Partager

Document IIF

Algorithmes d’apprentissage automatique et de criblage à haut débit pour l’optimisation de l’effet magnétocalorique dans les alliages de Heusler composés uniquement de métaux de type d.

Machine learning and high-throughput screening algorithms for optimization of magnetocaloric effect in all-d-metal Heusler alloys.

Numéro : 0026

Auteurs : BAIGUTLIN D., SOKOLOVSKIY V., BUCHELNIKOV V., TASKAEV S.

Résumé

This paper examines the application of regression models using active machine learning techniques to predict the structural and magnetic properties, as well as to estimate the magnetocaloric effect of all-d metal Heusler alloys. The accuracy of the model was determined by cross-validation using the coefficient of determination R2 and the root mean square error RMSE. The model predictions were compared with experimental data and the results of density functional theory (DFT) calculations. The resulting regression model exhibits high accuracy for structural properties, although difficulties in predicting magnetic moments are noted due to the limited representation of magnetic states in the training dataset. The model is capable of qualitatively predicting martensitic transition stoppages for Ni-Co(Fe)-Mn-Ti systems. An improvement of the model could be achieved by extending the training data set to include other possible magnetic states and types of structural disorder.

Documents disponibles

Format PDF

Pages : 6

Disponible

  • Prix public

    20 €

  • Prix membre*

    Gratuit

* meilleur tarif applicable selon le type d'adhésion (voir le détail des avantages des adhésions individuelles et collectives)

Détails

  • Titre original : Machine learning and high-throughput screening algorithms for optimization of magnetocaloric effect in all-d-metal Heusler alloys.
  • Identifiant de la fiche : 30032634
  • Langues : Anglais
  • Sujet : Technologie
  • Source : 10th IIR Conference on Caloric Cooling and Applications of Caloric Materials
  • Date d'édition : 24/08/2024
  • DOI : http://dx.doi.org/10.18462/iir.thermag.2024.0026

Liens


Voir d'autres communications du même compte rendu (29)
Voir le compte rendu de la conférence

Indexation