节点文献

广义回归神经网络在大坝安全监测数据分析中的应用

Research on the Application of General Radial Basis Function Neural Network in Analysis of Dam Safety Monitoring Data

【作者】 李钢

【导师】 徐晖; 邓念武;

【作者基本信息】 武汉大学 , 水工结构工程, 2005, 硕士

【摘要】 目前我国在建和已建成数量众多的大坝和水库,并且坝高和水库的规模有不断扩大的趋势,因此大坝和其他拦水建筑物的安全性尤显重要。 人工神经网络由于其具有强大的自我学习修正误差的能力和能够在理论上逼近任何非线性系统的特点,在大坝安全监测领域内得到了广泛的应用。 本文介绍了回归分析方法、大坝变形观测量的统计模型和人工神经网络的基本概念,并进一步介绍了径向基函数网络模型。该网络模型的输入到输出层的映射是非线性的,而网络的输出对可调参数而言是线性的,而输入矢量是直接映射到隐层的,一旦确定径向基函数的网络中心后这种映射关系就确定了,选用递归正交最小二乘算法可以快速确定网络训练中心和简化径向基网络训练结构。广义回归网络是径向基函数网络的一种重要变型,运用正则化理论确定网络隐层中心,用维数小于输入样本数目的格林函数实现函数逼近,达到简化网络结构和提高训练速度的目的。本文用广义回归网络模型建立大坝安全监测的径向基网络模型。和BP网络相比,广义回归网络避免了BP算法冗长计算过程和陷入局部极小值的可能,且具有更好的拟合精度和预报精度。实例分析证明,广义回归网络模型可以应用于实际的大坝安全监测数据处理工作中去并可以取得更优的分析结果。

【Abstract】 There are many dams and embankments built or building in our country, the height and size of these dams and embankment is increasing. So it’s very important of safety of these.Because of Neural Network’s powerful ability of studying and modifying errors by itself and its characteristic of approaching nonlinear system in theory, it has been applied to many fields.This paper describes regression analysis method , statistic model of dam distortion monitoring data and basic conception of Neural Network., and introduces Radial Basis Function Neural Network model .It’s mapping from input layer to output layer is non-linear , but according as adjustable parameter the network’s output layer is linear .Once the hub of this network is fixed on this mapping relationship is confirmed . In this paper Recursion Orthogonal Least Squares method is applied to fix on training hub rapidly and predigesting framework of the network. General Regression Neural Network is one transfiguration of RBFNN, witch applies theory of regularization and dimension of Green-function witch is less than input layer to predigesting the framework and compute rapidly. In this paper GRNN is applied to construct model of the large dam safety monitoring . Comparing to BP model ,GRNN model avoids appearance of iterative calculating process and local extremum of BP model, and has better combine and prediction precision . Analysis in this paper proves that , GRNN model can be practiced in data processing of the large dam safety monitoring and can get optimized calculation result.

  • 【网络出版投稿人】 武汉大学
  • 【网络出版年期】2006年 05期
  • 【分类号】TV698.1
  • 【被引频次】18
  • 【下载频次】784
节点文献中: 

本文链接的文献网络图示:

本文的引文网络