Journal of Advanced Concrete Technology Vol. 6, No.

1, 143-155, February 2008 / Copyright © 2008 Japan Concrete Institute 143

Scientific paper

Simulation of Chloride Diffusivity for Cracked Concrete Based on RBSM

and Truss Network Model
Licheng Wang1, Mitsutaka Soda2 and Tamon Ueda3

Received 11 May 2007, accepted 30 October 2007

Abstract
For concrete structures exposed to salt environment, the microstructure and cracks play a crucial role in the ingress of
chloride ions into concrete. In this study, concrete is simulated on the meso scale as a three-phase composite, i.e., ag-
gregate particles, mortar and the interfacial transition zone (ITZ). Because of the advantages in predicting cracks behavior
in concrete, Rigid Body Spring Model (RBSM) is employed to carry out the mechanical analysis to simulate the distri-
bution and width of microcracks. And then, the truss network model is adopted to evaluate the chloride diffusivity of the
cracked concrete. On the basis of the statistics analysis of diffusion coefficients of concrete and mortar determined ex-
perimentally, the diffusivity of ITZ is analytically clarified. The range of diffusion coefficient of ITZ estimated in this
paper is approximately 3-16 times of that of mortar depending on the different assumed thickness, which agrees well with
that of the previous experimental results. With the aim to validate the effect of microcracks on the diffusivity of concrete,
a series of the chloride ions penetrating analysis is numerically carried out on the concrete specimen under different stress
levels. The axial compressive and tensile loading conditions are investigated respectively and the effects of stress level on
chloride diffusivity of cracked concrete are examined. Results indicate that the chloride diffusivity is significantly de-
pendent on the stress level, but only considering the effect of cracks predicted by RBSM is not sufficient. So an empirical
equation which can account for the microstructure variation of concrete under loading is proposed. With it, a reasonable
estimation for chloride diffusivity of cracked concrete is achieved.

1. Introduction shrinkage, thermal gradients, freeze-thaw cycling, al-

In an environment exposed to seawater or de-icing salt,
the chloride-induced corrosion of reinforcing steel in
concrete structures has become one of the major causes
of deterioration in such reinforced concrete facilities as
marine structures and bridges. Thus, in the durability
design of these structures, the most important factor that
determines the initiation time of corrosion of steel bars
and service life of concrete structures is the transport
property of concrete; e.g., the diffusivity of ions and
gases. As a rule, the diffusivity of concrete is determined
experimentally (Care 2003). However, it is difficult to
precisely predict the diffusivity of concrete because the
concrete diffusivity is influenced by many parameters
such as the interfacial zone property between aggregate
particles and bulk cement pastes, as well as the micro-
structure of the cement paste itself (porosity and pore
structure) (Oh and Jang 2004).
Moreover, in most cases, cracks (or microcracks) are
present in concrete due to several causes such as
1 The interfacial transition zone (ITZ) formed at the vi-
PhD., Post doctor, Division of Built Environment,
Hokkaido University, Japan; Associate Professor, School
of Civil & Hydraulic Engineering, Dalian University of
Technology, China.
E-mail: wanglc@eng.hokudai.ac.jp
2
Engineer, Environmental Creation Bureau, Japan.
3
Dr. Eng., Professor, Division of Built Environment,
Hokkaido University, Japan.
kali-aggregate reaction and external loading, and their
combination. These cracks may form potential flow
channels which provide easy access to aggressive salts
and ions (Lim et al. 2000; Boulfiza et al. 2003), so cracks
in concrete can have negative effects on diffusivity, rate
of chloride ingress and hence reinforcement corrosion.
Moreover, it is argued that chloride ions usually cannot
transport through the gel pore of cement paste, so the
diffusivity of chloride ions in concrete is mainly per-
formed in the existing microcracks or the load-
ing-induced cracks (Xing et al. 2005a). Therefore, in the
study of diffusivity or substances transportation charac-
teristics of concrete, cracks must be taken into account
correctly to cover the practical working condition of
structures. However, due to difficulties in generating
desirable crack patterns in concrete specimens and
availability of appropriate methods for concrete diffu-
sivity measurements, a limited number of studies have
been done on the transport property of cracked concrete
(Wang et al. 1997), so till now the characteristics of
chloride ions transportation in cracked concrete have not
yet been made clear (Kato and Uomoto 2005).
cinity of aggregate particles in concrete is a very thin
layer, in which the microstructure differs significantly
from that of the bulk cement paste (Delagrave et al.
1997; Shane et al. 2000). The ITZ thickness is usually
reported in the range of 15-50μm or less than 80μm when
the aggregate size is varied from 0 to 20mm (Otsuki et al.
2006) and the typical values are in the order of 20μm
144 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008
(Bentz et al. 1997). Generally speaking, the presence of
aggregates in a hydrated cement paste matrix has two
opposite effects on the transport properties (Delagrave et
al. 1996; 1997). First, the addition of solid particles leads
to an increase in the tortuosity of matrix, so the chloride
ions have to move around the solid particles. This implies
that tortuosity (redirecting) effects reduce the transport
properties. Second, the presence of porous and connected
ITZ probably contributes to facilitating the movement of
ions. In order to investigate the effect of aggregate con-
tent and ITZ on the transport properties, usually the
mortar or concrete is treated as a three-phase composite
material in which aggregates are embedded in a matrix of
hardened cement paste, and the ITZ is around the ag-
gregates (Yang 2003; Care and Herve 2004; Kato and
Uomoto 2005) (see Fig. 1). Recent research shows that
the ITZ between the cement paste and aggregate, espe-
cially coarse aggregate, can more effectively explain the
phenomena of chloride diffusion (Takewaka et al. 2003).
Compared with a concrete, the mortar should be more
homogeneous because of its relatively small aggregate
size. Therefore, on the meso scale, the mortar can be
treated as a nominal homogeneous material and the
chloride diffusion coefficient of which is regarded as a
constant for a fixed water-to-cement ratio. But for the
ITZ around coarse aggregates, its effect on the diffusivity
of concrete should be taken into account and clarified
precisely.
The authors have showed potential of numerical ap-
proach with meso-scopic RBSM to simulate effects of
loading on diffusivity of concrete structures (Soda 2006;
Soda and Ueda 2006). The numerical approach is very
helpful to reduce the experimental time and errors and to
understand the phenomena itself. The object of this study
is to propose a numerical method which can appropri-
ately take into account the effect of load-induced crack-
ing on diffusivity of concrete through careful investiga-
tion of experimental evidences on diffusivity of concrete,
mortar, crack and ITZ. It is hoped that the numerical
approach in this study could facilitate the service life
prediction of concrete structures subject to salt attack.
PhaseⅠ: mortar
Phase Ⅱ: aggregate
Phase Ⅲ: ITZ
Fig. 1 Three-phase structure of concrete on meso-scale.
2. Truss network model based on RBSM
2.1 Outline
Generally speaking, the load-induced cracks appear
mostly at the paste-aggregate interface and extend into
the cement paste as the load is increased (Samaha and
Hover 1992). Research evidence indicates that the
characteristics of cracks after a concrete has been
unloaded are quite different from those while under load
because the cracks can close back partially or completely
upon unloading. Thus, the transport property of concrete
can be influenced by the test condition, i.e., whether the
test is carried out under loading or after removal of load
(Lim et al. 2000). However, in most of the previous
studies, the chloride transport property tests were usually
conducted on the completely unloaded specimen but not
on the loading concrete although some attempts were
made to study the progressive cracking in concrete. The
reason is that in most cases it is not easy and possible to
conduct salt attack tests on loaded concrete. It implies
that the results of chloride transport property test would
not reflect the actual crack characteristics of concrete,
because the salt attacks mostly take place under loading
conditions for practical concrete structures. Therefore,
RBSM must be a realistic method to characterize the
cracks of a concrete under loading.
2.2 Concept of RBSM
The RBSM is sometimes referred to as a discrete element
method, which was first developed by Kawai (1978).
One of the key notions of RBSM is its random geometry
based on Voronoi diagrams, so that the locations and
directions of crack propagation need not be anticipated
and adaptive remeshing is not needed (Bolander and Le
1999). In RBSM, the analytical model is divided into
polyhedron elements whose faces are interconnected by
springs. Each element has two translational and one
rotational degrees of freedom for a two dimensional
modeling simulation. The response of the springs can
simulate the interaction between elements instead of
consideration of the internal behavior of each element
(see Fig. 2) (Nagai et al. 2004).
The advantage of RBSM in simulating crack propa-
gation of concrete has been reported widely (Bolander
and Saito 1998; Nagai et al. 2004). In addition to the
simulation of random locations and directions, RBSM
can estimate the crack width by calculating the relative
displacement between rigid elements, which equals to
the elongation of the normal spring.
2.3 Truss network model method
In the present study, RBSM is used to predict the cracks
in concrete under loading, and consequently, the truss
network model is used to numerically simulate the chlo-
ride ions transport process. As shown in Fig. 3, each of
the Voronoi elements is linked by truss elements with the
nodes at the centers of Voronoi elements and the inter-
mediate points of Voronoi element boundaries. This is a
 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008
refined approach of truss network model proposed by
Nakamura et al. (2006), in which only the Voronoi nuclei
are linked to generate a truss element (Bolander and
Berton 2004). And then, it is assumed that substance (e.g.,
chloride ions) transport is performed along one dimen-
sional truss element in the network. Here, it should be
emphasized that the truss network model presented in
this paper is on the meso scale since concrete is treated as
a heterogeneous material, which is quite different from
the ideas of the previous studies regarding concrete as a
homogeneous material (Bolander and Berton 2004; Na-
kamura et al 2006). As a result, the type of truss elements
and the chloride transport through them become more
realistic. For an uncracked concrete, the cross sectional
area and the diffusivity of truss element are set according
to the following rules: for truss elements in aggregate,
both are assumed 0; for truss elements in mortar, the
cross sectional area is given according to the corre-
sponding Voronoi element area and the diffusion coef-
ficient is given according to experimental results of
mortar. For truss elements on the interface of two mortar
Voronoi elements, the cross sectional area is assumed 0
and no mass transfer occurs, but for those on the interface
of aggregate and mortar, i.e., the ITZ truss elements, the
kn
kt
t ter-to-cement ratio (w/c) and the temperature. Because
n
y into account in the present study.
x
Fig. 2 Schematic representations of the RBSM.
Node of centroid of Voronoi element
Node of boundary element
Truss element in Voronoi elements
Truss element on boundary of Voronoi elements
Fig. 3 Schematic representation of truss network model.
145
sectional area is determined by the thickness of ITZ and
the diffusion coefficient will be estimated based on the
method proposed in the following sections. After
cracking, the cross sectional area of truss element on
interface will be estimated based on the crack width
calculated by RBSM and the diffusion coefficient
through these cracks may become quite large compared
with that of bulk concrete (Nakamura et al 2006).
In the present study, volume faction and size distribu-
tion of coarse aggregate are assumed to be in accordance
with conditions generally applied in actual concrete. In
fact there are differences in cracking characteristics, such
as crack surface area and crack direction, between
three-dimensional space and two-dimensional space.
More precise three-dimensional analysis is necessary as a
further study.
3. Statistics of diffusion coefficient of
concrete (Dc) and mortar (Dm)
3.1 Diffusion coefficient of concrete
To a conventional concrete, the diffusion coefficient Dc
is not a real physical constant since the process of chlo-
ride ion transportation in concrete is considered to be
governed by a number of factors such as diffusion, ionic
interaction, water transportation, and chloride ion fixa-
tion (Kato and Uomoto 2005). However, according to the
extensive experimental investigations, it has been con-
cluded that the most important factors are the wa-
the results of experimental and numerical analysis are
scattering too much, in order to simplify the simulating
process, the influence of temperature on Dc is not taken
Several mathematical models based on the regression
analysis for relationship between Dc and w/c have been
proposed according to experimental investigation or
numerical analysis. Fig. 4 shows the effect of w/c on
chloride diffusivity as predicted by some previous mod-
els. Because these models are based on different ex-
perimental conditions and assumptions, in this paper,
they will be firstly summarized and argued separately
and then combined together to give a unified model.
Additionally, it should be pointed out that we are only
attempting to estimate the diffusion coefficient of chlo-
ride ions in a saturated concrete and are not accounting
for the chloride binding capacity with solid phases in the
cement paste as reported by many authors (Mohammed
and Hamada 2003; Martin-Petez et al. 2000). Certainly,
the modeling techniques presented in this study ought to
be extended to consider such effects by some improve-
ments.
When common Portland cement is used, JSCE (2002)
gives the expression as
log Dc = −3.9 ( w c ) + 7.2 ( w c ) − 2.5
2
(1)
in which the unit of Dc is in cm2/yr. The empirical equa-
tion proposed by Boulfiza et al. (2003), which is based
146 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008

on JSCE Specification and some other sources, is similar 14 JSCE 2002

to Eq. (1) only with 14.0 substituted for 2.5, and their Luciano and Miltenberger 1999
difference of the calculated value of Dc is quite marginal. 12 Hobbs 1999
Luciano and Miltenberger (1999) developed an em-

Dc ( x10-12m2/s)
Eq. (4)
pirical formulation based on the statistical analysis of a 10 3.5
database containing chloride diffusion coefficients from Dc=30.83(w/c)
all of the concrete mixture proportions. This model can 8
account for such factors as the amount of cement, fly ash,
silica fume, granulated blast furnace slag, coarse aggre- 6
gate, fine aggregate, temperature and curing time of
concrete. If only w/c is regarded as the dominant factor, a 4
simplified equation can be derived as
2
Dc = ( 7.466 + 30.52 × ( w / c − 0.45) )
2
(2)
0
where Dc is in mm2/yr. 0.2 0.3 0.4 0.5 0.6 0.7
Hobbs (1999) proposed the relation of chloride ion w/c
diffusion coefficient for mature concrete exposed to a Fig. 4 New models about relationship between Dc and
marine environment to w/c as follows: w/c compared with previous ones.

Dc = 0.04(1166 w / c ) × 10 −12 (3)

12 Otsuki et al. 1999 Eq.(5) for mortar
where Dc is in m2/s.
Diffusion coefficient ( x10-12m2/s)

The relationship between Dc and w/c of the above Care 2003 Eq.(4) for concrete
mentioned three models is depicted in Fig. 4 in order to 10 Delagrave et al. 1996
compare with each other. Obviously, they show a similar Jensen et al. 1999
tendency with the increase of w/c. For the sake of con- 8 Oh and Jang 2004
venience, the average of the above three models is cal- Li et al. 1998
culated and estimated with a new formula as follows: 6
Dc = 30.83( w / c ) 3.5
(4)
4
The comparison of the new model proposed by Eq. (4)
with previous models is also shown in Fig. 4. 2

3.2 Diffusion coefficient of mortar

0
In terms of the diffusion coefficient of mortar, although a 0.1 0.2 0.3 0.4 0.5 0.6 0.7
lot of experiments have been carried out with a similar w/c
method used in the measurement of concrete and the Fig. 5 Effect of w/c on Dm and comparison between Dc
relationship between Dm and w/c was once believed to be and Dm.
approximately exponential (Atkinso and Nickerson
1984), till now few attempts have been done on the re-
gression of this relationship. Many authors have reported
with the same w/c in most cases. This finding is validated
that the influence of aggregate on chloride ingress is
by the series of chloride transport tests conducted si-
mainly due to the presence of an ITZ, which results from
multaneously on concrete and mortar with the same w/c
interactions between aggregate and cement paste.
(Oh and Jang 2004), which is listed in Table 1. It should
Therefore the volume fraction and size distribution of
be pointed out that Eqs. (4) and (5) are to be applied only
aggregates will influence the diffusion properties of
when the values of Dc and Dm are unknown. That is be-
concrete, especially to mortar, the diffusion coefficient
cause of the small number but large scatter of available
must differ from that of concrete for lack of coarse ag-
test data and approximation in the regression for Eq. (5).
gregates. In this paper, based on the data collected from
Moreover, previous study has also shown that the cor-
previous studies, a regression analysis is performed for
relation between diffusion coefficient and w/c is very low
mortar to give a formula to express the relationship be-
for concrete and mortar than that for cement pastes be-
tween Dm and w/c. The regression result is written in Eq.
cause of the inclusion of aggregates (Hobbs 1999). Since
(5) and depicted in Fig. 5 to compare with that of con-
one of the central objectives of this paper is to show the
crete.
feasibility of the truss network model and to give an
Dm = 15.93(w / c ) − 2.98 (5) approximate estimation of the diffusivity of ITZ, it is
more reasonable to use the test data of Dc and Dm derived
Obviously, from Fig. 5, Dm is somehow higher than Dc from a same experiment in order to reduce the regression
 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008 147

error of Eq. (5). Therefore, in the following sections, the
experimental data from Oh and Jang (2004) with w/c of
0.55 is adopted (see Table 1) to demonstrate the analysis
process of chloride ions transport.
4. Diffusivity of ITZ
4.1 Summary of previous studies
Because of the higher porosity in the ITZ, its presence
might negatively impact the durability of cement-based
materials by forming fast-conduction pathways through
the material, thus allowing accelerated ingress and
as 4.86×10-3g/cm3. The unit means chloride concentra-
tion is in grams of chloride ions per cm3 of concrete. The
profiles of chloride concentration along three dotted lines
as shown in Fig. 6 are selected as the calculation results,
and they will be used to determine the diffusion coeffi-
cient of concrete according to the theoretical solution of
Fick’s second law as given by Eq. (6) if the initial chlo-
ride concentration of concrete is assumed to be zero:
⎡ ⎛ x ⎞⎤
C ( x, t ) = C s ⎢1 − erf ⎜ ⎟⎥ (6)
⎢⎣ ⎜ 2 D t ⎟⎥
⎝ c ⎠⎦

movement of aggressive ions. To predict accurately the

where C(x,t) is the chloride concentration at distance, x,
transport properties of concrete, the ITZ should be taken
from the exposed surface; Cs is the surface chloride
into account correctly. Based on this concern, numerous
concentration; t is the exposure time and in this paper, it
studies have been carried out to investigate effects of
is also the elapsed simulation time; erf is the error func-
aggregate and ITZ on the transport properties of mortar
tion. After a certain exposure time, in Eq. (6), only Dc is
and concrete.
the unknown parameter and can be calculated by the
Delagrave et al. (1997) argued that the increased tor-
regressing method with the given chloride concentration
tuosity of the matrix induced by the presence of aggre-
profile C(x,t).
gates appears more important than the influence of ITZ,
and the interconnection of ITZ was not found to lead to a
rapid increase of the chloride ions transportation coeffi- Table 1 Experimental data (Oh and Jang 2004).
cient; that means that the overall transport properties of w/c 0.35 0.45 0.55
the mortars are influenced more by the presence of ag- Dc (×10-12m2/s) 1.485 2.425 5.475
gregates than by the presence of ITZ. Yang (2003) Dm (×10-12m2/s) 3.563 5.526 6.644
pointed out that for mortar, the approximate migration
coefficient of ITZ (DITZ) is 1.6-1.3 times of that of matrix
when the thickness of ITZ was assumed to be 10-40μm. Table 2 Results of DITZ from references.
Using the three-phase composite material model, Care
and Herve (2004) proposed a new analytical model for Thickness of
Series DITZ/Dp References
ITZ(µm)
chloride diffusion coefficient in mortar and DITZ is esti-
1 20-40 1.6-1.3 Yang 2003
mated as 16.2 times of that of cement paste without ag- 2 30 16.2 Care and Herve 2004
gregate (Dp), which was simultaneously validated by the 3 10-30 5 Delagrave et al. 1997
experimental results. Oh and Jang (2004) reported that 4 20 4-8 Oh and Jang 2004
DITZ/Dp≈7, mostly in the range of 4 to 8, based on his 5 - 6-12 Breton et al. 1992
experimental data and analytic model, moreover, it co- 6 20 2-6 Bentz et al. 1997
incides well with the other researchers’ results (Dela- 7 15-30 2-7 Shane et al. 2000
grave et al. 1997; Breton et al. 1992; Bentz et al. 1997;
Shane et al. 2000). The diffusion characteristics of ITZ Cl-
predicted from different literatures are listed in Table 2.
Although the conclusions are not in full agreement with
each other, it can be generally summarized that DITZ is
usually less than 16 times of that of cement paste, espe-
cially in most cases no more than 10 times. Moreover,
although their experimental conditions are somehow
different, e.g., a few of them are under steady-state con-
dition and others are non-steady-state migration tests, it
is considered in this study that the diffusivity of ITZ in
comparison with that of cement paste is similar among
the different conditions.

4.2 Analysis model of a specimen

Figure 6 shows the analysis model and calculating con- Voronoi elements: 1011
ditions. The volume ratio of aggregate is about 40% and Truss elements: 10094
the particle size distribution of aggregates is set as 4, 6, 8,
10mm with the same ratio of 25%. On the diffusion
boundary, the surface chloride concentration is assumed (60×60 mm)
(60×60
Fig. 6 The model and analysis conditions of the specimen.
148 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008

With regard to the effects of aggregates on the trans- 9

port properties, it may be supposed that the calculating
results using truss network model will depend on the 8

m /s)
exposure time or the amount of aggregates that chloride

2
ion may probably pass by. The reason is that, with short 7

-12
exposure time, most of the chloride ions can only dis-

Dc (x10
tribute in the surface mortar layer, but not ingress deeper 6
due to the impermeable of aggregates. However, with the
increase of exposure time, the chloride concentration in
5
the surface mortar layer will reach to a balanced state and
more and more chloride ions will penetrate through the
mortar between aggregates and the ITZ. Thus, the in- 4
fluence of surface mortar layer will be decreased and the
0 50 100 150 200 250 300 350 400
predicted Dc will approach a constant. So that, the value
of Dc at the balanced state can be regarded as the actual Elasped time (days)
diffusion coefficient for a concrete. In this study, the Fig. 7 Effect of elapsed time in simulation on Dc.
values of Dc after a series of elapsed time are obtained so
as to determine a stable analysis period. Fig. 7 shows the
time-dependent characteristics of the chloride ions 6.3
penetration in concrete. It can be observed that when the
diffusion time is longer than 400 days, the diffusion m /s) 6.0
-12 2
Target Dc=5.47x10 m /s
2

coefficient will approach a constant with the calculating

conditions depicted in Fig. 6. Then, in the later analysis, 5.7
-12
Dc (x10

the calculating parameters will be based on Fig. 6 and the

exposed time will be set as 400 days. In the previous
studies, especially for a mortar, DITZ is usually expressed
as the times of that of cement paste (Dp). Therefore, for a
concrete, DITZ can be expressed using the similar rela-
tions as the times of that of mortar (Dm), which is written
as follows:
D ITZ = nD m
4.3 Determination method of DITZ
The experimental data in Table 1 with w/c=0.55 is used,
which means Dm and Dc are assumed as constants, i.e.,
6.644×10-12m2/s and 5.475×10-12m2/s. Thus, with the
target diffusion coefficient of concrete,
Dc=5.475×10-12m2/s, the purpose of this calculation is to
determine the approximate value of DITZ. However, to
date, the thickness and the diffusion coefficient of ITZ
cannot be predicted precisely although they have been
considered in a certain range of values. Therefore, only
for the sake of simplification, the thickness of ITZ is
assumed specially to be 20μm with consideration of the
usually proposed range shown in Table 2. So that as
depicted in Fig. 8, by varying n in a range of 1-16, a
series of values of Dc can be obtained and then one may
appropriately determine the value of n corresponding to
the target value of Dc. Thus, when the thickness of ITZ is
assumed as 20μm and target diffusion coefficient of Dc is
5.474×10-12m2/s, as shown in Fig. 8, the value of n is
predicted as 8.9.
4.4 Effect of the thickness of ITZ on its diffusion
coefficient
As expected, the value of DITZ is depending on the
thickness of ITZ, although the thickness of ITZ is typi-
5.4
5.1
4.8
4.5
2 4 6 8 10 12 14 16
(7)
n=DITZ/Dm
Fig. 8 Determination method of DITZ.
cally in the range of 15 to 50μm (Shane, et al. 2000). One
reason for this large range of thicknesses is that there is
no clear-cut transition from ITZ paste to matrix paste, so
the cut-off value is somewhat subjective (Shane et al.
2000). In this section, the influence of thickness of ITZ
on diffusivity of concrete is investigated by varying the
value of thickness of ITZ in the range of 10 to 50μm,
which is shown in Fig. 9. So it is observed that when the
thickness of ITZ is changed in that range, the values of n
are estimated in rather wide range (from 15.6 to 3.8) and
decrease with the increase of thickness of ITZ (see Fig.
9), which indicates that the calculation results with the
truss network model in this paper quite coincide with
those previous research conclusions as shown in Table 2.
5. Chloride ions transport through cracks
5.1 Review of previous studies
The testing schemes for evaluation transport property of
cracked concrete can be divided into three kinds based on
the loading method: uniaxial compressive test (Saito and
Ishimori 1995; Lim et al. 2000; Samaha and Hover 1992),
 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008 149

splitting test (Wang et al. 1997; Aldea et al. 1999; Kato 16

et al. 2005) and flexural test (Gowripalan et al. 2000;
Xing et al. 2005a). The effect of microcracking on 14

n=DITZ/Dm
transport property under uniaxial compression has been 12
studied by several researchers. However, there are some 10
conflicting views pertaining to their findings. Samaha
and Hover (1992) reported that microcracks at load lev- 8
els below approximately 75% of the maximum capacity 6
do not affect mass transport properties of the concrete,
4
but their influence become pronounced when the
maximum compressive capacity of the concrete is 2
reached, attaining a 10-20% increase of permeability 0
over unloaded specimens. Saito and Ishimori (1995) 10 30 2040 50
found that the chloride permeability of concrete sub- Thickness of the ITZ (μm)
jected to static compressive stresses up to 90% of the Fig. 9 Influence of the thickness of ITZ on its diffusion
ultimate strength is nearly equal to that of concrete be- coefficient.
fore loading. Furthermore, even concrete after reaching
the ultimate strength only exhibits a comparatively small
increase in the chloride permeability. Because the char-
acteristics of the microcraks are different when a con- 60 ο
Steel sample at 20 C
crete is under a load and when it is completely unloaded, Eq. (8)
Dcr ( x10 m /s)
Lim et al. (2000) pointed out that unloading or not can 50
2

influence the chloride permeability of a concrete.

-8

Using the feedback controlled splitting test method, 40

Wang et al. (1997) and Aldea et al. (1999) pointed out
that the degree of the permeability of cracked concrete
depends on the value of the crack opening (crack width
Wcr) in the concrete. But when a crack opening dis-
placement is less than 50μm, the crack opening had little
effect on the permeability. When the crack opening dis-
placement increases from 50 to 200μm, the cracked
concrete permeability increases rapidly with a great
magnitude in comparison with that corresponding un-
cracked concrete. Based on the similar test method, Kato
and Uomoto (2005) concluded that the diffusion coeffi-
cient of chloride ions through a crack (Dcr) increased
with increases in the crack width and is almost constant
when the crack width becomes wider than approximately
75μm. In the simulation model of Takewaka et al. (2003),
if Wcr is less than 50μm, the cracks seldom affect diffu-
sivity, but when the Wcr is between 50μm and 100μm,
Dcr was set 10 times of sound part (1×10-12m2/s), and
when Wcr is larger than 100μm, Dcr was set as 103 times.
Other experimental results indicate that no significant
chloride diffusion occurs in cracks having an opening
(crack width) below a critical value, which is approxi-
mately 30μm (Francois et al. 2005; Ismail et al. 2004).
5.2 Relationship between Dcr and Wcr
In order to establish the relationship between Dcr and Wcr,
in the present study, the experimental results of chloride
ions transport through the steel specimens with a slit
(Kato and Uomoto 20005) are adopted and formulated as
follows:
log Dcr = − 2.277 [1 + 1.311exp( − 20.6Wcr )) ]
The proposed Eq. (8) and experimental results about
Dcr are shown in Fig. 10. Obviously, in a certain extent,
30
20
10
0
0.00 0.05 0.10 0.15 0.20 0.25
crack width Wcr (mm)
Fig. 10 Relationship between Dcr and crack width Wcr.
Eq. (8) can approximately express the relationship be-
tween Dcr and Wcr reported in the previous studies.
6. Chloride diffusivity of cracked concrete
6.1 Compressive loading
A series of uniaxial compression analysis by RBSM are
carried out under different stress levels in order to obtain
different levels of damage in concrete. As stated in 4.3,
the thickness and diffusion coefficient of ITZ is set as
20μm and 8.9 times of that of mortar, respectively. Fig.
11 shows the crack locations in red lines at the peak
compressive stress, in which the penetration path is par-
allel to the loading direction. The crack width is obtained
from the analysis of RBSM and Dcr is determined ac-
cording to Eq. (8). Then the numerical calculation results
(8)
under different compressive stress levels up to the
strength capacity are shown in Fig. 12 to compare with
some available experiment data, in which and elsewhere
150 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008

Compressive loading Sample_1 Sample_2

Samaha and Hover 1992: w/c=0.31
1.5
Samaha and Hover 1992: w/c=0.53
Cl- 1.4 Samaha and Hover 1992: w/c=0.46
Lim et. al 2000

Dc'/Dc0
1.3

1.2

1.1

1.0

0.9
30 40 50 60 70 80 90 100
Stress level σ/fc (%)
Fig.12 Influence of compressive stress level on the diffu-
sion coefficient of concrete.

Calculation results based on Eq. (8)

Compressive loading 5
Fig.11 Loading and diffusion direction of the compressive 1.7 Calculation results assumed Dcr=10 Dm
specimen. Sakoi and Hiriguch 2006
1.6
Dc'/Dc0

in the text, Dc0 and D’c is the diffusion coefficient of
concrete before loading and cracked concrete, respec-
tively. In Fig.12, Sample_1 and Sample_2 are two ana-
lyzed specimens with the same size, the same volume
ratio of coarse aggregates, but with different positional
arrangement, because the coarse aggregate particles are
put in and located randomly in an area based on the con-
cept of RBSM (Nagai et al. 2004). Thus, in Fig. 12, the
difference of diffusivity between both models might be
attributed to the different stress level whether the critical
stress, representing the onset of unstable microcrack
propagation during loading process, is exceeded or not,
as reported by Lim et al. (2000). Moreover, although the
calculation results in the present study seem to be in good
agreement with those experimental data found in the
literatures (see Fig. 12), it must be emphasized that the
previous researches are based on tests conducted on
specimens after they have been completely unloaded.
Studies have proved that the microcracks can close back
partially or completely upon unloading depending on the
stress level at which the concrete is subjected. Especially
when the stress level is less than 50%, the microcracks
can almost recover 100%. However, from Fig. 12, it
appears that we cannot observe the visible effect of
unloading on the transport property of concrete by the
comparison between present calculations and the previ-
ous experimental results. One cause is supposed to be
due to the lower values of diffusion coefficient in cracks.
Therefore, another case with Dcr assumed as 105 times of
that of mortar (Soda 2006) is carried out, and the calcu-
lation results are shown in Fig. 13 to be compared with
the former case and the experimental results of Sakoi and
Horiguchi (2006), whose experiments were carried out
1.5
1.4
1.3
1.2
1.1
1.0
50 60 70 80 90 100
Stress level σ/fc (%)
Fig. 13 Comparison between calculating results and
Sakoi and Horiguchi’s (2006) experimental results.
under loading conditions. Obviously, even when Dcr is
assumed as large as 105 times of that of mortar, the cal-
culation results are still much lower than that of ex-
perimental results.
6.2 Tensile loading
The diffusion coefficient evaluation of cracked concrete
induced by tensile loading is also carried out. In Fig. 14,
the cracks distribution at peak tensile stress is illustrated.
Because the cracks generally propagate perpendicular to
the loading direction under uniaxial tension, the pene-
tration path may affect the diffusivity of cracked concrete.
Therefore, two cases, perpendicular and parallel to the
loading direction, are performed respectively for the
same stress level. Thereby, the influence of penetration
direction is considered by setting two vertical exposed
sides namely as case 1 and case 2 (see Fig. 14). Fig. 15
shows the effect of penetration direction on the diffusiv-
 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008 151

ity of tensile cracked concrete. It is observed that if

1.8 Case 1
penetration is perpendicular to the loading direction (case
1 in Fig. 14), i.e., parallel to the crack propagating di- Case 2

Dc'/Dc0
rection, the diffusion coefficient is quite larger than that 1.6
of parallel to loading direction (case 2 in Fig. 14);
moreover, with the increase of stress level, the influence
of penetration direction becomes more dominant. 1.4
In Fig. 16, the comparison of calculation results in this
paper with experimental data in previous literatures
1.2
shows noticeable deviation similar as those findings
under compressive loading conditions (see Fig. 13). In
order to find out the causes for such high deviations even 1.0
when Dcr is assumed as 105 times of that of mortar, the
amount and distribution of cracks under tensile stress 40 50 60 70 80 90 100
level of 60% is illustrated in Fig. 17. From Fig. 17, it is Stress level σ/ft (%)
found that the amount of cracks is so small that the cracks Fig. 15 Effect of diffusion direction on the diffusivity of
obviously have little effect on the diffusivity of concrete tensile cracked concrete.
no matter how large the diffusion coefficient of cracks is
set. However, the experimental facts indicate that even
the stress level is between 30~60%, the diffusivity of Calculation results based on Eq. (8)
5
concrete will undergo a considerable increase. For ex- 2.6 Calculation results assumbed Dcr=10 Dm
ample, Xing et al. (2005b) reported an increase of 20% Sakoi and Horiguchi 2006
2.4
and 90% at the stress level of 30% and 60%, respectively. Xing et al. 2005b
It means that only considering the effects of cracks on 2.2 He and Gong 2005
Dc'/Dc0

transport properties of damaged concrete is not sufficient 2.0

to accurately predict the diffusivity.
1.8
7. Diffusivity model of mortar and ITZ 1.6
before cracking 1.4
1.2
7.1 Basic assumption
The available experimental data shows that, for tensile 1.0
0 20 40 60 80 100
Tensile loading Stress level σ/ft (%)
Case 2

Fig.16 Influence of stress level on the Dc under tensile

Cl- loading conditions.

Tensile loading
Case 1

Tensile loading
Case 1: Perpendicular to loading direction
Case 2: Parallel to loading direction

Fig. 14 Loading and diffusion direction of the tensile Tensile loading

specimen. Fig. 17 Cracks distribution at tensile stress level of 60%.
152 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008
condition, even at lower stresses which are less than 50%
of the tensile strength, the diffusion coefficient increases
significantly compared with the same concrete which
never been loaded (He and Gong, 2005; Sakoi and
Horiguch, 2006; Xing, et al. 2005b). On the other hand,
some experimental facts also indicate that a compressive
stress, when it is less than a certain stress level, can re-
duce the diffusivity of concrete (Gowripalan et al. 2000;
Xing et al. 2005b). For example, in Gowripalan’s ex-
periment, a maximum reduction of chloride diffusion
coefficient of 38% was observed when compared with
the stress-free condition. These effects may attribute to
the propagation of originally existing microcracks and
the connectivity of the internal pore structures under
tensile stress (Sugiyama, et al. 1996). By using the
computer tomography (CT) scanning technique, it has
been observed that the air content in concrete is in-
creased upon loading due to new nucleation or growth or
to microcracking and the increase rate becomes lager or
nonlinear when the stress level is higher than 50% of the
concrete strength (Wong and Chau 2005). As for com-
pression, it is due to the closure of microcracks or cap-
illaries as well as the consolidation of pore structures that
exist in the direction of diffusion (Sugiyama, et al, 1996;
Gowripalan, et al. 2000).
Thus, in view of the microstructural variation of con-
crete mentioned above, the diffusion coefficient on meso
scale of mortar and ITZ should be corrected in terms of
their stress values in normal spring. Therefore, the dif-
fusion coefficient of mortar and ITZ (Ds) under stress
could be written as
Ds = λD0
where D0 is the diffusion coefficient of mortar or ITZ
before stressing; λ is the coefficient for stress effect,
which is expressed empirically using a simple equation
just to account for the variation tendency of diffusivity of
mortar and ITZ
⎡ b
⎤ structures and simulate the real working conditions, an
⎛σ ⎞
λ = ⎢1 + a⎜⎜ ⎟⎟ ⎥ for σ ≥ 0
⎢⎣ ⎝ ft ⎠ ⎥⎦
−1
⎡ ⎛σ ⎞
b
⎤
λ = ⎢1 + a⎜⎜ ⎟⎟ ⎥ for σ < 0
⎢⎣ ⎝ ft ⎠ ⎥⎦
where σ is the stress of normal spring under loading
conditions; ft is the peak strength of the normal spring for
mortar (ftm) and ITZ (fti), respectively; a and b are con-
stants, which should be experimentally determined.
7.2 Determining constant a and b
Generally speaking, in the micro-scale, the air void
content affects not only the durability of concrete but
also the strength. Wong and Chau (2005) experimentally
proved the compressive strength of concrete decreases
with increasing of air content. From the micro-scale
point of view, the reason that the strength of ITZ is lower
than that of mortar is due to the dense void content and
original microcracks in ITZ. For simplicity, in this paper
an assumption is given as: when the tensile stress in
normal spring between two mortar elements equals to
(ftm-fti), the diffusivity of mortar at this stress level is
supposed to be same as that of ITZ without stress, which
means λ=n (as in Eq. (7)) with σ=ftm-fti. Thus, for mortar
under tensile stress, substituting σ=ftm-fti into Eq. (10a),
we can get the following relation:
⎡ ⎛ ftm − fti ⎞ ⎤
b
n = ⎢1 + a⎜⎜ ⎟⎟ ⎥ (11)
⎢⎣ ⎝ f tm ⎠ ⎥⎦
For instance, when w/c=0.55, along the perpendicular
direction of the specimen in Fig. 6, n has been deter-
mined as 4.1, and the corresponding values of ftm and fti
are calculated as 3.17MPa and 1.51MPa, respectively,
using the equations for strength on meso scale presented
by Nagai, et al. (2004). Thus Eq. (11) yields the appro-
priate a as
a = 3.1 × (0.524 )
−b
(12)
The value of b is determined by sensitivity analysis in
order to best fit the experimental results. Fig. 18 shows
the comparison between calculating results with different
values of constant b and the available experimental data.
It is observed that the calculation results based on the
assumption as represented by Eq. (10) and (11) can
match the experimental findings better than that shown in
Figs. 13 and 16. Lastly, it must be pointed out that due to
(9)
the scattering features of experimental data caused by
diverse test conditions, the value of constant b should be
empirically fitted and determined.
8. Conclusions
To facilitate the durability design of reinforced concrete
(10a) attempt to evaluate the chloride diffusivity of load-
ing-damaged concrete is carried out on the basis of
RBSM and truss network model, the former and latter of
which are mechanical and transport analysis respec-
(10b) tively. The following conclusions can be drawn in this
study:
(1) On the meso scale, concrete can be described as a
three-phase composite, i. e., the mortar, aggregate
particles and interfacial transition zone (ITZ), in
which it is possible to adopt different material
models, such as mechanical and transport models,
for the three phases.
(2) The advantage of RBSM is to simulate the crack
propagation, representing on two aspects: one is that
it can express the random characteristics of cracks
location and direction; the other is that it can easily
estimate the crack width by calculating the elonga-
tion of normal spring. On the other hand the advan-
 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008 153
3.5 Eq. (10): b=0.5 2.4 Eq. (10): b=0.5
Eq. (10): b=1.0 Eq. (10): b=1.0
Eq. (10): b=3.0 2.2 Eq. (10): b=3.0
3.0

Dc'/Dc0
Dc'/Dco

Sakoi and Horiguchi 2006 2.0 Sakoi and Horiguchi 2006

Xing et al. 2005b
2.5 1.8
He and Gong 2005
1.6
2.0
1.4
1.5 1.2
1.0
1.0
0.8
0 20 40 60 80 100 0 20 40 60 80 100
Stress level σ/ft Stress level σ/fc

(a) Tensile loading (b) Compressive loading

Fig.18 Comparison between experimental data and calculation results with the newly proposed Eq. (10).

tage of the truss network model is its simplicity
which makes possible to conduct two-dimensional
transport analysis by one-dimensional governing
equation. Thus, combining the techniques of RBSM
and truss network model is a good method to con-
duct diffusivity analysis of damaged concrete.
(3) Based on the previous study and experimental data,
the statistics relationship of chloride ions diffusion
coefficients of concrete (Dc) and mortar (Dm) with
w/c are proposed respectively.
(4) The diffusivity of ITZ is numerically clarified,
which is about 3-16 times of that of mortar with
variation of its thickness in the range of 10-50µm.
The present conclusion agrees very well with that of
previous studies by other researchers.
(5) Based on the literatures’ findings, the diffusion
property of chloride ions in cracks is evaluated and
the formulation of diffusion coefficient in cracks
with regard to crack width is proposed according to
the experimental results.
(6) Chloride diffusivity evaluations of concrete under
compression and tension are numerically performed
at a series of stress levels, respectively. The calcu-
lation results indicate that when the stress level is
higher than a critical value, the diffusion coefficient
of a cracked concrete will increase rapidly. This
tendency can comparatively agree with that of ex-
perimental conclusions. However, compared with
those findings under loading conditions, a noticeable
deviation is found.
(7) The reported experimental results show that the
diffusivity of concrete may increase even at com-
paratively low tensile stress and decrease under
compression. Based on this fact, an empirical equa-
tion, which can account for the microstructure
variation of concrete under loading, is proposed in
terms of the stress of normal spring. Integrating this
equation into the calculation model, a reasonable
prediction of the chloride diffusivity of concrete is
achieved.
Acknowledgements
This study was supported by the 21st Century COE
Program ‘Sustainable Metabolic System of Water and
Waste for Area-based Society’ of Hokkaido University.
Also, financial support under the Center for Concrete
Corea, Korea to the Yonsei University of Korea and the
Grant-in-Aid for Scientific Research (A) No.19206048
from Japanese Government are gratefully acknowl-
edged.
The first author would also thank to the support by
the Program of Innovative Research Team of Education
of China (No. IRT 0518).
References
Aldea, C-M., Shah, S. P. and Karr, A. (1999). “Effect of
cracking on water and chloride permeability of
concrete.” J. of Materials in Civil Engineering, 11(3),
181-187.
Atkinso, A. and Nickerson, A. K. (1984). “The diffusion
of ions through water-saturated cement.” J. of
Materials Science, 19(9), 3068-3078.
Bentz, D. P., Detwiler, R. J., Garboczi, E. J.,
Halamickova, P. and Schwartz, L. M. (1995).
“Muti-scale modeling of the diffusivity of mortar and
concrete.” In: L.O. Nilsson and J. P. Ollivier, Eds,
Proceedings of Chloride Penetration into Concrete,
Saint-Rémy-lès-Chevreuse, France, 1995. RILEM
pulications, 85-94.
Bolander, J. E. and Berton, S. (2004). “Simulation of
shrinkage induced cracking in cement composite
overlays.” Cement & Concrete Composites, 26(7),
861-871.
Bolander, J. E. and Le, B. D. (1999). “Modeling crack
development in reinforced concrete structures under
service loading.” Construction and Building Materials,
154 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008
13(1-2), 23-31.
Bolander, J. E. and Saito, S. (1998). “Fracture analyses
using spring networks with random geometry.”
Engineering Fracture Mechanics, 61(5-6), 569-591.
Boulfiza, M., Sakai, K., Banthia, N. and Yoshida, H.
(2003). “Prediction of chloride ions ingress in
uncracked and cracked concrete.” ACI Materials
Journal, 100(1), 38-48.
Bourdette, B., Ringot, E. and Ollivier, J. P. (1995).
“Modeling of the transition zone porosity.” Cement
and Concrete Research, 25(4), 741-751.
Breton, D., Ollivier, J. P. and Ballivy, G. (1992).
“Diffusivity of chloride ions in the transition zone
between cement paste and granite.” In: J. C. Maso, Ed.,
Interfaces between Cementitious Composites. E & F.
N. Spon, London, 279-288.
Care, S. (2003). “Influence of aggregates on chloride
diffusion coefficient into mortar.” Cement and
Concrete Research, 33(7), 1021-1028.
Care, S. and Herve, E. (2004). “Application of a n-phase
model of the diffusion coefficient of chloride in
mortar.” Transport in Porous Media, 56(2), 119-135.
Delagrave, A., Bigs, J. P., Oliivier, J. P., Marchand, J.
and Pigeon, M. (1997). “Influence of the interfacial
zone on the chloride diffusivity of mortars.” Advanced
Cement Based Materials, 5(3-4), 86-91.
Delagrave, A., Marchand, J. and Samson, E. (1996).
“Prediction of diffusion coefficients in cement-based
materials on the basis of migration experiments.”
Cement and Concrete Research, 26(12), 1831-1842.
Francois, R., Toumi, A., Ismail, M., Castel, A. and Vidal,
T. (2005). “Effects of cracks on local diffusion of
chloride and on long-term corrosion behavior of
reinforced concrete members.” In: T. Zhao, F. H.
Wittmann and T. Ueda, Eds., Proc. of an International
Workshop on Durability of Reinforced Concrete under
Combined Mechanical and Climate Loads (CMCL),
Qingdao, Otc. 2005, China. Freiburg: Adeificatio
Publishers, 113-121.
Gowripalan, N., Sirivivatnanon, V. and Lim, C. C.
(2000). “Chloride diffusivity of concrete cracked in
flexure.” Cement and Concrete Research, 30(5),
725-730.
He, S. Q. and Gong, J. X. (2005). “Influence of flexural
loading on permeability of chloride ion in concrete.” J.
of Building Materials, 8(2), 134-138. (in Chinese)
Hobbs, D. W. (1999). “Aggregate influence on chloride
ion diffusion into concrete.” Cement and Concrete
Research, 29(12), 1995-1998.
Ismail, M., Toumi, A., Francois, R. and Gagne, R. (2004).
“Effect of crack opening on the local diffusion of
chloride in inert materials.” Cement and Concrete
Research, 34(4), 711-716.
JSCE (2002), “JSCE standard specifications for concrete
structures (construction).” Tokyo: Japan Society of
civil Engineering, pp.80. (in Japanese)
Kato, Y. and Uomoto, T. (2005). “Modeling of effective
diffusion coefficient of substances in concrete
considering spatial properties of composite materials.”
J. of Advanced Concrete Technology, 3(2), 241-251.
Kawai, T. (1978). “New discrete models and their
application to seismic response analysis of structures.”
Nuclear Engineering Design, 48(1), 207-229.
Lim, C. C., Gowripalan, N. and Sirivivatnanon, V.
(2000). “Microcracking and chloride permeability of
concrete under uniaxial compression.” Cement and
Concrete Composites, 22(5), 353-360.
Luciano, J. and Miltenberger, M. (1999). “Predicting
chloride diffusion coefficients from concrete mixture
proportions.” ACI Materials Journal, 96(6), 698-703.
Martin-Perez, B., Zibara, H., Hooton, R. D. and Thomas,
M. D. A. (2000). “A study of the effect of chloride
binding on service life predictions.” Cement and
Concrete Research, 30(8), 1215-1223.
Mohammed, T. U. and Hamada, H. (2003).
“Relationship between free chloride and total chloride
contents in concrete.” Cement and Concrete Research,
33(9), 1487–1490.
Nagai, K., Sato, Y. and Ueda, T. (2004). “Mesoscopic
simulation of failure of mortar and concrete by 2D
RBSM.” J. of Advanced concrete Technology, 2(3),
359-374.
Nakamura, H., Srisoros, W., Yashiro, R. and Kunieda, M.
(2006). “Time-dependent structural analysis
considering mass transfer to evaluate deterioration
process of RC structures.” J. of Advanced Concrete
Technology, 4(1), 147-158.
Oh, B. H. and Jang, S. Y. (2004). “Prediction of
diffusivity of concrete based on simple analytic
equations.” Cement and Concrete Research, 34(3),
463–480.
Otsuki, N., Nishida, T., Wattanachai, P. and Chin, K.
(2006). “Experimental study on influence of ITZ
around aggregate on Cl- diffusivity of concrete.” J. of
the Society of Materials Science, 55(10), 899-904. (in
Japanese)
Otsuki, N, Yodsudjai, W, Nishida, H. and Yamane, H.
(2004). “New test methods for measuring strength and
chloride ion diffusion coefficient of minute regions in
concrete.” ACI Material Journal, 101(2), 146-153.
Ping, X., Beaudoin, J. J. and Brousseau, R. (1991). “Flat
aggregate-portland cement paste interfaces, I.
electrical conductivity models.” Cement and Concrete
Research, 21(4), 515-522.
Saito, M. and Ishimori, H. (1995). “Chloride
permeability of concrete under static and repeated
compressive loadings.” Cement and Concrete
Research, 25(4), 803-808.
Sakoi, Y. and Horiguchi, T. (2006). “Loading effects on
chloride penetration of fiber reinforced concrete.” In:
Proceedings of the 2nd International Fib Congress,
Naples, 2006, Italy.
Samaha, H. R. and Hover, K. C. (1992). “Influence of
microcracking on the mass transport properties of
concrete.” ACI Materials Journal, 89(4), 416-424.
Shane, J. D., Mason, T. O. and Jennings, H. M. (2000).
 L. Wang, M. Soda and T. Ueda / Journal of Advanced Concrete Technology Vol. 6, No. 1, 143-155, 2008
“Effect of the interfacial transition zone on the
conductivity of Portland cement mortars.” J. of the
American Ceramic Society, 83(5), 1137-1144.
Soda, M. (2006). “Chloride diffusion analysis for
cracked concrete by RBSM.” Thesis (Master).
Hokkaido University.
Soda, M. and Ueda, T. (2006). “Chloride diffusion
analysis for cracked concrete by RBSM.” Proceedings
of JCI, 28(2), 953-958. (in Japanese)
Sugiyama, T., Bremner, T. W. and Holm, T. A. (1996).
“Effect of stress on gas permeability in concrete.” ACI
Materials Journal, 93(5), 1-8.
Takewaka, K., Yamaguchi, T. and Maeda, S. (2003).
“Simulation model for deterioration of concrete
structures due to chloride attack.” J. of Advanced
Concrete Technology, 1(2), 139-146.
Wang, K., Jansen, D. C. and Shah, P. (1997).
“Permeability study of cracked concrete.” Cement and
Concrete Research, 27(3), 381-393.
Wong, R. C. K. and Chau, K. T. (2005). “Estimation of
air void and aggregate spatial distributions in concrete
under uniaxial compression using computer
155
tomography scanning.” Cement and Concrete
Research, 35(8), 1566-1576.
Xing, F., Leng, F. G. and Feng, N. Q. (2005a). “Chloride
diffusivity of plain concrete subjected to sustained
flexural loading.” In: T. Zhao, F. H. Wittmann and T.
Ueda, Eds., Proc. of an International Workshop on
Durability of Reinforced Concrete under Combined
Mechanical and Climate Loads (CMCL), Qingdao,
Otc. 2005, China. Freiburg: Adeificatio Publishers,
133-137.
Xing, F., Leng, F. G. and Feng, N. Q. (2005b). “The
influence of long-term load on the chloride
permeability in reinforced cocnrete.” In: T. Zhao, F. H.
Wittmann and T. Ueda, Eds., Proc. of an International
Workshop on Durability of Reinforced Concrete under
Combined Mechanical and Climate Loads (CMCL),
Qingdao, Otc. 2005, China. Freiburg: Adeificatio
Publishers, 139-147.
Yang, C. C. (2003). “Effect of the interfacial transition
zone on the transport and the elastic properties of
mortar.” Magazine of Concrete Research, 55(4),
305-312.