AI-powered optimization of engineered cementitious composites properties and CO₂ emissions for sustainable construction

Machine learning (ML) approaches are ideal for meeting the complex requirements of structural elements, as they have shown tremendous potential for improving material performance and promoting sustainability in the building industry. Recently, ML techniques have become cost-effective and efficient for predicting how material properties influence the proposed mixture characteristics, cost, and time. ML approaches in civil engineering should be further developed to optimize the incorporation of fly ash and slag into mixtures. In this study, machine intelligence algorithms, including an adaptive neuro-fuzzy inference system (ANFIS), a hybrid genetic algorithm-artificial neural network (GA-ANN), and gene expression programming (GEP), were used to predict the tensile strain capacity (TSC) and tensile strength (TS) of engineered cementitious composites (ECCs). The dataset comprised samples for TSC and TS with 11 parameters. To assess feature relevance, a sensitivity analysis (SA) and a complete parametric analysis using the best-performing model were conducted. The results demonstrated that GEP achieved the highest prediction accuracy for TS and TSC, with R 2 values of 0.93 and 0.94 for training and 0.91 and 0.92 for testing, respectively. The models' performance was evaluated using MSE, RMSE, and MARE values for tensile strength and strain capacity during both training and testing phases, with GEP emerging as the best model, followed by ANFIS and GA-ANN. Sensitivity analysis was performed to determine the contribution of input variables. For CO₂ emission prediction, the GEP model achieved the highest accuracy with R 2 at 0.97, followed by ANFIS and GA-ANN, with ANFIS having the lowest accuracy with R 2 at 0.95. Additionally, a practical graphical user interface (GUI) tool was developed to apply the study's findings, supporting sustainability efforts in the construction industry. • Machine learning (ML) models optimize ECC properties, enhancing sustainability in construction. • GEP achieves the highest accuracy in predicting ECC tensile properties and CO₂ emissions. • CO₂ emissions predicted with R²= 0.97, advancing sustainable marine infrastructure. • ML models reveal critical input factors driving ECC performance and sustainability. • A user-friendly graphical user interface (GUI) supports efficient ECC design.