基于神经网络的混凝土预制桩单桩竖向极限承载力参数分析  

Parametric Analysis of Vertical Ultimate Bearing Capacity of Precast Concrete Pile Using Artificial Neural Network

作　　者：陈铁冰[1] 

机构地区：[1]深圳职业技术学院建筑与环境工程学院,广东深圳518055

出　　处：《公路交通科技》2007年第11期87-91,共5页Journal of Highway and Transportation Research and Development

基　　金：广东省自然科学基金资助项目(06028131);校重点科技资助项目(06KJd007)

摘　　要：应用BP神经网络,对混凝土预制桩单桩竖向极限承载力进行预测,并分析了各种参数对单桩竖向极限承载力的影响。通过影响因素分析,确定了桩径、桩长、入土深度、桩侧摩阻力加权平均值、桩端阻力平均值等参数对单桩竖向极限承载力有影响。对混凝土预制桩单桩静载试验资料进行分析和取样,将包含上述参数的样本与单桩竖向极限承载力形成数据对,采用三层神经网络进行训练,输入层为各参数,输出层为单桩竖向极限承载力,建立了混凝土预制桩单桩竖向极限承载力预测模型。研究表明,所建立的模型能够有效地预测混凝土预制桩单桩竖向极限承载力,通过参数分析,能够得出各参数对单桩竖向极限承载力的影响规律,从而确定比较合理的单桩设计参数。An approach of predicting vertical ultimate bearing capacity of precast concrete pile is presented using backpropagation （BP） artificial neural network （ANN） . Some parameters are analyzed to see how to influence vertical ultimate beating capacity of precast concrete pile. Through analyzing it is found that five parameters may influence vertical ultimate bearing capacity of precast concrete pile, including pile diameter, pile length, length accessing earth, weighted average frictional resistance at the side of pile and average frictional resistance at the end of pile.The ANN can be trained using a small set of parameters obtained from test data which may influence vertical ultimate bearing capacity of precast concrete pile,Three layers are adopted in ANN, which the input layer includes parameters mentioned above and the output layer is vertical ultimate beating capacity of precast concrete pile. The results show that the accuracy of vertical ultimate beating capacity of precast concrete pile predicted by the presented approach is validated by comparison with test data.The relationship between vertical ultimate beating capacity of precast concrete pile and parameters can be found through parametric analysis, and the reasonable parameters can be determined.

关 键 词：桥梁工程 单桩竖向极限承载力 神经网络 预制混凝土桩 参数分析 

分 类 号：U443.15[建筑科学—桥梁与隧道工程]