Title

OPTIMIZATION OF THE MECHANICAL PROPERTIES OF DP780 GI STEEL USING GENETIC ALGORITHMS AND MACHINE LEARNING: AN INDUSTRIAL APPLICATION

OPTIMIZATION OF THE MECHANICAL PROPERTIES OF DP780 GI STEEL USING GENETIC ALGORITHMS AND MACHINE LEARNING: AN INDUSTRIAL APPLICATION

Authorship

FINAMOR, FELIPE PEREIRA; ELIAS, EDUARDO POSSA; ALVES, GABRIEL GODINHO; CORRÊA, SILVIO; DRUMOND, JULIOVANY; FERREIRA, JETSON LEMOS; MEI, PAULO ROBERTO

FINAMOR, FELIPE PEREIRA

I am this author

ELIAS, EDUARDO POSSA

I am this author

ALVES, GABRIEL GODINHO

I am this author

CORRÊA, SILVIO

I am this author

DRUMOND, JULIOVANY

I am this author

FERREIRA, JETSON LEMOS

I am this author

MEI, PAULO ROBERTO

I am this author

DOI

10.5151/2594-5297-42302

Downloads

Abstract

In response to the growing demand for high-performance materials in the automotive industry, this study presents a hybrid methodology that integrates machine learning and multi-objective genetic algorithms to optimize the mechanical properties of galvanized DP780 steel. Using a dataset of approximately 14,000 industrial coils, predictive models were developed to estimate Yield Strength (YS), Ultimate Tensile Strength (UTS), and Total Elongation (EL) based on chemical composition and process parameters. Among the evaluated models, the LightGBM demonstrated superior predictive accuracy and was further analyzed using SHAP values to identify key influencing variables. Subsequently, the NSGA-II algorithm was employed to optimize processing conditions and alloying elements, aiming to enhance mechanical performance while increasing production speed. The optimized parameters were implemented on an industrial galvanizing line, resulting in a 12.5% increase in line speed and improved material properties. These results validate the effectiveness of AI-driven optimization strategies in industrial steel manufacturing, offering a robust framework for enhancing both product quality and operational efficiency

IN RESPONSE TO THE GROWING DEMAND FOR HIGH-PERFORMANCE MATERIALS IN THE AUTOMOTIVE INDUSTRY, THIS STUDY PRESENTS A HYBRID METHODOLOGY THAT INTEGRATES MACHINE LEARNING AND MULTI-OBJECTIVE GENETIC ALGORITHMS TO OPTIMIZE THE MECHANICAL PROPERTIES OF GALVANIZED DP780 STEEL. USING A DATASET OF APPROXIMATELY 14,000 INDUSTRIAL COILS, PREDICTIVE MODELS WERE DEVELOPED TO ESTIMATE YIELD STRENGTH (YS), ULTIMATE TENSILE STRENGTH (UTS), AND TOTAL ELONGATION (EL) BASED ON CHEMICAL COMPOSITION AND PROCESS PARAMETERS. AMONG THE EVALUATED MODELS, THE LIGHTGBM DEMONSTRATED SUPERIOR PREDICTIVE ACCURACY AND WAS FURTHER ANALYZED USING SHAP VALUES TO IDENTIFY KEY INFLUENCING VARIABLES. SUBSEQUENTLY, THE NSGA-II ALGORITHM WAS EMPLOYED TO OPTIMIZE PROCESSING CONDITIONS AND ALLOYING ELEMENTS, AIMING TO ENHANCE MECHANICAL PERFORMANCE WHILE INCREASING PRODUCTION SPEED. THE OPTIMIZED PARAMETERS WERE IMPLEMENTED ON AN INDUSTRIAL GALVANIZING LINE, RESULTING IN A 12.5% INCREASE IN LINE SPEED AND IMPROVED MATERIAL PROPERTIES. THESE RESULTS VALIDATE THE EFFECTIVENESS OF AI-DRIVEN OPTIMIZATION STRATEGIES IN INDUSTRIAL STEEL MANUFACTURING, OFFERING A ROBUST FRAMEWORK FOR ENHANCING BOTH PRODUCT QUALITY AND OPERATIONAL EFFICIENCY

Keywords

Optimization, Machine Learning, AHSS, Galvanizing

Optimization, Machine Learning, AHSS, Galvanizing