AI turns scrap impurities into useful ingredients for stronger recycled iron

Global metal production stands at 2 billion metric tons annually, with industry, infrastructure, and energy sectors driving demand. However, reliance on raw materials contributes to 40% of industrial greenhouse gas emissions and 10% of global energy consumption. Recycling metals presents a solution to mitigate these environmental impacts while reducing dependence on scarce raw materials.

Yet, metal scrap contains impurities and varying compositions, complicating the production of high-quality recycled alloys. An international team of scientists leveraged machine learning to transform these impurities into advantageous alloying elements, creating "dirty alloys" with enhanced mechanical properties and reduced production costs.

Traditional methods add virgin material to dilute impurities, undermining recycling's environmental benefits. The novel machine-learning approach linked alloy composition and processing conditions to desired performance without focusing on microstructural features. Using ductile iron as a model, the researchers tested 17 designs, with 14 exceeding performance benchmarks of existing recycled-iron materials.

Tensile strength increased by 15.2% at similar elongation, and elongation rose by up to 73.2% at comparable strength. Manufacturing costs were also 8.1% lower than the conventional blend of scrap and virgin iron. The framework's success extends the potential for recycled metals in transport, energy, and industrial manufacturing, with future developments aiming to integrate multi-agent AI systems for autonomous alloy development cycles.