You are on page 1of 9

649

Thermal damage of mass concrete:


experimental and numerical studies on the
effect of external temperature variations
N. Bouzoubaâ, M. Lachemi, B. Miao, and P.-C. Aïtcin

Abstract: According to a report published by the International Commission on Large Dams in 1984, approximately 34% of
all dams are damaged by thermal problems. In this paper, the effects of external temperature variations on mass concrete have
been investigated through laboratory experiments. A concrete block, instrumented with thermocouples and vibrating wire
extensometers and exposed to temperature variations on one face, was used to simulate the behavior of the downstream face
of a concrete gravity dam exposed to thermal cycles. A numerical procedure simulating the thermal effect was performed and
compared with the experimental results. Once the validity of the numerical procedure was established, it was used to
investigate the effects of the variation of outside temperatures on the thermal and mechanical behavior of a typical gravity
dam situated in northern Quebec.
Key words: dam, external temperature, finite element method, instrumentation, mass concrete, mechanical behavior,
monitoring, northern climates, thermal damage.

Résumé : Selon un rapport publié par International Commission on Large Dams en 1984, environ 34% des cas de
détérioration des barrages sont dus à des problèmes d’origine thermique. Dans ce papier, on a mis en évidence en laboratoire
les effets des variations de la température externe sur un béton de masse. Un bloc de béton, instrumenté avec des
thermocouples et des extensomètres à corde vibrante et exposé à des variations de température sur une face, a été utilisé pour
simuler en laboratoire le comportement du béton du parement aval d’un barrage poids soumis à des cycles thermiques. Une
analyse numérique simulant l’effet thermique a aussi été réalisée et les résultats de celle-ci ont été comparés aux résultats
expérimentaux. Une fois la validité de la procédure numérique établie, elle a été utilisée pour étudier les effets des variations
de la température externe sur le comportement thermique et mécanique d’un barrage poids typique situé au nord du Québec.
Mots clés : barrages, température externe, méthode des éléments finis, instrumentation, béton de masse, comportement
mécanique, suivi, climat nordique, endommagement thermique.

Introduction Dam, located in Tennessee, U.S.A., is another example of


thermal-induced cracking. A finite element analysis carried
Unreinforced concrete dams are the largest mass concrete out on the dam confirmed that the damage mechanism was a
structures subjected to various causes leading to cracking. It combination of two phenomena: thermal expansion of the con-
would be difficult to find a gravity dam that had not been crete on the downstream face and a temperature increase in the
affected by cracking of some kind. The majority of severe concrete due to cement hydration (Abraham and Sloan 1978).
cracking in dams, other than those caused by earthquake load- The Dam and Reservoir Damage Committee of the Inter-
ing, is due to climatic conditions. The Daniel–Johnson Dam national Commission on Large Dams published a report in
located in Quebec, Canada, is a typical example of a large dam which freeze–thaw cycles and external temperature variations
that has undergone thermal damage. In this case, cracks devel-
were cited as the principal causes of damage in concrete dams
oped due to temperature variations even before the reservoir
(ICOLD 1984). This is not surprising for dams constructed in
was completely filled (Tahmazian et al. 1989). The Fontana
northern climates where multiple factors can provoke rapid
temperature variations in the concrete structure. Upstream
Received February 12, 1996. faces are affected by seasonal water temperature cycles which
Revised manuscript accepted January 27, 1997. vary with the depth of the reservoir. Such temperature tends to
stabilize at 4°C toward the bottom of the reservoir. Down-
N. Bouzoubaâ. Canada Centre for Mineral and Energy
Technology, 405 Rochester Street, Ottawa, ON K1A 0G1, stream faces are subjected to daily and seasonal temperature
Canada. variations of the ambient air. Finally, during setting and hard-
M. Lachemi and P.-C. Aïtcin. Department of Civil ening of fresh concrete, the heat of hydration initially substan-
Engineering, Université de Sherbrooke, 2500 boulevard tially increases the internal temperature of the concrete which
Université, Sherbrooke, QC J1K 2R1, Canada. cools over time. Such thermal variations provoke nonuniform
B. Miao. School of Engineering, Université de Moncton, volumetric changes in the concrete, which are related to the
Moncton, NB E1A 3E9, Canada. thermal gradient between the center of the dam and its sur-
Written discussion of this article is welcomed and will be faces. When thermal stresses exceed the tensile strength ca-
received by the Editor until December 31, 1997 (address inside pacity of the concrete, cracks form in the dam which can lead
front cover). to further deterioration of the structure. Climatic conditions

Can. J. Civ. Eng. 24: 649–657 (1997) © 1997 NRC Canada


650 Can. J. Civ. Eng. Vol. 24, 1997

Fig. 1. Schematic representation of the experimental system used in the study.

continue to contribute to the damage over the service life of therefore be found in the same thermal state, but at different
the dam long after the effect of heat of hydration has dissi- times. If the ratio of dimensions of the two bodies is D1/D2,
pated. the time corresponding to obtaining the same thermal state in
Several researchers have investigated thermal stresses in both bodies can be expressed as
mass concrete. Some researchers focused their investigation 2
on fresh concrete properties and encountered certain difficul- t1  D1 
=
t 2  D2 
[2]
ties pertaining to the control of the experimental and numerical  
evolution of material properties. Research on the effect of ex- The typical dam used in this study for thermal simulation is a
ternal temperature variations on hardened concrete have not gravity dam with a thickness that varies between 10 and 15 m.
yielded consistent results that are supported by experimental The concrete block used in the laboratory tests to simulate the
data. Tests results obtained to date from instrumented dams temperature variation in mass concrete consists of a large cyl-
have not been able to dissociate thermal effects from static or inder with a diameter of 920 mm and a length of 1000 mm. The
dynamic effects. cylindrical form was chosen to simplify the modeling because
The objective of the research reported in this paper is to of its axisymmetrical geometry. The part of the dam simulated
highlight the thermal effects of climatic changes on concrete by the laboratory-scale experiment represents a cylinder meas-
determined experimentally by simulating the exposure of a uring 10 m in thickness and 9.2 m in diameter. The thickness
mass concrete structure in the laboratory to extreme tempera- of the laboratory block is therefore 1/10 that of the dam. On
ture variations, then numerically validating the measured re- this scale (1/10), the daily temperature cycle corresponds to a
sults using a finite element analysis. The finite element model cycle of 14.4 min.
validated by experimental studies can then be used to carry out
studies needed either to prevent thermal damage that can take Thermal cycles
place in concrete dam or to optimize rehabilitation programs The thermal cycles chosen for the laboratory tests were as
of damaged structures. follows:
• Summer conditions: a temperature cycle of 7/27°C was
Experimental program used for the air in contact with the downstream face of the
experimental block, and a constant temperature of 18°C
Simulation of a gravity dam was maintained on the upstream face.
Heat transfer within a mass, without any loss, is described by • Winter conditions: a temperature cycle of –10/10°C was
Fourier’s equation which relates the temperature, T, at a given used for the air in contact with the downstream face of the
point of the mass to the time, t: experimental block, and a constant temperature of 4°C was
∂T  ∂2T ∂2T ∂2T  maintained on the upstream face.
[1] = a  2 + 2 + 2
∂t  ∂x ∂y ∂z  Experimental setup
where a is the diffusivity of the material. The schematic of the experimental system is presented in
This equation shows that the diffusivity is sufficient to char- Fig. 1. The setup consists of four main parts: the experimental
acterize the heat transfer of a given material. Therefore, two block, the downstream temperature control, the upstream tem-
bodies that are geometrically equivalent and composed of the perature control, and the data acquisition device. The lateral
same material can have thermal states that will evolve in an surface of the block was covered with an insulating fiberglass
identical manner, although at different rates. Such bodies will mat measuring 150 mm in thickness in order to prevent heat

© 1997 NRC Canada


Bouzoubaâ et al. 651

Fig. 2. Position of thermocouples within the block.

Fig. 3. Position of extensometers within the block.

loss in the radial direction. The block was instrumented with 5 Because of the thermal inertia of the massive concrete
vibrating wire extensometers and 22 thermocouples. The loca- block, the duration of one thermal cycle lasted 2.4 h for
tions of the thermocouples and the vibrating wire extensome- summer conditions. For winter conditions, each cycle lasted
ters are shown in Figs. 2 and 3, respectively. It is important to for a much longer time. Therefore, the experimental system
note that the instrumentation was mostly concentrated near the was unable to produce the desired scaled thermal frequency
downstream face of the block, which was subjected to a greater of 14.4 min/cycle. Therefore, the testing lasted for a longer
variation of temperature. period of time, and the thermal gradients in the concrete
An aluminum reservoir having the same diameter as the block were less critical than those that could have taken
block was used to vary the temperature at the downstream face place if the 14.4 min/cycle has been achieved.
of the experimental block. The reservoir was linked by two
pipes to a specially developed programmable freeze–thaw box. Materials
Five parameters were defined for each cycle: the minimum The cement used for the manufacturing of the concrete block
temperature, the maximum temperature, the maximum dura- was a type 20M cement (a modified type 20 Canadian cement)
tion, the minimum duration, and the maximum number of con- known for its low heat of hydration. Such cement is currently
secutive cycles. A constant surface temperature on the used by Hydro-Québec for mass concrete construction. A natu-
upstream face of the experimental block was maintained using ral silicate sand and a coarse aggregate composed of a meta-
a second aluminum reservoir connected to a freezer with a morphic limestone were used. The coarse aggregate was
temperature control. A space between the reservoir and the composed of different classes of 60–40 mm, 40–20 mm, and
experimental block was provided to simulate the air to con- 14 mm nominal-size aggregate, in compliance with the recom-
crete heat transfer effects. The concrete block and aluminum mendations of an ACI 207 Committee.
reservoir system was wrapped in fiberglass insulation in order The various operating methods ranging from manufacturing
to minimize heat exchange to the surrounding environment. of the concrete to determining its thermomechanical properties

© 1997 NRC Canada


652 Can. J. Civ. Eng. Vol. 24, 1997

Table 1. Composition and properties of the concrete. Fig. 6. Concrete strains at 100 mm depth after 10 and 600 thermal
cycles.
Parameter Value
Water 122 kg/m3
Cement 203 kg/m3
Fine aggregate 775 kg/m3
Coarse aggregate 40–60 mm (30%) 386 kg/m3
Coarse aggregate 20–40 mm (30%) 386 kg/m3
Coarse aggregate 14 mm (40%) 514 kg/m3
Air-entraining agent 61 mL/m3

Slump 50 mm
Air content 7.4%
Unit weight 2310 kg/m3

Compressive strength 22 MPa


Modulus of elasticity 27 GPa
Flexural strength 3.1 MPa
Tensile strength 2.2 MPa were carried out in compliance with ACI and ASTM stand-
Thermal expansion coefficient 7.5 × 10–6 /°C ards. The thermal expansion coefficient was determined fol-
lowing an experimental procedure proposed by Bouzoubaâ
(1991).
The composition of the concrete and its measured fresh
Fig. 4. Concrete temperature readings due to the thermal cycles. properties are summarized in Table 1. Table 1 also presents
the principal thermomechanical properties of the concrete ob-
tained from cores and sawed beams taken from the block. The
tests were done at 91 days from samples extracted from the
block 1 day before the test. During the storage time between
coring, cutting, and testing, the samples were enclosed in
water-tight plastic sacks to prevent water loss.

Experimental results
The downstream face of the experimental block was submitted
to air temperature cycles (7/27°C) of 2.4 h duration. The tem-
perature variations in the block were recorded using thermo-
couples and vibrating wire extensometers. The temperature
variations at the surface and inside the concrete block are pre-
Fig. 5. Measured strains at different locations in the experimental
sented in Fig. 4. With temperature cycles of 7°C to 27°C, the
block.
temperature at the concrete surface varied from 11.5°C to
21.5°C, whereas it remained virtually constant toward the cen-
ter of the block.
Curves representing concrete strain at different locations in
the experimental block due to the variations of external tem-
perature have been incorporated in Fig. 5. From this the fol-
lowing can be observed:
• At 600 mm from the downstream face, the strain observed
was not in phase with that registered by extensometers at
other locations. This is because a drop in temperature at the
downstream face causes tensile stresses at the concrete sur-
face and compressive stresses further from the surface so
as to obtain a balance of forces in the system. However, in
areas further from the surface where there were no tempera-
ture variations, strain obviously did not occur.
• The amplitude of the strain recorded at 100 mm from the
surface was approximately the same as the one recorded at
200 mm, even though the amplitude of the temperature at
100 mm from the surface was greatly superior to the value
obtained at 200 mm. This behavior can essentially be attrib-
uted to three factors: surface cracks, creep in the concrete,
or malfunctioning of the vibrating wire extensometer.

© 1997 NRC Canada


Bouzoubaâ et al. 653

Fig. 7. Evolution of the calculated temperature versus time at Fig. 9. Comparison of calculated and measured temperatures at 100
different locations in the block. mm depth.

Fig. 8. Comparison of calculated and measured temperatures at the 4.5 W/(m⋅°C) and from 870 to 1080 J/(kg⋅°C), respectively.
surface. The values chosen for this study are 2.1 W/(m⋅°C) and
1000 J/(kg⋅°C). The temperature cycle conforms to that ob-
tained using thermocouples placed in the air at the downstream
face of the experimental concrete block.
Figure 7 presents the evolution of the calculated tempera-
ture in the concrete block over time at different depths. The
figure shows that the temperature at the surface varied between
10.5 and 21.5°C, whereas it was constant at the center of the
block. These results are quite close to those obtained experi-
mentally, as illustrated in Figs. 8 and 9. Figure 8 shows that
the calculated extremes were slightly offset in time as com-
pared with the measured results. This suggests that the dif-
fusivity value used in the model was a little higher than the
real value. A high diffusivity leads to a greater conductivity,
and a more conductive material leads to a faster balance with
extreme temperature values. Despite the slight variation in
In the present case, a malfunctioning of the vibrating wire temperatures, it can be concluded that the numerical model is
extensometer seems to be the cause of the lower than expected validated by the physical testing and temperature monitoring.
strain value, since no cracking was observed in the concrete
following an ultrasound and microscopic examination. Also, Thermal stress modeling
the effects of creep were negligible at 100 mm in depth, since
the amplitude was essentially the same after 10 and 600 ther- The stresses generated by temperature cycling in the concrete
mal cycles, as indicated in Fig. 6. Figures 5 and 6 show that were modeled using an isotropic linear elastic model (module
the maximum and minimum peak measurements of this LINE of the CESAR-LCPC calculation code). The ther-
extensometer are prevented, thereby causing a reduction in the momechanical properties of the concrete correspond to those
amplitude of measured strain. determined experimentally (Table 1). It is necessary to impose
the calculated temperature range to the block for each time
increment to obtain the corresponding stress range.
Numerical modeling of temperature range
Figure 10 shows the comparison between the measured and
Modeling and numerical calculations of thermal and mechani- calculated strains at 200 mm from the downstream face.
cal behaviors were carried out using the finite element code Table 2 provides the strain magnitude (maximum minimum)
CESAR-LCPC (Humbert 1989; Dubouchet 1992), which was in the concrete which was calculated and measured at five
developed at the Laboratoire Central des Ponts et Chaussées locations representing the positions of the vibrating wire ex-
located in Paris, France. The procedure includes modeling the tensometers. The results generally concur except at the
experimental block bidimensionally using appropriate mod- 100 mm site where the vibrating wire extensometer placed in
ules and comparing the numerical results with the experimen- the experimental block malfunctioned. Figure 10 shows that
tal ones. If the two results agree, the modules are therefore the model results are in agreement with the experimental
validated by experimentation and can be used to model an strains. Since the vibrating wire extensometer placed at
actual concrete structure. The temperature distributions in the 200 mm from the downstream face was perpendicular to the
experimental block were modeled during the cyclic testing in direction of the other vibrating wire extensometers, this justi-
the transient heat flow analysis using the DTLI module of the fies the bidimensional and axisymmetrical calculation used in
CESAR-LCPC code. this study.
According to the literature, the thermal conductivity and The experimental and numerical results for temperature
specific heat of mass concrete typically vary between 1.5 and and strain at the surface and inside the experimental block

© 1997 NRC Canada


654 Can. J. Civ. Eng. Vol. 24, 1997

Fig. 10. Comparison of calculated and measured strains at 200 mm Fig. 11. Size and finite element discretization of the evaluated
depth. dam-reservoir.

Table 2. Strain amplitude in the concrete.

Strain amplitude (µm/m)


Distance from the
downstream face (mm) Measured Calculated
100 12 37
200 12.5 15
300 3.5 5
600 1.5 2.5
900 0.8 1

demonstrate that the transient heat flow model and the iso-
tropic linear elastic model can provide an excellent prediction
of temperature and strain in the experimental block. Table 3. Thermal and mechanical parameters used in the model.

Parameter Value
Application to a typical gravity dam Specific heat 912 J/(kg ⋅ °C)
The gravity dam chosen is a dam measuring 90 m in height and Conductivity 2.62 W/(m ⋅ °C)
70 m in width which corresponds approximately to gravity Thermal expansion coefficient 8.6 × 10–6 /°C
dams built in Quebec (84 m high at Outardes 3, and 91 m at Exchange coefficient (convection and radiation) 27.4 W/(m2 ⋅ °C)
Manicouagan 2). Figure 11 gives the dam dimensions and the Modulus of elasticity 27 GPa
mesh used in the finite element model. A refined mesh was Unit weight 2400 kg/m3
used at the downstream surface of the dam, because this area Poisson’s ratio 0.2
was most affected by temperature variations. The thermal and
mechanical parameters chosen for the model are presented in Figure 13 shows the temperature variations at the down-
Table 3. stream face and at 3 m from the surface. It can be seen that the
Meteorological data from northern Quebec in the proximity amplitude of the thermal cycle diminished rapidly with
of the Manicouagan 2 and Outardes 3 dams were used to de- distance. This amplitude was 38°C at the face and 20°C
termine air temperature variations on the upstream and down- at 3 m from the surface. Also, it can be seen that the
stream faces of the studied dam, the reservoir water surface temperature was as low as –16°C during the month
temperatures at various depths, and solar radiation. Solar ra- of January (day 19).
diation was taken into account by adding the temperature vari- Figure 14 shows the distance corresponding to frost pene-
ation due to radiation flow to the air temperature. The average tration that varied between 4 and 6.5 m. It is important to note
daily air temperature varies between –17°C in January that the frost zone was defined as the area where the tempera-
and +22.7°C in July. When solar radiation is excluded from ture is below 0°C. Due to the thermal inertia of the dam, the
the calculation, this average daily temperature becomes deepest frost penetration was observed at the beginning of May
–18.5°C and +20°C. The initial temperature was defined ac- (day 123) with a lag time of 104 days after the surface tem-
cording to the Paul and Tarbox (1991) calculation. In the pre- perature reached its minimum value.
sent case, it was equal to 5.6°C or 1.8°C depending on whether The variations of the highest values for the principal
or not solar radiation was considered. Figure 12 presents the stresses as a function of time are shown in Fig. 15. It is noted
initial and boundary thermal conditions in the dam. that maximum tensile stress occurs on day 19 in January,

© 1997 NRC Canada


Bouzoubaâ et al. 655

Fig. 12. Initial and boundary conditions of the evaluated dam-reservoir.

Fig. 13. Temperature variations at the downstream face and at 3 m Fig. 14. Frost penetration in the dam.
from the surface.

conductivity and specific heat but uses a solar radiation flux


equal to zero.
The depth of frost penetration and the maximum tensile
stress were calculated for the three test parameters. The greater
the thermal conductivity value and the lower the specific heat
value, the deeper the frost penetration was. Also, when solar
which corresponds to the day when the surface temperature
radiation is not considered at the dam limits, the depth of frost
was lowest. In Fig. 16, the maximum tensile stress value was
penetration is greatly overestimated (Table 5). On the other
equal to 5.5 MPa, located at the toe of the dam near the down-
hand, the mechanical behavior of the dam seems insensitive to
stream face. Figure 17 shows the part of the dam most likely
the variations of the test parameters (Fig. 18), which is prob-
to be damaged by the effects of temperature cycles. This sec-
ably due to the lowest value of the initial temperature when
tion corresponds to the area where tensile stresses can be
the solar radiation is not taken into account.
greater than 3 MPa. This is almost equivalent to the tensile
The results obtained for the gravity dam investigated in this
strength of the concrete.
study are similar to those reported by Léger et al. (1993). How-
This study also considers the influence of various modeling
ever, unlike the latter case, the numerical models presented in
parameters on the predicted response of the typical gravity
this paper have been carried out in addition to laboratory ex-
dam. In particular, the sensitivity of the response to changes
perimentation which validates the numerical analysis.
in thermal conductivity, specific heat, and solar radiation was
evaluated. Three thermal states were studied relating to these
parameters (Table 4). The first corresponds to average values Conclusions
of thermal conductivity and specific heat. The second corre-
sponds to extreme values for thermal conductivity and specific In this study, an experimental system was developed to monitor
heat. The third corresponds to extreme values for both thermal the temperature variation as well as strain in a mass concrete block

© 1997 NRC Canada


656 Can. J. Civ. Eng. Vol. 24, 1997

Fig. 15. Principal stress variations versus time.

Fig. 16. Tensile stresses on day 19. Fig. 17. Area most likely to be damaged on day 19 by the effects of
temperature cycles.

caused by external temperature variations. This study also in- dam having the size of those constructed in the Manicouagan
cluded a numerical analysis which was used to examine the and Outardes regions and indicated the following:
behavior of a typical gravity dam. The most important conclu- — The maximum depth of frost penetration at the downstream
sions can be summarized as follows: face of the dam is approximately 6 m and is achieved at the
• The results obtained on a hardened concrete have shown beginning of May.
that the experimental setup works well despite the technical — The maximum principal tensile stress is equal to 5.5 MPa.
difficulties which were encountered during this study to de- Such stress takes place in mid January at the toe of the dam
crease the frequency of temperature cycles. near the downstream face.
• The temperatures calculated using a transient heat flow — The downstream face of the dam is the most likely to be
analysis at different locations in the block concur with the damaged. Damage thickness is limited to 1 m and is reduced
measured temperatures. with the increase in the distance from the bottom of the
• The linear elastic model can adequately predict strain in dam.
the experimental block when subjected to external temperature — The greater the thermal conductivity of the concrete and the
variations. Both thermal and mechanical models were validated lesser the specific heat value, the greater the frost penetra-
in this study. The models were then applied to an actual gravity tion becomes.

© 1997 NRC Canada


Bouzoubaâ et al. 657

Fig. 18. Principal stress variations for the three thermal states studied.

Table 4. Thermal states studied.


References
Conductivity Specific heat
State (W/(m ⋅ °C)) (J/(kg ⋅ °C)) Radiation Abraham, T.J., and Sloan, R.C. 1978. TVA cuts deep slot in dam, ends
1 2.62 912 Yes cracking problem. Civil Engineering, American Society of Civil
2 3.68 870 Yes Engineers, January, pp. 67–70.
American Concrete Institute Committee 207. 1987. Mass concrete. In
3 3.68 870 No
American Concrete Institute manual of concrete practice. Part I.
Detroit, Mich.
American Concrete Institute Committee 207.2. 1990. Effect of re-
straint, volume change, and reinforcement of cracking of mass
Table 5. Depth of the frost penetration and maximum tensile stress
concrete. Committee report. American Concrete Institute Materi-
for the thermal states studied. als Journal, 87(3): 271–300.
Bouzoubaâ, N. 1991. Coefficient de dilatation thermique du béton :
Frost penetration Maximum tensile
influence de la nature des granulats. M.A.Sc. thesis, Department
State (m) stress (MPa) of Civil Engineering, Université de Sherbrooke, Sherbrooke, Que.
1 4 – 6.5 5.5 Dubouchet, A. 1992. Développement d’un pôle de calcul: CESAR-
2 5–8 5.5 LCPC. Bulletin de Liaison des Laboratoires des Ponts et
3 7.5 – 10 5.6 Chaussées, 178, March–April, pp. 77–84.
Humbert, P. 1989. CESAR-LCPC : un code de calcul par éléments
finis. Bulletin de Liaison des Laboratoires des Ponts et Chaussées,
160, March–April, pp. 112–115.
— When solar radiation is not included in the boundary con- International Commission on Large Dams. 1984. Deterioration of
ditions of the concrete, the depth of frost penetration can be dams and reservoirs, examples and their analysis. A.A. Belkema
greatly overestimated. Publishers, Bookfield, Va.
It is important to note that the validated models can be used Léger, P., Venturelli, J., and Bhattacharjee, S.S. 1993. Seasonal tem-
to prevent thermal damage that can take place in mass concrete perature and stress distributions in concrete gravity dams. Part 1:
modelling. Canadian Journal of Civil Engineering, 20(6): 999–1017.
gravity dams by using some insulation systems either to mini-
Léger, P., Venturelli, J., and Bhattacharjee, S.S. 1993. Seasonal tem-
mize thermal tensile stresses or to reduce the extent of frost perature and stress distributions in concrete gravity dams. Part 2:
penetration. behaviour. Canadian Journal of Civil Engineering, 20(6): 1018–1029.
Paul, J.W., and Tarbox, G.S. 1991. Definition of critical thermal states
Acknowledgments in arch dams — a prerequisite for cracking analysis. Proceedings
of the International Conference on Dam Fracture, Boulder, Colo.,
This research program was carried out at the Université de pp. 643–657.
Sherbrooke under the sponsorship of Hydro-Québec. The Tahmazian, B., Yeh, C.H., Bonazzi, D., and Aubin, L. 1989. Non-
authors are especially grateful to Claude Faucher, technician at linear analysis of Daniel Johnson dam. Proceeding, International
the Department of Civil Engineering of the Université de Commission on Large Dams Symposium on Analytical Evalu-
Sherbrooke, for his assistance in the experimental part of the ation of Dam Related Safety Problems, Copenhagen, Denmark.
program.

© 1997 NRC Canada

You might also like