Computational Intelligence-Based Prognostication of Autogenous Healing in Engineered Cementitious Composites

This study introduces the latest methodology for autonomous healing capacity forecasting for Engineered Cementitious Composites (ECC) using ‎computational intelligence to enhance the durability and sustainability of concrete structures. Base models Adaptive Boosting Algorithm ‎(ADA) and Gaussian Process Regression ‎(GPR) are adopted, and the Seagull Optimizer (SOA) and the Subtraction-Average-Based Optimizer (SABO) ‎are introduced for the enhancement of their predictive capability. The voting ensemble technique is also ‎‎employed to combine the individual strength points for the enhancement of predictive ‎reliability. ‎The methodology is validated using the experiment data set, where the primary parameters like mineral admixtures and the initial crack width are researched for their impact on the ‎self-healing capability. Results verify the highest predictive capability for the ensemble model (AGSA) using the highest value for the coefficient of ‎determination (R² value: 0.9918), much superior when ‎compared against the individual models and the combination models. Sensitivity analysis using the ‎Shapley Additive ‎Explanations (SHAP) tool verifies the highest impact by the initial crack width (CWB), contributing by far the largest proportion (81.5%) towards the predictive ‎results. This study introduces the hybrid ensemble-learning technique for the self-healing ECC, contributing towards data-driven design for the field of construction engineering for the enhancement of the design and ‎production of stronger concrete materials.