Answers are grounded in this paper and cite their sources.
The maximum energy product, ( BH ) max , is the key performance metric of permanent magnets, governed by the nonlinear interplay between intrinsic magnetic parameters and thermal effects. Accurate prediction of ( BH ) max remains challenging because of the complex coupling among coercivity, magnetization, anisotropy, and temperature-dependent properties. In this work, we develop an explainable machine learning framework to predict ( BH ) max using experimentally accessible magnetic descriptors, including coercivity ( H c ), saturation magnetization ( M s ), remanent magnetization ( M r ), magnetocrystalline anisotropy constant ( K 1 ), Curie temperature ( T c ), and measurement temperature across a heterogeneous dataset comprising Nd-Fe-B, hexaferrite, Sm-Co, AlNiCo and other ferrites. Furthermore, principal component analysis (PCA) reveals that governing magnetic descriptors vary fundamentally across different classes of magnetic materials. PCA maps the magnetic property landscape into two primary axes, PC1 which represents the magnetization-dominated axis and PC2 which represents trade-off between thermal stability-coercivity. The mapping naturally separates different classes of magnetic materials in the PC1–PC2 space and reveals that ( BH ) max is not a random outcome, but rather a predictable optimization within this structured magnetic descriptor landscape. Class-specific feature importance analysis shows that different magnet families are governed by distinct dominant descriptors, confirming that global feature importance may partially obscure the importance of different magnetic descriptor in each class. Hence, the separation of magnetic classes enables the effective optimisation of targeted magnetic descriptor for functional property.
No takes yet. Share an insight, caveat, or question.
Nashier et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: