基于支持向量机和改进BP神经网络的路基边坡稳定性研究  

Study on Stability of Roadbed Slope Based on SVM and Improved BP Neural Network

作　　者：史笑凡 杨春风 王可意[3] SHI Xiao-fan;YANG Chun-feng;WANG Ke-yi(School of Civil and Transportation Engineering,Hebei University of Technology,Tianjin 300401,China;Key Lab of Civil Engineering,Hebei University of Technology,Tianjin 300401,China;Hebei University of Engineering,Handan Hebei 056000,China)

机构地区：[1]河北工业大学土木与交通学院,天津300401 [2]河北工业大学土木工程重点实验中心,天津300401 [3]河北工程大学,河北邯郸056000

出　　处：《公路交通科技》2019年第1期31-37,共7页Journal of Highway and Transportation Research and Development

基　　金：河北省交通运输厅科技计划项目(T2012129)

摘　　要：针对京新高速公路项目在建设中遇到的裂缝、滑移、倾倒等大量边坡稳定性问题,为了探讨边坡岩土体参数与边坡稳定性间的相关关系,以及保证研究项目路段在运营期间的行车安全,实现公路网尤其是山区公路的安全、高效、便捷运行,在已有研究的基础上,分别建立了支持向量机以及附加动量因子mc而改进后的BP神经网络两种边坡稳定性预测模型。通过引入45个训练样本,对5个工程边坡实例的安全系数进行预测计算,分析了两种模型的平均误差和最大误差,比较了两种模型的预测精度和适用范围,并且对京新高速公路胶泥湾至冀晋界路段的工程边坡稳定性进行了预测。结果显示,样本训练阶段,支持向量机和BP神经网络两种模型均具有较高的模拟精度,而BP神经网络更优;在样本预测阶段,支持向量机的预测精度明显优于BP网络;当随着样本容量不断增大时,两种计算模型的预测精度也逐渐提高;通过结果可以得出,支持向量机预测模型有较强的外推能力和预测计算的有效性,可以更好地描述边坡稳定性复杂的非线性关系,更适用于边坡稳定性的预测分析。Aiming at a large number of slope stability problems such as cracks, slips and dumping encountered in the construction of Beijing-Xinjiang expressway project,in order to explore the correlation between rock mass parameters and slope stability,to ensure the driving safety of the research project sections during operation,and to realize the safe,effective and convenient operation of the highway in mountain area,on the basis of the existing research,2 corresponding slope stability prediction models based on SVM and improved BP neural network with additional momentum factor mc are established respectively. By introducing45 training samples,predictive calculation of the safety factors for 5 engineering slope instances is conducted,the mean errors and maximum errors of the 2 models are analysed,the prediction accuracies and application ranges of the 2 models are compared,and the stability of the slope of Jiaoniwan to Hebei-Shanxi boundary section in Beijing-Xinjiang expressway is predicted. The result shows that( 1) at sample training stage,both the 2 models have higher simulation accuracy,and the BP network is the better;( 2) at sample prediction stage,the prediction accuracy of SVM is obviously better than that of BP network;( 3) as the sample size increases,the prediction accuracies of the 2 computational models increase gradually. From the result it can be drawn that the SVM prediction model has a strong extrapolation ability and the effectiveness of prediction calculation,it can describe the nonlinear relation of slope stability better,which is more suitable for prediction analysis of slope stability.

关 键 词：道路工程 边坡稳定性 支持向量机 路基滑坡 BP神经网络 

分 类 号：TU416.1[建筑科学—岩土工程]