Overhung and Between-Bearings Rotors: How the Support Layout Changes Balancing

The same impeller gets balanced differently depending on whether it sits between the bearings or is cantilevered out beyond them. The support layout decides where to put the sensors, how many correction planes to use (the rotor cross-sections where weights are fitted), how strongly those planes interfere with each other, and how many runs you'll spend. We go through the two main cases, and a third one that often gets mistaken for the first two.

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

Two questions that settle everything before the trial weight

A rotor doesn't behave the way it looks. The instrument doesn't see unbalance, it sees vibration at the bearing supports, and geometry sits between the mass and the supports: the distance from the weight to each support, and the sign of the moment that weight creates. So the first thing you do on site is look at the bearing housings, not the wheel.

Answer two questions. How many bearing supports are there in the shaft line, and which side of the mass do they sit on. Where can you physically fit a weight relative to those supports. From there, the whole job is set in advance: the number of planes, the measurement points, the expected phase behavior, and the number of runs.

The layout is determined by the position of the mass relative to the supports, not by the rotor's length. The L/D ratio (rotor length to diameter) only hints at the number of planes, and is covered in a separate article on the L/D rule. What matters here is where those planes end up relative to the bearings.

Between-bearings rotors: a beam on two points you can reason about

A weight fitted between the supports loads both of them in the same direction. The near support's reaction is larger, the far one's smaller, but they share the same sign. That's the main advantage of this layout: the response at the supports is easy to reason about and easy to separate. Shift the weight toward support 1, and its reading jumps more than support 2's does, exactly by the lever rule.

The phases at the two supports in this layout tell you the nature of the unbalance right away. The 1x phase is the angle that ties the vibration at running speed to the marker on the shaft. Close phases with comparable amplitudes mean the static component dominates, and one mass removes it. Phases diverging toward 180° reveal a couple component, which needs two planes and two masses at different angles.

Space the correction planes as far apart as possible within the rotor body: at the end face near support 1 and the end face near support 2. The lever arm between the planes works in your favor. The bigger it is, the smaller a mass it takes to compensate the same moment, and the more confidently the software can separate the contribution of the two planes.

Cross-coupling — the fact that a weight in one plane shows up at both supports at once — is present here, but stays moderate and predictable. The instrument accounts for it automatically through the influence-coefficient matrix; the mechanics of that calculation are covered in our article on the influence-coefficient method. The practical upshot: a between-bearings rotor usually settles within three runs, plus one trim run if needed.

ISO 21940-11 covers the split of tolerance between two correction planes separately, and dividing the tolerance evenly is only correct when the planes are positioned symmetrically about the rotor's center of mass. Check the applicable part and edition for your own machine.

Sources: ISO 21940-11:2016

Overhung rotors: one weight produces both force and moment

Now cantilever that same mass out beyond the supports. Let L be the distance between the bearings and a the overhang from the near support to the wheel's plane. The centrifugal force F from the weight on the wheel produces a reaction of F·(L+a)/L at the near support, and F·a/L at the far one, pointing in the opposite direction. The near support acts as the fulcrum of a lever, the far one as a tie-back.

This arithmetic gives you three things you'll see on screen. First: the 1x phases at the two supports on an overhung machine diverge by around 180° even with purely static unbalance in the wheel. On a between-bearings rotor you'd read that divergence as a sign of a moment; here it's normal for the layout. Second: the near support's reaction in terms of force is always larger. Third, and this is the most confusing one: vibration in mm/s at the far support often turns out to be higher.

There's no contradiction here. Vibration isn't force — it's force divided by the support's dynamic stiffness. The far support often sits on a taller or more compliant stand, carries less mass, or is itself closer to its own natural frequency. Add the whole assembly rocking on its frame: if the rocking node falls between the bearings, the far end traces an arc of a larger radius. That's why both channels are essential on an overhung machine, and a conclusion drawn from one support alone means nothing.

One more feature of the overhung layout: the bending moment from the overhang rocks the stand and the frame, and the axial 1x component on machines like this can be noticeable with no misalignment at all. Apply the familiar rule about high axial vibration more cautiously here, and cross-check against the signs in our article on telling unbalance apart from misalignment.

Why an overhung rotor almost always needs two planes

The temptation is understandable: the wheel is narrow, L/D is small, so one plane should do. On an overhung layout this rule often lets you down, and here's why.

The correction planes you can reach on the wheel are the front disc and the back disc. The unbalance's center of mass almost never sits exactly in one of them: the blades wear unevenly across their width, build-up collects toward the suction side, a crack sits near the root. Fitting a weight only on the back disc compensates the force but adds a moment that wasn't there before. The near support calms down, the far one climbs.

The reverse logic holds too. The overhang a on an overhung fan is often comparable to, or larger than, the bearing span L. That means any mass on the wheel produces a lever arm relative to the bearings, and the moment arises on its own, from the mounting geometry, not from the wheel's width. One plane removes one vector quantity, and you have two.

The practical conclusion is simple. If two planes are accessible on an overhung wheel, use two. If only one is accessible, work with one, reduce what can be reduced, and honestly write in the report that a couple component remains. The software does have a function to recalculate correction weights onto other accessible planes, but recalculation won't create a lever arm where none exists.

Worse convergence: how well-conditioned the problem is, and the order of runs

On an overhung rotor, both correction planes lie on the same side of both bearings, and the distance between them, b, is usually small compared with the overhang a. To the machine, these two planes look almost identical: a trial weight in the front plane and one in the back plane produce very similar changes in the vectors at both supports.

Mathematically, this is an ill-conditioned system: a solution exists, but a small measurement error changes it drastically. The two columns of the influence-coefficient matrix are close to each other, and the determinant is small. You'll recognize the symptom right away: the software outputs two large masses at almost opposite angles that add up to a small net correction. Formally everything checks out; physically you're hanging an extra kilogram on the wheel in the form of two weights that nearly cancel each other.

That's where the discipline required around runs comes from. Each trial run gives you one column of the matrix, and every column has to belong to the same machine in the same state.

1. 1

   Space the planes as far apart as possible

   The front and back disc of the wheel, not two neighboring spots on the same disc. Every centimeter of b improves conditioning and lowers the calculated masses.

2. 2

   Use the same trial weight in both planes

   The same mass at the same radius. That way you compare the columns directly and can see for yourself how distinguishable the planes are.

3. 3

   Start with the plane closer to the supports

   It gives a bigger, cleaner response. Check the criterion for a valid trial run: a change in 1x amplitude of at least 20–30%, or in phase of at least 20–30°, at at least one support.

4. 4

   Between runs, change nothing but the weight

   Not the speed, not the sensor positions and orientation, not the tachometer marker, not the damper, not the bearing temperature. Shift a sensor after the first trial run, and the matrix ends up built from two different machines.

5. 5

   Two columns nearly identical: change strategy

   First remove the static component with equal masses at the same angle in both planes, then handle the moment separately with equal masses at opposite angles. This split approach is more robust against error than solving the matrix directly.

6. 6

   Plan a trim run in advance

   For an overhung machine, one round of trim balancing (a touch-up with small weights based on already-known coefficients) is normal, not a sign of a mistake. Four trim rounds in a row is already a diagnosis: look for resonance, loose fasteners, or mass drifting on the wheel.

Check the check run at both supports, not just the one you started with. On an overhung rotor, a normal result is a drop at both supports at once.

Table: sensors, planes and cross-coupling

In every layout, mount the sensors on the bearing supports, as close to the bearing as possible, on a rigid fixing, and in the same direction from one run to the next. Assess the machine's overall condition in mm/s RMS against the applicable part of ISO 20816: the limits for measurements on non-rotating parts are written exactly for this.

Sources: ISO 20816-1:2016

Three supports or more: why this is a separate problem

The two-plane scheme rests on the model of a rigid rotor on two supports. A third support breaks that model in two ways at once.

First: the system becomes statically indeterminate. That means the support reactions depend not only on the unbalance, but also on exactly how the intermediate support is set in height and how well aligned the sections are. Raise the intermediate bearing by a few tenths of a millimeter, and the load distribution shifts along with the influence coefficients. That's why geometry, alignment and soft foot (feet not sitting flush against the frame) on a multi-support machine get put right before balancing, not after.

Second: the dimensionality of the problem changes. A three-support shaft is balanced in three planes, and you have to solve a system of three equations with three unknowns, using the influence coefficients from all three supports. The first run gives the initial vectors at all supports, then one trial run per plane. At least four runs for three planes, and five for four. Each trial run doesn't give you a single number — it gives a whole column of the matrix: you fit a weight in one plane and read the response at every support at once.

A two-channel instrument solves the problem for two planes. On a three-support shaft there are two honest paths. Either you set up a measuring system with the right number of channels and an algorithm for three or four planes, or you split the task by section, work two channels at a time, and check the result with a check run across all supports. The second path works, but repeatability drops, and that has to be agreed before the work starts.

• Don't treat an assembly made up of a motor, a gearbox and a driven machine as a single three-support rotor. It's several rotors in one shaft line, each with its own unbalance, plus couplings and misalignment between them. Balance one rotor at a time.
• A flexible rotor that changes shape at running speed isn't described by the rigid-rotor scheme, whether on two planes or three. Modal methods apply there, and that's a separate class of problem.
• See the ISO 21940 series for rotor classification and approaches to flexible rotors; determine the applicable part and edition for your own machine.
• Under the ISO 18436-2 classification, on-site two-plane balancing falls into a higher competence category than single-plane. Multi-support and flexible rotors sit higher still.

Sources: ISO 21940-11:2016 · ISO 21940-12:2016

Work out your layout in a minute, just by looking at the machine

You don't need a drawing or a datasheet for this. Trace the shaft line with your eye from the coupling or pulley to the impeller, and count the bearing housings.

• The wheel sits between two bearing housings: a between-bearings layout. A double-sided suction duct on a fan almost always means exactly this.
• Both housings on the same side of the wheel: an overhung layout. A single-sided suction duct, a free end carrying the wheel, an end cover.
• The wheel sits directly on the motor shaft with no bearings of its own: an overhung layout on the motor's own bearings. The overhang here is often larger than the bearing span, and the moment will be pronounced.
• There's an intermediate bearing housing between you and the wheel: three supports or more.
• Two coupling halves in the line: you're looking at two rotors and four supports. Weights don't fix misalignment.
• Measure three numbers with a tape: L between the bearings, a from the near support to the wheel's mid-plane, b between the accessible correction planes. Write them into the report.

These three numbers read like a forecast. The ratio a/L shows how strong the moment will turn out to be: the bigger the overhang, the more pronounced the phase opposition between the supports. The ratio b/L shows how well the software will be able to separate the two planes: a small b means large weights, sensitivity to error, and a likely trim run.

What this changes in practice, and how AXILINE helps

The support layout determines not just the method, but the shift planning too. A between-bearings rotor with two accessible planes is usually wrapped up in three runs. An overhung fan with a narrow wheel needs the same three runs plus one trim run, and almost always two planes. A multi-support shaft first needs a decision on the number of channels and planes, and it's better to make that decision before the site visit, not on site.

The tool for this job needs to be two-channel and phase-capable. The Balanset-1A measures both bearing supports at once with two accelerometers, takes the reference phase from a laser sensor against a reflective marker, and calculates single-plane and two-plane balancing by the influence-coefficient method — meaning it accounts for cross-coupling on its own. A few more things come in handy on overhung machines: fixed-position mode for blades or holes, the drilling-correction calculation, recalculating weights onto other planes, trim balancing from already-obtained coefficients, a polar diagram, and a report archive. For building into rigs and machines, there's a Balanset-1A OEM version without the case.

People operate the instrument, and on overhung rotors that shows: where cross-coupling is strong, the decision on the number of planes and the order of runs is worth more than calculation speed. AXILINE's engineers design and manufacture the Balanset instruments and balance with them on site themselves. Send us the machine type, the speed, photographs of the unit from both sides, and the three numbers L, a and b, and we'll work through your layout and propose an order of operations. If you'd rather do it yourself, advisory support on the instrument and the methodology is still available.

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

Frequently asked questions

Related content