基于回归和神经网络的NdFeB磁性能融合预测模型  

A Fusion Prediction Model Based on Regression and Neural Network for Nano-composite NdFeB Permanent Magnet Alloy

作　　者：胡旺[1] 李志蜀[1] 连利仙[2] 刘颖[2] 

机构地区：[1]四川大学计算机学院,四川成都610065 [2]四川大学材料科学与工程学院,四川成都610065

出　　处：《四川大学学报（工程科学版）》2004年第6期66-69,共4页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家863计划资助项目(2001AA324030)

摘　　要：为了进一步提高纳米复相NdFeB系永磁合金磁性能预测模型的精度和扩大适用范围,根据组合优化理论对均匀设计试验建立的纳米复相NdFeB系永磁合金磁性能的回归预测模型和神经网络预测模型进行了组合优化,提出了一种融合预测模型。结果表明,融合预测模型的精度最高,神经网络模型次之,而回归模型精度最低。融合模型的最大相对误差为2.2%,可以用于纳米复相NdFeB系永磁合金的成分优化设计。验证实验表明该模型具有很好的适应性。A fusion model, based on combination optimizing theory, was proposed to improve the precision and enlarge the range for the predict models of nano-composite NdFeB permanent magnet alloy. The fused model combines the regression model and artificial neural network (ANN) model, which are established by the uniform design experiments for nano-composite NdFeB permanent magnet alloy. The results showed that the prediction precision of the fusion model is the highest, ANN model in the second place, and the regression model in the end. The maximum relative precision of the fusion model reaches 2.2%. The verification experiments showed that the fusion model is applicable.

关 键 词：融合预测模型 回归 神经网络 均匀设计 永磁合金 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]