Data-Driven Framework for Identifying Representative Pore Size in Pervious Concrete through Mix Design Parameters

Abstract Pervious concrete is vital for low impact development, enabling infiltration, flood mitigation, and groundwater recharge. Yet, its performance is difficult to predict due to the complexity of its pore structure. This study numerically characterizes the pore size distribution and investigates its sensitivity to key mix design parameters, including aggregate size (AS), aggregate-to-cement (A/C) ratio, and compaction energy (CE). A total of 75 mix designs were created using three aggregate size ranges and five A/C ratios under five compaction levels. From 450 specimens (75 designs with 6 specimens each), pore characteristics were extracted using high-resolution panoramic imaging, analyzed through ImageJ and MATLAB, and interpreted via two complementary approaches: (1) S-curve-based logistic modeling, and (2) diameter-based distribution metrics ( <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="normal upper D 10" display="inline" overflow="scroll"> <mml:msub> <mml:mi mathvariant="normal">D</mml:mi> <mml:mn>10</mml:mn> </mml:msub> </mml:math> , <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="normal upper D 30" display="inline" overflow="scroll"> <mml:msub> <mml:mi mathvariant="normal">D</mml:mi> <mml:mn>30</mml:mn> </mml:msub> </mml:math> , <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="normal upper D 50" display="inline" overflow="scroll"> <mml:msub> <mml:mi mathvariant="normal">D</mml:mi> <mml:mn>50</mml:mn> </mml:msub> </mml:math> , <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="normal upper D 60" display="inline" overflow="scroll"> <mml:msub> <mml:mi mathvariant="normal">D</mml:mi> <mml:mn>60</mml:mn> </mml:msub> </mml:math> , <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="normal upper D 90" display="inline" overflow="scroll"> <mml:msub> <mml:mi mathvariant="normal">D</mml:mi> <mml:mn>90</mml:mn> </mml:msub> </mml:math> ). Advanced multivariate techniques, including hierarchical classification, nonlinear regression, and MANOVA, were used to identify mix design factors influencing pore structure and to enhance predictive modeling. Results revealed notable discrepancies between porosities derived from water displacement and image analysis, highlighting the importance of integrated characterization. Gaussian process regression, with kernel optimization, demonstrated strong predictive accuracy for pore-size metrics. Incorporating the logistic distribution factor <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="k" display="inline" overflow="scroll"> <mml:mi>k</mml:mi> </mml:math> into the diameter-based framework improved capture of pore variability and enhanced predictive performance. Among all parameters, <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="normal upper D 60" display="inline" overflow="scroll"> <mml:msub> <mml:mi mathvariant="normal">D</mml:mi> <mml:mn>60</mml:mn> </mml:msub> </mml:math> emerged as the most stable and representative indicator of pore structure. Compaction effort exerted the strongest overall influence on <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" alttext="normal upper D 60" display="inline" overflow="scroll"> <mml:msub> <mml:mi mathvariant="normal">D</mml:mi> <mml:mn>60</mml:mn> </mml:msub> </mml:math> , whereas aggregate size played a progressively greater role at coarser gradations. Excessive compaction disrupts the aggregate skeleton, driving paste migration and segregation, which creates uneven pore distribution, enlarges voids, and compromises both structural integrity and permeability. These findings establish a solid framework for optimizing mixed designs and improving performance prediction based on pore characteristics.