Can AI plan remanufacturing runs to cut changeovers?

Yes, for lumber and millwork remanufacturing, where moulders, resaws and rip lines change between profiles and sizes. Grouping runs of the same or similar products onto one setup, and sequencing the rest so each change is small, is a well-studied planning problem. Optimisation handles it well. The gap, in our experience, is the record: how many setups a line did, how long each took, what product it changed from and to, and which changes are cheap or expensive. Measure setups for a month and build a matrix of setup times between products. Then plan open orders against it, with delivery dates as a hard limit and an eye on the stock that grouping builds.

What the usual answer says

The usual answer says AI can group similar products, optimise sequences and plan around tooling to reduce changeovers. Better versions work from a table of changeover times between products, and one scheduler vendor describes refining it from history: "if recent batches of the same kind of changeover took 2 hours, the digital twin can adjust your scheduling to reflect this reality". That assumes the history exists.

The method is mature

Planning with setup times is one of the oldest topics in scheduling. Kuo-Ching Ying and colleagues, in a 2025 review that we read as its abstract, note that "the role of setup times in production planning and control was recognised in the late 1960s". Their review considered "over 2100 articles published between 1986 and 2024" and identified 22 research themes. Methods for sequence-dependent setups, where the time to change depends on what ran before, are well developed.

Practice is simpler. A 2024 study of sawmill order scheduling by Francisco Vergara and colleagues, in Maderas, notes that "in most cases static heuristics, such as earliest due date (E), longest processing time (L), and shortest processing time (S), are used because of their simplicity". In its tests, due-date sequencing left only 0.5% of orders late, against 17% for a product-level plan, so simple rules can protect dates well. They take no account of setups, which is our inference from how they work.

What the data can show

From Quarri's own work with a lumber and millwork manufacturer: grouping same-product remanufacturing runs onto one setup would remove about a fifth of measurable setups without missing a delivery date. That figure is identified, not yet banked.

What to record

Record every setup for a month: start and end time, machine, product before and product after, and a short reason code such as knives, profile, species, width or thickness. Where a machine logs its running time, the gaps between runs are a starting point for setup time. Check them against the operators' reason codes, since a gap can also be a break or a breakdown.

From a month of records, average the setup time for each product-to-product change. That matrix is the input the planning needs, and it shows which changes are cheap and which are expensive. Add open orders with product, quantity and promised date, so grouping never pushes an order late.

Then plan, and count the stock

With the matrix and the orders, an optimiser can group same-product lines across orders onto one run and sequence the rest along the cheapest changes. Grouping has a cost. Running an order early, or running more than today's orders need, builds stock. In a shop with many profiles, that stock can sit for months. So the plan should show both sides: setups saved, and stock built ahead of need, valued at cost. A cap on how far ahead the plan may run keeps the trade visible. A planner reviews and releases it, and it can be re-solved each day as orders arrive.

Track two numbers afterwards: setups per thousand board feet produced, and stock built ahead, both by machine and week. Together they show whether grouping is paying or just moving cost into the warehouse.

AI also helps read the inputs: orders that arrive by email, and setup reasons typed in free text by operators, sorted into consistent codes.

When it doesn't apply

Lines running one product for weeks have few setups to save. Shops where every order is a custom profile with its own knives may find little to group, though sequencing by species and dimension still helps.

Quarri for wood products is built around how a wood products plant runs, where the order book meets real capacity.

Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.