J Therm Anal Calorim (2016) 124:227239

DOI 10.1007/s10973-015-5131-x

Thermal analysis and adiabatic calorimetry for early-age concrete

members
Part 2. Evaluation of thermally induced stresses
Yun Lin1

Hung-Liang Chen1

Received: 20 July 2015 / Accepted: 24 October 2015 / Published online: 18 November 2015
Akademiai Kiado, Budapest, Hungary 2015

Abstract In this study, a finite element model was

developed to perform the stress analysis on early-age
concrete members to predict the thermally induced stresses
and the associated cracking risk. FORTRAN subroutines
were created for ABAQUS finite element program to
enable solution-dependent material properties in the thermal stress calculation. Youngs modulus development,
strength development, tensile creep, and compressive creep
behaviors at early age were experimentally obtained, and
these material properties were incorporated in the subroutines. Two 1.2-m concrete cubes were constructed with
embedded temperature sensors and vibrating wire strain
gages to verify the simulation results. Results showed that
the calculated temperature and strain values correlated well
with the measured field data. Additionally, visual cracks
were confirmed at the predicted locations on the concrete
cube. It is concluded that the method developed in this
study is capable of determining the thermally induced
stresses of early-age concrete members.
Keywords Mass concrete  Heat of hydration  Early-age
concrete  Thermal stress  ABAQUS

& Yun Lin

linyun2000@gmail.com
Hung-Liang Chen
roger.chen@mail.wvu.edu
1

Department of Civil and Environmental Engineering, West

Virginia University, P.O. Box 6103, Morgantown,
WV 26506-6103, USA

Introduction
The heat generation from cement hydration leads to a
temperature rise, especially at the core of a large concrete
member. At concrete surfaces, the temperatures are relatively lower due to surface heat loss from external ambient
cooling. The created temperature differential may lead to
high tensile stresses at the concrete surfaces and produce
surface cracking. The most common thermal control
practice is to limit the temperature differential between the
center and the surface of the concrete structures. However,
temperature differential is not conclusive enough to
determine the cracking risks due to thermal stresses. Nagy
and Thelandersson [1] pointed out that the development of
concrete Youngs modulus is very important in thermal
stress modeling. Gutsch and Rosatasy [2] suggested the
importance of tensile strength development and the tensile
creep behavior in terms of cracking potentials.
Lawrence et al. [3] reported that temperature differential
alone was not sufficient to determine thermal stresses.
Instead, a thermal stress analysis considering the changes
of concrete material properties, such as thermal expansion
coefficient, Youngs modulus and viscoelasticity should be
used.
During early age, the nonuniform temperature profile
distribution causes disproportionate thermal expansions
within the concrete body. The surface of concrete in lower
temperatures can be under high tensile stresses due to
relative thermal expansions from internal concrete. The
heating effect due to hydration and the cooling effect due
to surface heat loss occur simultaneously. Therefore, the
surface of concrete is under tension once concrete is set
until the hydration heat is fully dissipated to the environment. The reversal of stress may occur beneath the surface
of concrete when the concrete passes from the heating

123

228

Y. Lin, H.-L. Chen

Tensile stress and strength

phase to the cooling phase. Whether the high surface tensile stresses can cause cracking is depending on the stressto-strength ratio at the critical locations. During the
hydration process of the early-age concrete, both the
thermally induced stresses and the concrete strength are
being developed but at different rates. Cracks are most
likely to occur at the critical locations where tensile stress
exceeds the tensile strength. Figure 1 originally presented
by Tia et al. [4] depicts an example of thermal stress and
concrete tensile strength development. The cracking zone
in the figure refers to the time when tensile stress exceeds
tensile strength. In practice, this cracking time zone is most
likely to occur within 12 days after concrete placement,
depending on the member geometry, size, boundary
restraint and the ambient temperature variations.
The development of thermally induced stresses is a
complicated phenomenon which includes the variability of
temperature distribution, concrete thermal and mechanical
properties, and the viscoelastic behavior of early-age concrete. In recent years, finite element models have been used
to predict the thermally induced stresses of early-age
concrete members. Waller et al. [5] presented a model
using CESAR-LCPC which included two modules, TEXO
and MEXO, to perform the thermal analysis and stress
analysis on concrete structures. Wu et al. [6] described the
procedures calculating thermally induced stresses for a
wall element using ANSYS. Tia et al. [7] evaluated bridge
footing elements with wooden formwork using TNO Diana
software. Their research findings are very helpful to this
topic; however, the modeling procedure of the viscoelastic
behaviors due to tensile or compressive stresses was not
detailed enough for replication purposes.
Researchers have emphasized the importance of concretes viscoelasticity, which is crucial in calculating
thermal stresses. Bazants B3 model [8], which was

designed for long-term creep behaviors, has been widely

adopted recently to describe creep behaviors of early-age
concrete. stergaard et al. [9] improved the B3 model on
the early-age creep behavior by adjusting the aging term,
while Wei and Hansen [10] made an adjustment on the
later ages by modifying the flow term of B3 model.
However, the temperature, although often ignored in the
creep calculation, has a significant effect on early-age
concrete creep behavior. Using the equivalent age to consider the temperature effect in creep was suggested by
Bazant and Baweja [11]. Atrushi [12] also showed the
usage of equivalent age on the modified double power law
(DPL) with some experimental proof. Luzio and Cusatis
[13] validated the solidificationmicroprestressmicroplane (SMM) model considering moisture variation and
moisture diffusion associated with environmental exposure
and internal water consumption. This paper described the
thermal stress calculation of early-age concrete using a
modified B3 model considering the aging and temperature
effect in a variable loading and temperature environment.
A thermal stress calculation algorithm with experimental
verification is presented in this paper.

Concrete cube construction

To study the development of thermally induced stresses,
two 1.2-m concrete cubes were constructed (Fig. 2). For
both cubes, temperature measurements were taken at the
center of the cube and 5 cm from the side surface and the
top surface. The details about the temperature predictions
and measurements were presented in Lin and Chen [14].
Additionally, Geokon vibrating wire strain gages (Model
4200, gage length 15.25 cm) were installed to measure the
strain changes within a concrete member during the early

Cracking zone

Tensile strength
Tensile stress

Time
Liquid

Solid

Fig. 1 Thermal stress and tensile strength development with crack

initiation

123

Fig. 2 Pictures of internal sensor installations before cube 2 casting

Thermal analysis and adiabatic calorimetry for early-age concrete members

Table 1 Concrete mix design/kg m-3
Material

Cement

Water

CA

FA

AE/Lm

Quantity

335

139

969

844

0.067

-3

-3

WR/Lm
1.0

CA coarse aggregates, FA fine aggregates, AE air entraining agent,

WR High-range water reducer

ages after concrete placement. Strain measurements were

taken at location A for cube 1 and locations A, B, and C for
cube 2. The locations of these sensors are illustrated in
Fig. 2; locations A and B are 10 and 5 cm from the center
of the side surface, while location C is 2.5 cm from the
center of the top surface. The mix design and cement
chemical compositions used for the two cubes are shown in
Table 1 and Table 2. The cement chemical composition
shown in Table 2 was analyzed by the material testing
laboratory of West Virginia Department of Transportation
(WVDOT). Figure 3 shows the pictures of the vibrating
wire strain ages and the temperature loggers used and the
1.2-m concrete cubes.

229

the concrete strength and modulus at any given time, is a

function of the equivalent age, te. The equivalent age can
be calculated using the Arrhenius equation, Eq. (3.3) [15],
which is depending on the concrete temperature history and
the activation energy, Ea. The activation energy of this
particular mix design from Table 1 was determined to be
41,800 J mol-1 by Yikici and Chen [16] following ASTM
C 1074-10 procedures. The ultimate degree of hydration,
au, can be calculated using Eq. (3.2) [17]. The hydration
parameters, s and b, were two constants depending on the
mix design. s = 14.0 and b = 0.94 were determined from
the adiabatic temperature rise tests by Lin and Chen [14].
 b !
s
ate au exp 
3:1
te
1:031  w=c
0:194 w=c
 

t
Ea
1
1
te r exp

dt
R 273 Tr 273 Tc t
0
au

3:2
3:3

where s, b hydration parameters, R universal gas constant,

Tc(t) concrete temperature at time t, Tr reference temperature, 23 C, Ea activation energy/J mole-1.

Mechanical properties of early-age concrete

Concrete strength testing
In order to calculate the thermally induced stresses, accurate estimations of the concrete tensile strength and the
modulus of elasticity development are crucial. The degree
of hydration (a) calculated using Eq. (3.1), used to estimate

The compressive strength development curves obtained

from the compressive cylinder test results using 0.15 by
0.3 m cylinders cured at a constant temperature of 23 C

Table 2 Cement chemical composition/%

Components

CaO

SiO2

Al2O3

Fe2O3

MgO

SO3

Na2O

K2O

Percentages

62.3

20.22

4.8

3.1

2.51

3.0

0.034

0.76

Fig. 3 Pictures of vibrating

wire strain gage and
temperature logger and cube 1
and cube 2

123

230

Y. Lin, H.-L. Chen

(a)

(b)

35

Compressive strength/MPa

Compressive strength/MPa

Fig. 4 Relationship between

compressive strength and
degree of hydration

30
25
20
15
10
5
0
0

Tensile strength/MPa

y = 45.53x 1.71

25

R 2 = 0.9969

20
15
10
5
0

0.2

0.4

0.6

0.8

Youngs modulus development

3
2.5
2
1.5

Test result

1
Wight & Macgregor
0.5
0

Eqvalent age/day

Fig. 5 Splitting tensile strength test results in comparison with Wight

and Macgregor [19]

(Fig. 4a). Schutter [18] reported that concrete compressive

strength and degree of hydration (a) had a linear relationship. Results from the current mix design (Table 1) also
showed a strong linear correlation between the degree of
hydration and the compressive strength (Fig. 4b). This
linear relationship, Eq. (3.1.1), will be used to describe the
concrete strength at any given degree of hydration.


fc0 45:53a  1:71 a  0:04; fc0  0
3:1:1
Splitting tensile strength development was determined
according to ASTM C496 using 0.15 by 0.3 m cylinders.
Wight and MacGregor [19] presented Eq. (3.1.2) obtained
from the mean split cylinder strength (fct) from a massive
database. The curve described by Eq. (3.1.2) has a high
correlation with the current splitting tensile experiment
results; the comparison of the test results and the predicted
values using Eq. (3.1.2) are shown in Fig. 5. For modeling
purposes, the splitting tensile strength development can
also be expressed in terms of degree of hydration shown in
Eq. (3.1.3) by inserting Eqs. (3.1.1)(3.1.2).
q
fct 0:53 fc0 inMPa
3:1:2
p
3:1:3
fct 0:53 45:53a  1:71 a  0:04; fct  0

123

30

Degree of hydration

Equivalent age/day

35

In this study, accurately assigning elastic modulus for the

concrete under tension is essential for thermal stress calculations. The tensile modulus development curve was experimentally obtained. The specimens used in tensile modulus
test were 0.9-m-long dog-bone specimens, each with a
0.1 m 9 0.1 m mid-cross section (Fig. 6a). One vibrating
wire strain gage was embedded in the middle of the concrete
specimen. A steel hook was placed at each end in order to
apply tension. The specimen was loaded in direct tension nondestructively (Fig. 6b) using a force between 600 and 2400
Newton (approximately 10 % stress-to-strength ratio)
depending on the concrete maturity. The strain due to external
tensile stress was measured by the vibrating wire strain gage.
The specimen was loaded four times for each data point to
ensure the accuracy. Each loading lasted approximately 10 s
to minimize any creep effect. The tensile modulus values
shown in Fig. 6c was obtained based on the measured stressto-strain ratios. The relationship between compressive
strength and Youngs modulus for this particular mix design
can be determined using curve fitting method. A best-fit
exponential function as shown in Eq. (3.2.1) is used to
describe the development of the Youngs modulus. In
Eq. (3.2.1), the compressive strength fc0 can be expressed
with degree of hydration (Eq. 3.1.1). The elastic modulus can
also be expressed in terms of degree of hydration as in
Eq. (3.2.2). The Youngs modulus was assumed identical in
both tensile and compressive directions.
Ec 5407fc0

0:492

Ec 540745:53a  1:710:492

3:2:1
a  0:04; Ec  0
3:2:2

Thermal expansion coefficient

After performing the tensile modulus testing, the dogbone sample with embedded vibrating wire strain gage

Thermal analysis and adiabatic calorimetry for early-age concrete members

(a)

(b)
Rigid
Frame

Loaded
Concrete
Specimen
Vibrating
wire gage

Weights

(c)

30000

Tensile modulus/MPa

Fig. 6 Tensile modulus testing

setup and results

231

26000
22000
Test results

18000

Eq. (3.2.2)

14000
10000

Equivalent age/day

was reused to test the coefficient of thermal expansion

(CTE). The dog-bone specimen was submerged into a
temperature controlled water tank and placed on frictionless base provided by ball bearings. The specimen
was able to freely expand and contract due to temperature
changes. The strain data of the dog-bone specimen was
recorded, while the water temperature was controlled to
slowly rise and drop. The CTE test was repeated three
times, and the test results correlated closely with an
average thermal expansion coefficient of 8.53 microstrains per C at 28 days of age. The thermal expansion
coefficient is assumed to be a constant for simplicity. The
variation of CTE of concrete is difficult to measure
because of the temperature influence of the concrete
maturity especially at early age; it was shown by
McCullough and Rasmussen [20] that concrete CTE
variation after 24 h of age could be assumed negligible
where the CTE before 24 h decreased noticeably. It is
noted that CTE is also depending on the moisture level
inside concrete, and it was assumed that the moisture
level in the current concrete cube is close to 98100 %
[14].
Basic creep of early-age concrete under constant
load
The viscoelastic behavior of early-age concrete plays an
important role in calculating thermal stresses. The tensile
creep behavior of early-age concrete is complicated.

Tensile creep tests performed by Gutch and Rostasy [2]

showed pronounced viscoelasticity when load was applied
at early age. Umehara and Uehara [21] and Atrushi [12]
demonstrated the influence of temperature on early-age
tensile creep. stergaard et al. [9] and Atrushi [12] showed
the strong loading age dependency in the early ages.
Bazant and Baweja [11] presented a mathematical
expression of structural creep law (B3 model) shown in
Eq. (3.4.1). With experimentally determined empirical
constants (q1q4), B3 model was often found accurate in
terms of correlating with the experimental data.
et
r
h
i
t
0:1
q1 q2 Qt; t0 q3 ln 1 t  t0
q4 ln 0
t
3:4:1

J t; t0

where


r 1=r
Qf
Qt; t0 Qf t0 1
Z t; t0
h
i
2
4 1
Qf t0 0:086t0 9 1:21t0 9
h
i
1=2
0:1
Z t; t0 t0
ln 1  t  t0
r 1:7t0

0:12

8:0

t current age in days (t = 0 is when water is added to the

mixture), t0 loading age in days, q1, q2, q3, and q4 empirical
constants

123

232

Y. Lin, H.-L. Chen

q02 q2

t0
t 0  q5

3:4:2

Temperature is also an important factor, which has two

different effects on the creep behavior of early-age concrete. From the maturity concept, higher curing temperature will accelerate the hydration process and increase the
concrete maturity at the time of loading and therefore
decrease the specific creep. However, the creep deformation at early age increases significantly as the temperature
increases. Atrushi [12] stated that the increasing effect is
much greater than the decreasing effect. In Atrushis
experimental results, a significant increase in tensile creep
was found due to the effect of temperature increase. In
order to consider the effect of the temperature, the equivalent age concept was used by Bazant and Baweja [11] and
Atrushi [12]; the equivalent age was used to replace the
regular age in the places of the loading age and the loading
duration, where they found better agreements between the
theoretical and experimental results. Hence, the modified
B3 model can be expressed as shown in Eq. (3.4.3).

 ete
J te ; te0
r



t0
q1 q2 0 e Q te ; te0
hte  q5
0:1 i
te
q3 ln 1 te  te0
q4 ln 0
te

3:4:3

To measure the basic tensile creep of early-age concrete,

surface sealing is important because tensile loading can
significantly accelerate the drying effect and lead to more
load-induced drying shrinkage. In this study, during each
creep testing, two identical 0.9-m dog-bone specimens with
a 0.1 m 9 0.1 m cross section at the mid-span region were
used. The concrete molds and sensor installation were the
same as shown in Fig. 6. Both specimens were sealed with
epoxy paint plus four layers of plastic wraps immediately
after unmolding (1 h prior to the loading). Epoxy paint
creates an adhesion between the plastic wrap and the
concrete surfaces to further prevent surface drying. Both
specimens were kept in the same room with a controlled
temperature of 23 C and 50 % humidity level. One of the
specimens was loaded with a tensile stress of 0.13 MPa
(approximately 10 % stress-to-strength ratio), while the
other was kept free to deform. Although the loading

123

magnitude is small with respect to its tensile strength, the

specific creep was assumed to be un-affected.
Hauggaard et al. [22] reported that the specific creep
response of early-age concrete was found to be unchanged
when a stress-to-strength ratio is below 60 %. Similar
conclusion was drawn by Atrushi [12] in his tests up to
80 % stress-to-strength ratio.
The strain measurements for both specimens (loaded
and free) were taken using Geokon vibrating wire strain
gages. The difference in the monitored strain between the
two specimens divided by the loading magnitude is calculated to show creep compliance. The tensile creep test
for the specific mix design (Table 1) was performed three
times at three different loading ages (0.75, 1, and 10 days),
and the results are shown in Fig. 7. As shown in Fig. 7, all
of the three test results can be described by the modified B3
model (Eq. 3.4.3). The best-fit empirical constants are
shown in Table 3.
Although tensile and compressive creep models were
often assumed to be the same for simplicity, different
tensile and compressive creep behaviors were discovered
by numerous researchers [12, 23, 24]. The compressive
creep model of the current mix design was obtained from
an existing model by Atrushi [12] shown in Eq. (3.4.4).The
double power law (DPL) developed by Bazant and Osman
[25] has been widely used to model compressive creep
behavior for hardened concrete. Atrushi [12] modified the
DPL for early-age concrete by incorporating the temperature effect observed in early age. The equation includes
0
equivalent age at loading (te), the current equivalent age
0
(te), the Youngs modulus at loading (E(te)), and three
additional creep parameters (/, d and p). This modified

Creep compliance/ MPa1

stergaard et al. [9] found that for early-age tensile

creep, B3 model may underestimate the specific creep. The
early-age concrete exhibits much greater viscoelasticity.
They modified the q2 constant using Eq. (3.4.2) to amplify
the age dependency for early-age concrete, where q5 is
always less than the physical loading age (t0 ). In their
research, with a very early loading age at 16 h, the best-fit
value of q5 was found to be 14 h.

100
90
80
70
60
50
40
30
20
10
0

Eq. 3.4.3 (0.75 day)

Eq. 3.4.3 (1 day)
Eq. 3.4.3 (10 day)
Experiment (0.75 day)
Experiment (1 day)
Experiment (10 day)
0

10 11 12 13

Time/day

Fig. 7 Comparison of creep compliance between experimental

results and Eq. (3.4.3)
Table 3 Best-fit empirical constants for Eq. (3.4.3)
Constant

q1

q2

q3

q4

q5

Value

0.3

24.0

65.0

0.5

0.2

Thermal analysis and adiabatic calorimetry for early-age concrete members

233

double power law (M-DPL) is shown in Eq. (3.4.4).Since

the mix design used in this study (Table 1) was similar to
the Base-0 mix from Atrushi [12], same creep parameter
values were used for the current calculation. The values of
/, d, and p were 0.75, 0.2, and 0.21, respectively, obtained
from Atrushi [12]. The M-DPL with these pre-determined
creep parameters was used to account for compressive
creep of concrete in the FEM calculations.



p

1
J te ; te0  0  1 /te0d te  te0
3:4:4
E te

Figure 8b demonstrates a typical creep compliance J(t, t0 )

of a concrete specimen under constant loading. J(t, t0 ) is
defined as the ratio of the total strain (e(t) = etotal =
ecr ? eins) and the stress (J(t, t0 ) = e(t)/r). The creep
coefficient (Ccr) is defined as the ratio of the creep strain
(ecr) and the instantaneous strain (eins) due to the loading
(Eq. 3.5.3). Equation (3.5.4) can be derived according to
Fig. 8b. The growth of Youngs modulus is considered in
this model for early-age concrete as shown in Fig. 8b.
r
E
3:5:1
eins

Modeling of viscoelastic behavior

Eeff

Before cracking, the concrete material is usually assumed

linear elastic as in Eq. (3.5.1) for one-dimensional stress;
the elastic modulus is the ratio of the stress and the
instantaneous strain (eins). However, the early-age concrete
exhibits high viscoelastic behavior which causes a change
in effective modulus, and the analytical response is illustrated in Fig. 8. To simplify the creep calculation, an
effective modulus (Eeff) is used in the FEM modeling. The
effective modulus (Eq. 3.5.2) represents the ratio of the
stress and the total deformation as plotted in Fig. 8a.

Ccr t; t0

(b)

Eeff

total

ins

Creep compliance J (t,t )

(a)

cr(t )

1
E (t )
t

1
E (t )

ins(t )

Time t

(c)

n
n1

Load

n2

3
2
0

n 2 n 1
Time

n 1

r
r
E

etotal eins 1 Ccr 1 Ccr

ecr t etotal t  eins t

eins t
eins t

0
1
J t; t0 r  eins t J t; t  Et

Ccr t; t
1
eins t
Et
0
EtJ t; t  1

3:5:3

3:5:4

The creep behavior becomes more complicated when

the concrete member is under variable loading such as
thermally induced stresses which would change due to the
variations of the temperature gradient and the mechanical
properties. The variable loading problem can be solved
using the superposition principle. At time n, the total load
can be decomposed to n small increments (Drt). Each
loading increment has its individual loading time (t0 = i)
and loading duration (n - t0 = n - i) (Fig. 8c). Without
considering creep, the total stress at t = n, rtotal(n) can be
expressed as the summation of the load increments
(Eq. 3.5.5). When creep is considered, each loading
increment (Drt_cr) can be derived as in Eq. (3.5.6) and the
total load can be expressed as shown in Eqs. (3.5.7) or
(3.5.8). The actual overall stress release percentage due to
all of the load increments can be expressed as the ratio of
rtotal_cr(n) and rtotal(n). Thus, the overall creep coefficient
and effective modulus can be derived as in Eqs. (3.5.9) and
(3.5.10), respectively. In the FEM analysis, only basic
creep was considered. Basic creep refers to the strain
observed on sealed specimens due to sustained loading
[26]. In reality, all the surfaces of the two 1.2 m3 were
covered for entire 5 days after casting. The influences from
drying effect are assumed negligible.
rtotal n Dr1 Dr2 Dr3 . . . Drn2 Drn1
Drn
3:5:5

Load duration of

n2

Fig. 8 Illustration of effective modulus, the creep compliance and

variable load decomposition

3:5:2

rtotal n

n
X

Dri

i1

123

234

cr

Drt
Drt

0
1 Ccr n; t EnJ n; t0

rtotalcr n

rtotal
Ccr

3:5:6

Dr1
Dr2
Dr3

...
EnJ n; 1 EnJ n; 2 EnJ n; 3
Drn2
Drn1

EnJ n; n  2 EnJ n; n  1
Drn
3:5:7

EnJ n; n
n
X

Dri
3:5:8
E

n
J

t
n; t0 i
i1
Pn
rtotal n
Dri
 1 Pn i1 Dri  1
overall n
rtotal cr n
i1 EnJ n;i
cr n

3:5:9
Eeff n

E n
1 Ccr overall n

3:5:10

It is also noted that relationship between the creep

compliances and loading could become nonlinear if the
loading stress and strength ratio becomes very high [26];
the creep deformation would be further increased due to
nonlinear creep behavior at high stress-to-strength ratio
[12]. In this study, since only linear creep behavior is
considered, when stress-to-strength ratio is beyond 80 %,
the current creep model would overestimate the thermal
stresses, and this will be discussed in Sect. 5.

Discussion of maturity method on this application

The maturity method has been used to estimate in situ
concrete strength since late 1940s. Many researchers have
discovered that high-temperature curing may have a negative effect on the long-term concrete strength gain. Carino
and Lew [27] described the crossover effect due to hightemperature curing. They suggested that maturity method is

(a)
Compressive strength/MPa

Fig. 9 Strength development

curves for concrete cylinders
cured in three different
temperatures

more reliable on predicting the relative strength rather than

absolute strength. Tepke et al. [28] concluded that hightemperature curing affect the strengthmaturity relationship. Kim and Rens [29] experimented on three sets of
concrete cylinders in three different curing temperatures of
40, 50, and 60 C. Their results showed higher-temperaturecured samples exhibited lower ultimate strength at the
equivalent age of 28 days. In this study, three sets of concrete cylinders with the same mix design (Table 1) were
cured at 23, 40, and 50 C. The strength development curves
plotted versus equivalent age (Eq. 3.3) are shown in Fig. 9a.
Concrete cylinders cured at 23 and 40 C showed very
similar strengthmaturity relationship, while the cylinders
cured at 50 C showed lower strength. It suggests that
maturity method works for concrete with curing temperature
between 23 and 40 C but may not work for 50 C or higher
temperature. For verification purposes, another batch of
concrete with the same mix design was cast and cured at 23,
30, and 40 C. As shown in Fig. 9b, maturity method
worked quite well up to 7 days of equivalent age. In mass
concrete applications, the concrete temperatures are normally higher due to the relatively larger member sizes. At
the center of a mass concrete member, the temperature can
be kept higher than 50 C for an extended period. However,
since only the surface tensile stresses are critical, the temperature near the surface is of particular concern and the
temperature is usually much lower due to external heat loss.
For both 1.2-m concrete cubes constructed, the surface
maximum temperatures were about 4546 C and quickly
decreased after the maximum temperatures were reached.
To verify whether the concrete surface strength of these
two cubes can be predicted using the maturity method,
another compressive strength test was performed using a
set of concrete cylinders (0.1 m 9 0.2 m) cured in a temperature history similar to the surface temperatures experienced by the surfaces of the cubes (Fig. 10a). Similar to
Fig. 4a, the strength development curve from cylinders of

(b)

35

Compressive strength/MPa

Drt

Y. Lin, H.-L. Chen

30
25
20
15
23 C

10

40 C
5
0

50 C

30
25
20
15
23 C
10
30 C
5
0

Equivalent age/day

123

35

40 C
0

Equivalent age/day

Thermal analysis and adiabatic calorimetry for early-age concrete members

(b)

50

Compressive strength/MPa

(a)

45

Temperature/C

Fig. 10 a Surface temperature

histories of the two cubes and
the variable curing temperature,
b measured compressive
strength of the specimens cured
in different temperature
histories

235

40
35
30

Cube #1

25

Cube #2

20
15
0

Curing temperature
10

20

30

40

50

60

70

35
30
25
20
15

23 C

10

Match curing

23 C (Fig. 4(a))

80

30

Time/h

identical concrete cured at 23 C was also obtained. Figure 10b shows that the compressive strength of the cylinders with this variable temperature curing can be predicted
by the strengthmaturity relationship. These results indicate that maturity method may not accurately predict the
strength for long-duration high-temperature curing at a
constant 50 C or above, but it is applicable for the strength
prediction of the concrete experiencing short-duration
high-temperature curing, such as those experienced on the
surface of the 1.2 m3.

The computation of thermally induced stresses for earlyage concrete contains two parts: thermal analysis and stress
analysis. Thermal analysis was first conducted using 3-dimensional finite element method (FEM), and the calculated
temperature compared quite well with the experimental
measurements [14]. Figure 11 shows the temperature
comparisons between the FEM predictions and the measurements at the side surface (5 cm inside the surface) and
at the center of the 1.2-m concrete cube. The details of the

(a)

70

Experiment (Center)

Cube 1

(b) 70

Experiment (side)
FEM (Center)

50

FEM (side)

40

Ambient

Temperature/C

Temperature/C

60

30
20
10

90

120

150

180

Equivalent age/h

temperature calculations were described in Lin and Chen

[14]. The temperature profile results are used for the stress
analysis herein.
For thermal stress analysis, the complexity of variation
in material properties and viscoelastic behavior required us
to develop a subroutine USDFLD to account for the
change of material properties and viscoelastic behavior at
every time increment. For each individual element, the
program first calculates the equivalent age (te) and degree
of hydration (a) based on the calculated temperature history. The compressive elements and tensile elements are
treated independently. The difference in creep behavior
between elements in tension and compression is considered
(Sect. 3.4). The modified B3 model is used to describe the
tensile creep, and the modified double power law (M-DPL)
is used to describe the compressive creep behavior. To
simplify the analysis, it was programmed to check the
maximum principal stress in each element at every time
step to identify tensile and compressive elements.
The creep stress loading magnitude changes because of
temperature variation. Load decomposition and superposition rules were used to calculate the overall creep coefficient. The effective modulus was then used to incorporate

Simulation process and results

Fig. 11 Center and the side

temperature predictions of cube
1 and cube 2

60

Experiment (Center)

Cube 2

60

Experiment (side)

50

FEM (Center)
FEM (side)

40

Ambient

30
20
10
0

20

40

60

Time/h

80

100

120

20

40

60

80

100

120

Time/h

123

236

Y. Lin, H.-L. Chen

both elastic deformation and viscoelastic deformation due

to the creep behavior (Sect. 3.5). At each time step, the
thermal stress was computed for each element based on the
calculated thermal gradient, current elastic modulus, and
overall creep coefficient. Finally, the equivalent age,
degree of hydration, and stresses in principle directions for
each element were stored for the next time step. Figure 12
shows the programming algorithm of the stress analysis.
The analysis used a fixed time increment of 1 h. The entire
algorithm was executed for each individual element at
every time step. For simplicity, the Poissons ratio and
coefficient of thermal expansion (CTE) were assumed
constants. Poissons ratio was assumed to be 0.2. CTE was
experimentally obtained to be 8.53 microstrains per C
(Sect. 3.3). The frictional interaction between the bottom
of the concrete cube and the wood base was neglected for
simplicity.
Stress analysis was performed for both 1.2-m concrete
cubes using the ABAQUS program and the above FORTRAN subroutine. The model had 15,625 nodes and
13,824 elements using 3-D 8-node linear element (C3D8R)
with 5-cm element size. Results showed that due to the
temperature evolution and the thermal expansion, the inner
elements were expanding which caused the surface element
to be in tension. The calculated surface tensile stress contour patterns for the two cubes are similar. As shown in

Fig. 12 Algorithm of the stress

analysis

Fig. 13 (cube 1), the tensile strength of the concrete was

exceeded by maximum thermal stress at the center locations of the edges (shown as gray color) at 16 h after
concrete placement. The predicted tensile stresses and the
estimated tensile strength at the critical locations for both
cubes are compared in Fig. 14. The estimated tensile
strength history was calculated using Eq. (3.1.3) for the
concrete at that location; concrete strength was temperature
history dependent, and hence, location dependent. The
FEM result showed that cube 1 was likely to crack at these
locations because the maximum thermal stress exceeded
the tensile strength. From the experimental observation,
four large cracks with an approximate cracking length of
0.6 m were found at the center of the top edges (0.3 m on
the top surface and 0.3 m downward to the side surfaces).
Figure 13 shows the calculated stress distribution and the
actual crack locations for cube 1. No other crack was found
at side or bottom edges of cube 1. The reason can be the
nonuniform strength distribution due to vibration compaction; the top surface is typically found to be the weakest
part of the concrete cube [16]. For cube 2, the predicted
maximum stress was shown meeting the estimated tensile
strength at the critical locations (Fig. 14); however, no
thermal crack could be visually identified on any surface of
cube 2. The reason for cube 1 to have larger stress magnitudes in comparison with cube 2 was because of a larger

Start (time = n)
Load element temperature history and calculate its equivalent
age (te) [Eq. (3.3)].and degree of hydration () [Eq. (3.2)].

Yes

S(n1) > 0

No

Calculate tensile modulus,Et(n) [Eq.(3.2.2)].

Calculate compressive modulus,Ec(n) [Eq.(3.2.2)].

Calculate each individual tensile creep

compliance, J(t,t ) (t = 1,2,...,n) using
modified B3 model [Eq. (3.4.3)].
Calculate the overall creep coefficient,
Ccr_overall [Eq. (3.5.9)]
Calculate tensile effective modulus, Eeff(n)
using Calculate Et(n) and Eq. (3.5.10).

Calculate each individual compressive creep

compliance, J(t,t ) (t = 1,2,...,n) using M-DPL
model [Eq. (3.4.4)].
Calculate the overall creep coefficient, Ccr_overall
[Eq. (3.5.9)]
Calculate compressive effective modulus, Eeff(n)
using Calculated Ec(n) and Eq. (3.5.10).

Input Material properties:

Modulus: Eeff(n)
Thermal expansion coefficient: 8.53 C1
(Sect. 3.3)
Poissons ratio; 0.2 (constant)
Assemble global stiffness matrix and calculate element stress
based on the temperature profile and material properties.
Store history values of calculated element equivalent age (te) and stress ().
Time step Advance: n = n+1.
Return to start

123

Subroutine: USDFLD

Load element stress history (S) in principle direction.

Thermal analysis and adiabatic calorimetry for early-age concrete members

237

s/MPa
2.97
2.39
2.35
2.10
1.95
1.80
1.65
1.51
1.37
1.22
1.08
0.93
0.79
0.64
0.50
0.35
0.21
0.06

Fig. 13 Predicted stress field of cube 1 at 16 h

(b)

3.5

Tensile stress/strength/MPa

(a)
Tensile stress/strength/MPa

Fig. 14 Comparison of
calculated thermal stress and
estimated tensile strength

3
2.5
Tensile strength

Thermal stress

1.5
1

Cube 1

0.5
0

40

20

80

60

100

120

3.5
3
2.5
Tensile strength

Thermal stress

1.5
1
Cube 2
0.5
0

20

40

Time/h

(a)

(b)

250

Micro-Strain

Micro-Strain

150
100
Calculated

50

120

40

50

40

50

150
100
Calculated
50
0

Experiment

10

20

30

40

50

10

(d)

250

30

300
250

Micro-Strain

200
150
100
Calculated
50
0
0

20

Time/h

Time/h

(c)

100

250

Experiment
0

80

200

200

Micro-Strain

Fig. 15 Comparison of
calculated and measured strain
changes at the locations near the
concrete surfaces a cube 1
location A, b cube 2location
A, c cube 2location B and
d cube 2location C

60

Time/h

Experiment
10

20

30

Time/h

40

50

200
150
100

Calculated

50

Experiment

10

20

30

Time/h

123

238

ambient temperature drop at the night right after the cube 1

was constructed (see Fig. 11); an 18 C (Fig. 11) drop in
ambient temperature at the first night after cube 1 construction caused a significant increase in thermal stresses.
It is noted that the current FEM model assumes a creep
model that is linear to the applied stresses. The nonlinear
creep behavior due to high stress-to-strength ratio is not
considered in the current model. It was observed by Atrushi
[12] that when applied tensile stress/strength ratio was
about 80 %, the creep coefficient became nonlinear with
respect to the applied stress; more creep strain was
observed at higher stress-to-strength ratio. Therefore, it is
assumed that the current creep model is only able to estimate the allowable thermal stress up to 80 % of the tensile
strength. Because of the linear assumption, the estimation
of the stress at the level higher than 80 % is considered to
be conservative (the estimation is higher than the actual
stress value) using the current model. Although the gray
region of Fig. 13 shows thermal stress exceeded the tensile
strength, it can only be used as a qualitative indication of
high cracking probability. The current modified B3 model
is assumed to be only valid to obtain the thermal stress up
to 80 % of the tensile strength. The nonlinear creep
behavior due to high stress-to-strength ratio needs further
investigation.
The finite element calculated strains were verified with
the measurements. As mentioned in Sect. 2, concrete strain
histories were measured at several locations using vibrating
wire gages. For cube 1, the strain history was measured at
location A. For cube 2, strain values were measured at
locations A, B, and C. The locations are marked in Fig. 2.
The calculated strain histories at these locations were found
to be reasonably close to the measurements as shown in
Fig. 15. The best match was shown at location C of cube 2
(2.5 cm in depth at the center of the top surface) which was
far from any steel reinforcing bar (shown in Fig. 15d). The
calculated strain histories (Fig. 15a) showed some deviations from the experimental measurements, which were
possibly due to the restrains of concrete movement provided by the parallel steel reinforcing bars close to the
sensor during cube 1 testing. On the other hand, during the
cube 2 testing (shown in Fig. 15b, identical location), there
was no parallel reinforcing bar attached to the sensor. It is
noted that the current FEM calculation neglected the effects
of autogenous shrinkage and external restraint; therefore, it
was not included in the predicted strains in Fig. 15. The
autogenous shrinkage of this particular concrete mixture
was measured to be approximately 10 microstrains in sealed
condition after the first 7 days. Hence, the influence of the
autogenous shrinkage was neglected in the calculation. In
general, mass concrete structures are constructed using
concrete with low cement contents that would typically
produce low autogenous shrinkage.

123

Y. Lin, H.-L. Chen

Conclusions
This paper describes a method to perform thermal stress
analysis using ABAQUS program with the aid of user
subroutines. The developed subroutine program uses the
degree of hydration to estimate the variable elastic modulus and strength developments. Concrete tensile creep
and compressive creep behavior were included using a
step-by-step incremental calculation algorithm. The
influences from loading age and temperature effects were
considered in each time increment of the creep models. It
was assumed that the current creep model is able to calculate the thermal stress up to 80 % of the concrete
strength. The finite element simulations were verified by
the experimental data from two 1.2-m concrete cubes
testing. Strain deformations at the locations near the
concrete cube surfaces were measured and correlated
reasonably well with the calculated results.
The concrete cubes have high tensile stresses at the
surfaces, especially at the center of the edges. The tensile
strength development of the concrete at surface locations
can be estimated using the maturity method, and the
cracking risk could be assessed using the stress-to-strength
ratio obtained at the critical locations. Four visible cracks
were found perpendicular to the top four edges on cube #1
as predicted, due to a relatively high ambient temperature
drop at the first night after construction. The method
developed can be used to estimate the thermally induced
stress of concrete members so that precautions can be
implemented prior to concrete casting in order to prevent
unexpected cracking.
Acknowledgements The authors acknowledge the support provided
by the West Virginia Transportation Division of Highways and
FHWA for the research project WVDOH RP#257. Special thanks are
extended to our project monitors, Mike Mance, Donald Williams, and
Ryan Arnold of WVDOH. The authors also appreciate the assistance
from Alper Yikici, Zhanxiao Ma and Jared Hershberger and the
WVDOH concrete technicians from Material, Control, Soil and
Testing Division for the construction of the 1.2-m concrete cubes.

References
1. Nagy A, Thelandersson S. Material characterization of young
concrete to predict thermal stresses. Thermal cracking in concrete
at early ages. In: Springenschmid R, editor. E&FN Spon, 26,
ISBN: 0419187103, 1994.
2. Gutch A, Rostasy FS, Young concrete under high tensile stressescreep, relaxation and cracking. In: Springenschmidt R,
editor. Thermal cracking in concrete at early ages. Proceedings of
the international RILEM symposium, E & FN Spon, London,
1995. p. 111118.
3. Lawrence MA, Tia M, Ferraro CC, Bergin M. Effect of early age
strength on cracking in mass concrete containing different supplementary cementitious materials: experiment and finite-element
investigation. J Mater Civ Eng. 2012;24(4):36272.

Thermal analysis and adiabatic calorimetry for early-age concrete members

4. Tia M, Lawrence A, Ferraro C, Smith S, Ochiai E. Development
of design parameters for mass concrete using finite element
analysis. The Florida Department of Transportation, report
number: 00054863, 2010.
5. Waller V, dAloa L, Cussign F, Lecrux S. Using the maturity
method in concrete cracking control at early ages. Cement Concr
Compos. 2004;26:58999.
6. Wu S, Huang D, Lin FB, Zhao H, Wang P. Estimation of
cracking risk of concrete at early age based on thermal stress
analysis. J Therm Anal Calorim. 2011;105:17186.
7. Tia M, Lawrence A, Ferraro C, Do TA, Chen Y. Pilot project for
maximum heat of mass concrete. The Florida Department of
Transportation, report number: 00093793, 2013.
8. Bazant ZP. Mathematical modeling of creep and shrinkage of
concrete. Chichester: Wiley; 1988.
9. stergaard L, Lange D, Altoubat SA, Stang H. Tensile basic
creep of early-age concrete under constant load. Cem Concr Res.
2001;31(12):18959.
10. Wei Y, Hansen W. Tensile creep behavior of concrete subject to
constant restraint at very early ages. J Mater Civ Eng.
2013;25(9):127784.
11. Bazant ZP, Baweja S. Creep and shrinkage prediction model for
analysis and design of concrete structures: model B3. ACI SP
194. Detroit: American Concrete Institute; 2000. p. 183.
12. Atrushi DS. Tensile and compressive creep of early age. Ph.D.
dissertation, Trondheim, Norway: Department of Civil Engineering, The Norwegian University of Science and Technology, 2003.
13. Luzio GD, Cusatis G. Soli Solidificationmicroprestressmicroplane (SMM) theory for concrete at early age: theory, validation and application. Int J Solids Struct. 2012;50(6):95775.
14. Lin Y, Chen HL. Thermal Analysis and Adiabatic Calorimetry
for Early-age Concrete Members. J Therm Anal Calorim.
2015;122(2):93745.
15. Freiesleben Hansen P, Pedersen EJ. Curing of concrete structures.
Draft DEBguide to durable concrete structures, Appendix 1, 1985.
16. Yikici A, Chen HL. Effect of temperature-time history on concrete strength in mass concrete structure.TRB 92nd annual conference proceeding, Transportation Research Board of the
National Academies, Washington, DC, 2013.
17. Mills RH. Factors influencing cessation of hydration in water cured
cement pastes. Special Report No. 90, Proceedings of the Symposium on the Structure of Portland Cement Paste and Concrete,
Highway Research Board, Washington DC, 1966;406424.

239

18. Schutter GD. Fundamental study of early age concrete behavior

as a basis for durable concrete structures. Mater Struct.
2002;35:1521.
19. Wight JK, MacGregor JG. Reinforced concrete: mechanics and
design. 5th ed. Pearson Prentice Hall: Upper Saddle River; 2009.
p. 589.
20. McCullough BF, Rasmussen RO. Fast-track paving: concrete
temperature control and traffic opening criteria for bonded concrete overlays. US: FHWA 1998, Final Report.
21. Umehara H, Uehara T. Effect of creep in concrete at early ages on
thermal stresses, thermal cracking in concrete at early ages. In:
Springenschmidt R, editors. Thermal cracking in concrete at early
ages. Proceedings of the international RILEM symposium, E&FN
Spon, London, 1995. p. 7986.
22. Hauggaard AB, Darkled L, Hansen PF, Hansen JH, Christensen
SL, Nielsen A. HETEK, Control of early age cracking in concrete, phase 3: creep in concrete. Danish Road Directorate, Lyn
by, 1997.
23. Kanstad T, Bj/stergaard , Sellevold EJ. Tensile and compressive creep deformations of hardening concrete containing mineral
additives. Mater Struct. 2012;46(7):116782.
24. Hilaire A, Benboudjema F, Darquennes A, Berthaud Y, Nahas G.
Modeling basic creep in concrete at early-age under compressive
and tensile loading. The international conference on structural
mechanics in reactor technology, New Dekhi India, 2011;269,
p. 22230.
25. Bazant ZP, Osman E. Double power law for basic creep of
concrete. Mater Struct. 1976;9:49.
26. Bazant ZP, Xi Y. Continuous retardation spectrum for solidification theory of concrete creep. J Eng Mech. 1995;121(2):2818.
27. Carino NJ, Lew HS. The maturity method: from theory to
application. Washington: Structure Congress & Exposition; 2001.
28. Tepke DG, Tikalsky PJ, Scheetz BE. Concrete maturity field
studies for highway application. J Transp Res Board.
2004;1893:2636.
29. Kim T, Rens KL. Concrete maturity method using variable
temperature curing for normal-strength concrete mixes. J Mater
Civ Eng. 2008;20:20.

123