复杂应力路径人工冻土三轴剪切试验及本构模型
Elastic-plastic constitutive model of artificial frozen soil based on cascade-correlation artificial neural network
-
摘要: 构建人工冻土BP神经网络本构模型,利用冻土三轴实验数据对神经网络模型进行训练,并2次开发有限元本构程序。用BP神经网络本构模型能很好的反映人工冻土非线性,人工冻土三轴数值模拟值和实验值误差在2.43%范围内。通过深井冻结工程数值模拟表明:人工冻土BP神经网络本构模型能较好的描述复杂应力路径变形特征,数值结果和现场实测规律一致,且和实测位移误差在5.0%;应用BP神经网络冻土本构模型准确预测预报人工冻土帷幕应力场和变形场,为冻结工程设计与施工提供参考。Abstract: Elastic-plastic consistent stiffness matrix was derived based on a neural network constitutive model for frozen soil.Neural network model was trained by the triaxial test data and the traditional finite element constitutive model was replaced by the trained model.The calculated results of the same properties and moisture content of frozen soil were compared with the experimental results.It was found that this neural network constitutive model represents the nonlinear response of the material very well, can improve the numerical analysis results, and agrees very well with the measured results.The deep freezing engineering numerical simulation results show that the artificial frozen soil BP neural network constitutive model can well describe the complex stress path deformation characteristics.The numerical results and site measurement are consistent and the measured displacement error is 5.0%; The BP neural network constitutive model accurately predicts the stress field and deformation field of artificial frozen soil curtain, and provides some guidance for freezing engineering design and construction.