为系统研究风积砂地层中螺旋锚的竖向抗拔承载特性并建立相应的承载力预测模型，本文综合采用原位试验、数值模拟与机器学习方法，对原位试验与数值模拟结果进行对比验证，深入揭示螺旋锚的抗拔承载机制，并系统分析了锚盘直径、间距、埋深等关键参数对其承载特性的影响规律。进一步，以数值模拟数据为基础构建机器学习训练集，识别影响承载力的关键参数，并以此作为变量，拟合上拔承载力系数 N_qu 与土侧压力系数 K_u 之间的经验公式。研究结果表明：螺旋锚抗拔承载力随位移变化依次经历线性增长、非线性强化及近似线性渐进3个阶段；承载力随锚盘直径增大呈指数型增长，而与锚盘数量的相关性较弱；确定了锚盘深埋与浅埋的临界值为5D（D为盘径），揭示了多盘螺旋锚在锚盘间距为4D~5D存在整体圆柱剪切破坏与独立破坏模式的临界条件；提出了适用于风积砂地层中螺旋锚上拔承载力系数与土侧压力系数的理论计算公式，经现场试验验证，预测误差控制在5%以内，显示出良好的预测精度与稳定性。本研究可为砂土地区螺旋锚基础的理论完善与工程应用提供重要技术依据。

To systematically investigate the vertical uplift capacity of helical anchors in aeolian sand and establish a predictive model, this study employed an integrated approach combining in-situ tests, numerical simulation, and machine learning. The results of in-situ tests and numerical simulations were compared and validated, deeply revealing the uplift bearing mechanism of helical anchors and systematically analyzing the influence of key parameters such as helix diameter, spacing, and embedment depth on the bearing behavior. Furthermore, a dataset generated from numerical simulations was used to train machine learning models, identifying the most influential parameters on the uplift capacity. These parameters were then used as variables to formulate a predictive equation by fitting the relationship between the uplift capacity factor (N_qu) and the lateral earth pressure coefficient (K_u). The results indicate that the uplift load-displacement response undergoes a transition from linear to nonlinear, and then to a near-linear hardening phase. The uplift capacity increases exponentially with the helix diameter but shows a limited correlation with the number of helices. The critical value for deep and shallow embedment of helix anchors was determined to be 5D （D is plate diameter）, revealing that the transition between cylindrical shear failure and individual plate failure in multi-helix anchors occurs at a helix spacing of 4D~5D. A theoretical formula for calculating N_qu in relation to K_u for helical anchors in aeolian sand was proposed. Validation against field test results showed a deviation within 5%, demonstrating high prediction accuracy and stability. The findings of this research provide significant technical support for the refinement of design theories and the practical application of helical anchor foundations in sandy soil regions.