Simulation analysis of thermal stress of RCC dams using 3-D nite

element relocating mesh method

Yaolong Chen
*
, Changjiang Wang, Shouyi Li, Ruijun Wang, Jing He
Xi'an University of Technology, Southern Golden Flower Road, Xi'an 710048, Shaanxi Province, People's Republic of China
Received 9 January 2001; revised 5 April 2001; accepted 21 May 2001
Abstract
The 3-D nite element relocating mesh method is developed for simulation analysis of temperature and thermal stress distribution in a
roller compacted concrete dam during the construction period. According to the relation between specic properties and age of concrete,
some meshes are merged into a larger mesh or a few larger meshes when the age of the concrete is appropriate. Using this method, the total
number of elements and nodes were remarkably reduced when the dam height was increased. When the change in elastic modulus, creeps and
hydration heat is within the limits permitted by design criteria, the relocating of mesh will start. Using this method, a 3 D simulation analysis
of thermal stress in a roller compacted concrete (RCC) high dam can be realized by microcomputer and appeared at the construction site. On
the basis of real factors during the construction period, an engineer can predict the distribution of temperature and thermal stress in the RCC
dam. Therefore, engineers can take appropriate measures to control the concrete temperature to reduce the thermal stress and avoid crack
development within the dam. q 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Roller compacted concrete gravity dam; Simulation analysis; 3-D FEM; Thermal stress; Relocating mesh method
1. Introduction
During the construction period of a roller compacted
concrete (RCC) dam, the thickness of each layer is usually
0.30.5 m. Hence, an RCC dam with a height of 200300 m
may consists of several hundreds up to thousand layers. The
heat will be exchanged between the top surface of a new
placement layer and the circumference, and between the
bottom surface and the old layer or base rock. Since the
gradient of temperature and the stress in an RCC dam is
great in the vertical direction, when calculating thermal
stress in the dam during construction period, the mesh
sizes of the region has to be 0.30.5 m in order to reduce
the calculation error. The size of the mesh is usually same as
the thickness of the layer. After generating the mesh, it is
very difcult to carry out a 3 D nite element simulation
analysis by a microcomputer due to the great number of
elements and nodes [1].
When the thermal stresses in the Three Gorges Project
(TGP) concrete gravity dam were studied in 1989, a kind of
element having two layers or several layers was developed
[2]. On the basis of this principle, the relocating mesh
method is developed in this paper. Using the relocating
mesh method, the 3 D simulation analysis of the thermal
stress in a high RCC dam can be completed by micro-
computer and performed at the construction site.
2. Time for starting relocation of the mesh
According to the relationship between properties and age
of concrete, when the age of concrete is appropriate, some
meshes are merged into a larger mesh or a few larger
meshes. This method is called relocating mesh method.
Using this method, the total number of elements and
nodes were remarkably reduced when the dam height was
increased. The calculation time and the storage space of the
computer were greatly reduced.
The time for starting relocation of the mesh must be
established. It should assure that the error of relocating
mesh was controlled within the limits permitted by the
design criteria. The properties of concrete, such as elastic
modulus, creep and hydration heat depend on the age of
concrete. The inuence of these factors should be taken
into account.
In the upper concrete layers of RCC dam, the elastic
modulus, creeps and hydration heat change with respect to
time; each thin concrete layer is meshed as one layer of
element. In the lower concrete layers of RCC dam, the
Advances in Engineering Software 32 (2001) 677682
0965-9978/01/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved.
PII: S0965-9978(01)00025-4
www.elsevier.com/locate/advengsoft
* Corresponding author. Tel.: 186-29-328-1434; fax: 186-29-323-5545.
E-mail address: c21306@pub.xaonline.com (Y. Chen).
difference of elastic modulus, creeps and hydration heat
between each thin layer is very small; these thin layers
can be merged into a larger mesh. The calculating model
and discretized model of an RCC dam are shown in Fig. 1.
The relocating mesh model is shown in Fig. 2.
2.1. The elastic modulus effect
The elastic modulus of RCC at time t can be written as
Et E
0
1 2e
2lt
; 1
where E
0
is the nal elastic modulus, l is a coefcient of
placement temperature [3].
If t
i
, t
j
are the age of layers i and j of RCC, we obtain
Et
i
2Et
j

Et
j

# 1
1
; 2
where 1
1
is the acceptable error; Et
i
; Et
j
are the elastic
modulus of layers i and j of RCC, respectively.
If the layers from layer i to j were merged into a larger
layer, the average elastic modulus of these layers is used as
elastic modulus of the larger layer. The error of elastic
modulus of the relocating meshes is smaller than 1
1/2
.
Substituting Eq. (1) into Eq. (2) yields Eq. (3).
t
i
2
1
l
ln1 11
1
e
2lt
j
21
1
: 3
After ageing of RCC, the increment of elastic modulus of
RCC is very small. When the age of RCC is t
p1
, the differ-
ence of elastic modulus between layers is smaller than 1
1
,
and the layers whose ages are greater than t
p1
are merged
into a larger mesh.
The average elastic modulus of the layers is used as the
elastic modulus of the larger layer [8].
From Eq. (2) and let t ! 1; yields 1 11
1
e
2lt
j
2
1
1
! 0; and for t
j
t
p1
; we get Eq. (4)
t
p1
2
1
l
ln
1
1
1 11
1

: 4
The layers whose age is greater than t
p1
are merged into a
larger mesh by relocating.
2.2. Thermal insulation temperature rise effect
The thermal insulation temperature rise u (t) of RCC can
be written as
ut u
0
1 2e
2mt
; 5
where u
0
is the nal thermal insulation temperature rise, m
is the hydration heat coefcient of cement.
If t
i
, t
j
are the ages of layers i and j of RCC, respectively,
we can obtain
ut
i
2ut
j

ut
j

# 1
2
; 6
where 1
2
is the acceptable error; u(t
i
), u(t
j
) are the nal
thermal insulation temperature rise of layers i and j of RCC,
respectively.
Again, we obtain
t
p2
2
1
m
ln
1
2
1 11
2

: 7
It means that the layers whose age is greater than t
p2
are
merged into a larger mesh by relocating, and the error is
smaller than 1
2
.
2.3. The creep effect
The creep of RCC can be written as
Ct; t A 1Bt
2s
1 2e
2gt2t
; 8
where t is the time, t is the age of concrete; A, B, s, g are the
experimental coefcients in concrete creep.
In Eq. (8), let t ! 1; then the creeps whose age is greater
than t are
C1; t A 1Bt
2s
: 9
If t
i
, t
j
are the ages of layers i and j of RCC, respec-
tively, we can obtain
Ct
i
2Ct
j

Ct
J

# 1
3
; 10
where 1
3
is the acceptable error.
Y. Chen et al. / Advances in Engineering Software 32 (2001) 677682 678
Fig. 1. Finite element mesh for the calculating model and the discretized
model for RCC dam.
Fig. 2. Finite element mesh for the relocating model.
Substituting Eq. (9) into Eq. (10) we get Eq. (11)
t
p3

B1 21
3

A1
3

1=S
: 11
This implies that the layers whose age is greater than t
p3
are merged into a larger mesh by relocating and the error is
smaller than 1
3
[7,8].
For example, assume a RCC gravity dam, with height
156 m and maximum width at base 120 m, which is
shown in Fig. 1. Also, assume that the monthly concrete
placement is 6 m in height and the thickness of each thin
layer is 0.3 m, the total number of placement layers is 20.
Each layer is placed within 6 h and there is a long placement
interval of 5 days after continuously placing ve layers.
When the next 21st thin layer is placed, the age of concrete
in the bottom ten layers is 3022 days. The ten thin layers
are merged into a larger layer (3 m) whose average age is 26
days.
The relation between elastic modulus and the age of
concrete is
Et 29:961 2e
20:094t
:
The following can be obtained:
E26 2E22
E22
3%;
E30 2E26
E26
2:1%:
The above error for relocating is only 3%.
The relation between the thermal insulation temperature
rise of hydration heat and the age of concrete is
ut 16:871 2e
20:24t
:
The following can be obtained
u26 2u22
u22
0:3%:
The error for relocating is only 0.3%.
The relationship between the creep and the age of
concrete is
Ct; t 0:945 1159:539t
20:733
1 2e
20:2t2t
10
26
;
while t 3 days, we have
C26; 3 2C22; 3
C22; 3
1:2%;
while t 7 days, we have
C26; 7 2C22; 7
C22; 7
2:9%:
It shows that the maximum error for relocating is only
3%, and it is within the limits permitted by design criteria.
3. Finite element model
3.1. Foundation region
The thermal stress is caused by the change of temperature
within a constrained body. After the concrete of dam is
built, the hydration heat raises the temperature of dam [4].
Part of the hydration heat in the dam will be transmitted to
the foundation. Usually, the distance to which the hydration
heat is transmitted is less than 30 m. In this model, for
calculating the thermal stress, the depth of the foundation
is 60 m, the length from heel to upstream boundary is 60 m
and the length from toe to downstream boundary is 60 m as
well. This model is shown in Fig. 1.
3.2. Timetable of construction
According to the construction plan of the RCC dam, the
total RCC construction time is planed to be 33 months from
March of the rst year to November of the third year. The
dam was constructed to 156 m in height.
The numerical model used allows simulation of the actual
evolutionary construction process of the dam. After continu-
ously placing ve thin layers of RCC, usually, there is a long
placement interval, each layer takes 6 h, and the thickness of
the layer is 0.3 m. If the increase of the dam is 3 m in a
month, the placement interval will be 13.75 days. From
September of the rst year to May of the second year, the
average monthly increase of the dam is 6 m. The placement
interval is 6.25 days. From June to August of the second
year, the average monthly increase of the dam is 3 m. The
placement interval is 13.75 days. From September of the
second year to May of the third year, the average monthly
increase of dam is 6 m. The placement interval is 6.25 days.
From June to November of the third year, the average
monthly increase of the dam is 3 m. The placement interval
is 13.75 days.
3.3. Meshes and time step
The numerical model used for the 3-D thermal analysis
consists of a nite element discretization of the central
cross-section of the dam. The eight-node hexahedral
isoparametric elements and six-node pentahedral isopara-
metric elements are used in the numerical model. The
mesh represents the body of the dam (from elevation 0 to
156 m), plus the foundation (from elevation 260 to 0 m).
The dam is formed by 520 RCC layers 0.30 m thick. In the
3-D model, every layer of the dam is discretized into
10 120 1200 elements (X-direction, 10 elements;
Y-direction, 120 elements). The size of maximum element
is 2.2 m 1:0 m 0:3 m at elevation 0 m. In other words,
the length of the element in X-direction is 2.2 m, and the
length of the element in Y-direction is 1.0 m, and the height
of the element in Z-direction is 0.3 m. As the height of the
dam increases, the size of the elements become smaller.
Thus, the total number of elements of the dam body is
Y. Chen et al. / Advances in Engineering Software 32 (2001) 677682 679
520 10 120 624 000: The directions of the foundation
of the dam is 2.2 m 2:0 m 2:0 m. In other words, the
length of the element in X-direction is 2.2 m, and the length
of the element in Y-direction is 2.0 m, and the height of the
element in Z-direction is 2.0 m. The total number of
elements of the foundation of the dam is 10 120 30
36 000: The total number of elements of the numerical
model is 624 000(dam body) 136 000(foundation of the
dam) 660 000.
The RCC dam has transverse the joints every 22 m along
axial direction (X-direction) of the dam. The boundary
condition in the X-direction is that the two end joints of
monolith (two transverse joints) are as thermal insulation
boundary. In order to reduce the calculation time and
storage space of computer, the proposed relocating mesh
method is employed. After placing 20 layers of RCC, the
bottomtenlayers are mergedintoa greater layer andthe size of
element in the Z-direction is increased from 0.3 to 3 m. Thus,
the maximum size of the mesh for the fresh concrete layers is
2.2 m 1:0 m 0:3 m, while the maximum size for the old
concrete layers is 2.2 m 1:0 m 0:3 m. By using the
relocating mesh method for the simulation of the construction
process of RCC dam, when the dam lifts to an elevation of
156 m, except the upper most 10 layers of 0.3 m The thick-
ness of the other 51 layers is 3 m. The maximum total
number of elements of the dam body is 10 1200 151
1200 73 200: The maximum total number of elements of
the numerical model is 73 200(dam body) 136 000(foun-
dation of the dam) 109 200. It means that the total
number of elements of the numerical model is reduced by
about 83%.
The rst of March of the rst year is the starting time of
simulation, and the time step used is 0.25 day. The relocat-
ing mesh method and the construction procedures are
implemented in fortran by authors. The temperature and
the stress distributions in every layer are obtained using this
fortran code.
4. Temperature distribution during construction
4.1. Placement temperature equal to the average air
temperature
During the construction of the RCC dam, the temperature
controlling measures are not used and the RCC placement
temperature is simulated as the average air temperature. The
initial temperature of the activated elements is automati-
cally set equal to the corresponding placing temperature.
For the simulation case, the placing temperature of each
layer is being concluded as equal to the ambient tempera-
ture. The initial temperature of the foundation rock is
assumed to vary linearly from the ambient temperature at
the surface to 188C at the bottom of the model.
During the construction of the dam, the temperature at the
boundaries in contact with the air is automatically set equal
to the ambient temperature at the corresponding data. To
this end, the ambient temperature is automatically adjusted
to follow the average cyclic seasonal thermal variation in
the area. In the construction period of 33 months from
March of the rst year to November of the third year, the
variations of temperature distribution at the center of dam
are shown in Fig. 3, and the highest temperature of the dam
is 36.08C.
4.2. Placement temperature controlled at 158C
From Fig. 3, it can be seen that there are three high
temperature regions in the dam, and they occur in summer.
Y. Chen et al. / Advances in Engineering Software 32 (2001) 677682 680
Fig. 3. Temperature curve at the center for RCC dam in normal placement
temperature.
Fig. 4. Temperature curve at the center for RCC dam when placement
temperature is 158C.
In order to reduce the highest temperature within the dam,
the RCC placement temperature must be controlled and
reduced. When the RCC placement temperature is controlled
at 158C, the temperature distribution at the center of the dam
is shown in Fig. 4. It can be seen from Fig. 4 that there are
three high temperature regions and two low temperature
regions. The patterns of the temperature distribution in the
dam shown in Fig. 3 are similar to those in Fig. 4. However,
the highest temperature is 29.28C in Fig. 4, which is 6.88C
lower than the highest 36.08C in Fig. 3.
5. Thermal stress distribution during construction
5.1. Thermal stresses at placement temperature equal to the
averaged air temperature
5.1.1. Maximum stress during construction
During the construction period, stresses in the dam chane
with time to time. Engineers are particularly interested in
the maximum tensile stress during construction period
[5,6].
The maximum tensile stress in the axial direction of the
dam s
Xmax
varies from 0.60 to 1.30 MPa. The maximum
tensile stress occurs around the upstream surface in January
of the second year, meanwhile, the tensile stress around the
downstream surface s
X
is 1.20 MPa, and the tensile stress at
the center of dam is the smallest. Around the height of 18 m
of the dam, the tensile stresses around the upstream surface
and downstream surface are about 1.001.30 MPa.
The tensile stresses along the upstream to downstream
direction s
Ymax
varies from 0.30 to 0.69 MPa. The maxi-
mum tensile stress (0.69 MPa) occurs in the upstream
surface at a height of 16 m of the dam in January of the
second year. The tensile stresses at the elevation points of
the upstream surface are about 0.300.69 MPa. The tensile
stresses around downstream surface at the height of 12
27 m are 0.500.69 MPa. The tensile stresses inside the
dam are 0.450.50 MPa.
The tensile stress in the vertical direction s
Zmax
varies
from 0.80 to 1.92 MPa. The maximum stress occurs around
the upstream surface in January of the second year. At the
height of 24 m, the maximum tensile stresses around the
upstream surface are about 1.501.90 MPa. The tensile
stresses around the downstream surface are about 0.80
1.32 MPa. The tensile stresses inside the dam are the
smallest. When the height of the dam is over 36 m, the
tensile stresses are lower.
Fig. 5 shows the distribution of the maximum tensile
stresses along the height of the dam.
5.1.2. Variation of stresses during construction
During the construction period the stresses of each point
varies, the variations of the stresses in some region are very
large. The variations of stress along with time for three
points at a height of 1.5 m of the dam are shown in Fig. 6.
When the construction started in March, the maximum
compressive stress around the surface at a height of 1.5 m
dam occurs in August, but the maximum tensile stress
occurs in January. The variation of the compressive stresses
is similar to the variation of air temperature. The stresses
around the center of the dam vary slightly with the change in
air temperature. The stresses around the center of the dam
are varies with the age of concrete as well. At the age of 90
days, the variation of stresses in the RCC dam is great. After
the age of 100 days, the variation of stresses in the RCC dam
becomes smaller and smaller.
5.2. Thermal stresses at placement temperature controlled
at 158C
5.2.1. Maximum stresses during construction
The maximum tensile stress in the axial direction of the
dam s
Xmax
changes from 0.40 to 1.09 MPa along with the
height of dam. The maximum tensile stress occurs around
the upstream surface in January of the second year. At the
same level and time, the stress around the downstream
surface s
X
is 0.70 MPa. The tensile stress at the center of
dam is smallest. At a height of 15 m of the dam, the tensile
stresses around the upstream and downstream surfaces are
about 0.801.09 MPa.
The tensile stresses in the stream direction s
Ymax
change
from 0.08 to 0.46 MPa along the height of dam. The maxi-
mum stress occurs in January of the second year. The tensile
Y. Chen et al. / Advances in Engineering Software 32 (2001) 677682 681
Fig. 5. Maximum tensile stress curve of RCC dam in normal placement
temperature.
Fig. 6. Stress variation curve along with time at up face of 1.5 m height of
RCC dam in normal placement temperature.
stresses around the upstream surface are smaller than
0.30 MPa. The tensile stresses around the downstreamsurface
at a height of 24 m of the dam are 0.300.46 MPa. The
tensile stresses inside the dam are 0.200.35 MPa.
The tensile stress in the vertical direction s
Zmax
changes
from 1.04 to 1.76 MPa along the height of the dam. The
maximum tensile stress occurs in January of the second
year. At a height of 24 m of the dam, the maximum tensile
stresses at the upstream surface are 1.401.76 MPa. The
tensile stresses at downstream surface are 0.801.32 MPa.
The tensile stresses inside the dam are lower. When the
height of the dam is above 36 m, the tensile stresses of
every point are lower.
Fig. 7 shows the distribution of tensile stresses along the
height of the dam.
6. Conclusions
The 3 D analysis of the temperature and the thermal stress
in RCC high dam can be carried out using microcomputer
according to the thin layer placement construction plan. If
the time for starting relocation of the mesh is suitable, the
error caused by relocating the mesh for temperature and
stress distribution is controlled within the limits permitted
by design criteria. This method, for the 3 D FEM simulation
analysis of multi-layer concrete structure at a high RCC dam
using microcomputer, is appropriate and can be applicable
in engineering design.
The simulation analysis of the temperature eld in the
dam during the construction period suggests that when the
RCC placement temperature is the average air temperature,
the highest temperature of the dam is 36.08C; when the RCC
placement temperature is controlled at 158C, the highest
temperature of the dam is 29.28C. It is concluded that
reducing the RCC placement temperature is a very useful
method for controlling the highest temperature in the dam.
By changing the starting date and construction plan, the
temperature eld in the dam is changed, then simulating
the different construction programs, a better RCC construc-
tion plan can be determined.
The simulation analysis of the thermal stress, considering
the creep of concrete, shows that the stresses around the
surfaces of the dam are greater than in other areas, and
then change with the air temperature as well. When the air
temperature increases, the compressive stresses will occur.
When the air temperature decreases, the tensile stresses will
occur. The maximum tensile stress is 1.92 MPa in this case.
The tensile stresses inside the dam are lower than 0.7 MPa
whether RCC placement temperature is the average air
temperature or RCC placement temperature is controlled
at 158C.
References
[1] Xi'an University of Technology, The thermal stresses analysis of RCC
gravity dam in construction period. Xi'an, China, Report No: 26, 1995.
[2] Xi'an University of Technology, The thermal stresses analysis of SAN
GORGE gravity dam in construction period by using viscoelastic FEM.
Xi'an, China, Report No: 75, 1989.
[3] Barrett PR, Foadian. Thermal structural analysis methods for RCC
dam, The Third ASCE specialty conference on RCC. Rcc 3. New
York: ASCE, 1992 p. 407422.
[4] Ditchey EJ, Schrader. Mocksville dam temperature studies. ICOLD
Sixteenth Congress, vol. III, Q62, June 1988; San Francisco. p. 379
96.
[5] Hirose T, Nagayama, Takemura, Soto H. A study of control of
temperature cracks in large roller compacted dams. ICOLD Sixteenth
Congress, vol. III, Q62, 1988; San Fransico. p. 11935.
[6] Jackson HE, Thermal cracking in roller compacted concrete at Gales-
ville Dam. COLD Sixteenth Congress, vol. V, 1988; San Francisco.
p. 4625.
[7] Bofang Zhu, Ping Xu. Thermal stresses and temperature control of
concrete gravity dams without longitudinal joint. Proceedings of Inter-
national Conference held at University of Dundee, September 1999;
Scotland, UK. p. 12734.
[8] Bofang Zhu, Ping Xu. New methods for thermal stresses analysis
simulating construction process of concrete dam. Proceedings of
Tenth International Conference for Numerical Methods in Thermal
Problems, July 1997; Swansea, UK. p. 74352.
Y. Chen et al. / Advances in Engineering Software 32 (2001) 677682 682
Fig. 7. Maximum tensile stress curve of RCC dam when placement
temperature is 158C.