From 27 Minutes to 5 Seconds, and Only Half of It Was About Setup

A label finishing line with seven stations: uploader, manned Polar strip cutter, Polar AutoCut block cutter, bander, wrapper, sortation, packing.

Job-to-job changeover measured 27 minutes. After about eight weeks it measured 5 seconds. The stages were 27 → 12 → 8-9 → 5 sec, and the last stage came from a different mechanism than the first three.

Stage one: 27 to 12 minutes

Standard SMED. Internal work separated from external, external work moved off the stopped-line time, then motion reduction on what remained.

The largest single item was walking. The cutting lines were long, and changing the bander required walking around the back of it to reposition the heads. That walk had never been timed because it does not look like setup work.

Result: 12 minutes.

Stage two: 12 to 8-9 minutes

Two changes.

The bander heads were fastened with machine bolts, requiring tools and a trip to get them. They were replaced with hand knobs, and every head was refurbished at the same time.

The second change was procedural. Jobs were not run back to back through the bander, on the standing assumption that consecutive jobs risked a mix — labels from one job ending up in another job's bundle. Rather than argue the policy, mixes were attempted deliberately and repeatedly under production conditions. None occurred. The restriction was removed on that basis.

Result: 8-9 minutes, and no further movement from setup-side work.

What the remaining time consisted of

The packers at the end of the line needed three things before a new job could run: cartons, printed carton labels, and carton samples — labels from the job glued to each carton so contents can be verified visually.

None of that is quick, and none of it could start until the packing end knew the job was changing.

Notification was verbal. The cutters were several dozen feet from packing, separated by a trim system loud enough to make speech impractical. The sequence was: cutter finishes, cutter shouts, packer hears, packer begins preparation, line waits.

That interval was being recorded as changeover time and treated as setup. It was neither internal nor external setup work. It was the delay between the front of the line knowing and the back of the line knowing.

Stage three: rolling setup

The line was reorganized to change over in sequence rather than as a stop event. The uploader was cleared of the outgoing job first, then loaded with the incoming job while the outgoing job was still being strip cut. Each station changed over while the stations upstream of it still ran the previous job.

Job information was moved onto the product.

On the first load of a new job, the cutter sent a single sheet of that job down the line ahead of production. It reached the reject line, where a packer collected it. Sheets carried combo job numbers printed on them, so one sheet gave packing the job numbers it needed to print stickers, pull cartons and build samples. It arrived with approximately 20 minutes remaining in the current job.

End of job was signaled the same way. On the last bundle, the cutter applied round stickers over the UPC so the bundle would fail scanning and be rejected. A rejected final bundle meant the job was complete and the lines should switch to the next UPC.

Carton samples needed a holder. Mylar blocks were taken from maintenance stock, tested in the ovens to confirm they would not melt or ignite near the line, and fitted with bolted metal bull clips.

What 5 seconds measures

At the point of changeover the packing end has been prepared for 20 minutes, every station has already rolled to the new job, and the AutoCut is programmed and waiting.

The remaining internal work is: pause the rolling setup, visually check line clearance inside the mouth of the cutter, press go for the pusher after strip cut.

That is the 5 seconds. It is the complete internal time — the only work that still requires the line to be stopped. Everything else that made up the original 27 minutes now happens while the line is producing.

Notes

Untested restrictions. The mix policy had governed the line for years without anyone attempting to reproduce the failure it prevented. Attempting to cause it, and failing to, is what removed it. The attempt is cheap and rarely made.

Changeover time on a serial line often measures information latency. Individual stations can each be well optimized while the line still changes over slowly, because the binding constraint is when the far end of the line learns that the job has changed. Setup reduction does not reach it.

Signals carried on the product are robust. Material already travels the route in sequence, arrives at the right station, and is unaffected by ambient noise. In this case a deliberately rejected sheet and a deliberately rejected bundle carried the two signals the line needed.

Constraint. This work was done to keep the AutoCut fed, that station being the one that set line rate. The related effect on the pressroom is covered separately.