摘要
为研究合金成分对磁体性能的影响,将支持向量回归机应用于磁性能的预测。在小样本情况下应用MATLAB支持向量机工具箱,建立了磁性能的支持向量回归机模型,将合金成分作为输入变量对磁性能进行预测。选择合适的模型参数,最终建立了拟合误差、预测误差均较小的模型,对剩磁、矫顽力以及最大磁能积的预测相对误差平均值分别为1.94%、5.43%及4.34%。试验表明将支持向量回归机用于磁性能的预测是一种可行且有效的方法。
In order to study the effect of alloying compositions on magnetic properties,support vector machines（SVM）were used to predict magnetic properties.In small sample situations,a support vector machine for regression was establised in SVM toolbox of MATLAB.It was used to predict magnetic properties by taking alloy compositions as input.The model parameters were selected suitably to make the errors of fitting and predicting as small as possible.The mean absolute percent error（MAPE）between the measured value and predicted value of Br、H cj and（BH）_ max are 1.94%,5.43% and 4.43%.The results show that SVM is viable and effective in prediction of magnetic properties.
出处
《材料科学与工程学报》
CAS
CSCD
北大核心
2010年第3期453-457,共5页
Journal of Materials Science and Engineering
基金
四川省教育厅重点资助项目(2004A110)
关键词
稀土
纳米复相永磁体
支持向量回归机
结构风险
rare earths
nanocomposite permanent magnet
support vector machines regression（SVMR）
structure risk
