WORK ORDER — Maintenance and Improvement Record
Work order no: WO-A-2026-09 Date issued: 2026-09-11 Category: Fastening and assembly
Machine / cell: Four-spindle nutrunner station NR-2 in the engine assembly line
Status: Closed Engineer: P. Novak (automation) Total labour: 8.0 hours Parts cost: EUR 128.00
Reason for work: achieved tightening torque drifts below the clamp spec and the four spindles scatter unevenly
1. Fault description as logged by the shift team
Friday 10:30, operator record: the four-spindle nutrunner tightens the cover bolts to a nominal 120 N·m, and in the last two weeks the achieved torque has drifted down to 108-112 N·m on a growing share of parts, with one spindle consistently reading 8% below the others, and the quality office flagged the fastening records for the clamp band as concerning. The team had re-calibrated the transducers twice and re-frocked the spindle couplings, and the drift returned, which pointed away from the electronics and toward the mechanical joint the tool was actually tightening, the relationship between the nutrunner head and the bolts it drives.
The station is a four-spindle nutrunner with a common frame, each spindle driven by its own servo torque package through a universal coupling to the socket, and the frame indexes down onto the cover, whose four M10 bolts hold the cover to the housing face. The achieved torque is read by the transducer in each spindle and logged against the bolt number, so a single spindle reading low is a mechanical signature, a loose coupling, a worn socket or a damaged bolt, while a general drift across all four is a joint or a head-seating problem. I went in with a torque wrench, a socket inspection mirror and the fastening log rather than the calibration box, because the log had already told me the transducers were probably fine and the joints were not.
Evening note to the file: when a recalibrated tool still drifts, the energy is going somewhere the transducer cannot see, and the first place a torque loss hides is the mechanical path between the socket and the bolt.
2. On-site inspection record (Day 1)
I cross-checked all four spindles against a calibrated reference torque cell on the bench first, and all four read within 1.5% of the cell, so the electronics and the transducers were innocent, and the scatter was living in the mechanical path. I pulled the four sockets and inspected them under the mirror: socket three, the low spindle, had a worn detent bore that gripped the bolt head off-centre by about 0.8 mm, and its drive square had rounded corners so it slipped angularly before the bolt truly grabbed, and socket two had a cracked wall that twisted elastically under load, eating torque as strain instead of passing it to the bolt. I checked the universal couplings next, and couplings two and three had dry, worn pins that let each spindle back off by a few minutes of angle under the reaction, which is exactly the per-spindle scatter the log had been showing.
I measured the joint itself before changing anything, because a torque drift across all four could also be a lubrication or face-seating change: the cover bolts came in with no lubricant and a dry interface, while the drawing specified a light 10W-40 oil on the threads and the flange face mopped clean, and dry threads of M10 at 120 N·m consume a different share of the torque in friction, leaving less of it as clamp. The combination was the full picture: one worn-detent socket stealing its spindle’s torque, one cracked socket stealing energy as strain, two dry couplings adding backlash, and a dry joint quietly lowering the whole clamp band, four separate debts all reading as the single “torque drift” the line had been fighting with a calibration box.
Evening note: recalibration cannot fix a worn socket or a dry thread, because the transducer reports what reaches its own sensor faithfully, and the mechanical chain between the sensor and the bolt had four places to lose the struggle.
3. Corrective action taken
I replaced socket three with a new hex-socket pair matched to the bolt head, so the drive bore grips the M10 head square and on centre with no detent slop, and I replaced socket two with a solid-wall socket that passes the torque as shear to the drive flats instead of swallowing it as wall strain. I rebuilt couplings two and three, replacing the worn pins and re-greasing the U-joint yokes so each spindle drives its socket without backing off angularly under reaction, and I re-torqued the frame mounting to the drawing 90 N·m so the four spindles index square onto the cover and none of them heels against the bolt. On the joint side I put the cover bolts on the specified lubrication, a light 10W-40 wipe on the threads and a clean flange face, per the drawing, so the joint friction is back to the design value and the achieved torque lands as clamp rather than being eaten by dry friction.
I re-ran the station over four assemblies, sixteen bolts total, and the achieved torque came back to 119-121 N·m on every spindle, with the worst scatter 0.9% against the 8% drift before, and the angle-of-turn for each bolt landed in the drawing window, confirming the clamp was real and not just a higher reading on a still-lossy joint.
| Item | Before | After | Spec |
|---|---|---|---|
| Achieved torque vs 120 N·m | 108-112, one spindle -8% | 119-121 all spindles | 120 ±2 |
| Within-spindle scatter | up to 8% | 0.9% | ≤2% |
| Socket drive engagement | 0.8 mm off-centre, cracked wall | new square hex sockets | full flat |
| Coupling pins | dry, worn | rebuilt, greased | no backlash |
| Joint thread lubrication | dry | 10W-40 per drawing | as drawn |
I logged the socket condition into the daily change check, because a nutrunner socket is a consumable with a life measured in thousands of cycles, and a worn detent or a cracked wall is exactly the sort of small mechanical debt that shows up as a faster torque drift than any transducer drift ever would.
4. Acceptance result after the repair
The station ran a full week and 260 assemblies through the four spindles, and every fastening record landed in the 118-122 N·m band, with the angle-of-turn trace stable and no spindle re-check alarms, against the growing drift and the 8% low spindle before. Quality signed the re-certification with the clamp band verified by a calibrated cell on twenty five sampled fasteners, and the shift no longer sees the intermittent “torque low” alarms that used to pull the station down for rework. I closed the order with the station released, the procedure sheet updated for the daily socket check and the quarterly coupling and head-seating audit, so the mechanical chain that carries the torque is watched with the same care as the transducer that reports it.
5. Hours and parts list
- Labour — 8.0 h total: reference cell cross-check 1.5 h, socket and coupling inspection 1.5 h, socket and pin replacement 2.0 h, frame torque and joint lubrication set 1.5 h, verification across 260 assemblies 1.5 h
- Parts — hex socket pair set EUR 42.00, universal coupling pin and yoke kit EUR 35.00, U-joint grease EUR 9.00, frame bolts and 10W-40 lubricant EUR 12.00, socket inspection mirror EUR 30.00; total EUR 128.00
- Impact before repair — 8% low spindle, 108-112 N·m on up to 20% of parts; impact after — 119-121 N·m on 260/260 assemblies
6. Retrospective notes for the next engineer
When a nutrunner drifts after a transducer recalibration, do not reach for the calibration box a third time, go look at the sockets and the couplings, because the transducer reports faithfully what reaches it and everything between the sensor and the bolt can lose the struggle first. The failing sockets and the dry joints were the whole story: a worn detent socket and a cracked socket wall, two dry couplings and a dry thread, four small mechanical debts that together read as an 8% torque loss while the electronics sat there innocent. A torque-controlled station is only as honest as its mechanical chain, so treat the socket as a consumable on a life counter, check the couplings for backlash at the audit, and keep the joint lubrication on the drawing value, because clamp is the torque that survives the chain, and the chain is maintained in the bin, not in the software. When the four spindles agree again, the drawing clamp is finally real.
Attachments: reference cell cross-check, socket and coupling inspection photos, fastening log before and after, 260-assembly verification record (filed with WO-A-2026-09).
Glossary — nutrunner: a multi-spindle power tool for tightening fasteners; achieved torque: the torque logged by the spindle transducer; clamp band: the range of acceptable tightening values; U-joint coupling: the universal coupling taking the drive to the socket; angle-of-turn: the rotation measured after the seating point of the bolt.
Field note — where the applied torque actually goes on an M10 bolt
The arithmetic behind the drift is worth a permanent note, because it explains why a dry joint and a worn socket both read as torque loss and a recalibration fixes neither. On a torque-controlled M10 bolt the applied torque splits into roughly three slots by the classic relationship: about half of the applied torque goes to overcoming thread friction, about a third goes to the friction under the bolt head, and only a small share, usually ten to fifteen percent, is left to stretch the bolt into clamp. So at the nominal 120 N·m, only some 14 to 18 N·m of equivalent effort actually creates the clamp force, and every friction change in the joint shifts that split. When the threads run dry, the thread-friction share climbs, the clamp share shrinks, and the same achieved torque reading now holds less clamp than the drawing intended, which is precisely what shows up as the quality office’s concern even when the transducer reads nominally. The worn socket adds a different loss: it does not change the joint friction, it steals torque before it reaches the bolt at all, spinning in the detent and bending the cracked wall, so the spindle transducer reads 120 while the bolt sees less, and re-calibrating the sensor only made the tool report the theft more accurately. Repair the mechanical path and restore the drawing friction, and the torque comes back because the energy stops leaking before the bolt; keep the threads oiled, the sockets fresh and the couplings tight, and the split lands where the designer put it, with the clamp finally getting its fair share.
One closing operational note for the line: keep the angle-of-turn column in the fastening log written next to the torque, because a torque reading and an angle reading together tell the clamp story that torque alone obscures. A dry joint shows the same torque as an oiled joint at a much smaller angle of turn, because the bolt stops stretching early and the whole twist dissolves in thread friction, so the angle column catches the lubrication drift days before the torque band drifts below spec. The shift now reads both numbers on the first bolt of every batch, a two-second glance that has caught a dry-thread batch in time, and it is the cheapest torque insurance a nutrunner station can carry, because a bolt that stretches the right amount is a bolt that clamps the right amount, regardless of what the friction is doing in between.
