论文部分内容阅读
通过对广州市南沙地区大量软土物理力学试验和微结构分析,获取了40组软土试样的物理力学性质指标和微观结构参数。综合运用灰色关联分析的数据分析能力和人工神经网络的非线性映射功能,建立了软土物理力学性质指标与微结构参数的灰色关联-径向基神经网络模型。该模型利用灰色关联分析方法对数据进行预处理,提取重要因子作为网络的输入,而径向基神经网络充分利用样本数据信息,自适应确定隐含层节点个数、径向基函数中心、宽度以及网络的权系数。克服了传统RBF网络隐层节点数为样本个数,当数据较多时导致网络结构庞大、学习速度慢的缺点。通过模型A和模型B的实例研究表明,该方法简化了网络结构,提高了训练速度和预测精度,为软土物理力学性质与微结构参数关系的定量研究提供了一条有效途径。
Through physical and mechanical tests and microstructures analysis of a large amount of soft soil in Nansha district of Guangzhou, physical and mechanical properties indexes and microstructure parameters of 40 groups of soft soil samples were obtained. By combining the data analysis ability of gray relational analysis with the nonlinear mapping function of artificial neural network, a gray relational-radial basis neural network model of indexes of physical and mechanical properties of soft soil and microstructure parameters is established. The model uses gray relational analysis to preprocess the data and extract the important factor as the input of the network. The radial basis neural network makes full use of the sample data and information, and adaptively determines the number of hidden layer nodes, radial basis function center, width As well as the network’s weight coefficient. Overcoming the problem that the number of hidden layer nodes in the traditional RBF network is the number of samples and when the data is large, the structure of the network is huge and the learning speed is slow. The case study of model A and model B shows that this method simplifies the network structure, improves the training speed and prediction accuracy, and provides an effective approach for the quantitative study on the relationship between the physical and mechanical properties of soft soil and the microstructure parameters.