黄土颗粒体微结构的主成分分析方法研究  

Study on Principal Component Analysis of Microstructure of Loess Particles

作　　者：唐华瑞[1] 韩灵杰[2] 王杏杏[3] 高凌霞 

机构地区：[1]河南理工大学万方科技学院,郑州451400 [2]郑州科技学院土木工程学院,郑州450000 [3]天津城建大学土木工程学院,天津300384 [4]大连民族大学土木建筑工程学院,大连116600

出　　处：《武汉理工大学学报》2015年第10期78-83,共6页Journal of Wuhan University of Technology

基　　金：中央高校基本科研业务费专项资金资助项目(201502040402)

摘　　要：针对21个黄土样本,利用SEM图像采集其微结构参数,发现微结构参数众多,量纲与所表达的物理意义各不相同,各参数表示的信息多有重复。针对以上问题,提出了采用主成分分析法研究某域范围内颗粒体微结构参数的方法。结果表明,前3个主成分的综合贡献率为96%,可以反映所有原始微结构参数的大部分信息,不但减少了"信息重复",还降低了颗粒体微结构参数的维数。同时发现,第1主成分的分散性最大,随主成分次序增加,分散性逐渐减小,第9主成分分散性最小。降维之后,主成分指标减少,这些主成分指标可以用来建造本构方程等土力学关系,且可以通过建立的力学关系来研究土体微结构的再造过程。The microstructural parameters were collected through SEM images for the 21 loess samples, which were found to be numerous, the dimension and the physical meaning were different, and the information expressed by each pa- rameter was much repeated. In view of the above problems, the principal component analysis method was proposed to study the microstructural parameters of the particles in a certain range. The results show that the comprehensive contri- bution rate of the first 3 principal components is 96%, which can reflect most information of all the original mierostruc- tural parameters. Thus it not only reduces the information duplication, but also reduces the dimension of microstructural parameters on the particles. At the same time, the dispersion of the first principal component is the largest. With the in- crease of main components, the dispersion is gradually reduced, and the ninth main component is the least. After dimen- sion reduction, the principal component indices are reduced, and constitutive equation and other soil mechanics relation- ships can be established through these principal component indices. We can use the established mechanics relationship to study the mierostructural reconstruction process of soil.

关 键 词：黄土 微结构 主成分分析 累积贡献率 信息重复 分散性 

分 类 号：P642.131[天文地球—工程地质学]