摘要
为尝试采用遗传神经网络法解决无渗漏量资料的多目标渗流反分析问题,根据遗传神经网络的非线性映射特性,提出了基于遗传神经网络的初始渗流场反演方法,采用正交设计法设计渗流场参数样本,通过有限元分析获得钻孔水位样本,并利用遗传神经网络学习钻孔水位与渗流场各参数的非线性关系得到各参数的反演值。以卡拉水电站右岸坝区为例,反演了岩体和结构面的渗透系数和右岸边界水头,验证表明该方法在渗流场反演中具有较高的精度。
In order to solve multi-objective seepage back analysis problems with no leakage quantity data, genetic neu- ral network is proposed to back analysis initial seepage field based on the nonlinear mapping characteristics of genetic neu- ral network. Parameter sample of seepage filed is designed with orthogonal test. The sample of water level in the position of bore is obtained by using finite element method. The nonlinear relationship between water level in the position of bore and parameters of seepage field is trained with genetic neural network and then the parameter inversion value is obtained. Taking right bank dam site area of Kala hydropower station for an example, permeability coefficient of rock and structural plane and right bank boundary water head are inversion. The results show that the proposed method has high accuracy in back analysis of initial seepage field.
出处
《水电能源科学》
北大核心
2012年第12期74-77,共4页
Water Resources and Power
基金
国家重点基础研究发展计划(973计划)基金资助项目(2011CB013503)
新世纪优秀人才支持计划基金资助项目(NCET-09-0610)
关键词
水电工程
坝区
初始渗流场
反分析
遗传神经网络
hydropower engineering
dam site area
initial seepage field
back analysis
genetic neural network
