基于多尺度分割方法的斜坡单元划分及滑坡易发性预测

滑坡易发性预测可以有效预测潜在滑坡的空间位置,是滑坡危险性和风险性评价的基础。由于斜坡单元依据真实地形地貌划分和具有明确的地质特征意义,更多的学者尝试利用斜坡单元进行区域滑坡易发性预测。但是,如何高效准确地划分斜坡单元并考虑其内部环境因子的非均质性是制约斜坡单元应用的关键因素,也是目前研究中的难点。本文以江西省崇义县为例,首先,提取研究区域坡向和山体阴影图作为基础数据,采用多尺度分割(MSS)方法划分斜坡单元,并结合试错法和研究区域历史滑坡形态特征确定MSS方法的最优参数组合。然后,基于斜坡单元提取高程、坡度、剖面曲率等环境因子,分别导入支持向量机(SVM)和逻辑回归(LR)模型,构建Slope–SVM/LR易发性预测模型。通过变化值和标准差表征斜坡单元内部环境因子的非均质性,进而构建Variant Slope–SVM/LR易发性预测模型。最后,采用ROC曲线和频率比精度分析上述模型的预测精度。结果表明:1)当尺度、形状特征权重和紧致度权重参数分别取20、0.8和0.8时,研究区域斜坡单元的划分效果最好;2)Slope–SVM、Variant slope–SVM、Slope–LR和Variant slope–LR模型的ROC精度分别为0.812、0.876、0.818和0.839,相应的频率比精度分别为0.780、0.866、0.792和0.865,说明Variant slope–SVM/LR模型的预测精度高于Slope–SVM/LR模型。因此,MSS方法可以实现高效准确地自动划分斜坡单元,考虑斜坡单元内部环境因子的非均质性可以提高易发性预测结果的准确性。

铜尾砂和钨尾砂的抗剪强度及宏细观分析

铜矿和钨矿在经过开采和选矿过程后产生大量的尾砂,尾砂的抗剪强度直接影响到尾砂筑坝的稳定安全性和材料的创新利用。尾矿库现场取回尾砂的级配曲线不同于选矿产生的全尾砂,选用合适的尾砂抗剪强度参数以及对不同尾砂抗剪强度的比较和分析有助于评估尾矿库的溃坝安全性。为了深入研究铜、钨两种尾砂的抗剪强度及异同点,针对不同粒径范围的铜、钨尾砂进行了直剪试验、X射线衍射分析和扫描电镜分析。直剪试验结果表明,干燥尾砂表现出无黏性土的强度性质,相同粒组尺寸的铜尾砂抗剪强度要高于钨尾砂的抗剪强度;铜尾砂和钨尾砂在含水率为15%情况下的抗剪强度表现出一定的黏聚力,使其抗剪强度相对于干燥状态下提高了16.7%~83.8%;随着含水率从10%增加到20%,细粒钨尾砂的内摩擦角变化不大,含水量的增加使得钨尾砂的黏聚力较明显地减小并使得抗剪强度降低。X射线衍射分析结果显示铜尾砂和钨尾砂都以石英为主晶相,其余矿物成分大都是以硅酸盐为主的次生矿物;5000倍扫描电镜照片显示细粒尾砂颗粒的表面附着一定量的黏土矿物,黏土矿物的存在导致了尾砂抗剪强度随含水率增加而发生变化。

考虑裂隙迹长和开度相关性的随机裂隙网络数值模拟及渗流分析

基于蒙特卡罗法,引入裂隙迹线长度和开度的相关性函数,提出了一个改进的随机裂隙网络生成算法。利用MATLAB平台编制程序,实现二维裂隙网络的生成,通过与布尔模拟生成的真实裂隙场的比较,验证了改进算法的有效性和程序的正确性。然后借助离散裂隙网络渗流模型对生成的裂隙网络进行渗流分析。分别在考虑迹长开度相关性和不考虑二者相关性的条件下,计算了随机裂隙网络的渗流场,给出了下游边界出口裂隙流量分布和裂隙网络内部水压力的分布。计算结果表明,考虑裂隙迹线长度和开度的相关性能很好地反映裂隙网络渗流的结构特征,从而证明了所提方法的合理性。该模型是对随机裂隙网络模型的有效补充,可为岩体工程渗控分析提供理论基础和数值计算工具。

基于双重注意力机制的滑坡识别方法优化

随着计算机视觉技术的发展,通过卫星图像深度学习进行滑坡识别的研究正在逐步展开。通过引入双重注意力机制,提出了一种基于卷积神经网络的滑坡图像识别优化算法。基于统计的2200张滑坡图像数据集,探讨了10种网络结构及4种注意力机制对滑坡识别结果的影响,并通过比例为4∶1的训练集和测试集进行滑坡识别,验证了本文方法的有效性。结果表明:ResNet结构相较于其他网络结构表现更为优秀,就该算例而言,ResNet-101结构具有最高的召回率、精确率和F 1度量。融入了双重注意力机制的卷积神经网络相较于单个神经网络而言,滑坡识别的精确率更大,且滑坡边界分割结果更接近于真实的滑坡边界,其中,ResNet-101+DAN模型为最优模型。相较之下,单个神经网络无法克服图像噪声的影响,图像分割结果不佳。

基于耦合信息量法选择负样本的区域滑坡易发性预测

在利用机器学习(machine learning,ML)模型预测滑坡易发性时,选择合理的负样本对预测结果具有重要影响。现有研究大多从整个研究区或低坡度等特定属性区内随机选择负样本,往往不够准确或以偏概全,降低了易发性制图的可靠性。为解决这一问题,提出基于耦合信息量法(information value,IV)的ML模型开展易发性建模。以江西省瑞金市为例,采用IV法将环境因子的属性值转化为对滑坡贡献的信息量值,划定极低和低易发区,并从中随机选择出ML模型训练验证用的负样本数据,构建全新的信息量–支持向量机(IV–SVM)、信息量–随机森林(IV–RF)耦合模型,并预测瑞金滑坡易发性。进一步地,与从全区随机选择负样本的单独SVM和RF模型,以及从坡度小于2°的特定属性区内随机选择负样本的低坡度SVM和RF模型开展对比研究。最后,采用Kappa系数和ROC曲线等指标验证和比较建模结果,IV–SVM和IV–RF模型的Kappa系数分别为0.828和0.876,且对应的ROC曲线的AUC值分别为0.920和0.988,均高于单独SVM、RF和低坡度SVM、RF模型;同时,IV–SVM和IV–RF模型易发性概率分布的平均值较小而标准差较大。结果表明:1)IV–SVM和IV–RF模型具有比单独SVM和RF模型,以及低坡度SVM和RF模型更高的滑坡易发性预测精度,且更有效地反映了瑞金滑坡易发性分布规律;2)RF模型相较于SVM模型具有更高的预测精度;3)IV–RF等耦合模型能够解决单独模型存在的负样本采样不准确和低坡度模型存在的坡度因子对模型贡献度有误差的问题,其滑坡易发性预测精度更高,更加合适区域滑坡易发性预测建模。本文研究为机器学习预测滑坡易发性的负样本采样方法提供了新思路。

Verhulst反函数预测模型的改进及滑坡时间概率预测

滑坡时间预测是滑坡灾害防治的重要组成部分,但由于滑坡演化存在不确定性,准确预测滑坡的发生极为困难.Verhulst反函数模型是一种常用的滑坡时间预测模型,但其存在计算起始时刻选择不当会造成监测数据拟合质量差和预测精度低的问题.针对这一不足,提出了一种改进的Verhulst反函数模型(MVIF模型),并进行近实时概率预测分析.结果表明:(1)MVIF模型改善了原模型对计算起始时刻选择苛刻的问题;(2)MVIF模型预测精度较高,在滑坡进入中等加速变形阶段之后可进行可靠的预测;(3)预测滑坡时间与破坏概率结合提供了一种新的滑坡预报准则.该研究可为蠕滑型滑坡的预警预报提供有价值的参考.

考虑水力模型参数空间变异性土石坝边坡可靠度分析

土石坝边坡可靠度分析中大多只考虑了坝体材料抗剪强度参数的空间变异性,忽略了非饱和水力模型参数空间变异性的影响.本文发展了土石坝边坡可靠度非侵入式随机分析方法,探讨了水力模型参数空间变异性对土石坝边坡可靠度的影响规律,解释了土石坝边坡失效概率随水力模型参数n的变异系数增大而减小的原因.结果表明:非侵入式随机分析方法计算效率远远高于传统的蒙特卡洛模拟方法.参数敏感性分析表明土石坝边坡失效概率与坝体材料黏聚力、内摩擦角及饱和渗透系数的变异系数之间呈正相关关系,其中黏聚力变异性的影响最大,饱和渗透系数变异性的影响最小.此外,a的变异性对边坡失效概率基本没有影响,相反边坡失效概率随着n的变异系数增大而越小,这是因为当n的变异系数取值越大模拟得到的n值越接近于下限值(1.05),进而导致土石坝渗透系数越小,通过坝体的渗流量越小,从而土石坝边坡更安全.

不同滑坡边界及其空间形状对滑坡易发性预测不确定性的影响规律

滑坡边界及其空间形状通常呈现为半圆形、簸箕形等不规则多边形面,但部分滑坡易发性建模中采用理想化的滑坡点或缓冲圆作为滑坡边界,给易发性建模带来大量不确定性。为研究不同滑坡边界对建模不确定性的影响规律,以江西上犹县337个滑坡及10个环境因子为数据源,分别基于滑坡点(Point)、缓冲圆(Circle)和准确解译并绘制的多边形面(Polygon)建立滑坡边界及其与环境因子的频率比关联性;之后选取多层感知器(MLP)和随机森林(RF)来构建Point,Circle,Polygon-based MLP和RF等模型,最后采用ROC精度、易发性指数的分布规律及其差异性等开展建模不确定性分析。结果表明:(1)以滑坡点或缓冲圆作为滑坡边界将增加建模不确定性,而用准确的滑坡多边形边界显然能更有效保证建模精度和可靠性:(2)基于MLP和RF的滑坡易发性建模不确定性规律致,但RF模型预测精度高于MLP且其不确定性低于MLP:(3)以点和缓冲圆为滑坡边界得到的滑坡易发性结果在整体上也能反映研究区滑坡概率空间分布规律,在缺乏准确滑坡边界时可作为一种替补方案。

不同空间分辨率和训练测试集比例下的滑坡易发性预测不确定性

探讨空间分辨率和模型训练测试集比例等因素对滑坡易发性预测不确定性的影响具有重要意义。拟以江西省宁都县的滑坡为例,首先计算各环境因子在不同空间分辨率(15,30,60,90和120m)下的频率比值;然后将滑坡和非滑坡样本按照9/1,8/2,7/3,6/4和5/5的比例划分出不同的模型训练和测试数据集,由此得到25种组合工况下的模型输入和输出变量;再将其导入支持向量机(support vector machine,SVM)和随机森林(random forest,RF)模型中开展滑坡易发性预测,最后利用预测精度和易发性指数特征来分析各工况下的易发性建模不确定性。结果表明:(1)15m分辨率、9∶1训练测试集比例和RF模型工况下的易发性预测精度最高,各工况下重要的环境因子分别为高程、坡度和地形起伏度等;(2)随着分辨率的下降或训练测试集比例的减小,SVM和RF模型的预测精度均逐渐下降,易发性指数的均值变大且标准差减小;(3)所有工况下随着分辨率和训练测试集比例的减小,易发性预测精度均逐渐下降且其不确定性增大;(4)各工况下RF预测精度优于SVM,分辨率对RF模型的影响显著大于训练测试集比例,而在SVM中2种因素的影响相差不大。

Modeling of coupled deformation,water flow and gas transport in soil slopes subjected to rain infiltration

Rain infiltration into a soil slope leads to propagation of the wetting front, transport of air in pores and deformation of the soils, in which coupled processes among the solid, liquid and gas phases are typically involved. Most previous studies on the unsaturated flow and its influence on slope stability were based on the singlephase water flow model (i.e., the Richards Equation) or the waterair two-phase flow model. The effects of gas transport and soil deformation on the movement of groundwater and the evolution of slope stability were less examined with a coupled solid-water-air model. In this paper, a numerical model was established based on the principles of the continuum mechanics and the averaging approach of the mixture theory and implemented in an FEM code for analysis of the coupled deformation, water flow and gas transport in porous media. The proposed model and the computer code were validated by the Liakopoulos drainage test over a sand column, and the significant effect of the lateral air boundary condition on the draining process of water was discussed. On this basis, the coupled processes of groundwater flow, gas transport and soil deformation in a homogeneous soil slope under a long heavy rainfall were simulated with the proposed three-phase model, and the numerical results revealed the remarkable delaying effects of gas transport and soil deformation on the propagation of the wetting front and the evolution of the slope stability. The results may provide a helpful reference for hazard assessment and control of rainfall-induced landslides.

Uncertainty analysis of correlated non-normal geotechnical parameters using Gaussian copula

Determining the joint probability distribution of correlated non-normal geotechnical parameters based on incomplete statistical data is a challenging problem.This paper proposes a Gaussian copula-based method for modelling the joint probability distribution of bivariate uncertain data.First,the concepts of Pearson and Kendall correlation coefficients are presented,and the copula theory is briefly introduced.Thereafter,a Pearson method and a Kendall method are developed to determine the copula parameter underlying Gaussian copula.Second,these two methods are compared in computational efficiency,applicability,and capability of fitting data.Finally,four load-test datasets of load-displacement curves of piles are used to illustrate the proposed method.The results indicate that the proposed Gaussian copula-based method can not only characterize the correlation between geotechnical parameters,but also construct the joint probability distribution function of correlated non-normal geotechnical parameters in a more general way.It can serve as a general tool to construct the joint probability distribution of correlated geotechnical parameters based on incomplete data.The Gaussian copula using the Kendall method is superior to that using the Pearson method,which should be recommended for modelling and simulating the joint probability distribution of correlated geotechnical parameters.There exists a strong negative correlation between the two parameters underlying load-displacement curves.Neglecting such correlation will not capture the scatter in the measured load-displacement curves.These results substantially extend the application of the copula theory to multivariate simulation in geotechnical engineering.

A new variational inequality formulation for unconfined seepage flow through fracture networks

Darcy's law only applying to the flow domain is extended to the entire fracture network domain including the dry domain.The partial differential equation(PDE) formulation for unconfined seepage flow problems for discrete fracture network is established,in which a boundary condition of Signorini's type is prescribed over the potential seepage surfaces.In order to reduce the difficulty in selecting trial functions,a new variational inequality formulation is presented and mathematically proved to be equivalent to the PDE formulation.The numerical procedure based on the VI formulation is proposed and the corresponding algorithm has been developed.Since a continuous penalized Heaviside function is introduced to replace a jump function in finite element analysis,oscillation of numerical integration for facture elements cut by the free surface is eliminated and stability of numerical solution is assured.The numerical results from two typical examples demonstrate,on the one hand the effectiveness and robustness of the proposed method,and on the other hand the capability of predicting main seepage pathways in fractured rocks and flow rates out of the drainage system,which is very important for performance assessments and design optimization of complex drainage system.

The variational inequality formulation for unconfined seepage through three-dimensional dense fracture networks

The space block search technology is used to determine a connected three-dimensional fracture network in polygonal shapes,i.e.,seepage paths.After triangulation on these polygons,a finite element mesh for 3D fracture network seepage is obtained.Through introduction of the generalized Darcy's law,conservative equations for both fracture surface and fracture interactions are established.Combined with the boundary condition of Signorini's type,a partial differential equation(PDE) formulation is presented for the whole domain concerned.To solve this problem efficiently,an equivalent variational inequality(VI) formulation is given.With the penalized Heaviside function,a finite element procedure for unconfined seepage problem in 3D fracture network is developed.Through an example in a homogeneous rectangular dam,validity of the algorithm is verified.The analysis of an unconfined seepage problem in a complex fracture network shows that the proposed algorithm is very applicable to complex three-dimensional problems,and is effective in describing some interesting phenomenon usually encountered in practice,such as "preferential flow".

A relative permeability model for deformable soils and its impact on coupled unsaturated flow and elasto-plastic deformation processes.

Relative permeability is an indispensable property for characterizing the unsaturated flow and induced deformation in soils. The widely used Mualem model is inadequate for deformable soils because of its assumption of a rigid pore structure and the resultant unique dependence of the tortuosity factor on the volumetric water content. In this study, a unified relationship between the relative permeability and the effective degree of saturation was proposed for deformable soils by incorporating our newly developed water retention curve model into the original Mualem model, in which a new tortuosity factor was defined using the fractal dimension of flow paths and the mean radius of water-filled pores for representing the effect of pore structure variation. The modified deformation-dependent relative permeability model was verified using test data on five types of soils; the verification revealed a much better performance of the proposed model than the original model, which commonly overestimates the relative permeability of deformable soils. Finally, the proposed model was implemented in a coupled numerical model for examining the unsaturated flow and elastoplastic deformation processes in a soil slope induced by rain infiltration. The numerical results showed that the deformation-dependent nature of relative permeability has a remarkable effect on the elastoplastic deformation in the slope and that neglect of the deformation-dependent behavior of relative permeability causes overestimation of the depth of failure.

A new constitutive law for the nonlinear normal deformation of rock joints under normal load

In view of the deviation of the fitting results of the classical exponential model and the hyperbolic model （the BB model） from several experiment data during intermediate stress period, a new constitutive model for the nonlinear normal deformation of rock joints under normal monotonous load is established with flexibility-deformation method. First of all, basic laws of the deformation of joints under normal monotonous load are discussed, based on which three basic conditions which the complete constitutive equation for rock joints under normal load should meet are put forward. The analysis of the modified normal con- stitutive model on stress-deformation curve shows that the general exponential model and the improved hyperbolic model are not complete in math theory. Flexibility-deformation monotone decreasing curve lying between flexibility-deformation curve of the classical exponential model and the BB model is chosen, which meets basic conditions of normal deformation mentioned before, then a new normal deformation constitutive model of rock joints containing three parameters is established. Two main forms of flexibility-deformation curve are analyzed and specific math formulas of the two forms are deduced. Then the range of the parameters in the g-δ model and the g-2 model and the correlative influence factor in geology are preliminarily discussed. Referring to different experiment data, the validating analysis of the g-δ model and the g-γ model shows that the g-2 model can be applied to both the mated joints and unmated joints. Besides, experiment data can be better fit with the g-2 model with respect to the BB model, the classical exponential model and the logarithm model.