考虑剪胀的土的非线性本构关系神经网络模型

土的本构关系相当复杂，受诸多因素影响，这给土的本构关系模型的建立带来很大困难。提出了土非线性本构关系神经网络模型。该模型考虑了p对剪应变和q对体应变的影响，可有效地反映土的剪缩或剪胀现象，同时建模方便，直接利用试验结果对网络进行训练来确定模型参数。此外，将该模型纳入了有限元程序，通过对一基础沉降计算证实了该模型的有效性。

NONLINEAR NEURAL NETWORK MODEL OF CONSTITUTIVE RELATION OF SOIL WITH DILATION CAUSED BY SHEAR STRESS

The neural network model of soil is proposed to show the influence of p on shear strain and q on volume strain. Hence the volumetric expansion and shrinkage caused by shear stress can be reflected. At the same time, this model can be set up conveniently, in which the parameters of network can be directly determined by training the network with the results of triaxial compression tests. In addition, this model is introduced into FEM program. The settlements of a foundation are calculated by using neural network model, which demonstrated the effectiveness of the neural network model.