Efficacy of sustainable cementitious materials on concrete porosity for enhancing the durability of building materials

The degradation of concrete structures is significantly influenced by water penetration since water serves as the primary vehicle for the movement of harmful compounds. The process of capillary water absorption is widely recognized as a crucial indicator of durability for unsaturated concrete, as it allows dangerous substances to enter the composite material. The water absorption capacity of concrete is intricately linked to its pore structure, as concrete is inherently porous. The main goal of this work is to create an innovative predictive tool that assesses the porosity of concrete by analyzing its components using a machine-learning (ML) framework. Seven distinct batch design variables were included in the generated database: fly ash, superplasticizer, water-to-binder ratio, curing time, ground granulated blast furnace slag, binder, and coarse-to-fine aggregate ratio. Four distant ML algorithms, including AdaBoost, linear regression (LR), decision tree (DT), and support vector machine (SVM), are utilized to infer the generalization capabilities of ML algorithms to estimate concrete porosity accurately. The RReliefF algorithm was implemented to calculate the significant features influencing porosity. This study concludes that in comparison to the alternative techniques, the AdaBoost method demonstrated superior performance with an R 2 score of 0.914, followed by SVM (0.870), DT (0.838), and LR (0.763). The results of the evaluation of RReliefF indicated that the binder possesses a remarkable influence on the porosity of concrete.