土工基础

›› 2019, Vol. 33 ›› Issue (3): 304-307.

• 工程实录 • 上一篇    下一篇

基于GA-BP神经网络的PHC管桩桩侧摩阻力预测方法研究

张 洁1，石 荔2   

  1. （1.宝山钢铁股份有限公司，上海 201900；2.中冶集团武汉勘察研究院有限公司，武汉 430080）
  • 收稿日期:2019-02-25 修回日期:2019-03-13 出版日期:2019-06-20 发布日期:2019-08-16
  • 作者简介:张 洁（1983-），女，硕士，工程师，研究方向为基础工程、结构工程、防灾减灾及防护工程。

A Prediction Method Using GA-BP Neural Network for the Frictional Resistance of PHC Piles

ZHANG Jie1, SHI Li2   

  1. (1.Baoshan Iron & Steel Co., Ltd, Shanghai, 201900;
    2.Wuhan Surveying & Geotechnical Research Institute Co. Ltd., DFMCC, Wuhan 430080)
  • Received:2019-02-25 Revised:2019-03-13 Online:2019-06-20 Published:2019-08-16

摘要： 随着PHC管桩的广泛应用，为能够有效地预测单桩竖向极限承载力，分析了土的物理、力学参数对单桩竖向极限承载力的影响，通过影响因素分析选取贡献率较大的指标。结合现场静载荷试验数据，将遗传算法GA和人工神经网络BP有机地结合起来，利用GA-BP神经网络结合多元统计主分析主分量数学方法，对以粘性土和非粘性土为主的上海某地区的PHC管桩桩侧摩阻力进行了预测和检验，证明其预测精度良好、适用性强，验证了该方法的可行性，以此模型作为辅助控制手段进行合理预测，具有较大的工程实用价值。

关键词: GA-BP神经网络, PHC管桩, 桩侧摩阻力

Abstract: This paper presents a new method, which is called GABP method, that can effectively predict the axial resistance of prestressed highintensity concrete (PHC) piles. Various soil properties, such as, index properties, mechanical properties, on the axial capacity of PHC piles are evaluated and factors that have largest influence on the pile capacity are determined. This new method, which combines the genetic algorithm (GA) method and artificial neural network back propagation (BP) method, applies the statistical method of multi factors in GABP neural network system in the axial capacity predictions of PHC pile in cohesive and noncohesive soil in Shanghai. The results indicated that this method is successful.

Key words: GA-BP Neural Network, PHC Piles, Side Soil Resistance of Pile