基于有限元与LSTM机器学习模型的边坡稳定性预警分析

1.华南理工大学土木与交通学院,广州 510641;2.广东省地质环境监测总站,广州 510510

LSTM机器学习模型;有限元模拟;有限差分法;预警治理

Early Warning Analysis of Slope Stability Based on Finite Element and LSTM Machine Learning Model
ZHANG Boxiang1,SU Songlin1,SU Wenji1,WEI Pingxin2

1.School of Civil Engineering&Transportation , South China University of Technology , Guangzhou 510641, China;2.Guangdong General Station of Geological Environment Monitoring , Guangzhou 510510, China

LSTM machine learning model;Finite element simulation;Finite difference method;Early warning and management

DOI: 10.13512/j.hndz.2024.02.08

备注

从降水量与边坡含水量关系出发,找寻边坡物理力学性能与含水量的相关关系;然后进行某地区历史降雨量大数据分析,通过LSTM模型预测未来降水量,后利用FLAC3D进行有限元强度折减法分析模拟边坡破坏和位移情况并计算边坡稳定系数,找寻临界稳定系数对应的边坡含水量进而得知对应降水量大小。根据预测给出可能的危险时段,提出预警,以便于工程人员防护治理。研究发现危险情况集中于4月至9月部分时段,对于未来边坡预警和治理提供了新的可行应对方法。
Based on the relationship between precipitation and water content of the slope,the correlation between physical and mechanical properties of the slope and water content was explored. Then, the big data analysis of historical precipitation in a certain area was carried out,and the long short-term memory (LSTM) model was used to predict future precipitation. After that, the finite element strength reduction method was used to analyze and simulate the slope failure and displacement by FLAC3D, and the slope stability coefficient was calculated. The water content of the slope corresponding to the critical stability coefficient was found, and the corresponding precipitation was obtained. According to the prediction, the possible dangerous period was clarified for early warning, so as to facilitate the protection and management of engineering personnel. The results find that the dangerous situation is concentrated in some periods from April to September, which provides a new feasible response method for slope early warning and treatment in the future.
·