基于BP人工神经网络的SmFeN永磁材料工艺-磁性能关系预测

Prediction Model for Megnetic Propertites of Smfen Alloy Prepared by Hddr Based on Artificial Neural Network

采用均匀设计的方法设计了4因素16水平的HDDR工艺条件优化实验方案，建立了工艺参数与磁性能之间的神经网络数学模型，利用该模型结合微观结构的变化研究了各工艺参数以及多因素之间的交互作用对磁性能的影响规律，并对工艺条件进行了优化。结果表明：数学模型预测磁性能结果与实测结果吻合良好，剩磁Br的相对误差在2．85％以内，最大磁能积（BH）m的相对误差的在4．60％以内，内禀矫顽力Hcj的相对误差的最大值为6．0％，建立的数学模型有较好的可靠性。采用预测的HI）DR工艺条件，成功制备出了氮含量高、相组成单-的Sm2Fe17Nx稀土永磁材料，其粘结磁体的磁性能为：Br=0．562T，Hcj-1214kA／m，（BH）m-56kJ／m3和预测值吻合很好。

The 4--factors and 16--levels experiments are carried out by the uniform design theory and the relationship be- tween technique parameters and magnetic properties is established by artificial neural network （ANN） prediction model. The technique parameters are optimized by the ANN model. At the same time, the influences of single technique parameter or the interaction among parameters on magnetic properties are respectively discussed according to the curves ploted by ANN model. The result shows that the predicted and measured results are in good agreement. The relative errors are rather low, for exam- ple, 2. 85; for Br, 4. 6%for （BH） m, and6. 0; for Hcj . Sm2Fe17Nx rare permanent magnetic materials which are high nitrogen content and almost single--phase are prepared according to predictive conditions of HDDR. The magnetic properties of bonded magnets are as follows： Br=0. 562T, Hci=1214kA/m, （BH） m=56M/m3 , and they coincide well with predictive values.