Max Planck Institute for Sustainable Materials Develops AI Framework to Turn Scrap Metal Into Stronger Iron Alloys
By Ian Williams, manager
updated:
Recycling metal comes with a major catch. Unlike freshly mined ore, scrap arrives with unpredictable chemical mixes and stray elements that can weaken the final alloy, making it tough to produce high-grade materials that hold up under demanding industrial requirements.
Every year, global industry, construction, and energy networks consume roughly two billion tonnes of metal. Because manufacturers still rely so much on freshly mined raw materials, metal production drives about 40 percent of industrial greenhouse gases, 8 percent of total global emissions, and 10 percent of the world's energy use. Scaling up smarter recycling wouldn't just cut those emissions; it would also ease dependence on imported, hard-to-source raw materials.
Turning Scrap Impurities Into Strength
Metal recycling poses significant challenges because scrap, unlike primary raw materials, can have highly variable chemical compositions and contain unwanted elements. These inconsistencies may weaken the mechanical properties of recycled alloys and hinder the production of high-performance materials that meet strict industrial standards.
An international team of scientists has developed a machine-learning approach that makes use of this variability. Rather than removing impurities as contaminants, the team deliberately incorporates them as alloying elements. These “dirty alloys” are more tolerant of variations in scrap composition while offering improved mechanical properties and lower production costs. The findings were published in Science Advances.
Dirty alloy design workflow: The circular workflow begins with a database containing scrap composition, processing parameters, and microstructural data for known alloys (CPM-P). Using these data, a deep learning model generates optimized alloy designs. The proposed compositions are then screened to identify promising candidates. These designs are validated through compositional and microstructural analyses, as well as experimental melting, casting, and nodularization.
The resulting data are uploaded to the database, closing the design loop. The conventional approach to managing impurities in metal scrap is to dilute them by adding virgin material. However, this undermines the environmental benefits of recycling, as producing virgin metals requires substantial energy and generates significant CO₂ emissions. We therefore developed a new machine-learning strategy that turns these impurities from a problem into an advantage, enabling us to design alloys that not only tolerate contaminants but can even benefit from them, explains Liuliu Han, partner group leader at the Max Planck Institute for Sustainable Materials.
Instead of relying on slow microstructural checks, the system built with the Max Planck Institute for Sustainable Materials maps chemical compositions and factory settings straight to final metal strength. Using existing formulas as a starting point, a tree search algorithm shifts element ratios, temperatures, and curing times inside workable ranges to come up with new designs. Neural networks then predict how strong and ductile each candidate will be, weighing steady tweaks against riskier, unexplored mixes.
Long before melting any scrap, a secondary filter backed by a wide library of alloy data narrows the pool down to the strongest finalists. After lab crews cast those metals and measure how they hold up under strain, the test numbers feed back into the training set and sharpen the next batch of ideas. Ductile iron made an obvious test case because it handles stray elements with ease and accounts for 120 million tons of annual global output. Out of 17 AI generated recipes put through physical trials, 14 surpassed the best strength and stretch combinations ever recorded for recycled iron.
- Against that reference, tensile strength increased by up to 15.2 percent at comparable elongation.
- Elongation at fracture increased by up to 73.2 percent at comparable tensile strength.
- Manufacturing costs were 8.1 percent lower than those of the conventional route, which blends scrap with virgin iron.
- The team demonstrated that alloys made entirely from scrap can outperform materials produced through primary synthesis while reducing their environmental impact.
- The research team came from Shanghai Jiao Tong University, City University of Hong Kong, East China University of Science and Technology, Central South University and Germany’s Max Planck Institute for Sustainable Materials.
- The framework could be extended to recycled steels, aluminium and copper alloys for applications in transport, energy and industrial manufacturing.
- A proposed next step is to integrate multi-agent AI systems with automated experimental platforms.
- Specialized agents could coordinate alloy design, experimental planning, process control and interpretation of results.
- An autonomous development cycle could use experimental findings to guide subsequent design and manufacturing decisions while adapting to variations in scrap composition.
- Future advances could help identify ways to use impurities as beneficial alloying elements.