NordStrip Steel — 38% fewer off-gauge metres
A five-stand tandem mill with chronic tail-end off-gauge on automotive grades. Roll & Gauge in shadow, then assist, then bounded autonomy over 14 weeks.
Case studies
What actually changed on the line
Three deployments, with the baseline methodology, the failure modes we hit and the numbers we can publish. Full versions, including the ones that did not go smoothly, are available under NDA.
Baselines agreed before deploymentMethodology auditablePublished with permission
exit gauge1.982 mm
flatness I-unit3.1 IU
F5 roll force18.4 MN
interstand tension9.6 MPa
chatter margin0.24 · safe
anneal energy-11.4%
pass 6/7autonomy: assisted-writetwin delta 0.004 mmaudit id 7f31c
Built with steel and flat-metals producers across four continents
NordStrip Steel
Kalyon Metals
Ferrata Group
Aurum Flat Rolled
Meridian Coil
Volkan Çelik
Index
Three published cases
Each one starts with what was broken and ends with what it cost to fix.
Ferrata Group — 1.7 points of prime yield
A galvanizing line over-coating to stay clear of spec, with property misses on high-strength grades. Temper & Coat plus Yield & Energy.
Aurum Flat Rolled — the retiring roller
A hot-strip mill facing the loss of both engineers who could set up hard automotive grades. Schedule & Flow and the twin, deployed for succession as much as yield.
Case one
NordStrip Steel: the tail end nobody could fix
Automotive and appliance grades, five stands, and a loss everyone had stopped calling a problem because it had always been there.
The problem
Every coil lost metres at head and tail to off-gauge. The AGC reacted well but reacted late, and threading and tail-out are exactly where reaction is too slow. The mill had absorbed the loss into its yield expectation years earlier.
- Baseline12 weeks of production measured before deployment, agreed with the plant's own finance team.
- ScopeRoll & Gauge only. No twin, no other modules, deliberately.
exit gauge1.982 mm
flatness I-unit3.1 IU
F5 roll force18.4 MN
interstand tension9.6 MPa
chatter margin0.24 · safe
anneal energy-11.4%
pass 6/7autonomy: assisted-writetwin delta 0.004 mmaudit id 7f31c
What we did
Shadow mode for eight weeks, predicting gauge deviation from entry profile, roll thermal state and lubrication behaviour. Then assist mode for six weeks with every recommendation approved on the pulpit. Bounded autonomy from week 14, with a ±0.05 mm gap envelope.
- Acceptance rate94% of recommendations approved unmodified in the final four weeks of assist.
- What went wrongTwo weeks lost to a historian tag mapping error that made the first shadow results look better than they were. We found it; it cost trust we had to rebuild.
exit gauge1.982 mm
flatness I-unit3.1 IU
F5 roll force18.4 MN
interstand tension9.6 MPa
chatter margin0.24 · safe
anneal energy-11.4%
pass 6/7autonomy: assisted-writetwin delta 0.004 mmaudit id 7f31c
The result
Off-gauge metres per coil fell 38% against baseline. Two chatter events were prevented in the first quarter of autonomy, each historically worth a shift of lost production. Gauge standard deviation on the body of the coil fell from 2.9 µm to 1.8 µm.
- PaybackUnder six months on the Line subscription at the plant's own coil economics.
- NextRoll & Chatter added in the following quarter; the second line is in shadow now.
exit gauge1.982 mm
flatness I-unit3.1 IU
F5 roll force18.4 MN
interstand tension9.6 MPa
chatter margin0.24 · safe
anneal energy-11.4%
pass 6/7autonomy: assisted-writetwin delta 0.004 mmaudit id 7f31c
Case two
Ferrata Group, in numbers
A continuous galvanizing line that had been systematically over-coating for a decade because nobody could safely hold the target closer to spec.
1.7 ptsPrime-yield uplift on the galvanizing line
11.4%Reduction in reheat and anneal energy at constant throughput
3 g/m²Average coating-weight over-spend eliminated
1 qtrFrom first write to bounded autonomy
[PLACEHOLDER] Baseline methodology agreed contractually and audited by the customer's finance function.
Case three
Aurum Flat Rolled: buying autonomy for succession
Sometimes the ROI is not the ROI.
Aurum's hot-strip mill ran hard automotive grades that only two people could set up reliably. Both were within three years of retirement. The plant director's problem was not yield — it was that his mill was two resignations away from being unable to make its most profitable product.
Schedule & Flow and the mill twin went in together. The twin was calibrated on 412 historical coils across four grade families, then run in shadow against every new coil for ten weeks. Where its schedule differed from the senior engineer's, both were recorded, and the difference was reviewed every Friday with the engineer in the room.
That review is the actual product. Over ten weeks the disagreements narrowed, and the ones that remained were mostly cases where the engineer was compensating for a stand behaviour the twin had not yet learned. Feeding those back is how the craft got encoded.
Today the twin plans the schedule for those grades, the engineer approves it, and a junior engineer can run the setup with the twin's reasoning on screen. Passes per coil fell on several grades. But the plant director bought it for the succession problem, and that is how he measures it.
- Twin calibration412 historical coils across four grade families.
- Shadow period10 weeks with weekly disagreement review.
- OutcomeSetup craft captured, versioned and explainable to the next generation.
Methodology
How every number on this page was produced
If you cannot audit the baseline, the result is marketing.
| Step | What happens | Who signs it |
|---|---|---|
| Baseline definition | Metrics, measurement points and exclusions agreed in writing | Customer finance + Steelira |
| Baseline period | 8–12 weeks of production measured with Steelira read-only | Customer operations |
| Attribution rules | How grade mix, tonnage and campaign changes are normalised | Customer finance |
| Measurement | Same instrumentation, same calculation, post-deployment | Both parties |
| Audit | Customer or third party may audit the calculation at any time | Customer's choice |
Outcome-based commercial components use exactly this methodology. That is why we are careful about it.
Pattern
What the successful deployments have in common
Four things predict success better than anything about the mill itself.
- STEP 01
A named champion
A rolling or quality engineer who wants this to work and has the standing to say so publicly.
- STEP 02
An agreed baseline
Written down, signed, and boring. Every dispute we have ever had traces to a vague baseline.
- STEP 03
Weekly review in person
Someone from Steelira in the room with the engineers, every week, through shadow and assist.
- STEP 04
Willingness to promote
A plant that will actually turn on autonomy when the gate is cleared, rather than piloting forever.
Voices
From the three cases
“We ran Steelira in shadow for eight weeks on the tandem mill. It called eleven of the twelve off-gauge tail events before our AGC reacted. That was the moment the argument ended.”
“Our best roller retires in two years. Steelira is the first system that captured how he sets up a hard automotive grade — and then explained why, with the coil history to back it.”
“Prime yield moved 1.7 points on the galvanizing line and reheat gas fell double digits. Nothing else we bought this decade did both.”
Security
OT-safe by construction
Steelira writes to mill equipment, so security and safety are product requirements, not paperwork.
SOC 2 Type IIISO 27001IEC 62443 alignedGDPRSSO / SAMLData residency
- Segmented OT pathEdge runtime sits in a Purdue Level-2/3 DMZ with signed, rate-limited write-back and hard setpoint clamps.
- Deterministic fallbacksEvery control loop degrades to the mill's existing Level-2 setup on watchdog timeout — the mill never stops because Steelira does.
- Grade-IP isolationRecipes, grade specs and coil histories are tenant-scoped, encrypted at rest with per-tenant keys, and never used to train another producer's models.
- Assurance-grade auditEvery perception, recommendation, approval and actuation is written to an immutable, exportable log keyed to coil ID.
Get the unredacted case studies
The published versions leave out the grade specifics, the failure modes and the commercial detail. The full versions are available under NDA, including the deployments that took longer than they should have.