基于 PSO-BP神经网络的既有桩基极限承载力预测研究

Prediction of Ultimate Bearing Capacity of Existing Pile Foundations Based on PSO-BP Neural Network

摘要

针对既有桩基再利用中极限承载力评估困难的问题，本文通过13根既有桩基现场静载试验与数值模拟方法，对既有桩基极限承载力展开了分析。在此基础上，基于230组数据样本，采用BP、PSO-BP神经网络对单桩极限承载力进行预测，并通过决定系数（ R ²）、均方根误差（MAE）和平均绝对误差（RMSE）三个指标评价了预测结果。研究结果表明：9根破坏性试验桩的极限承载力均为桩基再利用设计值的2~3倍，安全储备较为充足。4根非破坏性试验桩回弹率均超过80%，结合数值模拟确定了其单桩极限承载力均大于桩基再利用设计值，验证了其再利用的可行性。预测模型对比分析，PSO-BP模型的决定系数较传统BP模型提升了196%，平均绝对误差与均方根误差分别降低了66%和62%，预测误差多控制在±2 000 kN以内。研究成果为既有桩基承载力高效评估与再利用提供了科学依据。

Abstract

Aiming at the difficulty in evaluating the ultimate bearing capacity of existing pile foundations for reuse, this paper analyzes the ultimate bearing capacity of existing pile foundations through on-site static load tests and numerical simulations involving 13 existing pile foundations. Based on 230 data samples, this study predicts the ultimate bearing capacity of single piles using backpropagation (BP) and particle swarm optimization-BP (PSO-BP) neural networks, and then evaluates the prediction results using three metrics: coefficient of determination ( R ²), root mean square error (MAE), and root mean absolute error (RMSE). The results demonstrate that the ultimate bearing capacities of all 9 destructively tested piles is 2 to 3 times the reuse design value for pile foundation, indicating a relatively sufficient safety margin. Furthermore, the rebound rate of the 4 non-destructively tested piles exceeds 80%, and numerical simulations confirm that the ultimate bearing capacity of each single pile is greater than the reuse design value, verifying the feasibility of their reutilization. A comparative analysis of the prediction models reveals that the coefficient of determination of the PSO-BP model increases by 196% compared to the traditional BP model, while the MAE and root mean square error RMSE decrease by 66% and 62%, respectively. The prediction errors are mostly controlled within ±2 000 kN. The research findings provide a scientific basis for the efficient evaluation of bearing capacity and reuse of existing pile foundations.

关键词

既有桩基 / 极限承载力 / 承载力预测 / PSO-BP神经网络

Key words

Existing pile foundation / ultimate bearing capacity / bearing capacity prediction / PSO-BP neural network

参考文献

[1] 李镜培, 徐子涵. 长期受荷桩极限承载力的预测[J]. 上海交通大学学报, 2021, 55(04): 380-386.

[2] 罗红星, 王兴国, 宋帅. 桥梁桩基极限承载力预测方法比较[J]. 水利与建筑工程学报, 2016, 14(04): 127-134.

[3] 张火兵. 静载试验在建筑桩基承载力检测中的应用探究[J]. 建设机械技术与管理, 2024, 37(06): 91-93.

[4] 宋士杰, 段志强, 郭伟, 等. 分层土中超长钻孔灌注桩侧摩阻力预测方法讨论[J]. 中国水利水电科学研究院学报(中英文), 2025, 23(04): 393-400+ 411.

[5] 周宇, 尚稳齐, 狄生奎, 等. 连续梁桥影响线识别与承载能力快速评估试验研究[J]. 振动与冲击, 2024, 43(07): 334-344.

[6] Thai P B, Nguyen D D, Thi Q A B, et al. Estimation of load-bearing capacity of bored piles using machine learning models[J]. Vietnam Journal of Earth Sciences, 2022, 44(4): 470-480.

[7] Reich Y . Machine learning techniques for civil engineering problems[J]. Computer-Aided Civil and Infrastructure Engineering, 1997, 12(4): 295-310.

[8] Xie C, Wang J, Wu Y H S . Application of deep learning in civil engineering: boosting algorithms for predicting strength of concrete[J]. Journal of Intelligent & Fuzzy Systems: Applications in Engineering and Technology, 2023, 45(5): 9109-9122.

[9] 郝小员, 刘汉龙, 高玉峰. 人工神经网络在水泥喷粉桩承载力设计计算中的应用[J]. 河海大学学报(自然科学版), 2002(04): 32-36.

[10] 王建华. 神经网络法预估水泥搅拌单桩沉降[J]. 土木工程学报, 1996(01): 55-61.

[11] 李刚. 基于神经网络的单桩竖向承载力预测[D].西安:西安建筑科技大学, 2003.

[12] Lee I M, Lee J H . Prediction of pile bearing capacity using artificial neural networks[J]. Computers and Geotechnics, 1996, 18(3): 189-200.

[13] Millán M A, Picardo A, Galindo R . Application of artificial neural networks for predicting the bearing capacity of the tip of a pile embedded in a rock mass[J]. Engineering Applications of Artificial Intelligence, 2023, 124: 106568.

[14] 李云峰, 闫思行, 冉斌斌, 等. 基于遗传算法改进BP神经网络的桩基横向承载力预测[J]. 科技创新与应用, 2024, 14(33): 30-33.

[15] 徐华荣, 刘刚. 基于修正的BP神经网络算法对桩基承载力预测方法研究[J]. 城市勘测, 2018(01): 163-167.

[16] 杨博, 张鏖, 苏荣华, 等. 基于遗传算法优化BP神经网络圆柱壳结构可靠度分析[J]. 工程力学, 2025, 42(S1): 16-22.

[17] Kerem A, Saygin A . Scenario-based wind speed estimation using a new hybrid metaheuristic model: Particle swarm optimization and radial movement optimization[J]. Measurement and Control, 2019, 52(5-6): 493-508.

[18] Moayedi H, Moatamediyan A, Nguyen H, et al. Prediction of ultimate bearing capacity through various novel evolutionary and neural network models[J]. Engineering with Computers, 2020, 36(2): 671-687.

[19] 浙江省交通工程管理中心. 《公路工程基桩检测技术规程》(JTG 3363-2019)[S].北京: 人民交通出版社, 2020.

[20] 陈似华. 大直径钻孔灌注桩载荷试验与承载性能分析[D].长沙:中南大学, 2009.

[21] 王俊林, 王博, 高有斌. 钻孔灌注桩荷载传递机理试验研究[J]. 河海大学学报(自然科学版), 2007(04): 430-433.

[22] 陈强华, 陈国铨, 谢汝彬, 等. 进入持力层不同深度对单桩承载力的影响[J]. 岩土工程学报, 1981(01): 16-27.

[23] 姬宇杰. 基于BP神经网络和IRMO算法的单桩竖向极限承载力预测研究[D].长沙:中南大学, 2023.

[24] 丁何杰. 高速公路改扩建既有桥梁桩基注浆加固机理及工程应用[D].泰安:山东农业大学, 2021.

[25] 冯世进, 柯瀚, 陈云敏, 等. 黄土地基中超长钻孔灌注桩承载性状试验研究[J]. 岩土工程学报, 2004(01): 110-114.

[26] 李沅朔. 后压浆钻孔灌注桩竖向承载性能试验研究[D].郑州:郑州大学, 2022.