摘要
在采用知识神经网络对微波器件进行建模优化时,先验知识大部分源于经验公式,而经验公式的推导十分繁杂,为避免先验知识获取困难,提出了一种将电磁仿真计算中精确网格剖分的计算模型作为教师信号,将粗糙网格剖分计算模型作为先验知识并运用粒子群算法对神经网络进行训练的方法,构建了相应的知识神经网络模型。对双层电磁带隙结构建模并进行优化设计,建模结果和优化效果均十分理想。说明了知识神经网络可以替代优化过程中的精确模型,减少优化所需的时间,证明所述方法的可行性与优越性。
When microwave devices are designed by knowledge-based neural network （KBNN） , the empirical formula is always used as a priori knowledge. However, it is difficult to derive the corresponding formulas. In order to avoid this problem, we use precise-mesh model of EM analysis as teaching signal and coarse-mesh model a priori knowledge to train the neural network by particle swarm optimization （PSO）. The proposed KBNN is applied to dual electromagnetic band-gap, and the results of modeling and optimization are good which proves that the KBNN can re- place precise model in optimization and reduce the computation time and shows effectiveness and superiority of this method.
出处
《计算机仿真》
北大核心
2017年第7期289-293,共5页
Computer Simulation
基金
国家自然科学基金项目(61401180)
