交指型缺陷接地结构共面波导的神经网络模型

Neural Network Model of Interdigital Defect Grounding Structure Coplanar Waveguide

提出一种新型交指型缺陷接地结构(DGS)共面波导,采用交指电容代替常规结构中的缝隙电容,交指电容DGS更能使电容的耦合加强,大大改进了DGS的特性。然后对交指型DGS共面波导建模,对DGS的尺寸和工作频率进行取样,作为多层感知器的输入样本,采用BP(反向传播)算法对该多层感知器进行训练,得到适合样本的神经网络模型。经过训练的神经网络模型在仿真过程中具有计算速度快、设计精度高、泛化能力强、仿真结果和实验测试结果一致性好的特点,从而验证了神经网络的有效性,为电路设计提供了理论依据。

A new kind of inter-digital Defected Ground Structure（DGS）coplanar waveguide adopting inter-digital capacitor to replace slot capacitor in regular structure was put forward,and inter-digital capacitor DGS can enhance coupling of capacitor better and improve the characteristics of DGS.Then inter-digital DGS coplanar waveguide was modeled,and dimensions and operating frequency of DGS were sampled,which was taken as the input samples of multi-layer perceptron,BP（back-propagation） algorithm was adopted to train multi-layer perceptron and get neural network model of suitable samples.Neural network model which has been trained has a lot of characteristics in the simulation such as fast calculation,high design precision,strong generalization ability,well consistency of simulation and experiment result.Those validate the efficiency of neural network and supply theoretical basis for circuit design.