InSAR与数值模拟协同的排土场边坡稳定性分析及形变预测研究

Slope Stability Analysis and Deformation Prediction of Dump Site by Combining SBAS-InSAR and Numerical Simulation

为探明历史滑坡治理后排土场边坡的形变规律,以大孤山露天矿排土场为例,首先利用短基线子集干涉(SBAS-InSAR)技术进行地表沉降时序监测,分析排土场边坡的主要沉降区域及稳定性影响因素;然后采用COMSOL软件建立典型沉降区域内部位移、安全系数对降雨的响应关系,对研究区的时空变形特征进行协同分析;最后通过搭建集成多类别损失函数的粒子群(PSO)优化长短期记忆网络(LSTM)预测模型,开展区域沉降时序预测。结果表明,中北部地区存在3处典型沉降区域,最大累积沉降量达295.8 mm,年平均沉降速率最高约为134.2mm/a;有效降雨量是边坡形变的主要影响因素,且随着前期降雨过程的持续,边坡稳定系数的下降速率最高约为0.025%。多类别损失函数集成的PSO-LSTM模型能反映排土场沉降波动趋势,且其预测精度综合评价指标(L_(total))为2.48 mm。研究成果可为后续排土场滑坡灾害防治提供理论依据。

To explore the slope deformation law of the dump site that landslide risk still exists after the treatment of open-pit mine dumps,the dump site in Dagushan open-pit mine was taken as an example.Firstly,the surface settlement time series monitoring was carried out by using the Short Baseline Subset and Interferometric Synthetic Aperture Radar(SBAS-InSAR)technique.The main settlement areas and stability influencing factors of the slope of the dump site were analyzed.Then,COMSOL software was used to establish the response relationship between internal displacement and safety factor of typical settlement areas and rainfall,and to conduct collaborative analysis on the spatiotemporal deformation characteristics of the study area.Finally,by constructing a Particle Swarm Optimization(PSO)with integrated multi class loss functions to optimize the Long Short Term Memory Network(LSTM)prediction model,regional settlement time series prediction was carried out.The results show that there are three typical settlement areas in the central and northern regions,with a maximum cumulative settlement of 295.8mm and a maximum annual average settlement rate of about 134.2mm/a.Effective rainfall is the main influencing factor of slope deformation,and as the early rainfall process continues,the maximum decrease rate of slop stability coefficient is about 0.025%.The PSO-LSTM model integrated with multi class loss functions can reflect the fluctuation trend of settlement in the dump site,and its prediction accuracy comprehensive evaluation index(L_(total))is 2.48mm.The research results can provide a theoretical basis for the subsequent prevention and control of landslide disasters in dump sites.