A caster sequence, a rolling programme and a paint line campaign obey completely different rules — and none of them are "minimise makespan". We build a scheduler per unit, on a shared model, so the whole route stays connected.
A tundish has a life. A ladle is either available or it is not. The scheduler treats physics as hard and commercial priority as soft — never the other way round.
When a caster goes down, the question is what to do with the twelve heats already tapped. The answer needs to arrive before the next one does.
Auto-sequence proposes, the planner disposes. Manual overrides are first-class and are recorded with a reason, not fought against.
Heat-level sequencing through EAF or BOF, ladle furnace, degasser and caster — the tightest coupling in the plant, and the one where a mistake costs a skulled ladle.
Balanced across furnaces rather than filled first-come.
Backwards from the caster, not forwards from the scrap bay.
Re-sequence proposals with the downstream impact shown before you commit.
Step sizes constrained by mould and segment settings, with flying-width-change capability where the caster supports it.
Sequence length capped by tundish age and expected wear, with planned change points rather than surprise ones.
Compatible grades cast back-to-back; transition slabs identified and dispositioned automatically.
Cut plans that maximise order coverage instead of producing well-formed stock nobody ordered.
Sequence building that treats the cast as the atomic unit. Because once you start a sequence, your options for the next four hours are largely gone.
The interesting tension in caster scheduling is between the long, clean, cheap sequence and the order book that wants six widths and four grades this week. We surface that trade-off explicitly — cost of transition against cost of lateness — rather than burying it in a weight in an objective function nobody can explain.
Longer sequences where the order book permits, shorter where lateness costs more.
Priced in tonnes of downgrade, so the trade-off is in a unit the plant understands.
Hot strip and plate rolling programmes that respect what the rolls are doing to themselves while they work.
Work rolls wear and pick up thermal crown as a campaign progresses, and that wear prints itself onto the strip. So width can only step down within a campaign, the hardest and widest material goes early after the warm-up coils, and every programme ends at a roll change. The scheduler builds programmes that fill that envelope while still hitting due dates — the whole craft is in doing both at once.
Downstream is where the order book gets specific and the campaign rules get awkward. Each line has its own idea of what a good run looks like.
Reduction schedules linked back to hot band availability, with width and gauge grouping to limit pass-line changes.
Coating-weight blocks and pot chemistry campaigns — the two things that decide how much you can reshuffle on the day.
Colour campaigns are the classic case: light to dark, minimise flushes, and never let an urgent order justify a colour change that costs an hour of line time.
Press and hammer loading where the furnace, not the press, is usually the real constraint — and where a die change can cost more than the job it enables.
The modelling approach transfers. Tell us the unit, its campaign rules and what a bad sequence costs you, and we can usually scope it in a workshop.
A perfect caster sequence against an impossible monthly plan is still a late order. Our planning layer sets the horizon the schedulers execute inside — and takes feedback when reality disagrees.
The fastest way to evaluate a scheduler is to point it at the week that went wrong. Send us a real order book and a real constraint set, and we will show you what it produces.