The IQF Freezing Tunnel Starts Talking

The problem

The step that decides food safety is the one that leaves no record.

The IQF tunnel is the critical step: if the temperature curve and residence time aren't right, the product core doesn't freeze in time and the batch's food safety is in question.

• The IQF tunnel is the critical step: if the temperature curve and the residence time are not right, the core of the croquette does not freeze in time and the batch's food safety is in question.
• A gourmet range makes that harder, not easier: a 20 g bite-size appetizer and a cylinder with a high-fat Iberian ham filling do not freeze the same, and the tunnel gets reset several times a day.
• The panel shows zone temperature and residence time live, but the manufacturer keeps the machine off the network and nothing ever leaves that screen.
• So the food-safety record is whatever somebody copied onto a sheet once a shift, and nobody can answer afterwards what the tunnel actually did while one specific lot was inside it. Which is the one question asked when a distributor returns a pallet of clumped product, or when the auditor picks a batch at random and wants its freezing evidence.

How it fits the IRIS system

Connect in photo mode on the panel — solved from outside the machine, not inside it.

A camera points at the tunnel's proprietary panel. iLEAN Connect reads every screen with OCR and an anchored LLM, without wiring anything into the original control or exposing it to the network. The temperature curve and residence time land in central memory, cross-referenced with whatever SKU was passing through the tunnel at that moment.

Nothing is wired into the original control, no port is opened and no software is installed on the tunnel. On the asset that carries the plant's food-safety argument, that is the difference between a case quality approves in a week and a project it never authorizes at all. The tunnel keeps running exactly as it did yesterday; what changes is that it now leaves a trace.

See the full IRIS architecture →

Before and after

Today's tunnel versus the tunnel being read

Aspect	Today	With iLEAN Connect
The freezing curve	On screen, then gone	Digitized run by run
Residence time per SKU	Set and forgotten	Recorded against the active SKU
Food-safety record	One reading a shift, by hand	Continuous, per batch
A complaint about clumped product	No way back	The curve of that exact lot
The tunnel's own control	—	Untouched and off the network
Replacement CAPEX	In the plan	Deferrable

Freezing curve visible only on-screen, never logged → Curve digitized and cross-checked against the SKU in production.

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, combining the avoided tunnel-replacement CAPEX with stronger food-safety documentation per batch.
• Freezing evidence per batch instead of one handwritten reading per shift.
• Drift in the tunnel visible as a trend, before it becomes a batch on hold.
• And a curve per SKU, which is what a range of 18 formats needs in order to be frozen properly.

Payback 5-10 months · avoids tunnel replacement CAPEX. Estimated payback is 5 to 10 months, combining avoided tunnel-replacement CAPEX with stronger food-safety documentation per batch. Estimate to validate.

And the fair question from the production manager

"Does a camera on a panel hold up as a food-safety record?" — the task is anchored: a fixed screen, fields in known positions and physical ranges known in advance, where the best models drop below 1.5% error [1]. On top of that, any frame whose value falls outside the possible range for the tunnel is discarded instead of accepted, which is exactly what would invalidate the series in front of an auditor.

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