The aggregate sizes used in this study included: 19 to 13.2 mm, 13.2 to 9.5 mm, 9.5 to 6.7 mm, 6.7 to 4.75 mm, and 4.75 to 2.36 mm. The physical properties of the aggregates are shown in Table 1. The aggregate size combinations for the top and bottom layers were selected on the basis of AutoCAD drawings such that there exists continuity of pores from the top to bottom layer. The layer combinations, as shown in Table 2, possessing a minimum porosity of 20% at the interface, were considered for the experimental study.

Fig. (2). Research outline.

Fig. (3). Novel Two Layered Pervious Concrete (2L – PC).

The ordinary Portland cement of Grade 43 was used as the binding agent. The polycarboxylic ether (PCE) based superplasticizer was used as an admixture to improve the workability of the mixture.

The pervious concrete mix design was carried out in accordance with the National Ready Mix Concrete Association (NRMCA) guidelines [31] but separately followed for the top and bottom layers. The samples were designed for the target design air voids of 25% considering w/c ratio of 0.30 and 0.33 for the top and bottom layers, respectively. The mix proportions are indicated in Table 3.

In order to prepare the specimens of cylindrical geometry (height 200 mm and diameter 100 mm), firstly, the cement and aggregates were mixed thoroughly to form a dry mix. Then the water containing the superplasticizer was added in small amounts as the mixing continued. The fresh mix was immediately poured into the cylindrical mold as a single layer occupying 170 mm (decided on a trial basis) and then compacted by giving 25 blows with a standard proctor hammer (2.5 kg). This allowed the mix to get compacted and attain the required layer thickness of 150 mm. The same procedure was adopted for placing the top layer as well. It was taken care that the time difference between the placing of two layers did not exceed 10 minutes. The specimens were demolded after two days of casting to ensure the proper setting and bonding. All the specimens were then water cured for seven days. The process followed is shown in Fig. (4).

Fig. (4). Specimen preparation.

The basic properties of PC specimen, namely density and porosity, were determined based on the buoyancy principle using the ASTM C1754-12 method [32]. The seven-day-cured specimens were shifted to the water tank and kept immersed for 12 hours. After 12 hours, the specimens were then continuously agitated inside the water vessel for a time period of 20 minutes to remove air bubbles. After verifying that there were no air bubbles, the specimen's submerged mass (M_sub) was recorded inside the water bath. Then the specimens were allowed to air dry 24 hours in order to note down the air dry mass (M_air) followed by oven drying for 5 days at 50 ^oC to record the constant oven-dry mass(M_dry). The effective porosity (φ_eff) and total porosity (φ_Total) were then calculated by using Equation (1) and Equation (2), respectively.

(1)

(2)

Where,

V_b = bulk volume of the specimen, m³

ρ = density of water, kg/m³

M_air = air dry mass of specimen, kg

M_{dry =} oven-dry mass of specimen, kg

M_sub = submerged mass of specimen, kg

The analysis of pore features at the interface of two-layered pervious concrete specimens was carried out by using 2D- scanned images. The cylindrical specimens were cut exactly at the interface of the top and bottom layers and sectioned into two slices, as shown in Fig. (5). On the obtained cut surfaces, in order to make the thresholding easy, the white painting was applied on the cement-aggregate matrix and then scanned using a flatbed scanner of resolution 600 dpi. The cut sections before and after the white painting are shown in Fig. (6).

Fig. (5). Sectioning of 2L-PC specimens.

Fig. (6). Cut surface before and after white painting.

The processing of acquired images was carried out using a freely available image processing tool called “ImageJ”. The images in color format (RGB) were first converted into Greyscale. The median filtering technique (radius = 5 pixels) was used to smoothen the paintbrush marks. The filtered images were converted to binary following the thresholding process to restrict the voids and solids in the matrix. In this study, the voids and solid matrix are shown with black and white color, respectively. The entire process is represented in Fig. (7).

As measurement in the images is expressed in terms of pixels, a length scale was set for the processed images by indicating the known diameter of the sample, i.e., 100 mm on the image. This ensured the calibration of the image, and all the measured parameters were expressed in millimeters. The pore parameters, such as porosity in the plane of the image, pore size in terms of area, and equivalent size of the pore assuming circular shape were calculated for each image. Although, the pore shapes are different from that of a circle, an equivalent radius can be calculated for every pore, assuming them to be circular in shape. Similar procedure was adopted for each of the sample, and pore area measurements were obtained considering both the top and bottom images.

The unconfined compressive strength (UCS) of the 7-day cured pervious concrete specimens were tested in a universal testing machine of 600 kN capacity, which is operated in a stress-controlled mode. The rate of loading was set at 0.25 MPa/s confirmed to ASTM C39-15 [15]. Cardboard of sufficient thickness was placed in between the loading surface and the specimen in order to overcome the limitations associated with surface preparation for uniform stress distribution. The test setup is shown in Fig. (8).

Fig. (7). Steps followed in image analysis.

Fig. (8). Compression test setup.

The abrasion resistance of the two-layered PC and conventional PC mixtures was determined using Cantabro loss Test. Three oven-dried samples of each gradation were placed in the Los Angeles abrasion testing machine and were run for 500 revolutions as per ASTM C1747 [23], as shown in Fig. (9).

Fig. (9). Cantabro test for abrasion resistance.

Due to the continuous impact, the aggregate particles get loosened, and the cylindrical specimens deteriorate as in pavement surface, which is subjected to vehicular loads. The abraded specimens were then passed through 1 inch (25 mm) sieve, and the weight of material retained on the sieve was noted down. The Cantabro abrasion loss was calculated using Equation (3).

(3)

Where,

W₁ = Initial Oven Dry weight of Specimens, kg

W₂ = Weight of specimens retained on the 25 mm sieve, kg

It is observed from the previous literature that the porosity controls various other fundamental and functional properties of PC and thereby the material response. The porosity is said to vary from one mixture to other and is highly dependent on the aggregate gradation and cement paste volume. Therefore, the main objective of this study was to understand the pore parameters like the size, shape, and distribution at the interface of the top and bottom layer as there is a high chance of blocking the path for water flow by the aggregates of the upper layer due to the size difference. The results from the image analysis indicated the presence of required pores at the interface in order to maintain functionality. The porosity at the interface obtained from the image analysis was also found to be in an excellent match with the average interface porosity predicted using the 2D-AutoCAD model for the same specimens (Fig. 10). It is also evident from Fig. (11 and 12), that there exists a difference in the pore size, area, and distribution at the interface of 2L-PC specimens. As seen in Fig. (12), the top layer shows slightly lower porosity except for B6, in which the total pore area at the top was found to be lower than the bottom. Lower porosity in the top layer may be attributed to the fact that this layer is made up of finer-sized aggregate and also the compaction energy imparted is higher for a given volume compared to the bottom layer. Further, higher porosity in the top layer for the B6 mixture can be attributed to the resistance given by 2.36 mm aggregate size for compaction due to finer size [25].

Fig. (10). Comparison of interface porosity results.

Fig. (11). Pore features at the Interface of Sample – 5.

Fig. (12). Pore area at the interface.

The equivalent diameter of pores were extracted from the image analysis. It varied in the range 0.050 mm to 15.53 mm in the top layers and 0.050 mm to 28.92 mm in the bottom layers. In order to understand the pore size distribution, statistical analysis was carried out to identify the distribution which would describe the distribution of pores. The well-known distributions such as normal, Gamma, Weibull, and lognormal distribution was fit to the pore sizes. The Anderson-Darling statistic (A-D) was used to check the hypothesis that the data followed a specific distribution. According to Anderson-Darling, the lower the value of A-D statistic, the better the fit. From the plots shown in Fig. (13), it is clear that the representative dataset of Sample B-5 bottom fits into 3 – Parameter Weibull distribution at a level of confidence of 95%. The observations were similar for the rest of the samples as well. In addition to this, the previous studies made by Deo and Neithalath (2010) [21] also proved that the pore size distribution in pervious concrete confirms to 2 – Parameter Weibull distribution. However, since there will be a minimum pore size, the three-parameter Weibull distribution will describe the distribution more accurately, with its location parameter, in addition to the scale and shape parameters [3, 16].

In order to understand the nature of the distribution, the pore sizes of both top and bottom layers of all the samples were modelled using the 3-Parameter Weibull distribution. The related probability distribution function (PDF) and cumulative density functions (CDF) are given in Equations 4 and 5, respectively.

(4)

(5)

For f(d) >0, d >0, β >0 and η >0. Here, d represents the pore diameter, β is the shape parameter, which defines the slope of the density curve, η is the scale parameter, which shows the flatness/ spread of the curve between any given interval and γ is the location parameter, which indicates the minimum value of the data points. The PDF and CDF curves for the pore distribution are shown in Fig. (14 and 15), respectively.

Fig. (13). Distribution Fitting for obtained Pore sizes.

Fig. (14a). Probability distribution plot for 2L-PC top layers.

Fig. (14b). Probability distribution plot for 2L-PC bottom layers.

Fig. (15a). Weibull Cumulative Distribution Plot for 2L-PC Top layers.

Fig. (15b). Weibull Cumulative Distribution Plot for 2L-PC Bottom layers.

Fig. (15a and 15b) show the cumulative distribution plots for the top and bottom layers of different samples, which indicate the probability of occurrence of particular pore size. In the layers made with aggregates of larger sizes (9.5 mm and 6.75 mm), the concentration of small pore sizes is relatively low, when compared to that of layers made with aggregates of small sizes (2.36 mm and 4.75 mm). The results of statistical pore distribution will be extremely helpful in simulating flow through pervious concrete using computational methods such as discrete element methods (DEM) and computational fluid dynamics (CFD).

The A-D statistics and the Weibull parameters obtained for the pore size datasets are listed in Table 4. The variation of these parameters with the aggregate size is presented in Fig. (16).

It may be observed from Fig. (16) that as the pore size increases, the shape parameter decreases, whereas the scale parameter increases with the increase in the pore size. This clearly indicates that in the case of the coarser aggregate PC mixtures (bottom layer), the distribution of the pores is heterogeneous, whereas in case of finer aggregate PC mixtures (top layer), it is rather homogeneous. As the scale parameter increases, there will be a tendency for deviation from the Darcy’s law as the pores are distributed heterogeneously, leading to increased velocity head [7]. Therefore this observation has a significant impact on the flow of water through the pervious concrete.

To understand this distribution better, the distinct or effective pore size “D₁₀” of the top layer was calculated from cumulative frequency distribution curves for each of the layers. The value of D₁₀ denotes the effective pore size, where 90% of pores have sizes larger than D₁₀. A good correlation was observed between the effective pore size and the aggregate size Fig. (17). It may be seen that D₁₀ increases with the increase in aggregate size.