The vintage multi-spindle lathe that now "talks"

The problem

The most expensive lathe in the plant forgets everything every cycle.

the multi-spindle lathe is the most expensive asset and the line's bottleneck. Its panel shows everything a machining supervisor would need to correlate cutting parameters with wear and out-of-spec parts — but that data dies on screen every cycle.

• A bar turning fleet often runs multi-spindle lathes that are 10 to 20 years old, with proprietary panels and no network port, because they still turn good parts and nobody wants to touch them.
• That panel shows RPM, feed rate, spindle load and estimated insert wear — exactly what a supervisor needs to explain an out-of-spec thread or a diameter drifting toward its limit.
• The data dies on screen at the end of every cycle, so nobody can correlate a cutting parameter with the parts that failed, and every root-cause discussion starts from opinions.
• Replacing one of those lathes costs hundreds of thousands of euros and weeks of downtime on the machine that is already the line's bottleneck.

How it fits the IRIS system

Connect in photo mode on the panel — the lathe stays itself, but now it talks.

Connect photo mode on the panel. A fixed industrial camera plus LLM-grounded OCR reads every value and structures a time series per lot/spindle, cross-referenced with the ERP.

Nothing is wired into the control and no software is touched. The camera reads what the operator already sees, and that is what turns fleet renewal from an urgency into a decision you can schedule — with data per spindle instead of a hunch about which machine is worn out.

See the full IRIS architecture →

Before and after

The silent multi-spindle versus the lathe on record

Aspect	Today	With iLEAN Connect
RPM, feed and spindle load	Visible, never stored	Time series per lot and spindle
Insert wear trend	The operator's feel	A curve you can query
Out-of-spec thread or diameter	Cause unknown	Root cause per lot in minutes
The lathe's control	—	Untouched, no ports opened
Fleet renewal	Pressed by lack of data	Deferrable with full visibility
Link to the ERP lot	None	Cross-referenced automatically

zero correlation between cutting parameter and out-of-spec part → root-cause analysis per lot in minutes. Fleet renewal becomes deferrable without losing visibility.

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, counted on avoided capex and scrap together.
• Avoided lathe-renewal capex: the machine keeps producing and stops being a blind spot in the fleet.
• Scrap reduction from correlating cutting parameters with defects, which is impossible today because the parameters are never stored.
• And root-cause analysis per lot drops from guesswork to minutes, which is what the OEM measures when a complaint arrives.

estimated payback 5-10 months, avoided lathe-renewal CAPEX, scrap reduction through parameter-to-defect correlation. *Estimate to be validated*.

And the fair question from the production manager

“Can a camera reading an old panel be precise enough?” — the task is anchored: fixed screen, fields in known positions and expected physical ranges for each spindle, where the best models drop below 1.5% error [1]. And any reading outside the physically possible range is discarded instead of stored, so the series stays clean enough to defend in front of a customer.

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