# On-site balancing of centrifugal fans: single-inlet and double-inlet

> Your centrifugal fan hums at the bearing housing, and you can only stop it between shifts. A single-inlet wheel hangs overhung, beyond both bearing housings, while on a double-inlet wheel the shaft runs through the hub. That's what decides how many correction planes you'll need to mark out and which inlet pocket you'll have to reach them through. We come to your site and balance the wheel in its own bearing housings, at operating speed and at your damper's normal setting. We're based in Vila Nova de Gaia, near Porto, and travel across Portugal.

**In short:** Yes, we balance centrifugal fans right on site, both single-inlet and double-inlet wheels. The rotor turns in its own bearing housings, and there's no need to remove the wheel or take the volute apart. Three conditions apply. The machine has to hold a stable speed at one damper setting. The correction plane — the spot on the wheel where balancing weights go — has to be reachable: an inspection hatch in the volute, a removable inlet cone, or an open inlet pocket, and on a double-inlet wheel both sides have to open. And most of the vibration has to come from the 1x component (vibration exactly at rotating speed — the signature of unbalance), not from bearings, a worn pulley, shaft misalignment, or pedestal resonance. We work out access and regime from your photos at the request stage, we measure the 1x share ourselves in the first half hour on site, and we tell you plainly if weights won't help here.

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

## Symptoms: what unbalance looks like on a centrifugal fan specifically

A centrifugal fan almost never comes to us with a built-in unbalance. Dust, abrasive wear, a blade repair, washing the wheel, or a loose hub fit on the shaft is what threw the wheel out of balance. The symptoms on this machine are recognisable, and they tell us a lot before we even visit.

Pay attention to which bearing housing shakes harder. On a single-inlet wheel, both bearings sit close together in one block, and vibration is usually higher at the housing nearer the wheel. On a double-inlet wheel, the housings sit apart, at the edges of the casing. If the amplitudes there are comparable and the 1x phases are close to opposite (phase being the angular position of the "heavy spot" on the rotor), you're dealing with couple unbalance: two unbalances at the ends of the rotor pulling in opposite directions. No single weight will remove it, however much you reposition it.

- [x] Noise and shaking increased after washing the wheel, replacing a blade, repairing the volute, or re-tensioning the belt
- [x] You can feel a knock once per revolution by hand on the bearing housing, while the inlet pocket or the ductwork may be humming
- [x] Vibration changes noticeably when the damper is moved or the VFD setpoint is changed
- [x] The amplitude jumped without any change in regime: looks like a chunk of buildup broke off, or the hub has shifted on the shaft
- [x] A scraping or knocking sound against the inlet cone is audible during run-up or coast-down
- [x] The belt runs hot and keeps slipping off, the pulley grooves are worn, and the pulley wobbles when turned slowly by hand
- [x] The bolts on the bearing pedestal and the volute feet keep working loose, and the frame's gusset plates are cracking
- [x] Bearings have started running hot, the grease darkens faster than usual, and the seals have started leaking
- [x] Balancing has already been done once, but the hum came back within a month

> A useful benchmark before you call us: overall vibration at the bearing housings, in mm/s RMS (root mean square, the standard measure of vibration level), has risen to about one and a half times the level at which the machine used to run quietly. The absolute limits depend on the applicable part and edition of ISO 20816, on the drive's power, and on whether the machine sits on a rigid foundation or on vibration isolators. A trend built from your own measurements tells you more than any table found online.

Sources: [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) · [ISO 13373-5:2020](https://www.iso.org/standard/62202.html)

## Wheel construction, and where its unbalance comes from

From here, everything comes down to two features: how many inlet sides the wheel has, and which way the blades curve. The first sets the bearing layout and the number of correction planes. The second decides what we use to fit weights, and where.

What unbalance actually is, and how static differs from couple unbalance, we cover separately, in more detail in our articles on rotor unbalance and on overhung versus between-bearings rotors. Here, we'll stick to what's specific to a centrifugal wheel.

### Single inlet: an overhung wheel

The shaft comes out of the bearing block, and the wheel hangs beyond both housings. The bearings sit close together, often 200–400 mm apart, between the wheel and the pulley. The lever arm from the wheel to the nearer housing is often longer than the distance between the housings themselves, so even a small residual unbalance produces a noticeable force on the bearings. A weight at the shroud affects both housings at once, sometimes the far one more than the near one.

### Double inlet: a wheel between the bearings

The shaft runs through the hub, the wheel is built from two halves back to back, and the bearings sit on their own pedestals at the edges of the casing. This is a classic between-bearings rotor with two natural correction planes, one per half. Access is through two inlet pockets, and both have to be opened. On the plus side, the two planes' effects separate cleanly here, and the calculation converges quickly.

### Backward-curved blades

Eight to sixteen relatively thick blades, often of an aerofoil profile. A hollow aerofoil blade has a hidden trap: water, condensate, or dust collects inside through an erosion pinhole, adding mass to one blade that's invisible from outside. A wheel like this gets dried out and its blades checked for pinholes first, otherwise the unbalance will drift from run to run.

### Forward-curved blades

A squirrel-cage wheel of the sirocco type: thirty or more short blades made of thin sheet metal, at a low peripheral speed. The pockets between the blades are deep and trap dust and fibre well, so buildup accumulates faster here than on a backward-curved wheel. The blade sheet is thin, so a weight can't be welded to it. We fit mass onto the back plate, a side ring, or the hub.

### Hub and tapered bushing

The wheel's fit on the shaft is the most common source of a drifting unbalance. A tapered bushing with loose screws lets the hub rotate slightly and creep down the taper, the key connection wears loose, and the weld between hub and back plate cracks. The sign shows up on the instrument: between two otherwise identical runs, the 1x phase shifts by tens of degrees and the amplitude changes in steps. That calls for retightening to the manufacturer's torque, or a repair — not balancing.

### Inlet pocket, inlet cone, and guide vanes

An uneven clearance between the inlet cone and the wheel's front ring causes rubbing once thermal expansion sets in, and rubbing reads exactly at 1x — it mimics unbalance well. Scatter in the inlet guide-vane angles and a skewed feed into the pocket produce an uneven flow: the blade-pass frequency rises, vibration becomes unstable, and none of it is fixed by weights at all.

### Belt drive and pulley

The wheel's speed isn't the same as the motor's, so we mount the reflective mark and aim the laser phase sensor at the fan shaft, not the motor. An eccentric or worn pulley, a warped pulley bushing, or an over-tensioned belt all produce vibration exactly at rotating speed. Skip that check, and weights on the wheel will end up compensating for pulley runout — a result that falls apart the moment the belt is next replaced.

> One frequency worth knowing for your wheel: the blade-pass frequency, equal to the number of blades multiplied by rotating speed. It's always present and normal on its own. A rise in it points to the inlet flow and uneven blade wear, not to a mass unbalance. Belt-drive vibration and pulley runout are covered in a separate article of ours.

Sources: [ISO 13373-5:2020](https://www.iso.org/standard/62202.html)

## Sensors: where we mount them, and why we never change direction

We mount vibration sensors on the wheel shaft's bearing housings, not on the volute, not on a bearing cap, and not on the frame. On an overhung single-inlet machine, both bearings sit in one cast or welded block, and we have to find mounting spots so each sensor sits directly over its own bearing, not in the middle of the block. We call the bearing nearer the wheel housing 1, and the farther one housing 2. On a double-inlet wheel, the housings sit apart at the edges of the casing, each on its own pedestal, and we fit one sensor per housing.

The mounting is rigid: a stud into a prepared pad, or failing that, a magnet on a flat spot cleaned down to bare metal. A threaded pad is better, because a magnet on paint or a rough cast surface cuts off the top of the frequency range and adds its own mounting resonance. A sensor stuck to the volute casing or a bolt reads the vibration of that part — and that's a completely different number.

The direction is radial, usually horizontal: on fans mounted on pedestals and frames, vertical stiffness is higher, so the peak usually shows up in the horizontal direction. That direction has to stay the same on every run. The reason is simple. The influence coefficient — the machine's response to a trial weight, which the instrument uses to calculate the correction — is tied to a specific point-and-direction pair: both the 1x amplitude and phase belong to the signal picked up on exactly that axis. Move the sensor 90 degrees, or to a different spot, and the numbers change, but the calculation keeps treating them as the same system. The weight ends up in the wrong place. That's why we mark the mounting spots with a pen, photograph them, and check the fit before every single run.

We measure the axial direction once, separately, before balancing. High axial vibration on an overhung wheel more often points to shaft misalignment at a coupled connection, or to a bearing cocked in its pedestal, rather than to unbalance.

> A subtlety for a double-inlet wheel: we measure both housings simultaneously, on two channels, and note which one is on the drive side. The 1x phase difference between the housings is exactly the sign we use to decide whether one plane is enough or two are needed.

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

## What we check before fitting the first weight

The walk-round takes fifteen to thirty minutes and decides whether it's even worth starting the procedure. Anything touched by hand is done on a stopped, locked-out machine.

- [x] Fasteners: bolts on the bearing pedestals, the volute feet, the frame, and the vibration isolators. Looseness produces a whole series of harmonics in the spectrum (2x, 3x, and higher) and an unstable phase — there's nothing to balance in this state
- [x] The wheel's fit: tapered-bushing screws torqued to the manufacturer's figure, play at the key, and the condition of the hub and its weld to the back plate. We paint a witness mark across the shaft-hub joint right away, so we can see any rotation after the trial run
- [x] Blades and discs: cracks at the blade root and in the welds to the back plate and shroud, bent leading edges, erosion thinning, pinholes in hollow blades, torn-off wear plates
- [x] Cleanliness of the wheel: both sides of the blades, the pockets between them, and the inner face of the shroud. A layer just two or three millimetres thick on one sector already produces a noticeable unbalance and is almost invisible from outside
- [x] The clearance between the inlet cone and the wheel's front ring, checked at four to six points around the circumference, plus signs of rubbing and the condition of the inlet guide vanes
- [x] The belt drive: pulley runout checked with a dial indicator, groove wear, alignment, and tension per the manufacturer's instructions
- [x] Operating regime: speed, damper position, VFD setpoint. We record it and keep it the same on every run
- [x] Supports and foundation: cracks in the pedestals and gusset plates, the condition of the vibration isolators, the state of the foundation grout, and how much the base under the machine gives
- [x] Safety: how the hatch or inlet pocket opens, how the start is locked out, who holds the key, and where a weight could fly off to if it came loose

> After that comes the measurement. We compare the overall vibration with the 1x component, look at the spectrum, and check phase repeatability between two identical runs. Balancing only lowers 1x. For example, 9 mm/s of overall vibration with only 2 mm/s at 1x isn't a job for weights, and we'll say so right away, not after three trial runs. The method for telling the causes apart is covered in our articles on reading a vibration spectrum and on identifying the cause of vibration.

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

## How the work goes on site

1. **Reviewing the wheel from photos** — You send us the wheel type, the inlet side (single or double), which way the blades curve and how many there are, wheel diameter and width, speed, power and drive type, and ambient temperature. Plus photos of the wheel from both sides, the bearing pedestals, the pulleys, and the inspection hatch. We tell you whether it can be balanced on site, how many correction planes we expect, and what to have open before we arrive.
2. **Mechanics, fit, and clearances** — On site, we start with the fasteners, the tapered-bushing screws, the condition of the blades and welds, the inlet-cone clearance, and the belt drive. This is also where we agree the damper setting we'll measure at, and the lockout procedure.
3. **Initial measurement at your regime** — Two accelerometers on the bearing housings, laser phase sensor on the reflective mark on the fan shaft. We record speed, overall vibration, the 1x amplitude and phase at each housing, the spectrum, and the time-domain signal. From this point on, the damper doesn't move until the work is finished.
4. **Blades as fixed positions** — The blades, the back plate's dedicated holes, and the gusset plates make up a ready-made grid of positions. We number them in the direction of rotation from the mark, measure the actual mounting radius with a tape, and enter it into the software. From then on the instrument gives a position number and a mass instead of an angle, splitting the mass between two neighbouring positions if needed.
5. **Calibrating for your wheel** — We fit the weighed trial weight as securely as we'll fit the permanent one. We count a run as valid when the 1x amplitude has changed by at least 20 or 30 percent, or the phase by at least 20 or 30 degrees. That's how the instrument gets the influence coefficient of your specific system: wheel, pedestals, frame, belt, foundation.
6. **Weights, plane by plane** — The software gives masses and position numbers separately for each plane. We fit them with a bolt through a dedicated hole, by welding on a plate where the manufacturer allows welding and the environment permits it, or by removing metal through drilling in a solid part of the hub or disc. Before starting, we turn the rotor by hand through a full revolution and check the clearances.
7. **Post-measurement and report** — Same regime and same speed as the initial run. If we don't hit the target right away, we do trim balancing — a short fine-tuning with small weights, usually one or two iterations run from the saved coefficients. Then we record the before-and-after figures at each housing and hand over the report.

> How many runs it takes and how long it lasts is covered in a separate article of ours on how long on-site balancing takes. What matters more here: every run happens at the same damper position and the same speed, otherwise the influence coefficients end up belonging to what are effectively different machines.

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

## One plane or two, and where those planes sit on the wheel

A correction plane on a centrifugal wheel isn't an abstract line, it's an actual sheet of metal. On a single-inlet wheel, that's the back plate near the hub and the shroud with its front ring, with the distance between them equal to the blade width. On a double-inlet wheel, the planes sit at each of the side discs, one per half, spaced across the full width of the rotor.

The formal rule looks at the ratio of L to D, where L is the distance between the available correction planes and D is the diameter in the zone where the weight goes. The rule itself is covered in our article on choosing the number of planes. Below is a practical breakdown specifically for centrifugal machines.

| Wheel type | Planes | Why |
| --- | --- | --- |
| Narrow single-inlet wheel, L/D under 0.5 | One | Behaves like a disc, dominated by static unbalance. One trial run is enough, weight goes at the back plate |
| Wide single-inlet wheel, sirocco-type drum wheel | Two | Width is comparable to diameter, otherwise a couple component remains. Planes: the back plate and the front ring at the shroud |
| Any double-inlet wheel | Two | The rotor is between-bearings and wide, and the two halves are unbalanced independently of each other. Each half gets its own plane |
| Access open on only one side | One, with a caveat | We calculate one plane and warn in advance that a noticeable residual may remain at the second housing. Usually a compromise, not a full result |
| After single-plane correction, the second housing is still high | Two | A direct sign of a couple component. We switch to a two-plane calculation, adding one more trial run |
| 1x phases at the two housings are close | One | The unbalance is close to purely static. This is the most common picture on an overhung wheel |
| Speed close to the first critical speed (at which a shaft starts to bend noticeably), a long shaft on a wide double-inlet wheel | Two, and not always enough | The rotor behaves as flexible, and the influence coefficients depend on speed. We check applicability against the current edition of the relevant part of the standard |

> The practical takeaway: anything that isn't a thin disc gets balanced in two planes with a two-channel instrument, as long as both planes are physically accessible. For a double-inlet wheel, that rule always applies, which is why we discuss access through both inlet pockets before the visit, not on site.

Sources: [ISO 21940-11:2016](https://www.iso.org/standard/54074.html) · [ISO 21940-12:2016](https://www.iso.org/standard/50429.html)

## Fitting weights on a centrifugal fan wheel

This is where the calculation ends and the fitting work begins. We choose the mounting method based on the metal thickness at the correction plane, what the manufacturer allows, and the environment the wheel runs in.

- A bolt through a dedicated hole in the back plate or a gusset plate. The most predictable option: washers are stacked to the required mass, the joint is removable, and the weight can be fine-tuned on the next visit. We lock the nut
- A welded-on plate on the back plate or a side ring, where the manufacturer allows welding and the zone conditions permit it. We never weld onto the thin blade sheet of a drum wheel: burn-through and a cracked weld are more dangerous than the unbalance you're removing
- Removing metal by drilling in a solid part of the hub, a boss, or the back plate. The software calculates the diameter and depth for the required mass. A working option wherever neither welding nor a bolt fits
- A rivet into thin sheet, when there are no dedicated holes and weakening the back plate with a through-bolt isn't desirable
- A magnetic weight only as a trial weight. It never stays on for the working shift, because it will come loose

> Three things that most often cost people their result. We use the actual radius, measured from the shaft axis to the weight's centre of mass, not the design figure off a drawing: a ten-percent error in radius is a ten-percent error in mass. A weight must not reduce the inlet-cone clearance, sit in the inlet flow, or catch on the guide vanes, so before starting we turn the rotor by hand through a full revolution. And a separate note on coatings: on a galvanised or painted wheel in an aggressive environment, welding destroys the protective layer, and corrosion comes back along with the unbalance — so there, we work with a bolt or by drilling. The mounting methods are covered in more detail in our article on fitting correction weights.

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

## When on-site balancing won't give a result

An honest conversation costs less than a useless site visit. Here are the cases on centrifugal machines where we either decline to balance or suggest something else first.

- No access to the correction plane. The volute is welded shut, there's no hatch, and the inlet cone can't come off without taking the ductwork apart. Then the wheel gets removed and balanced on a machine
- Only one pocket is open on a double-inlet wheel. We won't be able to mark out both planes, and single-plane correction on a wide between-bearings rotor will leave a couple-unbalance residual
- Play in the shaft fit, a loose tapered bushing, a crack at the blade root, critical thinning, or a pinhole in a hollow blade. That's a repair job, and balancing comes after it
- The machine won't hold its regime: the damper gets moved during the shift, or a VFD tracks speed to load. The influence coefficient comes out unreliable under those conditions
- Overall vibration is several times higher than 1x. Bearings, looseness, or shaft misalignment are driving most of the vibration. We'll run diagnostics and tell you exactly what to fix, but weights on the wheel won't change anything
- Shaft misalignment at a coupled connection. It's corrected by shaft alignment — mass on the wheel doesn't compensate for it
- A worn-out or eccentric pulley, worn grooves, a warped pulley bushing. Balancing masks this rather than fixing it, and the fix falls apart at the next belt change
- Resonance of the bearing pedestal, the volute, the frame, or the ductwork. The fix is to change the structure's stiffness, the mounting, or the speed
- Aerodynamics: flow separation at low flow rate, especially characteristic of forward-curved wheels, a skewed feed into the pocket, or scatter in the guide-vane angles. The symptoms look like unbalance, but the fix is different
- The wheel keeps getting dirty or worn down. We'll balance it, but the result only lasts until the next fouling cycle. Sort out the cleaning regime first

> If none of the causes listed above fits and the vibration still won't come down, take a look at our article on seven cases where balancing doesn't help, and our article on choosing between on-site and shop balancing.

## What you get, and how to book a visit

The result of the work isn't just a quieter hum — it's figures you can attach to an acceptance report, show to management, and compare again in six months.

Vibration diagnostics with a report costs 300 EUR per unit, balancing adds from 250 EUR, and the minimum invoice per visit is 500 EUR. The final figure depends on the number of rotors, the number of correction planes, how far away the site is, and whether diagnostics is needed before balancing. The calculator on the site gives you the exact figure for your machine. If there are several fans at the site, plan for them in one visit: the sensors and the mark are already in place, and the second machine costs substantially less than a separate trip.

- [x] Inlet side: single or double, and how access opens to each correction plane
- [x] Blade curve direction, blade count, wheel diameter and width, and the rotor's approximate mass
- [x] Wheel speed, drive power, and drive type: direct, coupled, or belt
- [x] Temperature and nature of the environment, whether the atmosphere is aggressive or explosive, and any manufacturer restrictions on welding
- [x] Whether there are dedicated weight-mounting holes on the back plate or gusset plates
- [x] Photos of the wheel from both sides, the bearing pedestals, the pulleys, the inlet cone, and the inspection hatch
- [x] Whether the wheel is clean or needs washing, and when it was last cleaned
- [x] Your shutdown window, the damper's normal operating position, and who's responsible for lockout

> Balancing is carried out by the engineers who design and manufacture the Balanset instruments and use them out in the field themselves. We work with the Balanset-1A: two accelerometers on the bearing housings, a laser phase sensor off a reflective mark, a two-channel USB module, and software on a laptop. The instrument calculates the correction in one or two planes using the influence coefficient method, works in a fixed-position mode keyed to the blades, calculates metal removal by drilling, shows 1x, phase, speed, spectrum, and the time-domain signal, assesses the tolerance by G accuracy class, and keeps an archive for reports. You end up with the initial and residual vibration at each housing in mm/s RMS, with a separate line for 1x, the speed and regime at the time of measurement, the mass, radius, and position number of every weight, before-and-after spectra, a note on mechanical condition, and your wheel's saved influence coefficients. You can buy the same instrument and balance it yourself. We're based in Vila Nova de Gaia, near Porto, and travel across all of Portugal.

Sources: [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) · [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/)

## Frequently asked questions

**How is balancing a double-inlet wheel different from a single-inlet one?**

The bearing layout and the number of planes. On a single-inlet wheel, the wheel is overhung, mounted beyond both bearings, and at moderate speed with a narrow wheel, one plane at the back plate is often enough. On a double-inlet wheel, the shaft runs through the hub, the rotor is between-bearings and wide, and the two halves are unbalanced independently, so there are always two planes, one at each side disc. There's one practical consequence: access is needed through both inlet pockets, and it's best to establish that before the visit. A two-plane job takes two trial runs instead of one — one extra stop.

**Do you need to remove the inlet cone or open the inlet pocket?**

Only if there's no other way to reach the correction plane. First we check your photos to see if there's an inspection hatch in the volute and dedicated holes on the back plate: if so, nothing needs removing. If there's no hatch, the inlet cone or the pocket cover usually comes off instead, which is noticeably cheaper than removing the wheel. One thing we always flag: once the cone goes back on, its clearance has to be set evenly around the circumference, otherwise you get rubbing — and rubbing produces vibration exactly at rotating speed, which looks exactly like the unbalance you just removed.

**Our wheel has forward-curved blades, around fifty of them. How do you fit the weights?**

A dense grid of blades actually helps here. We number the positions in the direction of rotation from the reflective mark, enter the count and the actual radius, and the instrument gives a position number with a mass instead of an angle, splitting the mass between two neighbouring positions if needed. An angular error is practically ruled out at that step size. What you can't do, though, is weld a weight onto a drum-wheel blade: the sheet is thin, and burn-through or a cracked weld is more dangerous than the original unbalance. We fit mass with a bolt or a rivet onto the back plate, a side ring, or the hub.

**Our fan runs at different damper angles. Which regime do you balance at?**

The one the machine runs in most of the time, and we run every pass at that same setting. The reason: the damper shifts the operating point and the aerodynamic load on the wheel, and on a belt drive it slightly shifts the speed too. An influence coefficient obtained with the damper partly closed effectively belongs to a different machine than one obtained with it open. If you have two genuinely different regimes and vibration behaves differently in each, we measure both, but we calculate the weights for the primary one. Whatever residual remains in the second regime, we note it in the report.

**The balance doesn't hold, and the wheel sits on a tapered bushing. What's wrong?**

Most likely, the mass isn't fixed in place. A loose tapered bushing lets the hub rotate slightly and creep down the taper, and the unbalance shifts on its own after that. This shows up on the instrument within five minutes: we run two or three passes in a row at the same speed and compare the 1x phase. On a properly seated wheel, the phase repeats within a few degrees. If it's drifted by tens of degrees, the screws need retightening to the manufacturer's torque first, or the fit needs repair, and balancing comes after. We paint a witness mark across the shaft-hub joint right away, so any rotation is visible to the eye.

**How much does balancing a centrifugal fan cost?**

Vibration diagnostics with a report costs 300 EUR per unit, balancing adds from 250 EUR, and the minimum invoice per visit is 500 EUR. The final figure depends on the number of rotors at the site, the number of correction planes, distance, and whether diagnostics is needed before balancing. A double-inlet wheel almost always comes out as a two-plane job, while a narrow overhung wheel is often single-plane. The calculator on the site gives you the exact figure for your machine. Planning several machines on one site into one visit is noticeably more economical.