1. Porosity And Permeability Evaluation Of Pervious Concrete Using 3D X-Ray Computed Tomography
A. Jagadeesh and G. P. Ong
National University of Singapore
Y. M. Su
National Kaohsiung University of Science and Technology

2. Introduction
Porous pavements are used on highways & streets to minimize storm-water run-off, wet weather accidents & tire/road noise
Increase in intrinsic permeability, wet pavement skid resistance & acoustic absorption coefficients
80% reduction in wet-weather accidents [Association of Japan Highway, 1996]
Tire/road noise reduction of 3 to 5 dB [Berengier et al., 1997]

3. Pervious Concrete
Special class of hydraulic cement concrete proportioned with sufficient interconnected voids that result in highly-permeable material [ACI, 2010]
Isolated voids is minimal due to use of uniform single-size coarse aggregates in mixture
Increased use as functional performance layers necessitates understanding of pore network properties and their relationships

4. X-Ray Computed Tomography in Pavement Research: Literature Review
Limited studies in literature:
Fluid flow simulation model for hot-mix asphalt using X-ray CT scan and Lattice Boltzmann method [Kutay et al., 2007]
Permeability tensor coefficients for asphalt using fluid flow micro-simulation [Masad et al., 2007]
Pore network parameters such as porosity, tortuosity & pore size determination [Coleri et al., 2012; Kuang et al., 2015, Biligiri, 2017]
Digital image processing thresholds [Masad et al., 1999; Abera et al., 2017]
High-resolution industrial X-ray CT is capable of producing detailed 3D imagery of Asphalt cores (Source: University of Texas at Austin)

5. Research Motivation & Objectives
Examine the use of medical-grade 3D X-ray computed tomography (CT) to determine effective porosity & permeability of pervious concrete
Examine performance of different segmentation threshold algorithms used in practice and research
Medical-grade vs. Industrial-grade micro-CT [du Plessis et al., 2016]
↑ X-ray voltages with industrial CT systems ⇒ ↑ penetrating power ⇒ ↑ image quality for dense objects
↓ cost and ↓scan time for medical-grade X-ray CT
Medical CT scans produce useful data in significantly reduced times especially for large no. of samples and moderate resolution
Ability to scan larger objects than typical microCT systems
↓ data set sizes ⇒ faster analysis with ↓ computing power

6. Pervious Concrete Mixture Nominal Maximum Aggregate Size (NMAS)
Materials
Two pervious concrete mixtures (P1 & P2) were produced in laboratory:
ASTM Type I cement as binding agent
Superplasticizer of 0.5% by weight of cement added to improve pervious concrete workability
Pervious Concrete Mixture
Nominal Maximum Aggregate Size (NMAS) P1
9.5 mm P2
12.5 mm
Note: Specific gravity and percent absorption of aggregates are 2.64 and 1.35% respectively

7. X-Ray CT Scan
Somatom Emotion 16-channel X-Ray CT scanner with 110kV energy was used in this study
A total of 300 section images of pixel size 1024 x 1024 were obtained at the interval of 0.7mm.
Voxel size of about 300 mm (cf. 150 mm for micro-CT) [Chandraparra & Biligiri, 2018]
Sample size of 150 mm diameter and 250 mm thickness scanned (cf. ~100 mm thickness for micro-CT) [Chandraparra & Biligiri, 2018]
Somatom Emotion from Siemens with sample
2D slice raw image of pervious concrete sample for P1 (left) and P2 (right)

8. Thresholding Algorithms
Thresholding of air voids based on grey-scale intensities was performed
Otsu bi-level and tri-level algorithm [Otsu 1979]:
Volumetric-based global minima algorithm [Zelelew & Papagiannakis, 2011]:
Intraclass variance for material 1
Intraclass variance for material 2
Grey-scale histogram
Experimental air void content
Computed air void content given threshold value
Reconstructed 3D model (right) from original CT 3D model before thresholding (left)

9. Threshold for Pervious Mixture P1 Threshold for Pervious Mixture P2
Thresholding Results
Air-void thresholds for the three algorithms for this study:
Algorithm
Threshold for Pervious Mixture P1
Threshold for Pervious Mixture P2
Otsu bi-level
1678
1792
Otsu tri-level
1471
1168
Volumetric
1647
1694
(a) Original image (b) Otsu’s bi-level (c) Otsu’s tri-level (d) Volumetric method

10. Constant Head Permeability Test
Permeability of pervious concrete samples was measured using the Association of Japan Highway (1996) standards
Permeability coefficient kT:
Percentage of inter-connected air voids or effective porosity:
Height of specimen
Volume of outflow water
Constant head
Cross sectional area
Time taken
Actual setup of constant head permeameter

11. Permeability Simulation Model
Fluid behavior in porous media can be modelled using the Navier Stokes equations & k-ε turbulence equations.
Standard properties of water at 25°C are adopted in this study.
Performed for P1 and P2 reconstructed volumes for the three algorithms (Otsu bi-level, Otsu tri-level & volumetric)
Continuity equation:
Momentum equation with k-e turbulence:
Sample simulation of fluid flow over pore network within pervious concrete specimen

12. Discharge Characteristics
Experiments show that ↑NMAS ⇒ ↑ dimensional geometry of pore network ⇒ ↑ vertical & horizontal permeability of pervious concrete
k-values: Otsu bi-level method > Volumetric method >> Otsu tri-level method
Error in k values: Volumetric method (17% to 28%) < Otsu bi-level method (22% to 56%) < Otsu tri-level method (45% to 56%)

13. Discharge Characteristics
Permeability anisotropy ratio: ratio of kv to kh values
Tortuosity: ratio of actual length of fluid flow to shortest distance from top to bottom of sample
↑ in air void threshold ⇒ opening of new interconnected air void channels ⇒ ↓ in tortuosity value ⇒ ↑ in permeability values
Similar findings to literature [Chandrappa & Biligiri 2017]

14. Discharge Characteristics
Velocity streamlines obtained from the simulation model
Reduced velocities were observed in the smaller pore channels and higher velocities at the larger pore channels.
Pressure losses in porous domain observed due to inertial effects and geometric features of pore network
Velocity streamlines of pervious concrete sample (obtained from permeability simulations)

15. Volumetric Characteristics
↑ in interconnected voids + ↓ in isolated voids ⇒ ↑ in permeability values
NMAS has a significant effect on permeability despite having the same porosity due to varied pore network structure
Error in effective porosity values: Volumetric method (<1%) < Otsu bi-level method (~5%) < Otsu tri-level method (25% to 35%)

16. Conclusion
Volumetric segmentation algorithm is considered to be predicting the permeability and effective porosity more closely to the experimental results compared to the Otsu’s bi-level and tri-level algorithms
Aggregate size and tortuosity have a significant influence on the permeability, despite having the same effective porosity.
Overall, it is expected that the present research will help to understand the pore network characteristics of pervious concrete using non-destructive evaluation and digital image processing.

17. University Town, National University of Singapore