Optimization trajectories for alloy recycling using factorization machines and annealing

This dataset accompanies the paper Data-driven multi-objective optimization for alloy recycling using factorization machines and quantum annealing by Thomas Plehn, Katrin Bugelnig, Silvana Tumminello, Daniel Barragan-Yani, and David Melching (2026), arXiv:2607.03208. The study investigates factorization-machine-based active learning for aluminium alloy recycling, comparing quantum annealing (QA) with simulated annealing (SA). Single-objective optimization targets yield strength; multi-objective optimization considers yield strength and thermal conductivity using data-driven Tchebycheff scalarization. The property values are computational estimates based on thermomechanical models. SOO denotes single-objective optimization and MOO denotes multi-objective optimization. S1/S2 use three mixing alloys, M1/M2 use four, and L uses five; the paired S and M subprojects investigate different mixture resolutions. The MOO records in this release contain four alloy-fraction columns and both target properties. When using these data, please cite both this dataset's Zenodo record and the related paper.