摘要
边坡稳定性的影响因素复杂多变,这对边坡稳定性评价和滑坡预测带来了巨大挑战。收集湖南省常吉高速、怀新高速和国内外文献中596组工程边坡实际数据,以坡高、坡角、容重、黏聚力、内摩擦角和孔隙压力比作为指标参数,建立了用于边坡稳定性评价和滑坡预测的样本数据库。通过引入改进CRITIC算法确定各指标参数在边坡稳定性评价中的权重,采用随机森林(Random Forest)算法补全样本数据中的缺失信息,分别利用基于Random Forest(简称RF)和XGBoost的集成算法,构建了基于改进CRITIC-RF算法和基于改进CRITIC-XGBoost算法的边坡稳定性评价和滑坡预测模型。将改进算法对未参与训练的120组边坡历史数据进行预测,并与常规的Random Forest、Adaboost、XGBoost和BP神经网络模型预测结果进行对比:结果以准确率为例,改进CRITIC-RF算法和改进CRITIC-XGBoost算法分别高达90%和89.2%,而常规的Random Forest、Adaboost、XGBoost和BP神经网络模型的准确率分别只有85.8%、83.3%、87.5%、83.3%。
The influencing factors of slope stability are complex and changeable,posing significant challenges to slope stability evaluation and landslide prediction.In this paper,596 sets of engineering slope data were collected at home and abroad.A database was established that incorporated index parameters such as slope height,slope angle,bulk density,cohesion,internal friction angle,and pore water ratio,which can be applied to evaluate slope stability.Prediction models for slope stability were established by the improved CRITIC algorithm,where the weight of each index parameter was determined,and then the missing information in the database was filled by Random Forest(RF)algorithm.Finally,the slope stability models were further proposed based on the improved CRITIC-RF and the improved CRITIC-XGBoost algorithms and used to calculate state values,respectively.The improved algorithm was applied to predict 120 sets of historical slope data that did not participate in training,and the prediction results were compared with conventional Random Forest,Adaboost,XGBoost,and BP neural network models:Taking the accuracy as an example,the improved CRITIC-RF algorithm and the improved CRITIC-XGBoost algorithm reached 90%and 89.2%respectively,while the accuracy of conventional Random Forest,Adaboost,XGBoost,and BP neural network models was only 85.8%,83.3%,87.5%,and 83.3%,respectively.
作者
胡惠华
张嘉睿
陈昌富
鲁光银
张根宝
林杭
陈鑫
HU Huihua;ZHANG Jiarui;CHEN Changfu;LU Guangyin;ZHANG Genbao;LIN Hang;CHEN Xin(Hunan Provincial Communications Planning,Survey&Design Institute Co.,Ltd.,Changsha,Hunan 410200,China;College of Civil Engineering,Hunan University,Changsha,Hunan 410082 China;School of Geosciences and Info-Physics,Central South University,Changsha,Hunan 410083 China;School of Resources and Safety Engineering,Central South University,Changsha,Hunan 410083,China)
出处
《公路工程》
2023年第6期74-83,共10页
Highway Engineering
基金
国家自然科学基金面上项目(52278349)
湖南省交通科技与进步计划项目(201003,202120)。
关键词
边坡评价
滑坡预测
机器学习
CRITIC
随机森森
slope evaluation
landslide prediction
machine learning
CRITIC
Random Forests(RF)
