Automated, Scalable AI for Real-Time Monitoring of Steel Continuous Casting System

Conventional machine learning methods used to monitor the caster operations are unable to seamlessly keep up with the physical changes in the system, causing model accuracy issues over time. This paper presents a novel approach that leverages supervised and unsupervised learning to automatically discover patterns that correlate to inefficiencies in the casting process. By correlating the model assessment with the live signals, a high-accuracy early warning system was achieved for identifying impending failure modes. This paper will showcase results from the real-world deployment of this approach and its efficacy in optimizing the caster performance.

Mr. Joe Porter | Falkonry Inc.
 

Keval Bhanushali | Falkonry Inc.
 

Shreebhooshan B. | Falkonry Inc.
 

Suhas Mehta | Falkonry Inc.
 

Nikunj Mehta | Falkonry Inc.