You are on page 1of 16

Engineering Structures 6263 (2014) 148163

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Application of air cooled pipes for reduction of early age cracking risk
in a massive RC wall
Miguel Azenha , Rodrigo Lameiras, Christoph de Sousa, Joaquim Barros
ISISE Institute for Sustainability and Innovation in Structural Engineering, University of Minho, School of Engineering, Civil Engineering Dept., Azurm Campus, 4800-058
Guimares, Portugal

a r t i c l e

i n f o

Article history:
Received 27 May 2013
Revised 2 December 2013
Accepted 13 January 2014
Available online 15 February 2014
Keywords:
Cement hydration
Service life conditions
Thermal shrinkage
Cracking
Numerical simulation

a b s t r a c t
The construction of massive concrete structures is often conditioned by the necessity of phasing casting
operations in order to avoid excessive heat accumulation due to cement hydration. To accelerate construction and allow larger casting stages (usually increasing lift height), it is usual to adopt internal cooling strategies based on embedding water pipes into concrete, through which water is circulated to
minimize temperature development. The present paper reports the use of horizontally placed ventilated
prestressing ducts embedded in a massive concrete wall for the same purpose, in line with a preliminary
Swedish proposal made in the 1990s. The application herein reported is a holistic approach to the problem under study, encompassing extensive laboratory characterization of the materials (including a technique developed for continuous monitoring of concrete E-modulus since casting), in situ monitoring of
temperatures and strains, and 3D thermo-mechanical simulation using the nite element method. Based
on the monitored/simulated results, it is concluded that the air-cooling system is feasible and can effectively reduce early cracking risk of concrete, provided adequate planning measures are taken.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
The combined effect of the exothermic nature of cement hydration reactions and the relatively low thermal diffusivity of concrete
leads concrete structures to endure temperature increases at early
ages, and eventually return to thermal equilibrium with the surrounding environment. These early temperature variations induce
volumetric changes in concrete that are partially restrained by
adjoining previously cast members, or even due to non-uniform
temperature distributions within a concrete member itself. Such
restraint to deformation may induce stresses that can be relevant
enough to induce early age thermal cracking in concrete, which
is usually unacceptable in view of aesthetics, durability and even
structural performance. Contractors usually attempt to avoid this
thermal cracking by adopting concrete compositions and construction schedules that maintain temperature gradients in concrete below prescribed limits, both along time and space [1]. It has
however been recognized that such approach leads to erroneous
conclusions, as several important issues are disregarded [1], such
Corresponding author. Tel.: +351 938404554; fax: +351 253 510 217.
E-mail addresses: miguel.azenha@civil.uminho.pt (M. Azenha), rmlameiras@
civil.uminho.pt (R. Lameiras), christoph@civil.uminho.pt (C. de Sousa), barros@
civil.uminho.pt (J. Barros).
http://dx.doi.org/10.1016/j.engstruct.2014.01.018
0141-0296/ 2014 Elsevier Ltd. All rights reserved.

as the degree of restraint to deformation and the actual mechanical


properties of concrete. In view of the limitations of these temperature-based criteria, it has been widely acknowledged [2] that
more realistic crack risk assessments can be made through multiphysics approaches that encompass numerical simulation of temperatures and corresponding stresses in concrete: thermomechanical analyses. The use of thermo-mechanical simulation
models allows the evaluation of alternative construction scenarios
(for casting procedures, concrete mixes, environmental conditions), and thus permits the optimization of construction without
compromising the safety in regard to thermal cracking. The numerical studies for the assessment of the optimum construction strategy frequently involve diminishing the temperature rise in
concrete at early ages. In fact, if the thermal variation is diminished, the corresponding volumetric changes also decrease, as well
as the developed stresses. The diminishment of early temperature
rises is usually achieved by partial replacement of cement by additions as y ash [3,4], or by cooling water/aggregates before mixing
operations [57], or even by introducing internal cooling pipes in
concrete with cooling uids (usually water) [812]. An attempt
to use air as the cooling uid in cooling pipes has been made in
the 1990s by Hedlund and Groth [8,9], who have shown the feasibility of such technique in thick columns. Nonetheless, no further
application of such technique was found in the literature, except

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

for an in situ use of air for localized cooling reported by Ishikawa


et al. [12].
Even though thermo-mechanical simulation approaches have
been applied to concrete structures for decades [1317], they involve several complexities and problematic issues particularly in
view of the assessment of material properties and model parameters (thermal and mechanical), which frequently demand specic
laboratory characterization: heat of hydration, thermal boundary
coefcients, creep of concrete, evolution of E-modulus and tensile
stress, among others. Even though several scientic works have
been done either on thermal simulation [18], thermo-mechanical
simulation [19] or monitoring the concrete behaviour at early ages
[2023], some combine the numerical simulation with experimental data obtained in laboratory [2426] or with temperature monitoring for partial validation [2731], whereas others go further
and additionally include in situ strain monitoring for validation
[32]. Nonetheless, in the scope of internal cooling of concrete
through the use of embedded pipes, no works were found to adopt
holistic approaches that simultaneously include material characterization, in situ monitoring of temperature/strain and thermomechanical simulation. The works that focus on concrete cooling
with embedded pipes are mostly limited to thermal [33], or thermo-mechanical analyses only [3437], and thermal [3840] or
thermo-mechanical analyses together with partial validation
through in situ monitoring [12,41].
The present paper pertains to a case study of the thermal stresses in the central wall of the entrance organ of a dam spillway.
Such wall is 27.5 m long, with a maximum width of 2.8 m and
height of 15.0 m, with attention being given to the most unfavourable construction phase in which a 2.5 m tall batch had to be made
(total 150 m3 of concrete). Due to the materials and equipment
available at the construction site, it was decided to attempt internal cooling of concrete with air-cooled prestressing ducts placed
longitudinally along the wall.
The present paper regards to the in-depth study of the early
age performance of concrete in the wall, encompassing laboratory thermal and mechanical characterization of concrete, as
well as in situ monitoring of temperatures/strains and the corresponding thermo-mechanical simulation with the nite element
method.
The extensive laboratory characterization included quantication of the heat of hydration evolution (isothermal and semi-adiabatic calorimetry), evaluation of compressive/tensile strength and
E-modulus through cube/cylinder testing, creep testing at several
ages and assessment of mechanical activation energy. In particular
regard to E-modulus testing, a methodology that allows
continuous measurement of concrete E-modulus since casting
(EMM-ARM [42]) was applied. This is a pioneering use of this
methodology for the purpose of supporting stress simulation on
concrete since its very early ages, with important advantages in
regard to previous approaches that tend to extrapolate values of
E-modulus at very early ages.
A relatively complete monitoring program has been carried out
in situ, involving the use of 20 temperature sensors and 7 vibrating
wire strain gauges embedded in concrete. Particular attention was
given to the evaluation of the effectiveness of the cooling system,
with temperatures being measured at several points along the prestressing ducts, and with air velocity measurements taken with
handheld anemometers.
Bearing in mind the information gathered with the laboratory
characterization and in situ monitoring, a thermo-mechanical
simulation was carried out with recourse to a three dimensional
nite element model. Such simulation model included the explicit
modelling of the cooling ducts, as well as the phased construction
of the wall. The simulation model was made with DIANA software
[43].

149

2. Thermo-mechanical model
The thermo-mechanical simulation approach presented here
has strong similarities with that described in a previous work
[30]. Nonetheless, some particularities are distinct, namely: (i) solar radiation is explicitly considered according to a model based on
the incidence angle of the sun beams; (ii) the effect of internal
cooling ducts is taken into account (iii) the evolution of mechanical
properties is simulated according to the equivalent age concept
(instead of the degree of hydration concept). The following subsections pertain to a general description of the modelling strategy
with specic emphasis on topics (i)(iii) mentioned above.
2.1. Thermal model
The calculation of temperature elds in concrete is based on the
heat balance equation, whose solution is made through the nite
element method [44]:

kr  rT Q_ qcT_

where k is the thermal conductivity, qc is the volumetric specic


heat and T is the temperature. Q_ is the volumetric heat generation
rate due to cement hydration, formulated as an Arrhenius type law
[45]:
Ea
Q_ Af aeRT

where A is a rate constant, Ea is the apparent activation energy, a is


the degree of heat development (ratio between the heat Q released
up to time t and the total heat Qnal released upon completion of cement hydration), R = 8.314 J mol1 K1 is the Boltzmanns constant
and f(a) is a normalized function for heat.
Thermal boundary conditions are applied through a prescribed
ux per unit area qT formulated as [46]:

qT hcr T b  T e

where hcr is a mixed convectionradiation boundary transfer coefcient, Tb is the boundary surface temperature and Te is the environmental temperature.
The simulation of thermal inputs associated to solar radiation in
concrete structures can be made with signicant accuracy through
the adoption of models that are readily used in meteorological sciences [31,47]. Such models can take into account the effects of the
spatial relationship between the earth and sun at a given time of
the day/year and thus predict the solar radiation that reaches a certain point on earth at sea level (i.e. low atmosphere). It is further
possible to compute the angle between the sunbeam and any arbitrarily oriented/inclined surface, and evaluate the intake of energy
throughout the day of such surface.
The calculation of the solar energy that reaches earth at sea level, qm, is based on the solar constant, q0, which represents the total radiation energy received from the sun at a distance
corresponding to 1 Astronomical Unit. Even though q0 varies
slightly throughout the year by 7%, it is usually acceptable to consider q0 = 1367 W m2. The estimation of qm can be done through
the following empirical equation [47,48]:
T l

qm q0  e0:99:4sinh

where Tl is the Linke turbidity factor that summarizes the turbidity


of the atmosphere (attenuation of the direct beam solar radiation)
and h is the solar elevation that corresponds to the angle between
the direction of the sunbeam and the idealized horizon. Tl is known
to usually vary between 3 and 7, whereas h can be calculated by
taking into account latitude, date and time. Further geometrical
considerations allow the calculation of the angle between an
incident sunbeam and the vector orthogonal to an arbitrarily

150

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

oriented/inclined surface, termed as i (see detailed description of


models to calculate h and i in [44,47]).
Based on the knowledge of qm and i at a given instant, and considering the absorvity of the material of the target surface (aS), it is
possible to calculate the radiation energy qS that is actually
absorbed:

Bearing in mind that the aim of the thermo-mechanical simulations is to assess the risk of cracking, the post-cracking behaviour is
not considered relevant and it is thus not simulated. Thus, linear
elastic behaviour (with creep) is considered for concrete both in
compression and in tension.

qS aS  qm  cosi

3. The Paradela dam spillway: description and monitoring

Another particularity of the present application in regard to previous works [30] is the use of prestressing ducts acting as cooling
pipes. A formulation is thus necessary to describe the added internal heat uxes that are caused by the presence of an embedded
cooling pipe, which can be expressed in the following energy balance equation, to be applied throughout the length z along the pipe
[49]:

_  cp
m

dT c
hc  P  T c  T w
dz

_ is the mass ow rate of the coolant (air in this case), cp is


where m
the specic heat of the coolant, hc is the boundary transfer coefcient between the coolant and the surrounding concrete, P is the
perimeter of the cooling pipe, Tc is the temperature in the cooling
pipe and Tw is the bulk temperature of concrete around the cooling
_ can be estimated through
pipe. The mass ow rate of the coolant m
the product of the cooling uid density (q) by the uid mean velocity (lm) and by the cross sectional area of the cooling pipe (Ac). The
implementation of Eq. (6) into a nite element software [43] brings
further nonlinearities due to the interaction between the cooling
uid and the surrounding concrete, which results in progressive
heating of the cooling uid along the pipe.
For updating age-dependent properties along time in the
mechanical model, the equivalent age of concrete teq is adopted.
Its formulation is based on an Arrhenius type equation established
for a reference temperature Tref (usually 293.15K) [50]. For a given
instant t, the equivalent age can be calculated as:

t eq

ERa

1  1
Ts T ref


ds

2.2. Mechanical model


The mechanical model is relatively similar to those adopted for
time-dependent mechanical analysis of hardened concrete, except
for some particularities associated to the facts that: (i) it is being
preceded by a thermal analysis (de-coupled), with imposition of
strains that are calculated with basis on the thermal dilation coefcient of concrete (aT) and the previously calculated temperature
eld; (ii) there is a strong evolution of mechanical properties
throughout the analysis, dully taken into account through the
equivalent age concept; (iii) the strong viscoelastic behaviour of
concrete at early ages makes it necessary to use creep formulations
that can provide adequate estimates within such time span. In specic regard to the last point (iii), basic creep of concrete was accounted for through the use of the Double Power Law (DPL),
which has a reasonably good performance on both early age and
long term time spans [51]:

Jt; t 0

1
/
m
n
1 t0 t  t0
E0 t 0 E0 t0

where J(t, t0 ) is the compliance function at time t for a load applied


at instant t0 , E0(t0 ) is the asymptotic elastic modulus, and /1, m and n
are material parameters. Since drying creep is negligible for an
application that only envisages early age behaviour, it was disregarded [52].

3.1. Overview
The Paradela dam, located in the North of Portugal, is a rockll
gravity dam built in the 1950s, with 540 m longitudinal development and maximum height of 112 m from foundation. Due to recent hydraulic problems in one of the dams spillways, it was
necessary to build a new complementary ski-jump spillway on
the right margin of the river [53]. The case study reported in this
paper concerns the cooling measures and assessment of cracking
risk in the construction of the central wall of the spillway entrance.
3.2. Description of the spillway entrance
3.2.1. Geometry and construction phasing
The spillway functions in free surface conditions, and has two
main entrances at the top level of the dam, each with 5.5 m width,
being separated by a hydro-dynamic shaped wall with maximum
width of 2.8 m and 17.4 m height. A three dimensional representation of the entrance region of the dam spillway is shown in Fig. 1a,
whereas its corresponding plan view at approximately mid-height
of the wall is depicted in Fig. 1b. The reinforcement of the middle
wall can be generally characterized by 16//200 mm placed vertically and 12//200 mm placed horizontally near each surface with
a concrete cover of 60 mm.
The construction of the wall was generally performed with
1.2 m tall construction phases, with empirically dened waiting
periods being dened by a target temperature in the core regions
during the cooling period (approximately 27 C, which corresponded to 17 C above average daily temperature during
construction). In order to minimize such waiting periods, an
air-cooling system based on ventilated prestressing ducts placed
horizontally was implemented, allowing lower peak temperatures
and faster return to temperature equilibrium with the surrounding
environment. The main scope of the present paper is the study of a
specic construction phase that corresponds to the zone of embedment of the xation parts of the sluice gates. Such xation parts
were approximately 2.5 m tall, and it was thus desirable to
perform a 2.5 m tall construction phase, labelled as 9th phase in
Fig. 2a. Due to its larger thickness, this construction phase is the
critical one in terms of peak temperatures and cracking risk, being
therefore the object of analysis.
3.2.2. Materials
The wall of the spillway entrance was generally cast using concrete of class C30/37 [54] with the composition labelled as S1-D32
in Table 1. In the 9th construction phase, due to increased complexity of reinforcement near the downstream extremity of the
wall (related to the salient concrete blocks), two slightly different
compositions with higher uidity were used in the vicinity of such
region, as shown in Table 1 (S3-D32 and S3-D16). In spite of this,
the areas of most interest to this study (thickest regions of the wall
and monitored sections) correspond to the upstream region. Therefore, and also taking into account the fact that the compositions
have similarities, all the characterizations and modelling in the
scope of this work pertain to mix S1-D32. Steel reinforcement
was S400C [54], with characteristic yield stress of 400 MPa.

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

151

Fig. 1. Entrance of the dam spillway: (a) three dimensional representation and (b) plan view of the central wall.

Fig. 2. (a) Construction phasing of the central wall (elevation) and (b) overall photo during the construction of the 9th phase.

Table 1
Mix proportions of concrete in the 9th phase of the spillway wall.
Components

C30/37 S1-D32 (kg/m3)

C30/37 S3-D32 (kg/m3)

C30/37 S3-D16 (kg/m3)

Gravel 1432 mm
Gravel 1016 mm
Gravel 48 mm
Sand
CEM I 42.5R
Fly ash
Plasticizer
Water

449
438
306
621
224
96
2.2
170

400
400
316
646
238
102
3.4
180

490
387
763
280
120
4
200

3.2.3. Cooling system


In view of the work reported by Hedlund and Groth [8,9], which
proposed the possibility of using ventilated prestressing ducts for
concrete cooling, and bearing in mind the easy availability of the
corresponding necessary equipment in the construction site of
the wall, it was decided to test the feasibility of this kind of cooling
technique. However, in view of practical limitations posed by contractor/owner, this pilot application of air-cooling system was
slightly different from that of Hedlund and Groth [8,9]. Instead of
placing the tubes vertically along the wall, they were placed horizontally, even though this implied more limited cooling capacity as
the length of tube along freshly cast concrete is longer. In the particular case of the 9th construction phase, a total of 6 prestressing
ducts of 90 mm diameter have been used, with their air intake
being made horizontally at the downstream extremity of the wall,
and the outtake made near the upstream extremity, on the top surface of the casting phase, in order to avoid a direct upstream
downstream potential leakage channel after construction. The

Concrete pour plan

overall path of the ducts is shown in the schemes of Fig. 3, with


the ducts labelled from T1 to T6. For ventilation, a 0.60 m diameter
fan was used, with 1200 m3/h ventilation capacity, that collected
air from the environment and blowed it into the ducts at an internal air speed of approximately 8.6 m/s (measured with anemometer at the outtake of the ducts). The fan had to be placed in the
downstream extremity of the wall due to practical constraints of
the contractor. The cooling system was only started at the age of
14 h after the end of casting operations to avoid introducing potentially undesirable vibrations to the freshly cast concrete before its
structural setting time. The ventilation system was disconnected
8.6 days after casting in view of the similarity of temperature between the walls core and the surrounding environment. After that,
the prestressing ducts were lled with mortar using standard procedures [55].
It is remarked that this cooling system had been previously tested
in the 8th phase of casting, with three prestressing ducts placed at
mid-height. Details on this test can be found elsewhere [56].

152

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

Fig. 3. Cooling system at the 9th construction phase: (a) plan view; (b) longitudinal section and (c) cross-section A0 A.

3.3. Monitoring and material characterization


3.3.1. General remarks
In order to better understand the effectiveness of the cooling
system, its inuence on the cracking risk, and assess the capabilities of the adopted numerical simulation strategy, an extensive
set of actions has been carried out, comprising in situ internal monitoring of temperatures/strains of concrete, in situ validation tests,
as well as laboratory material characterization.
3.3.2. Temperature monitoring
Temperatures inside the 9th construction phase have been
monitored with K-type thermocouples, aiming particularly at
assessing temperature proles in a region near the maximum
width of the wall (i.e. at approximately 5.3 m from the upstream
extremity: section AA0 as identied in Fig. 3a). The placement of
temperature sensors in section AA0 is depicted in the scheme of
Fig. 4a, where thermocouples are identied by the prex TC. Temperatures in the locations labelled as VW in Fig. 4a have also been
monitored with resistive temperature sensors, as these are the
locations of vibrating wire strain gauges, which contain internal
temperature sensors for strain compensation. The internal air temperature of ducts T1T3 has been monitored both in section AA0
and in neighbouring areas, as shown in Fig. 4b. Environmental temperature (dry-bulb) has been assessed with a thermocouple.
Monitoring was carried out since the instant of casting during a
period of 10 days, and the measurement frequency was set to 1
reading per each 30 min. The internally monitored temperatures
in concrete are shown in Fig. 5 for a vertical and a horizontal alignment of sensors that pass through sensor VW3. It is remarked that
the results of this gure and all upcoming gures of the paper

(either concerning experimental results or numerical simulation


results) have their corresponding time axis zeroed in regard to
the instant at which the casting operations of the studied phase
were nished. From Fig. 5 it can be seen that the initial temperature of concrete was 15 C, and the peak temperature was
approximately 42 C in the core regions (VW3 and VW5). Furthermore, it can be observed that the ascending branch of temperature
development is clearly affected at the age of 14 h, when the cooling
system is activated. In specic regard to the vertical prole of temperatures shown in Fig. 5a, the expectable behaviour was captured:
the core region has the highest peak temperatures (VW3, VW5),
whereas a decrease trend is seen towards the top surface. In fact,
sensors TC6 and TC7 exhibit maximum temperatures of 36 C,
while VW6 (near the top surface) has the lowest peak temperature
(circa 27 C). Near the bottom surface of this construction phase,
sensor VW1 highlights the importance of the heat storage effect
caused by the previously cast concrete: in fact, even though the
temperature peak is lower than that of the core regions, it occurs
later and the heat loss rate observed afterwards is lower than in
other regions. It should also be remarked that all sensors are almost in equilibrium with environmental temperature by the age
of 8 days.
In regard to the temperature development in the sensors located along a horizontal alignment, shown in Fig. 5b, it can be noticed that the sensors located in the vicinity of vertical boundaries
exhibit lower temperature variations (VW4 and TC5), with temperature peaks under 35 C. It is interesting to observe that TC5 is
placed nearer the surface (15 cm apart) than the case of VW4
(20 cm apart), and consequently the temperature peak of VW4 is
slightly higher. By looking at the temperature evolution after the
age of 8 days (heat of hydration has been dissipated), it can be seen

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

153

Fig. 4. Temperature and strain measurement devices: (a) location of sensors within section A0 A and (b) location of sensors within the tubes T1, T2 and T3. Units: [m].

Fig. 5. Monitored temperatures (two alignments that encompass sensor VW3): (a)
vertical alignment and (b) horizontal alignment.

that temperatures in VW4 remain higher than those of TC5 due to a


combination of two main reasons: VW4 is located in the vicinity of
a surface oriented to southeast, which is bound to receive more energy through solar radiation than the surface near TC5, which is
oriented to northwest; VW4 is 5 cm deeper than TC5, thus being
slightly nearer the inner core (higher thermal inertia).
The temperature evolution along the air inside duct T3 measured by sensors TC3, TC9 and TC10 (Fig. 4b) is shown in Fig. 6,
where the environmental temperature and the temperature in

VW5 (hottest region in concrete) are also represented for comparative purposes. The information provided by such gure allows the
clear identication of the instant at which air ventilation began
(14 h), as the rate of temperature rise is clearly disrupted inside
the duct. Furthermore, the rising temperature tendency along the
ducts length is identiable, as the temperature is consistently
higher in TC3 in regard to TC10, and in TC9 in regard to TC3. Taking
as example the temperatures recorded at the instant of peak temperature (1.88 days), TC9 measured a temperature of 32.2 C,
whereas TC10 indicated a temperature of 29.4 C. This represents
a shift in temperature of approximately 3 C in 3.5 m length of
duct. At three instants of this study, temperatures were measured
also at the entrance of T3 (x = 0) and at x = 1 m through the use of a
handheld temperature probe (PT1000). By joining such data with
the results of TC3, TC9 and TC10, it was possible to plot a temperature prole along the duct for t = 3.84 d, t = 4.56 d and t = 6.58 d
see Fig. 6b. It can be observed that the temperature at the inlet of
the tube matches the environmental temperature, and that the
heating of the air along the tube is strongly dependent on the combination of environmental temperature and internal temperature
of concrete. In fact, in the most unfavourable situation shown by
Fig. 6b, air was heated from 11 C at the air inlet to 26 C at a
point located 25 m away from the air inlet (t = 4.56 d). This increase of air temperature is bound to reduce its cooling capacity.
However, in spite of such diminishment of cooling capacity, the
temperature of the air in the hotter regions of the duct remained
at least 10 C below that of concrete during the periods at which
temperature in concrete was near its peak (see t = 1 d to t = 3 d in
Fig. 6a), showing that the heat removal potential was not negligible
at all.
The observed diminishment of cooling capacity along the length
of the duct highlights the fact that the adopted conguration for
the tubes does not maximize cooling capacity, which would be
conversely maximized if the length of tube inside concrete had
been minimized. Such goal could have been achieved by providing
a vertical arrangement for the tubes and introducing more individual smaller tubes.
3.3.3. Strain monitoring
Strain measurement was carried out with vibrating wire strain
gauges of metallic casing with 14 cm reference length (TES/5.5/T
Gage Technique). Past laboratory tests and in situ applications
[44,57,58] have shown that this kind of sensor is robust and adequate for strain measurement in concrete at early ages. The strain
gauges were placed at the locations identied in Fig. 4a (VW1
VW6), dully positioned in order to measure strains in the longitudinal direction of the wall. VW7 has distinct intents and shall be

154

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

Fig. 6. Monitored temperatures in duct T3: (a) Global results for TC3, TC9 and TC10 and (b) temperature prole along the duct at three selected instants. Units: [m].

specically addressed later. Measurements were taken with the


same datalogger and at the same sampling rate as adopted for
the temperature sensors.
One important issue to tackle is the zeroing of the measured
strains. In fact, before concrete sets and has enough stiffness to
drive the sensor into the same deformation state, the measurements taken by the sensor do not have any relevant physical meaning. It is thus necessary to assess the instant of solidarization (i.e.
the full bond) between concrete and the sensor. In a previous work
[57], the zeroing operation has been made by assessing the instant
at which two sensors with different casing (plastic and metallic),
placed under the same conditions inside concrete, started yielding
the same results. This means that both sensors are solidarized (as
the plastic sensor is bound to solidarize earlier due to its smaller
stiffness). Since the plastic cased sensor was not available, an alternative methodology for zeroing the data was implemented. By
interpretation of the ndings reported in [57] it can be considered
that the solidarization instant coincides with a progressive change
in the derivate of strain variation detected by the sensor, that can
be obtained by geometrical intersection of tangents of measured
strains, as shown in Fig. 7. It was decided to use such zeroing criterion in the scope of this research work. As a result of the application of such rule, the solidarization instant of each sensor VW1
VW5 was, respectively: 0.17, 0.19, 0.19, 0.20 and 0.17 days (due
to malfunctioning of the VW6, the strain results from this sensor
are not available). The solidarization instants seem coherent, since
they have a trend to increase with the distance of the sensor from
the bottom surface of the casting block, due to the natural delay in
its involvement by concrete during the casting process.
The measured strains in sensors VW1VW5 are shown in Fig. 8.
Even though the strain output is dependent on several factors that
interact with each other (thermal deformation, restraint, creep), it
is possible to nd a set of common points and reasoning. Overall,
all deformations are strongly commanded by the temperature variation, following the same kinetics. Sensors VW2, VW3 and VW5,
which are located in regions near the core of the walls cross section and had similar temperature development histories, also have
similar strain developments. This is bound to be caused by similar
thermal deformations and restraints for these locations of measurement. The smallest deformations are recorded in VW1, which
is located in the bottom of the casting phase (i.e. near the existing
concrete) and thus having less temperature rise (thus less expansion), while being more restrained by the existing concrete below
at lower temperatures. Finally VW4, which is near the surface and

Fig. 7. Methodology adopted for determination of the instant of solidarization of


VW sensor to concrete.

Fig. 8. Strains measured by sensors VW1VW5.

thus has lower temperature rises (maximum temperature of


33 C), also has a smaller deformation variation when compared
to VW2, VW3 and VW5.

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

In order to assess free deformations of the concrete used in the


construction (associated to unrestrained autogenous shrinkage and
thermal deformations), strain was measured in a concrete cylinder
placed in specially devised conditions. The concrete cylinder
(150 mm diameter and 300 mm tall), cast simultaneously with
the studied construction phase and by using the same concrete,
was cast into a special mould internally coated with a soft membrane (with lids also coated with such material). A strain gauge
was placed inside the mould to measure longitudinal strains
see photo of the open mould in Fig. 9a. After casting concrete into
such mould, it was placed horizontally inside the studied construction phase (during its casting procedures) at the location that is
identied as VW7 in Fig. 4a. This kind of procedure, here termed
as use of a no-stress specimen, has been reported in Choi et al.
[59], and it allows to measure free deformations of concrete, which
can in turn be used to assess the thermal dilation coefcient, TDC
(provided that the temperature inside the concrete cylinder is relatively uniform, and autogeneous shrinkage deformations are
known). Unfortunately, due to undetermined causes, the output
of the sensor could not be read during the rst 0.8 days, and thus
the reported data only starts at such age, as shown in Fig. 9b.
3.3.4. Heat generation and activation energy
In an extensive experimental program for the characterization
of the cements marketed in Portugal, Azenha [44] has reported a
library of heat generation obtained through isothermal calorimetry
under several temperatures for plain cement pastes with w/c = 0.5.
The cement used in the construction concerned in this paper was
also characterized (same supplier and manufacturing plant), and
the resulting information for calorimetry tests under 20 C, 30 C,
40 C and 50 C is shown in Fig. 10. A reasonable estimate of the
heat generated by concrete can be obtained by multiplying the
heat generation reported in Fig. 10 by the volumetric content of cement, which is of 224 kg/m3. By using the speed method algorithm
[44,60], the necessary data for the numerical simulation of heat
generation according to Eq. (2) was obtained: Ea = 37.31 kJ/mol,
A = 4.989  109 W/m3, Qpot = 8.295  107 J/m3, and function f(a)
characterized by the following set of data [a; f(a)] = [0.00; 0.00],
[0.05; 0.58], [0.10; 0.85], [0.15; 0.98], [0.20; 1.00], [0.30; 0.94],
[0.40; 0.69], [0.50; 0.41], [0.60; 0.22], [0.70; 0.13], [0.80; 0.07],
[0.90; 0.02], [1.00; 0.00]. Even though this data pertains to CEM I
42.5R of the same company that supplied the cement to this construction site, there may be deviations caused by inevitable variations in the characteristics of the cement. Also, the extrapolation
procedure mentioned above did not take into account the presence
of y ash in the mix (96 kg/m3), which may have non-negligible

155

Fig. 10. Heat generation rate of a cement paste with CEM I 42.5R and w/c = 0.5.

effects on the heat generation potential and hydration kinetics.


Therefore, in order to assess the potential importance of such deviations, a semi-adiabatic test was conducted in situ (simultaneously
with the casting operations) in a 30 cm edge concrete cube, duly
isolated by 2.1 cm thick plywood and 12 cm of polystyrene boards.
The results of such semi-adiabatic calorimetry test shall be addressed in Section 4, upon the simulation of its temperature development through the nite element method.
3.3.5. Complementary laboratory characterization
Compressive strength evolution was assessed with concrete
cubes (150 mm edge) at the ages of 1, 3, 7 and 29 days, whereas
tensile strength was measured with splitting tests on cylinders
(150 mm diameter and 300 mm tall) and at the same ages. It is remarked that testing at the reference age of 28 days was not possible due to laboratorial constraints. The evolution of both tensile
and compressive strength for concrete cured at 20 C (saturated
conditions) is shown in Fig. 11a (average results of three specimens
at each age). In order to assess the activation energy suitable for
compressive strength maturity estimations, a set of concrete cubes
was cured at 40 C with the compressive strength measured at the
same ages. The corresponding results are also shown in Fig. 11a. By
applying the equivalent age concept [61], together with the superposition method [60], it was possible to asses that the activation
energy based on mechanical testing has the value of 37 kJ/mol,
which is rather consistent with the activation energy obtained
through isothermal calorimetry for the same cement (yet without

Fig. 9. In situ determination of TDC: (a) overview of installed no-stress specimen and (b) evolution of strain and temperature evolutions in VW7.

156

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

Fig. 11. (a) Evolution of compressive and tensile strength of concrete. (b) Basic creep of concrete: specic creep for loading at the ages of 1.1, 3.3 and 7.3 days.

y ash), 37.31 kJ/mol [44]. Such coincidence in activation energy


for thermal and mechanical phenomena had already been reported
by Ulm and Coussy [62].
Basic creep was assessed in creep rigs on prismatic specimens
(sealed) with dimensions 15 cm  15 cm  60 cm, loaded at 30
40% of the concrete compressive strength and internally monitored
with vibrating wire strain gages. Such creep tests were conducted
at the ages of 1.1, 3.3 and 7.3 days, and the corresponding specic
creep curves are shown in Fig. 11b.
The experimental program also included a single specimen for
measurement of autogenous shrinkage. Such specimen was a
150 mm diameter and 300 mm long cylinder, internally instrumented with a vibrating wire strain gage, which was kept in its
formwork during the experiment and sealed with a plastic lm
on the top surface. Unfortunately, two factors contributed to render the results of this specimen unusable for this research: on
one hand, the monitoring only could be started at the age of 2 days
in the laboratory due to unavailability of datalogging system; on
the other hand, the measurements of autogenous taken since
t = 2 days were disturbed by an inefcient sealing, which promoted
undesired drying of the specimen. This was not considered a critical problem in view of the low values of autogenous shrinkage that
are usually expectable in concretes of low cement content and high
w/c ratio [6365].

testing setup and procedure applied for the concrete of this spillway application can be found in [56,66,67], as it corresponds to
an improved version of the originally devised test (a steel mould
is used). The collected results with EMM-ARM and cyclic compression tests on cylinders are shown in Fig. 12, where the feasibility of
EMM-ARM is conrmed in view of the resemblance of results. Also,
the richness of information that can be obtained through EMMARM represents an added value for the numerical simulation.

4.1.1. Geometry of the model and nite element mesh


A cross-sectional scheme of the model for simulation is shown
in Fig. 13a, where the construction stages considered in the analysis can be observed. The rst stage of the model encompasses all
concrete until the 7th phase of concreting (inclusive), considered
as hardened concrete, together with the 8th phase of concrete evaluated as freshly cast concrete. The second and third stages correspond to the 9th and 10th phases of concreting respectively. This
strategy diminishes the computational cost of the model without

3.3.6. Continuous monitoring of concrete stiffness


The evolution of elasticity modulus along time was measured
through compressive cyclic testing in concrete cylinders
(150 mm diameter and 300 m tall) at the ages of 1, 3, 7, 15 and
29 days. Concomitantly, E-modulus of concrete was continuously
assessed through a methodology termed as EMM-ARM (Elasticity
Modulus Measurement through Ambient Response Method). This
methodology has been developed by Azenha et al. [42] and consists
in a variant to classic resonant frequencies that allows the quantication of E-modulus continuously since the instant of casting of
the specimen inside the testing mould. The basic principle of
EMM-ARM is the following: the specimen is cast inside the testing
mould, which is in turn placed in simply supported conditions and
continuously subject to modal identication (using accelerometers) without any explicit excitation of the beam, as ambient vibration sufces. As concrete hardens, the rst resonant frequency of
the composite beam evolves, and the stiffness of concrete can be
inferred by applying the equations of motion. Details about the

Fig. 12. E-modulus of concrete assessed by compressive cyclic testing and EMMARM.

4. Numerical modelling
4.1. Geometry, mesh, materials, initial/boundary conditions and time
integration

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

157

Fig. 13. (a) Scheme of the simulation model and phasing; (b) section A0 A of the model and (c) nite element mesh.

relevant effect on the accuracy of results. For similar reasons, the


underlying subgrade is not modelled, as it is far from the construction phases of interest. Due to the geometrical symmetry of the
wall, a longitudinal plane of symmetry is considered, identiable
by a ZX plane in Fig. 13.
The simulation was made with a 3D nite element model comprising rectangular brick FE of 8 nodes (2  2  2 integration
scheme) for concrete in the thermal model, and coincident 20
nodes brick FE (3  3  3 integration scheme) in the mechanical
analysis. Convective boundaries were modelled with 4 node planar
elements (2  2 integration scheme), and the cooling ducts were
considered with linear elements of 2 nodes (2 point integration
scheme) [43]. The schematic representation of geometry, casting
phases and cooling duct location for section A0 A is shown in
Fig. 13b. The reader is reminded that phase 8 also had cooling
ducts, according to the description of Section 3.2.3. The longitudinal layout of all the ducts was considered straight on an horizontal
plane, in correspondence to simplications in the vicinity of the
extremities of the wall.
It should be remarked that due to the phased analysis, the mesh
evolved along time with some elements/boundaries being activated/de-activated (e.g. the convective top boundary of a given
phase is de-activated upon the beginning of the next casting
phase). The mesh adopted for this simulation is shown in
Fig. 13c, with a total of 18,738 elements and 66,037 nodes. As consequence of the symmetry simplication, it was considered that
the perimeter of the tube elements located in the symmetry plane

of the model was equal to half the actual perimeter of the tubes. As
solar radiation does not represent a symmetrical energy input to
the structure, the symmetry simplication adopted is not truly valid. However, as solar radiation has most of its effect near the surface, it was decided to keep the symmetry simplication by
considering the southeast half of the wall, which is most subject
to solar radiation effects.
4.1.2. Materials (thermal and mechanical properties)
The thermal conductivity and specic heat of concrete were
estimated with basis on the pondered average of the corresponding
thermal properties of the constituent materials of the mix [44,68].
The adopted values for k and qc for concrete were, respectively
2.40 W/m K and 2.4  106 J/m3 K. Even though it is known that
these thermal properties suffer variations during early ages [69
75], the adopted modelling approach considers them constant in
view of the conclusions of the parametric analyses reported by
Azenha [44], where a relatively small impact of considering evolving k and qc was found on computed temperatures in hardening
concrete.
In regard to the data for heat of hydration, the parameters mentioned in Section 3.3.4 were used. The adequacy of these parameters was evaluated through the semi-adiabatic calorimeter
described in the same section, whose behaviour was simulated
through a FE simulation model that explicitly considered the extruded polystyrene (XPS) and wood walls of the calorimeter. The
material modelling parameters for concrete coincide with those

158

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

herein described, whereas additional information and mesh representation are shown in Fig. 14a. The results of the simulation of the
calorimeter were quite coherent with those collected experimentally, as seen in Fig. 14b, leading to the conrmation that the strategy described in Section 3.3.4 to determine the heat generation
and activation energy was adequate for dening the heat of hydration modelling parameters adopted in the simulations.
The thermal dilation coefcient (TDC) of concrete was assessed
with basis on the no-stress specimen described in Section 3.3.3.
However, in order to obtain the thermal deformation and calculate
the TDC, it was necessary to subtract the autogenous shrinkage
deformation from the total deformation. As data on autogenous
shrinkage was not available, an estimate of the autogenous shrinkage evolution based on Eurocode 2 [54] was used. Another issue to
take into account is the fact that the TDC of concrete is not constant
during the rst hours of age. In fact, several authors have dealt
with this subject, and it generally agreed that the initial TDC tends
to be larger than that of hardened concrete, and tends to decrease
sharply within the rst 12 to 24 h of age, reaching then the plateau
level corresponding to hardened concrete [69,73,76]. The nostress specimen cast in the scope of this research cannot be used
to estimate TDC at concrete early ages, due to the absence of data
in the rst 8 h reported in Fig. 9b. Nonetheless, as the peak temperature occurred later than 24 h age, and full data is available for
temperatures and strains occurred in such period, calculations
could be made under the assumption that TDC was already at its
plateau value. The TDC was estimated between instants t = 2.0 d
(peak temperature) and t = 8.0 d (local minimum) as shown in
Fig. 9b, and the autogenous shrinkage strain variation in such period was estimated to be of 8.45 le (considering fcm = 42.3 MPa in
Eurocode 2 [54]). The estimated constant TDC to be used in the
numerical simulation was 11.07 le/C. Nonetheless, since it is
known that TDC varies during the rst 24 h of age, an alternative
formulation for the evaluation of the TDC was considered, based
on experimental evidence reported by Laplante and Boulay [77].
Therefore, this alternative formulation considers that during the
rst 16 h the TDC varies according to TDC(t) = 0.16t2  4.88t
+ 48.93 (t in hours), and remains constant after such age.
The E-modulus evolution of concrete for the simulation model
was directly extracted from EMM-ARM data reported in Fig. 12,
whereas creep modelling was made through the adjustment of
DPL parameters to the creep data experiments. The best-t creep
parameters and their adjustment to the experimental data are
shown in Fig. 11b. Prestressing ducts were modelled with

consideration of their inner perimeter of 283 mm (90 mm inner


diameter). Steel reinforcement was disregarded in temperature
calculations due to its low interference in temperature development [44,52]. Regarding mechanical eld simulations, steel was
not considered because post-cracking behaviour was not sought.
In the non-cracked stage, the similitude in TDC of steel in regard
to that of hardened concrete strongly minimizes the restraint to
concrete thermal deformation, thus rendering the effect of reinforcement negligible for the computation of thermal stresses at
early ages [52].
4.1.3. Initial/boundary conditions and construction phasing
In what concerns the boundary conditions in the thermal problem, it was assumed that an average wind speed of 2.5 m/s occurred (conrmed during 3 days with an anemometer) and the
resulting convection/radiation coefcient for concrete surfaces in
contact with the environment was estimated to be 15.25 W/m2 K
in accordance to the predictive formula of Branco et al. [78]. For
the particular case of formworks surrounding concrete, an equivalent boundary convection coefcient was adopted according to an
electrical analogy [46]. Bearing in mind that the formworks were
made of wood (kwood = 0.175 W/m2 K) and their thickness was
18 mm, the resulting equivalent boundary coefcient was 6 W/
m2 K. Formwork was applied to the vertical surfaces of each casting
stage during the rst 7 days of age, and removed afterwards. All
convective boundaries were subjected throughout the analysis to
the environmental temperature that was monitored in situ see
Fig. 5. The symmetry plane of the model was considered as an adiabatic boundary.
In what regards to solar radiation intake, the several surface
directions of the model were taken into account, and the absorbed
radiation was calculated according to the model described in Section 2.1. Absorvity of concrete was considered as 0.6 [79], the latitude was 41.77N, and the casting date was 28/03/2011. The Linke
turbidity factor was considered as 2.5, and its feasibility was conrmed by comparing computed solar radiation on horizontal surfaces with solar radiation data from a nearby weather station (Rio
Torto Station) in conditions of clear skies. The effect of cloudiness
was taken into account by normalizing the predictions of the
adopted solar radiation model according to information obtained
from the piranometer of the neighbouring weather station. The
diminishment of solar absorption caused by shadows cast by neighbouring objects was considered negligible throughout the entire
day, as the wall was one of the tallest elements in the landscape.

Fig. 14. (a) FE model for the semi-adiabatic calorimeter (geometry and general data/units in millimetres). (b) Calculated and recorded temperature in the geometrical centre
of the calorimeter.

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

Bearing in mind the construction phases taken into account in


this calculation model (identied in Fig. 13a), the initial temperatures were considered as follows. For the existing concrete at the
beginning of analysis, it was assumed that concrete was already
in thermal equilibrium with the environment, and so, the average
daily temperature of the preceding week (14.5 C) was considered
for the existing concrete. For the subsequent stages of construction,
the initial temperature of concrete was obtained from in situ monitoring (average value), which was also of approximately 14.5 C.
In regard to the cooling ducts, due to a limitation of the adopted
software, the inlet temperature had to be considered constant,

159

equal to the average environmental temperature during the time


in which the tubes were active. The following temperatures were
considered for each phase: 8th: 7.13 C and 9th: 13.91 C. The
internal convection coefcient in the ducts was obtained with basis on their internal air speed of 8.6 m/s, which according to the
studies of Hedlund and Groth [8] should correspond to a convection coefcient of 30.0 W/m2 K. The cooling duct elements were
activated at each construction phase when the surrounding concrete had age of 14 h.
Taking into account the directions of the axes of the coordinate
system presented in Fig. 13a, the mechanical boundary conditions

Fig. 15. (a) Measured and simulated temperatures for VW1, VW2, VW4 and VW5. (b) Temperature distribution within the thicker section of the wall for the instants that the
maximum temperatures are attained. (c) Calculated temperatures in the points where the highest temperatures are attained in the thicker section of the wall.

160

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

consisted in placing Z direction supports on the bottom surface of


the wall and Y supports in all the elements of the symmetry plane.
The time step strategy adopted for the model consisted in considering the initial time coincident with the casting instant of 8th
phase. Casting of the 9th and 10th phases were considered at the
relative instants t = 24 days and t = 41 days in accordance to the
actual construction. All analyses were conducted with a constant
time step of 1 h duration, even though some localized adjustments were necessary in view of construction phasing and ventilation activation. Nonetheless, all adjustments were carefully
made to assure that the duration of all time steps remained under
1 h.
4.2. Results and discussion
The presentation and discussion of results is centred in the 9th
construction phase, with particular emphasis for comparisons between monitored and simulated temperatures/strains. The temperature simulations in concrete were quite coherent with the
monitored ones, as it can be conrmed for a set of representative
locations (VW1, VW2, VW4 and VW5), whose results are shown
in Fig. 15. In fact the largest deviations regarding the monitored
temperatures during the entire calculation always remain under
4 C, thus providing conrmation of the feasibility of the modelling
strategy, particularly in regard to the cooling capacity of the ventilated ducts. Based on the condence gained on the thermal simulation model, a further numerical simulation was made, in which
the effect of the cooling ducts was disregarded. The corresponding
results for the hottest region of both models are shown in Fig. 15b.
It can be conrmed that the inclusion of the duct had a twofold effect: not only was the peak temperature diminished by 5 C with
benets for cracking safety, but also the return of the internal tem-

perature to thermal equilibrium with the environment was accelerated, with advantages for construction phasing.
The fact that the calculated temperatures matched well the
monitored ones is a solid starting point for the analysis of results
of the mechanical simulation, as any detected deviations are
bound to be solely attributed to issues in the mechanical simulation itself. The calculated and measured strains for the same set
of sensors that has just been discussed for temperature development are shown in Fig. 16. The experimentally measured strains
in this gure are represented by their value according to the
zeroing procedure mentioned in Section 3.3.3 (Fig. 7), but also
with a lower and upper bound related to possible uncertainties
in the instant for zeroing of the sensors output of 2 h. It can
be seen that all the computed strains with consideration of constant TDC underestimate the peak strain at 1.96 day, but the
post-peak kinetics seems to have been well captured. As the constant TDC assumption may lead to underestimations of early
strain development [57], a further calculation was made using
a plausible TDC evolution during the rst 24 h, as discussed in
4.1.2. The corresponding simulation results are shown in
Fig. 16, where a better overall t is seen between experimental
and calculation data (particularly for core regions). Even though
the variable TDC was not based on experimental evidence obtained in the scope of this research, it is feasible to assume that
a signicant part of the strain deviations regarding experimental
results can be explained by the variable TDC at early ages. It has
to be kept into consideration that another possible source of
deviation of results may be related to the instant at which measured strains were zeroed, which can be debatable. Nonetheless,
the adequate prediction of strains that was attained is a good
indication of the feasibility of the computed stresses which are
to be analyzed.

Fig. 16. Measured and simulated strains for VW1, VW2, VW4 and VW5.

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

Fig. 17. Principal tensile stresses at the point corresponding to the maximum
tensile stress in the wall (x = 8.5 m, y = 0.36 m, z = 11.7 m) for three simulation
models considered.

The discussion of cracking risk is now addressed by comparing


the computed principal tensile stress at the most unfavourable part
of the model (located in the core region: x = 8.5 m, y = 0.36 m,
z = 11.7 m), as shown in Fig. 17. This gure contains the equivalent
age-adjusted measured evolution of the tensile strength, and the
calculation results for the cases of constant and variable TDC, as well
as the case of inexistent cooling ducts and constant TDC. The rst
comment that can be made from Fig. 17, is that the consideration
of constant or variable TDC had very marginal effect on the results.
The reason for the very small difference is bound to be related with
the very low stress level that is induced during the rst 24 h, as the
E-modulus of concrete is still very small and creep/relaxation is very
high. Regardless of the comparison between these two models, it
can be observed from Fig. 17 that the ratio between the tensile stress
and the tensile strength of concrete at the most unfavourable instant is of approximately 0.9, which corresponds to a signicant
cracking risk. Nonetheless, even though the cracking risk was higher
than the desirable one, the structure did not present thermal cracks
neither later through-cracks (evaluations made until 2 years after
casting). It should be remarked that this point of stress analysis
was the most unfavourable one within the structure, and signicantly lower cracking risks were calculated for distinct regions,
resulting in a global scenario of much more cracking safety (compared to a single-point analysis).
Interestingly, a simulation of the same construction situation
without consideration of cooling ducts would yield to higher
cracking risk at the same location (as seen in Fig. 17), with the ratio
between the tensile stress and the tensile strength of concrete
reaching 1.2. Therefore, if the calculations made here are considered trustworthy, the use of the cooling ducts may have been the
differentiating factor that avoided a cracking scenario in this
concrete lift.

161

monitoring for temperatures and strains. A 3D thermo-mechanical


simulation of the construction phasing was shown, with input data
duly based in the laboratory characterization program.
In regard to previous works reported in the literature, the work
presented in this paper has its main original contributions in the
following elds: (i) the EMM-ARM methodology for continuous
monitoring of concrete E-modulus since casting was applied for
the rst time as a characterization tool for stress simulation in concrete at early ages, thus enhancing the quality of input data; (ii)
this is the rst reported application of horizontally placed air-cooling pipes, with its efciency being assessed and numerically simulated; (iii) the work reported here is relatively unprecedented in
view of its holistic approach, with the authors being involved in
all tasks of laboratory characterization (allowing sustainable estimates of material properties and modelling strategy for numerical
simulations), eld monitoring and numerical simulation with thermo-mechanical analysis.
It is nonetheless acknowledged that a signicant research effort
is still necessary in regard to the creep monitoring and modelling
at early ages, both in view of the effects of early hydration and in
view of temperature effects on creep. Even though in-depth analyses of these particular issues of creep have not been included in
this paper, the authors consider them to be of critical importance
for adequate stress simulation in hardening concrete, thus
demanding further research works.
The numerical simulation results were compared to those collected by the in situ monitoring and good coherences were observed
both in terms of temperature and strain, providing good prospects in
regard to the simulation capabilities of the models and the soundness of the experimentally obtained data. The monitoring/simulation results allowed concluding that the effectiveness of the air
cooling system with horizontally placed pipes is limited in view of
the signicant heating that air suffers along the rst meters of tube,
thus diminishing its capacity of cooling further regions of concrete.
Also, the duct efciency ends up being quite dependent of the environmental temperature, which cannot be easily anticipated during
planning. Nonetheless, in spite of the acknowledged limitations of
air cooling, it may prove quite feasible in relatively cool climates
and small lengths of embedment (e.g.: 10 m or less).
Furthermore, the risk of cracking on the studied construction
phase has resulted acceptable in most of its regions, even though
a non-negligible risk of internal cracking was observed in some regions. The fact that no surface or through cracking was observable
in the construction corroborates the cracking risk evaluation. It
was also concluded that the same construction phasing without
the use of the cooling ducts had a signicantly higher cracking
probability, thus conrming the usefulness of the cooling system.
Finally, it is also worth remarking that a parametric analysis
regarding the possibility of considering variable thermal dilation
coefcient of concrete along hydration has been carried out. The
outcome of such parametric analysis seems to point out a relatively
low impact of admitting the thermal dilation coefcient as constant along hydration on the corresponding computed stresses,
thus validating such simplication in this case study.

5. Conclusions

Acknowledgements

A case study regarding the assessment of the cracking risk of a


thick wall in the entrance of a dam spillway, internally cooled with
air-lled prestressing ducts, was presented in this paper. The use of
ventilated prestressing ducts is considered more straightforward
than water ducts due to the often easy availability of ducts and
fans in construction works associated to massive concrete
structures.
The case study has involved a comprehensive experimental
part, including laboratory characterization of materials and in situ

Funding provided by the Portuguese Foundation for Science and


Technology to the Research Unit ISISE, to the second author
through the PhD Grant SFRH/BD/64415/2009, and to the research
projects PTDC/ECM/099250/2008 and QREN Number 5387, LEGOUSE, is gratefully acknowledged. The kind assistance of the contractor (Teixeira Duarte S.A.) and the owner (EDP Eletricidade
de Portugal) are also deeply appreciated. The contribution of ngelo Costa to the experimental work here reported is also gratefully
acknowledged.

162

M. Azenha et al. / Engineering Structures 6263 (2014) 148163

References
[1] Bernander S. Practical measures to avoiding early age thermal cracking in
concrete structures. In: Springenschmid R, editor. Prevention of thermal
cracking in concrete at early ages: state-of-the-art report prepared by RILEM
Technical Committee 119. RILEM report 15. E&FN Spon; 1998. p. 255314.
[2] Springenschmid R. In: Thermal cracking in concrete at early ages: RILEM
proceedings. RILEM report 25. Routledge: Taylor & Francis; 1995.
[3] Du C. Dam construction concrete temperature control using y ash. Concr Int
1996;18:346.
[4] Utsi S, Jonasson J-E. Estimation of the risk for early thermal cracking for SCC
containing y ash. Mater Struct 2012;45:15369.
[5] Kurita M, Goto S, Minegishi K, Negami Y, Kuwahara T. Precooling concrete
using frozen sand. Concr Int 1990;12:605.
[6] Takeuchi H, Tsuji Y, Nanni A. Concrete precooling method by means of dry ice.
Concr Int 1993;15:526.
[7] St John J. Construction of the Hoover Dam Bypass. Concr Int 2011;33:305.
[8] Hedlund H, Groth P. Air cooling of concrete by means of embedded cooling
pipes. Part I: laboratory tests of heat transfer coefcients. Mater Struct
1998;31:32934.
[9] Groth P, Hedlund H. Air cooling of concrete by means of embedded cooling
pipes. Part II: application in design. Mater Struct 1998;31:38792.
[10] Roush K, OLeary. Cooling concrete with embedded pipes. Concr Int
2005;27:302.
[11] Maggenti R. From passive to active thermal control. Concr Int 2007;29:2430.
[12] Ishikawa S, Matsukawa K, Nakanishi S, Kawai H. Air pipe cooling system. Concr
Int 2007;29:459.
[13] Nobuhiro M, Kazuo U. Nonlinear thermal stress analysis of a massive concrete
structure. Comput Struct 1987;26:28796.
[14] Emborg M. Thermal stresses in concrete structures at early age. Doctoral
thesis. Lule University of Technology; 1989.
[15] Ishikawa M. Thermal stress analysis of a concrete dam. Comput Struct
1991;40:34752.
[16] de Borst R, van den Boogaard AH. Finite-element modeling of deformation and
cracking in early-age concrete. J Eng Mech 1994;120:251934.
[17] Torrenti JM, De Larrard F, Guerrier F, Acker F, Grenier F. Numerical simulation
of temperatures and stresses in concrete at early ages: the French experience.
In: Springenschmid R, editor. Thermal cracking in concrete at early ages.
RILEM proceedings 25, Paris; 1994. p. 2818.
[18] Fairbairn EMR, Ferreira IA, Cordeiro GC, Silvoso MM, Filho RDT, Ribeiro FLB.
Numerical simulation of dam construction using low-CO2-emission concrete.
Mater Struct/Mater Constr 2010;43:106174.
[19] De Schutter G, Vuylsteke M. Minimisation of early age thermal cracking in a Jshaped non-reinforced massive concrete quay wall. Eng Struct 2004;26:8018.
[20] Dolmatov AP, Neidlin SZ. Temperature control of the massive concrete of the
Krasnoyarsk hydroelectric station dam. Hydrotech Constr 1968;2:95660.
[21] Blinov IF, Shaikin YP, Galperin IR. Field studies of the temperature regime and
stress state of the concrete in arch dams subjected to various methods of
cooling. Hydrotech Constr 1980;14:1923.
[22] Torrenti J-M, Buffo-Lacarrire L, Barr F. CEOS.FR experiments for crack control
of concrete at early age. In: RILEM-JCI international workshop on crack control
of mass concrete and related issues concerning early-age of concrete
structures (ConCrack 3), Paris; 2012. p. 310.
[23] Cussigh F. Experience in limiting early age concrete temperature for DEF
prevention. In: RILEM-JCI international workshop on crack control of mass
concrete and related issues concerning early-age of concrete structures
(ConCrack 3), Paris; 2012. p. 7988.
[24] Benboudjema F, Torrenti JM. Early-age behaviour of concrete nuclear
containments. Nucl Eng Des 2008;238:2495506.
[25] Craeye B, De Schutter G, Van Humbeeck H, Van Cotthem A. Early age behaviour
of concrete supercontainers for radioactive waste disposal. Nucl Eng Des
2009;239:2335.
[26] Briffaut M, Benboudjema F, Torrenti JM, Nahas G. Numerical analysis of the
thermal active restrained shrinkage ring test to study the early age behavior of
massive concrete structures. Eng Struct 2011;33:1390401.
[27] Waller V, dAloa L, Cussigh F, Lecrux S. Using the maturity method in concrete
cracking control at early ages. Cem Concr Compos 2004;26:58999.
[28] Xiang Y, Zhang Z, He S, Dong G. Thermalmechanical analysis of a newly cast
concrete wall of a subway structure. Tunn Undergr Space Technol
2005;20:44251.
[29] Faria R, Azenha M, Figueiras JA. Modelling of concrete at early ages:
application to an externally restrained slab. Cem Concr Compos
2006;28(6):57285.
[30] Azenha M, Faria R. Temperatures and stresses due to cement hydration on the
R/C foundation of a wind tower a case study. Eng Struct 2008;30:2392400.
[31] Boutillon L, Linger L, Kolani B, Meyer E. Effects of sun irradiation on the
temperature and early age stress distribution in external concrete structure.
In: Toutlemonde F, Torrenti J-M, editors. RILEM-JCI international workshop on
crack control of mass concrete and related issues concerning early-age of
concrete structures. ConCrack 3, Paris; 2012. p. 18192.
[32] Zreiki J, Bouchelaghem F, Chaouche M. Early-age behaviour of concrete in
massive structures, experimentation and modelling. Nucl Eng Des
2010;240:264354.

[33] Zhu Y-M, Liu Y-Z, Xiao Z-Q, He J-R, Lin Z-H, Ma Y-F. Analysis of pipe-cooling
system in mass concrete. College of Water Conservancy & Hydropower Eng.,
Hohai Univ., Nanjing, China; 2004.
[34] James RJ, Dollar DA. Thermal engineering for the construction of large concrete
arch dams. In: The 6th ASMEJSME thermal engineering joint conference,
Hapena Beach, Hawaii; 2003.
[35] Xie H, Chen Y. Inuence of the different pipe cooling scheme on temperature
distribution in RCC arch dams. Commun Numer Methods Eng
2005;21:76978.
[36] Myers TG, Fowkes ND, Ballim Y. Modeling the cooling of concrete by piped
water. ASCE J Eng Mech 2009:137583.
[37] Yamamoto T, Ohtomo T. Practices for crack control of concrete in Japan. In:
Toutlemonde F, Torrenti JM, editors. RILEM-JCI international workshop on
crack control of mass concrete and related issues concerning early-age of
concrete structures (ConCrack 3). ParisRILEM Publications; 2012. p. 193200.
[38] Tanabe T-a, Yamakawa H, Watanabe A. Determination of convection
coefcient at cooling pipe surface and analysis of cooling effect. Proc JSCE
1984;34:1719.
[39] Yang J-K, Lee Y, Kim J-K. Heat transfer coefcient in ow convection of pipecooling system in massive concrete. J Adv Concr Technol 2011;9(1):10314.
[40] Yang J, Hu Y, Zuo Z, Jin F, Li Q. Thermal analysis of mass concrete embedded
with double-layer staggered heterogeneous cooling water pipes. Appl Therm
Eng 2012;35:14556.
[41] Kim JK, Kim KH, Yang JK. Thermal analysis of hydration heat in concrete
structures with pipe-cooling system. Comput Struct 2001;79:16371.
[42] Azenha M, Magalhes F, Faria R, Cunha . Measurement of concrete E-modulus
evolution since casting: a novel method based on ambient vibration. Cem
Concr Res 2010;40:1096105.
[43] TNO-DIANA-BV. Diana users manual release 9.2, Delft, The Netherlands;
2007.
[44] Azenha M. Numerical simulation of the structural behaviour of concrete since
its early ages. PhD thesis. University of Porto, Porto, Portugal; 2009.
[45] Reinhardt H, Blaauwendraad J, Jongedijk J. Temperature development in
concrete structures taking account of state dependent properties. In: Int conf
concrete at early ages, Paris, France; 1982.
[46] Faria R, Azenha M, Figueiras JA. Modelling of concrete at early ages:
application to an externally restrained slab. Cem Concr Compos
2006;28:57285.
[47] Breugel K, Koenders EAB. IPACS report BE96-3843/2001:31-1. Solar radiation:
effect on solar radiation on the risk of cracking in young concrete; 2001.
[48] Kasten F. A simple parameterization of the pyrheliometric formula for
determining the Linke turbidity factor. Meteorl Rdsch 1980;33:1247.
[49] Incropera FP, DeWitt DP, Bergman TL, Lavine AS. Fundamentals of heat and
mass transfer. Wiley; 2006.
[50] Freiesleben Hansen P, Pedersen EJ. Maturity computer for controlled curing
and hardening of concrete. Nordisk Betong 1977;1:1934.
[51] Bazant ZP, LHermite R. Mathematical modeling of creep and shrinkage of
concrete. John Wiley & Sons; 1988. 484 p.
[52] Japan Concrete Institute. JCI guidelines for control of cracking of mass
concrete. JCI; 2011. p. 153.
[53] Couto LT, Oliveira MS, Dias da Silva J, Magalhes AP. New Paradela dams
spillway: design and testing in reduced hydraulic model. 9 Congresso
Nacional da gua. Associao Portuguesa de Recursos Hdricos, Estoril,
Portugal; 2008 [in Portuguese].
[54] CEN. EN 1992-1 European Standard Eurocode 2: design of concrete structures.
Part 1: general rules and rules for buildings; 2004.
[55] CEN. EN 446:2007 grout for prestressing tendons: grouting procedures; 2007.
p. 20.
[56] Costa A. Thermo-mechanical analysis of self-induced stresses in concrete
associated to heat of hydration: a case study of the spillway of Paradela dam.
Master thesis. University of Minho, Guimares; 2011[in Portuguese].
[57] Azenha M, Faria R, Ferreira D. Identication of early-age concrete
temperatures and strains: monitoring and numerical simulation. Cem Concr
Compos 2009;31:36978.
[58] Yeon JH, Choi S, Won MC. In situ measurement of coefcient of thermal
expansion in hardening concrete and its effect on thermal stress development.
Constr Build Mater 2013;38:30615.
[59] Choi S, Won M. Thermal strain and drying shrinkage of concrete structures in
the eld. ACI Mater J 2010;107:498507.
[60] DAloia L, Chanvillard G. Determining the apparent activation energy of
concrete: Eanumerical simulations of the heat of hydration of cement. Cem
Concr Res 2002;32:127789.
[61] Carino NJ, Lew H. The maturity method: theory and application. Structures
congress & exposition. Washington (DC): American Society of Civil Engineers;
2001. p. 15.
[62] Ulm F-J, Coussy O. Strength growth as chemo-plastic hardening in early age
concrete. J Eng Mech 1996;122:112332.
[63] fdration internationale du bton. b Model Code 2010 (nal draft),
Lausanne, Switzerland; 2011. p. 653.
[64] Aitcin P-C, Neville A, Acker P. Integrated view of shrinkage deformation. Concr
Int 1997;19:3541.
[65] Hedlund H. Hardening concrete. Measurements and evaluation of non-elastic
deformation and associated restraint stresses. Doctoral thesis. Lulea University
of Technology, Lulea, Sweden; 2000.

M. Azenha et al. / Engineering Structures 6263 (2014) 148163


[66] Azenha M, Ramos LF, Aguilar R, Granja JL. Continuous monitoring of concrete
E-modulus since casting based on modal identication: a case study for in situ
application. Cem Concr Compos 2012;34:88190.
[67] Azenha M, Lameiras R, Costa A, Barros J. A cooling system based on ventilated
prestressing ducts for reduction of early age thermal cracking risk in concrete.
BE2012 Encontro Nacional Beto Estrutural 2012. University of Porto, Porto;
2012 [in Portuguese].
[68] Breugel K. Prediction of temperature development in hardening concrete. In:
Springenschmid R, editor. Prevention of thermal cracking in concrete at early
ages. Report 15, RILEM: E&FN SPON; 1998.
[69] Boulay C, Patis C. Strain measurements on concrete at early age (Mesures des
dformations du bton au jeune ge). Mater Struct 1993;26:30711 [in
French].
[70] Bastian G, Khelidj A. The thermophysical properties of a freshly cast concrete.
Propriets thermophysiques dun bton frachement coul. Bulletin de liaison
des LPC 1995;200:2535.
[71] De Schutter G, Taerwe L. Specic heat and thermal diffusivity of hardening
concrete. Mag Concr Res 1995;47:2038.
[72] Kim KH, Jeon SE, Kim JK, Yang S. An experimental study on thermal
conductivity of concrete. Cem Concr Res 2003;33:36371.

163

[73] Bjontegaard O, Sellevold E. Effects of silica fume and temperature on


autogenous deformation of high performance concrete. In: Jensen O, Bentz
D, Lura P, editors. ACI SP 220, Autogenous deformation of concrete. Farmington
Hulls, MI: American Concrete Institute; 2004.
[74] Bentz DP. Transient plane source measurements of the thermal properties of
hydrating cement pastes. Mater Struct/Mater Constr 2007;40:107380.
[75] Choktaweekarn P, Saengsoy W, Tangtermsirikul S. A model for predicting
thermal conductivity of concrete. Mag Concr Res 2009;61:27180.
[76] Viviani M, Glisic B, Smith IFC. Separation of thermal and autogenous
deformation at varying temperatures using optical ber sensors. Cem Concr
Compos 2007;29:43547.
[77] Laplante P, Boulay C. Evolution du coefcient de dilatation thermique du bton
en fonction de sa maturit aux tout premiers ges. Mater Struct
1994;27:596605.
[78] Branco F, Mendes P, Mirambell E. Heat of hydration effects in concrete
structures. ACI Mater J 1992;89:13945.
[79] Ruiz J, Schindler A, Rasmussen R, Kim P, Chang G. Concrete temperature
modeling and strength prediction using maturity concepts in the FHWA
HIPERPAV software. In: 7th International conference on concrete pavements,
Orlando, Florida, USA; 2001.

You might also like