Professional Documents
Culture Documents
(2011, A) Enhanced Shrinkage Model Based On Early Age Hydration and Moisture
(2011, A) Enhanced Shrinkage Model Based On Early Age Hydration and Moisture
11, 360-373, December 2013 / Copyright © 2013 Japan Concrete Institute 360
Scientific paper
Abstract
In this research, an existing multi-scale model for shrinkage behavior is enhanced to improve its applicability and preci-
sion. The enhancement work focuses on the initial autogenous shrinkage and intrinsic driving forces of the shrinkage at
the microscale. The existing model underestimates the rapid development of the autogenous shrinkage of cement with a
low water/cement (w/c) ratio. Hence, combined with self-desiccation, the autogenous shrinkage at an early age is first
discussed. As an approximation, a portion of the autogenous shrinkage at an early age is quantitatively calculated using
the chemical volume change and distance between cement particles, and that portion is added to compensate for the
underestimation in the existing model. Furthermore, in the existing model, capillary tension is assumed to be the princi-
pal driving force in all of the pores with various sizes, which would cause an overestimation of the long-term shrinkage
in the analysis. Therefore in the enhanced model, the driving forces are discussed and modified. The capillary tension is
assumed to only be active in relatively coarse pores, whereas the disjoining pressure dominates in fine pores on the
nanoscale. These two driving forces are quantified by the water status in the pores using the proposed formulas. With
the above enhancements, the autogenous and drying shrinkage behaviors with various w/c ratios and relative humidity
can be simulated reasonably.
forces of the shrinkage are quantified and applied to the ities are calculated according to the hydration degree,
cement matrix. In addition, a two-phase stiffness model chemically bound water, intrinsic porosity of the gel
is established to describe the deformation response. As grains, and surface area. The size distributions of the
long as the mix proportion, specimen dimensions, and capillary and gel pores are described by a Raleigh-Ritz
ambient conditions are input, the multi-scale constitu- distribution function.
tive model is theoretically capable of simulating auto- The water retaining capacity of the pore structure at a
genous or drying shrinkage simply by altering the given RH is calculated by considering the thermody-
boundary condition to a sealed or exposed condition. namic equilibrium of the liquid and vapor. First, as a
This methodology has been proven to be effective, and result of the surface tension, a capillary meniscus is
the shrinkage trends can be approximately traced. None- formed and a pressure difference is generated across the
theless, for practical application, an improvement in the vapor-liquid interface. Assuming the micropores are
precision is requisite, especially for low w/c concrete. cylinder-shaped, the pressure difference Pl (Pa) is de-
Therefore, in this study, the authors conduct a further scribed as
investigation of shrinkage on the basis of microscale
phenomena. Accordingly, the model is refined and im- 2γ
Pl = (1)
proved. Before that, the existing model is briefly intro- rc
duced in Section 2.
where γ is the surface tension of water (N/m) and rc is
2. Scheme of multi-scale constitutive the pore radius (m) where the interface is present. In
model addition, the pressure difference can also be written as
the Kelvin equation
2.1 Water equilibrium and transport in micro-
pores ρRT pvap
Pl = ln * (2)
The bases of the model are the pore structure and inter- M p
nal water status, which are simulated by a computa-
tional system called DuCOM (Durability COncrete where ρ is the density of liquid water (kg/m3), R is the
Model) (Maekawa et al. 2003, 2009). In this system, the gas constant (J/mol・K), T is the absolute temperature
hydration reaction, pore-structure formation, and mois- (K), M is the molecular mass of the water (kg/mol), pvap
ture equilibrium and transport in a cement matrix are is the actual vapor pressure, and p* is the saturated va-
coupled and simulated comprehensively. Figure 1 por pressure (Pa). The ratio p*/pvap is the relative hu-
shows the scheme of the pore-structure model. The mi- midity with the mark h. Combining Eq. 1 and Eq. 2, the
cropore information essentially consists of two parts: relationship between h and rc is obtained as
the porosity and pore size distribution. The micropores 2γM 1
in the model are categorized into three types according ln h = − ⋅ (3)
to the size and formation mode: capillary pores, gel RTρ rc
pores, and interlayer pores. During the hydration, gel
It is assumed that pores smaller than rc are totally
grains are generated inside and outside the cement parti-
filled by water. Pores larger than rc are partially filled
cles. The gel grains precipitating outside fill in large
because of the ink-bottle effect. Moreover, physically
voids, while the remaining voids are categorized as cap-
adsorbed water exists in the pores that are not filled by
illary pores. An individual gel grain is considered to
condensed water. According to the modified BET ab-
contain much smaller voids, called gel pores. Moreover,
sorption theory by Hillerborg (1985), the thickness of
gel grains consist of numerous layers, with extremely
adsorbed layer, ta, is calculated by the equation
tiny spaces between these layers, called interlayer pores
(the size is assumed to be 0.28 nm), each of which is 0.525 × 10−8 h
accessible only by a single water molecule. The poros- ta = (4)
(1 − h / hm )(1 − h / hm + 15h)
⎛ −γ M ⎞
hm = exp ⎜ ⎟ (5)
⎝ ρ RTr1 ⎠
where r1 = rc – ta is the residual value of the radius rc
minus the thickness of the adsorbed water layer. By
combining this with the pore size distribution, the satu-
ration of the gel or capillary pores S can be calculated
Fig.1 Pore-structure model in DuCOM. using the three portions of water as follows (Fig. 2)
Y. Luan and T. Ishida / Journal of Advanced Concrete Technology Vol. 11, 360-373, 2013 362
ta
r1 r Pore size distribution pressure acts. φcp is the capillary porosity and φgl is the
dV/dlog(r)
dv/d lnr gel porosity. Scp and Sgl are their saturations. Thus, β
Vapor equals the average saturation.
Trapped condensed water
(ink-bottle water) Adsorbed
Liquid
As drying continues and the internal RH falls to a low
water
water level, not only the condensed water but also the ad-
rc sorbed water starts to evaporate. According to statistical
thermodynamics, when the internal RH is lower than
rc : Liquid-vapor interface pore radius
Scrc : Saturation of pores smaller than rc SSrcc Sink, Sads 30%, the adsorbed water layer is only several molecules
S
Sads : Saturation due to adsorbed water thick or even vanishes (Brunauer et al. 1969; Wittmann
Sink : Saturation due to ink-bottle water 0 rc Pore radius
1973). The interaction between the solid molecules of
Drying process: Srcc +Sads+Sink
Scp, gl =S the gel grains, i.e., the Van der Waals force, becomes
non-negligible and the surface energy increases. Ac-
Fig.2 Moisture state in pore structure during drying proc- cordingly, the distance between the gel grains decreases
ess. and shrinkage occurs. The driving force from the sur-
face energy changes is expressed as stress σsd, and cal-
rc rmax culated as follows:
S = Sc + Sink + Sads = ∫ dV + ∫ f r dV + Sads (6)
rmin rc
σ sd = S sd ⋅ γ gl (8)
where Sc is the saturation fraction of the pores smaller
than rc. Sink is the saturation fraction of the ink-bottle where γgl is the surface tension of the gel grains, which
water in pores larger than rc. Sads is the saturation frac- is assumed to be 350 mN/m, according to Hori (1962).
tion of the adsorbed water. rmin and rmax are the mini- Ssd is the distribution density function (1/m). Ssd is re-
mum and maximum radii. fr is a coefficient reflecting lated to the total surface area of the gel pores and the
the ink-bottle effect. thermodynamic status of the absorbed water. It is quan-
The moisture transport in micropores involves water tified using the equation
and vapor movement. The flux potential is dominated Ssd = f (h) S pore (9)
by gradients of water pressure and temperature, and the
water conductivity is quantified based on the pore struc- where Spore is the specific surface area of the gel pores
ture and saturation. The details are omitted here. (1/m). h is the internal RH, and f(h) is the thermody-
namic function of the absorbed layer. Without the ad-
2.2 Driving forces of shrinkage in cement ma- sorbed water, the surface energy reaches its maximum
trix value, and f(h) is equal to 1.0. When the absorbed water
In the model, shrinkage is assumed to be caused by cap- is saturated, f(h) is assumed to be 0.0. Essentially, the
illary tension and surface energy change (Zhu et al. thermodynamic status falls between these two cases.
2004) (Fig. 3). The capillary tension is related to the Referring to the BET theory, f(h) is quantified using the
condensed water, whereas the surface energy change is equation
induced by the desorption of the absorbed water. As
indicated by Eq. 2, the formation of a meniscus causes a Vm − V1 1 − h / hm
negative pressure in the liquid water (the vapor pressure f ( h) = = (10)
Vm 1 − h / hm + 15h
is quite small and set to zero), which is the derivation of
the capillary tension. Therefore, the driving force from where V1 is the water volume of the first adsorbed layer
capillary tension, if written as the stress σsc, is given by nearest to the gel pore surface and Vm is the water vol-
ume when the first adsorbed layer is fully occupied by
φcp ⋅ Scp + φgl ⋅ S gl
σ sc = β ⋅ Pl = ⋅ Pl (7) water molecules.
φcp + φgl
2.3 Time-dependent deformation response un-
where β denotes the effective field in which the negative der shrinkage stress
When the stress obtained by Eq. 7 and Eq. 8 is applied
Condensed water Absorbed water
to a cement matrix, the strain deformation can be calcu-
lated using a time-dependent response law (Zhu et al.
Mechanism of capillary tension Mechanism of surface energy 2004; Asamoto et al. 2006, 2008). Concrete is idealized
Water
Atmospheric pressure
molecular Desorption as a two-phase composite, which consists of a cement
Surface
tension
of water
matrix and aggregate (sand and gravel). The aggregate
Gel
Pore
grains is assumed to be rigid and to exhibit an elastic deforma-
Negative
Negative
pressure pressure
Negative
pressure
tion with a linear law. The shrinkage caused by aggre-
Gel surface energy: low high gates is also taken into account (Asamoto et al. 2008). A
Negative pressure caused Distance between gels: long short
by liquid capillary tension Increases of gel surface energy
cement matrix is assumed to be an assembly of fictitious
clusters, referring to the solidification theory by Bazant
Fig.3 Scheme of driving forces in existing model.
Y. Luan and T. Ishida / Journal of Advanced Concrete Technology Vol. 11, 360-373, 2013 363
and Prasannan (1989). As the hydration proceeds, the initial days, after which the slopes tended to be gentle.
amount of clusters increases. The load capacity of a Conversely, the analysis shows a continuous increase
cement matrix is the summation of the capacities of all for a long period. Therefore, a great underestimation
of the clusters. An individual cluster exhibits a time- occurs at the early age. For low w/c concrete, extensive
dependent deformation relevant to the history. This previous studies (Yang et al. 1998; Miyazawa et al.
time-dependent deformation consists of elastic, viscoe- 1999; Nawa and Horita 2004; Jiang et al. 2005) found
lastic, viscoplastic, and plastic components, which are the distinct feature that the autogenous shrinkage devel-
simulated based on the water status in the simulated ops rapidly and drastically at an early age. As for the
capillary, gel, and interlayer pores. The details are omit- analysis, the water consumption and internal RH decline
ted here. Herein, it needs to be emphasized that the re- produced by the self-desiccation are both automatically
sponse law can treat creep as well as shrinkage defor- calculated by the coupled hydration and pore-structure
mation, because the load applied to the clusters can be formation models. Then, the shrinkage stress is calcu-
external or shrinkage stress. Attempts to modify the lated using the water status, after which the shrinkage
response law will influence both the shrinkage and the strain is obtained. For low w/c concrete, this calculated
creep deformation. The creep behavior has been verified self-desiccation is a gradual process rather than a drastic
in previous studies, and reasonable results have been development. This is the reason that the shrinkage in the
obtained. Therefore, this study essentially focuses on the analysis increases gradually. In the existing model, fac-
driving forces of shrinkage (see Section 2.2). The re- tors relevant to self-desiccation, such as the hydration
sponse law is not considered at all. degree, porosity, and internal RH, have been extensively
investigated and verified (Maekawa et al. 2003, 2009;
3. Discussion and modification of existing Ishida et al. 2007; Nakarai et al. 2007). Therefore, the
model overall self-desiccation process in the simulation could
be regarded as reasonable, and it is infeasible and ineffi-
3.1 Simulation of autogenous shrinkage using cient to tackle it. The drastic increase in shrinkage is
existing model often encountered in reality, and so it is necessary to
As introduced, the pore structure and internal water clarify the reason for it on the basis of the current self-
status are quantified at the microscale level, and the desiccation simulation and driving forces, and provide
driving forces of shrinkage are explicitly derived from an appropriate modification.
the pore pressure difference, saturation degrees, and
water volume of the absorbed layer, for any self- 3.2 Portion of autogenous shrinkage quantified
desiccation or ambient exchange. This implies that by chemical shrinkage
given the initial material and ambient temperature, by According to the JCI Autogenous Shrinkage Committee
setting the moisture exchange condition as isolation or (Tazawa, 1996), chemical shrinkage is defined as the
exposed, it is possible to freely simulate autogenous or absolute volume change in hydrates from unhydrated
drying shrinkage. In previous studies, this methodology cement and water. In contrast, autogenous shrinkage is
has proved to be effective, and the shrinkage trends defined as the macro-dimension or volume changes.
have been roughly verified. Here, the authors focus on When concrete is fluid, the chemical shrinkage can be
the precision. First, an analysis of autogenous shrinkage completely converted to a macro-volume change. After
with a low w/c is conducted and compared with a test the skeleton is completely formed, the matrix becomes
from Tazawa and Miyazawa (1997) (Fig. 4). The rigid and the macro-volume change can be resisted.
shrinkage value before the initial setting is deduced, as Therefore, nearly all of the chemical shrinkage is con-
was done in this test. An obvious difference can be ob- verted into internal pores. Actually, just after the initial
served. In the test, the shrinkage increased rapidly in the setting, a significant portion of the chemical shrinkage
is converted to a macro-volume change. This conversion
is a result of the combination of the self-desiccation and
1200 uncompleted skeleton. A meniscus is formed by self-
Autogenous shrinkage (micro)
Analysis Test
1000 desiccation in the forming skeleton, and the driving
23% 23% W/C ratio force of capillary tension is generated. Furthermore, just
800 30% 30% after the initial setting, the cement matrix is in the proc-
600
ess of transforming from a liquid to a solid. Thus, the
skeleton is far from completion. It could not resist even
400 a small driving force, and a relatively large deformation
200
occurs, resulting in the observed drastic shrinkage in-
Tazawa and Miyazawa 1997 crease. Theoretically, this drastic increase could be de-
0 scribed quantitatively by means of the shrinkage stress
0 20 40 60 80 100 and constitutive law at an early stage. However, in the
Age (days) analysis, although the self-desiccation and shrinkage
Fig.4 Analysis of autogenous shrinkage with existing stress can be simulated, it is very difficult to precisely
model.
Y. Luan and T. Ishida / Journal of Advanced Concrete Technology Vol. 11, 360-373, 2013 364
W/C = 30%
nous shrinkage at an early stage using chemical shrink-
0.02
age. This portion of the autogenous shrinkage can be
applied as the initial strain to compensate for the under- W/C = 50%
0.01
estimated portion in the existing model. In this study,
Blain value = 3100 cm2/g
the authors adopt this approach as an approximation. 0.00
Hence, the authors evaluate the underestimated por- 0 2 4 6 8 10 12 14 16
tion in Fig. 4 by means of chemical shrinkage. This por-
Average distance of cement particles 2⋅ δ m (μm)
tion from chemical shrinkage depends on the w/c, be-
Fig.6 Relationship between average distance 2δm and
cause lower w/c concrete usually exhibits a higher auto-
influential function f(δm).
genous shrinkage. The authors attempt to explain the
influence of the w/c based on a hypothesis at the micro-
JCI Committee’s definition, νch is calculated using the
scale (Fig. 5). At the beginning, the cement particles are
volume change from liquid water to chemically bound
dispersed in water, and the average distance depends on
water in the equation
the w/c. After the initial setting, a portion of the chemi-
cal shrinkage induces micropores, while the residual vch = Wch /(1/ ρl − 1/ ρ ch ) (12)
amount causes autogenous shrinkage. For a high w/c,
the average distance between the cement particles is where Wch is the weight of the chemically bound water
relatively large. Thus, the particles hydrate independ- in a unit volume of concrete (kg/m3). ρl and ρch are the
ently. For a low w/c, the distance decreases and some densities of the liquid water and chemically bound water
particles even contact each other. With the existence of (kg/m3), respectively. The value of ρch is assumed to be
self-desiccation, they may tend to move toward each 1.25 × 103 kg/m3. f(δm) is given by the empirical equa-
other more significantly by means of the connection or tion
accumulation of local deformations, resulting in a sig-
nificant macro-volume change. f (δ m ) = 0.045 ⋅ exp(− a ⋅ δ m b ) (13)
Thus, the portion of autogenous shrinkage εch quanti-
fied by chemical shrinkage can be described by the where a and b are constants with values of 1.2 × 10-4
equation and 6.0, respectively. For a normal w/c, f(δm) is ex-
tremely small (Fig. 6). As the w/c falls into a low range,
ε ch = vch ⋅ f (δ m ) (11) the particle distance 2δm decreases greatly. Thus, f(δm)
increases, and the contribution of chemical shrinkage
where νch is the chemical shrinkage. δm is the maximum becomes non-negligible. In the extreme case where δm is
thickness of the precipitation layer outside a single ce- zero, f(δm) reaches a maximum. Although it does not
ment particle (m), which is calculated using a stereo- occur in reality, the physical meaning is that 4.5% of the
logical model (Maekawa et al. 2009). 2δm has the physi- chemical shrinkage is converted to autogenous shrink-
cal meaning of the average distance between cement age.
particles. f(δm) is the influence function. Following the With this portion of autogenous shrinkage added as
Y. Luan and T. Ishida / Journal of Advanced Concrete Technology Vol. 11, 360-373, 2013 365
an initial strain to the existing model, the analysis in Fig. cases of low w/c. In general, the driving forces in the
4 is carried out again (Fig. 7). As anticipated, the existing model need to be specifically discussed.
shrinkage strains increase rapidly and agree with the test
results until 20 days. However, another feature in the 3.4 Discussion of mechanisms concerning driv-
test is that the shrinkage strains after the initial days ing forces
tend to be gentle or even stagnate. In contrast, the analy- According to Eq. 7, the driving force from capillary
sis strains continue to increase, with large slopes, over- tension is computed by multiplying the negative pres-
estimating the long-term values. The low w/c matrix has sure by the effective factor β. The negative pressure
a relatively large Young’s modulus and high deforma- dominates the water movement and is solved using the
tion resistance, which may be one reason for the stagna- governing equation of moisture transport. Past verifica-
tion of shrinkage in the test. Moreover, from the view- tions (Maekawa et al. 2003, 2009; Ishida et al. 2007)
point of porosity, for a low w/c, most of the capillary have already proven that the negative pressure is reli-
pores are filled in reality. It is rational to consider that in able. Therefore, the discussion focuses on β, which is
reality, the water in the capillary pores is inadequate to the average saturation of the capillary and gel pores. In
provide sufficient driving forces after this. However, the other words, the water in both the simulated gel and the
analysis implies that the driving force increases con- capillary pores is assumed to contribute to the capillary
tinuously. Thus, some discrepancy still seems to exist. tension. However, the capillary pores are usually on the
with w/c is found, and for a low w/c case, the shrinkage X Y
is greatly overestimated as well. Combined with Fig. 7, W Drying
surface
it is suspected that a common point exists for the auto- FEM mesh
genous and drying shrinkages with low w/c ratios. Al- Fig.8 FEM mesh for analysis under drying condition.
though it declines as a result of self-desiccation, RH still
lies in a relatively high-level field, whereas during the 1200
above drying, the internal RH is higher than 60% before
Shrinkage strain (micro)
micro scale, while the gel pores are much finer, and a The authors assume that disjoining pressure is appro-
large portion lies in the nanometer range. It is probable priate as the driving force in nanoscale pores. This the-
that such a large difference in pore size may induce sig- ory was originally proposed by Derjaguin (1940; 1974)
nificant property variation in the internal water. Thus, and defined as the action of the liquid absorbed in a
the capillary tension may only be the principal driving surface toward the surface. When it is applied to a ce-
force in a portion of the pores. On the other hand, as ment matrix, the disjoining pressure is assumed to be
mentioned in Sections 3.2 and 3.3, the driving force generated by the water in the narrow spaces, and acts as
may be overestimated for a low w/c. In these cases, rare the repulsive force toward gel particles (Powers 1968;
capillary pores remain, and the primary portion consists Bazant and Wittmann 1982). When the internal RH de-
of gel pores. Therefore, a natural inference is that the creases, the water starts to evaporate, and some disjoin-
capillary tension may be inappropriate as the driving ing pressure disappears. Accordingly, the repulsive force
force in gel pores at the nanometer scale. In addition, decreases. Thus, the distance between gel particles de-
while β is expressed by the average saturation, strictly creases and shrinkage occurs. For nanopores, because of
speaking, the absolute water volume seems to be a more the tiny size and tremendous surface area, the disjoining
appropriate index. With the same negative pressure and pressure can be thought of as the dominant mechanism.
saturation, it can be inferred that a matrix containing Furthermore, the surface energy change in the exist-
more water may exhibit larger shrinkage, because the ing model and its difference from the disjoining pres-
acting field of water increases. sure need to be clarified. The two theories both describe
Based on the above discussion, the driving force is re- the mechanism in nanopores. The surface energy change
established. Most of the capillary pores are at the micro is non-negligible only when the adsorbed water layer is
scale in reality. Therefore, it is reasonable to assume that extremely thin or completely desorbed. Instead, the dis-
capillary tension is active. Conversely, for gel pores at joining pressure is present as long as the water (con-
the nanometer scale, capillary tension may not be the densed and absorbed) exists. From the viewpoint of RH,
dominant mechanism. One proof is that in fine pores, the surface energy change is active at a very low RH,
physically absorbed water occupies a larger portion. whereas the disjoining pressure can dominate in a wider
Thus, there is a decline in the capillary tension, which is field. When the water is completely removed, the driv-
related to the decreased portion of condensed water ing force from the disjoining pressure reaches the
(Hansen 1987; Garci Juenger and Jennings 2002; Bentur maximum. In some sense, it can be regarded as being
et al. 1979). Beltzung and Wittmann (2005) pointed out equivalent to the surface energy change at complete
that in nanometer pores, the capillary tension is much desorption. In the enhanced model in this study, the dis-
smaller than the value calculated by the Kelvin equation. joining pressure is adopted to replace the surface energy
Additionally, in fine pores, the water flow may no change, as well as the capillary tension in nanopores.
longer be continuous. Water is confined by the nano-
space and becomes a set with only tens of molecules 3.5 Simulation of driving forces by capillary
(Bruni et al. 1998; Farrer and Fourkas 2003). Thus, wa- tension and disjoining pressure
ter movement may be discrete and closer to particles. In The newly proposed scheme in the simulated pore-
short, it is more reasonable to treat the driving force in structure is shown in Fig. 10. First, the radius rc intro-
nanopores separately. duced in Eq. 3 is used to distinguish the water distribu-
Skeleton Force F0
• Group S are saturated
• Equilibrium state rbr rc Pore radius log(r)
F0 dV/dlog(r)
dV/dlnr Low RH, rc<rbr
Water in Group L
Ft Capillary tension
Ft dominates
F0
• Group S starts to lose water
• F0 > Ft Driving force occurs
rc rbr Pore radius log(r)
tion at the unsaturated state. According to Fig. 2, in the where K is the intrinsic magnitude of the disjoining
simulated pore structure, pores smaller than rc are to- pressure and only depends on the material of the solid
tally saturated, whereas larger pores are partially filled surfaces and the liquid inside (Pa). λ is the decay length
by the ink-bottle and physically absorbed water. In order (nm). k is the distance between the surfaces (nm). Ma-
to separate the action fields, a so-called border radius rbr ruyama (2010) derived a K value of 4500 MPa for the
is suggested. In pores larger than rbr (abbreviated as cement matrix from a statistical thickness disjoining
Group L), the capillary tension dominates. As to pores model.
smaller than rbr (abbreviated as Group S), the disjoining Based on Eq. 16, if the pores are assumed to be cyl-
pressure dominates. Because rbr should be on the nano- inder-shaped, the disjoining pressure in an individual
meter scale, it is assumed to be 10 nm. At a high RH, rc pore depends only on the radius. The total repulsive
is larger than rbr, and thus, Group S is totally saturated. force should be a complex of the disjoining pressures in
The repulsive force from the disjoining pressure sus- all of the pores. Thus, integration should be performed.
tains the equilibrium with the skeleton force. Therefore, The pore size distribution, surface area, and water con-
the driving force only occurs in Group L by the capillary tent are necessary, and they can be simulated as intro-
tension. As RH decreases, the vapor-liquid interface duced in Section 2.1.
degrades toward nanopores, and rc becomes smaller First, when Group S is saturated, the repulsive force
than rbr. Group S starts to lose water, and a portion of maintains the equilibrium with the skeleton force.
the disjoining pressure disappears. Hence, the repulsive Therefore, the repulsive force equals F0 in Eq. 15, and is
force in Group S decreases and cannot sustain the bal- quantified as
ance with the skeleton force. As a result, the driving rbr
force in Group S is activated.
Hence, the shrinkage stresses are re-quantified. The ∫ Π(2r )dχ
rmin
driving force in Group L is given by F0 = φS ⋅ (17)
χ0
σ sc = A ⋅ VL ⋅ Pl = A ⋅ (Vcp _ L + Vgl _ L ) ⋅ Pl (14)
where φS is the volume fraction of Group S, as illus-
where VL is the water volume in Group L in the cement trated in Fig. 11. χ0 is the total surface area of Group S.
matrix (m3/m3). Vcp_L and Vgl_L are the water volumes in Following Eq. 16, П(2r) is the disjoining pressure in
capillary and gel pores larger than rbr (m3/m3). Therefore, pores with an arbitrary radius r. dχ denotes the infini-
Vcp_L and Vgl_L are subsets of VL and are illustrated in Fig. tesimal increment in the pore surface area for radius r.
11. A is an influential constant equal to 8.0. Hence, in Eq. 17, the integral value from rmin to rbr is the
For Group S, the driving force is derived from the resultant force of the disjoining pressure when Group S
skeleton force and repulsive force caused by the disjoin- is saturated. With the normalization by the surface area,
ing pressure. This is written as follows the repulsive force F0 is calculated and equals the skele-
ton force.
Δ F = F0 − Ft (15) Next, the repulsive force at the unsaturated state is
expressed by the equation
where F0 is the skeleton force, which equals the repul-
rbr
sive force when Group S is saturated. Ft is the repulsive rbr
10 nm. These values are different from Maruyama’s agreement with the test results for various w/c values.
data, because the approaches and thermodynamic vari- The phenomenon that the shrinkage becomes smaller
ables are not completely the same. for a lower w/c is grasped. This improvement was
For an unsaturated state, the internal water in Group S caused by the separation of the driving forces. For an
consists of two portions. All of the pores smaller than rc RH higher than 60%, the effect of the disjoining pres-
are water-filled, whereas larger ones are partially filled sure is not significant. Thus, the capillary tension domi-
by ink-bottle water. Therefore, using the coefficient fr in nates, which decreases as the w/c decreases.
Eq. 6, Eq. 18 can be expressed as Finally, for a normal w/c (50%), the results of analy-
ses using the existing and enhanced models are com-
rc rbr
pared. The test was carried out by Kanda et al. (2008)
∫ Π(2r )dχ + ∫
rmin rc
f r ⋅ Π(2r )dχ
and also followed JIS A-1129. The shrinkage was meas-
Ft = φS ⋅ (19) ured by the comparator and laser methods simultane-
χ0 ously, and identical results were obtained. As shown in
Fig. 15, the analysis with the existing model underesti-
By combining Eq. 15, Eq. 17, and Eq. 19, the driving
mates the value within tens of days, even though the
force can finally be written in relation to the stress σsd as
follows:
60
∫ Π(r )dχ − ∫ f r
⋅ Π(r )dχ 20
rc rc
= φS ⋅ 10
χ0
0
By combining this with the capillary tension in Eq. 14, 0.6 0.7 0.8 0.9 1.0 0.3 0.4 0.5
the total driving force σs is obtained as Relative humidity
Fig.12 Driving forces of shrinkage under various RHs.
σ s = σ sc + σ sd (21)
1200
The relationships between the driving forces and the
Autogenous shrinkage (micro)
Analysis Test
internal RH are shown in Fig. 12. The total forces in- 1000
23% 23% W/C ratio
crease as RH decreases. The capillary tension already 800 30% 30%
dominates at an RH field higher than 80%. Conversely,
the disjoining pressure is activated at around RH 80%, 600
and contributes principally at mid-low RH values. For a 400
low w/c, the capillary tension tends to decline. Thus, the
total driving force decreases. 200
0
4. Simulation of shrinkage behaviors using 0 20 40 60 80 100
enhanced model Age (days)
Fig.13 Analysis of autogenous shrinkage with enhanced
4.1 Verification of autogenous and drying model.
shrinkages
With the above enhanced model, the autogenous shrink- 1200
age in Fig. 7 is calculated again, and is shown in Fig. 13.
Shrinkage strain (micro)
After the initial days, the analysis results gradually in- 900
crease, which agrees better with the test results. In this
case, the simulated capillary pores are abundantly filled. 600
Thus, the water in Group L is insufficient to produce
adequate capillary tension. On the other hand, the inter- W/C ratio
300
nal RH remains in the high field, and Group S is almost 25% 35% 55% 65% Analysis
saturated. Therefore, the disjoining pressure is not active. 25% 35% 55% 65% Test
In the enhanced model, these factors are effectively 0
taken into account. 0 50 100 150 200 250 300 350 400
The shrinkages in Fig. 9 are also re-calculated and Drying Time (days)
shown in Fig. 14. Clearly, the results are in better Fig.14 Analysis of shrinkage behavior under drying con-
dition with enhanced model.
Y. Luan and T. Ishida / Journal of Advanced Concrete Technology Vol. 11, 360-373, 2013 369
Kanda et al 2008 the range of 95% to 20%. The water loss and length
600 change were measured simultaneously until the equilib-
rium state. In addition, some specimens were dried at
400 105°C, and the remaining weight was defined as the dry
mass. For the shrinkage, the length before drying was
by the enhanced model
by the existing model
set as the original length. Exactly the same conditions
200
test by comparator were utilized in the analyses.
test by laser The water contents and shrinkages at the equilibrium
0
states are shown in Fig. 16. For the water contents in
0 40 80 120 160 200 Fig. 16(a), although a slight deviation occurs for a w/c
Drying Time (days) of 55% with an RH of 80%, in general, they are well
Fig.15 Comparison of drying shrinkage precision be-
simulated. The shrinkages with the existing and en-
tween existing model and enhanced model. hanced models are shown in Figs. 16(b) and (c), respec-
tively. The results of another analysis with a w/c of 25%
final value is satisfactory. In contrast, the analysis with are plotted to make it easier to comprehend. With the
the enhanced model is consistent with the test for the existing model, almost the same results are obtained
whole process. regardless of the w/c. In contrast, with the enhanced
model, the difference in relation to w/c can be repro-
4.2 Shrinkage behaviors of cement matrix un- duced, which is consistent with the test results. The
der various RH values comparison of the driving forces in Fig. 17 reveals the
The above verifications of drying shrinkages were con- reason for the improvement. In the existing model, the
ducted at the standard RH of 60% from JIS A-1129. total driving force increases as w/c decreases. This is
Here, a wider RH range is investigated. The test results because all of the simulated gel and capillary pores are
from Maruyama (2010) are used for the verification. assumed to contribute to the capillary tension, and the
Cement paste specimens with w/c values of 55% and driving force is evaluated using pore pressure and their
40% were cast. These were 3 × 13 × 300 mm in size to average saturation. For a low w/c, the total simulated
alleviate RH gradients along the depth. The specimens porosity decreases greatly as a result of filling the capil-
were water-cured for 91 days, and can thus be regarded lary pores, while variation rarely occurs in the fine pores
0.30 0.000
Water content (g/g-dry mass)
-0.002
25%
-0.004
40% Analysis
W/C ratio 55%
-0.006
40%
Test
55%
-0.008
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Relative humidity
(c) Drying shrinkage with enhanced model
Fig.16 Analysis of water content and drying shrinkage in equilibrium state under various RHs.
Y. Luan and T. Ishida / Journal of Advanced Concrete Technology Vol. 11, 360-373, 2013 370
100
40 40
20 20
0 0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.5 0.6 0.7 0.8 0.9 1.0 0.3 0.4
Relative humidity Relative humidity
(a) Driving forces with existing model (b) Driving forces with enhanced model
Fig.17 Driving force of shrinkage under various RHs.
dV/dlog(r)
Total pore distribution (dV/dlnr)
55%
analysis.
0.08 W/C ratio 40%
W/C (%) 55 40 25
25%
Young’s modulus (GPa) 14.3 21.9 25.8
0.06
(Fig. 18). As a result, with the same RH and rc, it shows 0.04
a higher saturation but lower water contents. Following
Eq. 7, a higher driving force is obtained. The Young’s 0.02
moduli at 91 days in the analysis are shown in Table 2.
The increase in the driving force is neutralized by the 0.00
higher modulus. Thus, similar shrinkages are obtained, -9 -8 -7 -6 -5 -4
as shown in Fig. 16(b). With the enhanced model shown Log(r[m])
in Fig. 16(c), the discrepancy is effectively amended. Fig.18 Pore distribution at age of 91 days.
On the whole, it can be concluded that the shrinkage
0.000
behaviors can be appropriately simulated using the en-
Uniaxial volume strain (m/m)