Predicting compressive strength of pervious concrete with fly ash: a machine learning approach and analysis of fly ash compositional influence

Pervious concrete offers an eco-friendly solution for urban drainage by allowing rainwater to infiltrate the ground. However, designing strong yet permeable mixes remains a hurdle. This study investigates machine learning's potential to predict the compressive strength of fly ash-blended pervious concrete. Both mix design parameters (aggregate-to-binder ratio, fly ash-to-binder ratio, water-to-binder ratio and curing period) and the primary chemical composition of the fly ash (key oxides) are considered to predict the compressive strength. The Artificial Neural Network (ANN) emerged as the superior model, achieving high accuracy (RMSE similar to 2.33 MPa) in predicting strength. Notably, the ANN performed exceptionally well using only critical factors like curing time, water-to-binder ratio, and the content of specific fly ash oxides (CaO and SiO2). This approach significantly reduces the complexity of data acquisition while maintaining strong predictive power. By focusing on these crucial features, the model allows for higher fly ash replacement levels, leading to a more sustainable and cost-effective construction material with improved environmental benefits.