收稿日期: 2015-04-11
修回日期: 2015-08-03
网络出版日期: 2016-02-25
基金资助
“十二五”国家科技支撑计划项目“金属矿床高效地下开采关键技术研究及示范”(编号:2013BAB02B05)资助
Research on Rock Mass Blastability Based on Principal Component Analysis and RBF Neural Network
Received date: 2015-04-11
Revised date: 2015-08-03
Online published: 2016-02-25
为了对岩体可爆性进行更精确的预测分级,建立了主成分分析法与RBF神经网络相结合的评价模型。以某矿山岩石为例,将影响岩石可爆性的容重、抗拉强度、抗压强度和岩体完整性系数作为评价指标,统计矿山13种岩体的样本数据。对样本数据进行主成分相关性预处理,将输出结果作为RBF神经网络的输入变量,岩体的爆破等级作为输出变量,得到的结果精度更高。研究结果表明:预测结果的相对误差均控制在5%以内,与BP神经网络预测误差(16%)相比,所得到实际预测结果与期望值之间的相对误差分别降低了71.94%、86.65%、73.20%和76.62%,预测精度显著提高。该模型为岩体可爆性分级预测提供了一种更为完善的方法。
李夕兵 , 朱玮 , 刘伟军 , 张德明 . 基于主成分分析法与RBF神经网络的岩体可爆性研究[J]. 黄金科学技术, 2015 , 23(6) : 58 -63 . DOI: 10.11872/j.issn.1005-2518.2015.06.058
In order to predict the rock mass blastability classification more accurately,the evaluation model was established based on Principal Component Analysis and RBF Neural Network.Taking a mine rock mass for example,the four evaluation indexes(the rock mass density,compressive strength,tensile strength,the integrality index of rock mass) that affect the rock mass blastability were considered,and the sample data of 13 actual mine rock mass were counted.The sample data were processed by the method of principal component,the results were used as input factors of RBF network,and the level of rock mass blastability was used as output factor,the precision of rock mass blastability prediction can be more higher.The research results show that relative errors of predicting outcomes are all controlled within 5%,and compared with the prediction errors by BP neural network,the expected value relative errors of the four rock mass are reduced 71.94%,86.65%,73.20%,76.62%,respectively,the classification prediction accuracy are obviously improved.The model provides a better evaluation for the rock mass blastability classification analysis.
[1] 张德明,王新民,郑晶晶,等.基于模糊综合评判的矿岩体可爆性分级[J].爆破,2010,27(4):43-47.
[2] 璩世杰,毛市龙,吕文生,等.一种基于加权聚类分析的岩体可爆性分级方法[J].北京科技大学学报,2006,28(4):324-329.
[3] 毛健,赵红东,姚婧婧.人工神经网络的发展及应用[J].电子设计工程,2011,19(20):62-65.
[4] 刘思峰,蔡华,杨英杰,等.灰色关联分析模型研究进展[J].系统工程理论与实践,2013,33(8):2041-2046.
[5] 王淑红,李英龙,戈保梁,等.主成分分析法与神经网络在选矿建模中的应用[J].有色矿冶,2001,17(6):25-28.
[6] 冯岩,王新民,程爱宝,等.采空区危险性评价方法优化[J].中南大学学报(自然科学版),2013,44(7):2281-2288.
[7] 张钦礼,周碧辉,王新民,等.充填管道失效风险性预测精度研究[J].中南大学报(自然科学版),2014,45(8):2805-2811.
[8] 陈建宏,刘浪,周智勇,等.基于主成分分析与神经网络的采矿方法优选[J].中南大学报(自然科学版),2010,41(5):1967-1972.
[9] 韩红桂,乔俊飞,薄迎春.基于信息强度RBF神经网络结构设计研究[J].自动化学报,2012,38(7):1084-1090.
[10]王翠.基于灰色理论和RBF神经网络民航客运量预测方法研究[D].北京:北京交通大学,2008:32-33.
[11]薛剑光,周健,史秀志,等.基于熵权属性识别模型的岩体可爆性分级评价[J].中南大学学报(自然科学版),2010,41(1):252-256.
[12]李蓉,宋娟,何永延.基于属性识别理论的岩体可爆性分级方法[J].金属矿山,2008,(5):32-34.
[13]辛明印,璩世杰,陈煊年,等.南芬露天铁矿的岩体可爆性分级方法及其应用[J].工程爆破,2006,12(1):7-10.
[14]林杰斌,刘明德.SPSS10.0与统计模式建构[M].北京:人民统计出版社,2001:185-190.
[15] Wen S H,Chung D D L.Enhancing the vibration reduction ability of concrete by using steel reinforcement and steel surface treatment[J].Cement and Concrete Research,2000,30(2):327-330.
[16] Sinha S K,Pandey M D.Probabilistic neural network for reliability assessment of oil and gas pipelines[J].Computer Aided Civil and Infrastructure Engintering,2002,17(5):320-329.
[17] 韩红桂,乔俊飞.RBF神经网络的结构动态优化设计[J].自动化学报,2010,36(6):862-865.
/
〈 | 〉 |