摘要
边坡系统是一个影响因素众多、非常复杂的非线性系统,使得作为边坡内部力学现象外部表现的边坡变形同样具有很强的非线性特征,而神经网络所具有的高度鲁棒性、学习和联想记忆功能及数据挖掘等特性,对诸如存在内在联系的单时间序列的边坡位移预测有着较大的优势。以此为出发点,通过对单时间序列特点的分析,构造了基于单时间序列的神经网络预测模型,并以渝黔高速公路某边坡位移实际监测数据为例进行了计算。研究结果表明,通过挖掘边坡位移序列中的隐含信息,运用单时间序列BP神经网络进行边坡位移预测是完全可行的,预测平均误差仅为2.72%,预测结果与实际情况吻合度较高。最后通过与传统灰色理论预测方法进行对比发现,该方法预测效果明显提高,预测误差平均降低了近8倍。
Slope system is a very complicated nonlinear system which is influenced by many factors, thus, as an explicit behavior of the inherent mechanics phenomena, the displacement of slope are characterized with the strong non-linear of randomness and indetermination. Because neural network has the performances of the powerful robustness, learning and associative memory function, and data mining, so it has obvious advantage to predict placement of slope to the data of single time series which existing inner link. Take this as the starting point, the prediction model is estab- lished based on time - series and neural network after concluding and analyzing the real observation data. As a test, this model was used in displacement prediction of Yuqian high - way slope. The results of engineering case show that BP neural network is feasible to predict the data of single time series through mining crytic information the average error is 2.27%. After comparing with the method of the traditional gray theory, it is proved this model has the high precision and high fitting degree.
出处
《地下空间与工程学报》
CSCD
北大核心
2009年第6期1418-1421,共4页
Chinese Journal of Underground Space and Engineering
基金
云南省教育厅科学研究基金(07C40062)
云南省科技厅科技创新工程计划(2008KA001)
昆明理工大学博士基金(14118055)
关键词
边坡
时间序列
神经网络
位移预测
slope
time series
neural network
displacement prediction
