The spatial variability of rockfill properties is the key factor affecting stochastic seismic responses. Herein, an efficient probabilistic analysis method for stochastic seismic analysis named the generalized probability density evolution method (GPDEM) is proposed and combined with the random field generated by K-L decomposition to evaluate the stochastic seismic responses of concrete-faced rockfill dams (CFRDs) during earthquakes. The vertical plastic deformation is selected as a significant index to quantify the effect of rockfill spatial variability. First, a series of 2D random fields and a finite element model of the CFRD profile are established. Each realization of random fields is assigned to each grid depending on the coordinates. Then, four working conditions (different constitutive model parameters, cross-correlation coefficients, concrete-faced slab contact positions and earthquake intensities) are investigated and compared in CFRD dynamic calculations. Finally, the GPDEM-based vertical plastic deformation results are obtained, and probabilistic analyses are carried out to evaluate the influence of rockfill property uncertainty on the stochastic seismic responses. Multiple figures of the probability density function (PDF) and cumulative distribution function (CDF) of dynamic calculations are shown as examples to analyze the marked effect of rockfill spatial variability. The results demonstrate that the control parameter ($M_f$) is the most influential parameter of stochastic seismic responses in the GPDEM. The cross-correlation coefficient increases the probability of large vertical plastic deformation compared to zero. Moreover, the nonfaced-slab side is more sensitive to the effect of rockfill uncertainties. Increasing the seismic motion intensity causes a larger probability beyond the limit.

堆石料特性的空间变异性是影响随机地震响应的关键因素。在此，提出了一种有效的随机地震分析概率分析方法，称为广义概率密度演化法（GPDEM），并结合KL分解产生的随机场来评估面板堆石坝（CFRD）在地震过程中的随机地震响应. 选择垂直塑性变形作为量化堆石空间变异性影响的重要指标。首先，建立了一系列二维随机场和CFRD剖面的有限元模型。根据坐标将随机场的每个实现分配给每个网格。然后，在CFRD动力计算中对四种工况（不同的本构模型参数、互相关系数、混凝土面板接触位置和地震强度）进行了研究和比较。最后，获得了基于 GPDEM 的竖向塑性变形结果，并进行了概率分析，以评估堆石性质的不确定性对随机地震响应的影响。以动态计算的概率密度函数（PDF）和累积分布函数（CDF）的多幅图为例，分析堆石料空间变异的显着影响。结果表明，控制参数（${米}_F$) 是 GPDEM 中随机地震响应的最有影响的参数。与零相比，互相关系数增加了大的垂直塑性变形的概率。此外，非面板侧对堆石料不确定性的影响更为敏感。增加地震运动强度会导致更大的概率超出极限。