Using machine learning techniques for predicting autogenous shrinkage of concrete incorporating superabsorbent polymers and supplementary cementitious materials

Abstract

Superabsorbent polymers (SAP) are a very effective means of decreasing high-performance and ultra-high performance concrete autogenous shrinkage. However, their efficiency can hardly be predictable because of various parameters: SAP properties, supplementary cementitious materials (SCM) nature, and cement replacement ratios. This study provides a machine learning approach for predicting shrinkage/expansion in cementitious materials incorporating SAP and SCM. A dedicated database is built, and four machine learning models are compared. Extreme Gradient Boosting (XGBoost) model exhibited the highest accuracy. SHapley Additive exPlanations (SHAP) allowed the identification of the most influential inputs, and partial dependence plots provided quantitative information about their relative influence.