Inclusion monitoring for continuous casting
Real-time inclusion detection
With Event-based Metavision® systems, detect inclusion spikes and process instability during casting as they occur, enabling earlier intervention and reducing downstream losses.
THE GAP BETWEEN CASTING AND QUALITY AWARENESS
In continuous casting, quality issues are typically detected through offline sampling, creating a delay of 30 to 90 minutes between a process event and awareness of it.
During this time:
- Defective steel may already be cast
- Slabs may be stored or processed further
- Root causes become harder to trace
There is no continuous signal linking what happens inside the mould to the final quality outcome. As a result, most quality issues are identified after value has already been lost.
THE COST OF
OPERATING BLIND
€7K – €33K
Rejected slab (incl. downstream costs)
€20K – €100K
Customer quality escape
€100K – €300K
Breakout cascade event
€0.8M – €4M
Annual impact per caster (~200 kt)
FROM OFFLINE SAMPLING TO CONTINUOUS MONITORING
Detect inclusion spikes during casting
instead of discovering them 30–90 minutes later
See particle load, flux behavior, and flow in real time
not only through delayed lab analysis
Act before value is lost downstream
adjust casting speed, flux, or divert slabs early
Connect process events to quality outcomes
with a continuous signal across the full cast
WHY EVENT-BASED VISION
Monitoring molten steel requires capturing fast, subtle changes in a dynamic and extreme environment.
Event-based vision is inherently suited to this:
Continuous motion sensing
no gaps in particle flow detection
Sensitivity to subtle particle variations
detect inclusions and flow instability early
Real-time detection during casting
microsecond-level response
Reliable in extreme light and contrast
enabled by native high dynamic range
REVEALING THE SIGNAL
Prophesee applies event-based monitoring inline to provide continuous visibility into the casting process:
- Optical measurement of a representative sampled stream
- Real-time detection of inclusions, particles, and clusters
- Continuous comparison against a process-specific baseline
- Identification of multiple failure modes through one monitoring path:
- inclusion spikes
- flux anomalies
- clogging progression
This provides a continuous signal across the full cast, linking process behavior to quality outcomes in real time.
VALIDATION IN REAL OPERATING CONDITIONS
Validation is designed to assess detection capability and reliability under live steelmaking conditions. Prophesee’s validation protocol includes:
- Continuous monitoring across full casting sequences (4–12 hours)
- Operation at ~1,550°C and casting speeds of 0.8–2.5 m/min
- Detection of mixed particle populations (flux, inclusions, clusters)
- Signal quality monitoring during unattended operation
- Comparison with offline sampling and process context, including speed changes, grade transitions, and SEN events.
Result: An evidence-based assessment of signal quality, detection timing, and operational relevance before broader integration.
EXPLORE A PROOF OF CONCEPT
Evaluate real-time inclusion monitoring on your casting process and assess detection performance under real operating conditions.
Tell us about your project and we’ll get back to you shortly.
Explore other applications of event-based vision
RECOGNITION
