Shaft Misalignment: Types, Signs, and How to Check It

The coupling on the pump feels warm to the touch, there's black rubber dust inside the guard, and you're replacing the drive-end bearing for the second time this year. Vibration, meanwhile, is "tolerable," and the unit keeps running. That's usually what shaft misalignment looks like: a defect that rarely produces frightening numbers on a vibration meter, while it methodically takes the machine apart from the inside. Let's go through its three types, the damage mechanism, the signs you'll see in measurements, and the sequence for checking the geometry.

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

Misalignment Is an Assembly Defect, Not a Rotor Defect

A rotor can be perfectly balanced and the machine will still shake and chew through bearings. Because misalignment doesn't live in the rotor. It lives in the geometry of the unit: in the motor's position relative to the pump, in the shim pack under the feet, in the frame, in the anchor bolts, in the grout, in a rigidly connected pipe. You correct it not with weights, but by moving the machine on the foundation.

Keep the difference between the two terms straight. Misalignment is the cause, a state the machine is in. Alignment is the action you take to fix that state. Confusion between the terms regularly produces work orders asking to "balance it because it's misaligned" — but these are two different jobs with two different tool sets.

The scale involved here is unfamiliar to a fitter's eye. We're talking about hundredths of a millimetre at the coupling. A straightedge held against the coupling halves and a feeler gauge in the face gap will catch a gross installation error, but not the kind of misalignment that kills a bearing in six months. That's why "eyeballed" alignment and dial-indicator alignment produce two different machines in the end, even though both look correctly assembled.

• Installation done by straightedge and feeler gauge, without dial indicators and without calculating shims.
• Uneven thermal growth: a unit set to zero cold ends up misaligned once it reaches operating temperature.
• Foundation settling, cracks and delaminated grout, loosened anchor bolts.
• Strain from a rigidly connected pipe or duct: the pipe pulls on the nozzle and drags the machine out of position.
• Soft foot: tightening a bolt distorts the casing and pulls the bearing housings off a common axis.
• Replacing the motor or coupling with a different size without re-aligning by dial indicators.
• Worn fits, a bent shaft, runout of the coupling half on the shaft: the geometry drifts without the machine moving at all.

Misalignment often turns out to be the root cause disguised as several different defects at once: elevated vibration, coupling wear, premature bearing failure, seal leakage, broken coupling bolts. Fixing it can sometimes solve four problems in one action.

Three Types of Misalignment and the Four Numbers That Describe It

The textbook split into three types is useful because it tells you which way to move the machine and what to change in the shims. It's also misleading: pure cases almost never occur in the real world.

A real measurement doesn't give you a "type" — it gives you four numbers: offset and angle in the vertical plane, offset and angle in the horizontal plane. Aligning the unit means bringing all four within tolerance: you correct the vertical with shims under the feet, the horizontal by shifting the machine sideways. A fifth number, the axial gap between the coupling halves, also needs to be set to the coupling manufacturer's requirement, otherwise the flexible element works under preload or with play.

The angular component is specified not in degrees but as a gap difference over a length: millimetres per 100 mm of diameter, or over the length of a spacer. The same angle on a 300 mm coupling will produce three times the gap spread it would on a 100 mm coupling. That's why "raw" readings from one machine can't be carried over to another and compared directly.

What Misalignment Does to the Coupling, Bearings, and Seals

The mechanism is simple. The coupling joins two shafts that aren't sitting where they'd like to sit. Which means it's deformed. Deformation of an elastic body is a force: the stiffer the element and the larger the offset, the bigger the reaction. Over one revolution, that deformation goes through a full cycle twice, which is where the component at twice the running speed comes from.

You get two effects at once: a constant side load that continually presses the shaft in one direction in the stationary frame of reference, and an alternating component that loads the coupling, shaft, and bearings back and forth. The first wears out bearings and seals. The second builds up fatigue in the metal.

• The coupling heats up. The flexible element works through hysteresis: part of the deformation turns into heat. A warm guard and black rubber dust inside it are a direct sign that a vibration meter won't show you.
• The rubber ages and crumbles, the pins and bushings work loose, the teeth on jaw-type elements shear off on one side. Replacing the element without aligning the shafts means you'll be back to the same problem in a few months.
• The bearing loses service life. Life is calculated per ISO 281: for ball bearings it depends on the equivalent dynamic load raised to the third power. The arithmetic is unforgiving: a 26% increase in load and the calculated life is cut in half. The side load from misalignment adds to the working load and can easily produce that kind of increase.
• Seals wear unevenly. The shaft travels an elongated or figure-eight path, the lip seal wears down on one side, the mechanical seal loses flatness. A leak on one side and repeated replacement of the same seal is a conversation about geometry, not seal quality.
• Coupling-half bolts fatigue. The alternating bending moment loads the fasteners with a cycle twice per revolution. Typical failure points: flange coupling-half bolts, the fillet at the hub, the keyway.
• The shaft works under alternating bending. In the stationary frame the force is nearly constant, but inside the rotating shaft it travels around the cross-section, so every fibre of metal is alternately stretched and compressed.
• Part of the power goes into heating the coupling, bearings, and seals instead of useful work.

Pay attention to the check sequence on repeat failures. If the same bearing or the same seal fails a second time, replacing the part again is pointless until you've measured the alignment and checked for soft foot.

Sources: ISO 281:2007

Signs in the Measurements: What the Instrument Will Show

One caveat up front, without which everything else turns into guesswork. Vibration gives you a hypothesis, not a diagnosis. Misalignment is confirmed by direct measurement of the geometry, and the spectrum only tells you that it's worth picking up the dial indicators and spending an hour.

You need to compare points against each other, not just look at one. A two-channel instrument is more convenient here than sequential measurements: you get the phase at two points in a single run, under identical conditions, off the same reflective tape mark on the shaft.

• A noticeable 2x in the spectrum. Working threshold for suspicion: a 2x-to-1x ratio above 0.5. The 2x component appears because the coupling deforms twice per revolution.
• Strong axial vibration. One measurement along the shaft at the bearing housing next to the coupling takes two minutes and settles half the question. With angular misalignment, the axial reading is often comparable to the radial or even higher.
• A phase shift of around 180° across the coupling in the axial direction. Sensors sit on the motor bearing housing and on the driven-machine bearing housing, both along the shaft, oriented the same way. The two halves of the unit move toward and away from each other.
• A response to warm-up. The level drifts over the 20–60 minutes it takes to reach operating temperature, because the shafts move apart from thermal growth. Imbalance doesn't behave this way.
• A response to load. Readings drift together with pressure, flow rate, and the temperature of the pumped fluid.
• A figure-eight-shaped orbit. This sign is available on machines with sleeve bearings and permanently installed shaft-displacement probes; you won't get it from a portable vibration meter.

• Take three directions at each bearing housing on both sides of the coupling: horizontal, vertical, axial.
• Record overall vibration in mm/s RMS and the 1x amplitude separately. The difference between them already helps sort out the problem.
• Set Fmax so that 1x, 2x, and 3x land clearly within the spectrum, and record the 2x-to-1x ratio.
• Take the 1x phase at two points simultaneously, with identical sensor mounting.
• Repeat the measurement on the cold and on the warmed-up machine, at the same operating condition.
• Inspect the coupling, guard, feet, and seals. A visual inspection reveals signs that no spectrum will show.

A rigid flanged coupling can transmit enormous forces with almost no rise in 2x: it has nothing left to deform. The absence of 2x doesn't rule out misalignment. In that case the diagnosis rests on axial vibration, the phase shift across the coupling, and direct geometry measurement. A detailed, point-by-point breakdown of the signs is covered in our article on telling imbalance apart from misalignment.

Sources: ISO 13373-3:2015 · ISO 20816-1:2016

Soft Foot: the Most Common Cause of False Misalignment

The machine stands on four feet, but a plane is defined by three points, so one foot is always the "odd one out." It's exactly like a four-legged table on an uneven floor: three legs touch the floor, the fourth can be left hanging. The difference is that a table just wobbles, while a machine casing distorts when the bolt is tightened and pulls the bearing housings off a common axis.

This is where things get unpleasant. The shaft inside the machine is already tilted, even before the coupling comes into it. A laser system will honestly show you the misalignment. You'll honestly fix it, spending half a day. And within a shift the level comes back, because the cause was left under the foot. On top of that, the alignment itself won't converge: you move the machine, tighten the bolts, and the readings shift differently every time.

That's why soft foot is checked before alignment, not after — and rechecked after every time the shim pack is rebuilt.

• Parallel gap: the foot hangs evenly above the pad, and needs a shim matched to the gap thickness.
• Angular gap: the foot rests on one edge, the gap is wedge-shaped. Flat shims won't close it evenly; the pad needs to be worked on.
• Springing foot: the foot itself or the frame is deformed and flexes under tightening. The most stubborn case.
• Debris under the foot: dirt, a burr, a layer of paint, an old forgotten shim, a weld bead.
• A stack of a dozen thin foils. It acts as a spring and produces soft foot even with a perfect fit.
• External strain: a pipe, duct, cable gland, guard. It pulls on the casing the same way a hanging foot does.

• Loosen all the bolts and inspect the pads. Wear marks covering only part of the pad speak for themselves.
• Tighten all the feet to the specified torque. Place a dial indicator on one foot, loosen only that foot's bolt, and watch the lift. Repeat around for every foot.
• Use the machine manufacturer's requirements as the threshold. As an order-of-magnitude guide: a lift of a few hundredths of a millimetre is already reason to redo the shim pack.
• Use solid, one-piece stainless-steel shims covering the full area of the foot, no more than three or four sheets per pack.
• Disconnect the pipe flange and see whether the nozzle moves. If it does, the strain has to be relieved before alignment.
• Check the frame, anchor bolts, and grout: cracks, crumbly concrete, delamination, loosened bolts.

No dial indicator on hand? Work with a vibration meter wherever it's safe to do so: loosen one bolt at a time on the running machine and watch the 1x amplitude and phase. A noticeable reaction to a specific foot points to that foot. The method is cruder than using a dial indicator, but it doesn't require a shutdown.

How Geometry Is Checked and What Tolerances Look Like

Every method solves the same task: find the relative position of the two rotation axes and calculate how much to shim under each foot and how far to shift the machine sideways. They differ in accuracy, speed, and how many mistakes you manage to make along the way.

• Rotate both shafts together, not one at a time. Otherwise you're measuring coupling-half runout, not axis position.
• Take readings at four positions: 0, 90, 180, and 270°. Check consistency: the sum of the top and bottom readings should match the sum of the left and right readings.
• Measure bar sag separately, on a length of pipe, and subtract it from the readings. A long cantilevered bar deceives you more than it seems.
• Tighten the bolts to the specified torque before every reading. A half-tightened foot gives you nice-looking but useless numbers.
• Align in the thermal state the tolerance is specified for, or use cold offsets that compensate for the calculated thermal growth.
• Set the axial gap between the coupling halves to the coupling manufacturer's requirement. This is the number people forget most often.

A note on tolerances. There's no single number that fits every machine, and any table you find online is only a guideline. The tighter the tolerance, the higher the speed: going from 1500 to 3000 rpm roughly halves it. It also depends on the coupling type and the distance between its flexible planes: a coupling with a long spacer allows more offset, but the angle is calculated over the spacer length. For machines in the 1500–3000 rpm range, the order of magnitude is: axis offset measured in hundredths of a millimetre, angle in hundredths of a millimetre per 100 mm. Take exact values from the machine and coupling documentation — they matter more than any industry table. For contractual acceptance, record the measurement method, the machine's thermal state, and the base length the angle was calculated over.

Two Misconceptions: "The Flexible Coupling Will Absorb It" and "Let's Just Balance It"

The first misconception sounds logical. The coupling is flexible, that's what it's there for. Except it compensates for misalignment in the only way available to it: by deforming. And deformation of a flexible element is exactly the force that ends up in the bearings, shaft, and seals. A stiffer element produces a larger force for the same offset. A softer one produces a smaller force, but heats up and fatigues faster itself. There's no free option at this fork.

A coupling catalogue's "allowable misalignment" is the survival limit of the coupling itself — not a vibration criterion, not a bearing-life criterion, not a seal criterion. A unit that sits comfortably within the catalogue tolerance can still be steadily eating through its drive-end bearing. A rigid flanged coupling shows the opposite extreme: 2x barely rises, because there's nothing left to deform, and the entire force goes straight into the bearings and shaft.

The second misconception costs more. At running speed, an instrument sees a single combined 1x vector, made up of everything acting once per revolution. It doesn't distinguish, and can't distinguish, between genuine rotor imbalance and the contribution from misalignment. The weight the software calculates from a trial run will zero out the combined vector — meaning it will also compensate for a force that isn't its own. Vibration drops, a report gets written, everyone goes home.

• When someone later properly aligns the unit, the misalignment's contribution disappears, but the weight stays. It then starts acting as a genuine, physically added imbalance. Vibration ends up higher than it was before any of this work started.
• The misalignment hasn't gone anywhere. The coupling, seals, and drive-end bearing keep wearing out; you've just removed the indicator from the dashboard.
• The result doesn't hold. Misalignment shifts with warm-up and load, so the compensation was tuned for one thermal state and falls apart in another.
• If the correction was made by drilling, the mistake is irreversible. You can't weld the metal back on, and on a thin-walled impeller the holes also reduce strength.
• Imbalance can also shift between correction planes (the rotor cross-sections where corrective weights are placed): you improve one bearing and make the other one worse.

A practical tip: if the unit was already balanced before you got to it, and the misalignment was never corrected, find and remove the previously installed weights before aligning. Otherwise, after a proper alignment you'll end up with higher vibration and be hunting for its cause all over again. We cover the vector breakdown of this case in a separate article on imbalance versus misalignment.

Work Sequence: From the Foundation to the Verification Measurement

The sequence isn't arbitrary. Each step depends on the one before it, and if you swap two steps, you'll end up compensating one against the other and going in circles.

1. Step 1

   Measure and Inspect Before Touching Anything

   Take the baseline set at the running condition: overall vibration in mm/s RMS, 1x amplitude and phase, the 2x-to-1x ratio, the axial component at both bearing housings next to the coupling. Inspect the coupling, guard, feet, seals, and grout. These numbers and photos will later be the only proof that the work actually changed anything.

2. Step 2

   Foundation, Frame, Anchor Bolts, Fasteners

   Cracks, crumbly concrete, delaminated grout, loosened anchor bolts. As long as the foundation has a life of its own, the alignment won't hold for more than a few shifts. This is the cheapest step, and it's also the one that most often resolves the whole issue on its own.

3. Step 3

   Relieve Strain from Piping and Connections

   Disconnect the flanges and check whether the nozzles move. Strain from a rigidly connected pipe works exactly like soft foot, only it's harder to spot. While you're at it, check cable glands and rigidly attached guards too.

4. Step 4

   Soft Foot

   Check every foot with a dial indicator while the others stay tightened. Rebuild the shim pack with solid stainless sheets. Recheck after reshimming. As long as the casing distorts under tightening, there's no point going further.

5. Step 5

   Thermal Offsets and Coupling-Half Axial Gap

   Determine which thermal state you're aligning for. Set cold offsets that compensate for the calculated thermal growth, or plan a check on the warmed-up machine. Set the axial gap between the coupling halves to the coupling manufacturer's requirement.

6. Step 6

   Shaft Alignment in Two Planes

   Work with the method you've chosen: reverse indicators or a laser system. Bring all four numbers within tolerance, not just the one that's easiest to fix. Replace a worn flexible element, damaged pins, or bushings now, rather than putting it off until the next scheduled overhaul.

7. Step 7

   Warm-Up and Re-Measurement

   Bring the machine up to running condition, let it warm up, and take the same set of readings as in Step 1. Often this is where it all ends: 2x and the axial reading have dropped, the level is normal, there's nothing left to balance. Don't skip this measurement — it's exactly what separates necessary work from unnecessary work.

8. Step 8

   Balancing Decision and Report

   If 1x is still dominant and above the target level, in-situ rotor balancing is now appropriate. Record the measurement points and directions, the operating condition, the speed, and the initial and residual values in the report. Repeat the measurement two to four weeks later: that's precisely what distinguishes a lasting result from a temporary one.

If overall vibration is still much higher than 1x after alignment, keep looking: bearings, rubbing, hydraulics, looseness, support resonance. We have separate material on reading the spectrum and on situations where balancing won't help.

If There's No Time to Work It Out On-Site

You can separate misalignment from imbalance yourself. It takes a two-channel instrument with phase measurement, half an hour, and discipline in sensor placement. That's exactly what we built Balanset-1A for: two accelerometers, a laser phase-and-speed sensor using a reflective tape mark, a two-channel USB module with preamplifiers, integrators, and an ADC, and Windows software. The device shows overall vibration and 1x separately, amplitude and phase on both channels at once, speed, the time waveform, and the FFT spectrum — and then it runs single- and two-plane balancing using the influence coefficient method, calculates the tolerance by G class, and writes up a report.

Let's be honest about the device's limits. Balanset-1A does not measure alignment geometry. It shows the signs, helps you form a hypothesis, and confirms the result of the work with before-and-after numbers. For the geometry itself you need dial indicators or a laser alignment system.

When there's no time to deal with it yourself, we come out. Our engineers design and manufacture Balanset instruments and use them to do field balancing themselves, so you get the measurement and the work from the same people. The sequence on site is the same one described above: first measurement and separating out the causes, then mechanics, soft foot, and shaft alignment, then rotor balancing in its own bearings if needed, and finally a verification run and a report. If the measurement shows balancing isn't needed, we'll say so. We also provide consulting support on the instrument and the method: where to place sensors on your machine, which direction to choose, how to read the phases you get.

• Measurement at the bearing housings on both sides of the coupling: overall vibration, 1x, 2x, axial component, phase.
• Separating the causes: misalignment, imbalance, soft foot, looseness, resonance, bearings.
• In-situ rotor balancing in one or two planes, including splitting the weight across blades and fixed positions and calculating drilled-hole corrections.
• Tolerance calculation by G class and overall-vibration assessment by zone, with the applicable part of the standard cited.
• A report with measurement points, operating condition, and initial and residual values.

The applicable part and edition of the standard for your machine needs to be checked separately. The numbers, classes, and zones in our materials are a working guideline, not a ready-made acceptance criterion.

Sources: Balanset-1A manufacturer specification · Balanset-1A operation manual · ISO 20816-1:2016 · ISO 21940-11:2016

Frequently asked questions

Related content