G6.3 is fine for general service. For API 673 critical service we hold G2.5 — here is the arithmetic that makes the difference, and the balancing report that proves it.
“Dynamically balanced” on a fan quotation can mean a measured, recorded number — or it can mean nothing at all. The word only becomes a commitment when it carries a grade, a speed, and a residual-unbalance figure you can read off a test record. For general industrial fans we balance to G6.3. For API 673 critical service we hold G2.5 as our house balancing standard — and the reason is arithmetic, not marketing.
ISO 1940-1 (now ISO 21940-11) defines balance quality as a G-grade: the product of a rotor’s permissible specific unbalance and its angular velocity, expressed in mm/s. Lower G means finer balance. The grade on its own is abstract — what the balancing machine actually measures, and what the report actually states, is a permissible residual unbalance in gram-millimetres. That number falls out of the grade, the rotor mass and the service speed:
One rule governs the whole calculation: N is the maximum speed the wheel can see in service — the VFD ceiling frequency, not the nominal motor speed. Plug in a lower N and you silently inflate the tolerance.
Take a shrouded impeller of 120 kg running at a maximum 1,480 rpm. At the general-service grade G6.3, e_per = 9549 × 6.3 / 1480 = 40.6 g·mm/kg, so U_per = 40.6 × 120 ≈ 4,877 g·mm total. Re-run the same wheel at G2.5 and the tolerance collapses:
| 120 kg wheel · 1,480 rpm max | G6.3 (general) | G2.5 (critical) |
|---|---|---|
| Specific unbalance e_per (g·mm/kg) | 40.6 | 16.1 |
| Total residual U_per (g·mm) | 4,877 | 1,936 |
| Per-plane share (g·mm) | ~2,439 | ~968 |
| Correction mass at 600 mm radius (g) | ~8.1 | ~3.2 |
G2.5 is roughly 2.5× tighter than G6.3 (the ratio of the grades themselves), and the residual you are allowed to leave in the wheel drops from about 4,877 to about 1,936 g·mm. At a 600 mm correction radius that is the difference between chasing an 8 g imbalance and a 3 g one — well inside the resolution of a properly calibrated soft-bearing machine, but only if you set out to hold it.
A real impeller distributes its mass along the shaft axis, so correcting in a single plane can leave a couple that still shakes the bearings. Production balancing is therefore two-plane: measure and correct at both faces, and the per-plane residual — roughly half of U_per for a symmetric wheel between bearings — must sit at or below its share before the wheel passes. Correction weights are welded or mechanically fixed per the impeller material rules; nothing is tacked on where it could detach into the airstream.
API 673 (3rd Ed.) covers special-purpose fans for petroleum, chemical and gas service, and it does not invent its own balance limit — it invokes ISO 1940-1 directly. That is the honest chain: the standard points at the residual-unbalance method above, and critical-service fans are held to a finer grade because the vibration they can tolerate over years of continuous duty is far lower. To be precise about what we are claiming: G2.5 is our balancing house standard for API-class and critical-service wheels, verified on the shop machine and recorded — it is not, and we do not present it as, a witnessed API 673 mechanical run-test. Two different assurances; we are careful not to conflate them.
None of this counts unless it is written down. Our balancing report travels with the fan and states: impeller mass, service speed, specified grade, the U_per calculation, initial and final residual unbalance per plane, correction masses and their positions, the balancing machine ID and its calibration-due date, and operator plus QA signatures. A wheel that will not come inside tolerance is quarantined, not shipped “close enough.” If you are specifying a critical-service fan, the three questions worth asking any supplier are: what grade, at what maximum speed, and may I see the residual on the record?
Balance is a state, not a permanent property — it is lost to erosion, to one-sided deposit, and to a cleaning crew that knocks off a load-bearing weight. That is why we re-balance after any rotor work before restart, and why field trim-balancing is judged against ISO 20816 vibration zones. If a fan in service is shaking, first prove it is imbalance and not resonance — see our notes on field vibration diagnostics and resonance versus imbalance. And because N drives the whole tolerance, it pays to nail the service speed down first when specifying the duty point.
Talk to us about balancing to G2.5 →
Jitamitra Electro Engineering · Fan-engineering notes, written for the engineer.
Sources & basis. Grounded in our internal impeller-balancing SOP and its companion marketing brief: the ISO 1940-1 / ISO 21940-11 residual-unbalance method (e_per = 9549 × G / N; U_per = e_per × mass), the worked 120 kg / 1480 rpm example, two-plane correction practice, the controlled balancing-report format retained in shop use, and the SOP grade table (G6.3 general service, G2.5 for API 673 / critical service where API 673 invokes ISO 1940-1). Anonymised: no customer, job, family-type or internal record codes reproduced. Standards are named as methods; G2.5 is stated as our balancing house standard, with no claim of a witnessed API 673 mechanical run-test.
Flow, static, gas temperature, application — or attach a spec, GA drawing or a multi-fan schedule. Engineer to engineer.
ISO 9001:2015 quality system · performance-tested to IS 4894 / ISO 5801 / AMCA 210 method · witnessed FAT on request, at no cost.
*For our standard range, additional days required for special projects