The curing curve, without touching the oven

The problem

The curve that decides whether the varnish is cured disappears every cycle.

Replacing the oven costs over half a million euros and weeks of line downtime, so today nobody touches it or connects it. Its panel shows everything the quality lead would need to correlate the curing curve with migration results or complaints — but that data dies every cycle when the screen rotates.

• The varnish curing oven is typically 12 to 20 years old, isolated from the network, and its panel is the only place the temperature profile by zone is shown.
• That profile decides whether a food-contact interior varnish is properly polymerized: a run cured under profile can look fine and fail the migration test weeks later. By then the cans have been formed, filled and shipped by the packer client.
• Replacing the oven costs over €500,000 and weeks of line downtime, so nobody connects it and nobody touches it.
• When the screen rotates, the curve is gone. The quality lead is left without the one datum needed to correlate a complaint or a migration result with a specific run.

How it fits the IRIS system

Connect in photo mode on the oven panel — read from outside, nothing installed inside.

Connect photo mode on the panel. A fixed industrial camera points at the panel; an anchored-LLM OCR interprets each frame, structures the time curve per run, and cross-references it with the ERP to know which run was in the oven at each moment.

The curve already exists; it is just never kept. Reading it from outside turns a screen that refreshes every cycle into a history per run, and leaves the oven exactly as it was commissioned. It is the cheapest way to get the most important process datum of a food-contact can out of a machine nobody wants to touch.

See the full IRIS architecture →

Before and after

The curing oven as it is today versus the curing oven read by camera

Aspect	Today	With iLEAN Connect
Temperature profile by zone	Visible only while on screen	Stored per run, cycle after cycle
Which run was in the oven	Inferred from the shift report	Crossed with the ERP automatically
Migration result vs. curing curve	No correlation possible	Root cause per run in minutes
A run cured below profile	Found at the migration test	Flagged against the reference curve
Oven controls and software	—	Untouched, nothing installed
Oven replacement	A €500,000+ line in the plan	Deferrable indefinitely

Zero curve-to-complaint correlation → root-cause analysis per run in minutes. Oven replacement becomes indefinitely deferrable.

Impact estimate

Impact estimate — to be validated with your numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 5-10 months.
• Avoided oven replacement capex, which runs over €500,000 plus weeks of line downtime.
• Fewer migration nonconformities, through the curve-to-result correlation that cannot be made today.
• And a complaint investigation that starts from the actual curing curve of that run, not from memory. With a few months of curves, quality can also see drift in a zone before it shows up as a migration result out of spec.

Estimated payback of 5 to 10 months, avoiding oven replacement CAPEX (>€500,000) and reducing migration non-conformities through curve-to-result correlation. Estimate to validate.

And the fair question from the production manager

“Can a camera reading an old panel be trusted for food contact?” — the task is anchored: fixed screen, temperature fields in known positions and physically bounded ranges per zone, where the best models drop below 1.5% error [1]. Any frame whose value falls outside the possible range is discarded instead of stored, and the quality lead validates the curve before it is tied to a migration result.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.