grams. That takes the arithmetic out of gloved, oily hands.

Sources: Balanset-1A operation manual

Choosing the location: strength, flow, clearances, corrosion, hygiene

The instrument gives you the angle and radius. It doesn't know there's a housing there, or thin sheet metal, or a food-safety regulation. You choose the location, weighing seven considerations at once.

In a food-production plant, the logic flips. What's ideal on a crusher - a bolt, nut, and washer stack - is a poor choice in a plant washed down every shift: threads and gaps don't clean out. A ground-smooth continuous weld, or metal removal, fits better there. Confirm the requirements before the visit, not while standing over a weight that's already mounted.

Radius: why it pays to mount further from the axis

The instrument works with the product of mass and radius. 60 g at a 200 mm radius and 30 g at a 400 mm radius cancel out exactly the same imbalance. The direct conclusion: the further from the axis you mount the weight, the less metal you have to hang.

Less metal means a smaller plate, a shorter weld, an easier spot to find, and less of a protrusion into the flow. On a wheel where everything's already taken, that's sometimes the only thing that makes fitting the correction possible at all.

An honest caveat that usually goes unsaid. For the same mass-times-radius product, the centrifugal force from the correction mass is identical at any radius. The mounting is loaded the same either way, and 'further out means safer' is a myth. The real advantage is in the weight's physical size, in ease of fitting, and in how much easier it is to find room for a small plate.

Further from the axis isn't always possible. The rim is taken by factory slide blocks or previous weights. The wheel's periphery sits in an abrasive flow, and a weight there would wear away within months. The housing is right there. The blade section is off-limits per the manufacturer. Or, simply: the periphery is thin sheet, and only the hub has any real strength.

In that case you mount closer to the axis and recalculate the mass in inverse proportion to the radius. 40 g at 300 mm becomes 120 g at 100 mm. Three times the metal, three times the pad size, and the question of the base's strength comes up all over again.

Measure the radius from the rotation axis to the weight's centre of mass, not to the plate's edge and not to the bolt head. And always enter the actual radius into the instrument: otherwise the residual unbalance in g·mm and the grade-G check against ISO 21940-11 become fiction, even if the vibration has dropped.

Sources: ISO 21940-11:2016

When there's nothing to mount at the calculated point

A common situation: the software calls for a mass at a 63° angle in plane 1, and there's either nothing there, or there's a coupling, a heat shield, and no access whatsoever.

The first option is fixed positions. The software splits the calculated mass between two adjacent holes or blades and gives you a position number instead of an angle. The combined mass ends up slightly higher than the calculated value, because two vectors at an angle to each other don't add up arithmetically. How much extra depends only on the spacing between positions, and you can estimate it before the visit. We cover this in more detail in a separate article on angle referencing and fixed positions.

The second is recalculating for a different plane. The software can recalculate weights for other planes and radii: you tell it where you can actually mount the mass, and it gives you the equivalent. This isn't free. Shifting along the axis introduces a moment component - a force couple that rocks the rotor - and on a long rotor, the result will be worse than a correction placed in the originally calculated plane. On a short disc, you won't notice the difference.

The third option is a combination. Part of the correction goes on as mass wherever you can fit it, and the rest is made up by removing metal at the opposite point.

And the honest option. If there's genuinely nowhere to mount anything and the vibration is high, the correction gets done with the machine disassembled, in the shop, where both planes and both sides of the wheel are accessible. We cover the choice between on-site work and shop work separately.

Who does the welding, and how you sign off the job

Field balancing is really two different jobs in one visit. The measuring, calculating, and checking is done by the instrument in the engineer's hands. Fitting the correction is done by hands holding a welder, a drill, or a wrench. It's worth not blurring the two together in your head - it's usually the second part that lets people down.

The usual arrangement: your own fitters and welders do the mechanical work, we tell them where, how much, and how, and confirm the result with a measurement. That's faster and cheaper, because your people know the machine, the permit system, and the site's internal rules.

If you don't have that kind of staff, or they're tied up, we take on that work ourselves. At AXILINE this is a separate service added to the visit, and it needs to be agreed in advance: scope, weight material, hot work, access, coating restoration.

These are decisions to make before the visit, not on site. How weights get mounted on this machine. Whether there are spare factory holes. Whether welding is allowed, and whether a permit is needed. Who restores the coating. Whether there are food-hygiene requirements. We cover the full pre-visit preparation checklist separately.

If you're balancing it yourself, the Balanset-1A covers the entire calculation side: fixed positions with numbers instead of angles, drilling calculations from your drill diameter, recalculating weights for other planes and radii, trim balancing, and a report with actual masses. The mounting itself is still your job. But you'll know exactly how much metal, and where, instead of chasing a number by guesswork.

• The actual mass of every weight has been weighed and recorded in the report, along with its radius and position number.
• The mounting is permanent: a full weld, a bolt with thread-locking, or rivets. No tack welds, no magnets, no glue.
• Clearances have been checked by slowly turning the rotor through a full revolution, not just eyeballed through a hatch.
• The coating has been restored, chips and spatter cleaned up, and the housing closed and secured.
• A verification run has been made after everything cooled, confirming the result is within tolerance.
• Plane, zero-reference, and position markings are on the rotor, so the next check takes minutes, not hours.

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

Frequently asked questions

Related content