Recycled Aggregate Concrete Incorporating GGBS and Polypropylene Fibers Using RSM and Machine Learning Techniques

In this study, Response Surface Methodology (RSM) and machine learning models were used to predict the mechanical properties of recycled aggregate concrete (RAC) containing ground granulated blast furnace slag (GGBS) and polypropylene fibers (PPFs). The investigation focused on compressive strength (CS) and split tensile strength (STS) tests at curing periods of 7, 28, 56, and 90 days, with variations in the percentages of GGBS (0–50%), recycled aggregate (RA) (0–100%), and PPF (0–1%). The RSM model showed high accuracy in predicting both CS and STS, with statistically significant results (p-value < 0.0001). Among the machine learning models, the Gradient Boosting Machine (GBM) exhibited the highest performance, achieving an R2 value of 0.98961 during the training and testing phases for CS prediction. It also demonstrated strong results for STS prediction, with an MSE of 0.02773, MAPE of 2.69775, and R2 value of 0.99404 in the training phase, and an MSE of 0.14141, MAPE of 5.71691, and R2 value of 0.96947 during testing. The Stacked Ensemble Learning model performed similarly to GBM, with an R2 of 0.99251 during training for STS and 0.96619 during testing. However, GBM consistently outperformed the other models in terms of balancing low error rates and high R2 values across both datasets. The Distributed Random Forest model also provided strong performance but slightly higher error rates and lower R2 values than GBM. Overall, both GGBS and PPF significantly enhanced the mechanical properties and workability of the concrete, highlighting the importance of these additives in optimizing concrete performance.