高边坡开挖变形非线性时间序列预测分析.pdfVIP

第 25 卷 增 1 岩石力学与工程学报 Vol.25 Supp.1 2006 年 2 月 Chinese Journal of Rock Mechanics and Engineering Feb. ，2006 高边坡开挖变形的非线性时间序列预测分析 周家文，徐卫亚，石安池 (河海大学 岩土工程研究所，江苏 南京 210098) 摘要：在岩体高边坡开挖过程中，可以得到现场的位移监测数据，如何利用现场监测数据来预测高边坡的开挖变 形是一件很有实用价值的工作。根据高边坡开挖变形时间序列的非线性特征，应用局域法对三峡高边坡的位移进 行了预测分析。把局域法的思想引入到神经网络中去，按照寻找邻近点的原理构造出训练样本，通过神经网络得 到的预测值与局域法得到的预测值很接近，并且可以大大地节约计算时间。计算结果表明，对于岩土体工程中的 一维监测数据，通过非线性时间序列分析方法可以对其进行预测分析，该方法具有较高的实用价值。 关键词：岩石力学；开挖变形；非线性时间序列；局域法；混沌；神经网络 中图分类号：TU 457 文献标识码：A 文章编号：1000 –6915(2006)增 1 –2795 –06 APPLICATION OF NONLINEAR TIME SERIES ANALYSIS TO EXCAVATION DEFORMATION PREDICATION OF HIGH SLOPE ZHOU Jiawen ，XU Weiya ，SHI Anchi (Institute of Geotechnical Engineering ，Hohai University ，Nanjing ，Jiangsu 210098 ，China) Abstract ：In the course of high rock slope excavation，the deformation data in the locale can be monitored ，and it′s useful to predicate the excavation deformation of high slope with the monitor data. According to the nonlinear characteristics of the excavation deformation of high slope，the displacements of the high slope of the Three Gorges are predicted by local-region method. The idea of the local-region method is introduced to the neural network ，and the training samples are formed according to the theory of finding near points. The predicated displacements by the trained neural network are very close to those by the local-region method，and computational time is saving. The result shows that ，based on the one-dimensional monitoring data ，the displacement can be predicted by the method of nonlinear time series ，and the method has practical value. Key wo