The use of stone columns (SC) and deep cement mixing (DCM) columns to stabilize road subgrades are economical and reliable ground improvement techniques that have been extensively studied. The current study used an adaptive neural fuzzy inference system (ANFIS) hybridized with a particle swarm optimization (PSO) algorithm to develop models for estimating the ultimate bearing capacity (UBC) of soft soil improved with floating SC or DCM columns. A database of 86 physical modelling tests was used to create and train the ANFIS-PSO models. The input parameters used were the undrained shear strength of soft soil, area improvement ratio, and length-to-diameter ratio of the floating SC and DCM columns. The performance of the proposed ANFIS-PSO model has been validated using the testing data and a decision-making model for the design of the geotechnical properties of the floating SC and DCM columns has been proposed. The results show that the accuracy of the proposed ANFIS-PSO model in terms of the performance indices was satisfactory. The R², VAF, and MSE values were obtained for the testing sets of the optimized ANFIS-PSO predictive model. This ANFIS-PSO model can be used by geotechnical engineers for the design of floating DCM and SC columns that increase the UBC of a road subgrade.

使用石柱（SC）和深水泥搅拌（DCM）柱来稳定路基是经济且可靠的地面改良技术，已得到广泛研究。当前的研究使用了一种自适应神经模糊推理系统（ANFIS）和一种粒子群优化（PSO）算法相结合的方法，以开发模型来估算用浮动SC或DCM柱改善的软土的极限承载力（UBC）。使用86个物理建模测试的数据库来创建和训练ANFIS-PSO模型。所使用的输入参数为软土的不排水抗剪强度，面积改良率以及浮动SC和DCM柱的长径比。利用测试数据验证了所提出的ANFIS-PSO模型的性能，并提出了用于设计悬浮SC和DCM柱的岩土特性的决策模型。结果表明，所提出的ANFIS-PSO模型在性能指标上的准确性是令人满意的。的对于优化的ANFIS-PSO预测模型的测试集，获得了R ²，VAF和MSE值。岩土工程师可以使用此ANFIS-PSO模型来设计浮动DCM和SC柱，以增加道路路基的UBC。