Multi-Objective Optimization for Mold Flux Design via Machine Learning

Achieving high performance of mold flux across casting conditions and steel grades involves optimizing multiple objectives (e.g., viscosity, crystallization, thermal conductivity) with many variables (e.g., flux composition, temperature) simultaneously. This study proposes a multi-objective optimization (MOO) method based on machine learning (ML) and metaheuristic algorithms to efficiently optimize thermophysical properties for mold flux design. ML models for predicting thermophysical properties were developed using Artificial Neural Network, Support Vector Machine, and Random Forest algorithms. The selected ML models were utilized as the objective function. The MOO method was employed to obtain an optimal thermophysical properties from the Pareto front.

Authors:

Gibeom Kim | POSTECH
 

Sung Suk Jung | POSCO
 

Dae-Geun Hong | POSTECH
 

Session Chairs:

Andrea Aller | Nucor Steel Gallatin
 

Chase Ault | Steel Dynamics Inc. - Flat Roll Group - Butler