The vintage molder panel that starts talking

The problem

The machine that makes the most critical part keeps no memory.

The vintage molder often produces the most critical plastic or aluminum component in the product. Its panel shows temperature, pressure and cycle time, but that data dies at every screen refresh. Replacing the machine costs hundreds of thousands of dollars and weeks of downtime.

• The large-part molders in a white-goods plant — the ones shooting washer tubs, refrigerator liners or dryer drums — are often 15-20 years old and run a proprietary panel with no network output.
• That panel shows melt temperature, injection pressure, hold time and cycle time. Every value disappears at the next screen refresh, and only the operator ever sees it.
• So when a batch of tubs comes out warped or short, nobody can say whether the cycle drifted, which shift it was or which resin lot was loaded in the hopper.
• Replacing the machine means hundreds of thousands of dollars and weeks of downtime on the part with no second source, which is why the decision gets postponed year after year with no data behind it.

How it fits the IRIS system

Connect over the panel — the molder stays itself, but now it talks.

Fixed industrial camera over the panel + anchored- LLM OCR. iLEAN interprets every frame, structures the cycle curve by batch and cross-references it with the active ERP order, with zero physical changes to the molder.

Nothing is installed on the molder, its controller is not touched and no port is opened. The camera reads what the operator already reads, and that is enough to put every cycle next to the ERP order it belongs to. For maintenance and for the machine's builder, nothing has changed.

See the full IRIS architecture →

Before and after

The silent molder versus the molder that reports

Aspect	Today	With iLEAN Connect
Cycle curve per shot	Lost at each refresh	Stored and structured by batch
Link to the ERP order	None	Cross-referenced automatically
Warped or short tubs	Cause unknown	Root cause by batch in minutes
Original controller	—	Untouched, no ports opened
Machine replacement	Under discussion every budget	Postponable indefinitely
Downtime to install	—	Mounting a camera bracket

Zero cycle↔rejection correlation → root-cause analysis by batch in minutes. Machine replacement indefinitely postponable.

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, mostly from avoided replacement CAPEX on the molder.
• Cycle-to-rejection correlation that is impossible today: root-cause analysis by batch in minutes instead of arguments in the morning meeting. A short shot on a refrigerator liner can finally be traced to the hold pressure that caused it.
• Replacement of the molder becomes a choice made with data, not a necessity forced by the lack of it.
• And scrap on the most expensive injected part of the product — the tub, the liner, the drum — stops being an unexplained number at month end.

Estimated payback 5-10 months, avoided replacement CAPEX. *Estimate to validate*.

And the fair question from the production manager

"Can a camera really read a 20-year-old screen?" — the panel is fixed, the fields sit in known positions and each value has a physical range, so it is an anchored task where the best models drop below 1.5% error [1]. Any frame whose reading falls outside the possible range is discarded instead of stored, so a glare or a hand in front of the screen never becomes a false cycle. Process engineers see the same values the operator sees, only now with a history.

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