【摘要】 传统的位移反分析方法一般均需借助优化手段 ,本文尝试将人工神经网络这一新兴的非线性科学应用于基坑的位移反分析问题 ,以模拟基坑开挖的有限元程序为正演工具 ,以BP网络为反演工具 ,并通过样本的映射关系将正演和反演过程有机地结合起来。最后 ,通过算例验证了将人工神经网络和有限元法结合起来进行基坑位移反分析具有可行性更多还原

【Abstract】 An attempt is made to propose a method to predicted the settlement of a deep foundation pit in the paper. The method is the displacement back?analysis combined with BP neural networks. It is reputed that the artificial neural networks are suited to solve the nonlinear problem. A plastic FEM program is used as a tool in the forward process. The BP neural network is adopted in the backward process. It is reasonably combined the forward and backward process through mapping the realtionship of samples. A multi?Staged excavation of a deep foundation pit can be simulated. Finally, a practical example has shown that the settlements of a deep foundation pit are agreement very well with that of the measurement.更多还原