Professional Documents
Culture Documents
Seasonal Thermal Cracking of Concrete Dams in Northern Regions
Seasonal Thermal Cracking of Concrete Dams in Northern Regions
Seasonal Thermal Cracking of Concrete Dams in Northern Regions
Abstract: In several northern regions around the world, concrete dams are subjected to severe seasonal temperature variations, with up to
75°C changes from summer to winter. Those variations contribute to the degradation of the stiffness, strength, and durability of concrete dams.
Thermal stresses and related concrete cracking also need to be evaluated to ensure the structural stability of the dam and to define the initial
conditions for unusual or extreme load combinations, such as floods and earthquakes. This paper presents finite-element modeling procedures
for assessing the thermomechanical behaviors of concrete dams. The stress relaxation and mechanical properties that depend on the temper-
ature of the concrete are first investigated by considering the thermomechanical responses of simple notched beam models. Heat transfer and
thermomechanical analyses are then presented for a 31-m concrete gravity dam and a 214-m multiple arch dam located in Canada. It is shown
that temperature-dependent material properties do not significantly affect the structural response of the dam. Oblique cracks present on the
downstream face of the multiple arch dam are successfully reproduced by the proposed numerical model. Modeling of the related cracked
arch flexibility using equivalent linear continuum concrete constitutive models is also discussed. DOI: 10.1061/(ASCE)CF.1943-5509
.0000483. © 2014 American Society of Civil Engineers.
Author keywords: Gravity dam; Arch dam; Mass concrete; Thermal analysis; Finite elements; Smeared crack model; Thermomechanical
coupling; Thermal stresses; Temperature variations.
high tensile stresses and led to the initiation of cracks in at least parametric study to illustrate the effects of steel reinforcements near
four locations on the dam. Sheibany and Ghaemian (2006) studied the downstream face in complement to the Malm and Anders
the thermomechanical behaviors of the Karaj arch dam (168 m) (2011) study. The Buffalo Bill gravity arch dam (107 m), located
located in Iran. They conducted a 3D linear FE analysis of the in northwest Wyoming, showed long vertical cracks on the down-
dam to determine the variation in temperature throughout the year stream face because of extreme thermal gradients, with monthly
and the related thermal stresses using a sustained concrete elastic temperatures varying from −3–21°C (Boggs 1985). In previous
modulus equal to 65% of its instantaneous value. They found the studies, cracking and permanent irreversible displacements were
temperature load to be the main reason for cracking on the noted at the junction of the La Tuque gravity dam, located in
downstream face of the arch. Cracks were observed at the heel Quebec, and the intake (Caron et al. 2003). It was shown that
of the Daniel Johnson arch dam, located in Quebec, after its the cracking was because of an alkali-aggregate reaction and ther-
construction. These cracks, called plunging cracks [Fig. 1(c)], were mal effects that promoted cracking because of longitudinal geomet-
found to have been caused by stress concentrations attributable to ric discontinuities in the dam. An expansion joint was made by
geometric discontinuities (Veltrop et al. 1990). On the same dam, making a cut in the left part of the gravity dam. The purpose of
oblique cracks, as shown in Figs. 1(c and e), were observed in the the joint was to allow for free movement of the cyclic seasonal dis-
lower part of the downstream faces of many arches (Bulota et al. placements caused by thermal effects and to absorb the volumetric
1991). These cracks resulted from severe winter temperatures. expansion of the concrete. Volynchikov et al. (2011) analyzed the
Finite-element models have been developed to evaluate the Boguchany concrete gravity dam (79 m) in Russia, which experi-
temperature and stress distributions in the dam. Manual iterations ences severe climatic conditions. They suggested steel structural
for relieving excessive tensile stresses by reducing the elastic reinforcement of the downstream face of the dead end section,
Fig. 1. Daniel Johnson Dam: (a) dam view and characteristics; (b) arch 3-4 section; (c and d) crack patterns; (e) computed versus observed down-
stream face cracks
Constitutive Model for Thermomechanical Analyses Smeared Crack Constitutive Model with Fracture
Energy
Thermal Stresses An alternative FE analysis approach uses a criterion based on the
Downloaded from ascelibrary.org by UTFPR - Universidade Tecnologica Federal do Parana on 08/16/19. Copyright ASCE. For personal use only; all rights reserved.
Thermal stresses can be evaluated with the following relationship: concept of energy dissipation by the structure when undergoing a
fracture process (Bhattacharjee and Léger 1992; Calayır and
σT ¼ K r Es αðT − T r Þ ð1Þ Karaton 2005). This theory requires three tensile softening param-
eters: the fracture energy, Gf , defined as the amount of energy
with required to create a crack growth (Bažant and Prat 1988); the direct
Es ¼ Eð1 − ψÞ ¼ E=ð1 þ ϕÞ ð2Þ tensile strength, f t ; and the shape of the tensile softening diagram
(Rots and de Borst 1987). Herein, the fracture process is restricted
where K r = restraining coefficient of the structure; E = short-term to the tensile mode (mode I) that incorporates the strain-softening
(instantaneous) elastic modulus; α = coefficient of thermal behavior of concrete in the fracture process zone (FPZ). Thermal
expansion; T = applied temperature; T r = reference temperature cracks tend to be distributed (or smeared) over a significant
(at zero stress) often corresponding to the mean annual concrete concrete volume, as shown in Figs. 1(c and d). A smeared crack
temperature after a long operational time (Paul and Tarbox propagation model (SCM) that propagates cracks as a continuum
1991);ϕ = creep coefficient; and ψ = stress relaxation coefficient. front smeared over a representative volume associated with the
If the concrete structure is free to move, K r ¼ 0. Otherwise, if the Gauss points of the elements is used herein. In a SCM, after the
structure is restrained either externally or internally while assuming initiation of cracking, the isotropic stress-strain law is changed
a rigid foundation, the restraining factor is K r ¼ 1. Stress relaxa- to an orthotropic law with fixed principal n-, s-, and t-axes of
tion is considered using the long-term elasticity modulus Es , as orthotropy, in which n refers to the direction normal to the
defined by Eq. (2). The stress relaxation factor ψ is assumed to vary crack, and s and t refer to the directions tangential to the crack,
between 0.3–0.5 [U.S. Bureau of Reclamation (USBR) 2006, as indicated by the principal stress orientation. The use of a
1977]. To complement the dam analysis guidelines, experimental SCM avoids any change in the FE mesh topology along the
evidence on the importance of using a sustained concrete modulus evolving crack patterns. Smeared crack propagation models can
to adequately compute thermal stresses from imposed thermal be categorized into fixed or rotating crack models (Rots and
strain or a slowly applied thermal load has been provided by Blaawendraad 1989; De Borst 1997). The FSCM is used in this
Chantelois et al. (1999), Shkoukani and Walraven (1994), and study. In this FSCM, the crack orientation is defined at the crack
Vecchio et al. (1993). initiation based on the principal stress direction and remains
unchanged (fixed) for the rest of the analysis, which is usually
Constitutive Model Using SOM Cracking Criterion performed using incremental steps for the load application.
Very often, a tensile strength criterion based on the concept of SOM
is used for the determination of crack initiation and propagation in ABAQUS and ANSYS Concrete Model Assumptions
concrete structures. When the stress or strain occurring at the
crack-tip exceeds the tensile strength of the material, a crack is as- The ABAQUS FE computer program assumes that the tensile
sumed to propagate. Furthermore, a sudden stress drop is assumed response of concrete is characterized by a stress-strain relationship,
at fracture initiation or when principal stresses reach the tensile as shown in Fig. 2(a) using a FSCM. The compressive strength, the
strength of the material. Bažant and Cedolin (1979) criticized this inelastic strain, the tensile strength, the fracture energy, and the ex-
SOM criterion for its nonobjectiveness because it depends on the ponential shape of the tensile softening diagram are the defining
choice of the FE mesh. Moreover, a comparison of the computed parameters of the concrete model. All of these parameters can
tensile stress with the tensile strength of the material is often be temperature-dependent. In 2D applications, a four-node FE with
inadequate in a cracked structure because spurious results may incompatible modes is used to prevent shear locking.
Fig. 2. (a) Response of concrete to uniaxial loading under tension; (b) notched beam under three-point loading
In the ANSYS FE software, concrete is modeled through the 3D beams tested by Bazant and Pfeiffer (1987) is used. A variation in
solid element (SOLID65). The related triaxial concrete constitutive temperature induces no stress in this beam. Thus, the effect of tem-
model and failure envelope have been described by Willam and perature-dependent material properties on the peak failure load
Warnke (1975). The element SOLID65 takes into account both could be investigated. The model is a notched beam under a
the cracking and crushing failure modes. It is based on three-point loading, is 304.8 mm deep, 762 mm long, and
elastic-perfectly plastic concrete behavior in compression with a 38.1 mm thick, and is made with a maximum aggregate size of
brittle tensile fracture condition. This element with incompatible 12.7 mm [Fig. 2(b)]. Table 1 presents the material properties used
modes is defined by eight nodes with 3 degrees of freedom per node, (Bažant and Pfeiffer 1987).
and can be used with reinforcing bars (embedded smeared steel The compressive strength of concrete has been determined from
layers). The compressive strength and the tensile strength (each of laboratory tests, and the fracture energy value was calibrated by
which can be temperature dependent) are needed to define the con- Bažant and Pfeiffer (1987) to provide a good prediction of the peak
crete failure envelope. The concrete is initially considered to be an load resistance, as shown in the laboratory test. The beam is
isotropic material. Tensile cracking, which is a concern in this study analyzed using a displacement control increment of 10−6 m with
of a multiple arch dam for which compressive stresses are quite small, a maximum displacement of 1.5 × 10−4 m at the two surface nodes
is initiated using a SOM criterion at each Gauss point. The failure in the line above the notch [Fig. 2(b)].
criterion of concrete under tension [Eq. (3)] is expressed as a function The laboratory test showed the ultimate load to be
of the principal stress state, FðsÞ (Willam and Warnke 1975) Pu ¼ 7.78 kN, corresponding to the average responses of three
FðsÞ ¼ s1 − ft ; s1 ≥ s2 ≥ s3 ð3Þ specimens of the same size. The curve at temperature 20°C in Fig. 3
is taken as the reference numerical force-displacement response of
where ft = uniaxial tensile strength; and s1 = maximum principal the notched beam. The ultimate load causing a stress equal to σt ¼
stress. There is no consideration of the concrete fracture energy. After 2.89 MPa at the center of the beam is P ¼ 6.63 kN. This result
cracking, the material properties are adjusted using the smeared band gives a ratio of 1.17 (Pu =P), which is approximately 15% less
of cracks. than the laboratory test result. The difference between the two
values may be because of aleatory uncertainties related to material
properties that are not precisely known.
Thermomechanical Coupling
There are three categories of thermomechanical models for
concrete structures. In an uncoupled model, the transient heat flow
analysis is performed only once, irrespective of the cracks that are
likely to develop. Thermal strains and stresses and related cracking
are then evaluated by applying the computed temperature field
distribution and the temperature-independent mechanical proper-
ties of the concrete. Therefore, the thermal and mechanical
analyses are independent of one another. In a weakly coupled
model, the mechanical (cracking) analysis is conducted using
temperature-dependent material properties. Finally, in a strongly
coupled thermomechanical analysis, the transient heat flow
analysis is repeated between each mechanical analysis step to
evaluate the updated thermal field while considering the effects
of cracking on temperature intensity and distribution (e.g., reservoir
water penetration in a crack).
Fig. 5. Thermomechanical response of a constrained beam: (a) cyclic response; (b) tensile reaction
Fig. 6. Finite-element model of the La Tuque Dam, with the envelope of the minimum (November 15th to April 15th) and maximum (August 1st)
reservoir temperature distributions and ambient air temperature variations
Case Study I: Simulation of the Thermomechanical principal tensile stresses obtained in the dam based on a 1-day in-
Response of the La Tuque Gravity Dam cremental analysis. The maximum principal stress value is
1.79 MPa, which occurred on December 30th. When the mechani-
A set of simulations were performed for the 31 m La Tuque Dam to cal properties do not vary with temperature, the obtained maximum
show the effect of temperature dependent material properties on the value of the principal tensile stress is 1.7 MPa [Fig. 9(b)], and it is
structural response of such a system. remaining below the tensile strength such that no cracking occurs.
Another analysis was made by reducing the tensile strength to
Heat Transfer Analysis of the La Tuque Gravity Dam 0.675 MPa (a quarter of the measured value of 2.7 MPa) to compute
the crack penetration depth. Weak concrete lift joints and
A seasonal thermal analysis was first performed on the gravity dam. freeze-thaw deterioration cycles could lead to reduced tensile
Fig. 6 shows the FE model used and the climatic temperature con-
strength. This analysis was performed with and without
ditions of the air, reservoir, and foundation, which was actually
temperature-dependent material properties. When the mechanical
84.1 × 31 m in the model. The effect of solar radiation is taken into
properties of concrete are temperature-dependent, the results show
account by increasing the mean annual air temperature of the con-
cracked elements at a depth of approximately 1.8 m from the
crete air-interface by 5°C. This increase is based on solar radiation
downstream face. When the mechanical properties are constant,
data available from Tarbox (1977) for exposed surfaces at varying
the cracked element depth increases to 2.3 m from the downstream
slopes, orientations, and latitudes. An assumption of adiabatic
face [see Dontsi-Maken (2012) for more details]. The depth of the
boundary conditions, excluding the dam-foundation contact, was
made for the foundation boundaries. Table 1 shows the thermal crack elements is lower when the properties are temperature-
and mechanical properties of the concrete and foundation used dependent because the tensile strength increases (for a temperature
for the model (Léger and Seydou 2009). Thermal properties were range of 20 to −30°C). Displacements (considering self-weight,
considered to be constant with temperature. The initial temperature
was set to 4°C, which represents the yearly average concrete
temperature (Paul and Tarbox 1991). The transient heat transfer
analysis was carried out for 4.5 years using a step of 1 day to reach
a yearly steady-state temperature response. Fig. 7 shows the
January and July temperature distributions across section A of
the dam. Fig. 8(a) shows the temperature distribution for January
1st, when the temperatures varied from 6.9 to −10°C.
Fig. 8. (Color) La Tuque Dam: (a) January 1st temperature (°C) distribution; (b) January 1st displacements (m) based on temperature-dependent
material properties
Fig. 9. (Color) La Tuque Dam maximal principal stress distribution (Pa): (a) mechanical properties based on temperature; (b) constant mechanical
properties without temperature variations (note: Avg 75% indicates that the results computed from different elements are averaged if
they are within 75% of one another to produce a smooth plot; a discontinuity is shown if an element stress differs from the others by more
than 25%)
temperature load, and hydrostatic load) at the crest of the dam are assumed to be equal to the water (reservoir) temperature at this
also computed. Fig. 8(b) shows the computed displacements and depth. A transient heat transfer analysis was conducted. A mechani-
the maximal horizontal displacement [δðT°Þmax ¼ 6.5 mm] ob- cal analysis considering temperature load, hydrostatic pressure
tained at the crest of the dam for the winter (January 1st) condition load, and self-weight was also performed with mechanical proper-
when the mechanical properties vary with temperature. For this ties varying with temperature. The aim of this study was to evaluate
winter condition, the horizontal displacement value decreases by the effect of water penetration on the computed total stresses. Water
20% [δðT°Þmax ¼ 5.2 mm] when the concrete properties do not penetration affects the stress field by locally changing the
vary with temperature. Thus, a consideration of temperature- temperature of the concrete and the related thermal gradients.
dependent mechanical properties does not produce significant The principal stress results obtained for the half-open lift joint
differences in the results of this dam analysis. are similar to those obtained for the closed lift joint with a maximal
principal tensile stress of 1.7 MPa (occurring on January 1st) at the
downstream face. When water completely penetrates the joint, the
Analysis of Water Penetration in the Dam maximal tensile stress at the downstream face is reduced to
The effect of water penetration in a crack, with a lift joint assumed approximately 1.0 MPa. The stresses decrease by 40% when cold
along section A (Fig. 6), was studied for the La Tuque Dam using a water (still liquid) completely penetrates the joint compared to a dry
strongly coupled thermomechanical analysis. Leaking lift joints joint. However, this is only valid if the water flow is such that the
have been observed at the dam site, which motivates the study water does not freeze near the downstream face during winter
of this particular condition. Two cases are considered: a case with conditions. In most instances, seeping water can freeze, leading
reservoir water penetration up to the middle of the section to an ice plug, which has been observed at some locations at
(half-open lift joint), and the other with water penetration through the La Tuque Dam, and related transitory increases in uplift
the whole section (completely seeping lift joint). The concrete pressure. From a durability point of view, saturated concrete
temperatures at the corresponding nodes of the lift joint are subjected to freeze-thaw cycles is obviously not desirable. For
Quebec, Canada, is the largest multiple arch dam in the world, with
a maximum height of 214 m, a crest length of 1,315 m, and a gross ANSYS FSCM concrete constitutive model to predict the cracking
reservoir capacity of 140 × 109 m3 . The dam is composed of 13 of the thermal (oblique) downstream face of the dam for a known
arches and 14 buttresses [Fig. 1(a)]. Upstream face cracks, called severe thermal loading condition. A long-term objective of the
plunging cracks [Figs. 1(c and d)], were discovered during structural safety assessment of the Daniel Johnson Dam is to
reservoir filling, shortly after its construction in the mid to late use a discrete crack model to represent plunging cracks (allowing
1960s (Tahmazian et al. 1989). In 1968, a system of cracks was for effective modeling of the hydro-mechanical coupling as a func-
observed on the lower downstream faces of the arches. These tion of crack opening) and an equivalent linear continuum concrete
cracks were oriented with an oblique angle to the cantilever material with isotropic or anisotropic properties to represent the
direction and were then named oblique cracks (Veltrop et al. effects of multiple downstream oblique cracks on the stiffness
1990). These cracks were found to be caused by severe winter tem- properties and related stress magnitudes and distributions. A linear
peratures in the region, with an amplitude variation of up to 75°C. equivalent downstream face constitutive concrete model can
Each winter, there are approximately 50 freeze-thaw cycles at the simplify and accelerate the computations of the dam safety indexes
site (Bulota et al. 1991). Freezing and thawing of the water con- for several loading scenarios, including seismic ground motions.
tained in the cracks most likely contributed to the cracking process. However, none of the Daniel Johnson Dam analyses presented
Measurements of the variations in the crack openings showed that herein considered plunging cracks at this stage because this
they were particularly sensitive to sudden temperature variations. consideration is beyond the scope of this study. Four different
Finally, drilling investigations have been performed for many modeling strategies for the oblique cracks are thus considered
oblique cracks, and it has been found that the crack depths are ap- herein.
proximately 35% of the arch thickness, despite a small opening at 1. A linear analysis considering isotropic concrete properties with
the downstream face (Bulota et al. 1991). In this study, arch 3-4 of a constant sustained elasticity modulus (Es ¼ 20.7 GPa) over
the dam was selected to perform numerical predictions of the ob- arch 3-4. This analysis is the reference when no cracks are
served oblique cracks. Figs. 1(b and c) show the section character- considered.
istics of arch 3-4 and a developed view of the crack patterns as 2. A linear analysis considering an isotropically reduced
found after inspection. The cracks are concentrated in the bottom elasticity modulus over half the depth of arch 3-4 to represent
of the arch and have modified the stress distribution in the dam. A the added flexibility because of the oblique cracks in a global
decision was made to protect the downstream faces using insulation manner [Fig. 12(a)]. This is a simple approach that could
walls (installed in 1992) to prevent further extension of the oblique easily be combined with discrete crack modeling of the
cracks. However, in this study, the presence of the insulation walls plunging cracks in future studies.
is not considered. Therefore, a transient heat transfer analysis 3. A nonlinear smeared crack analysis considering the sustained
was made with the environmental parameters established before elasticity modulus and using a concrete material (SOLID65)
the installation of insulation walls (1972–1974). The FE program on the downstream half of arch 3-4 [Fig. 12(b)]. This analysis
ANSYS was used for all analyses because a detailed 3D model of could be used as an anisotropic oblique crack predictor to
the dam has been developed with this program by Hydro-Québec. modify the concrete material properties in a group of elements
to define a linear equivalent cracked continuum on a rational
basis in further analyses.
Thermal Analysis 4. A nonlinear smeared crack analysis considering a concrete
Fig. 10 shows the 3D FE model and the related boundary condi- sustained elastic modulus, using a concrete material (SO-
tions used for the heat transfer and structural analyses. The model LID65) on the downstream half of arch 3-4 with the addition
has nine parts, including the complete arch 3-4, buttresses 3 and 4, of the steel reinforcement present near the downstream face.
the mid arches 2-3 and 4-5, downstream and upstream foundations, This analysis is included to assess the effects of near-face steel
the rock transition zone, and the gravity dam at the top of the arch reinforcement on thermal cracking.
(Fig. 1). There are a total of 183,076 elements and 313,963 nodes. The loading conditions for analyses 1) to 4) are reservoir load
A transient heat flow analysis is performed with a step-basis of (hydrostatic pressure) with a water elevation of 350.95 m,
1 day over a long period to assure convergence. The thermal corresponding to February 21st, dead load, and winter thermal
material properties used are shown in Table 1. The reference load. The mechanical properties used are presented in Table 1.
temperatures are computed as the mean annual concrete tempera- The properties are considered to be temperature independent.
ture for the nine components of the dam with different exposures to The program ANSYS also requires a specification of the shear
ambient (water or air) conditions (Paul and Tarbox 1991). The transfer coefficient along cracks for concrete. This coefficient rep-
reference temperatures were set to 7°C for arch 3-4, 3°C for the resents a shear strength reduction factor for loads that induce shear
buttresses and the gravity dam, 7°C for midarches 2-3 and 4-5, across the crack face. The value of this parameter is set to 0.3, taken
5°C for the upstream foundation and transition zone, and 6°C from Dahmani et al. (2010). The reinforced concrete beam ana-
for the downstream foundation. The critical temperature state re- lyzed by Dahmani et al. (2010) is used as an initial benchmark
tained for the winter condition corresponds to February 21st, with for the nonlinear concrete smeared crack analysis presented in this
a minimum temperature of −15.1°C and a maximum temperature study. Smeared steel layers are used to model the reinforcement
of 6.5°C. Fig. 11(a) shows the temperature distribution through a [Fig. 12(b)]. The concrete material and steel reinforcement are
Fig. 10. (Color) Daniel Johnson Dam thermomechanical model: (a) boundary conditions; (b) air temperatures; (c) reservoir temperature;
(d) foundation temperature
Fig. 11. (Color) Temperature distribution (°C) (February 21st): (a) Z-S cut; (b) through section A-A
Fig. 12. (Color) (a) Distribution of the elasticity modulus Es over the arch for analysis 2); (b) element type modeling zones for analyses 3) and 4) and
reinforcing steel
discretized into elements with the same geometrical boundaries. 100 increments (load steps) to maintain accuracy and to reach
The steel reinforcement effects are averaged in layers within an convergence. The ANSYS program uses the full Newton-Raphson
element that considers their spatial orientation. The volumetric method for incremental load analyses. The stiffness matrix of the
ratios (VR) of the reinforcing steel along the transverse model is updated after each load increment to account for nonlinear
(VR ¼ 13.3%) and longitudinal (VR ¼ 6.8%) directions are behavior (concrete cracking).
provided for each element in the first FE layer along the down-
stream face of the dam. The steel properties are density
Cracking and Displacement Response Analysis
(ρ ¼ 7,850 kg=m3 ), elasticity modulus (E ¼ 200,000 MPa),
Poisson’s ratio (ν ¼ 0.3), and yield strength (Fy ¼ 400 MPa). In An initial smeared crack analysis was conducted using SOLID65
the nonlinear analysis, the total load applied was divided into concrete elements over all of arch 3-4 to compute the penetration,
3) and 4) were then performed by changing the elements of the Variable E (analysis type 2)
downstream half of the arch to elastic (SOLID185) concrete X −0.9 −1.1 −1.4 −1.4 −1.3
Y −19.2 −18.2 −15.6 −11.4 −7.5
elements. This decision was made to produce good comparisons
Z −13.1 −12.9 −12.1 −10.4 −8.5
with analyses 1) and 2). Fig. 1(e) presents the computed crack
pattern on the downstream face compared with the observed crack Smeared crack (SOLID65) (analysis type 3)
pattern. The result shows very good concordance between the two X −0.6 −0.7 −1.1 −0.9 −0.8
patterns, except for the large horizontal computed crack above the Y −17.4 −16.7 −13.9 −9.9 −6.7
Z −10.9 −10.1 −10.2 −8.8 −6.4
oblique ones at elevation of 312 m. The computed crack pattern
shown was obtained after the completion of analysis 4). Smeared crack with steel reinforcement (SOLID65) (analysis type 4)
Fig. 13 shows the pendulums’ positions in arch 3-4. There are X −0.6 −0.8 −1.1 −1.0 −0.7
five simple pendulums and one inverted pendulum in the founda- Y −17.4 −16.7 −14.1 −10.0 −6.7
tion. Displacements at the simple pendulums are always given Z −10.8 −10.4 −10.2 −8.8 −6.6
relative to the anchor of the inverted pendulum, which is considered
as a fixed point in the foundation. Hence, displacements at the
inverted pendulum from the model are subtracted from the which requires a sophisticated analysis when data become available
displacement value obtained at the simple pendulums to compute (Chouinard et al. 2006).
displacements. The numerical displacement values obtained for the
four analyses are presented in Table 3. There is very good concord- Stress Response Analysis
ance between the numerical displacements obtained with the four
analyses, with a maximum difference of ΔZ ¼ 2.5 mm. No at- Tensile stress distributions through a Z-S cut of the dam [Fig. 1(e)]
tempt is made to establish correlation between the computed were obtained after each analysis. There is a good concordance
and measured displacement values because (1) the presence of between analyses 2), 3), and 4). The presence of reinforcing steel
the plunging crack is not considered (this is beyond the scope does not produce significant differences, except on the skin of the
of this study); (2) there was no pendulum in place in 1972; and downstream face, as characterized by the first layer of the FE
(3) the reported numerical displacements are absolute without con- (0.2 m thick).
sideration of a measured (or computed) irreversible component, Fig. 14(a) shows the maximum principal stress distributions
along section A-A [Fig. 11(a)] for the four analyses. The section
depth is 6.19 m, with the origin located at the upstream face. The
stress responses for the four analyses are almost the same up to a
depth of 1.8 m, which is 30% of the section depth. From that point
on, the stresses for analyses 3) and 4), which are similar to one
another, diverge slightly from those of analyses 1) and 2). This
result arises because from that point on, some elements are cracked
through to the downstream face [Fig. 1(e)]. The stresses for
analyses a) and b) remain constant up to 3.10 m, which represents
50% of the section. A reduction in the elastic modulus from the
downstream face to half of the arch section for analysis 2) leads
to a reduction in the computed stress over that zone.
Fig. 14(b) presents the principal stress distributions along sec-
tion B-B, which is closer to the location of the plunging crack
[Fig. 1(c)]. This section is 6.3 m deep and goes from the upstream
face to the downstream face [Fig. 11(a)]. The stresses resulting
from analyses 3) and 4) are quite similar along this section, ranging
from 2.65 MPa at the upstream face (elastic elements) to 0.3 MPa at
the downstream face. The stresses for analysis 2) vary from 3 to
0.27 MPa at the downstream face. The stress distributions from
analyses 2), 3), and 4) are thus in very good agreement. These
results indicate that in section B-B (and even for section A-A),
an isotropic modification of the concrete stiffness properties could
be globally acceptable as a linear equivalent cracked continuum in a
first approximation of the dam’s thermomechanical behavior. The
equivalent isotropic model could later be refined by defining aniso-
tropic equivalent stiffness properties. Fig. 15(a) shows vector plots
of the principal tensile stresses projected on the downstream face
Fig. 13. Arch 3-4 pendulum positions
for analysis a) with a constant Es. The computed principal stresses
Fig. 14. Maximal principal stress (MPa): (a) along section A-A; (b) along section B-B
Fig. 15. (Color) Distribution of tensile principal stresses (MPa) projected on the downstream face by analysis type: (a) constant Es;
(b) isotropic Es reduction; (c) smeared crack; (d) smeared crack with steel
the steel reinforcement near the downstream face produces the ten- to develop an equivalent linear concrete continuum that could
sile stress pattern presented in Fig. 15(d). The maximum computed be efficiently used in more comprehensive analyses and
stress over the entire face is 1 MPa. Moreover, there are some small further studies involving the discrete representation of plun-
areas in which the computed stress is almost zero, meaning that ging cracks and several loading scenarios, including flood
some Gauss points exhibited complete cracking. There is a and earthquake loadings.
noticeable difference between analyses 3) and 4) regarding the
stress patterns presented on the upper part of the downstream face.
Introduction of the steel reinforcement on the downstream face pro- Acknowledgments
duces many elements that crack just in the first layer (0.2 m thick)
of the arch (downstream face). With the presence of steel, when the The authors gratefully acknowledge the financial support provided
concrete cracks, the steel transmits the applied load to neighboring by the Natural Sciences and Engineering Research Council of
elements, which then also become damaged, resulting in a more Canada (NSERC), the Québec Funds for research on nature and
diffused damage pattern. However, some oblique cracks localize technology (FQRNT), and Hydro-Québec, and the collaboration
and penetrate into the dam body over approximately 40% of the of Hydro-Québec engineers who provided field data and the multi-
section depth. ple arch dam FE model used in this study.
0887-381X(1988)2:1(13), 13–24. U.S. Bureau of Reclamation (USBR). (2006). “State-of-practice for the
Léger, P., and Seydou, S. (2009). “Seasonal thermal displacement of gravity nonlinear analysis of concrete dams at the bureau of reclamation.”
dams located in northern regions.” J. Perform. Constr. Facil., 10.1061/ Denver.
(ASCE)0887-3828(2009)23:3(166), 166–174.
Vecchio, F. J., Agotino, N., and Angelakos, B. (1993). “Reinforced con-
Malm, R., and Anders, A. (2011). “Cracking of concrete buttress dam due
crete slabs subjected to thermal loads.” Can. J. Civ. Eng., 20(5),
to seasonal temperature variation.” ACI Struct. J., 108(1), 13–21.
741–753.
Paul, J. W., and Tarbox, G. S. (1991). “Definition of critical thermal states
Veltrop, J. A., Yeh, C.-H., and Paul, W. J. (1990). “Evaluation of cracks in a
in arch dams: A prerequisite for cracking analysis.” Proc. Int. Electric
Power Research Institute (EPRI) Conf. on Dam Fracture, V. E. multiple arch dam.” Dam Eng., 1(5), 5–12.
Saouma, R. Dungar, and D. Morris, eds., Boulder, CO, 643–657. Volynchikov, A. N., Mgalobelov, Y. B., and Deineko, A. V. (2011). “Sub-
Rots, J. G., and Blaawendraad, J. (1989). “Crack model for concrete: Discrete stantiation of the design for the downstream and spillway faces of a
or smeared? Fixed, multidirectional or rotating?” HERON, 34(1), 3–56. concrete dam functioning under severe climatic conditions.” Power
Rots, J. G., and de Borst, R. (1987). “Analysis of mixed-mode fracture in Technol. Eng., 45(2), 91–95.
concrete.” J. Eng. Mech., 10.1061/(ASCE)0733-9399(1987)113:11 Willam, K. J., and Warnke, E. D. (1975). “Constitutive model for the
(1739), 1739–1758. triaxial behavior of concrete.” Proc., Int. Association for Bridge and
Sheibany, F., and Ghaemian, M. (2006). “Effects of environmental action Structural Engineering, Vol. 19, Istituto Sperimentale Modelli E Strut-
on thermal stress analysis of Karaj concrete arch dam.” J. Eng. Mech., ture (ISMES), Bergamo, Italy.
10.1061/(ASCE)0733-9399(2006)132:5(532), 532–544. Zhang, Y., and Ma, L. (1991). “Relation between the ageing of concrete and
Shkoukani, H., and Walraven, J. C. (1994). “Creep and relaxation of con- the ambient temperature.” Proc., 17th ICOLD Congress on Large
crete subject to imposed thermal deformations.” Creep and shrinkage of Dams, International Commission on Large Dams, Paris, France,
concrete, E&F Spon, London, U.K., 45–55. 257–268.