Why a Flexible Coupling Does Not Fix Misalignment

Why this matters

"It has a flexible coupling, it will take it up" has killed more bearings than any other sentence in a mechanical room. The coupling catalog prints a misalignment number, the shop treats that number as permission, and eighteen months later two bearings on two different machines are on their second replacement while the coupling that "handles it" looks brand new. The coupling looking fine is not evidence that the joint is fine. It is evidence that the coupling is the one part in the assembly designed to be deflected, and that everything on either side of it is quietly paying for the deflection.

Before you get near the joint

Open and lock the motor disconnect and prove the terminals dead with a live-dead-live check against a known live source before and after (29 CFR 1910.333(b)(2); NFPA 70E-2021, 120.5). Then work the mechanical side under 29 CFR 1910.147: confirm the rotor has stopped coasting, block it against rotation, and isolate, lock, relieve and drain any pressurized or hot line into the machine before a foot bolt moves. When you rotate a shaft by hand to take readings, turn it with a bar on a designated barring point or on the shaft itself with the guard hardware clear, keep fingers out of the gap between the hubs, and keep the rotor blocked whenever a hand is inside that gap rather than relying on the machine staying still. Machine surfaces that ran above about 140 F burn on contact, so read them with a non-contact infrared thermometer and let them cool before you work bare-handed.

The deflection is imposed by geometry, not by the coupling

This is the whole mechanism in one sentence. Two shaft centerlines that do not coincide are separated by a distance and an angle that exist whether the coupling is a solid flange or a rubber tire. The coupling does not get to choose how far it deflects; the machines choose that when they are bolted down. Flexibility changes only one thing: how much force it takes to hold that deflection.

Force equals stiffness times deflection. A flexible element has a low stiffness, so it produces a small force for a given offset. Small is not zero, and the force does not disappear - it goes somewhere. It is reacted at both ends of the coupling, in the bearings of both machines, once per revolution, for every revolution the set runs.

So the honest statement of what a flexible coupling does: it converts a geometric error into a modest cyclic force instead of an enormous one, and it survives the conversion. It does not remove the error, and it does not shield the bearings from the force it creates.

The catalog number is an element survival limit

The misalignment figure in a coupling catalog answers one question: how far can this element be deflected, continuously, before the element itself fails. It is a rating for a rubber spider, a steel grid or a disc pack. It says nothing about the reaction force that deflection puts into a bearing, and nothing about whether the machines can live with it.

The machines' alignment tolerance is a different number entirely, and it comes from somewhere else: the operating speed, the bearing arrangement, and the coupling's own stiffness. It is routinely several times tighter than the coupling's capacity. Get the coupling manufacturer's published alignment tolerance for the installation, not the element rating, and where the machine builder publishes one for the driven equipment, the tighter of the two governs.

Flexible is a spectrum, and a stiff flexible coupling exports more

Do not treat "flexible" as a single class. For the same imposed offset, the reaction force is proportional to the element's stiffness, and stiffness varies widely across families: a metal disc pack is radially far stiffer than an elastomer element in shear, and a grid or gear coupling accommodates by sliding rather than by elastic deflection, which puts the exported force under the control of the friction at the sliding surfaces instead of a spring rate. Two couplings can carry identical catalog offset ratings and export forces that differ several-fold at the same misalignment.

The practical consequence is one shops rarely anticipate: changing coupling type on an existing set, without changing the alignment, changes the bearing load. Moving from an elastomer element to a disc pack on a set that has always run a few mils out is a real increase in exported force, delivered to bearings that were coping.

Where the offset actually goes: the flex-plane spacing

A flexible coupling accommodates parallel offset by taking an angle at each of its two flex planes. For small angles, the angle at each plane is approximately the offset divided by the distance between the flex planes, in radians. That geometry, not the element material, is why spacer couplings tolerate more offset than close-coupled ones, and it is why a catalog offset number for a spacer coupling is meaningless without the spacer length attached.

Run the numbers on a single 12 mil parallel offset, which is 0.012 inch:

• Across a spacer coupling with 7 inches between flex planes: 0.012 divided by 7 is 0.001714 radians, which is 0.098 degrees at each plane.
• Across a close-coupled design with 2 inches between flex planes: 0.012 divided by 2 is 0.006 radians, which is 0.344 degrees at each plane.

Same machines, same offset, and the short coupling asks each flex element for 3.5 times the angle. The relationship is a small-angle approximation for pure parallel offset taken up by two flex planes at the ends of the span, and at these angles it is accurate to well under a percent. It stops being a good approximation once angular misalignment is also present, because then the two contributions add or subtract at each plane depending on direction.

One rule, two sets, opposite symptoms

The rule: align to the machines' tolerance, which is set by speed and bearing arrangement, never to the coupling's element capacity. Here is that single rule run against two sets that share the same 12 mil offset.

Set A, the spacer coupling on a 1,780 rpm pump. Each flex plane sees 0.098 degrees against an element rating of, illustratively, 0.25 degrees per plane. The coupling is at roughly 40 percent of its element capacity and will run for years. Nothing in the guard ever looks wrong. The rule still says this set is out, because the coupling capacity was never the constraint. The reaction force is still there, still cycling once per revolution, and it lands on the inboard bearings of both machines. The failure history is what tells you: the motor inboard bearing at fourteen months, the pump inboard at nineteen, both on their second replacement, coupling original. A shop reading the coupling as its evidence concludes the joint is fine and buys bearings forever.

Set B, a close-coupled jaw coupling at the same 12 mils. Each flex plane sees 0.344 degrees against the same illustrative 0.25 degree rating. Now the coupling is over its element limit as well, and it says so: the spider goes soft, backlash appears, crumbs collect in the bottom of the guard. Here the coupling fails first and the shop notices. The trap is what they do next - fit a higher-capacity coupling, because the coupling was the thing that broke. The crumbs stop. The bearings keep dying, on exactly the same schedule as Set A, because the misalignment never moved.

The same rule resolves both sets identically and both symptoms mislead. In A the coupling's good health is read as a clean bill. In B the coupling's failure is read as a coupling problem. In both cases the number that matters is the offset, and in both cases it was 12 mils.

What changes the answer

Speed is the condition that genuinely inverts the tolerance. Alignment tolerances tighten as speed rises, because the exported force cycles more often per hour and because rotor dynamics get less forgiving, so an offset that is unremarkable on a 900 rpm set is a real problem on a 3,570 rpm one. The other genuine inverter is bearing arrangement: a sleeve-bearing machine reacts a coupling force differently from a rolling-element one and brings its own axial-position constraint into the same joint.

What does not change the answer is the coupling's element rating, and that is the point of this article.

How to verify you have not been fooled by a healthy coupling

Two checks, in this order, on a locked-out machine. First, read the guard floor and the element before you read anything else: crumbs, fretting powder or a cracked disc root tells you the coupling is over its own limit, which is a fact about the coupling and not yet a fact about the alignment. Second, and regardless of what the first check showed, measure the actual offset and angle at the coupling and compare them to the published installation tolerance for that coupling at that speed, not to the element rating. If the coupling is clean and the offset is out of tolerance, you are looking at Set A and the bearing history is your evidence. Pull the last two bearing replacements out of the records for both machines before you decide the joint is fine.

References

• 29 CFR 1910.147 for mechanical isolation, stored energy and rotor blocking; 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5, for de-energizing and verifying the motor circuit.
• Coupling manufacturer documentation for the installation alignment tolerance at speed, the element misalignment rating, flex-plane spacing and published stiffness values; these are three different numbers in the same catalog.
• See related: "What Misalignment Does to a Bearing in Mechanical Terms"; "The Coupling Types and What Each One Tolerates"; "How to Tell Angular Misalignment From Parallel Offset".