论文部分内容阅读
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 the problem of multi-objective seepage back analysis with no leakage data by using genetic neural network, according to the nonlinear mapping characteristics of genetic neural network, an initial seepage field inversion method based on genetic neural network is proposed. The orthogonal design method The parameters of seepage field are designed, and the water level of the borehole is obtained by finite element analysis. The nonlinear relationship between the water level of the borehole and the seepage field is studied by genetic neural network to get the inversion value of each parameter. Taking the dam bank on the right bank of Kala Hydropower Station as an example, the permeability coefficients of rock mass and structural plane and the head of the right bank boundary are retrieved. The results show that this method has high accuracy in seepage field inversion.