没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
第 10 卷 第 24 期
2019 年 12 月
黑龙江科学
HEILONGJIANG SCIENCE
Vol 10
Dec 2019
基于 PCA - BP 神经网络的概率积分法参数算法研究
黄 晖
1
ꎬ池深深
2
ꎬ韩必武
1
ꎬ刘可胜
1
(1 淮南矿业(集团)有限责任公司ꎬ淮南 232001ꎻ2 安徽理工大学 地球与环境学院ꎬ淮南 232001)
摘要: 随着经济的快速发展ꎬ我国对煤炭资源的需求量也越来越大ꎬ近年来地下煤炭资源被大量开采ꎬ造成了地表沉陷、裂缝、粉
尘、固态垃圾等一系列环境问题ꎬ给矿区的生产生活带来严重威胁ꎮ 为最大程度开采煤炭资源并减少地表大面积沉陷ꎬ对开采沉陷
预计理论展开广泛研究ꎮ 常见的开采沉陷预测模型有力学模型法、概率积分法 + 时间函数模型法、基于非线性理论的预测模型法ꎬ
当前最常用的是概率积分法及其改进模型ꎮ 为更好地建立地质采矿条件与概率积分法预计参数之间的非线性关系ꎬ以 40 个典型
矿区的实测数据为样本ꎬ采用主成分分析法获取影响预计参数的主因素ꎬ作为 BP 神经网络输入层ꎬ采用 BP 神经网络解算概率积分
法预计参数ꎮ 主成分分析结果表明ꎬ影响概率积分法预计参数的地质采矿条件ꎬ按敏感性由大到小的排序为:松散层厚度、煤层倾
角、煤层厚度、采深、倾向采宽比、走向采宽比、推进 速度、覆岩 平均坚固性系数ꎮ BP 神经网络 预测结果表 明:预测相对 误差 在
- 3 80% ~ 10 00% ꎬ精度满足工程需要ꎮ 剔除敏感性较小的参数走向采宽比、推进速度、覆岩平均坚固性系数ꎬ并基于剔除后的数
据进行 BP 神经网络建模ꎬ预测对比结果表明其预测精度高ꎬ此方法行之有效ꎮ
关键词: 主成分回归分析ꎻBP 神经网络ꎻ概率积分法参数
中图分类号: TD821 文献标志码: A 文章编号: 1674 - 8646(2019)24 - 0001 - 05
收稿日期:2019 - 10 - 20
基金项目:国家自然科学基金资助项目(41474026)ꎻ淮南矿业
(集 团 ) 有 限 责 任 公 司 资 助 项 目 ( HNKY - JTJS
(2017) - 122ꎬHNKY - JTJS(2018) - 178)ꎻ中煤新集
刘庄矿业有限公司(ZMXJ - LZ - JS - 2018 - 25)
作者简介:黄晖 (1964 - )ꎬ男ꎬ高 级 工 程 师ꎬe - mail:dzyhangh
@ 163 comꎮ
Research on prediction of mining subsidence parameters based on BP
neural network and principal component analysis
HUANG Hui
1
ꎬ CHI Shen ̄shen
2
ꎬ HAN Bi ̄wu
1
ꎬ LIU Ke ̄sheng
1
(1 Huainan Mining Group Co ꎬ Ltd ꎬ Huainan 232001ꎬ Chinaꎻ
2 School of Earth and Environmentꎬ Anhui University of Science and Technologyꎬ Huainan 232001ꎬ China)
Abstract: With the rapid development of economyꎬ China’s demand for coal resources is also growing. In recent yearsꎬ
a large number of underground coal resources have been minedꎬ resulting in a series of environmental problems such as
surface subsidenceꎬ cracksꎬ dustꎬ solid wasteꎬ etc. ꎬ which pose a serious threat to the production and life of the mining
area. In order to maximize the exploitation of coal resources and reduce large ̄scale subsidence of the surfaceꎬ the theory
of mining subsidence prediction is widely studied. The common prediction models of mining subsidence include
mechanical model methodꎬ probability integral method + time function model methodꎬ and prediction model method
based on nonlinear theory. At presentꎬ probability integral method and its improved model are most commonly used. In
order to better establish the nonlinear relationship between geological mining conditions and probability integral methodꎬ
the principal variables affecting the prediction parameters were extracted from the measured data of 40 typical mining
areas by principal component analysis methodꎬ and the relatively small factors were eliminated. Then the BP neural
network was used as the input layer of BP neural network to predict the probability. The results of integration method and
principal component analysis show that the order of sensitivity of geological mining conditions affecting the parameters of
probability integration method is loose layer thickness > coal seam dip angle > coal seam thickness > mining depth >
inclined mining width ratio > strike mining width ratio > advancing speed > average hardness coefficient of overburden
rock. The BP neural network prediction results show that the prediction is relative. The error is between - 3. 80% and
10% ꎬ and the accuracy meets the engineering requirements. Remove the parameters with less sensitivityꎬ such as strike
width ratioꎬ advance speed and average firmness
coefficient of overburdenꎬ and the BP neural network
model is built based on the eliminated data. The
prediction results show that the prediction accuracy is
high and the method is effective.
Key words: Principal component analysisꎻ BP neural
networkꎻ Probability integral method parameters
1
资源评论
数据资源
- 粉丝: 113
- 资源: 23万+
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功