Although the influence of gradation on soil properties has been recognized, the grain sizes have not yet been fully considered in maximum dry density (MDD) evaluation and prediction. This study explores the impacts of full grain sizes on the MDD and evaluates MDDs through different particle size distribution (PSD) curves of gravelly soils. First, full grain sizes, d₁₀–d_100, were extracted from 90 gravelly soil samples and employed as input parameters to develop the neural model for MDD, by using genetic algorithm (GA) to optimize the back-propagation (BP) neural network. A mean impact value (MIV) method was then proposed to quantify the impact of each grain size, followed by the vibrating compaction tests for 22 artificially designated gravelly soil specimens to verify the model and evaluate the MDDs based on grain sizes. The model analysis and verification tests agreed with each other and clearly showed the intrinsic dependence of grain sizes on MDD for gravelly soils. As revealed by MIV analysis, d₅₀–d₁₀₀ and d₁₀–d₄₀ displayed positive and negative impact to MDD during compaction, behaving as the relatively coarse and fine grain sizes, respectively. Additionally, the relative impact weight showed that d₁₀₀ tends to have the largest impact to MDD, and high-sensitivity (HS), medium sensitivity (MS), and low sensitivity (LS) could be proposed to distinguish these grain sizes. As the MDD was found to correspond exactly to the unique soil gradation, the full grain sizes should be employed to precisely predict and correctly evaluate the MDD of gravelly soils.

Pervious concrete (PC) pavement has been widely accepted as a green infrastructure but is prone to clogging. This study comparatively investigated sand and clay clogging mechanisms of PC and vertical sediment distributions of sand-clogged and clay-clogged PCs. Clay and three sizes of sand were used to clog PC under two exposure methods (no drying and drying). X-ray computed tomography (CT) was used to scan the clogged samples before and after 30 pressure washing cycles. The clogged permeability and permeability after each washing cycle were measured. The clogging patterns of sand depend mainly on sand particle sizes relative to PC pore sizes. The applied fine sand, coarse sand, and medium sand cause easy-passing clogging, surface clogging, and full-depth clogging, respectively. After clay clogging, more than 77% of the total retained clay occurs within the vertical region 24–72 mm below the sample surface; the most clogging (the lowest-permeability layer) occurs at a depth of approximately 48 mm. The dried clay retained within the region 40–120 mm below the surface (especially within the lowest-permeability layer) is hard to wash away because the drying process increases the cohesion of internal clay particles and clay adhesion to the rough, tortuous pore wall of PC. The clogged normalized permeability of 0.154 and permeability recovery ratio of 4.91% in dried clay-clogged samples are lowest among all the samples. However, pressure washing readily washes away the retained undried clay. Accordingly, it is recommended that pressure washing is used to eliminate the clogging effect of dried clay before hot, sunny exposure conditions dry the retained clay. This study provides evidence for developing effective pavement maintenance strategies.