EXTENDED ABSTRACT: This presentation provides an overview of some of the work carried out by the speaker and his collaborators. From the groundbreaking work with Prof Wei Xu [1] in integrating computational thermodynamics and genetic algorithms (GAs) to maximise the strength of maraeging steels, to the integration of processing printing parameters to enhance marageing steels and aluminium alloys printability [2]; there is a proven ability for GAs to discover advanced alloys of improved properties. This has further been demonstrated to tailor the mechanical response of Ti-based high entropy alloys, where transformation induced plasticity (TRIP) and twinning induced plasticity (TWIP) have been designed for and attained with surprising temperature-dependent properties. The key to GA design is a judicious choice of physical metallurgical (PM) parameters to describe the relevant phenomena, e.g. precipitate nucleation and growth, TRIP, and TWIP. With the emergence of physical metallurgical knowledge graphs (PMKG), it has become possible to increase even further predictability [3], demonstrating that deep learning protocols for alloy design can be further enhanced [4], increasing significantly predictability. PMKG use has proven to be successful in the design of quench and partitioning steels [3]; its application to the design of high-speed rail alloys of improved wear properties is outlined here.

Keywords: alloy design; genetic algorithms; knowledge graphs; physical models; mechanical properties.
REFERENCES:
[1] W. Xu et al., Acta Materialia, 58, (2010) 3582-3593; [2] H. Eskandari Sabzi et al., Acta Materialia, 274, (2024) 120018;
[3] Y. Li et al., Materials Genome Engineering Advances, under review. [4] C. Shen et al., Journal of Materials Science and
Technology, 87, (2021) 258-268.