煤矿安全2025，Vol.56Issue(6)：144-154,11.DOI:10.13347/j.cnki.mkaq.20240525
融合DS-InSAR与THPF-LSTM的关闭矿井地表形变监测及预测
Fusion of DS-InSAR and THPF-LSTM for monitoring and predicting surface deformation in closed mines
摘要
Abstract
After the closure of a mine,the overlying rock layer and the ground surface will deform again,this affects the safe opera-tion of buildings(structures).Due to the lack of supervision of mine closure,the spatial and temporal evolution of surface deforma-tion and prediction and warning models are not well studied.To this end,we proposed a prediction model for surface deformation of closed mines combining distributed scatter interferometric synthetic aperture radar(DS-InSAR),temporal high pass filtering(THPF),and a long short term memory network(LSTM).Taking the 98-view Sentinel-1A uptrack image as the data source,firstly,the DS-In-SAR method combined with persistent scatterer(PS)and DS points was used to obtain the time-series surface subsidence informa-tion of the closed mines in western Xuzhou for the period from November 2019 to December 2022;then the THPF was used to de-compose the original subsidence sequences to obtain the high frequency and low frequency,and then,LSTM was used to complete the deformation prediction of the high and low frequency sub-sequence,and the predicted values of the high and low frequency sub-sequence were superimposed to obtain the final prediction result.The results show that:the density of DS-InSAR monitoring points is uniformly distributed,and the coefficient of determination between the measured deformation and the monitoring results reached 0.95;compared with the LSTM model,the maximum RMSE(root mean square error)of the THPF-LSTM model in the prediction points is 3.0,and the mean absolute error(MAE)is 2.4,and the maximum Adjusted R-Square was 0.9,which is better than 4.5,3.9,and 0.6 of the traditional LSTM model,and the comprehensive prediction accuracy of the model is improved by more than 20％,and it can accurately reflect the trend and volatility of the surface deformation of the closed mine,and can effectively improve the predic-tion accuracy of mine settlement in the short term.The method of this paper realizes the integrated analysis of monitoring and predic-tion of surface deformation in closed mines.关键词
地表形变预测/关闭矿井/DS-InSAR/LSTM神经网络/时间域高通滤波Key words
surface deformation prediction/mine closure/DS-InSAR/LSTM neural network/temporal high pass filtering分类
矿业与冶金引用本文复制引用
张建阳,范洪冬,朱向阳,孙明虎..融合DS-InSAR与THPF-LSTM的关闭矿井地表形变监测及预测[J].煤矿安全,2025,56(6):144-154,11.基金项目
国家自然科学基金资助项目(42274054,51774270) （42274054,51774270）
国家重点研发计划资助项目(2022YFE0102600,2017YFE0107100) （2022YFE0102600,2017YFE0107100）
