Model-predictive gauge control
Anticipates deviation from entry profile, roll thermal state and lubrication behaviour instead of reacting to the exit gauge.
Features
Everything the mill needed and never had
Predictive gauge, anticipated shape, surface classified at line speed, properties predicted before the lab, chatter caught before the cobble, and energy optimised against the schedule that is actually running.
Predictive, not reactiveExplainable to a metallurgistBounded by your envelope
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
Feature set validated on hot, cold and coating lines
NordStrip Steel
Kalyon Metals
Ferrata Group
Aurum Flat Rolled
Meridian Coil
Volkan Çelik
Control
Control features
The loops that decide whether a coil is prime.
Predictive shape control
Bending, shifting and selective cooling driven by a predicted shape profile through acceleration and grade change.
Inter-stand tension control
Tension trimmed for stability and gauge simultaneously, with looper coordination on hot mills.
TMCP path planning
Finishing and coiling temperature targets set per chemistry to land microstructure in the property window.
Skin-pass elongation control
Elongation held against a predicted property outcome, not a fixed recipe.
Coating weight targeting
Air-knife and line-speed control that holds coating just inside spec across width and length.
Perception
Sensing and prediction features
What the mill can finally see.
Strip-surface defect classification
Scale, scratches, slivers, inclusions, roll marks and coating faults classified by family, severity and coil-length position at line speed.
Flatness and shape sensing fusion
Shape-meter, thermal and vision data fused into one shape state that survives sensor dropout.
Property prediction
Yield strength, tensile and elongation predicted per coil from chemistry and the realised thermal path.
Chatter precursor detection
Third-octave and fifth-octave precursors detected from vibration and drive current with an explicit safety margin.
Roll wear and thermal crown
Wear and crown estimated continuously so roll changes are scheduled from state, not calendar.
Cobble risk scoring
Composite instability score across looper, tension, temperature and shape, surfaced before the operator would feel it.
Intelligence
Features that compound
The parts that get better every coil.
Correction-driven learning
Every time an engineer overrides a recommendation, Steelira captures the state, the proposal, the override and the outcome. That triple is the highest-value training data in the plant and it exists nowhere else today.
Similar-coil retrieval
“When did we last see this defect on this grade?” answered in seconds from pgvector over your own history.
Grounded metallurgy Q&A
Ask about a grade, a standard or a schedule and get a cited answer from your documents, not a guess.
Golden-dataset evaluation
Every model and prompt change is gated in CI against a golden dataset of your coils. Nothing reaches a mill because it looked better in a demo.
Per-mill memory
The system learns your stands, your rolls and your habits — and keeps that knowledge tenant-scoped.
Anonymised cohorts
Enterprise tiers benchmark against comparable mills without exposing a single coil.
Comparison
Point tools versus a closed loop
The honest version of the competitive picture.
| Dimension | Level-2 / AGC vendor | Surface inspection vendor | Steelira |
|---|---|---|---|
| Scope | Setup and reactive control | Detect and flag defects | Schedule → roll → inspect → anneal → coat |
| Loop | Setpoints held | Reporting only | Closed perceive-decide-act at the mill edge |
| Prediction | Model-based setup | None | Twin-predicted gauge, flatness and properties |
| Acts on the mill | Yes, within its own loop | No | Yes, across loops, inside your envelope |
| Learns from corrections | No | Limited labelling | Every override becomes training signal |
| Data moat | Vendor-held models | Image archive | Per-tenant mill-and-metallurgy record |
Incumbents are excellent at what they do. Steelira is the layer that makes them act together.
Assurance
Features that let you sleep
Autonomy in a hot mill is a trust problem before it is an AI problem.
- STEP 01
Safe operating envelope
Hard per-setpoint, per-grade, per-line clamps enforced in the runtime and signed by your engineers.
- STEP 02
Human-in-the-loop gates
Configurable approval checkpoints per loop and per autonomy level, with full attribution.
- STEP 03
Deterministic fallback
Watchdog timeout hands control straight back to your existing Level-2 setup.
- STEP 04
Immutable audit
Every perception, plan, approval and actuation logged against coil ID and exportable for claims.
Operations
Day-two features
The unglamorous things that decide whether a deployment survives its first year.
Signed OTA edge updates
Staged rollout rings, per-line rollback and fleet health for every mill edge node you run.
Observability
OpenTelemetry traces, Grafana dashboards and Langfuse traces for every agent decision path.
SSO, SCIM and RBAC
Role separation between operator, engineer, quality and admin, enforced end to end.
Offline tolerance
The edge keeps controlling and buffering through WAN loss; nothing about control depends on the cloud.
Export everything
Your coil records, models metadata and audit logs are exportable at any time, in open formats.
Named support
A mill success engineer and metallurgist assigned to the account on Plant and Enterprise tiers.
Proof
Features in production
“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.”
Integrations
Speaks the mill's language on day one
Pre-built connectors to the systems already running your line — read for perception, write for control, with the same audit trail on both.
Level-2 & control
- Primetals Level-2
- SMS X-Pact
- Danieli Automation
- TMEIC mill drives
- ABB Ability
- Siemens SIMATIC / PCS 7
Gauge, shape & surface
- IMS X-ray gauges
- Vollmer thickness
- Shape-meter rolls
- ISRA Vision surface
- Parsytec inspection
- Thermal line scanners
Plant systems
- Level-3 MES / QMS
- SAP PP-PI
- OSIsoft PI / AVEVA
- OPC UA & Classic
- Historians (IP.21, Canary)
- CMMS / EAM
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.
See these features on your coils
The fastest way to evaluate Steelira is shadow mode: six weeks, read-only, scored against your own production.