Improvement in Coating Weight Control Model Functionality Using Advanced Machine Learning Technique

Coating weight control models (CWCM) regulate either knife position or pressure, but not both, simultaneously. Consequently, operators must rely solely on their experience to manually choose one of the parameters, potentially causing coating defects. To enhance model performance, the existing CWCM was integrated with machine learning-based look-up tables to align knife positions with predicted pressures. Data from thousands of high-quality coils — covering pressure, position, knife height, effective aluminum and actual coating weight — was gathered. The XG-Boost algorithm with process know-how (hybrid modeling) was used to forecast knife positions, which were then validated during live production.

Authors:

Mr. Alex Solovyev | Stelco Inc., A Cleveland-Cliffs Company
 

Mr. Vipin Chandra Muthirikkaparambil | McMaster University
 

Mr. Alex Chen | Stelco Inc., A Cleveland-Cliffs Company
 

Mr. Jackson Moran | Stelco Inc., A Cleveland-Cliffs Company
 

Dr. Amiy Srivastava | Stelco Inc., A Cleveland-Cliffs Company
 

Mr. Alex Suciu | Stelco Inc., A Cleveland-Cliffs Company
 

Mr. John D’Alessio | Stelco Inc., A Cleveland-Cliffs Company
 

Mr. Keith Dennis | Stelco Inc., A Cleveland-Cliffs Company
 

Session Chairs:

Dan Drane | Steel Dynamics - The Techs Div.
 

Luis Garza Martinez | Cleveland-Cliffs Research & Innovation Center