Data-Driven Design of Recycling-Friendly Aluminium Alloys

Abstract

Improving material efficiency through the increased use of secondary material is an unambiguous goal for the aluminium industry. The main barrier in achieving this is the compositional mismatch between scrap and alloy. Among the many approaches to address this, one promising direction is to design alloys that compositionally match scrap streams, a practice called recycling-friendly alloy design (RFAD). However, previous efforts have largely focused on creating good scrap sinks without considering the system-wide nature of scrap consumption. In a supply chain governed by interdependent material flows, alloys must function not only as sinks but also as future sources of scrap.

In order to increase overall scrap consumption, we propose an RFAD framework combining dynamic material flow analysis, economic feedback, and plant level batch-design to include a systemic assessment of end-of-life material generation and consumption. Each of these components is integrated in a Bayesian optimisation framework with comprehensive exploration of the alloy design space to balance the pervasive source-sink tension of elemental limits. Finally, we include a natural language processing module to assist in identifying thermodynamic and property constraints used to filter the design space.

Using our framework, we find that the optimal alloys for the North American transport sector achieve a 7% increase in scrap usage, corresponding to 200-400 kt/year of additional scrap consumption. We also demonstrate that alloys designed for short lifetime sectors (e.g., transport) tend to be good sources whereas those designed for longer lifetime sectors (e.g., building & construction) are good sinks due to their limited capacity to influence the composition and therefore the consumption of the post-consumer scrap stream. The capabilities of our model also allow us to simulate RFAD under different future scenarios. In the automotive sector, adoption of electric vehicles is driving a decrease in cast alloy production, a traditional sink. Thankfully, we show that the lower the production of cast alloys by 2040, the higher the scrap usage attainable with RFAD along with better properties. In the can sector, unialloys have been theorised to increase recycling by lowering compositional variability. Our simulations show that RFAD for a can unialloy leads to an additional 3.2% scrap usage increase versus RFAD for a non unialloy. Finally, exploring the Pareto frontier between yield strength and scrap usage, we find that 2xxx alloys offer the best trade-off, suggesting the potential for high-strength alloys sourced from automotive scrap.