# Fileset

[REPM2025_P2-3_Hosoi.pdf](https://mdr.nims.go.jp/filesets/7c48ac47-b446-47b3-b836-bff6dafd8b6c/download)

## Creator

Hyuga Hosoi, Yamano Hayate, Kinoshita Akihito, Sakuma Noritsugu, Umetani Yusuke, Shoji Tetsuya, Thomas Schrefl

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Multi-objective optimization of magnet compositions by machine learning](https://mdr.nims.go.jp/datasets/7ac9fc1c-3495-4264-93c3-05f9f66ebaa4)

## Fulltext

●The demand for electric motors is increasing as aresult of the rapid production of electrified vehicles.●Rare earth elements (Nd, Pr) added to NdFeB magnetcould face imbalances in supply and demand. (Fig.1)We aim to rapidly develop various types of magnets optimizedfor different products, utilizing cost-effective elements (Ce, La).●Prepared magnetic powders (Fig.2)(Nd,Ce,La,Pr,Dy,Tb)13.55(Fe,Co,Ni)80.54(B,C)5.91Classified single-crystal-like powders were obtained.●The physical properties evaluation(Ms and Ha at 300～453 K)Ms: saturation magnetizationHa: anisotropic magnetic field●Regression analysis of Ms and Ha (Fig.3)Explanatory variables:Ce, La, Pr, Dy, Tb, Co, Ni, C, tempreture ⇒Good predictive accuracy (R2≧0.89)●Multi-objective optimization (Fig.4)Optimization method: genetic algorithm・Ms and Ha ※maximization search・cost ※minimization searchOptimal solutions were obtained and visualized.●Evaluation of the predicted values (Fig.5)The predicted values were achieved by experiment.Immediate proposal of magnet compositions suited to different application is possible in response to change in market conditions. These analysis (regressions, multi-objective optimization and so on) can be easily performed using WAVEBASE system. ※WAVEBASE: Could-Based Material Data Analysis serviceHyuga Hosoi1, Yamano Hayate1, Kinoshita Akihito1, Sakuma Noritsugu1, Umetani Yusuke1, Shoji Tetsuya1, Thomas Schrefl2(1. Toyota Motor corporation (Japan), (2. Donau University Krems (Austria)Multi-objective optimization ofmagnet compositions by machine learningBackground / ObjectiveResearch HighlightsKeywordsReferences(1) H. Yamano, et al: Efficient optimization approach for designing power device structureusing machine learning, Japanese Journal of Applied Physics, 62, SC1050 (2023).(2) M. Yano, et al: Material data analysis cloud service “WAVEBASE”, TOYOTA Technical Review, 69, 48 (2023).NdFeB, Machine learning,Genetic algorithm, WAVEBASEMulti-objective optimization040801201602002013/1/6 2015/1/6 2017/1/5 2019/1/5 2021/1/4 2023/1/4PrNdCe La FeCost [dollar/kg]Fig.2. Experiment procedureFig.3. Regression analysis results of Ms and HaFig.1. Trends in raw material costsR2=0.90RMSE=0.52Anisotropic magnetic field, HaSaturation magnetization, MsR2=0.89RMSE=0.05Ms[T] at 300KSample numberFig.4. Results of the multi-objective optimization Fig.5. Evaluation of the predicted valuesCost[dollar/kg]Ms[T] at 300KHa[T] at 300KNd2Fe14Bcost=0.73(Nd0.81Ce0.14La0.05)2Fe14Bcost=0.60(Nd0.91Dy0.09)2Fe14(B0.8C0.2)cost=0.87Nd2Fe14(B0.88C0.12)cost=0.73 スライド 1