The Paint-Booth Math: Sequence-Dependent Changeovers in EDGEBIC

Sequence-dependent setup time scheduling means the changeover cost of a job depends on what the machine ran immediately before it, and a scheduler that ignores this is not simplifying your plan, it is falsifying it. In a paint booth, White to Black is one 60-minute flush; Black to White is a 240-minute solvent purge. Same two colors, four times the cost, purely because of direction. EDGEBIC by User Solutions models this with a from-to setup matrix per work center, and this walkthrough runs one six-job queue through three schedules: the flat-setup plan that lies, the matrix plan that tells the truth, and the campaigned plan that wins the day back.

The Booth and the Queue

The shop has one paint booth, PAINT-1: a single instance, one Day Shift from 08:00 to 16:00, 8 hours of capacity. Three products run through it, the colors White, Black, and Red. Every job takes 1.0 hour of painting, and the routing carries a flat default setup of 30 minutes, the number the schedule uses when nothing better is known.

Six orders are due this week, one per day, all released Monday morning:

Sorted by earliest due date, the booth sequence is White, Black, White, Red, Black, White. Alternating light and dark: the worst possible order for a paint booth, and exactly what a due-date sort produces.

The real changeover times, measured at the booth, go into EDGEBIC's setup matrix for PAINT-1:

Any pair not in the table falls back to the 30-minute routing default. Same color back to back costs zero.

Schedule 1: The Flat-Setup Plan That Lies

With no matrix configured, every job gets the flat 30-minute setup. Six jobs at 1.5 hours each stack neatly down the Monday column:

Claimed finish: Monday 17:00, one shift plus an hour of overtime. The plan books 180 minutes of total setup.

Walk the same sequence through the real matrix and the changeover alone is 630 minutes, more than three times what the plan claims. The Black-to-White purge appears twice in this order, at 240 minutes each. The schedule is internally consistent and completely wrong, and nobody finds out until Monday 11:30 when the booth is still running solvent and the afternoon's jobs have nowhere to go. This is the core failure mode we describe in job shop scheduling challenges: a plan built on inputs that do not match the floor gets ignored, and the whiteboard takes over.

Schedule 2: The Honest Plan

Now the matrix is loaded and EDGEBIC resolves setup per job by asking one question before each: what did this booth run last? The first job finds an empty booth, a cold start, and uses the 30-minute default. Every job after that gets a real from-to lookup:

Look at MO-2003. Its 4-hour purge plus 1 hour of painting is 5 hours of work, but Monday only has 4.5 hours left after 11:30. The engine allocates 4.5 hours Monday afternoon and carries the remaining 30 minutes across the shift boundary to Tuesday morning. Setup is not a special case here; it is simply part of the work, and it spills across shifts like any other work would.

Honest finish: Wednesday 08:30. Two and a half shifts for six jobs, with 630 minutes of total changeover. Nobody enjoys seeing that number, but it is the number the floor was always going to live, and now it is on the schedule before you promise anything against it.

Schedule 3: Campaign the Colors, Win Back the Day

The matrix does more than make the schedule honest. It exposes the cost of the sequence, and the sequence is yours to change. Group the light colors first and the dark colors last, White, White, White, Red, Black, Black, and the expensive dark-to-light purge never happens at all.

For this schedule the planner also switches from product-level entries to setup families, the abstraction you reach for when the color list grows. White and Red join a Light Colors family; Black is Dark Colors. Changeover is then entered between families: light to light 10 minutes, light to dark 45, dark to light 200, dark to dark 0. Product-level entries still override a family rule wherever a specific pair misbehaves. Our explainer on setup families and sequence-dependent changeover covers when families pay for themselves; the short version is that eight families over 100 products replace ten thousand matrix cells with 64.

The campaigned run:

Finish: Monday 15:30, half an hour before end of shift. Total changeover: 90 minutes.

The Three Schedules, Side by Side

Same six jobs, same booth, same paint hours. The only variables are whether the schedule knows the real changeover cost and whether the sequence respects it. The matrix turns a fictional Monday finish into an honest Wednesday one; the campaign turns the honest Wednesday into a real Monday. That second step is the one that pays: the booth just gained most of a working day of capacity without a dollar of capital.

Every Setup Number Explains Itself

A schedule you cannot interrogate is a schedule you will not trust, so EDGEBIC stamps every scheduled operation with a setup source code and a plain-language reason, both visible as columns in the Job View:

When a supervisor asks why MO-2003 swallowed Monday afternoon, the answer is on the row: Black to White, 240 minutes, from the product matrix. And the audit survives rescheduling. If Black actually completed at 11:30 this morning and you reschedule, the engine reads the booth's last product from the recorded actuals, so the next White job still pays its full purge. The booth's contamination is physical; the schedule tracks it whether or not the original plan survived the morning.

One more honest boundary: EDGEBIC schedules the sequence you give it and prices every changeover truthfully, and in this walkthrough the planner chose the light-to-dark campaign. Finding better sequences automatically is what the optimization layer is for; see the EDGEBIC optimizer guide for how multi-run search evaluates alternative orderings with a guarantee of never doing worse than the baseline schedule.

Where to Start in Your Own Shop

You do not need to matrix the whole plant. Start with the one work center where sequence visibly hurts: the paint booth, the heat-treat furnace switching alloys, the extruder switching resins. Measure the from-to times you already know by heart, enter them in the Setup Matrix tab on that work center, and reschedule. The plan gets longer and truer on day one; the resequencing win comes the first time you group a campaign.

This post is part of the EDGEBIC worked examples series, where every claim traces to dated, hour-by-hour schedules. The same discipline runs through the rest of EDGEBIC: finite capacity, real constraints, numbers you can check. User Solutions has been doing this for manufacturers since 1991, from job shops to the US Navy. If your booth, furnace, or press is losing shifts to changeovers the schedule cannot see, bring your changeover times to a demo and we will build all three of these schedules with your data.