Shaft Alignment and Rotor Balancing: Fixing the Two Faults Behind Most Machine Vibration

Walk any plant with a vibration analyser and the findings list ends up dominated by the same two entries. Not exotic resonances, not cracked shafts, not oil whirl. Misalignment and unbalance. Between them they account for the majority of rotating equipment vibration problems we see in the field, and both are entirely correctable with hand tools, shims and arithmetic.

That is the good news. The bad news is that both faults are routinely corrected badly. Machines get "aligned" with a straightedge across the coupling and a squint. Fans get balanced by welding on whatever washer was in the fitter's pocket. The vibration drops a little, everyone goes home, and the bearings continue dying at three-year intervals while the plant wonders why.

We covered how to recognise these faults in a spectrum in our guide to vibration analysis for rotating equipment: unbalance lives at 1× running speed, misalignment at 2× with high axial content. This article is the other half of the job. What each fault physically is, what it does to bearings, seals and couplings, how the correction work is actually done, what tolerances to hold at what speed, and why an alignment that was perfect on Saturday is wrong by Monday once the machine is hot.

Shaft alignment and field balancing are not glamorous. They are also the highest-return maintenance work most plants can do, and this is the working knowledge needed to do them properly or to judge whether a contractor has.

What Misalignment Actually Is

Two coupled machines are aligned when their shaft rotational centrelines form a single straight line at operating conditions. Every real installation misses that ideal in two distinct ways, and the distinction matters because they are measured and corrected differently.

Parallel misalignment, also called offset, means the two centrelines are parallel but displaced from each other, like two lanes of a road. It is measured in millimetres or thousandths of an inch at the coupling centre.

Angular misalignment means the centrelines meet at an angle, like a road junction. It is measured as a gap difference across the coupling face, expressed per 100 mm of coupling diameter.

A real machine has both at once, and has them in two planes, vertical and horizontal. So a complete alignment measurement is four numbers: vertical offset, vertical angularity, horizontal offset, horizontal angularity. Anyone who reports your alignment as one number has not measured it.

The coupling deserves a word here. A flexible coupling exists to transmit torque while tolerating small residual misalignment. It is not a licence to skip alignment. A flexible element absorbing misalignment does so by flexing once per revolution, which at 3,000 rpm is 4.3 million cycles a day. The coupling's rated misalignment capacity is the limit at which it survives, not the limit at which the machines behind it do.

What It Does to Bearings, Seals and Couplings

Misalignment does its damage through force. When the centrelines do not line up, the coupling drags each shaft toward the other's axis once per revolution, and that reaction load passes straight into the nearest bearings, on top of the loads they were sized for.

The arithmetic is unforgiving. The standard bearing life calculation in ISO 281 makes basic rating life proportional to the cube of the load ratio for ball bearings. Run that in reverse: increase the load on a ball bearing by 25% and its calculated life roughly halves. Double the load and the life falls to about one eighth. A misalignment that adds a few hundred newtons of cyclic load to a small motor bearing is the difference between a bearing that lasts eight years and one that lasts one.

Heat is the visible symptom. The flexing coupling element and the overloaded bearings both dissipate the extra work as temperature, which is why a misaligned machine train so often shows a hot coupling guard and warm bearing housings on a thermal survey before it shows anything dramatic on a vibration route. There is an energy cost too; in our experience it is small, a low single-digit percentage at worst, and it is the least of your problems next to the bearing life.

Seals suffer differently. Misalignment deflects the shaft, and a mechanical seal is two lapped faces held together within microns. A shaft wobbling at the seal faces opens and closes that gap once per revolution, and the seal responds by leaking, sometimes within weeks of an otherwise tidy overhaul. On pumps, where the mechanical seal is usually the most frequently failing and most expensive consumable, a striking share of "seal problems" are alignment problems wearing a disguise.

Couplings themselves fail last and loudest: elastomeric elements crumble and shed black dust (find that dust inside a coupling guard and you have found a misaligned machine), gear couplings wear their teeth and lock up, disc packs crack. The U.S. Department of Energy's motor systems sourcebook lists misalignment and vibration among the recurring causes of chronic motor failure, and notes a cause we see constantly on site: pipework that does not quite meet the pump flanges, force-fitted during installation, dragging the pump off its alignment through pipe strain.

Bearings dying young, seals leaking, couplings shedding dust? Our maintenance and asset management team diagnoses the root cause before replacing the casualties.

The three casualties of misalignment, in the order they usually appear. Black dust inside a coupling guard is a diagnosis, not a housekeeping problem. Source: MIMAH engineering analysis; U.S. Department of Energy motor systems sourcebook.

The Laser Alignment Procedure, Done Properly

Laser shaft alignment replaced dial indicators for a simple reason: a laser and detector spanning the coupling measure the relative shaft positions to hundredths of a millimetre without bracket sag, reading parallax or arithmetic slips, and the computer converts the readings directly into shim and move corrections at the machine feet. The instrument, however, is the easy part. The job is won or lost in the preparation.

First, deal with what the laser cannot fix. Check the foundation and grout for cracks, check hold-down bolts, and check pipe strain by loosening the pump flange bolts and watching whether the pipe springs away. If the pipework moves the machine when disconnected, no alignment will survive reconnection.

Second, find the soft foot. Soft foot is the machine equivalent of a wobbly restaurant table: one or more feet do not sit flat on the base, so tightening the bolts distorts the casing, which moves bearings, distorts the shaft and makes alignment readings unrepeatable. The check is quick. With the machine loose-coupled, mount the measuring system, tighten all feet, then loosen one bolt at a time and watch the movement. More than about 0.05 mm of lift at any foot needs correcting with clean, full-contact shims before alignment proper begins. Use a few thick shims rather than a stack of thin ones; a pile of rusty shim stock is itself a spring, the so-called squishy foot.

Third, measure and correct. Rough-align the machine so the laser stays on the detector, rotate the shafts through the measurement sweep, and let the system compute the four numbers. Correct the vertical plane first with shims under the feet, because horizontal moves do not disturb it, then pull the horizontal plane in with jacking bolts. Re-measure after every move, and re-measure once more after final bolt torque, because torquing moves machines. Record the as-left values; an alignment that is not documented did not happen.

How good is good enough? In the absence of a manufacturer's figure, the general-purpose field tolerances used across industry tighten with speed, because the forces rise with it. As typical working values for short flexible couplings: at 1,500 rpm, offset within about 0.09 mm and angularity within 0.07 mm per 100 mm; at 3,000 rpm, offset within about 0.05 mm and angularity 0.05 mm per 100 mm; above 6,000 rpm, offset within 0.03 mm and tighter still on machines with journal bearings. A straightedge across the coupling rim can perhaps detect half a millimetre. That is ten times coarser than what a 3,000 rpm machine needs, which is why straightedge alignment is a rough-in method, not an alignment.

Aligned Cold, Crooked Hot

Here is the trap that catches even careful crews: alignment is defined at operating temperature, but it is measured on a cold machine. Metal grows when it heats, and machines do not all heat equally.

The physics is one line. Steel grows by roughly 12 microns per metre of height for every degree Celsius of temperature rise. Take a hot-service pump whose centreline sits 400 mm above its feet and whose casing runs 100°C above ambient once on line: the pump centreline rises almost 0.5 mm. Its motor, warmed only by its own losses, may rise a tenth of that. A train aligned perfectly cold is therefore half a millimetre misaligned hot, which is ten times the tolerance we just set for a 3,000 rpm machine.

The correction is deliberate cold misalignment. The manufacturer's data sheet often states cold offset targets, typically setting the cold machine low or high so that thermal growth carries it into line. Where no figures exist, they can be calculated from centreline heights and expected temperature rises, or measured directly by taking an alignment reading immediately after shutdown while the machine is still hot and comparing it with the cold reading. On critical hot machines, steam turbine trains above all, this hot-alignment verification is not optional refinement. It is the alignment.

The same thinking applies in reverse to machines that run cold, and to machines whose growth is horizontal because of asymmetric casing temperatures. The question to ask of any alignment report is always the same: aligned to what target, and where did the target come from?

Why a machine aligned cold runs crooked hot, and why deliberate cold offset exists. Steel grows roughly 12 microns per metre of height per degree Celsius. Source: MIMAH engineering analysis.

Unbalance: Static, Couple and Dynamic

Unbalance is simpler to picture than misalignment and subtler to correct. A rotor is balanced when its mass is distributed so that its centre of mass lies on the rotational axis and its principal inertia axis coincides with it. Casting voids, machining tolerances, corrosion, erosion, product build-up and lost balance weights all break that condition. The framework and vocabulary are set out in ISO 21940-1, and the distinctions matter because they decide how many correction planes you need.

Static unbalance means the mass axis is displaced parallel to the shaft axis: one net heavy spot. Set such a rotor on knife edges and it rolls until the heavy spot hangs at the bottom, hence the name. It is corrected in a single plane.

Couple unbalance means the mass axis crosses the shaft axis at the centre of gravity: two equal heavy spots at opposite ends, 180 degrees apart. The rotor balances perfectly on knife edges, then shakes violently at speed as the two ends whirl in opposition. It can only be corrected in two planes.

Dynamic unbalance is the general case, a combination of both, and it is what every real rotor has. The practical question is whether the couple component is big enough to matter, which broadly follows rotor geometry: narrow rotors such as single fan wheels and pump impellers usually behave statically and accept single-plane correction, while long rotors, multi-stage elements and rolls need two planes.

Why unbalance punishes speed so brutally is worth one calculation. The centrifugal force from an unbalance mass rises with the square of speed. Ten grams sitting at 300 mm radius produces about 74 N of rotating force at 1,500 rpm, and about 296 N at 3,000 rpm. Double the speed, four times the force, which the bearings absorb once per revolution, forever. This is why a machine sped up on a variable speed drive sometimes crosses from smooth to rough with no mechanical change at all.

Field Balancing and the ISO 21940-11 Grades

Balancing in situ is routine, provided the diagnosis is right. The influence coefficient method underneath every modern field balancer works in three steps: measure the original vibration amplitude and phase at the bearing, attach a known trial weight and measure the change, then compute from those two vectors the correction weight and angle that cancel the original. One plane for static behaviour, two planes with cross-effect terms for dynamic. A competent job on a fan is a few hours including the runs; the usual sizing rule for the trial weight is that its centrifugal force should be around ten percent of the rotor's static weight, enough to move the vibration measurably without risking the machine.

How much residual unbalance is acceptable is not a matter of taste. ISO 21940-11 defines balance quality grades, the familiar G numbers, where the grade is the permissible velocity of the rotor's centre of gravity in millimetres per second at operating speed. Smaller is finer. Grade G 6.3 covers most general machinery: fans, pump impellers, electric armatures, general machine parts. G 2.5 applies to gas and steam turbines, turbo-compressors and machine tool drives. G 1 covers grinding machine drives, while at the loose end G 16 covers cardan shafts and agricultural machinery and G 40 covers car wheels.

The grade converts to a hard number with one formula: permissible residual unbalance per kilogram of rotor mass, in gram-millimetres, equals 9,549 times the G number divided by the speed in rpm. So a 200 kg fan rotor running at 1,500 rpm balanced to G 6.3 is allowed 9,549 × 6.3 ÷ 1,500, about 40 g·mm per kg, roughly 8,000 g·mm in total, which at a 500 mm blade-tip radius is a 16 gram tolerance on the whole rotor. Halve it for the same rotor at 3,000 rpm. Specifying "balance grade G 6.3 at operating speed" on a repair order, and asking for the residual unbalance certificate, is a one-line habit that removes all argument about whether a rotor came back balanced.

One caution from the field: balancing treats the symptom of mass distribution, not its cause. A fan that needs re-balancing every few months has an underlying problem, usually build-up, erosion, or cracking, and the honest fix is finding it, not scheduling quarterly balancing forever. Whether the vibration is truly at 1× and behaving like unbalance is a diagnosis question, covered in the vibration analysis guide.

Chasing a vibration problem that keeps coming back? Our industrial engineering services cover alignment, field balancing and root cause analysis on turbines, pumps, fans and generators.

Converting an ISO 21940-11 balance quality grade into a number a balancing machine can work to. Source: ISO 21940-11:2016.

The Economics of Half a Millimetre

Set the two corrections against what they prevent and the case makes itself.

A precision laser alignment on a coupled pump set is a few hours of skilled work. The bearing failure it prevents costs the bearing, the seal that usually dies with it, the labour, and the downtime, which on an unspared machine is measured in production, not in parts. Because bearing life falls with the cube of load, the return is steeply non-linear: the difference between a sloppy 0.3 mm alignment and a precise 0.05 mm one is not 15% bearing life, it can be the difference between annual failures and a bearing that quietly reaches its design life. The same logic prices balancing: a few hours with a field balancer against bearings hammered by hundreds of newtons of entirely optional rotating force.

There is also a scheduling dividend. Machines aligned and balanced to known tolerances, with as-left records, trend flat on the vibration route, which means the monitoring programme spends its attention on genuine developing faults instead of background noise. Precision work at overhaul is what makes condition-based maintenance cheap to run; a plant full of rough-aligned machines keeps its analysts busy documenting problems that should never have existed. Vibration limits for judging the result are set out in ISO 20816-1, and a freshly overhauled machine should sit comfortably in the newly commissioned zone, not scrape into the acceptable one.

Frequently Asked Questions

A flexible coupling tolerates misalignment, so why bother with precision alignment? The coupling tolerates it; the bearings and seals pay for it. The coupling's rated misalignment is the limit at which the coupling survives, and it flexes at that duty millions of cycles a day while feeding the reaction forces into both machines. Align to the machine tolerance, not the coupling rating.

Do new or newly installed machines need aligning? Yes, always. Factory alignment of a packaged set does not survive transport, lifting, grouting and pipework connection. Force-fitted pipe flanges in particular can drag a pump visibly off alignment after the fitters have gone. Alignment is a commissioning activity, checked again after the first period at operating temperature.

What balance grade should I specify? For general plant machinery, fans, pump impellers and motor armatures, G 6.3 at operating speed per ISO 21940-11 is the standard specification and G 2.5 is the precision option. Turbine and turbo-compressor rotors are G 2.5 territory by default. Specify the grade and the service speed, and ask for the residual unbalance report; a grade without a speed is meaningless because the tolerance halves every time the speed doubles.

We balanced the fan and the vibration came back in three months. Why? Because the unbalance had a cause that is still active. Build-up on blades accumulates and sheds unevenly, erosion and corrosion remove metal unevenly, and a cracked blade or weld moves progressively. Repeated re-balancing of the same rotor is a symptom you should treat as an inspection trigger, not a maintenance strategy.

Are dial indicators obsolete for alignment? No. The reverse dial method in skilled hands achieves precision-class results and remains a sound fallback and cross-check. Laser systems earn their cost through speed, repeatability, built-in tolerance judgement and documentation, and on jobs with limited rotation, long spans or vertical machines they are dramatically easier. What is obsolete is the straightedge as a final method.

When can a rotor not be balanced in place? When the vibration is not actually unbalance (balancing a misaligned or cracked machine makes things worse), when the rotor cannot safely carry trial weights, when correction planes are inaccessible without dismantling, or when the rotor is flexible, meaning it runs above a critical speed and bends as it rotates. Flexible rotors, large turbine elements among them, are balanced at speed in a shop facility under ISO 21940 procedures.

Straight Shafts Are Cheap Insurance

Misalignment and unbalance are unusual among machine faults in that they are fully preventable at commissioning and fully correctable in service, with tolerances published, procedures standardised and instruments readily available. Most of the bearing and seal failures they cause are therefore voluntary. The plants that stop paying for them are the ones that treat precision alignment and specified balance grades as normal work, written into every overhaul scope and verified with numbers, not adjectives.

That standard of work is what our team applies across four decades of turbine, generator and rotating equipment experience, from steam turbine overhauls in Nigeria to root cause analysis at White Nile Sugar in Sudan, where the question "why does this machine keep eating bearings" has had an answer measured in tenths of a millimetre more often than not.

Have a machine train that eats bearings, leaks at the seal, or shakes more every year? Talk to our engineering team. We will measure the alignment and balance condition properly, correct them to written tolerances, and leave you the as-left records to hold the next contractor to.