1 5081654 PDF
1 5081654 PDF
1 5081654 PDF
Patrik Růžička
CFD analysis of water cooling heat exchanger for storage energy from exhaust pipe
AIP Conference Proceedings 2047, 020017 (2018); https://doi.org/10.1063/1.5081650
© 2018 Author(s).
Modeling of Boundary Layer and the Influence on Heat
Transfer with Help of CFD
1, a)
Patrik 5ĤåLþND
3RãWRYVNi.\ãLFH&]HFK5HSXEOLF
a)
ruzickapatrik94@gmail.com
Abstract. This work deals with different models of a boundary layer of fluid flow used in numerical simulation in
computational environment ANSYS. With the help of Computational Fluid Dynamics (CFD) was simulated the flow of
the fluid and was monitoring the influence of the boundary layers on heat transfer and pressure drop. The basics of the
boundary layer flow near a smooth wall, heat transfer and numerical simulation were described. Special attention was
focused on possibilities of creation layers in computational grid.
The analytical and numerical calculations of pressure losses and heat transfer coefficient were performed on straight
circular pipeline with different types of boundary layers. The results were compared with each other, then between two
different types of used grid, and evaluated, where the analytical results were set as referential values.
INTRODUCTION
The World is full of nature phenomena that could be considered unreasonable or illogical. Whether it is a flow of
the water in a river or a movement of the planet in space. Firstly, in ancient Greece, researches and scientists tried to
understand and describe the laws of this natural phenomena and a new science called physics emerged. In the
beginning it contained mainly reflections and observations of the surrounding world. Over the time, science itself
developed and became more complex, allowing physical phenomena not only to observe and to contemplate them,
but also to predict. For the description and the prediction were defined basic forms. Consequently, were compiled
mathematical equations and reasoning, which describe the behavior of different substances in different forms. Due
to the complexity of the mathematical equations describing fluid flow and heat transfer, earlier solutions had been
simplified for analytical calculation, or an experiment had been used to achieve results.
In the end of the twentieth century were numerical simulations increasingly used with the approach of IT. From
all of these three options (analytical calculation, experiment, numerical simulation), the analytical calculation is the
fastest but for more accurate results in short time is the numerical simulations the best possibility. However, it is
adequate to combine at least two methods for verifying the results. In recent years seems that the results of the
numerical simulation of specific phenomena are very accurate, almost equivalent to the reality in some cases, but it
is still only a description or a function describing the process of given phenomenon. The aim of this work is to
evaluate numerical simulations depending on the way of defining the boundary layer of the flowing liquid and the
choice of mathematical equations for calculation in the environment of ANSYS FLUENT Inc.
020021-1
TYPES OF FLOW
There are several physical and kinematic properties, based on which can be fluid flow divided. For viscous
liquids are used two types of flow. Laminar and Turbulent flow. Laminar flow is associated with low velocities.
Viscous forces are dominant and the flow has characteristic steady movement. The movement can be see like
infinitely thin layers, that shifting on each other and between the layers is formed frictional force. Turbulent flow, on
the other hand is associated with higher flow velocities and interrail forces are dominant in this case. It results in
random motion in all directions. This leads to swirling and other instabilities in the flow.
The type of flow can be set with help of Reynolds number, which expresses relation between inertial and viscous
forces. (1)
HYDRAULIC LOSSES
Important fact about fluid flow are hydraulic friction losses. In a circular pipe are fluid properties (viscosity,
density), flow viscosity and piping properties (wall roughness, length, pipe diameter) main factors influencing the
friction losses. In ideal case are these variables constant. The roughness of the walls is not influenced by a corrosion,
piping diameter by a sedimentation etc. There are many types of equations for different situations. For circular pipe
is most used the Weisbach equation. (1)
݈ ݓଶ
߂ ൌ ߣ௭
݀ ʹ
w - flow velocity, l - pipe length, d - pipe diameter, ߣ௭ - friction factor
The friction factor depends on Reynolds number. For turbulent flow in circular pipe:
Ͳǡ͵ͳ
ߣ௭ ൌ ర
ξܴ݁
For the solution is needed to know the temperature and pipe dimension. In the case of horizontal pipes with
turbulent flow, the mean fluid temperature is used. At the same time, the wall temperature must not exceed the
boiling point of the fluid due to evaporation of liquid in the boundary layer and generation of thermal resistance. The
equation for surface heat transfer coefficient was defined as: (3)
ሺߩݓሻǡ଼
ןൌ ܤ
݀ ǡଶ
w - IORZYHORFLW\ȡ± fluid density, d - pipe diameter, B ± fluid coefficient
The fluid coefficient B is known from experiments and depends on type of fluid and temperature and for most
fluids can be found in tables. The computational method of heat transfer coefficient in Ansys is completed described
in ANSYS Fluent Theory Guide (5).
OUTLET TEPERATURE
For analytical computation of outlet temperature is needed at first the heat flow, which can be achieved with use
of the Biot-Fourier law.
ܳ ൌ ߙ൫ܶ௪ െ ܶ ൯ܣ
Q ± heat flow, Tw ± wall temperature, Tf ± fluid temperature, A ± area
020021-2
The temperature difference can be achieved with use of the modified calorimetry equation.
ܳ
οܶ ൌ
ܿ
Q ± heat flow, c ± heat capacity
The computational method of outlet temperature in Ansys is completed described in ANSYS Fluent Theory
Guide (5).
BOUNDARY LAYER
The term boundary layer was introduced for a thin layer of viscous liquid close to the surface of the solid. Within
the boundary layer, the velocity of the flow moves from zero at the surface, where the liquid adheres to the wall due
to viscosity, to velocity at the edge of the boundary layer corresponding to the speed of the uninfluenced liquid
stream. The value of the thickness of the boundary layer depends on the frictional stress, expressing the interaction
between the molecules of the liquid and the solid. Maximum value of the frictional stress is on the surface of solid
and then decreases to zero at the infinite distance from the solid. The basic idea of the boundary layer was
introduced by Ludwig Prandtl. It talks about the boundary layer as a layer formed in stream, which has a relatively
low viscosity compared to the internal forces. FIGURE 1 illustrates velocity values (wf - of free stream, w ± local
YHORFLW\DQGVKHDUVWUHVVIJ:- DWWKHZDOOIJ- local shear stress). (2)
ENTRANCE REGION
In a small area in a pipe, where fluid enters, the flow is not fully developed because of the consecutive
development of the boundary. This area is called entrance region and has influence on the average flow velocity.
FIGURE 2 shows a pipe with the entering region and region with fully developed flow. (3)
020021-3
CFD
Computational Fluid Dynamics, known as CFD, represents a calculation of dynamic flows using a computer. For
the first time, the concept of CFD emerged in an academic research. In the beginning the practical use has been
found in astronautics and has subsequently expanded to complex problems in engineering practice in general. Fluid
flow can be described by a partial differential equation, which cannot be solved analytically except for cases with
some simplistic assumptions. In order to get the results, it is necessary to use discretization methods which specify
the differential equations using the system of algebraic equations. Than it can be solved using a computer.
Discretization is applied to the small areas in space and time. The resulting values, as in experiments depend on the
quality of the used tools, the quality of numerical calculations depend on the used discretization method. The basics
of a turbulence are described by Navier-Stokes equation. However, this equation cannot be used due to the
fluctuation of velocities over a wide period of time and space in direct numerical simulation, where it will take
decades to get the results with current available technology. For this reason, the equations have been simplified for a
certain spectrum of the turbulent flow. The most used simplification is time-centric (which, for practical reasons,
straightens the equation over time), resulting in Reynolds's Navier-Stokes (RANS) equations. In this way,
fluctuations from stream are eliminated and results in average values of velocity and pressure fields. However, the
process introduces other unknown shear stress which must be supplemented by an appropriate turbulence equation.
The quality of the numerical simulation depends primarily on the choice of these turbulent models and it is therefore
important to select the corresponding model as well as the corresponding computational grid for different cases.
ANSYS Fluent provides a choice between these turbulence models. Because of the complexity of turbulent models
are introduced only the models¶ titles, which are Realizable k-İDQG667N-ȦPRGHO
Relizable k ± İPRGHO
Two-HTXDWLRQ NLQHWLF HQHUJ\ N DQG LWV GLVVLSDWLRQ İ WXUEXOHQFH PRGHOV DOORZ WKH GHWHUPLQation of both, a
turbulent length and time scale by solving two separate transport equations. The term ³realizable´ means that the
model satisfies certain mathematical constraints on the Reynolds stresses, consistent with the physics of turbulent
flows. Robustness, economy, and reasonable accuracy for a wide range of turbulent flows explain its popularity in
industrial flow and heat transfer simulations. It is a semi-empirical model, and the derivation of the model equations
relies on phenomenological considerations and empiricism. (5)
SST k ± ȦPRGHO
The SST - model includes transport equations for the turbulence kinetic energy k and the specific dissipation rate
ȦDQGLQDGGLWLRQDFFRXQWVIRUWKHWUDQVSRUWRIWKHWXUEXOHQFHVKHDUVWUHVVLQWKHdefinition of the turbulent viscosity.
The use of a k-ȦIRUPXODWLRQLQWKHLQQHUSDUWVRIWKHERXQGDU\OD\HUPDNHVWKHPRGHOGLUHFWO\XVDEOHDOOWKHZD\
down to the wall through the viscous sub-layer, so it switches to a k-İEHKDYLRULQWKHIUHH-stream and thereby avoids
the common k-ȦSUREOHPWKDWWKHPRGHOLVWRRVHQVLWLYH
020021-4
SOLUTION
The experiment was a creation of a different types of the boundary layer near to the pipe walls on the specified
geometry using Total Thickness and Smooth Transition. Beside that the influence on total pressure drop and heat
transfer. Based on the theoretical knowledge was expected the turbulent flow with the given flow velocity, the
dimension of the pipe and the viscosity of flowing fluid. Due to the heating of the liquid from the warmer pipe wall
was expected an increase in the values of the velocity in the boundary layer. For numerical simulation was
considered a stationary flow without the influence of a weight of the fluid and also a constant wall temperature. The
values for calculating the pressure drop and the heat transfer are shown in TABLE 1.
Value Units
:DWHU )ORZLQJPHGLXP
ݓ ͳ ሾ݉ ή ି ݏଵ ሿ )ORZYHORFLW\
ܶ ʹͻ͵ǡͳͷ ሾܭሿ )OXLGWHPSHUDWXUH
ߩ ͻͻͺǡʹ ሾ݇݃ ή ݉ଷ ሿ )OXLGGHQVLW\
ߥ ͳǡͲͲͶ ή ͳͲି ሾ݉ଶ ή ି ݏଵ ሿ .LQHPDWLFYLVFRVLW\
ߟ ͳǡͲͲʹ ή ͳͲିଷ ሾܲܽ ή ݏሿ '\QDPLFYLVFRVLW\
݀ ͲǡͲͷ ሾ݉ሿ 3LSHGLDPHWHU
ܮ ʹ ሾ݉ሿ 3LSHOHQJWK
ܶௐ ͵ͷ͵ǡͳͷ ሾܭሿ 3LSHWHPSHUDWXUH
்ெ ൌ ͳͲͳ͵ʹͷ ሾܲܽሿ 3UHVVXUH
ߣ Ͳǡ ሾܹ ή ݉ିଵ ή ି ܭଵ ሿ +HDWWUDQVIHUFRHIILFLHQW
Results
For the high number of results and figures were chosen three simulations. All simulations can be found in a
thesis (6).
Numerical Computation 1
For the first calculation the "Smooth Transition" function with the number of layers 10 was used. Number of
computational cells 1 240 275, maximum cell size 3 mm. The grid is shown in FIGURE 4. The calculation time was
approximately 30 min.
020021-5
Average Results
TABLE 2 Results 1
Graphical results
FIGURE 5 (a) shows the pressure dependence on the pipe length. A linear pressure drop is apparent. The
deviation at the beginning of the pipe causes non-developed flow. The value of the heat transfer coefficient
FIGURE 5 (b) stabilized within the range of 3000-4000 ሾܹ ή ݉ିଶ ή ି ܭଵ ሿ at a distance of about 400 mm from the
input.
(a) (b)
FIGURE 5 Computation 1 Pressure along the pipe (a). Heat transfer coefficient along the pipe (b)
FIGURE 6 shows contours of temperature along the pipe. The outlet area of the pipe is shown in a cross-section
parallel to the direction of the flow and the cross-section along the right is normal to the direction of the flow in the
outlet. The temperature change is gradual throughout the tube. A more visible temperature gradient is located in the
area near the wall. The area with unchanged temperature in the center of the pipe almost disappeared.
FIGURE 7 shows the vector velocity field in the end region of the pipe. A wall area of up to 10 mm was
selected in the direction of the pipe axis. There is a decrease in the velocity near the wall. Moving away from the
wall, the speed stabilizes at a certain value.
020021-6
FIGURE 7 Computation 1 velocity vectors near the wall
Numerical Computation 2
For the second calculation the "Smooth Transition" function with the number of layers 15 was used. Number of
computational cells 1 634 933, maximum cell size 3 mm. The grid is shown in FIGURE 8. The calculation time was
approximately 35 min.
Average results
TABLE 3 Results 2
Graphical results
FIGURE 9 (a) shows the pressure dependence on the pipe length. A linear pressure drop is apparent. The
deviation at the beginning of the pipe causes non-developed flow. The heat transfer coefficient FIGURE 9 (b)
shows a very rapid approximation of the resulting value at a distance of about 300 mm from the inlet where it
stabilized in the range of 3000-4000 ሾܹ ή ݉ିଶ ή ି ܭଵ ሿ.
020021-7
(a) (b)
FIGURE 9 Computation 2 Pressure along the pipe (a). Heat transfer coefficient along the pipe (b)
FIGURE 10 shows contours of temperature along the pipe. The outlet area of the pipe is shown in a cross-
section parallel to the direction of the flow and the cross-section along the right is normal to the direction of the flow
in the outlet. Covering the boundary layer area with a higher number of grid layers has a noticeable influence. In
comparation to the numerical calculation 1, there is unaffected larger area of the flow around the pipe axis.
Subsequently, a larger temperature gradient is visible near the wall.
FIGURE 11 shows the vector velocity field in the end region of the pipe. A wall area of up to 10 mm was
selected in the direction of the pipe axis. Higher numbers of computational grid layers in the area near the wall give
a better idea of the speed profile. There is a clear decrease in the speed towards the wall.
020021-8
Numerical Computation 5
For the fifth calculation the "First Layer Thickness" function with the number of layers 1 was used. Number of
computational cells 221 091, maximum cell size 5 mm. The grid is shown in FIGURE 12. The calculation time was
approximately 20 min.
Average results
TABLE 4 Results 5
Graphical results
FIGURE 13 (a) shows the pressure dependence on the pipe length. A linear pressure drop is apparent. The heat
transfer coefficient FIGURE 13 (b) converges to its value slowly, due to the insufficient effect of the boundary
layer. The deduced value determined by the average differs from the previous cases by at least 150ሾܹ ή ݉ିଶ ή ି ܭଵ ሿ.
(a) (b)
FIGURE 13 Computation 5 Pressure along the pipe (a). Heat transfer coefficient along the pipe (b)
FIGURE 14 shows contours of temperature along the pipe. The outlet area of the pipe is shown in a cross-
section parallel to the direction of the flow and the cross-section along the right is normal to the direction of the flow
in the outlet. Due to the insufficient coverage of the boundary layer by the cells of the computational grid, the area
020021-9
near the wall with a temperature almost identical to the wall temperature increased considerably. This also affected
the average value of the outlet temperature from the pipe.
FIGURE 15 shows the vector velocity field in the end region of the pipe. A wall area of up to 10 mm was
selected in the direction of the pipe axis. Insufficient computational grid results in an inadequate idea of the rate of
velocity in the boundary layer and its effect on the total flow.
020021-10
Results of the analytical method were taken as the reference values for numerical computation.
TABLE 5 Results
ER (,,Entering Region³), FDR (,,Fully Developed Region³) region with fully developed flow. For computation 1-5 was used model ܵܵܶ݇ െ ߱. For 6-10
model ܴ݈݁݅ ݈ܾ݇݁ܽݖെ ߝ.
020021-11
CONCLUSION
As it can be seen the maximum deviation of the pressure drop values with application of k-İPRGHO2QWKHRWKHU
hand, the smallest deviation was for the same computational model with more grid layers in the area of boundary
layer. For Realizable k- İWKHYDOXHFRUUHVSRQGV rather to the heat transfer coefficient for the non-developed flow
region. Similar results were obtained with SST k-ȦDSSOLFDWLRQ2YHUDOOWKHKHDWWUDQVIHUFRHIILFLHQWVFRUUHVSRQGHG
rather to a fully developed flow.
When using First Layer Thickness function, the heat transfer coefficient was stabilized in the first quarter of the
pipe length in the range of 100-200 mm. When using Smooth Transition, the values have stabilized around their
average value at a distance of about 300-500 mm. The exception was the use of the First Layer Thickness function
with insufficient number of computational layers in the boundary layer where the influence of the liquid was not
sufficiently documented. The value of the heat transfer coefficient approximated to its diameter at a distance of
approximately 600-800 mm from the pipe inlet and continued decreasing.
More obvious differences were in the distribution of the flow temperature at the end of the pipe. In case of a
higher number of computational layers, the area of the heat-affected stream around the pipe axis increased. When
comparing these two computational models, the pressure drop was smaller when using SST k-Ȧ,QWKLVFDVHWKH
value of the heat transfer coefficient was close to the value for the fully developed flow area, while the Realizable k-
İYDOXHZDVUDWKHUIRUXQGHYHORSHGDUHD7KHUHIRUHWKH667N-ȦLVPRUHVXLWDEOHIRUXVHLQDIXOO\GHYHORSHGIORZ
area. The Realizable k-e model achieves more accurate values compare to the analytical solutions.
In the end, the real experiment will have the most accurate values and the values from analytical and
computational methods are mostly used for visualization. Complete work will have all of that three methods:
analytical computation, numerical simulation and experiment.
REFERENCES
1. --(ä(.%9È5$',29ÈD-$'$0(&0HFKDQLNDWHNXWLQ9\GSĜHSUDFRYDQp 3UDKDýHVNp
Y\VRNpXþHQtWHFKQLFNp,6%1-01-01615-3.
2. R. HEJZLAR, Mechanika tekutin 3UDKDýHVNiWHFKQLND- QDNODGDWHOVWYtý987,6%1-80-
7043-766-7.
3. C. KOTHANDARAMAN, Fundamentals of Heat and Mass Transfer 3rd ed. (New Delhi: New
Age International Pvt., 2006) ISBN 978-812-2426-427.
4. -+)(5=,*(5D03(5,ûComputational methods for fluid dynamics (New York: Springer, 2002) ISBN
35-404-2074-6.
5. ÄANSYS Fluent Theory Guide,³ 20 4 2017. [Online]. Available: https://uiuc-
cse.github.io/me498cm-fa15/lessons/fluent/refs/ANSYS%20Fluent%20Theory%20Guide.pdf.
6. 35ģä,ý.$ÄPortal ZCU,³ 2017. [Online]. Available:
https://portal.zcu.cz/StagPortletsJSR168/PagesDispatcherServlet?pp_destElement=%23ssSouboryStudentuDi
vId_41&pp_locale=cs&pp_reqType=render&pp_portlet=souboryStudentuPagesPortlet&pp_page=souborySt
udentuDownloadPage&pp_nameSpace=G530268&soubidno=110375 .
020021-12