ACI Materials Journal/March-April 2007 129

ACI MATERIALS JOURNAL TECHNICAL PAPER

ACI Materials Journal, V. 104, No. 2, March-April 2007.
MS No. M-2005-334 received December 5, 2005, and reviewed under Institute
publication policies. Copyright 2007, American Concrete Institute. All rights
reserved, including the making of copies unless permission is obtained from the copy-
right proprietors. Pertinent discussion including authors closure, if any, will be
published in the January-February 2008 ACI Materials Journal if the discussion is
received by October 1, 2007.
A class of high-performance fiber-reinforced cementitious
composites (HPFRCC), referred to as engineered cementitious
composites (ECC), was studied for its time-dependent properties.
The material exhibits a pseudo strain-hardening response with
multiple fine cracking in uniaxial tension. A series of experiments
on ECC specimens, as well as similar specimens without fibers,
was conducted to provide information about shrinkage, basic
creep, drying creep, and creep recovery of the material. Comparisons
with established predictive models for creep and shrinkage of
concrete were made. It was found that the ECC material developed
greater creep strains than a similar cementitious mixture without
fiber reinforcement. Surface cracking was observed to effect
shrinkage strain measurements and estimates of material
shrinkage behavior. Existing predictive models, while not
developed for such materials, can give a reasonable estimate of
creep and shrinkage behavior.
Keywords: cement paste; composites; creep; mortar; shrinkage.
INTRODUCTION
The high-performance fiber-reinforced cementitious
composite (HPFRCC) investigated in this paper is one that
was designed to exhibit a pseudo strain-hardening response
and multiple fine cracking in uniaxial tension. The material
investigated herein is also referred to as engineered cementitious
composites (ECC). The mixture design investigated is
composed of cement, silica fume, water, fine sand, and
roughly 2% by volume of short, randomly distributed polymeric
fibers. Typical mixture proportions are given in Table 1. One
potential application for ECC that has been investigated by
the authors is for use in hinge regions of segmentally
precast, post-tensioned concrete bridge piers.
1,2
In this
application, the ECC segments would be subjected to
considerable long-term compressive loads. Large creep
strains are likely to occur due to the post-tensioning and
because the ECC contains no coarse aggregate. Furthermore, a
1.5 to 2% by volume addition of fibers having no chemical
bond with the matrix material may change the
time-dependent response of the material. At the outset of
this research, it was unclear if the fibers would create paths
for easier flow of water or if they would control micro-
cracking within the matrix to reduce the flow of water
through the composite. Therefore, a testing program to
investigate the creep and shrinkage response was developed.
In addition, an ability to predict the long-term response of
ECC is needed for implementation of this material in structural
applications, particularly in prestressed concrete applications
where sustained loads can significantly influence global and
local structural response.
RESEARCH SIGNIFICANCE
ECC materials are being studied for several structural-scale
applications such as in low-rise shearwalls,
3
damping
devices in frame structures,
4
hinge regions of post-tensioned
bridge piers,
1,2
and various frame systems for seismic
resistance.
5,6
For many of these applications, and in particular
for applications involving prestressing, it is anticipated that
large creep strains may develop in ECC materials because
they contain no coarse aggregate. The time-dependent
behavior of ECC materials has been studied very little and
creep experiments on other fiber-reinforced materials have
led to conflicting results. Furthermore, currently accepted
(for example, codified) creep and shrinkage models have not
been calibrated for cementitious materials with or without
fibers that do not contain coarse aggregate. The experiments
reported herein provide needed information to estimate
time-dependent response of these ECC materials for potential
application to structures as well as to assess the applicability
of currently accepted concrete creep and shrinkage models to
ECC materials.
BACKGROUND
Creep and shrinkage of fiber-reinforced concrete is a topic
of considerable current research and is not yet fully understood.
Experimental studies on fiber-reinforced concrete have
reported conflicting findings on the effect of the fibers on
Title no. 104-M15
Creep and Shrinkage of High-Performance
Fiber-Reinforced Cementitious Composites
by Jon M. Rouse and Sarah L. Billington
Table 1Mixture proportions
P mixture
paste
PF mixture
ECC
(paste matrix)
M mixture
mortar
MF mixture
ECC
(mortar matrix)
Portland cement,
Type I
*
0.9 0.9 0.9 0.9
Silica fume
*
0.1 0.1 0.1 0.1
w/c 0.35 0.35 0.35 0.35
Fine aggregate
*
0 0 0.5 0.5
High-range water-
reducing admixture
*
0.0053 0.0074 0.0063 0.0084
Ultra-high-molecular-
weight polyethylene
fibers
*
0 0.011 0 0.014
Fiber volume,

%
0 1.6 0 1.6
*
By mass of cementitious material.

By total volume of mixture.

Note: P = paste; PF = paste with fibers; M = mortar; and MF = mortar with fibers.
ACI Materials Journal/March-April 2007 130
creep and shrinkage. For example, Balaguru and
Ramakrishnan
7
and Houde et al.
8
concluded that the creep
strains of concrete specimens reinforced with small amounts
of steel or polypropylene fibers were consistently higher
than strains measured in specimens without fibers. On the
other hand, Mangat and Azari
9
and Chern and Young
10
reported that steel fibers are effective in reducing both creep
and shrinkage in concrete. Mangat and Azari explained this
result with a model in which the steel fibers aligned with the
applied load act as compressive reinforcement for a cylinder
of the idealized surrounding matrix.
11
Zhang and Li
12
used
a similar model to make predictions for long-term shrinkage
of ECC. No information on the experimental behavior on
long-term shrinkage of ECC was provided in that study.
In the ECC materials studied herein, the fibers are
considerably finer than steel fibers and traditional polymeric
fibers used in fiber-reinforced concrete. The fibers have a
diameter between 8 and 38 m. This size lies between
typical capillary pore sizes (0.1 m and entrapped air [1 mm]
and is on the order of the aggregated C-S-H particles (1 to
5 m).
13
As such, the long-term response of ECC is expected
to behave differently than traditional fiber-reinforced
cement-based composites under compressive loads. For
example, Mangat and Azaris model assumed the mortar
surrounding the steel fiber to be a uniform, homogeneous,
and isotropic material.
11
This assumption would not hold at
the roughly 10 m length scale diameter of the polymeric
fibers used in ECC.
ECC materials (as well as other HPFRCC materials)
exhibit considerably more tensile ductility and are able to
bridge and control crack widths for larger material strains
than traditional fiber-reinforced concrete. It is expected that
cracking from creep and shrinkage will be better controlled
in ECC relative to traditional fiber-reinforced concrete. At
the outset of the study, it was unclear if the fibers would
limit long-term deformations through resisting crack
opening or increase long-term deformation due to larger
porous zones through which water may more easily flow.
To investigate the time-dependent response of ECC, creep
and shrinkage tests were conducted. A limited set of
permeability tests, not reported herein, were also conducted
on this material to investigate possible causes for creep and
shrinkage response observed.
1
Extensive surface crack
analysis on the shrinkage specimens was performed to evaluate
true shrinkage of the specimens. One of the primary motivations
for this study was to estimate prestressing losses for applications
of ECC to prestressed concrete structures. Therefore, the
creep results were compared with two existing analytical
models recommended by ACI and the Comit Europeen du
Bton (CEB-FIP) for predicting creep in concrete. It is noted
that these models are not intended to apply to the ECC materials
investigated herein. The potential strength or weakness in the
application of these models, however, is studied to identify
modifications or the need for new models for predicting the
time-dependent response of ECC to assist in its implementation
in structural designs.
CREEP AND SHRINKAGE TEST PROGRAM
The different mixture proportions used in the experimental
study are given in Table 1. Two different mixture designs for
the ECC were tested along with two corresponding mixtures
that were identical except that they did not contain fibers and
had minor differences in amounts of high-range water-
reducing admixture. The two ECC mixtures were the same
with the exception that one had a paste matrix consisting of
portland cement, silica fume, and water and the other had a
mortar matrix with fine sand added.
The fibers used in this study were ultra-high molecular
weight polyethylene (UHMWPE) fibers that were 6.4 mm in
length and approximately 8 m in diameter. The fine aggregate
used in the mortar specimens was silica sand uniformly
graded between particle sizes 0.3 mm and 0.15 mm (ASTM
sieves No. 50 and No. 70).
The study presented herein includes 12 triple-height
specimens, with each specimen representing three cylinders,
thus supplying nine data points (refer to Table 2 and Fig. 1).
Each specimen was a cylinder nominally 75 mm in diameter
by roughly 650 mm in length. The test set-up is similar to
that of ASTM C 512
14
wherein three cylinders are stacked
on each other and loaded. In this research, smaller diameter
cylinders were selected and therefore the three cylinders
were cast together in one specimen to ensure better stability
and alignment when loaded.
For each mixture design, one specimen was tested for
creep under a uniform compressive load and a companion
specimen that was not loaded was monitored for shrinkage.
In addition to the creep and shrinkage specimens of each
mixture design, four additional fiber-reinforced specimens
were sealed with a commercial paint-on epoxy vapor barrier.
Of these four additional specimens, two had a paste matrix
(one loaded for creep, the other monitored for shrinkage) and
two had a mortar matrix (one loaded for creep, the other
ACI member Jon M. Rouse is a Lecturer in the Department of Civil, Construction
and Environmental Engineering at Iowa State University, Ames, Iowa. He received his
BS in civil engineering from Iowa State University and his MEng and PhD in struc-
tural engineering from Cornell University, Ithaca, N.Y. His research interests include
behavior of concrete structures and fiber-reinforced materials.
ACI member Sarah L. Billington is an Associate Professor in the Department of Civil
and Environmental Engineering at Stanford University, Stanford, Calif. She is a
member of ACI Committee 341, Earthquake-Resistant Design of Concrete Bridges,
and Joint ACI-ASCE Committee 423, Prestressed Concrete, and is Co-chair of Joint
ACI-ASCE Committee 447, Finite Element Analysis of Reinforced Concrete. Her
research interests include experimental and computational evaluation of ductile
cement-based composites for structural design and retrofit applications as well as
fiber-reinforced polymeric composites made from renewable resources.
Fig. 1(a) and (b) Specimen geometry; and (c) creep frame.
ACI Materials Journal/March-April 2007 131
monitored for shrinkage). The sealed specimens were
included in the study to separate basic creep and drying creep
by comparing the response of the sealed specimens to that of
the unsealed specimens. Basic creep is defined as the
deformation that occurs while no moisture exchange is
allowed between the specimen and its environment.
All specimens were cured in a lime-saturated water bath
for 14 days (13 days for sealed specimens) prior to testing.
After curing, all specimens were stored and monitored in a
severe drying environment maintained at approximately
30 C (3 C) and roughly 45% (3%) relative humidity.
After 14 days of drying, each creep specimen was loaded
to 110kN (26 MPa). This load corresponds to approximately
40% of the 28-day compressive strength of the material, as
determined by compressive strength tests on 3 in. diameter
cylinders at 28 days. Load was maintained by a spring-type
creep frame, as recommended in ASTM C 512,
14
and was
applied and measured using a 267 kN hydraulic ram and load
cell. The sealant was applied to each sealed specimen 1 day
prior to load application (after 13 days of drying).
Changes in length in the specimens were monitored over
nine gauge locations on the surface of the specimen
three equally spaced around the circumference by three
along the longitudinal axis spaced at 150 mm (Fig. 1).
Measurements were taken using a hand-held mechanical
strain gauge with precision to 1.3 m. The creep specimens
were loaded for 125 days and then unloaded. Measurements
were taken throughout the loaded stage and for at least 45 days
after unloading.
CREEP AND SHRINKAGE RESULTS
The mean of the nine measurements per specimen for the
creep and shrinkage strains is shown for all 12 specimens in
Fig. 2. The average standard deviation of the nine strain
measurements for all specimens is 0.00058. The results show
that the ECC specimens exhibit considerably higher drying
shrinkage and creep strains than their corresponding
unreinforced companions. As expected, the specimens with
a mortar matrix undergo significantly less creep and
shrinkage than the corresponding paste specimens. This
effect is attributed to the replacement of paste volume with
aggregate. It is well known that the aggregate volume can be
assumed to behave elastically while the paste phase is the
primary agent of creep and shrinkage.
Mean elastic strain, elastic strain recovery, and creep strain
recovery may also be computed from the values in Fig. 2. These
values are given in Table 3. The much lower elastic strains
exhibited by the mortar specimens may again be attributed to
the addition of fine aggregate.
The results in Table 3 show that the specimens with fibers
consistently exhibited higher elastic strain recovery than the
corresponding specimens without fibers. This may be
attributable, at least to a small degree, to elastic strain energy
stored in the fibers when the specimens were compressed.
Note that the axial elastic modulus of the fibers of 66 GPa
(9600 ksi) is approximately five times that of the paste
matrix. Once load is removed, the energy stored in the fibers
would be expected to exert a restorative force on the specimens.
Shrinkage cracks that formed prior to loading may also have
Fig. 2Creep and shrinkage response of all test specimens.
(Note: time = 0, represents Day 14, that is, 14 days after
casting when drying commenced and measurements began.)
Table 2Test specimens
Specimen
*
Matrix
Load,
kN
Fibers,
% Sealant
PCFS Paste 110 1.6
Two-component, high-build,
thixotropic, vapor-proof epoxy coating
PCFU Paste 110 1.6 None
PCXU Paste 110 None None
PShFS Paste 0 1.6
Two-component, high-build,
thixotropic, vapor-proof epoxy coating
PShFU Paste 0 1.6 None
PShXU Paste 0 None None
MCFS Mortar 110 1.6
Two-component, high-build,
thixotropic, vapor-proof epoxy coating
MCFU Mortar 110 1.6 None
MCXU Mortar 110 None None
MShFS Mortar 0 1.6
Two-component, high-build,
thixotropic, vapor-proof epoxy coating
MShFU Mortar 0 1.6 None
MShXU Mortar 0 None None
*
P = paste matrix; M = mortar matrix; C = creep test; Sh = shrinkage test; F = contains
fibers; X = contains no fibers; S = sealed; and U = unsealed.
Table 3Elastic strains, elastic strain recovery,
and creep strain recovery
PCFU PCXU MCFU MCXU
Elastic strain 0.00253 0.00239 0.00147 0.00161
Elastic strain
recovery
0.00195 0.00146 0.00118 0.00105
% recovered 77.1 61.1 80.3 65.2
Creep strain
recovery
*
0.00057 0.00078 0.00054 0.00032
% recovered 14.4 26.4 23.1 17.1
*
Creep strain recovery 45 days beyond unloading.
132 ACI Materials Journal/March-April 2007
played a role in the additional strain recovery of the ECC
relative to the material without fibers. A specimen with
damage due to shrinkage cracking would likely be more
compliant when loaded than a specimen with no damage. By
reducing shrinkage cracking prior to loading, the fibers may
help enhance the elastic stiffness and therefore facilitate
more strain recovery of the specimens. The effects of surface
cracking are further discussed in the next section.
The influence of fibers on creep strain recovery is less
clear. While the mortar-based ECC Specimen MCFU recovered
more creep strain after 45 days than the corresponding
unreinforced Specimen MCXU, the opposite was observed
for the paste-based specimens. Additional experiments to
investigate this specific behavior are needed.
Figure 3 shows the total creep, basic creep, and drying creep
of the ECC specimens with a paste matrix and with a mortar
matrix. Basic creep is the creep strain measured in the sealed
specimens. Drying creep is calculated by subtracting the creep
strains measured in the sealed specimen from those measured in
the unsealed companion specimen for equivalent times.
Experiments on unreinforced materials were not conducted.
The ECC specimen with the paste matrix exhibits similar
magnitudes of basic and drying creep while drying creep
appears to reach a constant value approaching 0.002 after
90 days. The ECC specimen with the mortar matrix also
exhibits drying creep that approaches 0.002, but is slower to
reach this value. This difference in rates of drying creep is
attributed to the formation of surface cracks. The larger
early-age shrinkage strains observed in the paste specimens
relative to mortar specimens would cause a greater degree of
surface cracking by the time of loading. Surface cracking
prior to loading would then accelerate the drying creep by
providing flow paths for water movement in the test specimens.
The difference in magnitude of basic creep apparent between
the paste and mortar materials may be attributed to paste
being replaced by aggregate in the mortar.
DISCUSSION OF CREEP
AND SHRINKAGE RESULTS
The differences in total creep and shrinkage among the
paste and mortar specimens as discussed previously were
expected due to the replacement of paste with sand in the
mortar matrix. To understand the differences in response
between the ECC specimens and their unreinforced
companion specimens, two hypotheses were considered.
First, there may have been a greater degree of drying
shrinkage cracking on the surface of the unreinforced
specimens. Second, there may be an elevated permeability
of the fiber-reinforced specimens. A study of drying
shrinkage cracking was conducted and is presented herein.
Surface cracking serves to relieve tensile stresses as the
specimen dries and would lead to errantly low values of
measured strains. A schematic diagram illustrating this
effect is shown in Fig. 4. As a result, the total average
shrinkage relationship between the reinforced and unreinforced
specimens may be different than they appear in Fig. 2. To
assess this, surface cracking on four of the shrinkage specimens
(PShFU, PShXU, MShFU, and MShXU) was measured
using an x-y-z translation stage and a 50 magnification
crack scope with a scale etched on the lens.
A comparison of total strain measurements before and
after accounting for surface cracking is given in Table 4
(variables defined in Fig. 4). On average, the sum of the
crack openings between gauge points of the specimens
without fibers was over 2.5 times that of the fiber-reinforced
specimens. Because the surface crack survey was taken near
the end of shrinkage monitoring, the correction was extrapo-
Fig. 3Total, basic, and drying creep in ECC paste and mortar
test specimens.(Note: time = 0, represents Day 14, that
is, 14 days after casting when drying commenced and
measurements began.)
Fig. 4Measurements for shrinkage specimen surface
cracking.
Fig. 5Effect of shrinkage correction on shrinkage strains
for specimens with paste matrix. (Note: time = 0, represents
Day 14, that is, 14 days after casting when drying commenced
and measurements began.)
Fig. 6Effect of shrinkage correction on creep strains for
specimens with paste matrix.(Note: time = 0, represents
Day 14, that is, 14 days after casting when drying commenced
and measurements began.)
ACI Materials Journal/March-April 2007 133
lated back in time to the start of the experiment. This was
done by assuming that the crack openings measured at the
end of the test developed in time proportionately with the
measured deformations (that is,
c
/
c

ult
=
m
/
m

ult
).
The results of this correction for specimens with a paste
matrix are shown in Fig. 5 and 6. After making this correction,
it appears that the rate of shrinkage is initially higher for the
fiber-reinforced materials, but after some time, shrinkage in
the specimens without fibers exceeds that of the
fiber-reinforced specimens. A possible explanation for this
behavior is that, at early times when surface cracking is
minimal, the fiber-reinforced specimens are slightly more
permeable and more readily lose water to the environment.
As cracking develops on the surface of the specimens, water
may more readily be lost to the environment. Because the
specimens without fibers have more extensive surface
cracking, their shrinkage rates eventually overtake those of
the fiber-reinforced specimens that have much finer surface
cracks. It is also evident from Fig. 5 and Table 4 that,
after correcting for surface cracking, the difference in
shrinkage behavior between specimens with fibers and
the corresponding ones without fibers is significantly reduced.
The effect of this shrinkage correction on creep calculations is
considered as well. To calculate total creep strains, the sum
of the total strains measured in the shrinkage specimens and
the elastic strains measured upon loading of the creep specimens
are subtracted from the total strains measured in the creep
specimens. The effect of the correction, by increasing the
shrinkage strains, leads to lower values of calculated creep as
illustrated in Fig. 6 for the paste-matrix specimens, PCFU
and PCXU. Because the shrinkage correction is of greater
magnitude for the specimens without fibers, the difference in
creep strains between fiber-reinforced and corresponding
materials without fibers is magnified. Note that it is question-
able whether or not the shrinkage correction should be
applied when calculating creep strains. When load is applied
to the creep specimens, the elastic shortening of the specimens
is several times the magnitude of the sum of crack openings on
the shrinkage specimens. Therefore, the applied load likely
causes most of the shrinkage cracks to close and helps suppress
additional shrinkage cracking. Because the surface cracking
accelerates drying shrinkage by facilitating moisture loss, the
creep specimens (while loaded) were likely experiencing less
shrinkage than their corresponding shrinkage specimens. For
this reason, the shrinkage correction should probably not be
applied when calculating creep strains.
A second hypothesis for the differences in behavior
between the specimens with and without fibers is permeability.
It is well known that one of the primary mechanisms of
shrinkage and creep in cementitious materials is redistribution of
water in the gel and loss of water to the environment. The
higher the permeability, the more readily these processes can
occur. In concrete, a zone of higher porosity often exists at
the interface between an aggregate and the surrounding
matrix. Such a zone has also been observed around the fibers
in an ECC material similar to that studied herein.
15
Therefore, it
is reasonable to expect a higher permeability in ECC materials
because the porous zone may act as a conduit to promote
moisture migration. To investigate this possibility, a series
of permeability tests was conducted on plain, unreinforced paste
and ECC paste specimens.
1
The results were interesting but
inconclusive and are not reported herein. Further study of
permeability in these materials is recommended.
CORRELATION WITH CONCRETE
CREEP AND SHRINKAGE MODELS
Two common models used in current civil engineering
design to predict the creep and shrinkage of concrete are put
forth by ACI
16
and the CEB-FIP Model Code.
17
These
models are semi-empirical in nature and are calibrated for a
wide range of concrete mixtures but not for cement pastes,
mortars, or fiber-reinforced materials. They are examined
herein to see their potential applicability to ECC materials
and to shed light on necessary modifications or the need for
new models for ECC materials. A brief review of the models
is given to facilitate the review of their applicability to
ECC materials.
The ACI model to predict shrinkage after age 7 days for
moist cured concrete is
16
(1)
where t
m
is the time in days from the end of moist curing;
(
sh
)
t
is the shrinkage strain as a function of time; (
sh
)
u
is the
ultimate shrinkage strain in time; and is the correction
factors for nonstandard conditions.
The basic variables used by this model (including those to
estimate (
sh
)
t
and (
sh
)
u
) are drying time, ambient relative
humidity, minimum specimen dimension, slump, cement
content, and air content. For the case of the experiments, air
content and slump were not measured and thus were estimated
for the model. The range of values for air content considered
was 0 to 12% and for slump was 76 to 150 mm.
The ACI model to predict creep after age 7 days for moist
cured concrete is
16
(2)
where E
ci
is the modulus of elasticity of concrete at the time
of initial load; t is the age of concrete since mixing; t is the
age of concrete at loading; (C
u
) is the ultimate specific creep =
(
cr
)
u
/
c
; (
cr
)
t
is the creep strain as a function of time; and
is the correction factors for nonstandard conditions.
The variables considered in the ACI creep equation
(including those to estimate (
cr
)
u
) are age of concrete at
loading, applied stress, initial elastic modulus, ambient relative
humidity, minimum specimen dimension, slump, cement
content, and air content. Tables 5 and 6 give the values of the
various parameters and coefficients used and calculated to

sh
( )
t
t
m

sh
( )
u

35 t
m
+
------------------------- =

cr
( )
t
1 E
ci
( ) 1 t t ( )
0.6
C
u
( ) 10 t t ( )
0.6
+ ( ) ( ) + ( ) =
Table 4Shrinkage specimen surface crack
opening and strain
*
L
0

m

c

m

t
PShFU
Average 6.1303 0.0166 0.0027 0.0027 0.0032
SD 0.0786 0.0013 0.0009 0.0002 0.0003
PShXU
Average 6.1408 0.0122 0.0080 0.0020 0.0033
SD 0.0107 0.0020 0.0013 0.0003 0.0003
MShFU
Average 5.9944 0.0140 0.0020 0.0023 0.00274
SD 0.0625 0.0007 0.0002 0.0001 0.0001
MShXU
Average 6.0947 0.0116 0.0056 0.0019 0.0028
SD 0.0998 0.0017 0.0014 0.0003 0.0005
*
Variables are defined in Fig. 4.
Note: SD = Standard deviation.
134 ACI Materials Journal/March-April 2007
estimate shrinkage strains as per Eq. (1) and creep strains as
per Eq. (2), respectively. Note that E
ci
was measured during
the compressive strength tests and the average results of
three specimens were used for the model. Details are given
in Reference 16.
The CEB-FIP Model Code
17
for shrinkage strains
cs
specifies
(3)
where
cso
is the notional shrinkage coefficient (as a function
of cement type);
s
is the coefficient to describe shrinkage
development with time; t is the age of concrete (days); and t
s
is the age of concrete (days) at the beginning of shrinkage.
The coefficients in Eq. (3) are functions of the mean
compressive strength of the concrete at age 28 days, cement
type, the size of the member, ambient temperature, and
ambient relative humidity.
The CEB-FIP Model Code
17
for creep strains
cc
specifies
(4)

cs
t t
s
, ( )
cso

s
t t
s
( ) =

cc
t t
0
, ( )
c
t
0
( ) t t
0
, ( ) E
c28
=
where
c
is the applied compressive stress; (t,t
0
) is the
creep coefficient =
0

c
(t t
0
); E
c28
is the modulus of
elasticity at age 28 days;
0
is the notional creep coefficient;
c
is the coefficient to describe creep development with
time after loading; t is the age of concrete (days) at the
moment considered; and t
0
is the age of concrete at
loading (days) adjusted for the ambient temperature and
cement type.
Fig. 7Comparison of ACI
16
and CEB-FIP
17
models to
measured shrinkage strains.
Fig. 8Comparison of ACI
16
and CEB-FIP
17
models to
measured creep strains.
Table 7Variables used for model parameters
for shrinkage strain prediction (CEB-FIP)
17
Specimen
PShFU PShXU MShFU MShXU
h
e
,% 45 45 45 45
T, C 30 30 30 30
t
s
, days 14 14 14 14
h, mm 36.83 36.83 36.83 36.83
f
cm
, MPa 72.4 77.7 77.9 87.4
Table 6Variables used for model parameters
for creep strain prediction (ACI)
16
Specimen
PShFU PShXU MShFU MShXU
S
c
, in. 5 6 4 5
C, lb/yd
3
2434 2474 1882 1913
A
c
,

% 10 10 10 10
E
ci
, ksi 1499 2108 2305 2682
t, days 28 28 28 28
h
e
,% 45 45 45 45
T
m
, in. 2.9 2.9 2.9 2.9

c
, ksi 3.78 3.78 3.78 3.78
Table 8Variables used for model parameters
for creep strain prediction (CEB-FIP)
17
Specimen
PShFU PShXU MShFU MShXU
h
e
,% 45 45 45 45
T, C 30 30 30 30
h, mm 36.83 36.83 36.83 36.83

c
, MPa 26.1 26.1 26.1 26.1
f
cm
, MPa 72.4 77.7 77.9 87.4
E
c28
, MPa 10,336 14,535 15,895 18,492
Table 5Variables used for model parameters for
shrinkage strain prediction (ACI)
16
Specimen
PShFU PShXU MShFU MShXU
h
e
, % 45 45 45 45
T
m
, in. 2.9 2.9 2.9 2.9
S
c
, in. 5 6 4 5
C, lb/yd
3
2434 2474 1882 1913
A
c
, % 10 10 10 10
ACI Materials Journal/March-April 2007 135
The coefficients and adjustments in Eq. (4) are a function
of ambient relative humidity, the size of the member, mean
compressive strength of the concrete at 28 days, mean
modulus of elasticity at 28 days, cement type, and ambient
temperature. Tables 7 and 8 give the values of the variables
used to estimate shrinkage strains as per Eq. (3) and creep
strains as per Eq. (4), respectively.
Figure 7 shows the predictions of shrinkage strain and Fig. 8
shows the predictions of creep strain given by the ACI and
CEB models compared to the measured values on all of the
unsealed specimens. Included in the figures are the measured
values with the shrinkage strain corrected for surface
cracking, as discussed previously.
The results in Fig. 7 show that both models generally
underpredict the experimental values of corrected shrinkage
strains, as expected. For all four materials, the CEB-FIP
shrinkage model predicted slightly higher values than the
ACI model. This may be attributed primarily to the CEB-FIP
models accounting for the high ambient temperature (30 C)
used in the experiment. Note that with the ACI model, the
results from using the highest reasonable estimate for air
content (12%) and slump (150 mm) still underpredicts both
the creep and shrinkage for all four materials.
As with the shrinkage models, the ACI and CEB-FIP
creep models consider some similar variables with the
difference that, while the ACI model uses properties of the
fresh material (slump and air content), the CEB-FIP model
relies upon 28-day compressive strength and includes
consideration for ambient temperature. Figure 8 shows that
the ACI model consistently underpredicts the experimental
values of measured creep while the CEB-FIP model predicts
creep remarkably well for the four materials tested. The ACI
model underpredicts the fiber-reinforced materials to a
greater degree than the unreinforced materials because of the
reduction in slump arising from the fibers in the fresh
mixture. The fibers in the fresh mixture give the material a
cohesive, sticky consistency unlike that of normal concrete.
This cohesiveness gave the ECC a lower slump than the
corresponding materials without fibers, although the basic
mixture design was the same. Therefore, while the use of
slump in the model is most likely reflecting variations in
concrete mixture constituents, in this case, the slump change
is simply due to the addition of fibers. The CEB-FIP model
predicts higher values of creep than the ACI model in part,
again, due to its consideration of temperature.
The predictions of both models are reasonable based on
the experimental data presented. The models, however, were
not developed and calibrated for ECC materials and further
testing is needed to validate the general application of these
models to ECC materials. A more accurate model for
predicting creep and shrinkage strains should perhaps
account for the fiber volume in the material, cracking and
expected crack widths, and permeability.
CONCLUSIONS
Based on the experimental results, it was found that ECC
develops greater creep strains than an identical cementitious
mixture (paste or mortar) without fibers. As expected, the
addition of fine aggregate in ECC can significantly reduce
creep and shrinkage of the material. Paste matrix ECC
specimens exhibited higher values of basic creep than
mortar matrix ECC specimens due to paste replacement with
aggregate in the mortar. By the end of creep monitoring,
paste matrix ECC and mortar matrix ECC exhibited similar
values of drying creep, although the paste matrix ECC
drying creep appeared to be accelerated at earlier times.
Furthermore, the ECC creep specimens consistently
exhibited greater elastic recovery upon unloading than
the corresponding specimens without fibers.
The sum of drying shrinkage crack openings on the
surface of shrinkage specimens without fibers was found to
be over 2.5 times that of the corresponding specimens with
fibers (ECC). Surface cracking and the effect of fibers on
such cracking should be taken into account when predicting
shrinkage behavior. Because the applied load on the creep
specimens inhibits surface cracking, the cracking correction
should not apply to calculated creep values.
Although the ACI and CEB-FIP models for creep and
shrinkage were not developed for the materials studied
herein, it was found that they could predict the creep and
shrinkage behavior of ECC reasonably well, with the
CEB-FIP model giving more accurate results. Before these
models could be applied for the assessment of structural
designs using ECC materials, more experimental verification as
well as modifications to the models are needed.
ACKNOWLEDGMENTS
This research was partially sponsored by an award from the Civil Engineering
Research Foundation of ASCE, the National Science Foundation
(CMS-9984127), and a Cornell University Graduate Fellowship from Cornell
University. This support, as well as fiber donations from DSM in The
Netherlands, is gratefully acknowledged.
NOTATION
A
c
= air content, %
C = cement content, lb/yd
3
C
u
= ultimate specific creep (ksi
1
) = (
cr
)
u
/
c
E
c28
= modulus of elasticity at age 28 days, MPa
E
ci
= modulus of elasticity of concrete at time of initial load, ksi
f
cm
= mean compressive strength of concrete at 28 days, MPa
h = specimen radius, mm
h
e
= ambient relative humidity, %
L
0
= initial gauge length
MCFS = specimen name: mortar matrix, creep specimen, with
fibers, sealed
MCFU = specimen name: mortar matrix, creep specimen, with
fibers, unsealed
MCXU = specimen name: mortar matrix, creep specimen, without
fibers, unsealed
MShFS = specimen name: mortar matrix, shrinkage specimen, with
fibers, sealed
MShFU = specimen name: mortar matrix, shrinkage specimen, with
fibers, unsealed
MShXU = specimen name: mortar matrix, shrinkage specimen, without
fibers, unsealed
PCFS = specimen name: paste matrix, creep specimen, with fibers,
sealed
PCFU = specimen name: paste matrix, creep specimen, with fibers,
unsealed
PCXU = specimen name: paste matrix, creep specimen, without
fibers, unsealed
PShFS = specimen name: paste matrix, shrinkage specimen, with
fibers, sealed
PShFU = specimen name: paste matrix, shrinkage specimen, with
fibers, unsealed
PShXU = specimen name: paste matrix, shrinkage specimen, without
fibers, unsealed
S
c
= slump, in.
T = ambient temperature, C
T
m
= minimum specimen dimension, in.
t = age of concrete since mixing at given time, days
t = age of concrete at loading, days
t
0
= adjusted age of concrete at loading, days
t
m
= time from end of moist curing, days
t
s
= age of concrete at beginning of shrinkage, days

c
= coefficient to describe creep development with time
after loading
ACI Materials Journal/March-April 2007 136

s
= coefficient to describe shrinkage development with time

c
= sum of crack openings between pins at given time

c

ult
= sum of crack openings between pins at end of experiment

m
= measured shortening at given time

m

ult
= measured shortening at end of experiment

cc
= creep strain at given time
(
cr
)
t
= creep strain at given time
(
cr
)
u
= ultimate creep strain

cs
= shrinkage strain at given time

cso
= notional shrinkage coefficient

m
= measured strain at given time =
m
/L
0
(
sh
)
t
= shrinkage strain at given time
(
sh
)
u
= ultimate shrinkage strain in time

t
= corrected strain at given time = (
m
+
cr
)/L
0
(t,t
0
) = creep coefficient

0
= notional creep coefficient
= correction factors for nonstandard conditions

c
= applied stress (ksi for ACI equations, MPa for CEB-FIP
equations)
REFERENCES
1. Rouse, J. M., Behavior of Bridge Piers with Ductile Fiber-Reinforced
Hinge Regions and Vertical, Unbonded Post-Tensioning, PhD thesis, Cornell
University, Ithaca, N.Y., 2004, 298 pp.
2. Billington, S. L., and Yoon, J. K., Cyclic Response of Precast Bridge
Columns with Ductile Fiber-reinforced Concrete, Journal of Bridge
Engineering, ASCE, V. 9, No. 4, 2004, pp. 353-363.
3. Kim, K., and Parra-Montesinos, G., Behavior of ECC Low-Rise
Walls subjected to Displacement Reversals, High-Performance
Fiber-Reinforced Cement Composites, Proceedings of the Fourth International
RILEM Workshop, A. E. Naaman, and H. W. Reinhardt, eds., Ann Arbor,
Mich., 2003, pp. 505-515.
4. Fukuyama, H.; Suwada, H.; and Ilseung, Y., ECC Damper for Structural
Control, Proceedings of the JCI International Workshop on Ductile Fiber
Reinforced Cementitious Composites (DFRCC), Japan Concrete Institute,
Gifu, Japan, 2002, pp. 219-228.
5. Parra-Montesinos, G., and Wight, J. K., Seismic Response of Exterior RC
Column-to-Steel Beam Connections, Journal of Structural Engineering,
ASCE, V. 126, No. 10, 2000, pp. 1113-1121.
6. Fischer, G., and Li, V. C., Intrinsic Response Control of
Moment-Resisting Frames Utilizing Advanced Composite Materials and
Structural Elements, ACI Structural Journal, V. 100, No. 2, Mar.-Apr.
2003, pp. 166-176.
7. Balaguru, P. N., and Ramakrishnan, V., Properties of Fiber-Reinforced
Concrete: Workability, Behavior under Long-Term Loading, Air-Void
Characteristics, ACI Materials Journal, V. 85, No. 3, May-June 1988,
pp. 189-196.
8. Houde, J.; Prezeau, A; and Roux, R., Creep of Concrete Containing
Fibers and Silica Fume, Fiber-Reinforced Concrete Properties and
Applications, SP-105, S. P. Shah and G. B. Batson, eds., American
Concrete Institute, Farmington Hills, Mich., 1987, pp. 101-118.
9. Mangat, P. S., and Azari, M. M., Compression Creep Behavior of
Steel Fibre Reinforced Cement Composites, Materiaux et Constructions,
V. 19, No. 113, 1986, pp. 361-370.
10. Chern, J. C., and Young, C. H., Compressive Creep and Shrinkage
of Steel Fibre Reinforced Concrete, International Journal of Cement
Composites, V. 11, No. 4, 1989, pp. 205-214.
11. Mangat, P. S., and Azari, M. M., A Theory for the Creep of Steel
Fibre Reinforced Cement Matrices Under Compression, Journal of Materials
Science, V. 20, No. 3, 1985, pp. 1119-1133.
12. Zhang, J., and Li, V. C., Influences of Fibers on the Drying
Shrinkage of Fiber Reinforced Cementitious Composite, Journal of
Engineering Mechanics, ASCE, V. 127, No. 1, 2001, pp. 37-44.
13. Mehta, P. K., and Monteiro, P. J. M., Concrete: Structure, Properties,
and Materials, 2nd Edition, Prentice-Hall, Englewood Cliffs, N.J., 1993,
496 pp.
14. ASTM C 512, Standard Test Method for Creep of Concrete in
Compression, ASTM International, West Conshohocken, Pa., 1994, 4 pp.
15. Chan, Y.-W., and Li, V. C., Effects of Transition Zone Densification
on Fiber/Cement Paste Bond Strength Improvement, Advanced Cement
Based Materials, V. 5, No. 1, 1997, pp. 8-17.
16. ACI Committee 209, Prediction of Creep, Shrinkage, and Temperature
Effects in Concrete Structures (ACI 209R-92), American Concrete Institute,
Farmington Hills, Mich., 1992, 47 pp.
17. Comit Euro-International Du Bton, CEB-FIP Model Code 1990,
Thomas Telford, London, 1993, pp. 53-58.