A. Neves
(Applied Materials Technology, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany; Institute of Hydrogen Technology, Helmholtz-Zentrum Hereon GmbH (hereon), Max Planck-Str. 1, 21502, Geesthacht, Germany)
Hydride materials that can reversibly abs-/desorb hydrogen have been intensively investigated due to their potential as a hydrogen storage medium and functional materials for many applications. The dynamic reaction between hydride/hydride-forming material and the gaseous phase is complex and has several intermediary processes, making the insightful description of these phenomena impractical, especially for calculations. However, such hydride material formation/decomposition is assumed to occur in steps, and the slowest step characterizes the overall dynamic process. This step is called the rate-limiting step (RLS). With the help of experimental data and empirical equations that describe each different RLS, it is possible to model the most relevant processes better. In addition, differential equations can be obtained for the numerical design and optimization of the storage systems. However, this previous approach requires a time-consuming analysis of the experiments. Assessment is only possible with expert knowledge of materials behavior to determine the most likely RLS of each material at a specific combination of temperature and pressure. In order to overcome these fundamental limitations, a machine learning approach is used here for the first time, in which surrogate models are learned to describe the materials behavior.
This work conducts kinetic measurements with an AB2 hydride-forming alloy (Hydralloy C5) in different temperature and pressure conditions. The gathered data is employed to develop non-supervised machine learning models to identify the RLS. The results show an accuracy of more than 80% considering only the best-ranking model and close to 97% when very close second and third matches based on the R² values of the original analysis are considered. The application of machine learning methods to the kinetics of hydrides facilitates and accelerates the development of a streamlined, highly automated determination of the kinetic parameters and the kinetic equations for application in hydride-based system designs.
A. Neves
(Applied Materials Technology, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany; Institute of Hydrogen Technology, Helmholtz-Zentrum Hereon GmbH (hereon), Max Planck-Str. 1, 21502, Geesthacht, Germany)
C. Fritsch
(Applied Materials Technology, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany; Institute of Hydrogen Technology, Helmholtz-Zentrum Hereon GmbH (hereon), Max Planck-Str. 1, 21502, Geesthacht, Germany)
J. Jepsen
(Applied Materials Technology, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany; Institute of Hydrogen Technology, Helmholtz-Zentrum Hereon GmbH (hereon), Max Planck-Str. 1, 21502, Geesthacht, Germany)
J. Puszkiel
(Applied Materials Technology, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany; Institute of Hydrogen Technology, Helmholtz-Zentrum Hereon GmbH (hereon), Max Planck-Str. 1, 21502, Geesthacht, Germany)
M. Passing
(Institute of Hydrogen Technology, Helmholtz-Zentrum Hereon GmbH (hereon), Max Planck-Str. 1, 21502, Geesthacht, Germany)
O. Niggemann
(Computer Science in Mechanical Engineering, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany)
T. Carraro
(Applied Mathematics, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany)
T. Klassen
(Applied Materials Technology, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany; Institute of Hydrogen Technology, Helmholtz-Zentrum Hereon GmbH (hereon), Max Planck-Str. 1, 21502, Geesthacht, Germany)
V. R. Hosseini
(Applied Mathematics, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany)
W. Großmann
(Computer Science in Mechanical Engineering, Helmut Schmidt University (HSU), University of the Federal Armed Forces, Holstenhofweg 85, 22043, Hamburg, Germany)
There are no materials yet.
