ArcStream: AI-Powered Electric Arc Furnace Optimization
Real-time tracking, prediction, and optimization of steel chemistry, temperature, and tap-to-tap time — deployed at a major European steel producer.
The Challenge
The Complexity of Modern EAF Operation
Arc furnaces face increasing pressure to reduce energy consumption, electrode consumption and tap-to-tap time while producing higher-quality steel.
Unstable operation leading to extended tap-to-tap times
Unplanned downtimes from process instability
Manual decisions based on operator experience alone
Hard to predict final steel chemistry and temperature before the tap
Energy waste from non-optimal operating profiles
No standardized best practice across shifts and steel grades
The Solution
ArcStream — L2 Process Package for EAF
ArcStream provides real-time mass and energy balance, a dynamic EAF model forecasting steel temperature and chemistry, endpoint control targeting tap conditions, and a pre-encoded causal graph for what-if simulation.
Dashboard Showcase
Find Your Golden Heats
Plot every heat on throughput (TPH) vs energy efficiency (kWh/t). The model automatically identifies the "Golden Heats" — the top 12% performing heats under similar conditions that operators should replicate. Click any point to inspect its operating parameters.
Real-Time Steel Chemistry and Temperature Prediction
The Dynamic EAF Model forecasts steel temperature, weight, chemical composition, meltdown degree, and slag/steel weight continuously — closing the loop with endpoint targets before the tap.
From Heat Timeline to Shift Performance
Every heat is tracked as a Gantt — melting, refining, power-on/off, charging, slagging, tapping. Aggregated across shifts and steel grades, you spot systematic bottlenecks and reduce tap-to-tap time by up to 20%.
Safe & Efficient Meltdown
Live melt-down degree with Normal / Warning / Critical zones. Operators see charging, melting, and refining phases in real time — reducing energy waste and preventing dangerous operating conditions.
Process Models
ArcStream L2 Process Models
Four interconnected models covering the full heat cycle — from charge-in to tap.
Online Mass & Energy Balance
Real-time mass and energy balance across the full EAF heat cycle — continuously updated from sensor data and charge weights.
Dynamic EAF Model
Forecasting steel temperature, weight, composition, meltdown degree, and slag & steel weight from charge-in to tap.
Endpoint Control Model
Closed-loop control targeting tap temperature, chemistry and weight — reducing chemistry corrections and reblows.
EAF Causal Graph
Pre-encoded DAG of process variables enabling counterfactual analysis and what-if simulation for process engineers.
Results
Measurable Impact Across Every Heat
Deployed at a major European steel producer, ArcStream delivered quantifiable improvements across energy, electrodes, and throughput within the first months of operation.
-20%
Tap-to-tap time
-8%
Energy consumption (kWh/t)
+12%
Electrode life
100%
Heats traceable & analyzable
" ArcStream gave our operators the feedback loop they never had — we now optimize in real time instead of reviewing batches after the fact. "
Process Optimization Lead
European Steel Producer
Ready to Optimize Your EAF Operations?
Talk to our steel industry experts and see how ArcStream can be deployed at your site.