Facilitating recycling of 6xxx series aluminum alloys by machine-learning based optimization

In our research, we focus on reducing the number of grades in the 6xxx series of aluminum alloys. A higher number of alloy grades poses challenges in recycling and increases the complexity of the solution space. To address this, we designed a machine learning framework aimed at optimizing the number of grades within the 6xxx series.

We first collected data on the chemical composition, mechanical properties, service characteristics, and technological properties of various 6xxx series aluminum alloys under different tempering conditions (T5, T6, and T7). We then applied Principal Component Analysis (PCA) and k-means clustering to reduce the dataset into a few clusters and sub-clusters, each containing alloys with similar property ranges.

Next, we employed an optimization algorithm to select the optimal alloys within each sub-cluster that effectively represent the range of properties found in that sub-cluster. Additionally, we analyzed the metallurgical reasoning behind the clustering to ensure practical feasibility.

This approach not only demonstrates the potential of machine learning in alloy selection but also provides a strategic direction for improving aluminum recycling practices.