ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING MODELS FOR PREDICTING THE COMPRESSIVE STRENGTH OF CONCRETE: A SYSTEMATIC LITERATURE REVIEW

Abstract

Accurate prediction of concrete compressive strength is critical for ensuring structural safety and optimizing material usage, yet traditional empirical models often fail to capture the complex, nonlinear relationships inherent in concrete behavior. This systematic literature review was therefore designed to synthesize and critically evaluate the growing body of research on artificial intelligence and machine learning models developed for this purpose. We systematically examined peer-reviewed studies that apply supervised learning, ensemble methods, deep learning architectures, and hybrid models to predict compressive strength from diverse input features. The methodology involved a structured search and a rigorous screening process to identify relevant articles, followed by thematic analysis across eight proposed dimensions, including concrete material types, algorithm selection, explainable AI integration, non-destructive test data fusion, environmental curing effects, early-age prediction strategies, and optimization via metaheuristics. Our results reveal that ensemble trees and deep neural networks consistently achieve the highest predictive accuracy, particularly when combined with feature engineering and metaheuristic tuning, while hybrid models that incorporate experimental data and environmental factors further improve generalization. However, we found that many studies still lack interpretability assessments, and the influence of curing conditions and real-time monitoring remains underexplored. Consequently, this review concludes that although machine learning offers substantial promise for replacing or augmenting traditional testing, future research must prioritize model transparency, standardization of datasets, and integration of non-destructive testing modalities to enable practical deployment. By mapping current trends and gaps, this work provides a foundation for developing more robust and interpretable predictive frameworks in concrete technology.