摘要
为掌握采空区上方所建高速公路的变形趋势,解决老采空区上方地表变形监测数据较少,不易建立时序沉降预测模型的问题,利用D-InSAR(Differential Interferometric Synthetic Aperture Radar)技术对某高速公路进行了变形监测和分析,同时将其结果同地面实测数据相融合,并以LS-SVM(Least Squares-Support Vector Machine)为基础,建立了采空区上方高速公路变形预计模型,通过实例,验证了模型的正确性。具体过程:处理融合数据为等时间间隔,并将其趋势项去除,对余项进行平稳性、正态性及零均值处理;利用Cao方法计算嵌入维数,建立训练样本集,并进行LS-SVM学习训练;最后,采用训练好的模型对未来地表沉降进行预计。以511号监测点为研究对象,建立滚动预计方法,结果显示其最大下沉绝对误差3 mm,最大相对误差2.2%,取得了较为可靠的预计成果。
In order to obtain the deformation law of expressway above goal, solve not enough monitoring data for aban- doned mine to establish the subsidence prediction models, the fused deformation values of level measure and Differenti- al Interferometric Synthetic Aperture Radar（D-InSAR） technique were used to establish the prediction models based on Least Squares-Support Vector Machine（LS-SVM）. The details are as follows：the fused data were processed to get equal-time interval time series deformation values, whose trend items should be rejected, and the residues were pro- cessed by stationary, normality and zero mean;Using Cao method to calculate embedding dimension, and establishing sample set to train LS-SVM model;Finally, using the model to predict the land subsidence in the future. The rolling prediction results of the No. 511 point show that the maximum absolute error of subsidence is 3 mm, maximum relative error is 2.2%. Therefore, the predicting results are reliability.
出处
《煤炭学报》
EI
CAS
CSCD
北大核心
2012年第11期1841-1846,共6页
Journal of China Coal Society
基金
国家自然科学基金资助项目(41071273)
中央高校基本科研业务费专项资金资助项目(2010QNA21)
国土环境与灾害监测国家测绘局重点实验室开放基金资助项目(LEDM2011B07)
作者简介
范洪冬（1981-），男，山东新泰人，讲师。E—mail：cumtfanhd@163．com
