Machine Learning-Assisted Design of Recycled Aluminum Alloys for Lightweight Automotive Structures: A Comparative Analysis of Conventional Trial-and-Error Metallurgy and Data-Driven Alloy Optimization

Abstract

Recycled aluminum alloys are increasingly important for lightweight automotive structures because they reduce embodied energy, lower carbon emissions, and support circular manufacturing. However, secondary aluminum streams often contain variable impurity concentrations, especially iron, silicon, copper, and magnesium, which complicate mechanical-property control and structural reliability. This article comparatively evaluates conventional trial-and-error alloy development and machine learning-assisted recycled aluminum alloy optimization. Using a computational materials engineering framework grounded in alloy databases, thermodynamic reasoning, microstructure–property relationships, and supervised learning models, the study analyzes how data-driven alloy design affects tensile strength, ductility, impurity tolerance, processing stability, and sustainability outcomes. The comparative analysis demonstrates that conventional metallurgy remains valuable for mechanistic interpretation and process validation, but machine learning improves design-space exploration, composition screening, and prediction of property trade-offs under variable recycling conditions. The findings suggest that sustainable alloy engineering requires integration of metallurgical theory, materials informatics, casting-process control, and lifecycle-oriented manufacturing strategy. This article contributes to engineering scholarship by connecting circular economy objectives with computational materials design and lightweight structural engineering.