Case Study

Modeling the Piped Water Cooling of a Concrete

Dam Using the Heat-Fluid Coupling Method
Xing-hong Liu1; Yin Duan2; Wei Zhou3; and Xiaolin Chang4
Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Thermal analysis is one of the main components in the design and construction of roller-compacted concrete dams. In this article, the
heat-fluid coupling method is introduced to perform the thermal analysis of a pipe cooling system in mass concrete structures. Several typical
models and examples are selected to verify the performance of this method. The proposed method is then applied to a roller-compacted concrete
gravity dam called the Guandi Dam, which is still under construction in China. The actual climatic conditions and thermal properties of the
materials are considered in this analysis. The temperatures determined by numerical simulation are in good agreement with the actual monitored
values. The simulation results indicate that the proposed method can accurately simulate the cooling pipe state, the temperature rise along the
water flow, and the effect of directional changes of flow in temperature distributions. Moreover, cracking is shown to more likely develop
in concrete near the cooling pipes, mainly from the extreme temperature gradient and fluctuation during the cooling process of concrete.
The untimely application of water cooling systems during hot seasons will induce extreme tensile stresses and increase the risk of cracking.
DOI: 10.1061/(ASCE)EM.1943-7889.0000532. © 2013 American Society of Civil Engineers.
CE Database subject headings: Finite element method; Coupling; Thermal factors; Cooling; Water pipelines; Concrete dams.
Author keywords: RCC dam; FEM; Heat-fluid coupling method; Thermal analysis; Cooling pipes.

Introduction measure in the construction of concrete dams, such as in the Hoover

Dam. A great deal of research has been conducted to simulate the
Controlling temperature-induced cracking in concrete is one of the cooling effect of this water flow.
main concerns in the design and construction of concrete dams. In Cervera et al. (2000a,b) presented a numerical procedure for the
roller-compacted concrete (RCC) dams, the application of contin- thermal and stress analysis of the evolutionary construction process
uous casting with thin-layer rollers combined with the large use of of RCC dams. This procedure took into account the ambient tem-
harsh concrete with less cement content during construction makes perature, placing temperature, casting schedule, and, in particular,
these dams significantly different from conventional concrete dams the more relevant features of the behavior of concrete during the
with respect to the hydration heat temperature, cooling conditions, early stages after construction, such as hydration, aging, creep, and
and the main factors affecting thermal stress. Particularly, the quick damage. The U.S. Bureau of Reclamation (USBR 1949) conducted
construction process and large size of concrete lifts will induce a research on the calculation of the final stage of water cooling and
high temperature or extreme thermal gradient within the concrete presented an analytical solution using a two-dimensional (2D)
and cause significant thermal stresses during the cooling of the dam, program, and an approximate solution using a three-dimensional
which are sufficient for cracking (Zhu 1999). Thus, it is necessary to (3D) program with no heat source by using the separation-of-
simulate and analyze the temperature and stress fields during con- variables method.
struction and then suggest some effective temperature control meas- Zhu (1999) studied the calculation of the initial stage of water
ures to prevent cracking (Tatro and Schrader 1992; Zhu 2006). cooling and obtained an analytical solution using a 2D program and
Cooling by the flow of water through an embedded cooling pipe an approximate solution using a 3D program with a heat source
has become a common and effective artificial temperature control by using the integral transform method. He noted that polythene
pipes had fewer joints, which appeared to be more convenient than
1 steel pipes in construction, and he also presented four methods for
Associate Professor, School of Civil Engineering and Architecture,
Wuhan Univ., Wuhan 430072, China. E-mail: liuxhphd@163.com computing the effect of cooling by nonmetal pipes. In addition, he
2
Doctoral Candidate, State Key Laboratory of Water Resources and simplified the analysis and proposed methods to compute the
Hydropower Engineering Science, Wuhan Univ., Wuhan 430072, China equivalent radius and equivalent horizontal spacing of nonmetal
(corresponding author). E-mail: duanyin_0224@126.com pipes. Zhu (1991) treated water cooling as negative hydration heat
3
Professor, State Key Laboratory of Water Resources and Hydropower and presented a cooling equivalent algorithm based on the FEM. He
Engineering Science, Wuhan Univ., Wuhan 430072, China. E-mail: equally distributed the water cooling effect to every element in the
zw_mxx@163.com FEM analysis, easily obtaining an approximate distribution of the
4
Professor, State Key Laboratory of Water Resources and Hydropower temperature field. This method, which requires only simple pro-
Engineering Science, Wuhan Univ., Wuhan 430072, China. E-mail: gramming and exhibits high computational efficiency, has been
changxl@whu.edu.cn
widely used in recent years. However, this method suffers from
Note. This manuscript was submitted on April 29, 2011; approved on
August 20, 2012; published online on August 29, 2012. Discussion period oversimplification and cannot accurately reflect the cooling effect,
open until February 1, 2014; separate discussions must be submitted for the cooling water velocity, or the temperature variation.
individual papers. This paper is part of the Journal of Engineering Kim et al. (2001) developed a 3D FEM program to thermally
Mechanics, Vol. 139, No. 9, September 1, 2013. ©ASCE, ISSN 0733- analyze concrete structures with pipe cooling systems. Line ele-
9399/2013/9-1278–1289/$25.00. ments were adapted to model pipes, and internal flow theory was

1278 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013

J. Eng. Mech., 2013, 139(9): 1278-1289

 applied during the calculation of the temperature variation of the Basic Formulations and Solution Approach of
cooling water. Liu (2004) proposed a general analytical model to Temperature Field
address heat extraction from mass concrete by employing a rect-
angular array of cooling pipes. In this study, an analytical solution is Based on the energy balance principle, the general partial differen-
derived in terms of the physical parameters for heat diffusion and tial equation governing heat flow in a 3D solid medium is expressed
heat removal. The model can be employed to analyze a temperature as (Wu and Ronaldo 2001; Incropera and DeWitt 2002)
field whenever heat is removed using a rectangular pipe lattice and
ax ∂ T2 þ ay ∂ T2 þ az ∂ T2 þ ∂u ¼ ∂T
can provide guidance in choosing pipe sizes, pipe spacing, aggre- 2 2 2
(1)
gates of different heat diffusivity, types of cements, fresh concrete ∂x ∂y ∂z ∂t ∂t
temperature, and admixtures for reducing the heat generation rate.
However, this method still cannot accurately reflect the velocity and where T 5 concrete temperature; ax 5 lx =cr, ay 5 ly =cr, and
water temperature variation rise along the flow. az 5 lz =cr 5 heat diffusivity coefficients; lx , ly , lz 5 heat con-
Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Myers et al. (2009) developed a model for a simplified pipe ductivity coefficients in the x-, y-, and z-directions, respectively;
network, which was assumed to consist of a series of straight pipes r 5 material density; c 5 material’s specific heat; u 5 adiabatic
for simplicity. Heat transfer occurred in the pipe walls from the temperature rise of the concrete; and t 5 time.
concrete to the water. It is found that the concrete temperature The initial transient temperature can be represented as
depends primarily on the water temperature, flux, and pipe spacing.
This simplified model highlights the dependence on certain pa- T ¼ T0 ðx, y, zÞ (2)
rameters and aids in the design of cooling systems; however, it can
only negligibly consider the combined effects of multiple pipes and
complicated boundary conditions. The two main boundary conditions at the external surfaces are
The objective of this work is to present an accurate and applicable expressed as
prediction method for the analysis of the cooling effect exhibited by
TðtÞ ¼ f ðtÞ (3)
pipes in mass concrete structures. This work accounts for the fol-
lowing factors: concrete properties, ambient conditions, placing
∂T ∂T ∂T

temperature, lift thickness, casting schedule, and, in particular, the lx lx þ ly ly þ lz lz þ R ¼ 2b T 2 Tf (4)
actual pipe cooling system. Additionally, it considers the pipe layout ∂x ∂y ∂z
(pipe sizes and spacing), the cooling water velocity, the inlet tem-
perature and water-temperature rise along the flow, and the thermal where f ðtÞ 5 known temperature values on some boundaries; R 5
properties of cooling pipes by using a numerical simulation. Thus, heat flowing from the surface; b 5 film coefficient; Tf 5 ambient
the prediction of the thermal fields associated with cooling pipes can temperature; and lx , ly , and lz 5 direction cosines of the outward
be successfully used in practice. normal to the surface.
In this study, the heat-fluid coupling method is introduced to
perform the thermal analysis of pipe cooling systems in mass Simulation of Water Cooling with the Heat-Fluid
concrete structures and is applied to an actual RCC dam. A 3D FE Coupling Method
analysis is conducted using heat-fluid elements in the FE program
ANSYS (2004). To verify the rationality and applicability of the In the entire FE model, concrete is simulated by 3D solid elements,
heat-fluid coupling method, a comparison with several widely used and cooling water pipes are simulated by heat-fluid elements, as
methods is presented. Furthermore, this method is applied to verify shown in Fig. 1. Each heat-fluid element has four nodes, consisting
and predict the distribution of temperature during the initial stage of two primary (I, J) and two supplemental (K, L) nodes. The
of the cooling process in monolith No. 13 of the Guandi Dam, and coupling of the additional nodes of the heat-fluid element and the
the calculated results are in good agreement with the actual tem- nodes of the concrete solid element are used to simulate the con-
peratures monitored within the dam. vection heat of water and concrete.

Fig. 1. Heat-fluid element geometry

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013 / 1279

J. Eng. Mech., 2013, 139(9): 1278-1289

Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Concept diagram of energy conservation principle

In numerical simulations of fluid-structure interactions, the water where Tf 5 water temperature; Ts 5 temperature of concrete near
current in cooling water pipes is considered as a one-dimensional the cooling pipe; Qw 5 inflow volume of the water per unit time;
steady flow. As shown in Fig. 2, the heat flux of flowing water cw 5 water-specific heat; s 5 direction of flow; b 5 film coefficient
per unit time is expressed as of the cooling water convection; G 5 pDL=L 5 pD boundary
surface with cooling water per unit length; L 5 length of the heat-
qs ¼ Qw rcw Tf ðsÞ fluid element; and D 5 hydraulic diameter of pipe.
(5) According to the ANSYS (2004) release 9.0 documentation, the
∂qs
qsþds ¼ qs þ ds finite-element (FE) governing equations of heat-fluid elements,
∂s which satisfy the heat conduction equation and boundary conditions,
are deduced by the variational principle in the following form:
The heat absorbed by the flowing water per unit time is
•
∂qs ∂Tf ðsÞ
q2 ¼ qsþds 2 qs ¼ ds ¼ Qw rcw ds (6) ½C fTg þ ½KfTg ¼ fQg (9)
∂s ∂s

The heat supplied to the pipe per unit time by Newton’s cooling where [K] 5 heat conduction matrix of the heat-fluid •elements; {T}
law is 5 temperature vector; [C] 5 specific heat matrix; fT g 5 variable


 temperature rate vector; and {Q} 5 node heat transfer rate vector.
qþ ¼ 2bA Tf 2 Ts ¼ 2bðpDdsÞ Tf 2 Ts (7)
Matrices of [K] and [C] are calculated in the following forms:
According to the energy conservation principle, the heat supplied
to pipe is equal to the heat absorbed by the flowing water: q1 5 q2 . 2 3
1 0 0 0
Hence, the heat exchange between the flowing water and the con- 60
crete is expressed as rcAL 6 1 0 077
½C ¼ 6 7 (10)
2 40 0 0 05
∂Tf

Qw rcw ¼ 2Gb Tf 2 Ts (8) 0 0 0 0
∂s

2 3 k1 ¼ Als =L k2 ¼ bAI k3 ¼ bA J
k 1 þ k2 2 k4 2k1 þ k4 2k2 0
6 7 0 if flow is from i to j
6 2k1 2 k5 k1 þ k3 þ k5 0 2k3 7 k4 ¼
6 7
½K ¼ 6 7 Qw rcw if flow is from j to i (11)
6 2k2 0 k2 0 7
4 5 Qw rcw if flow is from i to j
0 2k3 0 k3 k5 ¼
0 if flow is from j to i

where A 5 pD2 =4 5 cross-sectional area of the pipe; AI 5 AJ Nu ¼ aRb Pr n ¼ bd=lp (12)

5 pD=2; and ls is the thermal conductivity coefficient. The
variables of k4 and k5 represent the energy change attributable to where Nu 5 Nusselt number; R 5 Reynolds number (R 5 dvr=m);
mass transport that is lumped at the outlet node, by which we can and Pr 5 Prandtl number (Pr 5 mcw =lw ). In addition, parameter
simulate the rise in water temperature along the flow. The variables k2 b 5 film coefficient of cooling water convection (the evaluated
and k3 represent the heat convection of cooling water, and the degree variable in the formula); d 5 pipe diameter; r 5 fluid density;
of heat exchanged by cooling water convection is mainly determined v 5 flow velocity; and m is fluid viscosity.
by the film coefficient of cooling water convection b, which can be Eq. (12), by Dittus and Boelter, is widely used in projects to
defined by the Dittus-Boelter equation calculate the turbulent film coefficient of cooling water convection.

1280 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013

J. Eng. Mech., 2013, 139(9): 1278-1289

 It is applicable to the smooth tube, and we define a 5 0:023, b 5 0:8, properties that are used are shown in Fig. 3. Accordingly, the FE
and n 5 0:4, when the water is heated. The restrictions on the model for the heat-fluid coupling method is built, and the thermal
use of Eq. (12) are R 5 10,000 2 12,000, Pr 5 0:7 2 120, and L=d distribution is simulated. The computational results of the concrete
$ 60, and the difference between water temperature and pipe surface temperatures determined by the heat-fluid coupling method, as well
temperature should be no more than 20–30C. as the approximate solution and analytical solution produced using
Generally, the diameter of the cooling pipe used in mass concrete the 2D program given in Zhu (1999), are presented herein. Table 1
construction is relatively fixed. The usual outer radius and inner presents the mean temperatures for the three solutions at different
radius of pipe is 0.016 and 0.014 m, respectively. The length of the times. The maximum temperatures occurred approximately six
pipe is commonly between approximately 100 and 300 m, and the days after pouring, and the maximum values are 16.97, 16.82, and
flow velocity of the cooling water is approximately 0:54 m=s. Thus, 16.49C, respectively, which show little difference. Moreover, the
we obtain L=d 5 7,142 2 21,428, Pr 5 0:45, and R 5 11,544, result calculated by the heat-fluid coupling method is shown in
which meet the use conditions of Eq. (12). Therefore, Eq. (12) is Fig. 4, which is in good agreement with the analytical solutions.
Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

reasonable and feasible for expressing the characteristics of forced

convection in the cooling pipe.
Additionally, some level of thermal resistance, which will A Simplified 3D Pipe Network Model
somewhat reduce the cooling effect, exists in the cooling pipe. Thus, For further verification, another example of a simplified 3D pipe
it is essential to simulate the heat transfer process of the cooling pipe, network model developed by Myers et al. (2009) is studied. The
especially for plastic piping, by using the equivalent film coefficient problem configuration is shown in Fig. 5. Water flows through a pipe
of cooling water convection bs in Eq. (13), to replace b as of radius a, which is encased in a cylindrical sleeve of concrete
(radius R).
1 In the following, we will use a 5 0:016, R 5 0:845 [to meet the
bs ¼ (13)
1=b þ hp =lp practical conditions of Eq. (12)] to represent the general layout of a
cooling pipe in a concrete dam; the other parameters and material
where hp 5 thickness of water pipe and lp 5 thermal conducti- properties used in the calculation are the same as those provided by
vity coefficient of the pipe (which, for metal piping, lp 5 160 Myers et al. (2009). The comparison between the temperatures
KJ=m2 × h × C; for plastic piping, lp 5 1:66 KJ=m2 × h × C). simulated by two different methods is shown in Fig. 6. As can be
seen, the peak temperatures for the points located at r9 5 0:845 m
and r9 5 0:405 m are basically the same. Moreover, the greatest
Model Verification difference in temperature after 150 h is no more than 62C, showing
good agreement between the two methods.
A 2D Model of Mass Concrete with an Internal Source of
Heat Cooled by Nonmetal Pipes
Simulation Analysis of the Guandi RCC Dam
Before the application of this method to a practical problem, its
performance is first verified using an example featuring a 2D model The Guandi hydropower station is the third cascade power station
of mass concrete with an internal source of heat cooled by nonmetal built on the Yalong River in the Sichuan province of China. The
pipes. The concrete in this example is simplified as a hollow cylinder Guandi Dam is a RCC gravity dam with a total volume of concrete of
with outer radius b. The embedded nonmetal cooling pipe has an 3,445,000 m3 , which includes about 2,927,000 m3 of RCC. The
outer radius c and inner radius r0 . The basic parameters and material maximum height of the dam is 168 m, and the crest length is 516 m.
The dam crest elevation is 1,334 m, and the corresponding normal
water level is 1,330 m. The dam consists of 24 monoliths in which
five surface outlets and two bottom outlets are used as a combined
flood discharge. The construction of the dam began in September
2009 and completed in September 2011. During the construction,
the maximal yearly pouring intensity of RCC amounted to
1,728,900 m3 , while the peak monthly pouring intensity reached
237,260 m3 (in June 2010). To expedite the construction progress,
each dam monolith was built in several concrete blocks, with no
longitudinal joint set, and each block was built as a series of thin
compacted layers (0.3 m) with a height of 3–6 m. Moreover, a large
number of embedded cooling pipes are used to lower the temperature

Table 1. Mean Temperature for the Three Methods

Mean temperatures (C)
Time Heat-fluid Approximate Analytical
(day) coupling method solution solution
1 7.08 7.12 7.24
2 11.62 11.67 11.77
4 16.06 15.98 15.89
6 16.97 16.82 16.49
Fig. 3. Cross section of equivalent concrete cylinder cooled by water 8 16.38 16.11 15.65
flowing in pipes 10 15.17 14.76 14.16

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013 / 1281

J. Eng. Mech., 2013, 139(9): 1278-1289

Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Comparison of temperature histories for the three methods

accurately represent the practical state of the cooling pipes shown

in Fig. 9.
The incremental construction of a RCC dam can be modeled
using the birth and death algorithm in the calculation. Initially, the
FE mesh of the dam and its foundation is generated according to
the actual construction process. Then, the pouring of mass concrete
is divided into several rolling layers along the vertical direction,
which are usually approximately 0.3 m in thickness. Next, the se-
quence of the fresh roller-compacted concrete of the dam is made
active and added to the previous stage of the calculation.

Environment Variables and Material Parameters

Climatic and Hydrological Conditions at the Guandi Dam Site
The Guandi Dam site belongs to the plateau climate zone in western
Sichuan, China. Influenced by the upper level westerlies and the
southwest monsoon, the dry-wet season here is rather extreme.
The main meteorological factors are summarized in Table 2 ac-
cording to the statistical data from the weather station at the Guandi
Dam site.
Fig. 5. Problem configuration
Concrete Mixture-Proportioning
Table 3 gives the material properties for RCC and conventional
within the dam. Undoubtedly, the evolution of temperature and concrete mainly used in restrained zones of the dam foundation.
thermal stress in the dam is a complex process resulting from
multiple factors such as concrete properties, external temperature Thermal Properties of Concrete and Cooling Water
variation, placing temperature, lift thickness, casting schedule for the The thermal and structural parameters of the main concrete and rock
concrete, and the use of some temperature control measures, par- foundation are summarized in Table 4. These parameters, which are
ticularly water cooling. Thus, the simulation and prediction of the the actual values supplied by the consulting company, are used in the
temperature distribution within the dam is necessary. analysis as the input data. Table 5 gives the thermal parameters of
the cooling water.
FE Modeling
Monolith No. 13 of the Guandi Dam is selected as a typical Adiabatic Temperature Rise of Concrete
structure to conduct the analysis in this study. As it is still under The adiabatic temperature rise of concrete is given by
construction during this analysis, the analysis is performed only
up to an elevation of 1,200 m. The 3D FE mesh model for the TðtÞ ¼ T0 ð1 2 expð2mt n ÞÞ (14)
simulation is shown in Fig. 7. Eight-node isoparametric elements
were used to analyze the concrete. The spacing of the cooling pipes where T0 5 maximum temperature of concrete under adiabatic
was designed to be 1:5 3 1:5 m (horizontal and vertical). Fig. 8 conditions; t 5 concrete age in days; and m and n 5 model
shows a detailed drawing of a heat-fluid element, which can parameters that represent the heat generation rate (Table 6).

1282 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013

J. Eng. Mech., 2013, 139(9): 1278-1289

Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Comparison of simulated temperatures for the typical two points

Meanwhile, in this example, we only simulate 9 m of monolith

No. 13. However, the total element and node numbers are as large as
54,720 and 61,659, respectively, including 1,731 of pipe element
(heat-fluid element) and 3,414 of pipe node. The total simulation
time period is approximately 135 days (from December 2009 to
April 2010), and the total number of time steps is 262. The computer
run time using this method to complete the calculation of the tem-
perature field on a PC, which has a 500-GB hard drive, 2 GB of
installed RAM, and a 2.8-GHz Core 2 Duo processor, is approxi-
mately 210 min. This run time may be acceptable because for the
moment we are only simulating a few typical layers.

Model Application for a Typical Stage of No. 13 Monolith

in Hot Seasons
Fig. 7. 3D FE mesh model In the seventh stage of construction of monolith No. 13 of the
Guandi Dam, with an elevation ranging from 1,194 to 1,200 m, the
concrete pouring was carried out during the hot season, beginning
Results and Discussion on June 8, 2010. The RCC was placed in 30-cm-thick horizontal
layers, and the construction of 6-m-thick layers ended on June 14,
2010. The concrete placing temperature and inlet water temperature
Comparison and Analysis of Temperature Simulated
were determined to be 19 and 11C, respectively, and the water flow
by the Heat-Fluid Coupling Method
was determined to be 1:4 m3 =h. Additionally, the surface heat
The first few stages of construction of monolith No. 13, which were transfer coefficient considering the heat preservation was approxi-
constructed during the cold season, are studied as a practical mately 3 KJ=m2 × h × C. As the spacing of the cooling pipes was
example for further comparison. designed to be 1:5 3 1:5 m (horizontal and vertical), the cooling
According to the monitoring records, the initial stage of water pipes will be embedded during the seventh stage at elevations of
cooling was applied after the completion of each concrete stage and 1,194.6 m, 1,196.1 m, 1,197.6 m, and 1,199.1 m. The 3D FE model
lasted approximately 21 days. The average inlet water tempera- for the simulation is shown in Fig. 7.
ture was 12C, and the water flow was approximately 1:2 m3 =h. In Because of some difficulties in the construction, the initial stage
addition, the progress and some other typical monitoring records of cooling is likely to commence only after the completion of the
up to April 2010 are listed in Table 7. seventh stage (on June 14, 2010), which will last for approximately
A comparison between the simulated temperatures and the 28 days. This will exert a bad influence on temperature and
actual monitored temperatures from the thermocouples installed at crack control. Thus, it is necessary to perform the simulation
elevations of 1,167 m, 1,167.5 m, 1,170 m, and 1,172 m are shown in analysis of the temperature and stress field so as to reveal the main
Fig. 10. The thermocouple locations installed at a number of selected problems experienced during construction to prevent cracks from
levels are shown in Fig. 9. It is clear from these plots that the developing.
temperatures simulated using the heat-fluid coupling method are in In the following, the effect of different water cooling schedu-
good agreement with the monitored temperatures, which further les on the temperature distributions in concrete are investigated.
illustrates the accuracy of the heat-fluid coupling method. Initially, the cooling was scheduled to begin on June 14, 2010

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013 / 1283

J. Eng. Mech., 2013, 139(9): 1278-1289

Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. FE model of cooling pipes

(untimely application of water cooling), while the alternative was cooling effect more uniform. This oscillation nearby the import
scheduled to begin soon after the final setting of concrete at the and export of the cooling pipes is a result of the cooling system
same elevation (timely application of water cooling). being employed.
Fig. 13 shows the temperature changes for section A-A (shown in
Schedule of Initial Stage Cooling Starts on June 14, 2010 Fig. 11) after several periods of RCC placement. Compared with
For the points T1 and T2 shown in Fig. 11, Fig. 12 illustrates the other locations, the temperature of the concrete near the cooling
temperature histories. For point T1 , which is separated from the pipes is obviously lower and fluctuates more greatly. Meanwhile, the
cooling pipes, the temperature rises rapidly from the generation of temperature gradient is much greater, especially in the regions near
hydration heat within the concrete during the first 7–9 days after the water inlet or outlet.
pouring, and the maximum predicted temperature reaches ap- After the completion of the thermal analysis, the stress analysis
proximately 33.9C. Then, the temperature starts to decrease can be performed. For the present case study, the safety factor of
gradually and finally reaches approximately 26C (9C cooler than crack-resistance Kf is assumed to be 1.65. The allowable tensile
without water cooling) after 28 days. For point T2 , which is near stress [s] is calculated using Eq. (14), which depends on the RCC
the water cooling pipes at an elevation of 1,196.1 m, the tem- tensile strength Rt ðtÞ, and the dam can be considered to be safe
perature is lower and fluctuates more greatly. Indeed, the direction against cracking when the thermal stress s is not greater than the
of flow in the cooling pipes is changed every 24 h to make the allowable value [s], as is shown in

1284 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013

J. Eng. Mech., 2013, 139(9): 1278-1289

 Fig. 9. Practical disposition of cooling pipes
Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Table 2. Main Meteorological Factors at the Guandi Dam Site

Month January February March April May June July August September October November December Average
Water temperature (C) 7.2 9.6 12.3 15.7 17.6 18 18.4 18.5 16.8 15.2 11.3 8.2 14.1
Air temperature (C) 11 14.6 19.1 21.8 22.8 22.7 23 22.9 20.6 18.7 14.6 11.1 18.6
Surface temperature (C) 11.8 17.1 22 26.2 25.5 25.5 26.5 25.5 23.7 21.4 16.6 12.4 21.2
Rainfall precipitation (mm) 4.5 2.2 5.5 31.1 85.2 251.7 265.5 216.7 146.9 56.5 9.1 2.6 89.8
Relative humidity (%) 70 57 51 55 68 83 87 88 88 85 83 79 74.5
Evaporation discharge (mm) 99.4 126.4 206.7 223.2 198.7 135.6 120.2 130.8 95.7 84.8 64.6 62.5 129.1
Wind speed (m=s) 0.7 0.6 0.6 0.6 0.4 0.5 0.6 0.5 0.5 0.6 0.6 0.4 0.6

Table 3. Main Concrete Mixture-Proportioning Used in Restrained Zones Table 5. Thermal Parameters of Cooling Water
of the Dam Foundation
Heat conduction Specific
Material consumption per 1 m3 R90 25 C90 25 Density coefficient heat Viscosity
Water (Kg) 84 102 kg=m3 W=mC KJ=ðKgCÞ Pa=s
Cement (Kg) 89.4 151.9 1,000 0.161 4.187 1.1
Flyash (Kg) 89.4 65.1
Sand (Kg) 823.5 698.2
Table 6. Adiabatic Temperature Rise Parameters
Stone (Kg) 1,631.6 1,662.9
Water-reducing agent (Kg) 1.252 1.519 Concretes R90 25 C90 25
Air-entraining agent (Kg) 0.107 0.026
T0 ðCÞ 18.3 23
Water/binder ratio 0.47 0.47
m 0.136 0.202
Sand ratio (%) 34 30
n 1.017 0.9
Flyash percentage (%) 50 30

Table 7. Progress and Some Other Typical Records in Construction

Table 4. Thermal and Structural Properties Placing
Concretes R90 25 C90 25 Rock Elevation range Placement schedule temperature (C)

Density kg=m3 2,719 2,716 3,000 EL1,166–1,168 m December 12–16, 2009 17.0
Heat conduction coefficient 2.41 2.66 1.92 EL1,168–1,174 m March 12–15, 2010 16.0
ðkÞ W=mC EL1,174–1,176 m April 11–13, 2010 16.3
Specific heat ðcÞ KJ=KgC 1.01 1.10 2.3
Coefficient of linear 7.11 6.80 —
expansion 1026 =C greater than the allowable value, meaning that cracking is likely to
Poisson ratio 0.2 0.2 0.24 occur.
Thus, it can be seen that the simulation using the heat-fluid
coupling method can reflect the influence of the cooling pipe state,
s # ½s ¼ Rt ðtÞ=Kf (15) water-temperature rise along the flow, and directional changes of
flow in the temperature and thermal stress distributions during the
Fig. 14 illustrates the thermal stress histories for points T1 and T2 . cooling of the concrete. More importantly, the proposed method
Fig. 15 shows the thermal stress changes in section A-A after several reveals the great temperature gradient and extreme tensile stresses in
periods of RCC placement. It can be seen that the tensile stress rises nearby cooling pipes.
gradually during the cooling of the concrete and achieves a maxi-
mum value when the temperature no longer decreases during the Schedule of Initial Stage Cooling Starts Soon after the Final
final water cooling phase. Additionally, the thermal stress of the Setting of Concrete
concrete near the cooling pipes is clearly greater and fluctuates more To fully investigate the effect of different water cooling schedules
wildly, mainly from the extreme temperature gradient and fluctu- on the temperature distributions in the concrete, an alternative
ation. The maximum tensile stress is 1.35 MPa, which is much working schedule is studied in which the initial stage cooling begins

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013 / 1285

J. Eng. Mech., 2013, 139(9): 1278-1289

Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Comparison curve of monitoring temperature and simulated values for some selected nodes: (a) nodal w13–7 at an elevation of 1,167 m;
(b) nodal w13–8 at an elevation of 1,167.5 m; (c) nodal w13–16 at an elevation of 1,170 m; (d) nodal w13–20 at an elevation of 1,172 m

1286 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013

J. Eng. Mech., 2013, 139(9): 1278-1289

 soon after the final setting of concrete at the same elevation. Figs. 16 allowable value. Thus, the timely application of water cooling can
and 17 show the temperature and thermal stress histories, re- be considered a much safer measure against cracking.
spectively, for the two featured points. Compared with the results of Furthermore, this analysis clearly reveals the cooling effect of water
the first initial cooling schedule, the temperature and thermal stress flow on the temperature variations within the concrete, especially for
for the second are obviously smaller; indeed, the maximum tensile the first 7–9 days after pouring, which is the best time to lower the
stress in the nearby cooling pipes is only 0.9 MPa, smaller than the temperature rising and decrease the maximum temperature within the
concrete. The timely application of water cooling will more efficiently
decrease the internal temperature of the concrete as well as decrease the
temperature and thermal stress gradient near the cooling pipes.
Meanwhile, the water cooling procedure that is not well controlled will
not only generate high temperatures but will also produce a great
Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

temperature drop and gradient, which will induce great tensile stresses
within the concrete, particularly in the region near the cooling pipes. In
addition, cracking is more likely to be initiated in the concrete near
cooling pipes during the cooling process.

Conclusions

Based on the work presented here, the following conclusions can be

Fig. 11. Model of cooling pipes and characteristic point position
drawn:

Fig. 12. Temperature variation for the first stating schedule of initial stage cooling

Fig. 13. Temperature changes for section A-A after several time periods of the RCC placement

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013 / 1287

J. Eng. Mech., 2013, 139(9): 1278-1289

Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 14. Thermal stress variation for the first stating schedule of initial stage cooling

Fig. 15. Thermal stress for section A-A after several time periods of the RCC placement

Fig. 16. Temperature variation for the second stating schedule of initial stage cooling

1288 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013

J. Eng. Mech., 2013, 139(9): 1278-1289

Downloaded from ascelibrary.org by Indian Institute of Technology, Delhi on 07/20/17. Copyright ASCE. For personal use only; all rights reserved.

Fig. 17. Thermal stress variation for the second stating schedule of initial stage cooling

1. The heat-fluid coupling method is introduced to perform the References

thermal analysis of a pipe cooling system in mass concrete
structures. The numerical results were compared directly to ANSYS. (2004). ANSYS rel. 9.0 documentation, ANSYS, Inc., Canons-
two theoretical models. It is found that the proposed method is burg, PA.
highly accurate. Cervera, M., Oliver, J., and Prato, T. (2000a). “Simulation of construction
2. The simulation of the temperature field by the heat-fluid of RCC dams. I: Temperature and aging.” J. Struct. Eng., 126(9),
coupling method can reflect the influence of the cooling pipes, 1053–1061.
water-temperature rise along the flow, and directional changes Cervera, M., Oliver, J., and Prato, T. (2000b). “Simulation of construction of
in flow on temperature distributions. Moreover, the computa- RCC dams. II: Stress and damage.” J. Struct. Eng., 126(9), 1062–1069.
Incropera, F. P., and DeWitt, D. P. (2002). “Transient conduction.”
tional results are in good agreement with the monitored tem-
Chapter 5, Fundamentals of heat and mass transfer, 5th Ed., Wiley,
peratures, which shows that this method is generally effective New York, 289–296.
and is excellent for temperature simulation and forecasting. Kim, J. K., Kim, H. K., and Yang, J. K. (2001). “Thermal analysis of
3. During the concrete’s cooling process, the temperature of the hydration heat in concrete structures with pipe-cooling system.” Comput.
concrete near the cooling pipes is obviously lower and fluc- Struct.., 79(2), 163–171.
tuates more wildly; the temperature gradient is also much Liu, C. (2004). “Temperature field of mass concrete in a pipe lattice.”
greater. Accordingly, the thermal stress of concrete near the J. Mater. Civ. Eng., 16(5), 427–432.
cooling pipes is obviously greater and fluctuates more wildly, Myers, T. G., Fowkes, N. D., and Ballim, Y. (2009). “Modeling the cooling
which makes cracking more likely to be initiated in the of concrete by piped water.” J. Eng. Mech., 135(12), 1375–1383.
concrete near the cooling pipes during this time. Tatro, S., and Schrader, E. (1992). “Thermal analysis for RCC—a practical
4. The simulation of construction of the seventh stage of mono- approach.” Proc., Roller Compacted Concrete III, ASCE, Reston, VA,
389–406.
lith No. 13 of the Guandi Dam during hot seasons suggests
U.S. Bureau of Reclamation (USBR). (1949). “Cooling of concrete dam.”
that approximately the first nine days after pouring represent
U.S. Dept. of the Interior, Washington, DC.
the best time to lower the rise in temperature and decrease the Wu, Y., and Ronaldo, L. (2001). “Numerical implementation of tem-
maximum temperature within RCC. perature and creep in mass concrete.” Finite Elem. Anal. Des., 37(2),
5. Finally, this method also has certain deficiencies. Though it can 97–106.
simulate cooling pipes well, the construction of the FE model is Zhang, X.-F., Li, S.-Y., Chen, Y.-L., and Chai, J.-R. (2009). “The de-
still complex, and the total number of elements and nodes is velopment and verification of relocating mesh method for the compu-
difficult to control, which produces great computational costs. tation of temperature field of RCC dam.” Adv. Eng. Softw., 40(11),
However, the computation efficiency can be improved by 1119–1123.
adopting methods such as the relocating method (Zhang et al. Zhu, B.-F. (1991). “Equivalent equation of heat conduction in mass
2009) or multiscale method, which will be further researched. concrete considering the effect of pipe cooling.” J. Hydraul. Eng., (3),
28–34.
Zhu, B.-F. (1999). “Effect of cooling by water flowing in non-
Acknowledgments metal pipes embedded in mass concrete.” J. Constr. Eng. M., 125(1),
61–68.
This work was sponsored by the National Science Fund for Dis- Zhu, B.-F. (2006). “Current situation and prospect of temperature control
tinguished Young Scholars of China (No. 50725931) and the Na- and cracking prevention technology for concrete dam.” J. Hydraul. Eng.,
tional Natural Science Foundation of China (No. 50909078). 37(12), 1424–1432.

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2013 / 1289

J. Eng. Mech., 2013, 139(9): 1278-1289