This is a lightly edited thread from our files, names and one supplier changed, figures unchanged. It shows how a set screw holding-torque question travelled from design to the supplier to production and back, and how the final answer was a compromise nobody loved and everybody shipped.
From: M. Varga, Design — To: R. Ott, Procurement; cc F. Lind, Production
Subject: RE: set screws on the 20 mm coupling hub — do we trust them?
Morning all. The new index drive uses one 20 mm shaft and one coupling hub, and the current drawing bites the shaft with two M6 grub screws 120 degrees apart. The applied torque at the hub is 8.5 N·m with occasional shock to maybe 12, never above that. Before this gets carved into the BOM I would like the holding-torque number we claim, because IÊve seen hub screws walk on smaller jobs. If the published holding torque at the screw tip is 9.4 N·m on a freshly degreased shaft, our 12 N·m shock is already above it, and I would rather add a flat and a shoulder than explain a slipped hub to the customer.
[Inline comments from F. Lind, Production: flat or slot vs no flat changes everything, and the 120 degree split puts one screw into the gap on a two-flat hub — check the drawing before we argue.]
From: R. Ott, Procurement — To: M. Varga; cc F. Lind
Subject: FW: set screw holding torque table from the fastener house
I asked the fastener supplier for their own M6 set screw holding torque on EN1A shafting and on case-hardened shaft, degreased and oiled, and they came back with a table I’ve forwarded. The honest bits: the published 9.4 N·m assumes a harder shaft and a clean point penetration. On our 20 mm soft shaft they would only guarantee 5.5 to 7 on two screws. That és not a complaint about the supplier, it is a complaint about the joint. I would like design to tell us which surface we are biting into, because the table does not cover a shaft with a through-key below the screw.
From: M. Varga, Design — To: R. Ott; cc F. Lind
Subject: RE: set screw holding torque table from the fastener house
Good table, and I owe everyone the drawing detail. The hub runs on a plain 20 h6 shaft, no keyway, ground shaft, and the drawing currently shows a two-flat arrangement, flats 5.8 mm wide, 90 degrees apart intended, but the CAD copy that went to you has them at 120. That is my mistake and it is exactly the kind of detail that turns a fine holding calculation into a slipped hub, so let us settle the geometry first. Two flats, 90 degrees apart, one screw on each flat, avoids the corner problem; with 120 degrees one screw sits on a flat and one on the round and the two screws fight each other in the joint.
The bigger question is whether two flats are even the right answer on a shock-loaded 20 mm shaft. The supplier’s table gives us about 5.5 to 7.0 N·m on the soft shaft with two screws, and our shock is 12. That is a factor of roughly two on the wrong side, and no amount of Loctite rescues a joint where the published number is already half the load. Options I see, in order of preference: one, move the torque take by a plain DIN 6885 key with a pinch collar so the set screws only locate and do not drive; two, keep two flats but step the hub hub bore so the shoulder takes the axial push while the screws locate; three, drop to a 2-screw on 12 mm with a case-hardened shaft, which changes the whole joint. I would like production to confirm which one makes the assembly line happiest, because I can model the key joint today but I cannot model the wrench.
From: R. Ott, Procurement — To: M. Varga; cc F. Lind, quality
Subject: RE: set screw holding torque table from the fastener house
Forwarded your options to the fastener house with the corrected drawing, their reply is pasted below: on a plain 20 h6 shaft with two flats and a light interference, they rate the two-screw joint for torque transmission up to roughly 6 N·m continuous and 8.5 on momentary peak, and they flatly refuse to rate 12. They also note that a flat is only as good as the flat, a wide flat that runs the full hub width gives the screw tip a bigger seating surface, and a spot that is too narrow lets the tip ride the edge and lose a third of the holding. Quality has asked for a formal joint review before the BOM freezes, which I think is reasonable, because the screw that holds our index hub is not a place to be brave.
From: T. Weber, Wälzl fasteners — To: R. Ott
Subject: RE: set screw holding torque table for 20 h6 shaft
Hello Mr. Ott. With the corrected drawing we confirm the figures below for a DIN 916 M6 cup-point set screw into a flat, torque to the tightening value in the last column, degreased mating, shaft 20 h6 EN1A. Two screws at 90 degrees, each on its own flat, together transmit about 5.8 to 6.2 N·m continuous on a 20 mm shaft, and about 8.5 N·m for a momentary transient, assuming the flats are full width and the screws are tightened to the listed torque, M6 to 4.5 N·m. We cannot rate 12 N·m shock through two set screws on this shaft without a key. Where the design needs more, please use a key and let the set screws act as positioners, which is the arrangement we sell the most of in index drives. A 5.8 mm wide flat on a 20 mm shaft gives the cup point a seating width of roughly 4 mm of effective contact, and that is the geometry behind the number we quote.
From: F. Lind, Production — To: M. Varga; cc R. Ott
Subject: RE: set screw holding torque — what the line wants
Design asks what the line wants. Answer: the joint that does not make us clock the hub position. With a two-flat 90 degree set screw design we currently clock the hub to a scribe line, tighten, torque stripe, repeat on a bad day, and a slipped hub on a Saturday is the kind of event that costs a customer call. If a key plus two set screws means the key carries the torque and the screws only stop axial drift, the line wants that, the tooling is a 5 mm broach or a slot mill we already own. I would add one operational requirement: whatever we do, we need the two screws torqued in the right order and to the printed value, and I would put a torque value on the drawing instead of ’tighten firmly’. The 4.5 N·m number from the supplier is a drawing number, not a suggestion.
From: M. Varga, Design — To: R. Ott; cc F. Lind, quality
Subject: RE: joint review — the compromise
After the joint review I would like to propose what we actually agreed, because it is the compromise I said nobody would love. We keep the two-flat 90 degree set screw arrangement for positive axial location and we add a 5 mm key so the shock torque at 12 N·m rides on the key, not on the screw tips. The set screws then only hold the hub from drifting axially, which the supplier’s table covers with margin, and the key carries the 8.5 continuous and the 12 shock with a calculated shear well inside the steel. The drawing will call both, the screws torqued to 4.5 N·m in sequence, a Loctite on the screws, and a torque stripe. The plate arrangement stays, the shoulder carries the axial push from the drive, and nothing about the 20 h6 shaft changes.
From: R. Ott, Procurement — To: M. Varga; cc F. Lind, quality
Subject: RE: joint review — the compromise, closed
Closing this thread with the agreed outcome so nobody has to scroll. The index hub on the 20 h6 shaft will be a 5 mm DIN 6885 key joint carrying all transmitted torque, plus two M6 DIN 916 cup-point set screws at 90 degrees on two flats for axial location, torqued to 4.5 N·m in the sequence stamped on the drawing, Loctite 243 on both, torque stripe over the hub-shaft line. The supplier has confirmed the arrangement, production has confirmed the tooling, quality has signed the joint review, and the BOM is frozen with these three parts, key, hub, screws, all standard stock. Attachment: updated hub drawing DRW-2041 rev C with the keyway and the flat geometry, and the signed joint review form.
Attachment: DRW-2041 rev C (hub with keyway, two flats at 90°, screw torque 4.5 N·m, Loctite 243); supplier holding-torque table M6 DIN 916; joint review form JR-71 signed by Design, Production, Quality.
Honest note after the thread — the mistake that nearly shipped was not the screw, it was the drawing: a two-flat hub at 120 degrees instead of 90, which puts one screw on the round and quietly halves the holding torque while the CAD looked fine. A set screw joint is only trustworthy when three things are true at once, the flat is full width under the screw tip, the shaft is the hardness the table assumes, and the tightening torque is a printed value not a mood. The supplier table was the honest referee, it said the number we wanted was half what we needed, and the compromise, key for torque, screws for position, is the arrangement that index drives have used for decades because it lets each element do the job it is actually good at. Worth keeping in the design rules when the next coupling hub lands on the desk.
Reference data — DIN 916 M6 cup-point set screw, tightening torque 4.5 N·m, holding on two flats 5.8 mm wide: continuous ~6 N·m, momentary ~8.5 N·m on a 20 h6 EN1A shaft. Key joint 5 mm DIN 6885 with 8.5-12 N·m: shear well inside limit. Shaft 20 h6, ground, degreased at assembly.
Glossary — set screw (grub screw): a headless fastener that bites into a flat or shaft to hold a hub; cup point: the cone-shaped tip of DIN 916 that concentrates the clamping load; holding torque: the tangential torque a set screw joint can transmit before slip; flat: the milled land on a shaft giving the screw tip a seating surface; BOM: bill of materials.
Field note — the arithmetic behind a set screw holding torque, kept short
For anyone who joins the thread late, the holding torque of a cup-point set screw boils down to one useful model: the axial force the screw tip presses into the flat, times the friction at the tip-shaft interface, times the shaft radius, summed over the screws that actually share the load. The axial force is the tightening torque divided by the thread pitch mechanics, so an M6 torqued to 4.5 N·m pushes with a fairly constant axial load, and what changes the outcome is the interface, a soft shaft and a narrow flat give the tip a small, plastic seating area and a lower usable friction, a hard shaft and a full-width flat give the tip a controlled bite and a higher holding. That is why the same M6 screw reads 9.4 N·m in the optimistic catalogue and 6 in the honest table, not because the supplier changed, but because the joint changed. And it is why the two-flat geometry matters so much on a 20 mm shaft: two flats at 90 degrees put both screw tips on steel, while a sketchy 120 degree split steals one screw, and the pair quietly becomes a single screw in the calculation nobody checked. When in doubt, let the key carry the torque and the screw carry the position, and the holding-torque argument stops being an argument at all.
