BP人工神经网络在富水性评价中的应用
|
摘要
为了对研究区含水层进行富水性评价,并减轻评价结果对水文孔的依赖程度,通过对地质及水文地质资料的分析,确定出四个富水性主控因素,分别为含水层厚度、岩芯采取率、脆性岩厚度比和风化影响指数。引入具有自主学习、非线性映射能力的BP人工神经网络,将25组经量化、归一化处理的主控因素数据作为网络学习样本、以实测单位涌水量为预测目标,通过反复训练学习,实现了对主控因素到单位涌水量映射关系的精确模拟。最后,使用训练好的神经网络对研究区富水性进行了预测,并引入灵敏度分析方法分析了预测结果对主控因素的敏感性。
In order to evaluate the water abundance of the study area,and to reduce the dependence of the evaluation results on the hydrographic bore. After making the best of geological and hydrogeological data,four main controlling factors is established,which are aquifer thickness, core recovery percentage, thickness ratio of brittle rock and weathering influence index. BP artificial neural network with autonomous learning,nonlinear mapping ability is introduced into the study,the data of 25 groups of main control factors treated by quantification,normalization is used as a network of learning samples,taking the measured specific well yield as the prediction target, the simulation of the mapping relationship between the main control factor and the specific well yield has been realized by repeated training. Finally,the trained neural network is used to predict the water abundance,and the sensitivity analysis method is introduced to analyze the sensitivity of the prediction results to the dominant factors.
In order to evaluate the water abundance of the study area,and to reduce the dependence of the evaluation results on the hydrographic bore. After making the best of geological and hydrogeological data,four main controlling factors is established,which are aquifer thickness, core recovery percentage, thickness ratio of brittle rock and weathering influence index. BP artificial neural network with autonomous learning,nonlinear mapping ability is introduced into the study,the data of 25 groups of main control factors treated by quantification,normalization is used as a network of learning samples,taking the measured specific well yield as the prediction target, the simulation of the mapping relationship between the main control factor and the specific well yield has been realized by repeated training. Finally,the trained neural network is used to predict the water abundance,and the sensitivity analysis method is introduced to analyze the sensitivity of the prediction results to the dominant factors.
引文
[1]樊亚红,刘文连,曹福明.改进突水系数法在先锋露天矿底板突水评价中的应用[J].煤炭工程,2016,48(1):114-117.
[2]武强,王洋,赵德康,等.基于沉积特征的松散含水层富水性评价方法与应用[J].中国矿业大学学报,2017,46(3):460-466.
[3]韩超,泮晓华,李国梁,等.基于GIS多源信息集成的含水层富水性模糊层次分析法[J].水文地质工程地质,2012,39(4):19-25.
[4]国家安全生产监督管理总局,国家煤矿安全监察局.煤矿防治水规定[M].北京:煤炭工业出版社,2009.
[5]武强,樊振丽,刘守强,等.基于GIS的信息融合型含水层富水性评价方法——富水性指数法[J].煤炭学报,2011,36(7):1124-1128.
[6]王广弟,刘雨琪,刘红艳.含水层富水性分区的无量纲与可拓层次分析法研究[J].煤炭工程,2016,48(S1):145-147.
[7]董军,胡上序.混沌神经网络研究进展和展望[J].信息与控制,1997,26(5):360-368.
[8]朱大奇.人工神经网络研究现状及其展望[J].江南大学学报(自然科学版),2004,3(1):103-110.
[9]徐佩华.基于人工神经网络方法的锦屏一级水电站枢纽区高边坡稳定性分区研究[D].长春:吉林大学,2006.
[10]武强,陈佩佩,董东林,等.基于GIS与ANN耦合技术的地裂缝灾情非线性模拟预测系统[J].地震地质,2002,24(2):249-257.
[11]刘业娇,任玉辉,任静静,等.基于BP神经网络的煤矿安全管理绩效评价[J].煤炭工程,2013,45(2):135-137.
[12]王增银.岩层“富水性”及评价方法探讨[J].地质科技情报,1987(3):44-46.
[13]傅荟璇,赵红.MATLAB神经网络应用设计[M].北京:机械工业出版社,2010.
[14]朱大奇,史慧.人工神经网络原理及应用[M].北京:科学出版社,2006.
[15]武强,周英杰,董云峰,等.基于地理信息系统与人工神经网络耦合技术的产油潜力评价模型[J].中国石油大学学报(自然科学版),2004,28(5):18-22.
[16]陈太聪,韩大建,苏成.参数灵敏度分析的神经网络方法及其工程应用[J].计算力学学报,2004,21(6):752-756.
[2]武强,王洋,赵德康,等.基于沉积特征的松散含水层富水性评价方法与应用[J].中国矿业大学学报,2017,46(3):460-466.
[3]韩超,泮晓华,李国梁,等.基于GIS多源信息集成的含水层富水性模糊层次分析法[J].水文地质工程地质,2012,39(4):19-25.
[4]国家安全生产监督管理总局,国家煤矿安全监察局.煤矿防治水规定[M].北京:煤炭工业出版社,2009.
[5]武强,樊振丽,刘守强,等.基于GIS的信息融合型含水层富水性评价方法——富水性指数法[J].煤炭学报,2011,36(7):1124-1128.
[6]王广弟,刘雨琪,刘红艳.含水层富水性分区的无量纲与可拓层次分析法研究[J].煤炭工程,2016,48(S1):145-147.
[7]董军,胡上序.混沌神经网络研究进展和展望[J].信息与控制,1997,26(5):360-368.
[8]朱大奇.人工神经网络研究现状及其展望[J].江南大学学报(自然科学版),2004,3(1):103-110.
[9]徐佩华.基于人工神经网络方法的锦屏一级水电站枢纽区高边坡稳定性分区研究[D].长春:吉林大学,2006.
[10]武强,陈佩佩,董东林,等.基于GIS与ANN耦合技术的地裂缝灾情非线性模拟预测系统[J].地震地质,2002,24(2):249-257.
[11]刘业娇,任玉辉,任静静,等.基于BP神经网络的煤矿安全管理绩效评价[J].煤炭工程,2013,45(2):135-137.
[12]王增银.岩层“富水性”及评价方法探讨[J].地质科技情报,1987(3):44-46.
[13]傅荟璇,赵红.MATLAB神经网络应用设计[M].北京:机械工业出版社,2010.
[14]朱大奇,史慧.人工神经网络原理及应用[M].北京:科学出版社,2006.
[15]武强,周英杰,董云峰,等.基于地理信息系统与人工神经网络耦合技术的产油潜力评价模型[J].中国石油大学学报(自然科学版),2004,28(5):18-22.
[16]陈太聪,韩大建,苏成.参数灵敏度分析的神经网络方法及其工程应用[J].计算力学学报,2004,21(6):752-756.