Cooling Tower Fan Balancing: Blades, Hub, Gearbox and Drive Shaft

A cooling tower cell trips out on its vibration switch, usually on the hottest day of the year, when the load is highest. The fan ring is humming, the deck is shaking, and the gearbox has started leaking oil. The fan wheel, several metres across, turns slowly, but its mass sits out on a large radius, and the imbalance here comes from things an ordinary fan never sees: water trapped inside a fiberglass blade, buildup on one side of the wheel, a lost factory balance weight, a pitch angle that has drifted. We climb up to the fan deck — the platform on top of the cooling tower — work out how much of the vibration comes from mass, how much from aerodynamics, and how much from the drive, and balance the wheel right on its hub. We are based in Vila Nova de Gaia, near Porto, and travel throughout Portugal.

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

Symptoms: how an unbalanced cooling tower fan behaves

The fan axis is vertical, so the centrifugal force from the imbalance rotates in a horizontal plane. It travels into the gearbox bearings, into the support frame, into the fan ring, and on into the tower structure. The structure is light and flexible, so everything vibrates together, and the complaint usually comes not from a mechanic but from the dispatcher whose cell has just tripped.

A second feature: cooling towers almost always carry a vibration switch or vibration sensor wired into an emergency shutdown. Here, imbalance is noticed not by the hum but by the trips. Until the setpoint is exceeded, the machine runs on quite happily with a growing imbalance for months, chewing through the gearbox bearings the whole time.

• The vibration switch trips more and more often: first once a month, then on every start-up or every switch to high speed
• Vibration increased after a blade replacement, blade repositioning, hub repair, or wheel washing
• The fan ring hums, the deck shakes, and the vibration can be heard in neighboring cells and in the space below the tower
• Oil is leaking from the gearbox, fits have worked loose, and the fasteners on the support frame or the beams under the gearbox have loosened
• The flexible couplings on the drive shaft run hot, the disc packs crack, and the coupling-half bolts shear
• A factory balance weight has shifted or gone missing from the hub or a blade root fitting, with rust streaks left underneath
• Blade tips are tracking at different heights, and contact marks have appeared on the inside of the fan ring
• A fiberglass blade sounds dull when tapped and weighs noticeably more than its neighbors: water has collected inside it
• The machine runs seasonally, and after the winter shutdown the vibration is nothing like it was in the autumn

The cheapest diagnostic trick on a cooling tower: compare vibration across all cells in the same operating mode, at the same points on the gearbox housing. The cells are identical, so whichever one stands out from the rest is your candidate for work. The absolute limits depend on the applicable part and edition of ISO 20816 and on your vibration switch's setpoint, and the spread between neighboring cells tells you more than any table.

Sources: ISO 13373-5:2020

How the drive is built, and where the imbalance comes from

A typical induced-draft cooling tower layout looks like this. The motor sits outside the airstream, on the deck or on an outboard frame, and turns at 1000 or 1500 rpm. From it, through flexible couplings, a long drive shaft — often a composite tube — runs to a right-angle gearbox under the fan hub. The gearbox reduces the speed several times over, and the wheel turns slowly, typically in the range of roughly 90 to 250 rpm. The blades are large and light, made of hollow fiberglass or pressed aluminum; they are held in the hub by root clamps with an adjustable pitch angle.

The main consequence for measurement follows from this. One machine carries two different rotational frequencies at once: the fan's and the shaft-and-motor's. Until you have separated them in the spectrum (vibration broken down by frequency), it is too early to talk about wheel imbalance: the vibration switch reacts the same way to either one.

Why angles, droop and masses come first, and balancing comes after

This is the main way a cooling tower differs from any other fan, and the order here is not negotiable. A spread in pitch angle creates an aerodynamic force at the running frequency. A spread in blade-set mass creates a mass force. A single measurement at operating speed cannot tell them apart, yet they are fixed in different ways. Fit the weights first and set the angles afterward, and the correction has to be recalculated from zero.

The factory matches a blade set not just by mass but by static moment — mass multiplied by the distance to the centre of gravity. That is exactly why replacing a single blade with a new one from a different batch almost always ends in vibration, even when the scales show similar figures.

• We measure the pitch angle of every blade with a digital protractor at the same radius from the root, from the hub plane. The manufacturer's tolerance on cooling towers is tighter than on an ordinary axial fan: a spread of half a degree already noticeably changes blade thrust.
• We check tip droop. We slowly turn the wheel by hand and measure the gap from each blade tip to a fixed mark on the fan ring. We keep the spread within the manufacturer's tolerance, otherwise the blades run at different heights and create a couple-imbalance component — a pair of forces that rocks the wheel.
• We weigh the blades if there is any suspicion of water or a replacement. A blade with water inside sounds dull when tapped and turns out to be kilograms heavier than its neighbors. A blade like that gets dried and resealed, not compensated with a weight.
• We change the operating mode. Mass imbalance does not depend on airflow: switch to low speed or close the louvres, and the amplitude changes predictably, with the square of the speed. The aerodynamic part behaves differently when the mode changes.
• We check phase repeatability. Phase is the angle showing where the heavy spot sits on the wheel; the instrument measures it from a mark on the shaft. A fixed mass gives a stable phase from one start to the next, while a loose root clamp or water sloshing inside a blade makes the phase drift by tens of degrees.

In practice this looks unglamorous and works reliably: torque the root clamps to the manufacturer's instructions, set the angles equal, bring the droop into line, and only then switch on the instrument. Often half the vibration is already gone at this step, leaving balancing a neat remainder to deal with.

What we check before the first weight

Everything involving hands-on work is done on a stopped, locked-out and mechanically restrained machine. The general method for separating imbalance from other causes is covered in our article on diagnosing the cause of vibration. What follows here is specific to cooling towers.

• Pitch angles, tip droop, root-clamp torque, and the condition of the mating surfaces in the hub
• Blade integrity: gelcoat cracks and fiberglass delamination at the root, fretting marks under the clamps, aluminum corrosion, and the condition of the drain holes
• Factory balance weights: whether they are all present per the logbook, properly torqued, and whether the fasteners under them have corroded
• Radial clearance from the blade tips to the fan ring all the way around, and any contact marks on the ring
• Gearbox: oil level and condition, water or emulsion in the oil, output-shaft play, and the spectrum for bearings and gear mesh
• Drive shaft: runout, condition of the flexible couplings and disc packs, coupling-half bolt torque, and whether a guard is fitted
• Motor-to-gearbox alignment, and the condition of the motor base frame and its fastening to the deck
• Support structure: cracks in the beams under the gearbox, loosened anchors, the condition of the fan ring and its mounting, and a coast-down check for resonance
• Operating mode: speed, the step of a two-speed motor or the VFD setpoint, louvre position, and whether the neighboring cell is running

Then comes the measurement. We mount two accelerometers (vibration sensors) on the gearbox housing horizontally, in two mutually perpendicular directions — along the drive shaft axis and across it — on rigid pads near the bearing housings. We stick the reflective phase mark on the low-speed shaft or the hub, and aim the laser phase sensor up from below. One quirk specific to cooling towers: at 150 rpm, the fan's running frequency is 2.5 Hz, below the standard 10–1000 Hz band that the ISO 20816 zones are built on. So we record the measurement band explicitly in the report and do not plug these figures into a table meant for a different band.

Sources: ISO 13373-3:2015

Height, wind, and locking out the fan

Working on a cooling tower means working at height in a wet, slippery environment. We bring our own fall-protection gear, but the organisational side is on your end, and it needs to be agreed before the visit, not on the day of the work.

• Locking the fan is mandatory, and it is not the same thing as switching off the motor. Draft through the tower and wind can spin the wheel in free rotation, so before anyone climbs up to the hub, the shaft is mechanically restrained with the factory locking pin or brake.
• Start-up lockout with tags, held by a single responsible person, accounting for the automation: the cell can start on its own based on return-water temperature.
• Neighboring cells. We will not work next to a running fan sharing the same deck plane: it has its own airflow, its own noise, and its own spray.
• Decking, ladders and guardrails. Wet fiberglass is slippery, and an open hatch in the fan ring is a fall hazard. We keep tools on lanyards, because a dropped wrench falls straight into the basin.
• Wind window. A fan with blades several metres long is effectively a sail. In strong wind the load on the wheel varies, the amplitude and phase drift from one run to the next, and the influence coefficient — the machine's response to the trial weight, the basis of the whole calculation — comes out unreliable. In thunderstorms and gale-force wind we do not work at all.
• Temperature and icing. There is no point balancing a wheel with ice on it: that is temporary mass that will disappear with the thaw.

The requirements for start-up, lockout, weight fastening, and the throw-off zone are covered in a separate article on safety during on-site balancing. On a cooling tower, the only things added are height and water — but those are exactly what determine how long each stop-and-start cycle takes, so plan your window with margin.

How on-site balancing proceeds

1. Inquiry

   Reviewing the machine from photos

   You send us the cooling tower type, fan diameter, number and material of the blades, wheel and motor speeds, the drive layout, whether it has a VFD or two speeds, and photos of the hub, root fittings, gearbox and deck. We reply with where the correction plane will be (the rotor cross-section where the weights will go), what we will fasten them with, and what access and locking you need to arrange.

2. Mechanics

   Angles, droop, clamps, drive

   We work through the checklist on the locked-out machine. We set the pitch angles and bring the tip droop into line, torque the root clamps, and check the couplings and drive alignment. If the cause turns up here, the next steps may not even be needed.

3. Run 0

   Baseline measurement and frequency separation

   Sensors on the gearbox housing, mark on the low-speed shaft. We record speed, overall vibration (the total level across all frequencies), the running-speed component with phase, and the spectrum. We separately check the components at the drive shaft's rotational frequency: if the shaft or a misalignment is providing most of the energy, it is too early to balance the wheel.

4. Marking

   Blades as fixed positions

   We number the blades in the direction of rotation and enter them into the software as fixed positions. From then on the instrument works in blade number and mass, not angle and direction of reckoning, splitting the mass between two neighboring positions where needed. On a four-blade wheel, this removes the single most common weight-placement mistake.

5. Trial weight

   A small mass at a large radius

   The wheel is light and the radius is large, so the trial weight is usually small. We fit it at the factory mounting point near the blade root or on a hub arm, and fasten it just as securely as we will fasten the permanent one. We count a run as valid when the amplitude changes by 20–30 percent or the phase by 20–30 degrees.

6. Correction

   Weights by position, blade reshuffling if needed

   The software outputs a mass and a position number. If the calculated mass exceeds the manufacturer's limit for a single point, we spread it across several root fittings, or first rearrange the blade set by moment and make up the remainder with a small weight. Any remainder after the check run is removed with trim balancing — a fine adjustment using the coefficients already calculated, with no new trial runs.

7. Check

   After measurement, both speeds, report

   Check run in the same mode and at the same points. If the machine is two-speed or on a VFD, we take readings at both operating speeds: correction on a rigid wheel holds at either one, but a structural resonance can sit at only one of them. The report gets the before-and-after figures, blade angles, and the masses and positions of the weights.

Sources: Balanset-1A operation manual

Correction planes and what to fasten weights with on this rotor

The geometry here is simple: hub width is much smaller than wheel diameter, the length-to-diameter ratio is clearly small, and the imbalance is practically pure static imbalance — the wheel is simply heavier on one side. So there is one correction plane, at the hub: the hub arms, the blade root fittings, the factory weight mounting points. One trial run, one weight installation.

A second plane appears not on the wheel but in the drive. The drive shaft is a separate rotor, balanced in two planes at its coupling halves and at its own rotational frequency. So on a cooling tower it is more accurate to talk not about two planes of one rotor, but about two separate jobs, done one after the other with different sensors. The rule for choosing the number of planes is covered in our article on single-plane versus two-plane balancing.

Fastening on this rotor is tightly constrained by the material. Welding is never used on a fiberglass blade, and on an aluminum blade or a cast hub it distorts the metal and changes the structure at the weld. Drilling the blade is out too: you open up the cavity, and water will get in through the hole. That leaves mechanical fastening, and the first thing we look for is whatever the manufacturer already built in.

• Fasteners are stainless or hot-dip galvanized only. Plain carbon steel corrodes within a season inside a cooling tower, the weight loses mass and washes away, and the imbalance comes back in one jump.
• We thread-lock and lock-nut everything, because vibration plus constant moisture is the perfect recipe for self-loosening.
• Any hole opened in the composite gets sealed. Water inside the blade will undo the entire job within a couple of months.
• A magnetic weight is only acceptable as a trial weight on a steel hub, and it is removed before the machine is handed back.
• We log every weight fitted — its mass, radius and position number — in the fan's logbook, so that at the next repair it is not mistaken for a stray part and thrown away.

Correction-weight fastening methods in general are covered in a separate article on weight fastening. For a cooling tower, only the mechanical-fastening part of it applies; everything about welding on plates falls away along with the blade material.

Sources: ISO 21940-11:2016

When on-site work will not work, or will not help

An honest list saves you a wasted visit. Here are the cases where we stop and suggest something else first.

• A blade has taken on water or delaminated. Drying, repair or replacement comes first, balancing after. There is no point compensating a variable mass with a weight.
• A crack at the blade root or in the hub. Balancing would just mask it, and a blade that comes off at operating speed destroys the fan ring and can leave the tower entirely.
• Play in a root clamp or a worn blade fit. The phase is unstable in this case, the influence coefficient is unreliable, and repair comes first.
• The pitch angles cannot be adjusted and the blades are deformed. Aerodynamic unevenness cannot be evened out with mass; the wheel gets repaired or replaced.
• The drive is producing the vibration: motor-to-gearbox misalignment, worn couplings, or bearing or gear-mesh defects in the gearbox. We will run the diagnostics and tell you exactly what needs fixing, but weights on the blades will not change any of it.
• Resonance in the fan ring, the beams under the gearbox, or the tower frame. On coast-down you see a narrow peak near the operating speed, with a phase reversal. This is fixed with structural stiffening or a speed change, not with weights.
• Speed will not hold: a VFD is controlling the fan by water temperature and constantly changing speed. We only balance if process control can hold a fixed frequency for the duration of the work.
• There is no access to the hub and root fittings, there is no hatch in the fan ring, and there is no provision for locking the fan. In that case we discuss partial disassembly or removing the wheel.
• Weather. Strong wind, thunderstorms, icing. This is not a whim: with a shifting wind load, the measurement result is unreliable.

If vibration does not drop after a correctly done balancing job, or comes back quickly, the cause is usually not the weights. Seven such scenarios are covered in our article on why balancing doesn't help, and the choice between working on-site and removing the rotor to a shop is covered in our article on on-site balancing versus shop balancing.

What you get, and how to book a visit

By the end you get not just a quiet cell but figures you can attach to your acceptance record and compare against a season later. The report includes: initial and residual vibration at each measurement point, with direction and measurement band noted; the fan's running-speed component and the drive shaft's rotational-frequency components, listed separately; the measured blade pitch angles and tip-droop spread; the mass, radius and position number of every weight fitted; the blade arrangement, if we matched blades by moment; before-and-after spectra; and a mechanical-condition summary of the drive. We give a zone assessment under the applicable part of ISO 20816 with a direct reservation about edition and band, because the fan's running frequency sits below the standard range.

Your machine's influence coefficients are kept on file. The next balancing job on the same fan — after a blade replacement or a seasonal overhaul, say — needs no trial runs and takes noticeably less time. The work is carried out by the engineers who design and manufacture the Balanset instruments, and who do the on-site balancing themselves. Vibration diagnostics with a report cost EUR 300 per unit, balancing adds from EUR 250, the minimum invoice for a visit is EUR 500, and the calculator on the website gives you an exact figure for your machine. On a cooling tower with several cells, count them as one visit: the sensors and method are already set up, and the second cell comes out noticeably cheaper than a separate call-out.

• Cooling tower type and number of cells, fan diameter, number and material of the blades
• Wheel and motor speeds, drive layout, drive shaft length, coupling type
• Speed control: two-speed, VFD, or fixed speed, and who can hold a fixed mode for the work
• Whether there are factory mounting points for balance weights, and whether the fan's logbook has survived
• How access to the fan deck and hub is arranged, what locks the fan, and who is responsible for start-up lockout
• Photos of the hub, root fittings, gearbox, couplings and support beams, plus a view of the wheel from inside the fan ring
• The vibration switch setpoint and any trip records you keep
• Your window for taking the cell offline, and acceptable weather for the work

We are based in Vila Nova de Gaia, near Porto, and travel throughout Portugal. Send us the answers to the list above, and we will tell you straight: whether your fan can be balanced on-site, whether the angles need setting and a blade needs drying first, or whether the problem is in the drive after all and needs shaft alignment and a gearbox repair. If your data shows that weights will not help, we will tell you that before the visit, not after.

Sources: ISO 20816-1:2016 · Balanset-1A manufacturer specification

Frequently asked questions

Related content