A genetic algorithm (GA) was used in this study to develop a standard penetration test (SPT)-based design method for the axial capacity of driven piles. A total of 72 pile load tests was collected from literature and divided into two groups based on their measurements. The first group had the load-transfer distribution measurements for extracting both the unit side and tip resistances. These unit resistances were correlated by the GA with soil measurements and pile properties to develop the design method. The second group, where only the total capacity measurements were available, were used to validate the new design method and compare its performance with three existing SPT-based design methods. The new GA-derived design method considers nonlinear relationships with the effective stress and pile length and provides an unbiased prediction with a low coefficient of variation (COV) of 40.0 %, while the three existing methods overestimate the capacity by a factor of 1.62 to 1.65 with a high COV of 40.3 % to 52.8 %, which could result in an under design of pile foundations. This study shows that the GA was able to obtain complex relationships with great accuracy and the new design method can be applied to new cases reasonably well.

本研究使用遗传算法 (GA) 开发了一种基于标准贯入试验 (SPT) 的打入桩轴向承载力设计方法。从文献中收集了总共 72 个桩荷载试验，并根据它们的测量结果分为两组。第一组进行了负载转移分布测量，用于提取单元侧和尖端电阻。GA 将这些单位电阻与土壤测量值和桩特性相关联，以开发设计方法。第二组只有总容量测量可用，用于验证新设计方法并将其性能与三种现有的基于 SPT 的设计方法进行比较。新的 GA 派生设计方法考虑了与有效应力和桩长的非线性关系，并提供了具有 40.0% 低变异系数 (COV) 的无偏预测，而现有的三种方法将容量高估了 1.62 到 1.65 倍COV 高达 40.3 % 至 52.8 %，这可能导致桩基础设计不足。这项研究表明，遗传算法能够非常准确地获得复杂的关系，并且新的设计方法可以很好地应用于新案例。