基于ACO-SVM的桥梁基础群桩轴力预测  

Prediction on Axial Force of Pile Group in Bridge Foundation Based on ACO-SVM

作　　者：黄伟杰[1] 吴叶[2] 陈志坚[1] 俞俊平[1] 

机构地区：[1]河海大学地球科学与工程学院,南京210098 [2]东南大学材料科学与工程学院,南京211189

出　　处：《长江科学院院报》2016年第1期121-125,共5页Journal of Changjiang River Scientific Research Institute

基　　金：国家"十一五"科技支撑资助项目(2006BAG04B05);国家重点基础研究发展计划(973计划)项目(2002CB412707)

摘　　要：由于大型深水群桩基础受到复杂的环境影响,其基桩轴力的变化与环境因素之间呈现复杂非线性关系。利用在解决小样本、非线性、高维数方面具有很强能力的支持向量机,对苏通大桥群桩基础轴力实测数据进行分析,预测了一段时间内轴力的变化。并采用了蚁群算法(ACO)寻找模型最优参数,由此建立了ACO-SVM模型,避免了人为选择参数的盲目性。为方便对比,建立了传统SVM与RBF神经网络预测模型,对比了ACO-SVM,SVM,RBF这3个模型的预测结果。研究表明,与传统SVM,RBF的预测结果相比,ACO-SVM模型具有更高的可信度和预测精准度,且具有更强的泛化能力,在大型深水群桩基础的轴力预测中具有一定的工程应用价值。As for large-scale pile group foundation with deep water,relationship between axial force of pile shaft and environmental factor is complex and nonlinear due to complex environment. In light of advantages of support vector machine（ SVM） method in solving small sample size,nonlinearity,and high dimension,we use the method to analyze measured data of axial force in pile group foundation of Suzhou-Nantong bridge,and to predict axial force for a period. Then,we look for optimal parameters by using ant colony optimization（ ACO） and establish ACO-SVM model,which can avoid optionally choosing parameters. Meanwhile,we establish prediction models based on traditional SVM and RBF neural network and compare prediction results of the 3 models. The results show that,COSVM model is of high reliability,high accuracy and strong generalization ability,superior to SVM and RBF. Finally,CO-SVM model can be applied to predict axial force in large-scale pile group foundation with deep water.

关 键 词：深水群桩基础 支持向量机 蚁群算法 轴力预测 ACO-SVM模型 

分 类 号：U443.15[建筑科学—桥梁与隧道工程]