# How to Balance a Fan: The On-Site Procedure > The fan started humming after the impeller was cleaned, and nobody will let you pull the wheel off: ductwork, housing, belts, the shift in full swing. The good news is that a fan is balanced right in its own bearing housings, in two or three runs. The bad news is that half the wheels that get carted off "for balancing" first need washing, tightening, and inspecting. Below is a procedure that cuts out everything that doesn't matter. **In short:** A fan is balanced on site, in its own bearing housings. You mount two vibration sensors on the housings, a laser phase sensor on a reflective marker on the fan shaft, run the machine at its running speed, and compare overall vibration (the total level across all frequencies) with the 1x running-speed component — vibration exactly at the rotation frequency. If 1x accounts for most of it, the unbalance is real: next comes the trial weight, the calculation, mounting the weights on the blades as fixed positions, and a verification run. Before any of that, you need a clean wheel, tight fasteners, intact blades, and a belt drive in good working order. Source: https://axiline.pt/en/articles/how-balance-fan/ Publisher: AXILINE · Vila Nova de Gaia, Portugal · +351 931 831 229 · axilinegeral@gmail.com ## What Makes a Fan Different From Other Rotors Fans and induced-draft fans generate more balancing requests than anything else, for a simple reason: a large wheel, an open flow path, and dirty air. Here, unbalance isn't a built-in defect — it's a result of operation. That said, balancing a fan is convenient. Speeds are low: 700–1500 rpm for most centrifugal fans, occasionally 3000. The rotor is rigid: the running frequency sits well below the first critical speed — the speed at which the rotor hits its own resonance. That's why influence coefficients (the link between "weight added" and "vibration changed") stay stable, and the calculation converges in one trial run per plane. Residual unbalance isn't just noise. It's a rotating force that loads the bearings, the wheel's fit, and the welds right alongside belt tension and rotor weight. An ISO 281 life calculation won't see it unless you put it in yourself. ### Overhung Wheel On most fans, the wheel sits out beyond both bearing housings. A weight near the shroud affects both at once, sometimes the far one more. Hence the typical picture: after correction in one plane, a noticeable residual remains at the second housing. ### Belt or Direct Drive Put the marker and laser sensor on the fan shaft, not the motor's. With a belt drive the wheel turns at a different speed, and with the marker on the motor the instrument will pick out the wrong frequency. ### Buildup and Erosion Dust, ash, fiber, and resin settle on the blades unevenly, and abrasive particles wear down the leading edges. Because of this, unbalance shifts over weeks — on dirty gas streams, over hours. ### Blades as Ready-Made Positions The blades, the factory holes, and the gussets on the back disc form a ready-made grid of fixed positions. In this mode the instrument doesn't tell you "137°" — it tells you "position Z4, 8 g," or splits the mass between two adjacent blades. > One frequency worth identifying right away: blade-pass frequency, BPF = z · f_rotation, where z is the number of blades. It's always present and is normal by itself. A rise in it points to inlet flow problems or uneven blade wear, and weights won't fix it. Sources: [ISO 13373-5:2020](https://www.iso.org/standard/62202.html) · [ISO 281:2007](https://www.iso.org/standard/38102.html) ## What to Check Before You Reach for the Weights The temptation is understandable: show up, mount the sensors, throw on a weight. On fans this ends with the machine humming again a week later. A walk-around takes fifteen minutes and decides whether the procedure is even worth starting. First, the things you check by hand on a stopped and locked-out machine. - [x] Housing and mounting fasteners: bearing housing bolts, casing feet, and frame bolts. Looseness produces a comb of multiple frequencies — 2x, 3x, and higher — and an unstable phase (the vibration angle relative to the shaft marker). There's nothing to balance in this state - [x] Wheel fit on the shaft: play in the key, a worn hub, a loose taper bushing. Any play means repair, not balancing - [x] Blades: cracks at the root and in welds, bent edges, thinning from wear, torn-off wear plates - [x] Wheel cleanliness on both sides of the blades and inside the shroud. A 2–3 mm layer over just one sector already produces noticeable unbalance, and from the outside it's barely visible - [x] Belt and pulleys: runout checked with a dial indicator, groove wear, parallelism, tension per the manufacturer's instructions. An eccentric pulley produces vibration right at 1x and masquerades as unbalance - [x] Support structure: cracks in the frame and gussets, loose grouting, foundation settling, a flexible platform - [x] Guards and safety: how the access hatch opens, what locks out the start circuit, who holds the key. A weight mounted without lockout isn't balancing — it's an injury waiting to happen > The point people forget most often is the process condition. Balance at the speed and damper position the fan actually runs at. A reading taken with the damper throttled and operation with it open will give different vibration: both the operating point and the aerodynamic load change. ## A Dirty, Worn, or Cracked Wheel This is the main reason a fan's balancing job doesn't hold. The calculation assumes the rotor's mass is fixed and constant. An induced-draft fan wheel with an ash buildup doesn't meet that assumption. The rule is simple: wash first, measure second. Clean the whole wheel, not just where the buildup is visible. Partial cleaning often makes things worse: you strip the deposit off one sector and end up with more unbalance than you started with. ### Buildup After cleaning, vibration sometimes goes up instead of down. That's normal: you removed mass that happened to be balancing the wheel. The lesson isn't "don't clean" — it's "balance after cleaning." ### Erosion Abrasive particles wear the edges faster wherever local flow speed is higher. The wheel loses mass unevenly and keeps losing it: you mount a weight, and a month later the unbalance is back. At critical thinning, the blades get repaired or the wheel gets replaced, and balancing becomes the final step. ### Crack A crack at a blade root, in a weld, or in the back plate is a stop-work condition. Balancing would mask it: vibration would drop, and the one external warning sign would disappear. The blade gets repaired per the manufacturer's procedure, or the wheel gets replaced. ### Deposit Breaking Off A sudden jump in 1x during operation with no change in operating conditions is the typical picture of a chunk of buildup breaking off — amplitude and phase change together. If this happens after balancing, saved influence coefficients let you recalculate the correction without new trial runs. > How to tell a stable unbalance from a shifting one in five minutes: run two or three times in a row at the same speed and compare 1x amplitude and phase. On a clean, rigid rotor the phase repeats within a few degrees. If it wanders by tens of degrees, look for a cause: buildup, a loose fit, resonance, thermal shaft bow. ## Sensors, the Marker, and the Baseline Run Mount the two vibration sensors on the fan's bearing housings, as close to the bearing itself as possible and on bare metal, not on the guard. Mounting must be rigid: a stud into a prepared pad, or a magnet on a cleaned, flat spot. A sensor on the guard will show the guard's vibration, and that's a completely different number. The direction is radial, usually horizontal: on frame-mounted fans, horizontal vibration is most often the largest. Keep the same direction across all runs, or the phases won't be comparable. Measure axial separately, once: high axial vibration points more toward misalignment. The laser phase sensor is aimed at a reflective marker. Stick it on a clean, degreased spot on the fan shaft — exactly one strip: a second one will double the reading. Make sure the laser isn't catching a glint off the shaft neck or the keyway, and cross-check the speed against the nameplate or a calculation from pulley diameters. The first run is the baseline. Let the machine settle into its operating condition, then record the speed, overall vibration in mm/s RMS (root mean square), the 1x amplitude and phase at each housing, and the spectrum — the breakdown of vibration by frequency. - [x] The share of 1x in overall vibration. 9 mm/s overall with 2 mm/s on 1x isn't a job for weights: a perfect balancing job would remove two, leaving almost nine - [x] Repeatability of the 1x phase between two runs — within a few degrees - [x] Phases at the two housings: close values point to static unbalance (a heavy spot on one side of the wheel), opposite-phase values point to couple unbalance (two heavy spots at opposite ends, turned against each other) - [x] Spectrum: does 1x dominate, or is there a strong 2x nearby, a comb of harmonics, an elevated blade-pass frequency, subsynchronous components (components below the rotation frequency) - [x] Resonance: on coast-down, amplitude falls off smoothly, with no sharp peak and no roughly 180° phase reversal > Reading spectra and separating overall vibration, 1x, and phase are covered in separate articles. Here, the conclusion is enough: balancing a fan makes sense when the peak at the rotation frequency accounts for most of the vibration. ## One Plane or Two Work out L/D, where L is the wheel width between the possible correction planes and D is the diameter in the zone where the weight is mounted. A full breakdown of this rule is in a separate article on one plane versus two; below is the practical cut for fans. | Wheel type | Planes | Why | | --- | --- | --- | | Narrow centrifugal wheel, L/D under 0.5 | One | Behaves like a disc, static unbalance dominates. One trial run is enough | | Wide wheel, double-inlet, L/D over 0.5 | Two | Otherwise couple unbalance remains: a single mass can't reduce vibration at both housings | | Axial fan, blades on a hub | Usually one | The wheel is short. But check the blade-setting angle: scatter there produces an aerodynamic imbalance, not a mass one | | Wheel between two bearing housings | Two | Planes are accessible from both sides, both housings are loaded comparably | | After correction in one plane, the second housing is still high | Two | A sign of a couple component — switch to a two-plane calculation | > Speed matters too. A fan at 1000–1500 rpm is a rigid rotor, and balancing in its own bearing housings works as intended. A machine running close to its first critical speed behaves like a flexible rotor: two planes don't always handle it, and the approach is different. Check applicability against the current edition of ISO 21940-12. Sources: [ISO 21940-12:2016](https://www.iso.org/standard/50429.html) ## Step by Step: From the Baseline Run to the Report 1. **Baseline vibration** — The machine at running speed and in its working condition. The instrument records speed, 1x amplitude and phase at both housings, overall vibration, and the spectrum. This is where you decide: balance, or look for a different cause. 2. **Positions and radius** — Stop and lock out the machine. Number the blades or holes in the direction of rotation: Z1, Z2 … Zn. Measure the radius where you'll actually mount the weight, and enter the number of positions and the radius into the software. From here the instrument answers with a position number, not an angle. 3. **Trial weight in the first plane** — Weigh the weight, mount it at position Z1 as securely as you'll mount the permanent one, and enter the actual mass and radius. The criterion for a valid trial run: the 1x amplitude changed by at least 20–30%, or the phase by at least 20–30°; the instrument flags validity itself. If the change is small, increase the weight, enter the new mass, and repeat. 4. **Trial weight in the second plane** — Only for two planes. Move the weight as the software instructs and repeat the measurement. In total: one plane means two runs, two planes mean three. 5. **Correction masses and positions** — The software outputs the mass and position of the weight for each plane. In fixed-position mode it splits the mass across two adjacent blades on its own, and for correction by metal removal it calculates the hole diameter and depth. If there's nowhere to mount the weight in the calculated plane, there's a way to recalculate for a different plane. 6. **Fitting the correction weights** — The zero reference is wherever the trial weight sat. You remove the trial weight or leave it in place strictly as the software instructs. The radius is the one you entered: mount the weight closer to the center and it will act weaker than calculated. 7. **Verification run** — The same speed and the same condition as the baseline run. 1x should drop several-fold. If you miss the target, the software will suggest adding small weights to the ones already in place; this is trim balancing, usually one or two iterations. 8. **Recording the result** — Record the baseline and residual vibration by housing and direction, the speed, the condition, and the weight masses and position numbers. The data stays in the instrument's archive. Without the baseline figures, you won't see a trend on the next round. > "Within tolerance" in the software means one thing: residual 1x is below the target value you entered. The machine's overall condition is judged by overall vibration in mm/s RMS (root mean square) in the 10–1000 Hz band, per the applicable part of ISO 20816, while the quality of the rotor's balancing is judged by residual unbalance in g·mm/kg against balance quality grades G (balancing accuracy grades) from ISO 21940-11. The instrument will calculate the grade-G tolerance if you enter the rotor mass, speed, and radius. Check separately which part and edition of the standard applies to your fan: belt-driven machines and small units carry exceptions. A breakdown of the three tolerances is in a separate article. Sources: [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) · [ISO 21940-11:2016](https://www.iso.org/standard/54074.html) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/) ## What to Use to Mount the Weight on a Fan Wheel A weight that comes loose isn't a ruined balancing job — it's a punctured housing. Mount the correction mass as if it's going to stay there for the wheel's entire service life. | Method | Where it fits | What to watch for | | --- | --- | --- | | Bolt and nut through a factory hole | Back-disc holes, gussets, wheel rim. Usually the best option | The nut must be locked. Don't drill a new hole in a blade without the manufacturer's approval: it's a load-bearing part | | Welding on a plate | Steel wheels where welding is allowed and there's access | Heat distortion of a thin blade, wheel warping, ruined coating. Requires a qualified welder. Not for aluminum, and not for explosion-hazard zones | | Magnetic weight | Temporary use only: trial weight, method setup | Comes off under vibration, loses holding force when heated, doesn't hold at all on aluminum or stainless steel | | Metal removal by drilling | Massive hub, thick back disc, when there's nowhere to add mass | The removal point sits 180° from where the weight would have gone. You're responsible for controlling the remaining wall thickness | | Factory weights and washers | Wheels with a manufacturer-supplied balancing rim | A limited set of masses — sometimes you'll need to combine washers | > The radius in the software must match the radius the weight actually ended up on. An error in the radius is a direct error in the mass: 250 mm instead of 350 mm gives about 70% of the expected effect. If you land on a different radius, enter the actual one and recalculate. ## Common Mistakes on Fans ### Marker on the motor shaft With a belt drive, the instrument reads the motor's speed and picks out the 1x of the wrong machine. ### Sensor on the guard You're measuring the sheet metal's vibration: the numbers are big and have nothing to do with unbalance. The sensor belongs on the bearing housing, on bare metal. ### Balancing a dirty wheel The result lasts until the next buildup, or until the first chunk of it breaks off. ### Different conditions between runs The damper is throttled on one run and open on the next, the VFD speed drifts. The influence coefficient will come out wrong, and the correction random. ### Trial weight too small The readings barely changed, and the software calculates the correction from noise. The 20–30% or 20–30° rule exists exactly for this case. ### Weight mounted as a mirror image The angle-reference direction got mixed up, and vibration doubled instead of dropping. Fixed-position mode removes this error entirely: there's no angle and no direction there, just a blade number. ### Weights instead of alignment A strong 2x plus noticeable axial vibration means misalignment. That's fixed with shaft alignment — mass on the wheel won't help here. ## How Long It Takes and When to Call in Engineers The estimate below is typical, not a guarantee: on any given machine, access to the wheel and its condition decide everything. ### You can handle it yourself An accessible hatch, factory holes for weights, a clean wheel, steady speed, 1x dominating, and a repeating phase. A fan like this balances with the standard procedure, and the software walks you through it step by step. ### It's worth calling in engineers Vibration doesn't hold after balancing, the phase doesn't repeat run to run, there's a frame or ductwork resonance nearby, the machine is critical and downtime is costly, or you need a report with an ISO-based assessment. And separately — when you're not sure the cause is unbalance at all: half the work here is diagnostic. - Walk-around and mechanical checks: 15–30 minutes - Mounting the sensors and marker: 10–20 minutes - Baseline run with settling to operating condition: usually 5–15 minutes - Trial run per plane, including stopping and mounting the weight: 20–40 minutes — beyond that, everything depends on how convenient the access hatch is - Correction weights and verification run: 30–60 minutes - In total, for an accessible fan in one plane — usually 2–4 hours including stops. Two planes add one more cycle. If the wheel needs washing, add time for cleaning and measure again > AXILINE comes to your site with the Balanset-1A and balances the fan in its own bearing housings: no wheel removal, no dismantling ductwork, with a measurement before and after. The kit includes two accelerometers, a laser phase sensor, a two-channel USB module with preamplifiers and an ADC, and Windows software. The instrument calculates correction in one and two planes, splits mass across blades, calculates drilling, evaluates tolerance against grade G, and stores results for the report. The same instrument can also be bought and used on your own: it's designed and manufactured by engineers who do their own on-site balancing with it and provide consulting support for your machine. Sources: [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/) · [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) ## Frequently asked questions **Can a fan be balanced without removing the wheel?** Yes, that's the standard method. The wheel is balanced in its own bearing housings at running speed: sensors on the housings, a laser phase sensor on the shaft marker, weight mounted through the access hatch. You need to remove the wheel if there's no access to the correction planes, the rotor is flexible, the wheel is going in for repair anyway, or you need grade-G acceptance on a balancing machine. **How many runs does balancing a fan take?** One plane means two runs: baseline and trial. Two planes mean three. Plus a verification run, and if you don't hit the target right away, one or two short trim cycles. Rebalancing the same fan later using saved influence coefficients skips trial runs entirely. **Where do you mount the weight if the wheel has no holes?** In order of preference: factory holes and gussets on the back disc, welding on a plate where the manufacturer allows welding, or metal removal by drilling in a massive part of the hub or disc. Magnetic weights only work as trial weights. Don't drill a blade for a bolt without the manufacturer's approval: it's a load-bearing part. **Vibration went up after cleaning the wheel. Is that normal?** Yes, that happens often. The buildup sits unevenly and can partly offset the wheel's own unbalance. Remove the buildup, and the offset disappears. The correct order really is: clean, measure, balance. Doing it the other way around gives a result that lasts only until the next wash. **We found a crack in a blade. Should we balance it?** No. Balancing would lower the vibration and remove the one warning sign that made the crack visible from outside. The blade gets repaired per the manufacturer's procedure, or the wheel gets replaced, and only then does balancing happen. A blade that tears off at running speed destroys the housing. **The software says "within tolerance," but the fan still makes noise. Why?** "Within tolerance" means residual 1x is below the target value you entered. The rest of the vibration hasn't gone anywhere. The noise could come from blade-pass frequency with poor inlet flow, aerodynamic stall at low flow, a bearing, the belt drive, or a guard or duct resonance. Judge the machine by its overall vibration and spectrum, not by one line in the software.