基于DNN-NSGA-Ⅱ的高填方加筋边坡参数优化研究OA
Parameter Optimization of High-fill Reinforced Slope Based on DNN-NSGA-Ⅱ
以福建某典型高填方加筋边坡为研究对象,提出一种集成深度神经网络(DNN)与非支配排序遗传算法(NSGA-Ⅱ)的智能化优化设计方法,用于实现高填方加筋边坡支护设计的多目标协同优化.首先,通过有限元模拟生成样本数据,构建以关键设计参数为输入、稳定性响应指标为输出的 DNN 代理模型;随后,将该代理模型嵌入NSGA-Ⅱ框架,实现以最小化水平位移、加筋材料用量与最大化安全系数为目标的多目标寻优.通过对 Pareto 前沿解集的分析与典型方案提取,验证所提方法在兼顾边坡安全性与经济性方面的有效性,可为高填方边坡优化设计提供理论支撑与工程参考.
Taking a typical high-fill reinforced slope in Fujian Province as a case study,an intelligent optimization framework that integrates a deep neural network(DNN)with the Non-dominated Sorting Genetic Algorithm Ⅱ(NSGA-Ⅱ)is proposed for the multi-objective optimization of slope reinforcement design.Firstly,the finite element simulations are conducted to generate training samples for a DNN surrogate that maps key design parameters to stability response metrics.Then,the surrogate is embedded in the NSGA-Ⅱ search to obtain Pareto-optimal solutions with objectives of minimizing horizontal displacement and reinforcement quantity while maximizing the factor of safety.The analysis of the Pareto front and extraction of representative designs confirm the effectiveness of proposed method in balancing safety and economy,providing theoretical support and engineering guidance for optimizing high-fill reinforced slopes.
ZHA Wenhua;TAN Xuejian;XU Tao;XU Yuanxin;LAI Siling;JI Chao
School of Civil&Architecture Engineering,East China University of Technology,Nanchang 330013,Jiangxi,ChinaSchool of Civil&Architecture Engineering,East China University of Technology,Nanchang 330013,Jiangxi,ChinaSchool of Civil&Architecture Engineering,East China University of Technology,Nanchang 330013,Jiangxi,ChinaSchool of Civil&Architecture Engineering,East China University of Technology,Nanchang 330013,Jiangxi,ChinaFujian Makeng Mining Co.,Ltd.,Longyan 364021,Fujian,ChinaGeographic Information Engineering Brigade of Jiangxi Provincial Geological Bureau,Nanchang 330013,Jiangxi,China
交通工程
高填方边坡加筋设计多目标优化深度神经网络非支配排序遗传算法
high-fill slopereinforcement designmulti-objective optimizationdeep neural networkNon-dominated Sorting Genetic Algorithm Ⅱ
《水力发电》 2026 (1)
45-51,7
江西省"双千计划"支持项目(DHSQT22021002)东华理工大学研究生创新专项基金项目(DHYC-202419)
评论