随着我国西部大开发战略的深入推进，黄土地区所涉及的边坡地质环境越发复杂，且工程实践中的黄土大多为非饱和状态。降雨是引发滑坡的关键因素，因此研究非饱和黄土边坡渗流特性具有理论和工程意义。为了对黄土边坡入渗机理深入了解，本文将光滑粒子法（SPH）引入渗流计算中，以陕西吴起天然黄土边坡为研究对象，设计室内土水特征曲线试验并结合非饱和渗透系数间接预测法，探讨黄土的渗透特性。基于SPH法推导质量守恒的渗流方程离散格式，编写Matlab串行计算程序求解典型非饱和渗流问题，验证SPH法的可行性。最后实现黄土边坡非饱和瞬态渗流数值求解，并与有限元模拟软件的渗流规律对比分析。开展工作如下：

（1）采用压力板仪对黄土土样进行土-水特征曲线试验，得到不同干密度下重塑非饱和黄土基质吸力与含水率的关系曲线，分析初始干密度对土样水力特性产生的影响。选用VG模型对试验数据处理拟合得到土-水特征曲线，为预测渗透系数提供数据支持。通过结合VG模型和Childs & Collis-Geroge模型，基于水渗透系数函数的计算技术，间接预测非饱和黄土在基质吸力变化时的渗透系数。通过研究非饱和渗透系数与基质吸力或含水率的相互影响，构建其数学模型，为渗流控制方程SPH离散化提供可靠依据。

（2）总结光滑粒子法中典型偏微分方程的积分近似和粒子近似并对其进行精度分析，使其满足非饱和渗流问题的求解。将非饱和渗透系数与体积含水率（基质吸力）的关系式引入渗流方程中，得到基于CCG模型的渗流控制方程的SPH离散格式。设置方程求解的边界和初始条件，依据串行流程图编写Matlab计算程序，通过一维竖向入渗算例得到基于SPH算法的渗流控制方程数值解，并与解析解对比验证其可行性。

（3）基于工程应用对边坡入渗SPH模型进行研究，利用串行计算程序得到不同深度处粒子点的体积含水量在降雨条件下的变化趋势，并分析其求解域内随时间变化的分布特性。研究结果表明：土体体积含水率响应速度随雨强增大而增大，其变化幅度也随雨强而增大。在相同降雨强度条件下，随着时间的推移和降雨入渗深度的增加入渗速率越来越缓慢，利用SPH法计算所得含水率变化曲线均表现为斜率首先变大随后逐渐趋于平缓。

（4）采用Geostudio软件的SEEP/W模块，建立了与SPH法相同的非饱和边坡模型，基于渗流场计算结果研究不同降雨强度下黄土边坡体积含水率分布情况。通过分析不同位置的特征点，得到其入渗速率和深度相关信息。研究结果表明：在不同位置处，入渗深度存在显著差异，其中坡脚的入渗深度最大，接着是坡顶，最小为坡中。连续降雨条件下，靠近坡脚处出现暂态饱和，并逐渐沿着坡面向上扩散。两种方法计算结果对比表明坡中、坡顶浅层特征点含水率整个模拟降雨期间SPH计算结果与模拟结果拟合较好；深层特征点体积含水率SPH计算结果在初期增长速度相较于数值模拟较快，降雨后期两者计算值误差较小。坡脚特征点由于坡面径流的影响，降雨前期SPH计算结果的增长速率小于模拟结果。各特征点SPH计算结果和模拟结果均呈现先增大随后逐渐变缓。

With the deepening of China's western development strategy, the geological environment of slope in loess area is more and more complex, and most of the loess in engineering practice is unsaturated. Rainfall is the key factor causing landslide, so it is of theoretical and engineering significance to study the seepage characteristics of unsaturated loess slope. In order to deeply understand the infiltration mechanism of loess slope, this paper introduces the smoothed particle method (SPH) into the seepage calculation. Taking the natural loess slope of Wuqi in Shaanxi Province as the research object, the indoor soil water characteristic curve test is designed and combined with the indirect prediction method of unsaturated permeability coefficient to explore the permeability characteristics of loess. Based on the SPH method, the discrete format of the seepage equation of mass conservation is derived, and the Matlab serial calculation program is written to solve the typical unsaturated seepage problem to verify the feasibility of the SPH method. Finally, the numerical solution of unsaturated transient seepage of loess slope is realized and compared with the seepage law of finite element simulation software. The work is as follows :

(1) The soil-water characteristic curve test of loess soil samples was carried out by pressure plate instrument, and the relationship curve between matrix suction and water content of remolded unsaturated loess under different dry densities was obtained. The VG model is used to fit the test data to obtain the soil-water characteristic curve, which provides data support for predicting the permeability coefficient. By combining VG model and Childs & Collis-Geroge model, based on the calculation technology of water permeability coefficient function, the permeability coefficient of unsaturated loess is indirectly predicted when the matric suction changes. By studying the interaction between unsaturated permeability coefficient and matrix suction or water content, a mathematical model is constructed to provide a reliable basis for SPH discretization of seepage control equation.

(2) The integral approximation and particle approximation of typical partial differential equations in smooth particle method are summarized and their accuracy is analyzed to satisfy the solution of unsaturated seepage problems.The relationship between unsaturated permeability coefficient and volumetric water content ( matric suction ) is introduced into the seepage equation, and the SPH discrete format of the seepage control equation based on the CCG model is obtained. The boundary and initial conditions of the equation solution are set, and the Matlab calculation program is written according to the serial flow chart. The numerical solution of the seepage control equation based on SPH algorithm is obtained by one-dimensional vertical infiltration example, and its feasibility is verified by comparison with the analytical solution.

(3) Based on the engineering application, the SPH model of slope infiltration is studied. The change trend of volumetric water content of particle points at different depths under rainfall conditions is obtained by serial calculation program, and the distribution characteristics of the solution domain with time are analyzed. The results show that the response speed of soil volumetric water content increases with the increase of rainfall intensity, and the change range also increases with the increase of rainfall intensity. Under the same rainfall intensity conditions, with the passage of time and the increase of rainfall infiltration depth, the infiltration rate becomes slower and slower. The water content curve calculated by SPH method shows that the slope first increases and then gradually tends to be gentle.

(4) The SEEP/W module of Geostudio software was used to establish the same unsaturated slope model as SPH method. Based on the calculation results of seepage field, the distribution of volumetric water content of loess slope under different rainfall intensities was studied. By analyzing the feature points at different locations, the infiltration rate and depth-related information are obtained. The results show that there are significant differences in infiltration depth at different locations. The infiltration depth at the foot of the slope is the largest, followed by the top of the slope, and the smallest is the middle of the slope. Under continuous rainfall conditions, transient saturation occurs near the toe of the slope and gradually spreads upward along the slope. The comparison of the calculation results of the two methods shows that the SPH calculation results and the simulation results are well fitted during the whole simulated rainfall period. The SPH calculation results of the volumetric water content of the deep feature points grow faster in the initial stage than the numerical simulation, and the error between the two calculation values is small in the later stage of rainfall. Due to the influence of slope runoff, the growth rate of SPH calculation results in the early stage of rainfall is less than that of simulation results. The SPH calculation results and simulation results of each feature point increase first and then gradually slow down.