The dryer that now talks

The problem

The variable that decides the product leaves no trace.

replacing the dryer would cost hundreds of thousands of euros and weeks of downtime. Meanwhile, quality can't correlate moisture with issues because the data is never logged.

• Moisture decides almost everything downstream: whether the mill clogs, whether the powder cakes in the sachet, whether the product holds its best-before date, whether mold risk appears in the jar three months later.
• The dryer or the mill that governs it is often more than a decade old, with a proprietary controller the manufacturer keeps off the plant network by design. Replacing it costs hundreds of thousands of euros and weeks of downtime the season will not allow.
• Its screen shows the curve in real time, and that curve dies with every screen change. What survives is a final figure written into the logbook, if anyone writes it, and a figure is not a curve.
• So when a customer complaint arrives, quality cannot correlate the incident with what happened in the dryer, because the data was never logged in the first place. The analysis stops before it starts.

How it fits the IRIS system

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

Connect photo mode on the panel + OCR with a grounded LLM. Each reading is structured and cross-referenced with the batch being processed at that moment.

It does not integrate with the controller, does not open ports and installs nothing on a machine that has been running for a decade. The camera is mounted, aimed and calibrated, and the equipment never learns it is being read. On the machine the whole plant depends on, that is the difference between an approvable case and a project maintenance never authorizes.

See the full IRIS architecture →

Before and after

Today's dryer versus the dryer being read

Aspect	Today	With iLEAN Connect
The moisture curve	Dies with the screen change	Digitized batch by batch
What survives a drying run	A final figure in the logbook	The whole curve, tied to the origin lot
Moisture-to-incident correlation	Impossible	Root cause in minutes
Caking in the sachet	Diagnosed by guesswork	Traced back to its drying run
The controller's software	—	Untouched, no revalidation
Replacement capex	On the plan	Deferrable

zero moisture↔issue correlation → root-cause analysis by batch in minutes.

Impact estimate

Estimated impact — to validate with your own 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, with the avoided capex of replacing the dryer on top.
• Root-cause analysis by batch in minutes instead of a correlation nobody in the plant is currently able to make at all.
• Drying drift detected as a trend over successive runs, before it turns into an out-of-spec batch or a caking complaint three months later.
• And the equipment keeps running exactly as it is: no weeks of downtime in the middle of the campaign and no hundreds of thousands of euros brought forward in the investment plan.

estimated payback of 5 to 10 months, also avoiding equipment replacement CAPEX. *Estimate to validate*.

And the fair question from the production manager

"Does a camera pointed at a screen hold up in front of an auditor?" — here the task is anchored: a fixed panel, fields in known positions and physically bounded ranges, where the best models drop below 1.5% error [1]. And the system discards any frame whose moisture value falls outside the possible range instead of accepting it, which is exactly what would invalidate the series for a certification body.

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