|
|
|
| New adaptive RBF neural network for micro-strip antenna modeling |
| HUANG Yuan-jun 1， LOU Ping 1， WU Zhi-jun 2， LIN Xiao-feng 3 |
1. Institute of Mechatronics and Automobile, Jiaxing Vocational Technology College, Jiaxing 314036, China;
2. Jiaxing Glead Electronics Co., Ltd., Jiaxing 314003, China;
3.College of Electrical Engineering, Guangxi University, Nanning 530004, China |
|
|
|
Abstract The relationship between the resonance frequency and the size of the structure is multiparameter and nonlinear, which makes it difficult to establish accurate model when design of a rectangular microstrip antenna. To solve this problem, a new way of RBF (radial basis function) neural network algorithm was proposed by combination the methods of dynamic adaptive clustering and the pseudo inverse weighting, which possessed several advantages of unfixed neuron number on hidden layer, by means of adaptive adjustment, good approximation effect, and small amount of calculation, fast learning. Then combined with the better approximation properties of RBF neural network for multivariable and nonlinear functions , it was used for modeling the resonance frequency of the rectangular micro-strip antenna. Experiment results showed that the established neural network model was obviously superior to the result of literature in both accuracy and speed. Thus a new effective method to improve the design efficiency is provided for the developer of antenna design.
|
|
Received: 28 May 2014
Published: 28 October 2014
|
新型自适应RBF神经网络应用于微带天线建模
矩形微带天线设计中，为解决谐振频率与尺寸结构之间存在多参数、强非线性关系而建立其精确模型的困难，提出一种动态自适应聚类与伪逆求权值相结合的新型RBF神经网络算法，其优点为：无需事先固定隐层神经元个数，以自适应调整的方式，获得较好的逼近效果，且计算量小、学习速度快.同时，结合RBF神经网络对于多变量、非线性函数有较好的逼近特性，将其用于矩形微带天线的谐振频率建模.实验表明，建立的神经网络模型无论是在精度还是速度上都明显优于已有文献的结果，从而为天线设计开发者提供了一种新的有效方法，提高设计效率.
关键词：
微带天线,
神经网络,
非线性模型,
自适应
|
Viewed |
|
|
|
Full text
|
|
|
|
|
Abstract
|
|
|
|
|
Cited |
|
|
|
|
| |
Shared |
|
|
|
|
| |
Discussed |
|
|
|
|
