A pump coupling failed every three months. It was a spider (jaw) coupling, 55 mm size, connecting a 15 kW motor to a centrifugal pump. The rubber spider looked chewed on the edges. The customer replaced the spider, aligned the shafts “by eye” with a straightedge, and the new spider lasted another three months. The pattern was steady. The pump bearings were fine. The motor bearings were fine. The coupling was the only thing that died, which told us the coupling was carrying the misalignment that the shafts should never have seen.
What a flexible coupling actually tolerates
A jaw coupling is flexible in a specific way. It allows angular misalignment up to about 1 degree, parallel offset up to about 0.2 mm, and axial movement of 2-3 mm. Those are the design limits. The rubber spider deforms to accommodate that movement, and it’s meant to run slightly warm because it’s working. What kills it is running beyond those limits. When a coupling is pushed past its rated misalignment, the spider gets loaded unevenly. One lobe carries the whole torque while the others barely touch. The rubber gets hot, the edge wears, and it fails. The failure pattern on the old spider — chewed edges, one lobe worn flat — was the classic signature of angular misalignment beyond spec.
The straightedge check was the problem. A straightedge on the coupling hubs tells you angular alignment but it’s insensitive to parallel offset. Two shafts can look perfectly straight on a straightedge and still be offset by 0.5 mm in parallel, which is two and a half times the coupling’s rated limit. The motor had shifted on its base over time. The baseplate was warped, and the feet were shimmed with 0.5 mm of paint buildup. The straightedge said fine. The coupling said otherwise.
The fix was a dial indicator and half a day
We pulled the coupling apart and mounted a dial indicator on the motor hub, sweeping the pump hub face and rim. Face reading: 0.15 mm of angular misalignment over the sweep diameter, about 0.4 degrees. Rim reading: 0.55 mm of parallel offset. Both beyond spec. The motor feet were re-shimmed. The angular error went to 0.02 mm over the face, and the offset to 0.05 mm. That’s within the coupling’s rating with a good margin.
The new spider ran for 14 months, at which point it was replaced on the maintenance schedule with no visible wear. The quarterly failure pattern was gone. The cost of the fix: a dial indicator borrowed from the tool crib, some shim stock, and four hours of two people’s time. The cost of the old pattern: a $30 spider every three months plus a couple of hours of unplanned downtime each time, and a pump that was running with extra bearing loads it didn’t need.
The alignment spec that matters
For a jaw coupling up to 55 mm, the practical spec is 0.1 mm TIR on the face (angular) and 0.1 mm TIR on the rim (parallel), measured with a dial indicator. That’s tighter than the coupling’s own limit but it leaves margin for thermal growth and baseplate flex when the motor heats up. If you’re doing a motor alignment with a laser system, the spec is usually written as 0.05 mm per 100 mm of coupling diameter, which works out about the same.
The standard excuse for skipping the dial indicator is “it’s a flexible coupling, it absorbs it.” That’s the exact reasoning that kills the spider. The flexible coupling is a wear item, and it’s designed to wear slowly at its rated misalignment. Running it beyond rating makes it wear fast. The coupling is a warning system. When it fails regularly, it’s not the coupling that’s defective. It’s the alignment that’s defective, and the coupling is just the cheapest part in the line, so it dies first.
Why the pump bearings were fine
An interesting part of this failure: the pump and motor bearings were healthy even though the coupling was being hammered. That’s normal. The bearings are stiffer than the coupling, so under moderate misalignment the coupling absorbs most of the movement. The bearing loads go up, but not enough to fail quickly. The coupling, being the softest component, takes the punishment. That’s why the coupling dies first. It’s also why the failure is easy to misread — the bearings look fine, so the coupling brand gets blamed.
The chewed spider wasn’t a defective part. It was a coupling running at 0.4 degrees of angular and 0.55 mm of parallel misalignment — way beyond its rating. A straightedge can’t see parallel offset. Use a dial indicator or laser, keep face and rim TIR under 0.1 mm, and the coupling stops being the consumable that fails quarterly. Flexible doesn’t mean forgiving.
