摘要
为研究高填方边坡雨季长期稳定性预测问题,以仁遵高速填方边坡为例,基于有限元计算和梯度回归树模型进行稳定系数长期拟合预测。研究结果表明:边坡稳定性变化在一年内分为平稳期、雨季波动期和缓冲期,其中雨季波动期稳定性最差,黏性土较砂性土边坡稳定系数劣化更为迅速,恢复更慢;基于梯度回归树模型,结合外界环境因素和内部土体材料对稳定系数进行预测,与实际值对比得到R^(2)为0.927,MSE为0.021,MAPE为3.877,拟合程度较高,且在雨季波动期的预测优势尤为突出;通过与线性回归模型、回归树、随机森林和XGB模型进行对比,得到该模型具有高拟合性、灵活性、强鲁棒性以及泛化能力强的优点,能够在高填方边坡稳定系数预测中广泛使用。
In order to study the prediction model of the stability coefficient of high fill slope,taking Renzun high speed fill slope as an example,the stability prediction research was carried out based on the finite element model and gradient lifting tree model.The results show that the slope stability changes in a year can be divided into stationary period,fluctuating period of rainy season and buffer period.Among them,the fluctuating period of rainy season has the worst stability.The stability coefficient of cohesive soil slope deteriorates more rapidly and the recovery is slower than that of sandy soil slope.Based on the gradient regression tree model and combining external environmental factors and internal soil materials,the stability coefficient was predicted,and the determination coefficient R2 was 0.927,root-mean-square error MSE was 0.021,and the absolute square-root error MAPE was 3.877.The predicted value was in good agreement with the actual value,and the prediction advantage was especially outstanding during the rainy season fluctuation period.By comparing with linear regression model,regression tree,random forest and XGB model,it is concluded that the model has the advantages of high fitting,flexibility,strong robustness and strong generalization ability,which provides a new idea and exploration for the prediction of stability coefficient of high fill slope.
作者
赵建祥
毕鹏飞
惠亚强
ZHAO Jianxiang;BI Pengfei;HUI Yaqiang(China Communications Road and Bridge Construction Co.,Ltd.,Beijing 101121,China)
出处
《水利水电技术（中英文）》
北大核心
2021年第S02期406-411,共6页
Water Resources and Hydropower Engineering
基金
重庆市自然科学基金面上项目(cstc2020jcyj-msxmX0218)
关键词
梯度回归树
高填方边坡
稳定系数预测
gradient regression tree
high fill slope
stability coefficient prediction
作者简介
赵建祥(1983-),男,高级工程师,学士,从事公路施工技术研究工作。E-mail:20999353@qq.com
