基于Relief算法与ReLU核ELM的煤矿开采最大下沉预测模型研究  

Study of the Maximum Subsidence Forecast Model of Coal Mining Based on Relief Arithmetic and ReLU Core ELM

作　　者：魏勇[1,2] 唐延东[1] 喻强[1] 杨春兰[3] WEI Yong;TANG Yan-dong;YU Qiang;YANG Chun-lan(School of Architectural Engineering,Sichuan Engineering Vocational Technical College,Deyang 618000,China;State Key Laboratory of Geological Disaster Prevention and Geological Environment Protection,Chengdu University ofTechnology,Chengdu 610059,China;School of Architectural and Environment,Xihua University,Chengdu 610039,China)

机构地区：[1]四川工程职业技术学院建筑工程系,四川德阳618000 [2]成都理工大学地质灾害防治与地质环境保护国家重点实验室,四川成都610059 [3]西华大学土木建筑与环境学院,四川成都610039

出　　处：《煤矿开采》2018年第5期68-72,21,共6页Coal Mining Technology

摘　　要：尝试引入Re LU function核的ELM算法及Relief Algorithm对开采区最大下沉量进行预测。首先基于Relief Algorithm对现场岩移数据进行筛选优化;然后通过隐含层数目循环实验选出预测精度较高的ELM预测模型隐含层数目;再筛选优化后的参数为输入,最大下沉为目标分别建立基于Re LU function核、igmoid function核、Radial basis function核及Hardlim function核的ELM预测模型;最后对4种模型的预测结果进行对比分析。结果表明:采厚、平均采深、走向长度和倾向长度与最大下沉关系显著;以Re LU function核、隐含层神经元数目为57的ELM的预测结果精度显著优于对比组。The maximum subsidence of goaf was forecast by ELM arithmetic that introduced ReLU function core and Relief Algorithm.First,stratum movement data were filtrate based on relief Algorithm,and then hidden layers number of ELM forecast model with higher forecast precision was picked by circulation experiment of hidden layer number,and the picked parameters were input,ELM forecast model for the maximum subsidence was built,which based on ReLU function core,igmoid function core,Radial basis function core and Hardlim function core,at the last,the forecast results of the four model were contrastive analysis.The results showed that relationship between mining thickness,mean mining depth,strike length,dip length and the maximum subsidence was obviously,and ELM forecast results precision was better than others,which core was ReLU function and hidden layers nerve cell was 57.

关 键 词：开采最大下沉 RELIEF ALGORITHM ReLU FUNCTION ELM 

分 类 号：TD325.2[矿业工程—矿井建设]