Integrating compaction energy into predictive models for pervious concrete permeability

Pervious concrete is an innovative material that effectively manages stormwater runoff. This study aims to develop predictive models for the permeability of pervious concrete, with a particular emphasis on the role of compaction energy. A comprehensive modelling approach was adopted, incorporating Linear Regression, Two-Factor Interaction model, Quadratic model, and Artificial Neural Network to examine the relationships between permeability and key mix design variables. Experimental data were collected by varying aggregate size, aggregate-to-cement ratio, and compaction energy, while maintaining a constant water-to-cement ratio to ensure zero slump. Results show that compaction energy has a pronounced effect on permeability, with reductions exceeding 90% after 448.8 J of compaction. Among the models tested, the Artificial Neural Network demonstrated superior predictive capability, achieving a high coefficient of determination and low root mean square error, with predictions of 2.58 mm/s for the test data set and 2.0 mm/s for the training data set following hyperparameter tuning. These outcomes highlight the importance of incorporating compaction energy into permeability prediction models and demonstrate the potential of machine learning to capture complex parameter interactions. The findings provide a reliable framework for material design, enabling engineers and practitioners to optimise Pervious concrete performance and contribute to sustainable urban infrastructure development.