Fans are the most commonly balanced machines in the field, and for good reason: an impeller picks up imbalance constantly — dust and process residue build up unevenly on the blades, blades erode or corrode, and repairs (a welded patch, a replaced blade) shift mass. The good news is that a fan is also one of the most rewarding rotors to balance in place: access is usually reasonable, the rotor is rigid, and a session from first measurement to verified result typically fits in an hour.
This guide walks through the in-situ procedure: instrument setup, the measurement runs, weight placement, and acceptance. It applies to radial and axial industrial fans, exhausters, blowers, and cooling-tower fans alike.
Before you balance: rule out everything else
Imbalance is only one of several reasons a fan vibrates. Balancing a fan with a failing bearing or loose anchor bolts wastes a shift and helps nothing. Spend ten minutes first:
- Clean the impeller. Caked dust is removable imbalance — wash or scrape it off before adding permanent weights, or the next cleaning will throw the rotor out of balance again.
- Check the blades for cracks, erosion, and loose patches.
- Rock the shaft to feel for bearing play; listen to the bearings at speed.
- Check anchoring — frame bolts, vibration isolators, duct connections.
- Measure first, conclude second. Take a vibration reading in Vibration Meter mode (F5): compare the overall RMS (Vs) with the 1× component (Vo). Vibration dominated by 1× shaft speed points to imbalance; a strong 2× line on the spectrum suggests misalignment; a raised broadband floor suggests bearings. If the picture is not clearly 1×, see When Balancing Doesn’t Help before going further.
Setting up
Sensors. Mount the vibration sensors on the bearing housings — as close to the bearings as you can get, on clean, flat metal, axis perpendicular to the shaft. How many you need follows from the number of planes (see Planes below): two-plane work uses one sensor per bearing support — X1 near the impeller-side bearing (plane 1), X2 near the drive-side bearing (plane 2) — while single-plane work needs only X1, on the bearing nearest the impeller. On newer instrument revisions the same channels are labeled Ch-1 and Ch-2, with the tachometer on Tacho. A magnetic base needs a snug, paint-free spot; the sensor must not touch anything that moves.
Tachometer. Stick a piece of reflective tape on the shaft or pulley, set the laser tachometer on its magnetic stand 50–500 mm away, and aim the beam at the tape. Two things matter: the stand must not move between runs, and direct sunlight or a bright lamp shining into the optics can corrupt the pulse — shade it if needed.
Planes. For a narrow, disk-like impeller (width much smaller than diameter), single-plane balancing is usually enough. For wide impellers, squirrel-cage wheels, and rotors with two bearing spans, balance in two planes. When in doubt, start with one plane — if residual vibration stays high with a clean 1× signature, switch to two.
The measurement runs
The influence-coefficient method builds its data from two or three short runs at a steady operating speed:
- Run 0 — baseline. Start the fan, let the reading stabilize, record amplitude and phase.
- Run 1 — trial weight in plane 1. Stop, fix a trial weight of known mass at any convenient blade in plane 1, note its position. Run again. The reading must change noticeably — at least 20–30% in amplitude, or 20–30° in phase. If it barely moves, the trial weight is too light, and the software says as much — trial weight too small, increase it and repeat the run. Increase it and re-run rather than continue with weak data, but stay inside the k = 0.05–0.10 band of the formula below: k = 0.10 is the ceiling, and simply doubling a weight already sized at k = 0.07 puts you well past it.
- Run 2 — trial weight in plane 2 (two-plane balancing only). Move the same weight to plane 2 and repeat.
A reasonable starting point for the trial mass:
mtrial = k × M × g / ( R × ω² )
where M is rotor mass in kg, R the mounting radius in meters, ω = 2πN/60 the angular speed in rad/s, g = 9.81 m/s², and k = 0.05–0.10 — the share of the rotor’s own weight you allow the trial weight’s centrifugal force to reach. With those SI inputs the answer comes out in kilograms; multiply by 1000 for grams. Straight to grams, in the units you actually read off the machine: m(g) ≈ 9 × 10⁸ × k × M(kg) / ( R(mm) × N(rpm)² ). For a 60 kg impeller at 1000 rpm with weights at a 400 mm radius and k = 0.07, that is 0.0094 kg — about 9 g. Heavier rotor → more mass; faster rotor → much less.
Do not size the trial weight as a percentage of rotor mass: the centrifugal force it generates scales with speed squared, so a "safe" percentage at 500 rpm is a hazard at 3000 rpm.
The software can do this arithmetic for you, with two conditions. First, the First trial weight estimate window belongs to the modern unified balancing form: you have to select it in Settings (F4) → Balancing form → Modern, and that group only appears on the newer firmware. A fresh install runs the classic forms, which have no such window — so do not expect the estimator to be on screen by default. Second, the window computes nothing until you type the Rotor mass (kg) and Mounting radius (mm) into it and Run 0 has been recorded; before that it returns neither a mass nor an angle.
What it returns depends on the Support type you choose. On Rigid supports it is exactly the k = 0.07 form of the formula above. On Soft supports — the default — it sizes the weight from the measured Run 0 amplitude instead, and uses the k = 0.07 centrifugal value only as a ceiling. It also proposes an angle, but that angle is referenced to the tachometer mark, not to a trial weight, and it is only meaningful when the sensors are mounted coaxially with the tachometer and the reflective mark sits at the tacho trigger point. Read it as a starting point; the 20–30% response check on Run 1 is still what tells you the weight was right.
The software then computes the correction: a mass and an angle for each plane.
Placing the correction weights
The angle convention causes more failed balances than any other single mistake: the angle is counted from the trial-weight position, in the direction of rotation. Counting it against rotation puts the weight in a mirror-image position and the vibration goes up, not down.
Ways to attach the correction mass on fans: welded plates (most permanent), bolted balance weights on a flange, or clip-on weights on blade edges where the design allows. If adding mass at the computed spot is impossible, switch the correction method to Remove mass — grinding or drilling is sometimes the only option on a closed wheel. Do not flip the angle yourself: the software already adds the 180° when you select "Remove mass", so grind exactly at the angle it displays. Adding another 180° puts the correction on the wrong side of the wheel.
Remove the trial weight before the check run unless you told the software it stays on the rotor — a fallback for a weight you genuinely cannot get back off (a tack-welded plate, a bolt you can no longer reach), not a shortcut. The checkbox sits under the Leave on rotor label, one per plane in two-plane mode; the classic form’s result page then flags that plane Remain trial weight.
Verifying and accepting the result
Run the fan once more — this check measurement is Run T (Trim) in the software. Two acceptance views:
- Vibration severity — compare the bearing-housing reading against ISO 20816 zones (formerly ISO 10816) for your machine class. Zone A/B is the normal target for a fan returned to service.
- Residual unbalance — check the result against an ISO 21940-11 grade (formerly ISO 1940-1). G6.3 is the grade the standard lists for fans and pump impellers; tighter grades apply only where the machine’s own specification calls for them. The built-in ISO 1940 tolerance calculator turns grade, rotor mass, and service speed into permissible residual gram-millimeters — divide by your correction radius to get grams.
If the first correction lands close but not inside the target, repeat Run T: the software refines the correction from the latest data without starting over. One trim run is routine; needing three or more usually means something else is wrong — loose weights, a moved tachometer, or a fault that is not imbalance at all.
Save the PDF report before you pack up: before/after amplitudes, weights, and angles. The influence coefficients are archived with the session — the next time the same fan needs attention, load them with Apply coefficients and the software skips the trial runs, computing the correction straight from Run 0.
Worked examples
Real jobs with numbers and timing: a cooling-tower fan stabilized in 25 minutes and a generator cooling fan balanced without dismantling. For choosing the instrument itself, see Choosing a Field Balancer.