摘要
为研究复杂地质体海水入侵渗透系数的不确定性,将非线性动力学神经网络用于海水入侵的三维含水层参数的逆反分析,建立了基于监测水头值的海水入侵渗透系数直接识别方法.采用正交设计试验方法进行三维数值试验产生代表性的样本数据,通过误差逆传播学习算法建立多层的神经网络模型,表示观测水头和渗透系数隐式的复杂非线性映射关系,通过观测水头信息直接预测含水层渗透系数.该方法可利用现有的海水入侵三维有限元模拟程序.算例表明该法具有良好的分析精度,可满足工程要求.
An immediate identifying method was established based on water head values to study the uncertainty characteristic of complex geology permeability coefficient of .salt water intrusion. Nonlinear dynamics of artificial neural network (ANN)was used in inverse back analysis of 3 - D aquifer parameters. Representative samples data was produced by 3 - D numerical simulation using orthogonal design test method. Neural network model mapping the nonlinear relation between monitoring hydraulic heads and parameters of aquifer was constructed by error back propagation arithmetic. Then the permeability coefficients of 3 D aquifer could be identified by ANN model. This method can utilize the large scale sea water intrusion finite element program. Computing samples indicate that the proposed method has good analysis precision and can meet the requirements of engineering designs.
出处
《大连海事大学学报》
EI
CAS
CSCD
北大核心
2007年第4期105-108,共4页
Journal of Dalian Maritime University
关键词
海水入侵
渗透系数
BP神经网络
逆反分析
sea water intrusion
permeability coefficient
BP neural network
inverse back analysis