The pervious concrete cube samples were drilled to extract cores of 70 dia. x 100 mm height. Some cores were cut into 70 dia x 35 mm thick discs, which were then used for determination of pore volume results or PSDs, as per the VM, LTM and IA procedures.
The falling head permeability test was done on various pervious concrete cores, while a limited number of core specimens were analysed using X – ray microCT to quantify pore connectivity. Figure 3 shows cube samples along with cores and disc specimens that were prepared for conducting the various tests.
3.1 Tests and measurement methods
3.1.1 Measurement of pore volume and pore size distribution
Flat surfaces of the 70 dia. x 35 mm thick disc specimens, were polished using successively finer abrasives to produce a satisfactorily smooth surface finish. Pore volume and PSD tests were done on the disc specimens, employing all three different methods comprising the VM, LTM and IA. It may be recalled that the VM and LTM only measure pore volume, while IA determines PSD.
(a) Volumetric method
The VM’s standard procedure given in ASTM C1754 (2012) (Montes et al., 2005) was used to determine pore volume of pervious concrete, based on low drying temperature. The procedure involved oven - drying the 70 dia. x 35 mm thick disc specimens for 24 hours at 40oC, followed by weighing in air (WD) and in water (Ws). Pore volume (Pv) was calculated using Eq. (2).
Pore volume (%) = [1–((WD – WS)/ ρw)/ VT] × 100 (2)
where WD - oven dry weight (g), WS - submerged weight (g), ρw - density of water (g/cm3), VT - total volume of specimen (cm3).
(b) Linear - traverse method
The LTM procedure given in ASTM C457 (2016) is a standard microscopical technique for determining the void content of normal concrete. In the present study, the method was innovatively employed to determine pore volume of pervious concrete. The measurement procedure involved drawing parallel lines across flat surfaces of disc specimens. The total sum of line lengths traversing across void components at the disc surface (Ta), is expressed as a proportion of the total sum of all line lengths drawn across the surface (Tt). Figure 4 shows parallel lines spaced 2.5 mm apart, drawn across flat surfaces of disc specimens. For each specimen, Ta and Tt were determined then used to calculate void content (PLT), as per Eq. (3).
\(\text{Void}\hspace{0.33em}\text{content (\%)}\hspace{0.33em}=\hspace{0.33em}\frac{{\sum }_{1}^{n}\text{T}{\text{a}}_{i}}{{T}_{t}}\) x 100 (3)
where Tai – length of a line traversing a void component (i), Tt – total length of all lines traversing across the whole disc surface, n – total number of void components encountered across the disc surface.
(c) Image analysis
Upon completion of pore volume measurements done on disc specimens as per the VM and LTM, the same specimens were then prepared for IA. The IA preparation procedure involved coating the flat surfaces of disc specimens using white paint. After the coating had dried, painted surfaces of the specimens were scanned using a flatbed scanner to acquire processed images.
Figure 5 shows some representative scanned images of disc specimens. White areas of the binary (black /white) images indicate solid components, while the irregular black spots are pores. It can be seen in Fig. 5, that aggregate sizes significantly influence pore interconnectivity. As the aggregate size used in pervious concrete reduced, pore sizes correspondingly became more numerous and smaller in size (Fig. 5a of specimen G6.7). In contrast, use of larger aggregate sizes in pervious concrete led to fewer but larger pore sizes (Fig. 5b,c of specimens G9.5, G13.2). It is also evident that the pores within each specimen, are randomly distributed from small to larger sizes.
In the present study, the IA free software Image J which is freely available at http://rsbweb.nih.gov/ij/, was used to conduct PSD analysis (Ferreira and Rasband, 2012). Pore area is determined as the number of pixels occupying a given pore space x area per pixel. The calculated area of a pore is converted into a circle of an equivalent area. The circle’s diameter is taken as the equivalent pore diameter. A histogram is then plotted for various pore sizes occurring across the surface, resulting in a cumulative frequency distribution curve of pore sizes i.e. the PSD curve. Threshold pore diameter is obtained as the pore size (dc) taken at 50% proportion of the cumulative frequency distribution curve (Neithalath et al., 2010).
IA employs the underlying assumption that the fraction of pores determined based on 2D area analysis is equivalent to the 3D volume fraction. However, various researches (Diamond and Leeman, 1995; Eshel et al., 2004) have shown that this assumption may be responsible for the observed IA’s overestimation of pore volume or PSD.
3.1.2 Permeability measurement
A falling - head permeameter shown in Fig. 6 was built in - house within the laboratory, in accordance with details given in Das (1988). The permeameter comprised a 450 mm long graduated cylinder of 72 mm inner dia., connected to a U – pipe fitted with a valve. The test set - up involved inserting a saturated core specimen inside a rubber tube of 65 mm dia., during which the tubing was circumferentially stretched to tightly hold the 70 mm dia. core against radial water flow. Clamps were used to attach the sample in position, as shown in Fig. 6. Water was added into the graduated cylinder until its level at both the cylinder and at the drain pipe, reached the same height, then the valve was closed. Afterwards, the graduated cylinder was filled with water up to the maximum height level of 350 mm. Each test run involved opening the valve and measuring the time (in seconds) taken for water to drain through the core specimen, from the initial hydraulic head (h1) of 350 mm to the lower head (h2) of 50 mm. Three test runs were done for each core specimen. Permeability was calculated using Eq. (4) (Das, 1988).

where k - permeability (m/s), a - cross section area of the tube (m2), L - height of the core sample (m), A – cross section area of the core sample (m2), t - time taken for water level to drop from the height of h1 to h2 (seconds), h1 - upper height level of water (m), h2 - lower height level of water (m).
3.1.3 Pore connectivity determination
A limited set of core specimens were analysed using X – ray microCT to quantify pore volume, connected porosity, isolated porosity and pore connectivity. Of the methods employed in the present study, microCT was the only technique capable of directly quantifying pore connectivity, while also characterizing the spatial pore distribution features (du Plessis and Boshoff, 2019).
3.2 Discussion of measured pore parameters and clogging effects
3.2.1 Pore volume results
Figure 7 gives the pore volume results determined as per the VM, LTM and IA. It can be seen that the pore volume results obtained are within 16 to 30%, which is the typical range of porosity values for pervious concretes (Ekolu et al., 2016). The mixtures made with dolomite aggregate, gave pore volume results that were generally lower than the corresponding values obtained for the samples made with the other aggregate types.
It is also evident that the pervious concrete samples G9.5C, G13.2C and S6.7C, which had been used to treat AMD in turn giving rise to pore clogging, showed porosity values that were similar to those of the control samples (G9.5, G13.2, S6.7) that had not been exposed to AMD. These observations show that the VM and LTM were unable to determine pore clogging effects. It is known that clogging occurs as discrete blockages within small pores, forming bottlenecks inside an interconnected pore network. Such small discrete blockages are unlikely to significantly affect the bulk pore volume results determined using the conventional methods.
3.2.2 Pore size distribution results
IA was employed to determine the PSDs of pervious concretes, as shown in Fig. 8. It can be seen that the pore volume (PI) results determined using IA, are generally within the range of 15 to 35%. These values are similar to the pore volume results that were determined using the VM and LTM (Sect. 3.2.1). The observed maximum pore sizes (mps) in the pervious concretes, were generally between 10 to 15 mm. Moreover, the mps values increased with corresponding increase in the aggregate sizes used to prepare the pervious concretes. These findings are consistent with the visual observations discussed in Sect. 3.1.1c, showing that pervious concretes made with larger aggregate sizes, depicted correspondingly greater mps values.
3.2.3 Statistical comparison of pore volume results
Figures 9 gives pairwise comparisons for the pore volume results comprising Pv, PLT and PI, that were determined in accordance with the VM, LTM and IA, respectively. Evidently, Pv and PLT show strong agreement with values lying along the line of equality (Fig. 9a). In the converse, PI results are consistently higher than the corresponding Pv or PLT values, as seen in Fig. 9b,c. The tendency of IA to overestimate porosity values or PSDs, relative to those determined using conventional methods, is well - established (Diamond and Leeman, 1995). The assumption employed in IA, which considers the 2D area fraction of an irregularly shaped pore as equal to volume fraction of an equivalent pore sphere, is responsible for the method’s tendency to overestimate porosity (Sect. 3.1.1c). It is known that a non - spherical particle measured across all orientations, gives a cross - sectional area that is larger than that of a sphere which has an equivalent volume as the particle (Eshel et al., 2004).
Statistical analysis was conducted on pore volume results of the paired methods. Three (3) statistical parameters were employed comprising the ratios Pv/PLT, PI/Pv, and PI/PLT; root mean square of errors (RMS) and coefficient of variation of errors (CVE). By definition, RMS = \(\sqrt{\text{(Residua}{\text{l)}}^{2}\text{/N}}\), where Residual is the difference between corresponding porosity values of the paired methods, while N is the total number of paired data values. CVE is the ratio of RMS to mean of all paired data values, expressed as a percentage. It can be seen in Table 3 that mean values of the ratios Pv/PLT, PI/Pv and PI/PLT are 1.10, 1.18 and 1.28, respectively. The Pv/PLT ratio of 1.10 is quite close to 1.0, which indicates strong agreement between results of the VM and LTM. Considering the average values comprising PI/Pv = 1.18 and PI/PLT = 1.28, it is evident that the IA method gives pore volume results that are about 20 to 30% higher than the corresponding values determined in accordance with the VM or LTM.
CVE values (Table 3) for the pairwise comparisons comprising (Pv, PLT), (Pv, PI) and (PLT, PI) are 19.9, 29.9 and 34.1%, respectively. Clearly, these observed CVE values fall within the range of 20 to 50%, which is the typical accuracy level of recognized prediction models (Bazant and Baweja, 1995; Ekolu, 2018; 2020).
Figure 10 shows residuals for pore volume results of the paired methods. Again, the pair (Pv, PLT) shows the closest agreement of results, giving both the lowest CVE of 19.9% and a small spread of \(\pm\)5%. It is also evident in Fig. 10 that there was no heteroscedasticity such as fanning out or convergence of data points. Also plotted in Fig. 10 are the 95% confidence limits. It can be seen that the residuals for each pair of methods, all fall within the confidence limits. The foregoing statistical analysis thus shows that the three (3) porosity measurement methods comprising the VM, LTM and IA, may be employed interchangeably. However, the IA technique gives pore volume results that are 20 to 30% higher than corresponding values determined using the conventional methods.
Table 3
Statistical pairwise comparisons for pore volume results: Pv, PLT and PI are porosity values determined using the volumetric method (VM), linear – traverse method (LTM) and image analysis (IA), respectively.
Sample ID
|
Pv (%)
|
PLT (%)
|
PI (%)
|
Pv/PLT
(Pv, PLT)
|
PI/Pv
(PV, PI)
|
PI/PLT
(PLT, PI)
|
D6.7
|
15.0
|
18.45
|
15.2
|
0.81
|
1.01
|
0.82
|
D9.5
|
22.7
|
15.76
|
21.8
|
1.44
|
0.96
|
1.38
|
G6.7
|
23.4
|
23.99
|
32.1
|
0.98
|
1.37
|
1.34
|
A9.5
|
27.1
|
22.14
|
18.9
|
1.22
|
0.70
|
0.85
|
G9.5
|
17.9
|
18.45
|
26.3
|
0.97
|
1.47
|
1.43
|
G9.5C
|
21.1
|
15.76
|
30
|
1.34
|
1.42
|
1.90
|
G13.2
|
22.6
|
29.19
|
34.5
|
0.77
|
1.53
|
1.18
|
G13.2C
|
30.6
|
27.42
|
35
|
1.12
|
1.14
|
1.28
|
S6.7
|
26.7
|
21.92
|
36.2
|
1.22
|
1.36
|
1.65
|
S6.7C
|
23.3
|
21.06
|
19.5
|
1.10
|
0.84
|
0.93
|
Mean
|
|
|
1.10
|
1.18
|
1.28
|
RMS
|
|
|
(4.4)
|
(7.5)
|
(8.2)
|
CVE (%)
|
|
|
(19.9)
|
(29.9)
|
(34.1)
|
3.2.4 Pore clogging effects
It was found as discussed in Sect. 3.2.1, that the conventional pore volume measurement methods comprising the VM and LTM, are unable to quantify pore clogging effects. However, other findings show that the falling head permeability test and IA technique, are able to effectively determine pore clogging effects in pervious concrete.
It can be seen in Fig. 11 that permeability results of the core specimens G9.5C, G13.2C, S6.7C reduced by 51.2, 16.8, 50.6% relative to corresponding values of control specimens G9.5, G13.2, S6.7, respectively. Evidently, pore clogging was responsible for the observed permeability reductions of nominally about 20 to 50%, depending on the type and size of aggregate used in pervious concrete.
IA results also showed that the pervious concrete samples (G9.5C, G13.2C, S6.7C) which had been used to treat AMD thereby giving rise to pore clogging, showed significantly smaller dc values relative to values of the corresponding control mixtures (G9.5, G13.2, S6.7) that were not exposed to AMD. For example, dc of the control specimen G9.5 reduced from 1.30 mm before pore clogging occurred to 0.90 mm after clogging in G9.5C, a significant decrease of 31%. Similarly, the 1.00 mm dc value of the control specimen G13.2 reduced by 35% to 0.65 mm of G13.2C.
3.2.5 Effects of aggregates on permeability results
Pervious concretes showed higher permeability with increase in size of the aggregate used, as seen in Fig. 11. For example, pervious concretes made with granite aggregate of various sizes comprising G6.7, G9.5 and G13.2, exhibited permeability values that correspondingly increased with increase in aggregate size, giving 5.6, 8.2 and 10.1 mm/s, respectively. Also, permeability results of pervious concretes made with the dolomite aggregate were significantly lower than corresponding values of the mixtures made with granite aggregate. For example, samples D6.7 /D9.5 gave permeability values of 0.9 /1.8 mm/s, which are 5 to 6 times less than the corresponding values comprising 5.6 /8.2 mm/s of samples G6.7 /G9.5, respectively. Pore volume results also showed a similar trend depicting this effect of the dolomite aggregate (Sect. 3.2.1). Of the various aggregate types used in the present study (Table 1), granite aggregate was found to be most suitable as it consistently gave the highest permeability results. Accordingly, a pervious concrete mixture prepared with granite aggregate expectedly exhibits a relatively more efficacious hydraulic conductivity. These observations also corroborate the earlier findings reported in Ekolu et al. (2016).
3.2.6 X - ray micro - computed tomography
X - ray microCT was employed for direct quantification of pore connectivity (Sect. 3.1.3). The total pore volume, connected porosity, isolated porosity and pore connectivity results of the samples G6.7 /G6.7FA, were determined to be 28.75 /19.10%, 28.65 /18.24%, 0.10 /0.86% and 99.7 /95.3%, respectively.
It is evident that incorporation of 30% FA into CEM I mixture (Sect. 2.1), altered the pore characteristics of pervious concrete, leading to reduction in pore connectivity from 99.7% for G6.7 to 95.3% for G6.7FA. This observed effect of FA on pore connectivity may partly be attributed to increase in paste volume of mixtures, owing to the relatively lower density of the supplementary cementitious material. More importantly, it is notable that the sample G6.7FA gave a significantly higher isolated porosity of 0.86%, compared to the much smaller 0.10% value of G6.7. The higher isolated porosity of G6.7FA is directly attributed to discrete pore blockages typically resulting from incorporation of FA into concrete mixtures. It is the implied presence of FA – induced discrete pore blockages, that majorly accounts for the significantly reduced pore connectivity of sample G6.7FA.
Figure 12 gives microCT slice 2D images of the core samples G6.7 and G6.7FA. The black features seen within sample geometry are the irregular - shaped pores, while the dark gray irregular shaped features are aggregate particles. The surface – coating seen on aggregate particles is hardened paste, while the bright dotted spots in the paste are alite particles of unhydrated or partially hydrated cement grains. The slice microCT images of G6.7 (Fig. 12a,b) and of G6.7FA(Fig. 12c,d), both show high pore connectivity levels, consistent with the measured values comprising 99.7 and 95.3%, respectively.