基于LS-SVM的粘结NdFeB永磁体磁性能预测

Performance Prediction of Bonded NdFeB Permanent Magnet Based on Least Square Support Vector Machine

基于粘结NdFeB永磁体制备工艺优化实验,建立了一个最小二乘支持向量机(LS-SVM)算法模型用于工艺参数的优化。以粘结剂含量、固化温度、固化时间以及单位压制力大小四个工艺参数为影响因数,以剩余磁感应强度Br、矫顽力Hcj和最大磁能积(BH)m为影响对象,通过最小二乘支持向量机算法模型建立起影响因素与影响对象之间的复杂的非线形关系。针对多影响对象,提出了一种γ和σ选择算法;以均匀设计试验结果为样本进行训练,用训练好的模型进行预测。结果表明,LS-SVM模型的实验结果与预测结果吻合良好,二者相对误差很小,对比ANN模型预测结果,LS-SVM模型具有更高的精度和运算速度,具有很好的实用性。

Based on the optimized preparation process of bonded NdFeB magnets,a least squares support vector machine model was built to optimize process parameters.With Binder content,thermosetting temperature,thermosetting time and pressure as influence factors,Br,Hcj and（BH）m as influence object,a new algorithm was proposed for selecting and,and Uniform Design experiment was taken to get data samples.Experimental results show that the relative error of magnetic properties between their measured value and predicted value are very small,and compared with ANN model,the prediction accuracy and prediction speed are much higher.