一种新型缺陷接地结构的BP神经网络模型

The BP Neural Network Model of a Novel Defected Ground Structures

组合式非周期性缺陷接地结构(CNPDGS)是在微波电路的接地金属平面上人为地蚀刻出特殊形状的非周期性"缺陷",改变接地电流的分布,从而改变传输线的频率特性。本文针对一种新型的具有双阻带特性的CNPDGS,在场分析的基础上建立了其BP神经网络模型,将其结构尺寸和频率作为输入样本,传输系数参数作为输出样本,采用贝叶斯正则化算法对神经网络进行训练。神经网络训练完成后,在学习范围内将其结构尺寸和频率输入到神经网络模型,从输出端立即得到准确的传输系数。最后通过了实验验证,进一步说明了神经网络模型的正确性和有效性,为DGS的分析和设计提供了新的有效途径。

Combinatorial nonperiodic defected ground structures （CNPDGS） are the structures that are etched on the cireuit~ ground plane with nonperiodic units of special form, the ground current distribution can be changed, and the frequency properties of the transmission lines can be influenced. For a novel CNPDGS with the characteristic of double stop band. Artificial neural network （ANN） model is developed on the basis of FDTD analysis for the first time. The structure size of the CNPDGS and the frequency are defined as the input samples of the ANN model, the parameters of transmission coefficient are defined as the output samples. As the ANN model has been trained with the Bayesian Regularization algorithm, the transmission coefficient of the CNPDGS at any arbitrary sizes and the frequencies within region trained can be obtained quickly from the ANN model. Finally, the ANN model has been approved by results of experimentation. It is also showed that the ANN model is very effective. The ANN model will provide powerful approach for the analysis and design of defected ground structures （DGS）.