In this paper, a hybrid model CvBlock-SVR combining convolutional neural network (CNN) and support vector regression (SVR) is proposed for predicting the properties of magneto-thermal effect materials. The chemical composition descriptors are constructed by Magpie and the material features are augmented using a convolutional block (CvBlock), and the augmented features are subsequently modelled with SVR to accurately predict the transition temperature (T_tr) of the material. Training the pervasive CvBlock-SVR model on a dataset of all magneto-thermal effect materials achieved R² = 0.871 and MAE = 24.937 K for Ttr prediction.The model was also successfully applied to the prediction of materials space such as Gd-Al, Gd-Co-Al, and Fe_93-x-yZr₇B_xCu_y, which verified its high accuracy and wide applicability. It is shown that the CvBlock-SVR model not only provides accurate prediction results, but also has wide applicability and is suitable to be applied to the prediction of the properties of different types of magneto-thermal effect materials.

Jian Lan. Performance Prediction of Magnetocaloric Materials Using CNN and SVR. Academic Journal of Materials & Chemistry (2025), Vol. 6, Issue 2: 57-62. https://doi.org/10.25236/AJMC.2025.060208.

[1] Gutfleisch O, Willard M A, Brück E, et al. Magnetic materials and devices for the 21st century: stronger, lighter, and more energy efficient [J]. Advanced materials, 2011, 23(7): 821-842.

[2] Guillou F, Pathak A K, Paudyal D, et al. Non-hysteretic first-order phase transition with large latent heat and giant low-field magnetocaloric effect [J]. Nature communications, 2018, 9(1): 2925.

[3] Park I, Jeong S. Development of the active magnetic regenerative refrigerator operating between 77 K and 20 K with the conduction cooled high temperature superconducting magnet [J]. Cryogenics, 2017, 88: 106-115.

[4] Brown G. Magnetic heat pumping near room temperature [J]. Journal of Applied Physics, 1976, 47(8): 3673-3680.

[5] Gottschall T, Skokov K P, Fries M, et al. Making a cool choice: the materials library of magnetic refrigeration [J]. Advanced Energy Materials, 2019, 9(34): 1901322.

[6] Zarkevich N A, Zverev V I. Viable materials with a giant magnetocaloric effect [J]. Crystals, 2020, 10(9): 815.

[7] Janiesch C, Zschech P, Heinrich K. Machine learning and deep learning [J]. Electronic Markets, 2021, 31(3): 685-695.

[8] Xue L, Shao L, Li Z, et al. Utilization of high entropy in rare earth-based magnetocaloric metallic glasses [J]. Journal of Materials Research and Technology, 2022, 18: 5301-5311.

[9] Huo J, Zhao D, Bai H, et al. Giant magnetocaloric effect in Tm-based bulk metallic glasses [J]. Journal of non-crystalline solids, 2013, 359: 1-4.

[10] Quintana-Nedelcos A, Leong Z, Morley N. Study of dual-phase functionalisation of NiCoFeCr-Alx multicomponent alloys for the enhancement of magnetic properties and magneto-caloric effect [J]. Materials Today Energy, 2021, 20: 100621.

[11] Gómez-Polo C, Socolovsky L, Denardin J, et al. Magnetocaloric effect in FeCr soft magnetic nanocrystalline alloys [J]. Journal of Magnetism and Magnetic Materials, 2007, 316(2): e876-e878.

[12] Gschneidner Jr K, Pecharsky V K. Magnetocaloric materials [J]. Annual review of materials science, 2000, 30(1): 387-429.

[13] Nam C. Magnetocaloric properties predicted by deep learning with compositional features for bulk metallic glasses [J]. Journal of Non-Crystalline Solids, 2024, 624: 122723.