Static and Dynamic Balancing: The Difference in Practice

You're standing next to the machine, you have a two-channel instrument, and the chief engineer asks: balance it statically or dynamically? Textbooks answer that through the axes of inertia; on site, the answer is shorter and more practical: how many correction planes you use, and how many runs you'll pay for it in. Below we go through the difference in terms of the actual work, not definitions: what physically remains in a long rotor after one mass, where the task really is purely static, and why knife-edges don't replace a phase measurement.

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

The difference is in the number of correction planes, not the measurement mode

Let's start with what usually gets confused. “Static balancing” and “dynamic balancing” don't describe how you measure — they describe how many correction masses you place, and in which planes. A correction plane is the rotor cross-section where a weight can be fitted or removed. In ISO 21940 terminology, the same procedures get more honest, unambiguous names: single-plane balancing and two-plane balancing. The words “static” and “dynamic” are left over from old shop-floor practice and still live on in everyday speech.

Single-plane balancing removes the resultant centrifugal force. The rotor carries excess mass, and its force rotates along with the rotor, loading the supports. You place one mass in one plane so that the total force comes close to zero. All you can compensate with a single mass is force.

Two-plane balancing removes two things at once: force and moment. Only a couple of forces creates a moment, so a single mass is physically powerless against it — not for lack of skill. You need two masses in two planes spaced apart along the rotor's length, and the instrument calculates them together, solving a system of two equations with an influence-coefficient matrix. An influence coefficient is the machine's response, measured on a trial run: how much, and in which direction, a known weight in a given plane changes the vibration at each support.

That gives you a practical selection criterion: does the couple component have a lever arm. A thin disc has almost no lever arm, and one mass is enough. An elongated or overhung rotor always has a lever arm, and any asymmetry along its length produces a moment.

We've covered the L/D rule (rotor length to diameter ratio) and the signs that one plane isn't enough in detail in a separate article on choosing the number of correction planes. What matters more here is this: the words “static” and “dynamic” themselves say nothing about whether the rotor is spinning or standing still.

Sources: ISO 21940-11:2016

The misconception: “dynamic” means “at running speed”

The phrasing you hear most often: static balancing is at rest, dynamic is at speed. It sounds logical, and it almost always gets in the way of the work.

Here's what actually happens. You arrive on site for a fan with a narrow impeller. You mount accelerometers on both bearing supports, aim the laser sensor at the reflective marker, and spin the machine up to its running 1500 rpm. You measure the amplitude and phase of 1x — vibration at the rotor's rotational frequency (phase is the angle that ties this vibration to the marker on the shaft) — and fit one weight to one blade. The rotor was spinning the whole time. There was one correction plane. By plane count, this is static balancing, carried out in a dynamic regime, at running speed, with a phase measurement. There's no contradiction here — the contradiction is only in the terms.

And the reverse case. A rotor with two correction planes spaced apart, on a balancing machine, spins at 300 rpm but will run at 3000. This is full two-plane balancing, even though the speed is nowhere near running speed. It's valid because a rigid rotor keeps its shape and mass distribution across the whole range up to the first critical frequency (the speed at which the rotor hits resonance), so a correction found at low speed also works at high speed.

What genuinely does depend on speed is something else: whether the rotor behaves rigidly, and how close you are to resonance. Balancing at a frequency where a support or frame goes into resonance is pointless regardless of the number of planes. The amplitude is inflated there, the 1x phase wanders, and the influence coefficients come out unreliable.

• “Static” and “dynamic” are about the number of correction planes: one or two.
• “At rest” and “at speed” are about the measurement method: gravity versus centrifugal force.
• These two pairs of concepts overlap but don't coincide. A single-plane job can be solved on knife-edges just as well as at running speed with phase.
• The only thing speed genuinely rules out is balancing inside the resonance zone.

This exact confusion is where the frequent request “do us a dynamic one, not a static one” comes from. Clarify what the person actually means: two-plane balancing, or balancing at running speed. These are different things — with a different price in downtime.

What physically remains in a long rotor after one mass

This is the core of the whole topic. Picture an elongated rotor carrying two unbalanced masses in two different planes along its length, pointing roughly in opposite directions. The center of mass may sit almost on the rotation axis, so the static component is small. But when the rotor spins, both centrifugal forces act at different points along the shaft and don't lie on the same line. They form a couple — a moment — that rocks the rotor about its center of mass: one bearing support goes up while the other goes down at the same instant.

Now fit one mass in one plane. You've created a single force. It can add to the moment in such a way that 1x drops nicely at the near support. At the far support it adds the other way, and things get worse there. You haven't removed the unbalance — you've pushed the vibration from one end of the rotor to the other. A month into operation, that comes back as a complaint about a bearing running hot on the side you didn't measure.

Here's what that looks like on a two-channel instrument's screen. Don't look at a single number — look at the two 1x vectors from both supports at once. Comparable amplitudes, close phases — the static component dominates, and one mass will help both supports. Comparable amplitudes, phases roughly 180° apart — the moment dominates, and single-plane correction won't get it. Phases 30–150° apart — you're looking at an ordinary mixed case, and it's solved in two planes.

A typical illustrative set of numbers for a dominant moment: the drive-side support at 7.8 mm/s with a phase of 32°, the opposite one at 6.9 mm/s with a phase of 211°. This is a hypothetical example for understanding the picture, not the result of an actual reading. What matters here isn't the magnitude, it's the phase relationship.

Assess the result at both supports at once, and by overall vibration — the total level across all frequencies — in mm/s RMS (root-mean-square value) over the 10–1000 Hz band, not just at the support where you got a nice-looking number. Zones A, B, C and D under the applicable part of ISO 20816 are covered in a separate article on the limits.

Sources: ISO 20816-1:2016

Comparison table: one plane versus two

The column with the number of runs is the price of the decision. An extra plane costs you one additional trial run — that is, a stop, fitting a weight, and a run-up. It only makes sense to save on it once the measurement has confirmed there's practically no moment.

Sources: ISO 21940-11:2016

Where the task really is purely static

There are rotors where two-plane balancing is excessive not out of laziness but because of geometry. They all share one thing: the mass is concentrated in one narrow plane, so the moment has almost no lever arm.

A grinding wheel. Flat, thin, all the mass in the plane of the disc. Balanced with a single mass in balancing grooves or with segments in the flange.

A thin pulley or drive disc. The lever arm is small, but the radial force from the belt and groove runout are often more important than the unbalance itself, so a pulley is worth checking together with the tension.

A circular saw blade. A thin blade, one plane by definition.

A flywheel and a brake disc. Parts that are compact in length, with mass at the periphery, where the correction radius is large and the weight comes out small.

A narrow fan impeller and a small blower wheel. Here it's already worth checking the second support, because an overhung mounting adds a moment even to a narrow wheel.

And a subtlety people forget. Even on a thin disc, one mass creates a small moment if the correction plane doesn't pass through the center of mass. Weld a weight onto one side face of the disc instead of the middle of the rim, and you get a lever arm equal to half the thickness. On a slow-running flywheel that's unnoticeable. On a high-speed wheel, it no longer is.

• The rotor's mass is concentrated in one narrow plane, L/D up to roughly 0.5.
• The rotor is rigid, running speed is noticeably below the first critical frequency.
• The correction plane passes through the center of mass, or very close to it.
• The 1x phases at the two supports are close, and the amplitudes are comparable.
• The second correction plane is physically inaccessible, and there's nothing to correct there anyway.

That last point comes up constantly on site visits: you only have access to one side of the impeller. In that case, single-plane correction isn't a choice, it's a condition of the job. An honest conclusion in that situation sounds like this: we reduced the vibration as far as one plane allows, and what's left is the couple component.

Knife-edge balancing: what it gives you and what it doesn't

It's an old, useful method. The rotor is laid with its journals on two horizontal knife-edges or on roller supports, allowed to turn freely, and the bottom point is marked. That's the heavy spot. A trial mass is placed opposite it, and the process repeats until the rotor stops spontaneously turning from any starting position.

What you get. The direction of the heavy spot, and a rough estimate of the mass, arrived at by trial and error. The result doesn't depend on speed or on the stiffness of the supports, because you're working with gravity, not centrifugal force. For a narrow disc, for pre-installation assembly, for screening out a gross mass skew, this is enough.

What you don't get. First, the moment. A pair of equal masses in two planes on opposite sides of the axis is balanced under gravity, so the rotor sits quietly on the knife-edges, and a static check won't catch that pair. A rotor with a substantial couple unbalance will pass a knife-edge check as sound, and then shake both supports out of phase on the machine.

Second, sensitivity. For the rotor to turn, the moment from the unbalanced mass, m·g·r, has to exceed the resistance moment in the supports. That sets a threshold: you can only detect unbalance greater than the friction moment M_fr divided by g. Anything smaller is hidden by friction. The threshold depends on the finish and geometry of the journals, on the condition of the knife-edges, on how level they are, and you don't know it in advance. That's why a knife-edge balancing result can't be translated into a G grade and written into a report.

Third, phase. By measuring the amplitude and phase of 1x on a spinning rotor, you get the response of the whole system — rotor, bearing supports, frame, foundation. That's exactly why a trial weight yields an influence coefficient, and the instrument calculates the mass and angle in one pass, without trial and error. Knife-edges simply don't give you this information at all.

A separate case: a rotor with no journals of its own, mounted on an arbor. The arbor's own eccentricity introduces unbalance and spoils the result. In spin balancing this is cancelled out with index balancing — two measurements with the rotor turned on the arbor. On knife-edges, arbor eccentricity simply adds to the rotor's unbalance, and there's nothing to separate the two.

The sensible order is this. Knife-edges are a rough preliminary step: they remove a large mass skew and save you the first trial run. Base the final assessment and the tolerance on measuring the amplitude and phase of 1x on the spinning rotor.

How to tell which job you're facing in a single run

1. 01

   Check the mechanics before balancing

   Foundation bolt tightness, soft foot (a machine foot that doesn't sit flush against the frame), play in the supports, the condition of the impeller's fit on the shaft. While you're at it, compare overall vibration against 1x. If overall vibration is several times higher, the main source isn't unbalance, and the number of planes doesn't matter.

2. 02

   Fit two sensors, not one

   One accelerometer on each bearing support, rigidly mounted, on a clean pad, in the same radial direction. Without the second channel you won't see the couple component and you'll be reduced to guessing.

3. 03

   Take the initial run and compare phases

   Record the amplitude and phase of 1x at both supports at stable running speed. Close phases point to a dominant static component; a divergence of around 180° points to a moment; an in-between divergence points to a mix.

4. 04

   Assess the geometry

   Calculate L/D from the distance between the accessible correction planes and the diameter at the weight-mounting radius. Compare the result against the phase picture. If they agree, the decision is obvious. If they disagree, trust the measurement.

5. 05

   Choose a mode and stick with it

   One plane, if the geometry and phases point to a static case and the second plane is inaccessible anyway. Two planes in every other case. Don't switch partway through the procedure: the influence coefficients are built around the scheme you chose.

6. 06

   Check the result at both supports

   A check run and the residual 1x at both supports. If one support comes into tolerance and the other doesn't, you're looking at an uncompensated moment, and you need the second plane.

When two planes are no longer enough

The claim that “two planes are always enough” is true with a caveat, and it's worth knowing that caveat before the site visit. It holds for a rotor that behaves as a rigid body at running speed — that is, one that runs noticeably below its first critical frequency and doesn't change shape under centrifugal forces.

A flexible rotor behaves differently. Its axis bends, the bend shape depends on speed, and the mass distribution in space stops being constant. A correction found at one speed works worse, or not at all, at another. Rotors like this are balanced by modes, that is, by bending shapes, with more than two planes and measurements at several speeds. This is a separate methodology, and it has its own part of ISO 21940 devoted to rotors with flexible behavior.

The second case where two planes fall short by design: shaft trains on three or more supports, long multi-section rotors, lines with several impellers on one shaft. There, the task doesn't reduce to a 2x2 matrix, and it's solved either section by section, or by a calculation with a larger number of planes.

There's a simple, useful sign to look for on site. If, after a careful two-plane balancing job, vibration is in tolerance at running speed but grows out of proportion with a small change in speed, you're not dealing with pure rigid-rotor unbalance. Look for support resonance or flexible rotor behavior, and don't keep adding mass.

An honest answer to the client in a situation like this is worth more than a fourth trial run. If the rotor is flexible or the supports are in resonance, additional weights won't improve the result, and sometimes will make it worse.

Sources: ISO 21940-12:2016

What the instrument does, and when it's simpler to call in a site team

The Balanset-1A covers both jobs with one kit: two accelerometers on the bearing supports, a laser phase sensor referenced to a reflective marker, a two-channel USB module with preamplifiers, integrators and an ADC, and Windows software. You choose single-plane or two-plane mode, and from there the instrument calculates the correction by the influence-coefficient method.

What's relevant here: two channels show the amplitude and phase of 1x at both supports at once, so you make the decision on the number of planes from the data, not from the rotor's appearance. The software calculates the allowable residual unbalance by balance quality grades G, recalculates weights for other correction planes, and works in fixed-position mode, outputting a blade or hole number instead of an angle. Index balancing for rotors on an arbor and trim balancing — a touch-up using saved influence coefficients with no new trial runs — are both supported. A polar diagram, FFT spectrum, time waveform and a report archive are all there too. For building into your own rigs and machines, there's a Balanset-1A OEM version without the case.

You can handle the job yourself if you have a two-channel instrument, access to the correction planes, and the ability to stop and start the machine several times. If stopping the machine is costly, the rotor is overhung and awkward, and the vibration history is unclear, it's more sensible to call in a site team. AXILINE's engineers, who design and manufacture the Balanset instruments and balance with them on site themselves, come to your site, measure vibration at both supports, separate unbalance from misalignment and resonance, choose the number of planes based on the actual data, and leave you a report with the residual values. Advisory support is also available if you want to do the work yourself and need help working through a specific 1x and phase picture.

• One plane: initial run, trial run, correction, check run. Usually faster.
• Two planes: initial run, two trial runs, correction, check run. One extra trial run, against the risk of pushing vibration onto the second support.
• A repeat visit to the same machine: saved influence coefficients remove the trial runs, leaving just a trim correction.
• Preparation on the client's side (access, sensor mounting pads, a marker on the shaft, the ability to run the machine) cuts time on site more than any instrument settings can.

Sources: Balanset-1A manufacturer specification · Balanset-1A operation manual

Frequently asked questions

Related content