【摘要】 为推进湿陷性黄土地区基坑工程中可回收锚杆的工程应用，以土体含水率（w）、土体干密度（ρ_d）和锚固体直径（d）作为试验影响因素，开展了湿陷性黄土的湿陷试验以及可回收锚杆微段单元体拉拔试验，获得了黄土湿陷系数（δ_s）和锚-土界面剪应力-剪切位移曲线随试验影响因素（w、ρ_d和d）的变化规律。试验结果表明：湿陷系数随干密度的增大呈减小趋势，随含水率的增大呈减小趋势；锚-土界面峰值剪切强度则随干密度的增大呈增大趋势，随含水率的增大呈减小趋势；单元体试样边界效应与土样软硬程度有关。最后，基于随机森林算法，选取土体含水率、土体干密度、湿陷系数和锚固体直径4个影响因素作为输入变量，建立了湿陷性黄土中可回收锚杆锚-土界面峰值剪切强度预测模型，其训练集和测试集的可决系数（R~2）分别为0.971、0.915，表明所建立预测模型效果较好。更多还原

【Abstract】 Collapsibility tests of loess and the pullout test of element-scale sample of retrievable anchors were carried out to promote the engineering application of recoverable anchors in the foundation pit project in the area of collapsibility loess. Specifically, the collapsibility loess was utilized as soil material in the pullout tests where the water content(w), the dry density of soil （ρ_d）, and the diameter of anchors(d) were considered as the influential factors. The variation of the collapsibility coefficient of loess(δ_s) and shear stress-shear displacement curves at the anchors-soil interface with different factors(w, ρ_d, and d) were obtained. The results show that the collapsibility coefficient decreases with the increase of ρd as well as w. The peak shear strength of the anchor-soil interface increases with the rising of ρd, and these values decreases with the elevation of w. Besides, the boundary effect of the elementscale specimen is proved to be related with the plasticity of the soil sample. Finally, a model based on the random forest algorithm is established with w, ρ_d, δ_s and d as input variables for the prediction of the peak shear strength of the anchor-soil interface of retrievable anchor rods in collapsibility loess. The determination coefficients(R~2) of the training dataset and the test dataset are 0. 971 and 0. 915, indicating a good performance of the established RF model in anchor-soil interfacial strength forecast.更多还原