渗透系数反演的CHNN模型方法

Inverse analysis of percolation coefficients by CHNN model

利用连续型Hopfield神经网络 (ContinuousHopfieldNeuralNetwork ,简称CHNN)的反馈特性 ,结合实测资料和数值计算 ,构建了岩土体渗透系数的人工神经网络反演模型 ,通过网络神经元状态的变迁而最终稳定于平衡状态 ,从而得到渗透系数反演优化计算的结果。经实例验证 ,效果较好。

Based on the inverse characteristics of the continuous Hopfield neural network(CHNN) model, combining with the observed data and numerical calculation results of groundwater level, an artificial neural network inverse analysis model for percolation coefficients of rock and soil body is established. Through employing the properties of self-astringency of net-neural unit to finally trend towards a balance status, an inverse optimal result can be found. It is verified from an illustration that the computed results are in good agreement with the observed data.