基于BP神经网络的草根加筋土本构模型  

A Constitutive Model of Grassroots-reinforced Soil Based on BP Neural Network

作　　者：陈昌富[1] 彭钊[1] 刘怀星[1] 

机构地区：[1]湖南大学岩土工程研究所,湖南长沙410082

出　　处：《水土保持通报》2008年第3期93-96,共4页Bulletin of Soil and Water Conservation

基　　金：教育部高校博士点基金(20050532021);湖南省科技攻关项目(03GKY3129)

摘　　要：用室内三轴试验方法得到了草根加筋土的应力—应变关系,并基于试验结果建立了素土和混和草根加筋土的BP神经网络本构模型。模型计算结果与试验结果对比分析表明,该神经网络本构模型具有很高的拟和精度和良好的泛化能力,能充分体现岩土材料的非线性关系。利用莫尔—库仑准则拟合得到的加筋土的强度指标分析了草根加筋土的护坡机理。研究结果对于合理建立草根土的本构关系和深入认识植被护坡机理具有指导意义。Laboratory triaxial tests were carried out to obtain the stress-strain relationship of grassroots-reinforced soil （GRS）. BP neural network constitutive models of soil and GRS reinforcement in mixing were established based on test data. The result from comparing predicted values and measured values shows that the network constitutive model has good fitting precision and good generalization ability and can fully describe the non-linear relationship of geo-materials. The shear strength indexes of GRS fitted by Mohr-Columb criterion may be used to analyze the mechanism of GRS protection of slope. The research results are of importance for establishing the constitutive model of GRS and understanding the mechanism of vegetation protection of slope.

关 键 词：草根加筋土 BP神经网络 本构模型 莫尔-库仑准则 

分 类 号：TU411.7[建筑科学—岩土工程]