基于神经网络的大型深基坑工程土体参数反演

发布时间：2018-05-25 15:45

本文选题：深基坑 + 人工神经网络　；参考：《水力发电学报》2015年07期

【摘要】：有限元方法常用来分析及预测深基坑工程在不同工况条件下的稳定性问题,而分析结果的可靠性取决于计算参数的准确性和模型边界条件的合理性。本文针对大型深基坑工程提出了基于人工神经网络的土体参数反演方法。建立深基坑工程土体参数反演的神经网络时,先由监测数据数量和待反演参数数目确定神经网络的拓扑结构,然后采用遗传算法优化网络的初始连接权值和激活函数阈值,再利用有限元模型计算建立神经网络的训练样本,将实际监测数据输入训练后的网络后即可进行土体参数的反演计算。本文以天津文化中心交通枢纽工程为例,详细介绍了基于神经网络的大型深基坑工程土体参数反演的步骤,并利用反演得到的参数进行其他工况条件下的数值计算,其结果验证了该参数反演方法行之有效。
[Abstract]:The finite element method is often used to analyze and predict the stability of deep foundation pit engineering under different working conditions. The reliability of the analysis results depends on the accuracy of the calculation parameters and the rationality of the boundary conditions of the model. In this paper, an artificial neural network based soil parameter inversion method is proposed for large deep foundation pit engineering. In the establishment of neural network for soil parameter inversion in deep foundation pit engineering, the topological structure of neural network is first determined by the number of monitoring data and the number of parameters to be retrieved, and then the initial connection weight and activation function threshold of the network are optimized by genetic algorithm. Then the training sample of neural network is established by finite element model calculation, and the parameters of soil can be inversed after the actual monitoring data are input into the trained network. Taking the transportation hub project of Tianjin Cultural Center as an example, this paper introduces in detail the steps of soil parameter inversion for large-scale deep foundation pit engineering based on neural network, and carries out numerical calculation under other working conditions by using the parameters obtained from inversion. The results show that the parameter inversion method is effective.
【作者单位】：天津市市政工程设计研究院;香港科技大学土木及环境工程系;清华大学水沙科学与水利水电工程国家重点实验室;
【基金】：国家自然科学基金青年基金项目(51109117) 清华大学自主科研计划(20111081125) 天津市市政公路管理局科技计划项目(2010-18)
【分类号】：TU753

【参考文献】