Starting a partly filled furnace keeps wafers moving; waiting may fill the batch or reach a cheaper electricity period. AIMS5.0’s 22 October update describes genetic programming that learns dispatching rules for this choice: candidate formulas are tested and recombined, with better-performing versions retained.[1][2]
A related 29 May paper by Daniel Schorn and Lars Mönch models two processing stages. It balances electricity cost with lateness weighted by each job’s priority. Wafers have time limits between operations because excessive delays can cause oxidation or contamination. Searching for rules happens offline; applying them is comparatively quick.[2]
Consider this invented planning example, not a project result. Two compatible lots fit one furnace run. Lot A is ready at 10:00 and must start by 11:00; lot B is ready at 10:45. The tariff falls at 10:30. Ignore warm-up changes and other work in the queue.
Three possible start times
| Choice | What happens |
|---|---|
| Start at 10:00 | A proceeds alone before the lower tariff begins. |
| Wait until 10:45 | A and B fit together; A still meets its latest start. |
| Wait until 11:15 | Both are ready, but A’s processing limit is missed. |
The paper tests generated scheduling problems, rather than reporting a measured factory electricity saving. Its heuristics can leave time-limit violations, and it assumes a specified tariff pattern. Cheaper operation alone cannot establish that fewer kilowatt-hours were used.[2]
Sources & further reading
- Scheduling with the energy bill in mind: AIMS5.0 use case 9 AIMS5.0 · 22 Oct 2026 · Accessed 29 Oct 2026
- A genetic programming approach to solve energy-aware flexible flow-shop scheduling problems with batching and time constraints Flexible Services and Manufacturing Journal · 29 May 2026 · Accessed 29 Oct 2026