Professional Documents
Culture Documents
Abstract
Control of thermal cracking in young concrete is of great importance to ensure a desired service lifetime and function of
a structure. Young concrete is here defined as the period up to approximately 100 days after casting. Making reliable
predictions about thermal stresses, and thereby cracking risks, the creep behaviour forms an important part of the mate-
rial modelling. Up until now few studies have been made to investigate how different creep modelling influences calcu-
lated thermal stresses. Existing creep models for young concrete are often pure mathematical expressions with no direct
relation to the material behaviour and thereby complicated to understand and use in a more practical context.
In this paper a new basic creep model primarily aimed for early age purposes is outlined. The formulation with its
model parameters, which have an easy to understand meaning in the material behaviour, is based on piece-wise linear
curves in logarithm of time and therefore denoted the Linear Logarithmic Model (LLM). Comparison with experimental
creep data and other more commonly used creep formulations for young concrete is made to achieve an opinion about
the accuracy of the new model. The new model is a flexible and robust formulation that can model the behaviour of
both young and mature concrete. The robustness enables it to make reliable creep modelling with very few test data.
Another advantage with the LLM formulation is that the appearance of negative relaxation in linear viscoelastic
modelling is very small and negligible with respect to thermal stresses. This means that the original formulation may be
used directly in a thermal stress analysis for young concrete without any adjustment for negative relaxation.
The LLM formulation shows very good agreement directly with experimental creep data and indirectly with meas-
ured thermal stresses, whereby the formulation has been used to model the viscoelastic behaviour of the concrete. The
formulation also has the best correlation with experimental data compared to other commonly used creep models that
have been analysed in this paper.
temperature gradually falls and the concrete structure not be calculated with acceptable accuracy without a
starts to contract causing increasing tensile stresses. If correct consideration of the creep behaviour. According
the concrete tensile deformation capacity is exceeded to Bosnjak (2000) and Atrushi et al. (2001), creep de-
the high tensile stresses may lead to failure causing creases an elastically induced stress in the order of 40-
cracks, often permanent, through the structure. 50% for a fully restrained concrete specimen. Westman
A prerequisite for calculations of self-induced stresses, (1999) states that the early age behaviour during the
and thereby estimations of cracking risks, is that follow- first three days after load application is the most crucial
ing concrete properties are known and modelled in a part of the modelled creep response when computing
correct way: thermal stresses whereas the modelled creep behaviour
heat of hydration after this period is less important. Bosnjak (2000) shows
temperature influence on the hardening process on the other hand that a correct prediction of creep dur-
thermal dilatation ing the entire hardening period of the concrete is neces-
shrinkage sary for reliable stress calculations.
creep and relaxation Some linear constitutive relations with and without
strength development history integrals for calculation of thermal stresses re-
fracturing mechanics (non-linear stress-strain be- quire relaxation data instead of creep (see for example
haviour at high tensile stresses) Trost 1967, Baant and Wittmann 1982, Baant 1988 or
When applying mathematical models for creep and Larson 2000). An important presumption when estab-
shrinkage of young concrete the total strain tot that lishing a relaxation function, from a known creep func-
evolves self-induced stresses during the hydration proc- tion or data from tests, is that tensile stresses can not
ess may for a uniaxial case be regarded as the sum of develop after long time when a compressive strain has
stress independent and stress dependent strains as (see been applied, i.e. negative relaxation values can not
further e.g. Baant and Wittmann 1982, Baant 1988 appear. Consequently, all creep models have to be
and Baant and Xi 1999) checked and, if necessary, adjusted for negative relaxa-
tion, which not always has been the case.
Stress independent Stress dependent
tot = [ ] + [
vol visc + fract + rest ] (1) 1.3 Existing formulations of creep functions for
early age concrete
It has been common in the past to formulate creep func-
where vol is a volumetric strain related to tem-
tions from the loading age of about 2 days or more and
perature and shrinkage, [-]
regard them as representative for young concrete. A
visc is strain due to viscoelastic behaviour
famous and well working family of creep formulas in
(creep and relaxation), [-]
this group are designed as
fract is strain related to fracturing mechan-
ics, [-]
1
rest is strain related to restraint , [-] J (t load , t 0 ) = (1 + F (tload , t 0 )) (2)
E0
1.2 How creep influences thermal stresses
There exist numerous methods to model the creep be- where E 0 is an infinite modulus of elasticity which is
haviour in young concrete (see for example Trost 1967, constant and formally valid for tload 0.
Baant 1972, Baant and Wu 1974, Baant and Panula The most famous functions F (tload , t0 ) are the
1978, Baant and Wittmann 1982, Baant and Chern following
1985a and 1985b, Baant 1988, Emborg 1989, Kanstad
1990, Baant and Xi 1999, Westman 1999, De Schutter DPL: Double Power Law (Baant and Osman, 1976)
2001, Gutsch 2001, Pane and Hansen 2001, Sakata et al. TPL: Triple Power Law
F (tload , t0 ) =
2001, Zi and Baant 2001 or Hagihara et al. 2002). (Baant, 1977 and Baant and Chern, 1985a)
Many of these models are complex mathematical for- LDPL: Log Double Power Law
mulations including sets of parameters that have no di- (Baant and Chern, 1985b)
rect relation to the material behaviour and thereby are All of these functions have been frequently used by
complicated to understand and use in a more practical researchers when modelling concrete behaviour, but in
context. Besides, several of the formulations also have the original form none of them are able to reflect very
some dependency among the free parameters, which early age concrete behaviour, i.e. loading ages below 2
means that they will give different solutions depending days. Different researchers have solved that in different
on the start position of the free parameters and/or on the ways. Emborg (1989) and later Westman (1999) added
amount of test data available in the regression procedure. separate functions for short durations at very early ages
Modelling of creep at early ages has been treated by to the TPL. These models are denoted Modified Triple
many authors, but very seldom in respect of how creep Power Law (MTPL). MTPL works well, but involves a
influences self-induced stresses. Westman (1999) and lot of (about 15) free parameters to be determined. Kan-
Bosnjak (2000) have shown that thermal stresses can stad (2000) substitutes the parameter E 0 with a general
174 M. Larson and J. Jonasson / Journal of Advanced Concrete Technology Vol. 1, No. 2, 172-187, 2003
function E0 (t0 ) using the DPL, and then determines the other identifying mix parameters. This test series is cho-
free parameters in comparison with creep tests, which sen as it has several loading ages at early ages.
here is denoted Modified Double Power Law (MDPL). The deformation of sealed cylindrical specimen was
Pane and Hansen (2001) start from the LDPL and sub- measured under constant compression, which was less
stitute E0 with a general function E0 (t0 ) and replace than approximately 40 % of the compressive strength at
the power function for the loading age with a more gen- loading. This implies a linear behaviour with deforma-
eral function, which here is denoted Modified Log tions assumed to be proportional to applied stress. Load
Power Law (MLPL). In both these later cases (MDPL independent deformations, such as shrinkage and ther-
and MLPL) the static elastic modulus E0 (t0 ) is not ex- mal dilatation, were measured simultaneously on
plicitly defined or taken care of in a consistent way in unloaded companion specimens stored under the same
the formulation. This creates a formal discrepancy, al- curing conditions.
though the formulation might work well apparently, The total time dependent deformation (t , t0 ) under
especially if very short load durations are not analysed compression may be expressed by a compliance formu-
carefully. lation as
This paper will focus on the creep behaviour of harden- where J (tload , t0 ) is the creep compliance defined as
ing concrete and how the choice of model influences the the response at time duration tload = t t0 . t is time
calculated thermal stress development. The main objec- after mixing with stress (t0 ) acting since loading age
tive is to formulate a new basic creep model whereby t0 .
following preconditions shall be fulfilled: The creep compliance J (tload , t0 ) was determined
I. The model shall include the behaviour of both from the test results as the difference per unit stress be-
young and mature concrete in one formulation with the tween the loaded and unloaded specimen and thus con-
same accuracy at all loading ages and load durations. sidered as the single effect of the applied stress. As the
II. The model shall in its own formulation have the temperature was kept constant (at about 20 C) in these
shape of the actual material behaviour enabling a reli- tests and the specimens were sealed (i.e. no exchange of
able modelling with few test data. water with the environment), by definition, basic creep
III. All model parameters shall be well recognized in is considered.
the creep behaviour, i.e. every parameter must have a
meaning easy to understand in the material behaviour. 3.2 Stress tests
IV. The appearance of negative relaxation (stress re- Thermal stress tests have been performed by Westman
versed values) in linear viscoelastic modelling has to (1999) and Hedlund (2000) at LTU in Sweden. The
vanish or become practically negligible with respect to fresh concrete specimen, constituting a 1 m long beam
thermal stresses. with a square cross section of 150 mm, is placed in a
V. The new creep model shall have the same (or bet- rather stiff steel frame and by regulation forced to
ter) correlation to experimental data as other more com- 100 % restraint. One end of the specimen is fixed into
monly used creep formulations. the frame while the other is free to deform in its longi-
tudinal direction. The length of the specimen is kept
constant throughout the test by means of a servo-
3. Laboratory tests
hydraulic cylinder connected to the free end of the
specimen. The force from the cylinder is directly pro-
3.1 Creep tests
portional to the stress inside the concrete specimen.
The comparison of creep formulations is performed The deformations that give rise to self-induced
based on results from creep tests performed by Westman
stresses in a hardening concrete are mainly related to
(1999) and Hedlund (2000) at Lule University of
temperature and shrinkage. An authentic temperature
Technology, LTU, in Sweden. The cement type used in development inside the specimen, which here corre-
the concrete mix is given in Table 1 together with some
sponds to the temperature that would occur in a 0.7 m
LTU66 is the denotation used at Lule University of Technology (Hedlund, 2000). w0 is the mixing water content
and C is the cement content. The cement is Degerhamn Std P from Cementa AB.
M. Larson and J. Jonasson / Journal of Advanced Concrete Technology Vol. 1, No. 2, 172-187, 2003 175
thick wall, is achieved by blowing temperate air around The basic assumption of the new formulation is that
the specimen. The shrinkage arises in the specimen it- the compliance development J (tload , t0 ) is described
self due to the fact that fresh, then gradually hardening, by piece-wise linear curves in logarithm of time span
concrete is used in the test. after loading tload . In the simplest application there are
only two linear curves (see for instance Baant and Wu
4. Description of basic creep models 1974, Jonasson 1977 and Persson 1998) according to:
I. Linear curve with inclination a1 from tload = t0
4.1 New basic creep formulation to tload = t1 , which here is called short-term creep.
II. Linear curve with inclination a2 from tload = t1
To formulate a modulus of elasticity E (t0 ) based on
and further on, which here is called long-term creep.
creep tests a quasi-instantaneous elastic deformation
have to be defined by the choice of an elastic time The principles outlined above are illustrated in Fig. 1.
A description of the creep behaviour with two linear
duration t0 , which gives
curves for each loading age will according to Fig. 1
need the following functions ( J 0 , a1 and a2 ) and pa-
1 rameters ( t0 and t1 ):
J (t load , t 0 ) = + J (t load , t 0 ) (4)
E (t 0 ) J 0 or E = 1 J 0 modulus of elasticity
with t0 load duration for definition of
1 modulus of elasticity
E (t 0 ) = (5) t1 limit between short-term and long-
J (t 0 , t 0 ) term creep
a1 inclination (or logarithmic creep
where J (tload , t0 ) is the creep part associated with the rate) of short-term creep
definition of the elastic modulus in Eq. (5). a2 inclination (or logarithmic creep
rate) of long-term creep
The time duration defining the elastic modulus may The limit for practical application between short-term
be chosen within 0 < t0 0.01 d (Neville et al., 1983), and long-term creep t1 is chosen as a typical value
and here is the definition from Westman (1999) chosen from evaluation of creep tests, which is further dis-
as cussed in section 5.2.
The inclinations a1 and a 2 of the linear curves in Fig.
t 0 = 0.001 d (6) 1 are defined by
Fig. 1 Illustration of creep development as two linear With t0 and t1 fixed and E or J 0 known, the addi-
curves in logarithmic time scale for a specific loading tional needed information is the inclinations a1 and a2 .
age t0 (I = short-term creep and II = long-term creep). The fundamental properties of these inclinations are
176 M. Larson and J. Jonasson / Journal of Advanced Concrete Technology Vol. 1, No. 2, 172-187, 2003
1 1
a1 J (tload , t0 ) = + ( F (tload , t0 ) F0 (t0 , t0 ) )
ai E (t0 ) E0 (12)
a2
For the DPL from Baant and Osman (1976) the
consistent formulation with an elastic modulus is
J (tload , t0 )
(13)
ts
t0
28 d
=
1
( )(
+ ref ( t0 ) + ( tload ) ( t0 )
E (t0 ) Eref
m n n
)
Fig. 2 Fundamental behaviour of function a1 and a2 in
the creep formula presented in Eqs. (7) to (10). where ref , , m and n are free parameters to be de-
M. Larson and J. Jonasson / Journal of Advanced Concrete Technology Vol. 1, No. 2, 172-187, 2003 177
termined by regression when comparing calculated and In for example Byfors (1980), Gutsch (1998) and
tested creep values. Eref is a reference modulus of elas- Bjntegaard (1999) comprehensive testing of the
ticity, which in this study has been chosen as the value modulus of elasticity from very early ages has been
at 28 days of age. t0 is the loading age for definition made. All results show the same tendency with a very
of modulus of elasticity, earlier discussed when intro- rapid (linear in logarithmic time scale) development of
ducing Eqs. (4) to (6). the E-modulus which then abrupt flats out to a much
A problem with the DPL is that it gives too high final slower increase (Fig. 4).
slope of the calculated long term creep curves giving The modulus of elasticity at time t0 equivalent age
raise to negative relaxation values and an improved may be modelled as
model called the Log Double Power Law (LDPL) was
therefore introduced by Baant and Chern (1985b): E (t 0 ) = E ref E (t 0 ) (16)
J (tload , t0 ) =
1 0
+
E0 E0
(( )
ln 1 + 1 ( t0 ) + ( tload )
m n
) (14)
where Eref is a reference value, which here is chosen as
the modulus of elasticity at 28 days age.
If the concept of describing the material behaviour by
in which 0 , 1 , , m and n are model parameters. piece-wise linear curves in logarithm of time described
By introducing a general function, m2 (t0 ) , instead of in section 4.1 is introduced, E may with denotations
the power function for the loading age, 1 (t0 ) m + , according to Fig. 4 be expressed as
the consistent formulation with an elastic modulus
based on Eq. (14) becomes 0 for t0 < ts
b1 log ( t0 ts ) for ts t0 < t B
1 + m2 ( t0 ) ( tload )m3 E (t0 ) = (17)
b1 log ( t B ts ) + b2 log ( t0 t B )
1 for t B t0 < 28 d
J ( tload , t0 ) = + m1 log (15)
E (t 0 ) 1 + m ( t ) ( t )m3 1 for t0 28 d
2 0 0
with m1 , m2 (t0 ) and m3 as model parameters. A commonly used expression for the modulus of elas-
ticity is the equation given in CEB-FIP MC90 whereby
One small change has been introduced in Eq. (15) E , here modified by introducing the apparent setting
compared with Eq. (14) as log, i.e. 10log, is used instead time ts , may be expressed as
of ln. This only means that the model parameters have
different numerical values. The formulation in Eq. (15)
0.5
t t
is here denoted Log Power Law (LPL). E (t 0 ) = exp s 1 1 / 0 s
(18)
In Pane and Hansen (2001) the corresponding 28 ts
formulation to Eq. (15) was used with m2 ( t0 ) ( t0 ) 3
m
45 45
40 40
35 35
30 30
E [GPa]
E [GPa]
25 25
Measured Measured
20 20
Not used
15 Eq. (17) 15
Eq. (17)
10 Eq. (18) 10
Eq. (18)
5 5
Eq. (19) Eq. (19)
0 0
0.01 0.1 1 10 100 0.01 0.1 1 10 100
t 0 [d] t 0 [d]
(a) (b)
Fig. 5 Measured modulus of elasticity according to Byfors (1980) compared to calculated development by means of Eqs.
(16) to (19). In (a) all data have been used in the regression while in (b) only data after 1 day.
formulation, with respect to negative relaxation, are relation to the experimental data.
practically identical due to the fact that the adjustment
in question mostly lay outside the measured data range. 5.2.3 Linear logarithmic model LLM
In the new basic creep LLM model the inclinations a1
5.2.2 Modulus of elasticity and a2 give the increase in creep compliance
As a first evaluation step for all studied methods, an E- J (tload , t0 ) , and they are determined from the meas-
modulus model is fitted to the measured instantaneous ured compliance as shown in Fig. 9. The development
deformation. 1 J (0.001 d, t 0 ) . The calculated E- of the inclinations as a function of the loading age t0 is
modulus development E (t0 ) and the measured values modelled with Eq. (10). It seems possible to set the limit
from concrete LTU66 are shown in Fig. 8. As can be between short-term and long-term creep t1 as a con-
seen the model according to Eq. (18) has, as previously stant value, which in this study has been chosen to be
also shown, difficulties describing the rapid change in 0.1 days. The end values of short-term creep a1 give
the E-modulus development at early ages. The new for- consistent results if they are set to 6010-12/(Palog-unit)
mulation given by Eq. (17) has in this case the best cor- as a maximum ( a1max ) and 0.110-12/(Palog-unit) as a
180 M. Larson and J. Jonasson / Journal of Advanced Concrete Technology Vol. 1, No. 2, 172-187, 2003
LLM I LLM II
0.62 0.62
0.75 0.75
140 0.83 140 0.83
1 1
120 2 120 2
4 4
100 7 100 7
J [10 /Pa]
J [10 /Pa]
80 80
-12
-12
60 60
40 40
20 20
0 0
0.001 0.01 0.1 1 10 100 0.001 0.01 0.1 1 10 100
t load [d] t load [d]
DPL LPL
0.62 0.62
0.75 0.75
140 0.83 140 0.83
1 1
120 2 120 2
4 4
100 7 100 7
J [10 /Pa]
J [10 /Pa]
80 80
-12
-12
60 60
40 40
20 20
0 0
0.001 0.01 0.1 1 10 100 0.001 0.01 0.1 1 10 100
t load [d] t load [d]
Fig. 6 Measured and calculated creep compliance at different loading ages (days) for concrete LTU66 with adjustment for
negative relaxation. Note that LPL does not need any adjustment.
minimum ( a1min ). Regarding the long-term creep a 2 the uncertainties in the evaluation (see a2 at t0 = 0.62 and
maximum end value should be significantly smaller 0.75 days in Fig. 9 (b) but as the formulation parameter
than that for short-term creep and here a maximum a2 has a physical meaning of creep rate it is easy to sort
value ( a2max ) of 3010-12/(Pa log-unit) gives sound re- out and exclude these values from the evaluation. This
sults. The minimum end value ( a2min ) will however vary is done here and is one important advantage of the pre-
depending on the modulus of elasticity at 28 days age sent creep formulation.
Eref , which is further described in Larson and Jonasson The LLM formulations will also result in negative re-
(2003) and Larson (2002). These experiences can to- laxation values but not near as extent as for DPL (or
gether with Eq. (6) be summarised as TPL). Just a minor adjustment is needed by introducing
the limit values (t0 ) age = 2 days and (tload ) age = 10 days
t0 0.001 d = days in Eq. (11).
t1 0.1 d = days The LLM model has been applied to the following
-12 two formulations of the elastic modulus:
max
a1 60 10 /(Pa log-unit) (21) LLM I: E-modulus according to Eq. (17), which also
amin
0.1 -12
10 /(Pa log-unit) is a linear-logarithmic model of the devel-
1
opment of the elastic modulus.
a max
2 30 10-12/(Pa log-unit) LLM II: E-modulus according to Eq. (18) from CEB-
FIP MC90.
When a creep test is too short there might be some As can be seen in Table 3 the LLM formulations
M. Larson and J. Jonasson / Journal of Advanced Concrete Technology Vol. 1, No. 2, 172-187, 2003 181
LLM I LLM II
0.62 0.62
0.75 0.75
140 0.83 140 0.83
1 1
120 2 120 2
4 4
100 7 100 7
J [10 /Pa]
J [10 /Pa]
80 80
-12
-12
60 60
40 40
20 20
0 0
0.001 0.01 0.1 1 10 100 0.001 0.01 0.1 1 10 100
t load [d] t load [d]
DPL LPL
0.62 0.62
0.75 0.75
140 0.83 140 0.83
1 1
120 2 120 2
4 4
100 7 100 7
J [10 /Pa]
J [10 /Pa]
80 80
-12
-12
60 60
40 40
20 20
0 0
0.001 0.01 0.1 1 10 100 0.001 0.01 0.1 1 10 100
t load [d] t load [d]
Fig. 7 Measured and calculated creep compliance at different loading ages (days) for concrete LTU66 without adjustment
for negative relaxation.
show good agreement with all loading ages except at the prerequisite of = 0 only a minor adjustment for nega-
youngest loading age 0.62 days. The LLM II formula- tive relaxation is needed by introducing the limit values
tion also have some difficulties describing the creep
development at the loading age 1 day, which is related 50
to the modelling of the E-modulus according to Eq. (18).
Both LLM formulations have, however, the overall best 45
correlation to the experimental data of the studied creep 40
models. 35
30
E [GPa]
0.62 d 60
140
0.75 d a1 Eq. (10)
120 50
0.83 d
a2 Eq. (10)
40 a1
80 2d
-12
4d a2
60 30
Not used
7d
-12
40 20
a
20
a
0 10
0.001 0.01 0.1 1 10 100 0 ts
t load [d] 0.1 1 10 100
t 0 [d]
(a) (b)
Fig. 9 (a) Measured creep compliance and measured inclinations together with b) measured and calculated inclinations
for concrete LTU66.
5 Measured 5 Measured
LLM I LLM I
4 4
LLM II LLM II
3 DPL 3 DPL
LPL LPL
Stress [MPa]
Stress [MPa]
2 2
1 1
0 0
-1 0 2 4 6 8
-1 0 2 4 6 8
-2 -2
-3 -3
Time [d] Time [d]
Fig. 12 Measured and calculated thermal stress devel- Fig. 13 Measured and calculated thermal stress devel-
opment for concrete LTU66 with creep formulations ad- opment for concrete LTU66 without adjustment for nega-
justed for negative relaxation. tive relaxation.
184 M. Larson and J. Jonasson / Journal of Advanced Concrete Technology Vol. 1, No. 2, 172-187, 2003
Table 4 Variation coefficients (Eq. 23) for the studied models with adjustment for negative relaxation.
Stress state N LLM I LLM II DPL LPL
Compression 27 0.05 0.05 0.04 0.11
Tension 42 0.05 0.05 0.06 0.08
Table 5 Variation coefficients (Eq. 5.8) for the studied models without adjustment for negative relaxation.
Stress state N LLM I LLM II DPL LPL
Compression 27 0.04 0.05 0.04 0.11
Tension 42 0.05 0.05 0.07 0.08