低温养护下偏高岭土基地聚物固化土抗剪强度参数预测OA

[Purposes]Traditional indoor test methods for investigating the shear strength parameters of geopolymer-solidified soil under low-temperature curing are challenged by numerous influencing factors,long experimental cycles,and high resource consumption.This study aims to address the prediction of highly nonlinear shear strength parameters of geopolymer-solidified soil in cold regions under multiple influencing factors using machine learning models.[Methods]Based on 180 sets of experimental data,eight machine learning models were built.Four performance metrics were employed to quantitatively evaluate the generalization ability of these models.Furthermore,the influence of input parameters on the shear strength parameters of geopolymer-solidified soil in cold regions was analyzed.[Results]Among the eight machine learning models,the IVY-XGBoost model demonstrates the best predictive performance for the shear strength parameters of geopolymer-solidified soil in cold regions,with correspondingly small relative prediction errors.Parameter sensitivity analysis reveals that curing age is the most important input feature for both cohesion and the internal friction angle.[Conclusions]The IVY-XGBoost model can not only significantly shorten experimental cycles and reduce resource consumption,but also provide a reliable basis for the mix design and performance optimization of geopolymer-solidified soil in cold regions.