Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 13: Data-driven Materials Science: Big Data and Workflows I
MM 13.8: Talk
Tuesday, March 10, 2026, 12:15–12:30, SCH/A251
Predictive and interpretable machine learning models for thermodynamics tuning of metal hydrides for hydrogen storage — •Sinan S. Faouri1, Kai Sellschopp2,3, Claudio Pistidda3, and Paul Jerabek3 — 1Mechanical and Industrial Engineering Department, Applied Science Private University, Amman, Jordan — 2Department of Chemical and Process Engineering, University of Canterbury, Christchurch, New Zealand — 3Institute of Hydrogen Technology, Helmholtz-Centre Hereon, Geesthacht, Germany
Metal hydrides remain among the most promising materials for solid-state hydrogen storage due to their tunable thermodynamic behavior. However, predicting key properties such as equilibrium pressure and hydrogenation enthalpy remains challenging, especially across diverse alloy systems. In this work, we explore feature-based machine learning strategies to model these thermodynamic quantities from elemental descriptors and derived structural features. The study combines experimental and computational data to identify the most relevant predictors governing hydrogen absorption thermodynamics. Particular attention is given to the relationship between atomic-scale size parameters, electronic features, and their collective influence on pressure-enthalpy correlations. The results demonstrate that data-driven approaches can reveal non-obvious structure-property relationships and guide the search for alloys with optimized storage performance. The presented framework offers a step toward integrating machine learning with physical insights for accelerated discovery of functional hydrides.
Keywords: Metal hydrides; Machine learning; Thermodynamics; Hydrogen storage; Equilibrium pressure