Performance prediction of sintered NdFeB magnet using multi-head attention regression models
Abstract The preparation of sintered NdFeB magnets is complex, time-consuming, and costly. Data-driven machine learning methods can enhance the efficiency of material synthesis and performance optimization. Traditional machine learning models based on mathematical and statistical principles are effe...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2024-11-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-024-79435-7 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850244458781605888 |
|---|---|
| author | Qichao Liang Qiang Ma Hao Wu Rongshun Lai Yangyang Zhang Ping Liu Tao Qi |
| author_facet | Qichao Liang Qiang Ma Hao Wu Rongshun Lai Yangyang Zhang Ping Liu Tao Qi |
| author_sort | Qichao Liang |
| collection | DOAJ |
| description | Abstract The preparation of sintered NdFeB magnets is complex, time-consuming, and costly. Data-driven machine learning methods can enhance the efficiency of material synthesis and performance optimization. Traditional machine learning models based on mathematical and statistical principles are effective for structured data and offer high interpretability. However, as the scale and dimensionality of the data increase, the computational complexity of models rises dramatically, making hyperparameter tuning more challenging. By contrast, neural network models possess strong nonlinear modeling capabilities for handling large-scale data, but their decision-making and inferential processes remain opaque. To enhance interpretability of neural network, we collected 1,200 high-quality experimental data points and developed a multi-head attention regression model by integrating an attention mechanism into the neural network. The model enables parallel data processing, accelerates both training and inference speed, and reduces reliance on feature engineering and hyperparameter tuning. The coefficients of determination for remanence and coercivity are 0.97 and 0.84, respectively. This study offers new insights into machine learning-based modeling of structure-property relationships in materials and has potential to advance the research of multimodal NdFeB magnet models. |
| format | Article |
| id | doaj-art-4183d760fe8f4f7696bb412e6469fafc |
| institution | OA Journals |
| issn | 2045-2322 |
| language | English |
| publishDate | 2024-11-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-4183d760fe8f4f7696bb412e6469fafc2025-08-20T01:59:43ZengNature PortfolioScientific Reports2045-23222024-11-0114111210.1038/s41598-024-79435-7Performance prediction of sintered NdFeB magnet using multi-head attention regression modelsQichao Liang0Qiang Ma1Hao Wu2Rongshun Lai3Yangyang Zhang4Ping Liu5Tao Qi6Department of Rare Earth, Jiangxi University of Science and TechnologyDepartment of Physics, Ganjiang Innovation Academy, Chinese Academy of SciencesDepartment of Technology, Jiangxi Guanying Intelligent Technology Co.Ltd.Department of Physics, Ganjiang Innovation Academy, Chinese Academy of SciencesDepartment of Physics, Ganjiang Innovation Academy, Chinese Academy of SciencesDepartment of Technology, Jiangxi Guanying Intelligent Technology Co.Ltd.Department of Physics, Ganjiang Innovation Academy, Chinese Academy of SciencesAbstract The preparation of sintered NdFeB magnets is complex, time-consuming, and costly. Data-driven machine learning methods can enhance the efficiency of material synthesis and performance optimization. Traditional machine learning models based on mathematical and statistical principles are effective for structured data and offer high interpretability. However, as the scale and dimensionality of the data increase, the computational complexity of models rises dramatically, making hyperparameter tuning more challenging. By contrast, neural network models possess strong nonlinear modeling capabilities for handling large-scale data, but their decision-making and inferential processes remain opaque. To enhance interpretability of neural network, we collected 1,200 high-quality experimental data points and developed a multi-head attention regression model by integrating an attention mechanism into the neural network. The model enables parallel data processing, accelerates both training and inference speed, and reduces reliance on feature engineering and hyperparameter tuning. The coefficients of determination for remanence and coercivity are 0.97 and 0.84, respectively. This study offers new insights into machine learning-based modeling of structure-property relationships in materials and has potential to advance the research of multimodal NdFeB magnet models.https://doi.org/10.1038/s41598-024-79435-7Sintered NdFeBMachine learningDeep learningMulti-head self-attention mechanismXGBoost |
| spellingShingle | Qichao Liang Qiang Ma Hao Wu Rongshun Lai Yangyang Zhang Ping Liu Tao Qi Performance prediction of sintered NdFeB magnet using multi-head attention regression models Scientific Reports Sintered NdFeB Machine learning Deep learning Multi-head self-attention mechanism XGBoost |
| title | Performance prediction of sintered NdFeB magnet using multi-head attention regression models |
| title_full | Performance prediction of sintered NdFeB magnet using multi-head attention regression models |
| title_fullStr | Performance prediction of sintered NdFeB magnet using multi-head attention regression models |
| title_full_unstemmed | Performance prediction of sintered NdFeB magnet using multi-head attention regression models |
| title_short | Performance prediction of sintered NdFeB magnet using multi-head attention regression models |
| title_sort | performance prediction of sintered ndfeb magnet using multi head attention regression models |
| topic | Sintered NdFeB Machine learning Deep learning Multi-head self-attention mechanism XGBoost |
| url | https://doi.org/10.1038/s41598-024-79435-7 |
| work_keys_str_mv | AT qichaoliang performancepredictionofsinteredndfebmagnetusingmultiheadattentionregressionmodels AT qiangma performancepredictionofsinteredndfebmagnetusingmultiheadattentionregressionmodels AT haowu performancepredictionofsinteredndfebmagnetusingmultiheadattentionregressionmodels AT rongshunlai performancepredictionofsinteredndfebmagnetusingmultiheadattentionregressionmodels AT yangyangzhang performancepredictionofsinteredndfebmagnetusingmultiheadattentionregressionmodels AT pingliu performancepredictionofsinteredndfebmagnetusingmultiheadattentionregressionmodels AT taoqi performancepredictionofsinteredndfebmagnetusingmultiheadattentionregressionmodels |