SmartMelt
Real-time thermochemical intelligence for EAF, BOF & Induction Furnace steelmaking
SmartMelt fuses first-principles thermochemistry with machine learning to predict hot-metal composition, temperature, and endpoint in real time — so shopfloor operators cut alloy cost, energy, and tap-to-tap time without guesswork.
One model, built from metallurgy — not just data
SmartMelt couples mass-and-energy balance thermochemistry with supervised ML classifiers, trained on real heat data, to give operators a live read on melt chemistry and temperature before the sample even reaches the lab.
Hot metal composition & temperature, minutes before tap
Classification and regression models trained on plant heat logs predict carbon, silicon, phosphorus, sulfur, and bath temperature — flagging off-spec heats before they happen.
- Live composition & temperature classification (C, Si, P, S, Mn)
- Time-series forecasting of FeO, silicon, and bath temperature drift
- Endpoint blow-time recommendation for BOF converters
Turn every heat into a tuning opportunity
SmartMelt closes the loop — recommending alloy additions, oxygen lance profiles, and power input adjustments derived from CFD and thermochemical simulation, not static lookup tables.
- Alloy addition optimization to hit spec at minimum cost
- Energy & electrode consumption benchmarking heat-over-heat
- 3D CFD-informed guidance for lance angle & slag practice
Induction Furnace melting, tuned for scrap-DRI charge mixes
Induction Furnace (IF) shops — widely used across Indian steel industries — run without a converter or refining step, so getting the charge mix and power profile right the first time matters more than anywhere else. SmartMelt models the coreless IF's electromagnetic melt cycle to track bath temperature rise and residual element buildup from mixed scrap and sponge-iron charges in real time.
- Live bath temperature and power-to-melt efficiency tracking through the melt cycle
- Residual/trace element forecasting (Cu, Cr, Ni, Sn) from scrap-DRI blends
- Charge-mix recommendation to hit spec without a downstream refining stage
Live endpoint carbon tracker
Streaming prediction of dissolved carbon through the blow, updated every 15 seconds from off-gas and thermal signals.
Alloy cost model
Least-cost ferroalloy charge mix recalculated per heat.
Thermochemical core
Mass & energy balance solved from first principles, not black-box only.
Plant-data feedback loop
Models retrain continuously on incoming heat data from JSW & AMNS-class plants, closing the sim-to-plant gap.
Built for EAF and Oxygen Steelmaking, at industry scale
SmartMelt's models are tuned separately for each process route — because an EAF heat and a BOF blow behave nothing alike on the shopfloor.
Electric Arc Furnace
Power-on to tap, SmartMelt tracks bath chemistry and energy input to shrink tap-to-tap time and cut electrode + power costs across scrap-to-liquid melting.
Trained on heat logs from EAF shops running DRI-scrap blends, with live slag basicity and FeO tracking to protect refractory life.
| Endpoint temperature accuracy | ±12°C |
| Carbon prediction accuracy | ±0.015% |
| Typical energy saving | 18–25 kWh/t |
| Tap-to-tap time reduction | 4–7 min |
Basic Oxygen Furnace
SmartMelt models the oxygen blow dynamically, predicting dephosphorization and decarburization endpoints to cut over-blow and re-blow rework.
Off-gas analysis and sub-lance data feed a continuously updated bath model instead of a static static-lookup end-point chart.
| Endpoint C accuracy | ±0.02% |
| Re-blow rate reduction | 30–40% |
| Oxygen consumption saving | 3–5 Nm³/t |
| Yield improvement | +0.3–0.6% |
DRI, MIDREX & COREX
Rotary kiln and shaft-furnace DRI models track metallization and gangue chemistry, extending our ironmaking modeling work into gas-based and coal-based reduction routes.
Same thermochemical core used in our ferroalloy smelting and blast furnace CFD work, adapted for reduction kinetics.
| Metallization prediction | ±1.5% |
| Reductant consumption saving | 5–8% |
| Kiln residence optimization | Yes |
| Applicable routes | MIDREX, COREX, Rotary Kiln |
Induction Furnace Steelmaking
SmartMelt extends its thermochemical core to coreless induction furnace (IF) melting, widely used across India's secondary steel sector for scrap and sponge-iron based production.
Models track bath temperature rise, power-to-melt efficiency, and residual/trace element buildup from mixed scrap-DRI charges — critical for IF shops feeding downstream billet casters without a converter or refining step.
| Bath temperature accuracy | ±10°C |
| Power-to-melt efficiency gain | 4–6% |
| Residual element tracking | Cu, Cr, Ni, Sn |
| Charge mix optimization | Scrap + DRI/sponge iron blends |
From plant data to shopfloor decision in four steps
Data ingestion
Sensor, off-gas, and lab data streamed from DCS/SCADA into SmartMelt's process layer.
Thermochemical modeling
Mass & energy balance engine computes bath state from first principles.
ML prediction
Trained classifiers refine composition & temperature forecasts against plant history.
Operator action
Alloy, lance, and power recommendations pushed straight to the pulpit display.
Bring SmartMelt to your shopfloor
ExtractMet works directly with plant process teams to pilot SmartMelt on live EAF, BOF, or Induction Furnace Steelmaking operations — starting with a data audit and a model built on your own heat history.