Temperature measurement in an electric arc furnace is one of the most dangerous routine tasks in steelmaking. An operator approaches the furnace with a thermocouple lance, inserts it through a slag door or dedicated port, waits for a reading, and retreats — all within a few meters of molten steel at 1,600°C or higher.
The task takes under two minutes. But it happens 8–12 times per heat, across three heats per shift, every shift of the year. The cumulative exposure is significant, and the consequences of an incident — a slag eruption, a lance failure, a misstep on a wet floor — can be catastrophic.
The Case for Automation
Automated temperature measurement isn't new. Continuous optical pyrometry, disposable sub-lance systems, and robotic sampling arms have all been deployed at various plants worldwide. But adoption remains low, particularly at smaller EAF operations, for three reasons:
- Cost. Robotic systems designed for BOF (basic oxygen furnace) environments cost EUR 500K–1.5M installed and require specialized maintenance.
- Reliability. Optical systems struggle with dust, fume, and slag interference. Accuracy degrades over a campaign, and operators lose trust.
- Integration. Standalone measurement systems that don't feed directly into Level 2 automation create extra manual steps, reducing the net time savings.
PRAX's Approach
We've been working with two partner plants to deploy a measurement system that addresses all three barriers. The core idea is simple: a ruggedized robotic arm, mounted outside the direct splash zone, that handles the same disposable thermocouple cartridges operators use today.
By keeping the consumable the same, we eliminate the accuracy debate — the measurement is identical to the manual method. The robot just removes the person from the hot zone.
The arm is integrated directly into the Level 2 system. When the automation calls for a temperature check, the robot executes the measurement and writes the result to the same data historian the operator would have used. No new screens, no new workflows.
Early Results
At the first pilot site, the system has completed over 3,000 measurement cycles with 97.2% first-attempt success rate. The 2.8% failures are almost entirely cartridge-related (bent thermocouples, blocked ports) — the same failure modes operators experience.
Operator feedback has been overwhelmingly positive, not because the measurement is faster (it isn't — the robot is slightly slower than a skilled operator), but because it eliminates the most physically demanding and psychologically stressful task in their shift.
The safety case speaks for itself: zero minutes of human exposure in the hot zone per heat, with no loss of measurement quality.