WORK ORDER — Maintenance and Improvement Record
Work order no: WO-A-2026-02 Date issued: 2026-09-11 Category: Robot 7th axis
Machine / cell: Welding Cell WC-2, articulated robot on a 6 metre linear track
Status: Closed Engineer: S. Varga (robotics) Total labour: 11.0 hours Parts cost: EUR 402.00
Reason for work: weld seam wander on long seams tracking the 7th axis motion
1. Fault description as logged by the shift team
Monday 07:20, operator record: the WC-2 robot lays a weld seam that drifts sideways partway down the long rails of a chassis sub-frame, the weld starts on the joint line and wanders up to 1.6 mm off the seam by the end of the 1.8 metre rail length, and the quality office has rejected the last four chassis for lack of fusion on one wall. The robot itself welds a short test piece perfectly, point-to-point, no drift, so the complaint points at the seventh axis, the 6 metre linear track the robot rides while it follows the long seam. The track is a rack-and-pinion drive with a toothed belt and two guide rails, built two years ago, cycled around 80,000 metres of travel since.
Tuesday I arrived with the laser interferometer and the track alignment records, because a wander that grows along the axis has two classic parents, a track that is not straight in plan, or a pinion that is rocking against the rack under load. The robot’s own repeatability is 0.06 mm and off the track it proves itself, so I drew the chalk line: the seam wander is the track telling us, through the weld, that its carriage is not moving on a true straight line.
Evening note to the file: a weld is a perfect 1:1 recorder of a track fault, every wobble of the carriage gets welded into the part and shipped to quality, so the interferometer today is the honest instrument, not the torch.
2. On-site inspection record (Day 1 afternoon)
I set the laser interferometer along the track and swept the carriage over its full 6 metre travel at the welding feed speed, 12 mm per second, and the plot was a textbook fault signature: the carriage lateral position ran within 0.05 mm of the straight line over the first 2.8 metres, then climbed steadily to a 1.4 mm peak around the 3.9 metre mark, held high for about a metre, and dropped back to within 0.06 mm by the end of travel. A rail that is bent in plan produces exactly that shape, a slow hump rather than a scatter, and the hump sat right where the weld seam had been wandering. I marked the floor under the hump and the bracket positions along the far rail, and pulled the track covers to inspect the mounting.
The first visible clue came off the far guide rail: the rail clamp over the marked hump had a 0.4 mm gap under its foot at one end, and the M8 mounting bolt there was loose enough to turn by finger, a bolt torqued originally to 24 Nm sitting at barely 4 Nm after two years of thermal cycles and weld spatter dust getting under the clamp. A loose clamp lets the rail sag in plan, and the sag is exactly the hump the interferometer drew. The second clue came off the pinion: I wiped the rack teeth over the hump zone and found a wear flat on the tooth flanks, the rack being pulled sideways by the sagging rail so the pinion ran against the rack side instead of the root, shaving a bright flat on the tooth face and pointing the carriage one way and then the other as it crossed the hump.
I checked the rack straightness with a taut wire and a feeler gauge over the suspect metre, and measured a 1.2 mm sideways bow in the rail at the loose clamp zone, which matched the 1.4 mm interferometer peak almost exactly, the remaining scatter coming from the rack tooth flats. So the diagnosis wrote itself: a loose rail clamp turned a straight track into a humped track, the humped track dragged the rack sideways, the pinion shaved the rack teeth, and the wandering weld was the whole chain reporting itself through the torch. The robot hardware was never guilty, it was riding a track that had quietly bent in plan.
Evening note: an unstraight track gets welded into every part before anyone sees the rail, and the seam wander that burned four chassis was one loose M8 clamp sitting at the feet of a hump that took me forty minutes to find and a laser to believe.
3. Corrective action taken
I released the two clamps either side of the hump, lifted the rail free of the mounting plate, and reground the mounting face flat to remove the spatter blisters that had raised the plate under the loose clamp, checking the face with a straightedge and a 0.02 mm feeler, then cleaned the clamp feet and the rail underside until the rail lay dead on the plate. I re-tightened the clamps in the proper sequence, working from the fixed end toward the free end so the rail could not be pushed out of line by accumulated bolt drag, torquing each M8 clamp bolt in two passes to its 24 Nm spec with a calibrated wrench rather than one pass, then re-swept the carriage with the interferometer before I let the cell run. The re-sweep after the first re-torque still showed a 0.25 mm hump, which traced to a chip pushed under one clamp foot, so I lifted that clamp, blew the chip out, and re-torqued again; the second sweep ran flat to within 0.04 mm over the full 6 metres, which is inside the track drawing tolerance of 0.06 mm per metre, so I declared the straightness job done and moved to the rack.
The rack teeth over the hump had a light wear flat, a bright polish rather than a deep pit, so I did not swap the rack but ran a worn-teeth check, measuring the tooth root-to-pitch-line wear with a micrometer against the new rack sample, and the deepest worn tooth lost 0.11 mm against a wear limit of 0.20 mm, so the rack keeps its place with a note to re-measure at the next 20,000 metre service. I replaced the drive belt as a precaution, the old belt had a visible 1.2 mm stretch mark and the pinion had been hammering through it over the hump, and I re-set the belt tension to the factory 180 Hz by the belt tension gauge, then greased the rack with the polyurea grease the track builder specifies, one clean pass over the whole 6 metres.
| Inspection / action item | Finding | Action | Result |
|---|---|---|---|
| Rail flatness sweep (laser interferometer) | 1.4 mm hump at 3.9 m | Re-clamp with face re-grind | 0.04 mm over 6 m |
| Clamp bolt torque | 4 Nm in lieu of 24 Nm | Torque in two passes | 24 Nm calibrated |
| Rack tooth wear | 0.11 mm max at hump | Keep, re-check at service | Within 0.20 mm limit |
| Drive belt | 1.2 mm stretch | Replaced, tensioned to 180 Hz | Factory spec |
Track builder service interval note: our maintenance programme had checked the pinion and the belt but never the rail clamp torque, because the drawing packs the clamps away under the covers and the audit sheet skips them; that blind spot is exactly how a straight track sleeps into a humped track without anyone noticing until the weld walks.
4. Acceptance result after the repair
I ran the cell release test the written way: three production chassis welded back to back on the re-aligned track, the first seam checked by the shop floor gauge at 0.02 mm deviation from the joint line over the full 1.8 metre rail length, the second and third at 0.03 mm and 0.02 mm, against the 0.5 mm acceptance limit for long seams, and quality released all three parts the same day instead of the usual re-work loop. Weld current reads the same as before the repair, the seam profile came out constant end to end, and the robot’s own axis data shows the carriage holding its programmed path to within 0.04 mm without any of the old sideways correction kicks. The cell ran a full shift, 14 parts through the seam station, and each one tracked the joint line, so I closed the work order and logged the rail clamp torque as a recurring service item, a quarterly check with a calibrated torque wrench, so the hump does not get two years to grow again.
5. Hours and parts list
- Labour — 11.0 h total: inspection and laser sweep 3.5 h, rail face grind and re-torque 2.5 h, rack check and belt change 2.0 h, re-sweep and cell release test 3.0 h
- Parts — drive belt EUR 120.00, polyurea grease 0.5 kg EUR 28.00, clamp bolts and shims EUR 54.00, rack wear gauge rental EUR 200.00; total EUR 402.00
- Wear note — two of the four chassis rejected earlier were economically re-welded over the seam, two were scrapped at the supplier claim
6. Retrospective notes for the next engineer
If a long weld wanders and the robot welds clean point-to-point, distrust the track before you distrust the torch, because a 1.4 mm rail hump reads through the weld as a quality defect and fools the operator into changing weld parameters that were never wrong. The loose bolt told the whole story but two years of audit sheets never looked at the clamp torque, and the interferometer was the only fast way to point at the metre of rail that mattered. A straight track under honest clamps holds a 0.04 mm sweep across 6 metres of travel, and the rack, the belt and the pinion all behave, so the seam stays on the joint line and the quality office stops vibrating. Keep the clamp torque on the quarterly sheet, because the cheapest thirty-minute check is worth more than the four rejected chassis and the one scrapped weld cell day.
Attachments: laser interferometer sweep plot, track flatness drawing, rack wear measurement sheet, cell release test record (hard copy filed with WO-A-2026-02).
Glossary — seventh axis: external linear axis that moves a robot along a track; rack and pinion: toothed bar and wheel converting rotation to translation; hump in plan: lateral bow of a rail seen from above; polyurea grease: long-life synthetic grease for rack drives; cell release test: run-off proving a repaired cell makes conforming parts.
Field note — the alignment arithmetic behind the re-torque
Worth writing down why a 0.04 mm sweep after the repair corresponds to the weld behaviour we saw, because the numbers connect the otherwise separate worlds of track metrology and weld quality. A rail clamped off-flat by spatter raises the guide way under the carriage, and the carriage rolls a fixed rocker geometry over the high spot, so a 1.4 mm rail bow does not translate one-to-one into seam wander, it is modified by the carriage wheel base of 480 mm and the weld torch offset of 210 mm ahead of the lead wheel, which stretches and shrinks the apparent lateral error as the node passes. The measured seam wander of 1.6 mm is consistent with the 1.4 mm bow once you add the pinion side-thrust kick at the rack flats, so the laser plot and the weld plot are two views of the same hump. After the clamp repair, the last seam checked on a 1.8 metre rail stayed within 0.03 mm of the joint line, and the shop floor gauge card records pass on every check point; the robot controller shows the carriage axis correction term dropping from 1.3 mm to 0.05 mm in the same shift, which is the most convincing single number, because the servo is no longer fighting a hump it cannot see. For the next engineer, the intervention rule is: if the weld walks along a long seam and the robot welds clean point-to-point, run the interferometer before changing any weld parameter, and put the rail clamp torque back on the quarterly audit where it belongs, because a straight track is the cheapest quality machine we own.
