Machine Learning Model Accurately Predicts Strength of Soil Stabilized with Geopolymer
-
The study investigates using boosting-based machine learning models like gradient boosting (GB) and adaptive boosting (AdaBoost) to predict the unconfined compressive strength (UCS) of clayey soil stabilized with geopolymer.
-
A database of 270 samples was used with 8 input variables related to properties of the soil, geopolymer binder and treatment conditions. The GB model showed highest accuracy in predicting UCS compared to AdaBoost and previous models.
-
The GB model had R2 of 0.980, RMSE of 0.969 MPa for the testing data. The ground granulated blast furnace slag content of the geopolymer binder was found to have the strongest correlation with UCS.
-
The developed GB model demonstrates potential for rapidly and reliably predicting strength of geopolymer stabilized soils, enabling wider application in civil engineering projects involving soil improvement.
-
Further research can explore using larger datasets and more advanced machine learning approaches like deep learning to improve generalization capability of the models across wider range of geopolymer-treated soils.