Work log — Labeller project, timing belt tension week
Engineer: D. Marchetti (machine commissioning)
Project: LBL-2300 hot-melt labeller, servo-fed label feed
Log period: Day 01 to Day 05
Document ref: WL-2026-0911-TB01
Day 01 — morning stand-up, the job before the log
Stand-up at 08:30, four of us. The label feed belt on the LBL-2300 has been slipping at the exit roller since the line trial, and the label pitch jumps by 3 to 4 mm every twenty or so labels. My job for the week: get the timing belt and its tension right, write down the procedure, and hand the measurement method to the maintenance techs so they stop guessing with the push-test thumb.
Morning session I spent with the drive drawing open. The feed uses an 8M profile, 25 mm wide, 384 teeth on the driven pulley and 40 on the small one, a 9.6:1 reduction to the feed roller shaft. The belt runs 1,040 mm centre distance with an idler on the slack side. Before touching anything I measured the existing tension with a frequency meter: the span vibration came back at 43 Hz on the longest open span, and the chart rating said the target for this drive is 62 Hz at 240 N static tension. So the belt was sitting maybe 40 percent under, which explains the jump, the belt was not stretching the labels, it was riding the pulley creep.
I set the meter reading aside and did the arithmetic properly, because the frequency number means nothing without the mass-per-metre of the belt. The belt maker’s datasheet gives 0.157 kg/m for the 8M-25, and the formula f = (1 / 2L) times the square root of (T / m) comes back to the measured 43 Hz when I plug in a tension of 106 N instead of the rated 240 N. The ratio works out. So the practical fault is a loose belt carrying the whole label pitch error, and the fix tone is to bring it to the 240 N band and then watch the driven pulley for creep under load.
Evening note: I keep a tension history on the belt since install, three data points, 62 Hz on day one, 54 Hz after the first shift, 43 Hz now. The belt had no sign of tooth wear when I pulled the guard, so it is stretching or the idler is walking. Something to check tomorrow before I re-tension.
Day 02 — the re-tension and the idler that was lying
Started by checking the idler axis with a straight edge before touching the belt, and that was the find of the day. The tension idler on the slack side was 0.9 mm out of square with the pulley faces, cocked toward the driven side. A belt that runs onto a misaligned idler walks and the walking scrubs the teeth off the edge, and it also shows up as a slow tension loss, which is exactly what the three-point history said. The idler base plate had been slotted at the factory and never locked, the slot allowed it to twist under the belt load. I pulled the plate, set it square with a dial against the pulley face, and locked the two hex screws at 20 N·m.
Then the re-tension. I released the motor plate, backed the belt off, and re-set it in two passes rather than one hard pull. First pass to 180 N, run the belt one full revolution by hand and let the teeth seat, second pass to 240 N measured on the frequency meter at the longest open span, 62 Hz target, and I settled it at 61.9. The pitch-drive spec wants 250 N on the tight side at running load, and 240 N static is the book number, so I logged 240 N static / 250 N dynamic as the pair.
Afternoon I ran the belt at the line’s 200 labels per minute with no glue, just counting pitch on the encoder. The jump dropped to zero on a 500-label run, encoder reads steady at 1,000 pulses per label. I left it running while I went to the afternoon stand-up and it held. Then I remembered the evening note about tooth wear and the second measurement: 62.4 Hz after two hours of running, so the set is staying, no creep. Good day. The one thing I did not do was write the SAG-check procedure; that is Day 03.
Evening thought: the failure taught me the two measurements belong together. The frequency meter tells you the tension, and the straight edge tells you whether the tension will survive. On this machine the idler was the root cause and the belt was the symptom, and had I only re-tensioned, Day 05 would be a repeat.
Day 03 — writing the SAG-check so the shop stops guessing
The maintenance techs asked for a method they can run without the frequency meter, and I spent the morning turning the belt away from the instruments and toward a feeler and a rule. The standard SAG check: apply a 50 N midspan force with a spring gauge, measure the belt deflection, and compare against the 1.6 mm per 100 mm of span datum. On this drive the longest open span is 420 mm, giving a SAG reference of 6.7 mm at 240 N tension. I tried it against the frequency meter and the two agree within 0.3 Hz, close enough for a field check. I wrote the two-number method into the sheet: 50 N gauge, 6.7 mm sag on the long span, and a re-check after the first overnight run.
Mid-morning I ran the full production cycle with glue and the 1,000-pulse encoder still reading clean at 200 labels per minute. Pitch registers plus or minus 0.4 mm against the 0.5 mm spec, so the label feed is now inside limit, and the previous 3 to 4 mm jump is gone. I also re-torqued the two idler plate screws after the run, 20 N·m, and marked the slot position with a paint line so a future loosening is visible at a glance.
Evening note: two things I would do differently. First, tension history should start on install day and not on failure day, three data points was lucky. Second, the SAG procedure must give the same answer as the meter or the shop will not trust it, and checking that agreement took me an hour I should have assumed away.
Day 04 — a full shift, then the small surprise on the motor
Full shift run, 3,200 labels, zero pitch complaints, belt steady at 62.0 Hz at the end of the day from 61.9 in the morning. Then the surprise: the feed motor frame felt warm at 44 C against a usual 38 C. Not a fault, but I logged it, because a belt that was just re-tensioned from 106 to 240 N adds a measurable motor load, and a 6 K rise on this small servo is inside its 60 C limit but worth a re-check at next week’s PM. The tension increase was paid for in motor temperature, and I want the month’s data before I call the 250 N dynamic number final.
Day 05 — handover and the numbers that close the job
Handover day. I filed the procedure sheet with the two methods (frequency meter 62 Hz, SAG 6.7 mm at 50 N), the idler squareness check with the paint mark, and the operation record for the running week. The maintenance handover ran three labels through at 200 per minute with pitch inside 0.5 mm, and the tech signed. Motor at 43 C under the week’s last run, still inside limit, logged for the PM sheet.
Closing the log with the numbers the job owed: 8M-25 belt, 384/40 teeth, 1,040 mm centre distance, tension settled at 240 N static measured 61.9 to 62.4 Hz over the week, SAG 6.7 mm at 50 N on the 420 mm span, idler squared and locked at 20 N·m with a paint mark, pitch error from 3 to 4 mm down to plus or minus 0.4 mm. The techs have a procedure they will actually use, which was the real deliverable.
Attachment A — parameter record table
| Item | Value |
|---|---|
| Belt profile | 8M, 25 mm wide |
| Teeth, driven / small pulley | 384 / 40, ratio 9.6:1 |
| Centre distance | 1,040 mm |
| Static tension target | 240 N (book), 250 N dynamic |
| Frequency target | 62 Hz on longest span |
| SAG check | 6.7 mm at 50 N on 420 mm span |
| Idler torque | 20 N·m, squared and paint-marked |
| Pitch result | +/- 0.4 mm vs 0.5 mm spec |
Attachment B — check list for the next tension job
1. Measure existing frequency and log it before touching anything. 2. Check idler squareness with a straight edge before re-tensioning. 3. Set belt in two passes, seat the teeth by hand between passes. 4. Measure frequency at the longest open span only. 5. Run one production cycle, re-measure, and log the drift. 6. Re-check motor temperature after the new tension settles.
Glossary
Static tension: the belt pretension set with the drive stopped. Frequency meter: an instrument that measures belt span vibration and converts it to tension. SAG check: a field deflection measurement using a spring gauge and a rule. Slack side: the belt span that runs from the driven to the driver pulley under load. Idler squareness: the alignment of the idler axis relative to the pulley faces.
Week-end reflection — the log as a fault record
I wrote this log partly to keep my own head straight, and it turned into the fault record the line actually needed. Read it back, and the sequence is clean in a way the days were not: the pitch jump pointed east, the frequency history pointed at the idler, and the straight edge caught the idler before I wasted a belt. The technician’s question at handover was the one worth answering on the sheet, not what the tension was, but how do I know it is still right on a Tuesday morning. That is what the SAG check and the paint mark answer, a 30-second check instead of a 240 N guess.
One more number for the file: the belt maker’s datasheet says the working tension range for this drive is 220 to 300 N, so the settled 240 N sits low but usable, and the 250 N dynamic number stays logged as a watch item tied to the motor temperature trend. If the motor climbs past 50 C over the next two PM cycles, I pull the tension back to 230 N and accept the small extra creep rather than cook the drive. That decision is written in the maintenance plan, not carried in my head.
The log closes with the standing rule I dropped into the procedure sheet: on this labeller, tension is checked on the same day as idler squareness, never one without the other. The first check without the second found nothing, and the second without the first would have held for exactly one shift.
A final arithmetic note no one asked for but every line will thank me for: at 200 labels per minute the feed roller turns 1.6 times a second through the 9.6:1 reduction, so the driven pulley carries about 33 rpm at the belt, and a 240 N static figure translates to roughly 180 N effective tension on the tight side once the servo torque is subtracted. None of that changes the setup, but when a future technician sees the belt at 43 Hz again he will know it drifted a full 60 percent of the working window before the pitch error showed up, and that the 62 Hz datum is not a preference, it is the floor.
