融合时序InSAR与MIDAS的露天边坡稳定性分析及预测

Analysis and Prediction of Slope Stability in Open-pit Mines by Integrating Time Series InSAR Technique and MIDAS

针对当前矿区地表形变监测技术监测范围小、周期长、采样率低等问题,以龙首露天矿为工程背景,融合SBAS-InSAR技术、MIDAS数值模拟与长短期记忆(Long Short Term Memory,LSTM)网络,提出了一种边坡稳定性的分析与预测一体化方法。利用SBAS-InSAR技术获取研究区2014—2023年地表垂直向形变时序反演结果,并分析其时空演化特征与形变机理。以典型形变剖面为研究对象,采用MIDAS GTS NX软件模拟边坡在强震作用下的稳定性,并分析边坡破坏规律及形变特征。采用贝叶斯优化算法(bayesian optimization,BO)优化LSTM网络,搭建并优化预测模型用于矿区地表形变预测。结果表明:南侧边坡垂直向形变相对严重,沉降速率达176.3 mm/a,累积沉降量达1489 mm;在强震数值模拟中边坡产生严重位移变形并最终失稳;基于SBAS-InSAR监测结果对各预测模型进行精度验证,验证结果表明BO-LSTM模型的预测精度最优,平均绝对误差与均方根误差至少降低了18%和16%。采用该模型预测矿区未来地表垂直向形变,预测结果表明,未来2 a内矿区形变速率放缓,边坡处于稳定状态。

Aiming at the problems of small monitoring range,long cycle and low sampling rate of surface deformation monitoring technology in mining area,taking Longshou Open-pit Mine as the engineering background,an integrated method of slope stability analysis and prediction is proposed by combining SBAS-InSAR technology,MIDAS numerical simulation and Long Short Term Memory(LSTM)network.SBAS-InSAR technology was used to obtain the inversion results of surface vertical deformation time series in the study area from 2014 to 2023,and its spatial and temporal evolution characteristics and deformation mechanism were analyzed.Taking the typical deformation profile as the research object,MIDAS GTS NX software was used to simulate the stability of the slope under strong earthquake,and the failure law and deformation characteristics of the slope were analyzed.The bayesian optimization(BO)algorithm is used to optimize the LSTM network,and the prediction model is built and optimized for the prediction of surface deformation in the mining area.The results show that the vertical deformation of the south slope is relatively serious,the settlement rate is 176.3 mm/a,and the cumulative settlement is 1489 mm.In the numerical simulation of strong earthquakes,the slope has serious displacement deformation and eventually loses stability.Based on the SBAS-InSAR monitoring results,the accuracy of each prediction model is verified.The verification results show that the BO-LSTM model has the best prediction accuracy,and the mean absolute error and root mean square error are reduced by at least 18%and 16%.The model is used to predict the future vertical surface deformation of the mining area.The prediction results show that the deformation rate of the mining area will slow down in the next two years,and the slope will be in a stable state.