摘要
研究了移动荷载作用下径向基神经网络在地基土的参数识别中的应用。针对径向基神经网络的不足,采用压缩映射递阶遗传算法来确定网络参数(连接权、隐节点中心和宽度),改进算法的收敛性。基于频率响应,采用主元分析方法,有效减少了网络的输入节点。结果表明该方法是可行的。
This paper investigated a new estimation method of soil properties under moving load by using radial basis function（RBF） neural network. A contractive mapping and hierarchical genetic algorithm for RBF neural network was proposed to train network parameters such as connection weights, centres and widths. In addition, the improving algorithm converged faster and could avoid falling into local minima. This research used the frequency response data as input to RBF neural network. By applying a principal component analysis （PCA）, the number of input nodes was reduced largely. This method was proved to be practical in estimating the parameters.
出处
《武汉理工大学学报》
CAS
CSCD
北大核心
2005年第12期53-56,共4页
Journal of Wuhan University of Technology
基金
湖北省自然科学基金(2003ABA015)
关键词
移动荷载
地基土
径向基神经网络
压缩映射递阶遗传算法
主元分析
moving load
soil
radial basis function（RBF） neural network
pricipal component analysis（PCA）
ground vibration
