Production of Zero Defect (ZD) Slabs in Steel Continuous Casting

PROJECT LEAD: ArcelorMittal

PARTNERS: Purdue University, Missouri University of Science & Technology, Rensselaer Polytechnic Institute

PROBLEM STATEMENT: Absence of predictive maintenance and real-time quality prediction tools increases the overall energy intensity of the steelmaking process via increased unplanned turnarounds (UPTA) and product defects.

PROJECT GOAL: Improve steel slab quality and continuous caster productivity using Smart Manufacturing (SM) methodologies and technologies to address the top two KPIs of the continuous casting process –yield (minimize defects/rejects) and uptime (reduce UPTA).

TECHNICAL APPROACH: 

• Employ the CESMII-based SM/big-data platform to capture process and quality data
• Extend and scale-up the in-house caster condition monitoring application “Caster Health Monitor” (CHM) into this new SM platform for predictive tools
• Develop real-time hybrid predictive models for slab defects and quality using the SM platform
• Build an interactive, SM-platform-tethered, VR-based interface of a digital twin prototype of the continuous casting process for shop-floor deployment by integrating all the above developments

KEY TASKS AND MILESTONES:

• Define CESMII-directed platform specification 
• Configure CESMII-based SM Platform for project use 
• Implement and prove effectiveness of new advanced sensors in the pilot caster 
• Extend, scale-up and demonstrate a first running version of CHM application on the new SM Platform 
• Complete, deliver and validate an enhanced version of the CHM application as a true predictive maintenance tool 
• Deliver a hybrid (data-driven/machine learning based + physics-based such as CFD) slab quality prediction model 
• Deliver and deploy a first version of a digital twin of the pilot caster on the shop floor 

POTENTIAL IMPACT: 

• A 0.2% savings in yield (from reduction of defects) is equivalent to an annual savings of $90M for the whole US steel industry (plus 2.68 PJ of energy savings per year equivalent to about 22 million gallons of gas savings, enough to power ~ 70,000 typical American homes for a year).
• Predictive maintenance tools alone could save at least $2M per caster strand per year (there are hundreds of strands in the US)