摘要
露天矿山大规模开采引发的地表形变严重威胁了周边基础设施的稳固性及附近民众生命财产安全,形变演化趋势的精准预测对于保障矿山安全运营具有重要意义。针对当前形变监测技术的时空采样率低、成本高,以及数据处理过程中影响因子筛选困难、趋势预测精度欠佳等问题,以辽宁省鞍山市露天矿集中分布区为工程背景,提出了一种融合自回归差分移动平均(Autoregressive Integrated Moving Average,ARIMA)模型—长短期记忆网络(Long Short-Term Memory,LSTM)模型的多源因素融合回归的露天矿地表形变演化趋势分析方法。首先,利用短基线子集干涉测量(Small Baseline Subset Interferometric Synthetic Aperture Radar,SBAS-InSAR)技术开展2020年1月—2022年4月期间研究区地表形变的长时序监测,获取该时段内地表形变时空分布特征。然后,耦合因子分析及灰色关联分析法提取形变主影响因子,基于皮尔逊相关系数(Pearson)验证影响因子的筛选效果,同时考虑地表相邻点位形变的联动效应,构建了多源异构数据融合回归序列。在此基础上,引入自回归差分移动平均(ARIMA)模型改进的长短期记忆网络(LSTM)模型开展形变趋势预测,并采用平均绝对误差(Mean Absolute Error,MAE)、标准误差(Root Mean Square Error,RMSE)以及平均百分比误差(Mean Absolute Percentage Error,MAPE)评估所提方法的预测性能。结果表明:监测期内东鞍山矿东部、大孤山矿中部以及鞍千矿东部沉降相对严重,年均沉降速率最高达166.41 mm/a。耦合因子分析及灰色关联度法提取的影响因子合理可靠,融合高程、地形起伏度及累积降雨量等因子的形变序列更贴合矿区地表真实形变过程。与ARIMA-LSTM模型相比,基于多源因素融合回归模型的预测误差MAE、RMSE、MAPE分别降低了48.0%、16.7%和25.5%,预测精度有所改善且能够有效反映形变累积的整体趋势。
Surface deformation triggered by large-scale mining in open-pit mines seriously threatens the stability of the surrounding infrastructure and the safety of the lives and properties in the neighborhood.Accurate prediction of deformation trends is significant to ensure the safe operation and maintenance of mines.The current deformation monitoring technology has low spatial and temporal sampling rates and high cost.It isn′t easy to screen influencing factors during data processing.And the accuracy of deformation prediction is poor.To address these problems,a multi-source factor fusion regression method for analyzing the trend of surface deformation evolution in open-pit mines is proposed,which is based on a combined model of Autoregressive Integrated Moving Average(ARIMA)model and Long Short-Term Memory(LSTM)model.The concentrated distribution area of open-pit mines in Anshan City,Liaoning Province,was taken as the engineering background.Firstly,Small Baseline Subset Interferometric Synthetic Aperture Radar(SBAS-InSAR)was utilized to carry out time-series monitoring of surface deformation in the study area from January 2020 to April 2022 to obtain the characteristics of the temporal and spatial distribution of surface deformation during this period.Then,the factor analysis and gray correlation analysis were integrated to extract the main influence factors of deformation.Based on the Pearson correlation coefficient,the screening effect of the influence factors was verified.Meanwhile,the linkage effect of the deformation of neighboring points on the surface was considered.Based on this,a multiple-source heterogeneous data regression series was constructed.On this basis,the ARIMA model was adopted to modify the LSTM model to carry out the deformation trend prediction from the sequence of multisource factor regressions.Mean Absolute Error(MAE),Root Mean Square Error(RMSE)and Mean Absolute Percentage Error(Mean Absolute Percentage Error,MAPE)were adopted to evaluate the predictive performance of the proposed method.The results show that the subsidence of eastern Donganshan Mine,central Dagushan Mine and eastern Anqian Mine are relatively serious during the monitoring period,and the average annual subsidence rate is up to 166.41 mm/a.The influence factors extracted by coupling factor analysis and grey correlation degree method are reasonable and reliable,and the deformation sequence integrating elevation,topographic relief and accumulated rainfall is more suitable for the real deformation process of the mine surface.Compared with the ARMI-LSTM model,the prediction errors MAE、RMSE、MAPE of the multi-source fusion regression model are reduced by 48.0%,16.7%and 25.5%,respectively,and the prediction accuracy is improved and can effectively reflect the overall trend of deformation accumulation.
作者
李如仁
李梦晨
葛永权
刘明霞
LI Ruren;LI Mengchen;GE Yongquan;LIU Mingxia(School of Transportation and Geomatic Engineering,Shenyang Jianzhu University,Shenyang 110168,China;School of Civil Engineering,Shenyang Jianzhu University,Shenyang 110168,China;Department of Rail Transportation Affairs,Shenyang Center for Urban and Rural Construction,Shenyang 110014,China)
出处
《金属矿山》
北大核心
2025年第1期186-197,共12页
Metal Mine
基金
国家自然科学基金项目(编号:51774204)
辽宁省交通科技项目(编号:202338)。
