小样本监测信息露天矿边坡变形预测模型对比分析  

Comparative Analysis of Slope Deformation Forecast Models for Open-pit Mine Under Small Sample Monitoring Information

作　　者：吴浩[1] 阮明浩 张宏 张建华[1] 叶海旺[1] 董元锋[1] 

机构地区：[1]武汉理工大学资源与环境工程学院,武汉430070 [2]武汉市江夏区土地交易中心,武汉430200

出　　处：《武汉理工大学学报（交通科学与工程版）》2014年第3期544-546,551,共4页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家自然科学基金项目(批准号:40901214);中国博士后科学基金资助项目(2013M531749);精密工程与工业测量国家测绘地理信息局重点实验室开放基金项目(批准号:PF2011-20);中央高校基本科研业务费专项资金项目(批准号:2013-IV-040);国家级大学生创新创业训练计划项目(批准号:20131049708009)资助

摘　　要：利用GM(1,1),BP神经网络和灰色BP神经网络组合三种模型,分别对不同样本容量的小样本监测信息进行测试,对比分析预测结果的准确性与稳定性.结果表明,边坡变形预测模型受既有监测信息的样本容量影响较大,样本容量的增加有利于边坡变形预测模型精度的提高.在既有监测信息较少的情况下,GM(1,1)模型预测精度虽然最高,但缺乏稳定性;单一BP神经网络模型的预测精度由于样本较少,其精度较差.从预测结果的稳定性和精度两个方面综合对比来看,灰色BP神经网络组合模型更适用于小样本监测信息情况下的露天矿边坡变形趋势的预测.Three models,including GM（1,1）model,BP neural network model,and the combined model based on grey and BP neural network,were chosen to forecast small sample monitoring with different sample size and their accuracy and stability of forecast results were compared.The results show that the sample size has a significant effect on slope deformation forecast model,and the slope deformation forecast model can achieve a higher accuracy with the sample size increasing.In the case of small sample monitoring information,the accuracy of GM（1,1）model for small sample is relatively high,but its stability is lower and the forecast accuracy of the single BP neural network model is poor because of fewer sample.From the integrated analysis of the forecast results in the stability and precision,the combined model based on gray and BP neural network,is more suitable for slope deformation tendency forecast under small sample monitoring information than other two methods.

关 键 词：边坡变形预测模型 小样本监测信息 GM(1 1)模型 BP神经网络模型 灰色BP神经网络合模型 

分 类 号：TD176[矿业工程—矿山地质测量]