Leading the Intelligent Factory:  Chapter 7

Run the same way on every shift

Targets from your own history, and a score that shows where you run to them.

Spread 1

Four crews, one process

The pressure. The same grade comes out a little different from one shift to the next. Each handover moves the process. Quality that depends on who is on shift is quality you cannot promise a customer.

What goes wrong today. Targets live in a binder, a spreadsheet, or the heads of your best operators. Each crew has its own way to run. A setting moves for a good reason on Tuesday and never moves back. After a few months, people can say where the process is running, but not where it is supposed to run.

What AI changes. This is the decision of where everything is set. A machine can build a target and limits for every key tag from your own run history, by grade and by speed. It then scores every tag, every hour, against those limits. One number shows how closely the plant ran to its own standard. A short list shows which tags make most of the misses.

What still needs people. For each tag on that list, one decision: run to the target, or reset the target. Only your team can say which is right.

Upper limitTargetLower limitOut of band: 11 hoursIn band 37 of 48 hours. Score for this tag: 77%One tag, one dot per hour, two daysIllustration, not data
Figure 7.1. How the score is built. Hours inside the band, divided by hours run, summed over every key tag.

Spread 2

A real example: one month on a paper machine

A paper machine scored 155 key tags, hour by hour, for one month. That is about 100,000 tag-hours over 646 run hours. The machine ran inside its own limits 89.8% of the time.

Day shift scored 89.9% and night shift 89.6%. When one shift had a weak day, the other usually did too. That points at the process, not at the crews.

The misses were not spread across the machine. Eleven tags made 65% of them, and most were in the wet end. Each of the eleven sat at a steady offset all month. That most likely means one of two things: the target no longer fits, or the level moved and stayed. Either answer clears the tag.

The work list

11 decisions, not 155

Run to the target or reset it on eleven tags, and the rest of the machine scores 96.2%.

Hours out of band in one month, top 15 of 155 tags646 run hours. A bar near 640 means the tag was off target all month.A1 vacuum638A2 vacuum638A3 vacuum638A4 vacuum638Refiner A power631Top liner screen dP599C3 vacuum593Top liner slice flow593C1 C2 vacuum592Layer 2 slice flow591E7 vacuum562E6 vacuum309Top liner freeness222Draw group 4-3186Dryer steam header pressure18265%of all missesfrom 11 tagsThe other 140 tags together: 26% of the misses.
Figure 7.2. One month, 155 tags scored. Eleven tags, in orange, make about two-thirds of all misses.
Compliance score, %Share of all missesWet end82.061%Stock preparation85.121%Dry end96.015%Press96.62%On-machine coating98.11%Utilities98.80%
Figure 7.3. The same month by machine section. The wet end carries 61% of the misses.

Also in this chapter

What you run, and in what order

Running to target is half of consistent quality. The other half is the schedule. Every grade change costs off-spec tons, and a large step costs more than a small one.

Most schedules come from the order book and from habit. The cost of each change is rarely measured, so the sequence is inherited, not designed.

A machine can measure what each change has cost in your own history, grade pair by grade pair. With that table, your scheduler can group grades and order them so each step is small. The orders are the same. The off-spec is not.

Inherited order26 lb8 t42 lb6 t33 lb5 t26 lb6 t36 lb4 t42 lb29 toff-spec5 grade changesDesigned order, same tons of each grade26 lb5 t33 lb4 t36 lb6 t42 lb15 toff-spec3 grade changesIllustration, not data. Green: grade runs by basis weight. Orange: off-spec tons at each change.
Figure 7.4. The same week of orders, run in two sequences.

Questions for Monday morning

  1. Where are our targets written down, and does every crew use the same ones?
  2. How many of our key tags ran inside their limits last month?
  3. Which few tags make most of our misses?
  4. When a tag sits off target for a month, who decides: run to it, or reset it?
  5. Do we know what each grade change costs us in off-spec tons?

In practice. Smart Centerlines builds targets and limits from your run history, then reports a compliance score every month by week, shift, section and tag.