Divergent Hydration Kinetics: Type-Aware Artificial Intelligence Framework for Compressive Strength Prediction and Sustainability Evaluation in Concrete Composites | RISE Research

This paper develops a 'type-aware' machine-learning framework for predicting the compressive strength of concrete and evaluating its sustainability, arguing that existing data-driven models wrongly assume homogeneous material behaviour and ignore the compositional differences between conventional and sustainable concrete. By engineering a new 'type' attribute for each concrete instance in its datasets, the study builds models that respect how the intrinsic properties of each concrete class govern microstructural stability and, in turn, strength. A key empirical finding drives the analysis: age correlates positively with compressive strength in conventional concrete but negatively in sustainable concrete, revealing that the two types follow divergent hydration-kinetic pathways. The paper explains this through the role of Supplementary Cementitious Materials (SCMs) and their differing capacity to generate Calcium Silicate Hydrate (C-S-H) gel. It trains and compares Linear Regression and Random Forest models, finding that most strength relationships are non-linear and that Random Forest substantially outperforms the linear baseline, confirming the complex predictor-target interactions in concrete. The work then translates its predictive results into user-friendly graphical interfaces that let engineers estimate compressive strength and assess sustainability, including for carbon-negative materials such as Ferrock and Rice Husk Ash. Its focus is on showing that incorporating material classification into AI-driven modelling improves both predictive accuracy and real-world usability, contributing a practical, scalable framework for sustainable construction analysis. Spanning civil and chemical engineering, chemistry, AI and environmental engineering, the paper connects life-cycle and sustainability concerns to a domain-informed modelling approach intended to be genuinely usable by field practitioners and decision-makers.