Implementation of Backpropagation Neural Network for Prediction Magnetocaloric Effect of Manganite

Abstract

In the field of magnetic cooling technology, there is still much to learn about the magnetocaloric properties of magnetic cooling materials. Research into magnetocaloric manganites exhibiting a significant maximum magnetic entropy change in the vicinity of ambient temperature yields encouraging outcomes for the advancement of magnetic refrigeration apparatus. Through a combination of chemical substitutions, changes in the amount of oxygen present, and different synthesis techniques, these manganites undergo lattice distortions that result in pseudocubic, orthorhombic, and rhombohedral structures instead of perovskite cubic structures. The present investigation used backpropagation neural networks (BPNNs) to investigate the correlations among maximum magnetic entropy change (MMEC), Curie temperature (Tc), lanthanum manganite compositions, lattice properties, and dopant ionic radii. Simbrain 3.07 was used to execute the BPNN model, and the suggested model accuracy was examined using coefficient determination. As a result, the model's predicted values for the mean absolute error, root mean square, and coefficient correlation for MMEC are 0.012, 0.022, and 0.9861, respectively. The model predicts that the Curie temperature mean absolute error, root mean square, and coefficient correlation will be 0.015, 0.021, and 0.9947, respectively. Based on these results, BPNN has the potential to be applied in predicting the MMEC and Tc of manganite as preliminary decision during experiments.