寒区偏高岭土基地聚合物固化土抗压强度预测

Prediction of Compressive Strength of Soil Solidified by Geopolymer Based Metakaolin in Cold Area

摘要:

为了探究季冻区电石渣激发偏高岭土基地聚合物固化土抗压强度变化规律及配比设计，通过180组试验数据构建数据库，以偏高岭土含量、电石渣含量、养护温度和养护时间为输入参数，固化土抗压强度为输出目标，利用机器学习方法建立输入参数与输出目标之间的回归关系。选用多种机器学习算法（KRR、ANN和GPR）建立电石渣激发偏高岭土基地聚合物固化土抗压强度预测模型，通过决定系数（R2）、均方根误差（RMSE）和平均绝对误差（MAE）对模型性能进行评估，并结合敏感性分析解释预测结果。研究结果表明，GPR模型的预测误差相对较小，具有较好的预测精度，可较好地预测电石渣激发偏高岭土基地聚合物固化土的抗压强度；敏感性分析发现，养护温度和养护时间对抗压强度的影响显著。相关研究结果可为电石渣激发偏高岭土基地聚合物在季节冻土工程性质改良中的应用设计提供参考依据。

Abstract:

In order to investigate the change rule and ratio design of compressive strength of polymer solidified soil at the base of calcium carbide slag-activated metakaolin in the seasonal frozen soil area, this study constructed a database using 180 sets of experimental data, including metakaolin content, calcium carbide slag content, curing temperature, curing time, andthe compressive strength of the solidified soil. The regression relationship between the input parameters and the output target (the compressive strength of the solidified soil) was established by various machine learning algorithms (KRR, ANN, and GPR) to predict the compressive strength of polymer solidified soil with calcium carbide-activated metakaolin base, and the performance of the model was evaluated by the coefficient of determination (R2), root-mean-square error (RMSE) and mean absolute error (MAE). Finally, the prediction results were interpreted by the parameters sensitivity analysis. The results show that the GPR model exhibits lesser prediction error and better prediction accuracy, and can better predict the compressive strength of polymer-solidified soils with calcium carbide-activated metakaolin base. The sensitivity analysis reveals that the effects of the curing temperature and curing time on the compressive strength are significant. The results of this study can provide a reference basis for the application design of the calcium carbide-activated metakaolin base polymer in the improvement of seasonal frozen soil engineering properties.