摘要
近年来,边坡稳定性预测得到了广泛的研究,及时、准确的预测可以有效地预防边坡破坏灾害的发生。提出了一种基于相关向量机(relevance vector machine,RVM)的边坡稳定性预测模型,结合京-新高速公路高堑边坡工程实例,通过对比支持向量机(support vector machine,SVM)模型、径向基函数(radical basis function,RBF)神经网络模型和RVM模型的拟合及预测结果来分析其可行性。结果表明:相较于SVM模型和RBF神经网络模型,RVM模型的3种预测指标值均是最小的。其中,平均绝对误差(mean absolute error,MAE)分别降低了86.02%和22.11%,均方根误差(root mean square error,RMSE)分别降低了72.05%和1.09%,相对均方误差(relative root mean square error,RRMSE)也分别降低了75.89%和21.13%,表明RVM是一种预测边坡稳定性的稳健工具,该方法能较为准确地预测出不同指标下的边坡安全系数。
In recent years,slope stability prediction is widely studied,and timely and accurate prediction can effectively prevent the occurrence of slope failure disasters.The fractal model of prediction for high cutting side slope deformation was proposed based on relevance vector machine(RVM).According to the instance of high cutting side slope project in Jing-Xin highway,the feasibility of the fractal model was analyzed by results of fitting and prediction which were drawn from comparing support vector machine(SVM)model,radical basis function(RBF)neural networks model and RVM combined model.The results show that comparing to SVM model and RBF neural networks model,the mean absolute error(MAE)of slope safety factor which was predicted by RVM model is reduced by 86.02%and 22.11%,respectively,the root mean square error(RMSE)is reduced by 72.05%and 1.09%,and the relative root mean square error(RRMSE)is reduced by 75.89%and 21.13%,respectively.It is concluded that RVM is a robust tool to predict the slope stability,and the method can accurately predict the slope safety factors under different indexes.
作者
孙吉书
夏健超
王建平
李伟华
SUN Ji-shu;XIA Jian-chao;WANG Jian-ping;LI Wei-hua(School of Civil Engineering and Transportation, Hebei University of Technology, Tianjin 300401, China;Jingxin Expressway Management Office of Zhangjiakou City, Zhangjiakou 075000, China)
出处
《科学技术与工程》
北大核心
2021年第28期12234-12242,共9页
Science Technology and Engineering
基金
河北省交通运输厅科学技术项目计划(T-2012131)
天津市交委科技计划(2021-24)。
