岩土结构的位移大都具有非线性动力学演化特征，针对目前所采用的时间序列分析方法中存在的模型选择和参数确定困难问题，用人工神经网络建模来取代传统的分析方法，提出了一种能自动确定输入时步长度和网络模型结构的遗传神经网络方法，而且也解决了神经网络结构选择和“过训练”问题，提高了网络学习的效率和网络预测的能力·该方法可以对位移演化规律进行智能识别，获得全局最优解· Most of the displacements of geotechnical structures have nonlinear dynamic evolution characteristics. In view of the problems of model selection and parameter determination in the present time-series analysis methods, artificial neural network modeling is adopted to replace the traditional analysis methods and a This method can automatically determine the genetic-neural network method of inputting the time-step length and network model structure, and also solves the problem of neural network structure selection and “over training”, improves the efficiency of network learning and the ability of network prediction. Displacement evolution of intelligent identification, access to the global optimal solution ·