Development of autogenous shrinkage prediction model of alkali-activated slag-fly ash geopolymer based on machine learning

Abstract

This paper developed an autogenous shrinkage prediction tool with high accuracy through machine learning for alkali-activated slag-fly ash geopolymer. The influencing factors of autogenous shrinkage of activated slag-fly ash geopolymer paste and mortar were analyzed. The results show that Extreme Gradient Boosting (XGB) algorithm achieves the best prediction performance with R2 of over 0.90 and strong generalization ability for predicting the autogenous shrinkage of alkali-activated slag-fly ash geopolymer paste and mortar. The decrease in W/B, alkali dosage and slag content can reduce the autogenous shrinkage of alkali-activated slag-fly ash geopolymer paste, while increasing W/B and decreasing alkali dosage are beneficial to mitigate the autogenous shrinkage of alkali-activated slag-fly ash geopolymer mortar. The Graphical User Interface (GUI) used for autogenous shrinkage prediction of alkali-activated slag-fly ash geopolymer paste or mortar was designed, which can directly be used for predicting autogenous shrinkage on the premise of knowing the synthesis parameters of geopolymer. This prediction tool can prejudge the autogenous shrinkage of alkali-activated slag-fly ash geopolymer materials instead of preforming cumbersome autogenous shrinkage test, which will significantly reduce the workload if we only need to know the autogenous shrinkage.