Multiple AI predictive models for compressive strength of recycled aggregate concrete

To address the growing concerns about the environmental impact and construction costs, there has been an increasing interest in the use of recycled aggregates in concrete applications. Among the mechanical properties of concrete, compressive strength (fc) is particularly significant. This study explored the estimation of the compressive strength of recycled aggregate concrete using various machine-learning techniques. In this study, 'Genetic Programming' (GP), 'artificial neural networks' (ANN), and 'Evolutionary Polynomial Regression' (EPR) were employed to predict the 28-day compressive strength of recycled aggregate concrete. The considered predictive inputs encompass a range of factors, including cement, fine aggregate, recycled fine aggregate, coarse aggregate, recycled course aggregate, water, water-cement ratio, and superplasticizers, which produced 476 data entries. Among the models developed, the hybrid ANN-based model demonstrated superior performance compared with the other models. A rigorous assessment of the model performance was conducted through diverse statistical calculations, such as spearman correlation and internal consistency, relative importance of input parameters, sum of squared error (SSE) and the coefficient of determination designated as R-squared (R-2). To reinforce the evaluation, a Taylor diagram and marginal histogram were employed as assessment parameters. Considering the statistical error analysis, Taylor diagram, and marginal histogram, the ANN-hybrid model was capable of accurately estimating the compressive strength (fc) of recycled aggregate concrete. The adopted machine learning models have the potential to conserve material resources and reduce the technical labor involved in determining the compressive strength of recycled aggregates in concrete applications.