Multi-objective gradient-based intelligent optimization of ultra-high-strength galvanized TRIP steels

• Published in: International journal of advanced manufacturing technology 2023-09, Vol.128 (3-4), p.1749-1762
• Main Authors: Flor-Sánchez, Carlos O., Reséndiz-Flores, Edgar O., Altamirano-Guerrero, Gerardo, Salinas-Rodríguez, Armando
• Format: Article
• Language: English
• Subjects: Algorithms; Bainitic transformations; CAE) and Design; Computer-Aided Engineering (CAD; Cooling rate; Engineering; Heat treating; Heat treatment; Hot dip galvanizing; Industrial and Production Engineering; Martensitic stainless steels; Mechanical Engineering; Mechanical properties; Media Management; Multiple objective analysis; Optimization; Original Article; Pareto optimization; Process parameters; Regression models; Room temperature; Support vector machines; TRIP steels
• ISSN: 0268-3768; 1433-3015
• DOI: 10.1007/s00170-023-11953-6

In this paper, a novel gradient-based algorithm named Kernel-based hybrid multi-objective optimization (KHMO) is implemented and coupled with a support vector regression (SVR) model to efficiently optimize the production of a cold rolled hot-dip galvanized TRIP steel. For this purpose, several heat treatments using an isothermal bainitic transformation (IBT) temperature compatible with continuous hot-dip galvanizing were performed. The most significant processing parameters (cooling rate after intercritical austenitizing ( C R 1 ), isothermal holding time at the galvanizing temperature in the bainitic region t 2 , and last cooling rate to room temperature ( C R 2 )) were thus optimized to achieve the required mechanical properties values. In general, SVR model fits in a satisfactory manner the highly non-linear relationship between experimental parameters and resulting mechanical properties; hence, it is used as objective function. Besides, KHMO algorithm reveals an outstanding performance since it found a dense and spread Pareto front. Moreover, the processing window to manufacture TRIP-aided martensitic steels is suggested in a range of 57–63 ∘ C/s, 33–37 s, and 1–2 ∘ C/s for C R 1 , t 2 , and C R 2 , respectively. The developed computational methodology for modeling and optimization of operating parameters is successfully applied for the first time in the experimental processing of advanced TRIP steels.