Machine learning driven design of high-performance lightweight Aluminium alloys

Guest Speaker：Assoc.Prof. Ishwar Kapoor, School of Engineering, University of Warwick, UK

Inviter: Professor Leyun Wang

Date & Time: Thursday, 13th March, 14：00-15：30

Venue: Yiucheng Lecture Hall（500）, Xu Zuyao Buildingi

Biography:

Ishwar is the Associate Professor, Director of Student Projects at the University of Warwick, UK. He leads and manage a portfolio of project modules across Engineering and Warwick Manufacturing Group (WMG) undergraduate and postgraduate programmes. Ishwar graduated with PhD in Engineering, WMG, University of Warwick, UK in 2020 and Bachelor of Technology in Mechanical Engineering with minor in Electronics and Communication Engineering, Indian Institute of Technology (IIT) Guwahati, India in 2016. During his PhD, he worked with Tata Steel Netherlands and the UK to develop new Advanced High Strength Steel (AHSS) grades for European and International Automotive markets. Currently, his research interests cover Materials Informatics Lightweight Alloys and Steel Product Development from scrap metals. He has been awarded one of the most outstanding articles published in steel research international journal (2024) and Institute of Materials, Minerals & Mining Adrian Normanton Medal for the best paper on the topic of steelmaking or casting (2022). Ishwar is the Senior Fellowship of the Higher Education Academy (SFHEA) and leads and manage a portfolio of project modules across Engineering and Warwick Manufacturing Group (WMG) undergraduate and postgraduate programmes. He leads and oversees initiatives to connect students with innovative projects, fostering collaboration with external stakeholders, including industry partners.

Abstract:

Aluminium (Al) alloys are widely used in the automotive, aerospace, and marine industries due to their high specific strength, thermal conductivity (TC), and recyclability. However, improvements in strength often come at the cost of reduced TC. Machine learning (ML) is increasingly applied in materials science, particularly for property prediction and alloy design. This work presents a joint project between SJTU and Warwick using ML techniques to develop predictive models for TC and ultimate tensile strength (UTS) of Al alloys, employing eXtreme Gradient Boosting (XGBoost) and Support Vector Machine (SVM) algorithms. Physical descriptors derived from alloy composition were used, with feature engineering performed using Lasso and Gini Impurity methods. Guided by these models, an Al-2.64Si-0.43Mg-0.10Zn-0.03Cu alloy was designed, achieving a TC above 190 W·m⁻¹·K⁻¹ and a UTS exceeding 220 MPa. Experimental results closely matched predictions, and microstructure analysis indicated that the fragmented and spherical Si phases, with minimal non-spherical Si, were critical to the improved properties. This study highlights the potential of ML in accelerating alloy development and bridging computational predictions with experimental validation.