Your most expensive asset, connected without touching it

The problem

The data that explains an out-of-tolerance diameter dies on the screen.

The tube mill is a physical bottleneck and one of the most expensive assets to replace. Its panel holds the information that would explain why a batch comes out with an out-of-tolerance diameter deviation, but that data dies every time the screen changes view.

• The tube mill shapes the strip through its forming stands before the seam is welded. It is often 15-25 years old, with a proprietary panel and zero connectivity.
• Replacing it costs hundreds of thousands of dollars and weeks of downtime, so it stays — and it stays blind. It is the physical bottleneck of the line, and everything downstream depends on how it formed the strip.
• Its panel shows roll pressure, line speed and the readings that would explain why a batch comes out with an out-of-tolerance diameter deviation. Each time the screen changes view, that value is gone.
• So when sizing or final inspection finds ovality or diameter drift, nobody can tie it back to what the forming stands were doing at the time. The 8D ends with a probable cause instead of a proven one.

How it fits the IRIS system

Connect in photo mode on the panel — nothing is wired into the machine.

Connect photo mode over the panel. Fixed industrial camera. OCR with a grounded LLM interprets each reading, structures it into a time series per batch, and cross-references it with the active ERP program.

The machine stays exactly as it is: same control, same panel, same operator. What changes is that every reading it shows becomes a time series tied to the batch and the ERP program, and root cause stops being a guess. On the most expensive asset in the plant, that is the difference between an approvable project and one nobody dares to start.

See the full IRIS architecture →

Before and after

The blind tube mill versus the tube mill being read

Aspect	Today	With iLEAN Connect
Roll pressure and line speed	Visible only while on screen	Time series per batch
Diameter deviation at sizing	Found with no cause attached	Crossed with the forming readings
Root cause for a batch	Days of guessing	Minutes
The original control	—	Untouched, no ports opened
Scrap from out-of-range forming	Counted after the fact	Traced to the stand reading behind it
Replacing the tube mill	On the CAPEX list	Deferrable indefinitely

zero correlation between forming parameters and downstream defects → root-cause analysis per batch in minutes. Machine replacement becomes indefinitely deferrable.

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.
• Avoided replacement CAPEX, which for a tube mill runs to several hundred thousand dollars, plus the weeks of downtime a replacement would cost.
• Less scrap from out-of-range forming, because the reading behind it is finally on record and can be acted on before the next batch.
• Root cause per batch goes from zero correlation between forming and downstream defects to an answer in minutes.

estimated payback 5-10 months, avoided replacement CAPEX (several hundred thousand USD), reduced scrap from out-of-range forming. *Estimate to validate*.

And the fair question from the production manager

“Can a camera reading an old panel be trusted in front of an OEM?” — the task is anchored: a fixed screen, values in known positions and physical ranges your process engineer already knows, where the best models drop below 1.5% error [1]. Any reading outside the physically possible range is discarded rather than stored, so the series you show a customer holds only plausible values. And the operator never has to write a reading down again.

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