Experimental study and machine learning model to predict formability of magnesium alloy sheet [version 1; peer review: 2 approved, 1 approved with reservations]

• Published in: F1000 research Vol. 11; p. 1118
• Main Authors: Viswanadhapalli, Balaji, V.K, Bupesh Raja, Nagaraju, Krishna Chythanya
• Format: Journal Article
• Language: English
• Published: London Faculty of 1000 Ltd 2022; F1000 Research Limited; F1000 Research Ltd
• Subjects: Aerospace and automotive applications; Cold; Crystals; Deformation; Ductility; Energy efficiency; eng; Etching; Formability; Forming Limit Diagram; Grain boundaries; Learning algorithms; Machine learning; Machine learning model; Magnesium; Magnesium alloy sheet; Magnesium alloys; Mechanical properties; Metal forming; Steel alloys; Stretch forming; Temperature; Tensile strength; Yield stress; Forming Limit Diagram; Stretch forming; Formability; Machine learning model; Aerospace and automotive applications; Magnesium alloy sheet
• ISSN: 2046-1402; 2046-1402
• DOI: 10.12688/f1000research.124085.1

Background: Magnesium alloy is not only light in weight but also possesses moderate strength. Magnesium AZ31-H24 alloy sheet has many applications in the automotive and aerospace industries. Experimental stretch forming tests are performed on this sheet to measure the material's formability by constructing forming limit diagrams. Methods: Several tests of Nakazima were carried out on rectangular samples at 24, 250, 350°C and 0.01, 0.001 mm/s using a hemispherical punch. The work done to predict the formability of magnesium alloys has not been recorded in recent literature on machine learning models. Hence, the researchers of this article choose to explore the same and build three models to predict the formability of magnesium alloy through Random Forest algorithm, Extreme Gradient Boosting, and Multiple linear Regression. Results: The Random Forest showed high accuracy of 96% in prediction. Conclusions: It is concluded that the need for physical experiments can be greatly minimized in formability studies by using machine learning concepts.