基于IGA-AK代理模型的边坡安全系数预测与滑动面定位研究  

Research on slope safety factor prediction and sliding surface localization based on IGA-AK surrogate model

作　　者：龚礼岳 方如胜 胡献竹 诸葛启寅 陈斌 GONG Liyue;FANG Rusheng;HU Xianzhu;ZHUGE Qiyin;CHEN Bin(Wenzhou Mass Transit Railway Investment Group Co.,Ltd.,Wenzhou Zhejiang 325000,China;College of Civil Engineering and Architecture,Wenzhou University,Wenzhou Zhejiang 325000,China;Qingtian County Water Conservancy Investment and Development Co.,Ltd.,Lishui Zhejiang 323000,China)

机构地区：[1]温州市铁路与轨道交通投资集团有限公司,浙江温州325000 [2]温州大学建筑工程学院,浙江温州325000 [3]青田县水利发展投资有限公司,浙江丽水323000

出　　处：《中国安全生产科学技术》2024年第S1期169-178,共10页Journal of Safety Science and Technology

基　　金：国家自然科学基金项目(51808407);温州市基础性软科学研究项目(R2020013)

摘　　要：为提高边坡安全性分析的效率并解决传统方法计算成本高和忽略滑动面位置的问题,提出1种新的代理模型,该模型采用有限差分法-强度折减法(FDM-SRM)计算剪切应变增量,以快速评估边坡安全系数并准确定位滑动面。研究结果表明:基于免疫遗传算法结合主动学习Kriging(IGA-AK)的代理模型能够有效预测边坡最小安全系数,并提供滑动面位置的准确定量,可替代极限平衡法或数值模拟结果。研究结果可为边坡滑坡问题的可靠度分析和风险评估提供1种快速且准确的方法。To improve the efficiency of slope safety analysis and solve the problems of high calculation cost and neglect of sliding surface position in traditional methods,this paper proposes a new surrogate model.This model is based on the finite difference method⁃strength reduction method(FDM-SRM)to obtain shear strain increments,which enables rapid computation of slope safety factors and accurate sliding surface localization.The results show that the proposed surrogate model based on the combination of immune genetic algorithm and active learning Kriging(IGA-AK)can efficiently predict the minimum safety factor of the slope and locate the critical sliding surface,and the predicted results can be used as substitutes for limit equilibrium method or numerical simulation results.The research results provide a fast and accurate approach for advancing the reliability analysis and risk assessment of slope sliding problems.

关 键 词：边坡稳定 临界滑动面 安全系数 免疫遗传算法 主动学习Kriging 

分 类 号：P642.22[天文地球—工程地质学] TU433[天文地球—地质矿产勘探]