Foundries: Scheduling Furnace Batches and Casting Lots

Foundry lot scheduling comes down to two things the shop already knows physically: a furnace runs one charge per chamber per cycle, and castings must cool before they can be machined. A schedule that packs multiple jobs into one furnace chamber, or sends hot castings straight to the mill, does not match the floor. EDGEBIC by User Solutions runs a furnace as parallel chambers with one job each per day, and models the cooling delay before castings move downstream, so the plan reflects how the foundry actually pours and cools.

For the furnace and cooling mechanisms at the category level, see continuous-process equipment scheduling and lot streaming explained. Pair this with tool-and-die setup-heavy scheduling and chemical processing scheduling. The full map is at how different industries use EDGEBIC.

The two constraints that define foundry scheduling

Foundry work has a rhythm a generic schedule gets wrong. A melt or heat-treat furnace is a batch resource: each chamber takes one charge, runs a cycle, and cannot take another until it is done. A furnace with three chambers is not one machine with three times the capacity; it is three parallel batch units, each running one job. And once metal is poured or heat-treated, the castings are hot: they cannot be handled or machined until they cool, which is a real delay measured in hours.

A schedule that treats the furnace as pooled hours will try to stack several jobs into one chamber, which is physically impossible. A schedule that treats machining as ready the instant casting ends will send hot parts to the mill, which the floor simply will not do. Both errors produce a plan the shop ignores. Modeling the two constraints correctly is what makes a foundry schedule usable.

The furnace: one job per chamber per day

EDGEBIC models a furnace as a work center with multiple instances, one per chamber, and a mode that gives each instance one job per day. When that mode is on and the furnace has more than one chamber, the engine assigns each new job to its own chamber rather than pooling capacity. Three chambers run three different jobs in parallel; a fourth job that day waits for the next open chamber-day.

The engine tracks each chamber individually, so it never double-books a door. When a job is assigned a chamber, it keeps that chamber for the duration, including across days if the cycle spans them, which matches how a charge sits in one furnace door until its cycle completes. This is exactly the behavior a foundry needs from a batch furnace: the plan schedules jobs into chambers the way the shop loads charges into doors.

Consider a heat-treat furnace with three chambers on a day shift. Three jobs arrive Monday morning needing 6, 7 and 5 hours. The engine puts the first job in chamber 1, the second in chamber 2, and the third in chamber 3, all running Monday in parallel. If a fourth job needs the furnace that day, it takes the first chamber that frees up, or the next chamber-day. The plan reads like the furnace log, because it is built on the same one-charge-per-door logic.

The cooling delay: castings cannot move hot

After pouring or heat treatment, castings need to cool before they can be handled and machined. EDGEBIC models this as a transfer delay in hours on the casting step: a flat lag added after the pieces are ready to move, before the downstream machining step is allowed to start.

If a batch needs two hours to cool, you set a two-hour transfer delay, and the engine holds the machining start two hours past the point the castings are ready. This is deliberately distinct from other timing tools. It is not a shift-aware queue buffer, and it is not a calendar-day transit to an outside vendor. It is a flat, hours-based physical delay for cooling, which is exactly what a foundry needs between the furnace and the mill.

A worked handoff: precision castings move from the foundry to machining with a 25-piece transfer batch, an hour of setup, and about 0.8 hours of run per piece, plus a two-hour cooling delay. The engine computes when the first 25 castings are ready, then adds the two hours of cooling, and only then does machining start. Without the cooling delay, machining would be planned to start the moment the castings came off, which the floor would never do. With it, the plan matches the real handling constraint.

Overlapping the lot: transfer batches

Foundries pour in lots, and a large order does not have to finish pouring before machining begins. A transfer batch lets machining start once a set number of castings has accumulated, rather than after the whole lot is poured.

You set the number of pieces that must accumulate before the first sub-lot moves downstream. Machining then begins when that many castings are ready, while the foundry continues pouring the rest of the lot. The cooling transfer delay composes on top, so each sub-lot still waits its cooling time before machining. For a large casting order, this compresses the total time materially, because the mill works the first castings while the foundry is still pouring, instead of the two stages running strictly in series. For the concept on its own, see what a transfer batch is.

There is a sensible guard here: if the transfer batch is set larger than the order quantity, the engine caps it at the order size, so a small urgent order does not wait for pieces that will never exist. A report flags that case so you can lower the batch for that order and get real overlap.

Loading the whole foundry against real capacity

The furnace and the cooling delay sit inside a finite-capacity plan that loads every work center against its real hours. Molding, pouring, shakeout, cleaning and machining each have their own capacity, and the finite-capacity engine will not stack work beyond what a resource can hold. When the furnace is the constraint, which it often is, its chamber-days set the beat for the whole plan, and you can identify and schedule around it as the bottleneck.

For jobs committed to a ship date, the plan can be pulled to finish just in time; for a foundry feeding downstream machining, the goal is usually to keep the furnace loaded and the castings flowing. Either way, the plan reflects the furnace's batch rhythm and the cooling reality rather than an idealized continuous flow. That accuracy is what lets a foundry planner promise a machined-and-finished date rather than just a poured date, because every step past the furnace, the cooling wait and the machining that follows, sits in the same plan against real capacity.

Common mistakes in foundry scheduling

Modeling the furnace as pooled hours. A three-chamber furnace treated as one machine with triple capacity will have jobs stacked into a single door. Use multiple instances with one-job-per-chamber-per-day so each chamber holds one charge.

One-job-per-day set on a single-chamber furnace. The mode only takes effect when the furnace has more than one chamber; on a single chamber it behaves like a normal single machine. Set the instance count to the real number of chambers.

Cooling modeled as a queue or transit. A shift-aware queue buffer or a calendar-day transit does not represent a flat cooling lag correctly. Use a transfer delay in hours for cooling between the furnace and the mill.

A transfer batch larger than the order. It cancels the overlap and behaves like waiting for the whole lot. The engine caps it and flags it; lower the batch for small orders to get real overlap.

Rolling it out

1. Model each furnace as a work center with an instance per chamber, and turn on one-job-per-chamber-per-day.
2. Set the cooling delay as a transfer delay in hours on the casting step, so machining waits for the castings to cool.
3. Set a transfer batch on large casting orders so machining overlaps pouring by sub-lot, with cooling applied to each sub-lot.
4. Confirm the other foundry work centers carry their real shifts and capacity, so the whole plan loads honestly.
5. Run the schedule and read a furnace day, checking that each chamber holds one job and the parallel loading matches your furnace log.
6. Read a casting-to-machining handoff, confirming the cooling delay holds machining and the transfer batch overlaps the lot.

Two neighboring trades run the same cycle-not-piece-rate arithmetic: scheduling technical ceramics around kiln cycles and firing batches and, for burnout and investment casting on a much smaller part, scheduling jewelry and precious metal manufacturing around casting and setters.

Bring a furnace, its chamber count and a casting routing with real cooling times to a demo of metal fabrication scheduling software, and we will build the furnace and cooling plan with you.