A Machine Learning Approach for Modelling Cold-Rolling Curves for Various Stainless Steels.

Area of Science:

• Materials Science
• Metallurgy
• Computational Materials Science

Background:

• Stainless steel exhibits cold-work hardening, influencing its mechanical behavior during forming.
• Predicting strain-hardening properties is crucial for manufacturing flat stainless steel products via cold rolling.

Purpose of the Study:

• To develop predictive models for stainless steel mechanical properties.
• To forecast tensile strength, yield strength, hardness, and elongation based on composition and cold reduction.

Main Methods:

• Utilized artificial neural networks (ANNs) and multiple linear regression (MLR).
• Cold-rolled various stainless steel grades (austenitic, ferritic, duplex) at laboratory scale.
• Determined mechanical properties via tensile testing to create a predictive database.

Main Results:

• ANN models demonstrated high accuracy in predicting mechanical properties.
• Models successfully correlated chemical composition and cold thickness reduction with mechanical outcomes.
• Established a robust database for machine learning application.

Conclusions:

• Machine learning models offer valuable tools for designing new stainless steel grades.
• These models can aid in optimizing cold-forming processes for stainless steel.
• Predictive capabilities enhance material engineering and product development.