摘要
隧道工程围岩的类别是评价隧道工程地质条件的一个综合性量化指标,是进行隧道工程建设的基础。为快速、有效地判别围岩类型,提出了将因子分析(FA)与改进的支持向量机(GA-SVM)相结合的隧洞围岩分类模型。首先,根据岩体岩性、地质构造、岩体结构等特性,选取岩石质量指标、完整性系数、单轴饱和抗压强度、纵波波速、弹性抗力系数和结构面摩擦系数6个指标作为隧洞围岩分类的初始判别指标。其次,采用因子分析理论对原始指标变量进行属性约简,提取了公共因子,减少了判别指标之间信息冗余。最后,利用遗传算法(GA)优化支持向量机(SVM)的惩罚因子C和核函数参数σ,并将提取出的公共因子作为GA-SVM模型的输入变量,建立了基于因子分析的GA-SVM隧洞围岩分类模型。将现场勘测的29组围岩数据作为训练数据,另用7组数据作为测试数据,同时将该模型分类结果与SVM、BPNN、FDA模型分类结果进行了对比。结果表明:因子分析可以有效地提取围岩分类指标,降低指标间信息重复度;利用GA优化参数可以提高SVM模型的精度与泛化能力;用该模型预测隧洞围岩的类别与实际分类相吻合,其错误预测率为0,研究成果可为隧洞围岩快速分类提供一种新思路。
The class of surrounding rock in tunnel engineering is a comprehensive quantitative index for evaluating tunnel engineering geological conditions, it is the foundation for the construction of tunnel engineering. In order to identify the class of surrounding rock quickly and effectively,a kind of classification model of tunnel surrounding rock by combining factor analysis（ FA） with the improved support sector machine（ GA-SVM） is put forward. First,on the basis of the characteristics such as rock mass lithology,geological structure and rock mass structure,6 indexes are selected as the initial identification indexes for tunnel surrounding rock classification. These 6 indexes are rock quality index,integrity coefficient,uniaxial saturated compressive strength, longitudinal wave velocity, elastic resistance coefficient and friction coefficient of structural plane respectively. Then, adopting the theory of factor analysis, the attribute reduction of the original index variables is carries out and the common factors are extracted to reduce the information redundancy between the identification indexes. At last,the genetic algorithm（ GA） is put into use to optimize the penalty factor C and the kernel function parameter σ of support vector machine（ SVM）,the common factors are extracted as the input variables of GA-SVM model,and the tunnel surrounding rock classification model based on GA-SVM and factor analysis is established. Twenty-nine groups of surroundingrock data obtained by the site survey are used as the training data and another 7 groups of data are utilized for testing. Meanwhile,the classification result produced by the GA-SVM model is compared with the results obtained through the models of SVM,BPNN and FDA. The result shows that（ 1） the factor analysis can effectively extract the classification index of surrounding rock,and also can decline the index multiplicity between indexes;（ 2） the accuracy and generalization ability of SVM model can be improved by using GA to optimize the parame
作者
温廷新
于凤娥
邵良杉
田煜晨
WEN Ting-xin;YU Feng-e;SHAO Liang-shan;TAN Yu-chen(Institute of System Engineering,Liaoning Technical University,Huludao Liaoning 125105,China;School of Business Administration,Liaoning Technical University,Huludao Liaoning 125105,China)
出处
《公路交通科技》
CAS
CSCD
北大核心
2018年第9期63-70,共8页
Journal of Highway and Transportation Research and Development
基金
国家自然科学基金项目(71371091)
辽宁省社科基金资助项目(L14BTJ004)
关键词
隧道工程
隧道围岩分类
遗传算法
因子分析
支持向量机
tunnel engineering
tunnel surrounding rock classification
genetic algorithm
factor analysis
support sector machine
