Asphalt Internal Structure Characterization with
X-Ray Computed Tomography and Digital Image
Processing
Ibrahim Onifade, Denis Jelagin, Alvaro Guarin, Bjorn Birgisson,
and Nicole Kringos
Department of Transport Science (TSC)
Division of Highway and Railway Engineering
KTH Royal Institute of Technology
Brinellvägen 23, SE-100 44 Stockholm, Sweden
onifade@kth.se
Abstract. In this paper, detailed study is carried out to develop a new workflow
from image acquisition to numerical simulation for the asphalt concrete micro-
structures. High resolution computed tomography scanned images are acquired and
the image quality is improved using digital image processing techniques. Non-
uniform illumination is corrected by applying an illumination profile to correct the
background and flat-fields in the image. Distance map based watershed segmenta-
tion are used to segment the phases and separate the aggregates. Quantitative analy-
sis of the micro-structure is used to determine the phase volumetric relationship and
aggregates characteristics. The result of the quantitative analysis showed a very high
level of reliability. Finite Element simulations were carried out with the developed
micro-mechanical meshes to capture the strength and deformation mechanisms of
the asphalt concrete micro-structure. From the micro-mechanical investigation the
load transfer chains, higher strength characteristics and high stress localization at
the mastic interface between adjacent aggregates was shown.
Keywords: X-ray computed tomography, digital image processing, finite element
method, image based modeling.
1 Introduction
Asphalt concrete (AC) is a heterogeneous material which consists of mastic (binder
and fines), aggregates and air-voids. The distribution of the air-voids in the matrix,
the interaction between the aggregates and the mastic, and the properties of the ag-
gregates and the mastic plays a vital role in determining the mechanical behavior of
the asphalt concrete. Mainly, the aggregate properties determine the strength char-
acteristics, the mastic determines the durability characteristics and the air-void is
related to the rate of moisture damage and rutting in the asphalt concrete. The be-
haviour or response of the AC is highly dependent on the temperature and rate of
loading. At low temperatures, it exhibits the characteristics of an elastic material
N. Kringos et al. (Eds.): Multi-Scale Model.
&
Charact. of Infrastruct. Mater., RILEM 8, pp. 139–158.
DOI: 10.1007/978-94-007-6878-9_11 © RILEM 2013
140 I. Onifade et al.
while at high temperatures above the glass transition temperature, the response is
viscoelastic.
The micro-structure of AC is very complicated and it is defined by the grada-
tion of aggregates, the orientation and number of contacts of aggregate particles,
the properties of aggregate-binder interface, the voids structure, the chemical con-
stituent of the bitumen, the texture of the stones, the adhesion between aggregates
and mastic among others. [Wang, 2010]. Understanding the complex mechanical
interaction that exists between the constituents of the asphalt concrete requires a
reliable way to characterize the AC micro-structure.
In this context, mastic is referred to as the mixture of binder and fines. The main
content of the binder is bitumen which is obtained from the fractional distillation of
crude oil. The bitumen is made up of complex chains of hydrocarbon which makes
it difficult to model the material using it’s true chemical constituent. The hydrocar-
bon chains are very sensitive to changes in the environment (temperature), the rate
of application of external forces or loads can also lead to a possible rearrangement
of the chemical structure of the material. To overcome this difficulties, the mastic
is usually represented using springs and dashpots to model the rate dependent re-
sponse of the material. The Prony series, which is made up of springs and dashpots
connected in series is used to model the mastic behavior in this study. In this study,
the aggregates are considered as a linear elastic material and, as such, the param-
eters required for modeling the aggregates behavior are the Young’s Modulus, the
Poisson’s ratio and the density.
Under loading condition, there may exist a rearrangement of the internal structure
of the asphalt concrete mixture depending on the magnitude and duration of loading.
The most important aspects to consider in the modeling are the contact between
adjacent aggregates and the interface between the aggregate and the surrounding
mastic. The morphological properties (shape, angularity and texture) determines the
load transfer between aggregate particles at contact and the bonding at the interface
between the mastic and aggregate particles. These properties are captured using the
X-ray CT and quantified using an image processing software (Avizo).
In the past, the micro-structure of the AC has been simplified or over-idealized.
Bazant et al. [Z.P. et al., 1990] and Schlangen and van Mier [Schlangen and van
Mier, 1992] have represented the aggregates as rigid spherical particles while others
Wittmann et al. [Wittmann et al., 1985] and Wang et al. [Wang et al., 1999] have
used algorithm to generate a random micro-structure image of the asphalt concrete
micro-structure. Most of these past investigations were based on 2D analyses of the
asphalt concrete micro-stucture due to the complexities in generating or accurately
representing the 3D micro-structure of asphalt concrete.
New techniques show a possible way to capture the micro-structure of the AC
to generate models for numerical simulation, one of which is X-Ray Computed To-
mography (CT). There has been a number of recent attempts to use X-Ray CT to
investigate the internal structure of AC and to investigate its impact on the AC me-
chanical properties. Coleri et al. [Coleri et al., 2012b] used the X-ray CT to study
the changes in AC micro-structure using full-scale test sections and Heavy Vehicle
Simulator (HVS) loading, and X-ray CT images taken before and after HVS testing.
Asphalt Internal Structure Characterization Using X-Ray CT and DIP 141
Coleri et al. [Coleri et al., 2012a] also used the X-ray computed tomography (CT)
and digital image processing to generate the internal micro-structure of the asphalt
mixtures and study the effectiveness of 2D and 3D models for the simulation of the
shear frequency sweep at constant height (FSCH) test. Zelelew [Zelelew and Pa-
pagiannakis, 2011] used automated digital image processing (DIP) algorithm called
Volumetrics based Global Minima (VGM) thresholding algorithm for processing
asphalt concrete (AC) X-ray computed tomography (CT) images. The thresholding
algorithm utilizes known volumetric properties of AC mixtures as the main criterion
for establishing the air-mastic and mastic-aggregate gray scale boundary thresholds.
Bhasin et al. [Bhasin, 2011] used X-ray CT images to study the 3-dimensional dis-
tribution of the mastic in asphalt composites. You et al. [You et al., 2012] devel-
oped a three-dimensional (3D) micro-structure-based computational model to pre-
dict the thermo-mechanicalresponse of the asphalt concrete using a coupled thermo-
viscoelastic, thermo-viscoplastic, and thermo-viscodamage constitutive model. You
et al. [You et al., 2008] studied the dynamic modulus from the stress-strain response
under compressive loads for two-dimensional 2D and three-dimensional 3D micro-
structure-based discrete element models of asphalt mixtures. Masad et al. [Masad
et al., 2005] developed an approach for constitutive modeling of the viscoplastic be-
havior of asphalt mixes and measured the micro-structure damage with X-ray com-
puted tomography and image analysis techniques. However, very limited amount of
work has been done so far for 3D image-based modeling of AC while some studies
only considered the distribution of the aggregates and the mastic phase [You et al.,
2012]
The air-voids, mastic and aggregates phase are considered in this study. The in-
terface between the mastic phase and the aggregate phase is challenging to model
when considering the mechanics of the mastic phase which is highly anisotropic.
Further research is required to adequately understand the interaction at the boundary
between the mastic and the aggregates. However, the spatial location of the contact
points between adjacent aggregates is determined in this study and referred to as the
contact geometry.
2 Objectives and Scope
The present study is aimed at developing the workflow from image acquisition to
simulation for accurate characterization of the AC micro-structure. The main ob-
jectives of this study are to develop procedures for: (1) Segmentation of the three
different phases in AC and determination of their volumetric relationship. (2) De-
termination of air-voids phase distribution with depth. (3) Determination of aggre-
gates particle size gradation and distribution. (4) Determination of the distribution
of contact zones between aggregates. (5) Micromechanical simulation using finite
elements method (FEM).
The steps involved in this study are summarized in Figure 1
142 I. Onifade et al.
Fig. 1 Process workflow
3 Experimental Data and Scanning Procedure
In this study, the KTH X5000 CT X-ray scanner is used to obtain the detailed micro-
scopic structure of the porous asphalt concrete core sample for further visualization,
characterization and analysis. The X5000 CT scanner is a seven-axis universal x-ray
imaging system designed for the inspection of large objects. It can accommodate a
variety of part shapes, sizes and weights. It can produce X-ray intensities of up to
450kV.
The asphalt concrete core sample with a diameter of 100mm and a height of
80mm as shown in Figure 2 is scanned. The sample is scanned at an energy intensity
of 225kV without beam filtration. The scanning resolution is 1949 x 1799 with a
slice thickness of 59microns and a total of 1932 slices.
The x-ray scanning process includes sample preparation, warming-up the scan-
ner, pre-scan settings, scanning, detector calibration and CT calibration. Beam hard-
ening artifact is manifested in CT images with brighter edges than the center of the
image. Beam hardening artifact reduces the quality of the scanned image and hence
affects the phase segmentation results.
There are a number of possible techniques to reduce the beam hardening in the
scanned image which includes the use of X-ray beam that is energetic enough to
ensure that beam hardening is negligible, use of filters, increased exposure time
among others [Ketcham and Carlson, 2001]. In this study, the beam hardening
artifact is corrected using the background and flat field correction feature in Avizo
Fire. The flat field is computed using the bkgimg command in Avizo which computes
a background image from a gray level image and a binary mask (or no mask for all
pixels of the image) using second order polynomial. The intensities of the 3D input
image are then scaled according to the normalized intensities of the flatfield images.
Asphalt Internal Structure Characterization Using X-Ray CT and DIP 143
Fig. 2 Porous asphalt concrete sample used
in the present study
The input image gets brighter at pixels where the flatfield is dark and vice versa. In
this way non-uniform illumination is compensated for [Avizo, 2009]. Other digital
image processing and analysis is also performed using Avizo Fire application.
4 Digital Image Processing (DIP)
Image processing mainly involves editing and enhancement of digital image with
the aim of improving the quality of the image or to extract relevant information.
DIP is also used for the identification and segmentation of the different phases in
the AC micro-structure.
The different techniques used in improving the quality of the acquired image in-
clude contrast enhancement, illumination correction and filtering to reduce noise in
the image. Avizo background and flat field correction tool is used to correct non-
uniform illumination in the acquired CT image. Non-uniform illumination correc-
tion helps to achieve improved segmentation results especially when the threshold
based segmentation is used.
Filters are mainly used to reduce noise and thereby improve image quality. Dif-
ferent types of filters are used in image processing depending on the expected result
or outcome. It is important to note that the nature of the mastic makes the segmen-
tation process of the asphalt concrete sample a little cumbersome. Considering the
fact that the mastic is a mixture of bitumen and fines, the threshold based segmen-
tation becomes difficult as part of the mastic is identified as aggregates. This is as a
result of the CT attenuation of the constituents materials in the asphalt concrete mix.
Image filtering can be used to overcome this problem and thus improve the image
segmentation results. Non-local means filter is used in this study to reduce the noise
in the image and help improve segmentation as it smoothens regions inside objects
while preventing smoothening near the edges.
Segmentation is the process of separating pixels with the same gray level value
from those pixels with a different value. This is a very important step in AC image
