露天矿边坡变形的LMD-Elman时序滚动预测研究  

Research on LMD-Elman-based time-series rolling prediction of slope deformation in open-pit mine

作　　者：李胜[1] 韩永亮[1] 杨宏伟 刘永睿 

机构地区：[1]辽宁工程技术大学矿业学院 [2]煤科集团沈阳研究院有限公司 [3]辽宁省科学技术协会

出　　处：《中国安全科学学报》2015年第6期22-28,共7页China Safety Science Journal

基　　金：国家自然科学基金资助(51004063);辽宁省高等学校优秀人才支持计划(LJQ2011029)

摘　　要：为准确预测边坡变形,采用局部均值分解(LMD)与Elman神经网络相结合的方法,构建基于LMD-Elman的露天矿边坡变形的滚动预测模型。通过LMD将时序样本分解为多个分量,利用Elman神经网络对各分量进行滚动预测,再叠加各预测值重构最终理论预测值。以某露天矿边坡实际监测数据为例,进行仿真预测。结果表明:监测数据自身携带诱导边坡变形及失稳的重要信息,基于LMD-Elman的滚动预测能有效揭示边坡变形的波动性、趋势性和周期性特征;模型预测结果的平均绝对误差为0.056 8 mm,平均相对误差为2.756 8%,与单一Elman模型相比,预测精度显著提高。To predict the slope deformation accurately,a rolling prediction model for open-pit mine slope deformation based on LMD-Elman is built,which combines the LMD and Elman neural network. Firstly,time-series samples are decomposed into multiple components by LMD. Then,rolling forecast of each component is carried out by Elman neural network. Finally,a theoretical prediction value is obtained by superposition of the forecasts. Taking the practical monitoring data on a certain open-pit mine slope as an example for making a simulation prediction. Results indicate that monitoring data itself carries important information on induction of slope deformation and instability, that the prediction model based on LMD-Elman can effectively reveal the volatility,trending and periodic characteristics of slope deformation,that the average absolute error of the model prediction result is 0. 056 8 mm,the average relative error is2. 756 8%,and that compared comparing to the single Elman model,the model improves the accuracy significantly.

关 键 词：露天矿边坡 局部均值分解(LMD) ELMAN神经网络 仿真预测 变形 

分 类 号：X936[环境科学与工程—安全科学]