Slope failure is a major challenge in engineering construction and geological hazard prevention in mountainous regions of China. When the spatial variability of geotechnical parameters is considered, a fixed geometric search pattern was used in the traditional limit equilibrium method. Potential slip surfaces cannot be accurately captured, and the computational efficiency is low. These shortcomings cannot meet the demand for rapid and accurate stability assessment in large-scale slope projects. An intelligent analysis method was proposed by integrating deep learning with rigorous limit equilibrium theory. Convolutional neural networks (CNN) and long short-term memory networks (LSTM) were coupled, and an attention mechanism was introduced. This method was used to learn the spatial distribution characteristics of randomly varying strength parameters and to predict slip surfaces and safety factors. The results show that the CNN-LSTM-Attention model achieves a precision of 0.996 8 in predicting slip surfaces and an R2 of 0.972 5 in predicting safety factors. The predictions of the deep learning model are highly consistent with those obtained by the limit equilibrium method. It is concluded that the proposed model provides intelligent support for slope stability evaluation in engineering practice.
随机场框架下数值方法得到了广泛应用,随机场有限元法(random field finite element method, RF-FEM)为考虑岩土参数空间变异性的边坡可靠度分析提供了有效工具,但其计算成本高昂的缺点严重制约了其工程适用性。为此,Wang等[6]提出采用卷积神经网络(convolutional neural networks, CNN)作为RF-FEM的代理模型,通过少量随机场样本训练即可高精度预测蒙特卡罗模拟结果(误差<5%),显著提升了多层土体边坡可靠度分析效率。Cao等[7]通过构建Monte Carlo模拟的替代模型,可在保证边坡安全系数预测精度的同时,显著降低了传统随机场无网格法分析的计算成本,为复杂地质条件下的高效稳定性评估开辟了新途径。相较于RF-FEM,现有主流极限平衡软件(如Slide2D、Slope)通过耦合随机场理论与圆弧滑裂面搜索算法,随机场极限平衡法(random field limit equilibrium method, RF-LEM)在工程实践中同样得到广泛应用。RF-LEM显著降低了考虑参数空间变异性的可靠度分析计算成本。然而,这种基于固定几何模式的搜索方法存在明显局限,对于非均质边坡,预设的圆弧滑裂面难以准确捕捉受岩土参数空间变异控制的真实破坏路径,导致失效概率评估存在系统性偏差。
XuQuan, HouJing, YangJian, et al. Fine analysis of rock slope stability based on synthetic rock mass technology[J]. Rock and Soil Mechanics, 2025, 46(7): 1-10.
LiXianwei, GaoXuguang, WangJipeng, et al. Experimental study on rainfall-induced slope instability model based on 3D laser scanning technology[J]. Science Technology and Engineering, 2025, 25(19): 7995-8002.
ZhuDayong, QianQihu. Critical slip fields of slopes satisfied all conditions of limit equilibrium for slices[J]. China Civil Engineering Journal, 2000(5): 68-74.
[9]
ZhengH. A three-dimensional rigorous method for stability analysis of landslides[J]. Engineering Geology, 2012, 145: 30-40.
[10]
WangZ Z, GohS H. Novel approach to efficient slope reliability analysis in spatially variable soils[J]. Engineering Geology, 2021, 281: 105989.
[11]
CaoX, LinS, DongM, et al. A surrogate model for the rapid prediction of factor of safety in slopes with spatial variability[J]. Mathematics, 2025, 13: 1604.
LiBin, FanJiawei. Deep learning-oriented literature review and experimental analysis of ship target detection[J]. Science Technology and Engineering, 2025, 25(27): 11427-11450.
[14]
ChenJ, CheA, WangL, et al. Dynamic response prediction for slope during earthquake using modal analysis and deep learning methods[J]. Computers and Geotechnics, 2024, 174: 106618.
[15]
HuangM, NishimuraS I, ShibataT, et al. End-to-end time-dependent probabilistic assessment of landslide hazards using hybrid deep learning simulator[J]. Computers and Geotechnics, 2025, 178: 106920.
[16]
ZhengH, YangZ L, SunG H. Extremum solutions to the limit equilibrium method subjected to physical admissibility[J]. Natural Hazards, 2013, 65(1): 79-96.
ZhengHong, CaoXitailang, ZhangZhihong. Study on the distribution of normal stress on the slip surface in the global analysis of slope stability[J]. Journal of Beijing University of Technology, 2024, 50(11): 1312-1325.
SunCong, LiChunguang, ZhengHong. Searching critical slip surfaces of slope based on global stability analysis[J]. Rock and Soil Mechanics, 2013, 34(9): 2583-2588.
[21]
ZhengH, ChenQ. Dimension extending technique for constitutive integration of plasticity with hardening-softening behaviors[J]. Computer Methods in Applied Mechanics and Engineering, 2022, 394: 114833.
YeJianbin, LiChao, XiaoQiong, et al. Optimization of urban rail train operation plan based on convolutional neural network[J]. Science Technology and Engineering, 2025, 25(27): 11793-11799.
[27]
DaoF, ZengY, QianJ. Fault diagnosis of hydro-turbine via the incorporation of Bayesian algorithm optimized CNN-LSTM neural network[J]. Energy, 2024, 290: 130326.
[28]
LiW, LiY, GargA, et al. Enhancing real-time degradation prediction of lithium-ion battery: a digital twin framework with CNN-LSTM-Attention model[J]. Energy, 2024, 286: 129681.
[29]
LiuS H, WuY D, HuangR. Prediction of drilling plug operation parameters based on incremental learning and CNN-LSTM[J]. Geomechanics and Geotechnics for Geo-Energy and Geo-Engineering, 2024, 234: 212631.
[30]
ChenJ, ZhengL, CheW, et al. A method for measuring carbon emissions from power plants using a CNN-LSTM-Attention model with Bayesian optimization[J]. Case Studies in Thermal Engineering, 2024, 63: 105334.