The foam machine panel, digitized without touching the machine

The problem

The machine that sets the cabinet's power draw keeps no memory of its cycles.

The foam machine defines the cabinet's insulation and, with it, the power draw and energy rating of the finished unit. It's usually a critical asset 10 to 20 years old, with a proprietary panel isolated from the network, showing pressure, mold temperature and cure time that nobody logs. If a lot comes out with the foam cycle off spec, the symptom — high power draw, failing energy-efficiency test — shows up weeks later, with no way to correlate it to the cycle that caused it.

• Polyurethane foaming defines the cabinet's insulation, and with it the power draw and the energy rating of the finished cooler. It is one of the most critical steps on the line. A cabinet with poor foam density or voids looks exactly like a good one once the liner and the outer skin are closed.
• The foam machine is usually 10 to 20 years old, with a proprietary panel isolated from the network. Pressure, mold temperature and cure time are on screen and nobody logs them.
• When a lot is foamed off spec, nothing visible happens on the line. The symptom — high power draw, a failed energy-efficiency test — appears weeks later.
• By then there is no way to correlate the failing units with the foam cycle that produced them, so the root-cause discussion turns into opinions about the machine. The usual outcome is a general tightening of the process window, which costs material and cycle time on every lot.

How it fits the IRIS system

Connect in photo mode on the foam panel — read from outside, the machine stays untouched.

Connect photo mode on the panel: a fixed industrial camera points at the panel, OCR plus a grounded language model reads each frame and structures the foam cycle by lot, linked to the serial numbers of the cabinets that came out of it.

Nobody opens the foam machine's controller, nobody asks the original manufacturer for a protocol and nothing is installed on the machine. A camera reads what the operator already sees, and every foamed cabinet inherits the cycle that insulated it. On the machine that defines the energy rating of every cooler, that is what makes the case approvable without a capex request.

See the full IRIS architecture →

Before and after

The silent foam machine versus the foam machine being read

Aspect	Today	With iLEAN Connect
Pressure, mold temperature, cure time	On the panel, never logged	Structured cycle by cycle
Foam cycle ↔ cabinet serial number	No link	Each cabinet carries its cycle
Unit failing the energy test	Weeks later, unexplained	Traced to its foam cycle
Root-cause analysis	Opinions about the machine	Minutes, by lot
The foam machine's controller	Closed and proprietary	Untouched, no ports opened
Replacing the foam machine	In the capex plan for data	Not needed to get the data

Zero correlation between foam cycle and power draw → root-cause analysis by lot in minutes.

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.
• Fewer coolers rejected in energy-efficiency testing, because a cycle drifting off spec becomes visible while it is happening. Each rejected unit at energy testing is a cooler that was fully built, charged and tested before anyone knew.
• From zero correlation between foam cycle and power draw to root-cause analysis by lot in minutes.
• And the foam machine keeps running as it is: getting its data does not depend on a replacement project. The replacement can be planned on its merits, not because the data is missing.

Estimated payback of 5-10 months, cutting units rejected in energy-efficiency testing. Estimate to validate.

And the fair question from the production manager

“Can a camera on an old panel really be trusted for the energy rating?” — reading gauges and displays in fixed positions with known physical ranges is an anchored task, where the best models drop below 1.5% error [1]. And a frame whose pressure or temperature falls outside the machine's possible range is discarded, not stored, so it never contaminates the series.

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