Glass-forming ability (GFA) remains a little-understood property. Experimental work on bulk metallic glasses (BMGs) is guided by many empirical criteria, often not of significant predictive value. This work aims to utilise machine-learning techniques both to produce predictive models for the GFA of alloy compositions, and to reveal insights useful for furthering our theoretical understanding of GFA. The produced machine-learning models simultaneously predict the liquidus temperature, glass transition temperature, crystallisation temperature, critical casting diameter, and probability of forming a BMG, for any given alloy.
The incomprehensible size of composition space means even coarse grid-based searches for interesting alloys are infeasible unless constrained, requiring prior knowledge. Genetic algorithms provide a practical alternative, by rapidly homing in on fruitful regions and discarding others. Competition, recombination, and mutation are applied to a gene pool of alloy compositions, with the goal of evolving towards excellent BMG candidates as predicted by the machine-learning models.