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 The traditional methods of displacement back analysis generally need to optimize means. This paper attempts to apply artificial neural network, a newly emerging nonlinear science, to the problem of back-analysis of foundation pit displacement. The finite element program simulating pit excavation is used as a forward modeling tool , The BP network is used as the inversion tool, and the positive inversion and the inversion are organically combined by the sample mapping. Finally, an example is given to demonstrate that it is feasible to combine the artificial neural network with the finite element method