滑坡变形的回归-神经网络预测模型研究  

Research on Prediction Model of Regression-Neural Network for Landslide Deformation

作　　者：王秀丽[1] 李恒凯[2] 

机构地区：[1]江西理工大学经济管理学院,江西赣州341000 [2]江西理工大学建筑与测绘工程学院,江西赣州341000

出　　处：《人民黄河》2012年第7期90-92,共3页Yellow River

基　　金：国家自然科学基金资助项目(41061041);江西省教育厅科技研究项目(GJJ11472)

摘　　要：受多种因素影响,滑坡变形具有趋势性和随机性的特点。从滑坡变形监测数据着手,将监测数据分离成趋势值和随机值,建立了滑坡变形的回归-神经网络预测模型。该模型采用逐步回归方法对滑坡变形的趋势值进行预测,用BP神经网络预测方法对滑坡变形的随机值进行预测。利用金沙江乌东德坝址区金坪子滑坡TP06点高程位移变化实测数据,对该模型进行了验证。结果表明:预测误差不超过11%,具有较高的预测精度。Because affected by many factors,it has trend and randomness for the monitoring data of landslides deformation.This paper started from the monitoring data for landslide deformation,the data was separated into trend values and random values,established a landslide deformation prediction model of regression-Neural network.The model used stepwise regression to predict trend values of landslide deformation and used BP neural network to predict to predict the random value of landslide deformation.The actual monitoring data for elevation displacement variation of Jinpingzi landslide TP06 point in Wudongde dam site area of the Jinsha River was used for model verification.The results show that the error of 6 issue data to predict is less than 11%,which is of high precision.

关 键 词：预测模型 滑坡变形 逐步回归 BP神经网络 

分 类 号：X43[环境科学与工程—灾害防治] P208[天文地球—地图制图学与地理信息工程]