基于人工神经网络的NiZn铁氧体结构不敏感性能的预测模型

Prediction model for structure-insensitive properties of NiZn ferrites based on artificial neural network

为了系统研究配方对铁氧体电磁性能的影响,制备了一系列Mn2+、Ge4+和Si4+替代的NiZn铁氧体材料,建立了铁氧体配方与结构不敏感性能之间的人工神经网络预测模型。利用所建立的模型研究了ZnO对NiZn铁氧体3个结构不敏感性能居里温度、磁饱和强度及介电常数的影响规律,以及多个组分的交互作用。结果表明:模型的预测结果与实验结果吻合良好,二者的相对误差较小。ZnO含量的增加会导致铁氧体居里温度下降,但会提高饱和磁化强度和介电常数。NiO和ZnO的交互作用对铁氧体的结构不敏感性能影响明显。利用模型得到的铁氧体性能-成分等值线图对寻找最佳配方有较高参考价值。

In the present paper, a series of Mn^2＋, Ge^4＋ and Si^4＋ substituted NiZn ferrites were prepared by conventional ceramic processing in order to study the effect of components on magnetic properties of NiZn ferrite materials. A model on the correlation between composition and structure-insensitive properties of NiZn ferrite materials was developed by using artificial neural network （ANN）. The influences of ZnO content or interaction among components on curie temperature, saturation magnetization and dielectric constant are respectively discussed based on the ANN model. The results indicate that the predicted values from the trained network outputs track the measured values very well, and their relative errors are rather low. With increasing the ZnO content, the curie temperature of the ferrite go down, whereas both the saturation magnetization and the dielectric constant will go up. The interaction among NiO and ZnO play a great role on the three structure-insensitive properties of NiZn ferrite. The contour map of composition versus properties obtained by using ANN model will be a effective approach to optimal component of ferrites.