基于PCA和ANFIS的黄土湿陷系数预测  

Prediction of Loess Collapsibility Coefficients Based on PCA and ANFIS

作　　者：安宁[1] 

机构地区：[1]陕西铁路工程职业技术学院,陕西渭南714000

出　　处：《深圳职业技术学院学报》2011年第1期28-31,共4页Journal of Shenzhen Polytechnic

基　　金：陕西铁路工程职业技术学院科学研究基金资助项目

摘　　要：运用人工智能原理,提出一种基于主成分分析法(PCA)与自适应神经模糊推理系统(ANFIS)的黄土湿陷系数预测方法.首先通过主成分分析对黄土的物理指标提取主成分,以消除变量间的相关性和减少模型输入量的目的;再利用神经网络的高自适应性和模糊推理系统的推理能力建立ANFIS模型,提出一种新的黄土湿陷性预测方法.通过实测数据和预测数据的对比分析,平均误差0.29%,最大误差20%,在工程上可以接受的范围,实例说明这种预测方法是可行的.Guided by artificial intelligence theory, a prediction method of loess collapsibility coefficients was proposed based on principal component analysis （PCA） and adaptive neuro-fuzzy inference system （ANFIS）. The principal component of physical indicators of the loess was extracted based o an analysis of principal components to eliminate the correlation between variables and reduce the amount of input. An ANFIS model was then established using the high adaptability of neural network and the reasoning faculties of fuzzy inference system, offering a new prediction method of collapsible loess was suggested. By comparison of the measured data and the predicted data, the average error was 0.29% and the maximum error was 20%, both of which can be accepted in actual projects. Its applications also indicated that this prediction method was feasible.

关 键 词：主成因分析 模糊神经网络 ANFIS 黄土湿陷性 MATLAB 

分 类 号：TU411.2[建筑科学—岩土工程]