Predicting hydrogen storage capacity of V–Ti–Cr–Fe alloy via ensemble machine learning

The V–Ti–Cr–Fe quaternary alloy is a promising hydrogen storage material for excellent performances, but it is difficult to take reliable multi-factor synergistic effects into account by means of experiments. At present, using the data-driven innovation method of ensemble learning, the structure-property relationship of V–Ti–Cr–Fe alloy is built and the maximum hydrogen absorption capacity is accurately predicted as well through 19 features covering the composition and various crystal parameters with the mean square error of 0.187. The feature importance ranking indicates that valence electron concentration, lattice constant, and Z/r3 play a critical role in the prediction. The genetic algorithm is furtherly used to propose 3 optimal composition ranges, which are proved to be accurate by experiments with relative errors of around 1%. The present work could provide an effective way for accurate and rapid prediction of hydrogen storage capacity and rational design of high-performance alloys.