摘要
软土的工程性质很大程度上取决于它的内部结构。通过对广州南沙地区软土的物理力学试验获取了土的物理力学性质指标,利用扫描电镜分析和图像处理技术获取了软土的微观结构参数。运用Matlab神经网络工具箱编程,建立了软土工程性质指标与微观结构参数的RBF神经网络模型。通过两个分析模型(模型Ⅰ和模型Ⅱ)的实例研究表明,RBF网络模型具有结构简单,计算速度快,精度高,泛化能力强、性能稳定的优点。该方法可以作为软土宏微观关系建模的有效途径和软土微观结构试验的有效补充,可为软土工程可靠性分析和软基处理设计提供参考依据。
The engineering properties of soft soil depend on microstructure characteristics. Through a large number of physieo-mechanical tests, microstructure analysis and Image processing technology of soft soils in Nansha area, Guangzhou, China, the physico-mechanical indexes and microstructure parameters are obtained. RBF networks models for the relationship between engineering properties and microstrueture parameters of soft soil are established through radial basis function neural networks and Matlab neural network toolbox. Compared with BP neural networks, the empirical results of two models （ model Ⅰ and model Ⅱ） indicate that RBF neural networks have advantages of a simple structure, fast computation, high accuracy and strong generalization ability. This method can be used as a supplementary way for microstructure test of soft soil, and provide an efficient way to quantitative study about relation- ship between macro engineering properties and microstructure of soft soils. Moreover, the method can give a good refer- ence for the reliability analysis of soft soil engineering and ground treatment design.
出处
《地下空间与工程学报》
CSCD
北大核心
2013年第4期777-782,共6页
Chinese Journal of Underground Space and Engineering
基金
获国家自然科学基金(51178122)
广东省自然科学基金(S2011040004133)
广东省大学生创新实验项目(118450088
118450089)
关键词
RBF神经网络
软土
土的工程特性
微结构参数
MATLAB
Radial Basis Function neural networks
soft soil
engineering properties of soil
microstrueture parameters
Matlab
