AGMA 9002: bores and keyways for flexible couplings

1. What the standard covers

That narrow scope sounds like faint praise until you consider the alternative. Without a dimensional standard every coupling hub is a bespoke part: a replacement means going back to the original maker with the original drawing, and a shaft reground on site at two in the morning has nothing to be measured against.

AGMA 9002 exists to prevent that, and it works as a pair with AGMA 9008. The first standardises the interface between the shaft and the hub. The second standardises the flange dimensions where the two halves of the coupling meet. Together they turn a coupling from a bespoke assembly into something closer to a commodity — one whose parts can be sourced, replaced and interchanged years after the original purchase order has been archived.

1.1 What it stays silent on

The scope is narrow by design, and everything outside it is genuinely absent rather than merely implied. AGMA 9002 says nothing about torque rating, because rating depends on tooth geometry, material and heat treatment the document does not address. Nothing about balance procedure or residual limits. Nothing about overspeed testing, material traceability or the documentation package. Nothing about how much misalignment the coupling will take.

Reading it as a quality specification is the most common way engineers misuse it. A supplier who answers a technical query with “our couplings are AGMA 9002 compliant” has told you the bores are dimensioned to a recognised class, and has told you nothing else at all.

2. Getting torque from shaft to hub

Torque has to cross from the shaft into the hub somehow, and only two mechanisms are available: a key working in shear and bearing, or friction generated by an interference fit. Most industrial couplings use the first. Most high-torque and reversing applications end up needing the second. A great many use both at once without anyone having decided which is actually carrying the load.

2.1 Clearance fit and a key

The default across general industrial service. The hub slides on by hand or with light persuasion, a parallel key transmits the torque, and a setscrew or retaining plate stops the hub walking off axially. Quick to fit, quick to remove, forgiving of a shaft that has seen better days.

The weakness is load reversal. A clearance fit lets the hub rock in the bore by the width of that clearance, and every reversal drives the key against one flank of the keyway and then the other. The result is a slow hammering that rounds the keyway corners, fretts the bore and eventually shears or rolls the key. Couplings on reciprocating compressors and reversing mill drives fail this way routinely — and the report almost always records a key failure, when the real fault was specifying a clearance fit for a duty that could not tolerate one.

2.2 Interference fit

An interference fit removes the clearance entirely. The bore is cut smaller than the shaft, the hub is expanded thermally or hydraulically to get it on, and once it cools the elastic recovery grips hard enough that friction carries a substantial share of the torque. Where the interference is heavy enough the joint is keyless and friction carries all of it.

The trade is assembly and removal. A hub fitted with 0.3 mm of interference on a 300 mm bore is not coming off with a three-legged puller and optimism; it needs the same heat or hydraulic pressure that put it on. Where that equipment is not available at site, an interference fit turns a two-hour coupling change into a two-day one — a consideration that belongs in the specification, not in the outage report.

4. Getting the hub on

Choosing an interference is a balance between grip and stress. Too little and the joint slips under peak torque, polishing the bore and destroying both hub and shaft in the process. Too much and hoop stress in the hub approaches the yield strength of the material, and the hub can crack — sometimes on assembly, sometimes months later from a notch nobody noticed.

Common practice for coupling hubs runs from roughly 0.5 to 1.5 micrometres of diametral interference per millimetre of shaft diameter: the lower end for light duty, the upper end where a keyless joint carries full torque on friction alone. Those are practice figures, not clause values. The governing standard and the coupling manufacturer both have a view, and on a critical machine the two should be made to agree before anything is cut.

4.1 Heating a hub

The ceiling on that number is the heat treatment of the hub, not the arithmetic. Hub teeth are commonly through-hardened or surface-hardened, and heating past the tempering temperature of the material draws that hardness back out. The teeth look identical afterwards and wear out in a fraction of their expected life. This is why induction heaters with controlled temperature limits have largely displaced oxy-acetylene torches for the job, and why any procedure containing the phrase “heat it until it goes blue” should be stopped where it stands.

5. The keyway nobody budgets for

A keyway is a stress raiser cut into the most highly stressed region of a rotating shaft, and it gets far less attention than it earns. A profiled keyway cut with an end mill leaves a sharp-cornered pocket that concentrates bending stress by a factor of roughly two to two and a half. A sled-runner keyway cut with a side mill, which runs out gradually at each end, is meaningfully gentler at around one and a half.

On a shaft that is fatigue-limited rather than statically limited, that difference decides whether it lasts the life of the machine or cracks from the keyway corner. The keyway is also where fretting debris and corrosion collect — and a keyway recut oversize on site to clean up damage leaves a shaft with materially less remaining life than the drawing implies.

6. What to put on the drawing

More than a bore diameter. State the nominal bore and the tolerance class, so the fit is unambiguous rather than implied. State the keyway dimensions and the tolerance on width — a key that is a push fit in one keyway and a hammer fit in the other loads unevenly from the first revolution.

Where an interference is intended, state it explicitly in millimetres or thousandths rather than leaving it to be inferred from a class designation the reader may not have to hand. Note the required assembly method and its temperature limit. And name the standard with its revision: standards get revised, and a drawing citing a document without a year cites nothing enforceable at all.