final thesis
TRANSCRIPT
CHAPTER 1: INTRODUCTION
1.1 Background of Research
The rate of computer system development is hampered due to the limitation of
current cooling system. Researches have been done to find a more efficient cooling
technology to address the thermal management issues. Forced-air cooling using the
traditional fan sink system will continue to be used for electronics cooling because of its
cost, reliability, and its familiarity to the design engineer as for now. However, given
the current and futuristic dissipation trends in chip design, it is evident that hybrid
cooling systems, containing both traditional forced air cooling and an advanced cooling
system that enables the local removal of high heat fluxes, will be the practical solution
in thermal management [1].
In this study, water cooling through serpentine minichannel is suggested as an
option to address this problem. Direct cooling through minichannel enables heat
removal directly from a chip surface more efficiently and effectively. Serpentine
minichannel offers higher heat transfer rate due to its large ratio of heat transfer surface
to fluid flow volume and augmentation of single-phase convective heat transfer through
series of right-angle turns. The cooling is further increased as the lower temperature at
the heat sinks allows more heats to be dissipated from the microprocessor unit.
.
1.2 Problem Statement
The rapid development in electronic components to provide higher
computational speed, functional density and quality by exponentially increasing the
transistors in micro-chips has led to the increase in the microchip heat fluxes. The
1
current method of passive cooling by air has almost reached its limit about 100W/cm2.
To obtain the optimum performance, maintaining the functional temperature is very
critical. Thus, it became a rule of thumb among the computer system designers to
assume that the increasing temperature of 10oC can double the failure rate of an
electronic component [2].
1.3 Limitation of Study
A mini-channel heat sink with bottom cross section size of 20mm X 20mm is
chosen for analyzing single-phase laminar flow of water as coolant through serpentine
mini-channel. Bottom cross section size of 20mm X 20mm is chosen based on reference
to X.L.Xie's work [3] which is based on water cooled mini-channel heat sinks.
The range of channel dimensions is limited from 0.5 mm to 2 mm in accordance
to the higher order of minichannels for ease of fabrication. The inlet velocity is
introduced in order to maintain the laminar flow. Inlet velocities ranging from 0.1 m/s
to 0.4 m/s are specified. The effects of design geometry parameters on pressure drop
and total surface heat flux are studied. Design of Experiment Method Orthogonal Array
test run is used in the analysis to reduce the number of overall design sample from sixty
four to sixteen without sacrificing the validity of the data and its results in achieving the
optimized design. Heat flux of 100W/cm2 is assumed to be released by processor as it is
the common parameter used in research [3,4]. To reduce time and complexity of
processing, only channel is analyzed. All the analyses were done by using
Computational Fluid Dynamic software of FLUENT.
2
1.4 Research Objectives
To determine the nearly optimized heat sink model for laminar flow which gives high
heat transfer performance with acceptable range of pressure drop by numerically
analyzing the design geometry and inlet velocity parameters using Computational Fluid
Dynamics Method through Design of Experiment Approach .
1.5 Research Methodology and Work Plan
Research methodologies of this work are as follows:
Stage 1: Literature Review
Past research works related to the present work reviewed .The current issues and future
need of thermal management in chips are studied. Water cooling system is one of the
problems in the area of the research. Serpentine minichannel is addressed as a promising
cooling solution. Analysis method and techniques of CFD and sampling method of
Design of Experiment (DOE) is reviewed. The sources for the literature review include
books, journals, creditable online resources and previous thesis work.
Stage 2: Setting Design Parameters
Design geometry parameters which influence the heat transfer rate and pressure drop are
identified and its range is set for heat sink with bottom dimension 20mm X 20mm. Inlet
velocity is calculated using Excel programming on the entire possible design parameters
to include only fully developed laminar flow range in the analysis.
3
Stage 3: Analysis
Material properties, boundary condition, solver formulation and basic equation specified
for the samples in Fluent. Analysis is iterated manually till convergence of residual
achieved. For more accurate results, the results of outflow pressure and total surface
heat flux are iterated till the results converged.
Stage 4: Verification of Analysis
Grid independence studies were done to verify the results of the analysis. Each design
sample is analyzed and remeshed using smaller meshing size. The resulted values from
subsequent meshing size are compared to find the results which are independent from
the grid applied.
1.6 Contribution of Study
This research would serve as an option for the search of a light miniature heat
sink which is able to withstand the increasing amount of heat flux produced by
processor. The design parameters which affect the performance of the serpentine
minichannel heat sink in a fully developed laminar flow is detailed and analyzed to find
the nearly optimized design. This study can be used as a reference for further
development of serpentine minichannel heat sink. The model developed and its design
parameters can be further studied to include turbulent flow by increasing the velocity
flow, to find the optimum serpentine design parameters for the suggested heat sink
dimensions.
4
1.7 Organization of the Dissertation
The research report was divided into sections as below.
a) Introduction
This chapter presents the importance of the study, problem statement and the
objective of the research stated. The research scope and the limitation and
contribution of the study are briefly stated.
b) Literature review
This chapter presents the current issues and future need of thermal management
in electronic industry of utilization of chips. Researches on water cooling system
and minichannel cooling solutions were reviewed. Advantages of adapting
serpentine channels in minichannel were stated. Utilization of CFD in solving
relevant problems of cooling system were also reviewed.
c) Analysis Using Computational Fluid Dynamics
This chapter presents detailed tasks in accomplishing the research. The tasks
includes modelling, meshing, applying boundary conditions and solving using
commercial CFD software of FLUENT.
d) Results and Discussion
This chapter presents the results of the simulations using CFD. Effects of
design parameters and the inlet velocity on total surface heat flux and pressure
drop are discussed
5
e) Conclusion and Recommendation
This chapter summarizes the important finding of work and further studies of
work are presented.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
6
16.
17.
18.
CHAPTER 2: LITERATURE REVIEW
2.1 Thermal Management in Industry
Moore’s Law states that every 12-18 months the transistors will be double in
microprocessors as the number of components on a chip grow annually with a factor 1.5
– 2 5. The growth of the electronics industries contributes to the need of the increasing
packaging capacity of microprocessors as predicted by Moore. However, the industries
are incapable of suspending the laws of physics whereby higher computing performance
is accompanied by more heat generation. As a result , thermal management is becoming
increasingly critical to the electronics industry 6.
7
Figure 2.1: Number of Transistors in a Chip
(Source: home.fnal.gov)
2.1.1 Current Issues and Future Needs
Increase in chips area density is the key to further improve processors
capability. When chip size remains constant, the increasing number of transistors
enables the microprocessor to move to the next generation of performance. The size of a
chip does not significantly increase during the recent years. At present the chip is about
1 cm2. The bigger surface area would lead to possibility of containing an undesired
impurity or a defecr. Larger chips would therefore cause higher rejection rate. A bigger
packing density is much desirable as it may improve performance of the system by
smaller transistors, shorter on-chip interconnections, and less inter-chip connections.
The costing impact will also be lower.
8
From 1971 to 2007, development of using transistors on a chip has massively
increased. By the end of the decade, it is estimated that a square centimeter of
microprocessors could produce an amount of heat roughly equivalent to 1,000 degrees
Fahrenheit[7]. Most of the conventional computer cooling system works by attaching a
heat sink unit on the spot that has high power density. The heat sink is usually cooled by
an axial fan. Air is forced to blow on it and the heat is removed through conduction and
convection.
Initially, passive cooling was enough to keep CPUs running in a stable
condition. But as development of chips evolved, to have more power density increases,
the heat sink size and airflow become more constrained. Unfortunately, development of
the cooling technology does not scale exponentially. As a result, processors went from
no heat sinks needed in the 1980s, to moderate-size heat sinks in the 1990s, to today’s
huge heat sinks, need more dedicated fans to increase airflow over the processor [8].
The size increment of heat sink is also non proportional with its performance
due to spreading resistance. For an example, as the length of aluminium heat sink
increases from 50mm to 100mm, the weight will double from 133g to 266g. However,
the improvement of the performance is only around 16%. This method is no longer
adequate for the recent development of chips [9]. Moreover, the usable space for the
finned heat sink remains limited 10.
A microprocessor working at its best performance needs to be at a lower
maximum temperature. In normal operation, it is good to keep the temperature to be half
of the specified maximum temperature. The functional temperature limit is specified in
accordance to the performance requirements. Operation exceeding the functional
temperature limit can degrade the system performance or cause unexpected failures.
9
The absolute maximum temperature limit is the highest temperature at which a
portion of the component can be safely exposed. Temperatures exceeding the limit can
cause physical destruction or may result in irreversible changes in operating
characteristics. Higher temperatures result in premature failure of the devices in the
system.
The latest 0.18 micron technology of Intel for miniaturization combined with
well known MOS (Metal Oxide Semiconductor), and VLSI (Very Large Scale
Integration) technologies for packaging an enormous number of components on single
chip has led to an important issue on heat dissipation consequently it leads to higher
mean operating temperatures, localized hot spots, and adverse thermal gradient. The
thermal problem caused by the increasing heat dissipation is as a result of the trend of
miniaturization of modern electronics [11].
Some industry analysts were predicting that effective thermal solutions will
become a major constraint for the reduction of cost and time-to-market 12. With all the
issues faced with a conventional heat sink, a more effective computer cooling system
must be well addressed in future.
2.2 Cooling System
To ensure that the processor is maintained within functional and absolute
maximum limits the industry depends on functional cooling techniques to help cool
their rapidly advancing chips. Without proper cooling, performance and power will be
sacrificed for lower temperatures and stability, thus inhibiting the development of even
higher speed chips. Ineffective cooling could lead to overstress in electrical
components, causing the computer to fail prematurely, typically at the spot where the
10
heat is dense. Though forced-air cooling using the traditional fan sink system will
continue to be a work horse for electronics cooling because of its cost, reliability, and its
familiarity to the design engineer. However, given the current and futuristic dissipation
trends in chip design, it is evident that hybrid cooling systems, containing both
traditional forced air cooling and an advanced cooling system that enables the local
removal of high heat fluxes, will be the practical solution in thermal management.
Indirect cooling methods in the cooling system, although can remove fairly high
heat flux but it poses the difficulty of integrating them with the main systems. Incropera
13], noted that “the most fruitful approach to enhance the performance of cooling
technologies is likely to be one which reduces the thermal path between the electronic
packages and the cooling fluid”. For this reason, Mallik et al 14 stated that the direct
cooling strategies may present the best alternative solution.
An overview of leading-edge advanced cooling systems are as follows7.
2.2.1 Active Heat Sinks
Active heat sinks are the solution to minimize the ducting and leakage problems present
in forced convection air cooling. By making the heat sink an integral part with the fan,
leakage is non-existent. However, care must be taken in the design of active heat sinks
since the performance of the fan is now affected by the presence of the heat sink
attached to it and how the active heat sink is located within the global system. It may
not be appropriate to use a lumped analysis model based on the known fan curve to
represent the fan within the active heat sink. Actual non-uniform flow into the fan and
the nature of the flow dictated by the heat sink at the fan exit pose a different operating
11
scenario than the fan curves obtained during typical fan tests. Additionally, the fins in
the active heat sink may be designed based on a certain air speed that may differ from
that provided by the fan after being installed within the active heat sink. Thus, the
design of the active heat sink should involve the conjugate design of both the fan and
the heat sink.
2.2.2 Air Jet Impingement
The concept of using a concentrated jet for localized high heat flux cooling is similar to
that used for metal quenching. Jet impingement offers not only the ability to remove
high heat fluxes but also the ability to target hot spots or uneven heating. In addition,
the jet placement is not a crucial factor with respect to the cooled part. A concentrated
jet does not spread out in a conical fashion as a typical spray would and that makes its
design simpler. The drawback is that a high pressure head is needed that would be
converted to high kinetic energy of the jet. Also, there may be some noise concerns
becauseof the high speeds. These cautions can be analyzed up front at the time of design
to weigh out the benefits and risks.
2.2.3 Micro Channels
Micro channels are based on a very simple heat transfer concept: the heat transfer
coefficient for laminar flow is inversely proportional to the hydraulic diameter. This
means that the smaller your channel is, the higher your ability to draw heat from the
source. Micro channels typically have sizes in the 5 to 100 μm range leading to a heat
transfer coefficient of that may reach 80,000 W/m2K. They are typically etched on the
die surface in the shape of rectangular grooves. There are commonly two main problems
when designing a system of micro channels: pressure drop and flow uniformity across 12
the channels. The smaller hydraulic diameter results in a higher heat transfer coefficient
on the one hand but higher pressure-drop on the other hand. This would require higher
pumping power. One solution for that is called “stacking” – instead of having a single
layer of micro channels on top of the heat source, you may have two, three, or more
stacks.Studies have shown that most often two or three stacks are a good compromise
between heat transfer behavior and pressure drop. Flow non-uniformity across the micro
channels would result in non-uniform cooling, which may have implications on both the
performance and reliability.
2.2.4 Heat Pipes
Heat pipes are now the darlings of portable electronics cooling. They offer a high
degree of flexibility in design and have proven to be extremely reliable since they are
passive with no moving parts. Their heat transfer characteristics are superb, offering
effective conductivities up to several thousands of that of copper, enabling the transfer
of heat with minimal temperature gradient. Keep in mind when designing or selecting a
heat pipe for a certain system, the known limits must be taken into consideration. These
limits include the capillary limit, boiling limit, sonic limit, entrainment limit, and
flooding limit. Depending on the design of the heat pipe, its orientation within the
system and the heat flux applied to it, it may hit one of its limits and fail to perform its
cooling duties.To ensure the proper function of the heat pipe within the system, the
dynamic operation of the system with the heat pipe has to be analyzed under different
conditions to ensure continuous performance.
2.2.5 Spray Cooling
13
Spray cooling has the promise of extracting heat fluxes in excess of 100 W/cm2.
Typically a liquid is sprayed directly on the die, which makes the use of a dielectric
fluid essential, where it gains heat, converts to vapor and is cooled far away from the
heat source, condenses and then re-pumped to be sprayed again. Sprays may be
generated through different mechanisms, such as nozzles (pressure sprays or
atomization sprays) or even through the use of inkjet-inspired technology1.The latter
has the advantage of being able to target non-uniform heat sources and avoid “pooling”
of liquid on cooler parts of the heat source. Design variables include nozzle design,
spacing between nozzle exit and target, spray flow, liquid properties,and heat flux – all
of which must be analyzed closely to avoid potential problems
2.2.6 Immersion Cooling / Direct Contact Cooling
The terms “immersion” or “direct” are used to describe this approach because the
working liquid comes into direct contact with the chip. The liquid may be moving
passively due to natural convection or be driven by a pump. The liquid may also
undergo partial phase change in which case much higher heat fluxes may be attained.
Typically, excessive boiling should be avoided in order to reduce the creation of large
bubbles that would lead to local hot spots. Instead, the preferred condition is that of sub-
cooled boiling where the bubbles are small enough to re-condense into the main flow.
14
2.2.1 Water Cooling Techniques
Water cooling is not new to computers. Some early computers and many large
mainframes used water cooling systems for years but eventually gave way to air cooling
after chip technology made them much smaller. Water cooling system of desktop
personal computer has again gained its attention in the recent years mainly due to its
capability on performing spot cooling on graphic processor unit (GPU) and high
performance CPU. This cooling system can provide the coolant directly to the spot
which needs to be cooled and this is the key benefit. It is also one of the best options for
heat source that is nonuniform.
The other advantages of using water over air cooling are due to higher values of
specific heat capacity, density and thermal conductivity. Water cooling through
channels transfers the heat directly from a chip surface more efficiently because it bring
fluid into intimate contact with the channel walls and in return it brings fresh fluid to the
walls and remove fluid away from the walls as the transport process is accomplished .
2.3 Minichannels
In the early 1980’s Tuckerman and Pease [15] introduced the concept of micro-
channel heat sinks. It was demonstrated that laminar flow in micro rectangular channels
has higher heat extraction abilities than turbulent flow in conventional sized flow
channels. This discovery drove an entire new research field as it offers a reliable system
with good price/performance ratio.
Employing smaller channel dimensions results in higher heat transfer
performance due to the increase of heat flux dissipation but it is accompanied by a
higher pressure drop per unit length. The higher volumetric heat transfer densities
15
require advanced manufacturing techniques and lead to more complex manifold
designs. An optimum balance for each application leads to different channel
dimensions. These high levels of heat dissipation require a dramatic reduction in the
channel dimensions, matched with suitable coolant loop systems to facilitate the fluid
movement away from the heat source.
2.3.1 Channel Classification
Classification of channels is proposed by Mehendale et al. 16 as shown in Table 2.1.
Table 2.1: Proposed Classification of Channels
1 > Dh > 100 m Microchannels
100 m > Dh > 1 mm Meso-channels
1 > Dh > 6 mm Compact Passages
Dh > 6 mm Conventional Passages
Dh : Hydraulic Diameter
The classification scheme is later modified and more general scheme based on the
smallest channel dimension is presented in Table 2.2.
Table 2.2: Channel Classification
0.1 m D Nanochannels
1 D > 0.1 m Transitional Nanochannels
10 D > 1 m Transitional Microchannels
200 D > 10 m Microchannels
3 mm D > 200 m Minichannels
D 3 mm Conventional Channels
D: Channel Diameter. For non circular channels, the minimum channel dimension is used.
16
2.3.2 Advantages of Minichannels
The microchannel flow geometry offers a large surface area of heat transfer and
a high convective heat transfer coefficient. A cooling system for a microscale device
might require cooling channels of a few tens of micrometers as compared to more
conventional sized channels with 1–3mm flow passage dimensions. However, it is hard
to implement it into the compact/slim design of computers or consumer electronic
devices. The major difficulty is driving water with high pressure head, which is required
to pump the coolant fluid though the channels. A normal channel could not give such
high heat flux although the pressure drop is very low. Thus, an idea formed that water-
cooled minichannel with characteristic lengths within 0.2~3 mm [17] can be used in
heat sink with a high heat flux and a mild pressure loss. Copper heat sinks with
integrated microchannels and minichannels are expected to dominate heat sink
applications in future.
2.3.3 Recent Studies in Minichannels
As the channel size becomes smaller, some of the conventional theories for
fluid, energy, and mass transport need to be revisited for validation. There are two
fundamental elements responsible for departure from the “conventional” theories at
microscale. Differences in modeling fluid flow in small diameter channels may arise as
a result of uncertainty regarding the applicability of empirical factors derived from
experiments conducted at larger scales and uncertainty in measurements at microscale,
which includes geometrical dimensions and operating parameters.
Microchannel cooling technology was first put forward in 1981 by Tuckerman
and Pease [18], who employed the direct water circulation in microchannels fabricated
in silicon chips. They were able to reach the highest heat flux of 7.9 MW/m2 with the
17
maximum temperature difference between substrate and inlet water of 71 °C. However,
the penalty in pressure drop was very high, i.e. drop by 200kPa with plain
microchannels and 380kPa with pin fin enhanced microchannels. Later, Philips [19]
analyzed the heat transfer and fluid flow characteristics in microchannels in details and
provided formulations for designing microchannel geometries. Recently, Kandlikar et
al. made a series studies on the direct liquid cooling technology by microchannels [20-
22].
Convective heat transfer and fluid flow in minichannel and their application in
the cooling technology of electronic devices have attracted great attention of researchers
in recent years. Gael et al. [23] indicated that the heat conduction in the walls of
mini/micro-channels makes the heat transfer to be multidimensional, and the axial heat
conduction in the walls can not be neglected. The surface roughness effects on pressure
drop in single-phase flow in minichannels were investigated by researchers in [24-27].
Gao et al. [28] made experimental investigations of scale effects on hydrodynamics and
the associated heat transfer in two-dimensional mini and microchannels with channel
height ranging from 0.1 - 1.0mm. Their results showed that the conventional laws of
hydrodynamics and heat transfer can be applied to channels with height larger than
0.4mm. Wang et al. [29] experimentally examined the frictional characteristics inside
minichannels (Dh = 0.198-2.01mm) with water and lubricant oil as the working fluids,
and the tests were performed in both round and rectangular configurations. The test
results indicated a negligible influence of viscosity on the friction factor if the hydraulic
diameter is greater than 1.0mm. The measured data can be well predicted by the
conventional correlation in both laminar and turbulent flow conditions. Agostini et al.
[30] presented an experimental study of friction factor and heat transfer coefficient for a
vertical liquid up flow of R-134a in minichannels. Downing et al.[31,32] experimentally
18
investigated the single- and two-phase flow pressure drop and heat transfer
characteristics in straight and miniature helical flow passages with R-134a as a working
fluid. Debray et al. [33] performed the measurement of forced convection heat transfer
coefficients in minichannels. Reynaud et al. [34] measured the friction and heat transfer
coefficients in two-dimensional minichannels of 1.12mm to 0.3mm in thickness and
experimental results are in good agreement with classical correlations relative to
channels of conventional size. Liu and Mui [35] proposed a microprocessor package
with water cooling in which a narrow water jacket was used to cool a thermal spread
attached to the silicon die backside for an efficient cooling. Schmidt [36] described a
microprocessor liquid cooled minichannel heat sink and presented its performance as
applied to a microprocessor (IBM Power 4) chip. Yazawa and Ishizuka [37] gave an
analytic model for laminar flow and conducted a numerical study to optimize the
channel in cooling spreader on a smaller and transient heat source. It was concluded that
when small pumping power was available, a deeper channel with a thicker base was the
best profile for the miniature channel coolers, and the best cooling performance was
found at 0.0586K/W for 0.03W pumping power.
2.4 Serpentine Minichannels
A method to study fully-developed flow and heat transfer in channels with
periodically varying shape was first developed by Patankar et al. [38] for the analysis of
an offset-plate fin heat exchanger. Webb and Ramadhyani [39] and Park et al. [40]
analysed fully developed flow and heat transfer in periodic geometries. We recently
characterised the thermo-hydraulic performance of serpentine passages with a circular
channel cross-section [41] and showed that the establishment of Dean vortices at the 19
bends in this geometry give rise to significant heat transfer enhancement which, in the
absence of the creation of recirculation zones, can be achieved with a very small relative
pressure-drop penalty.
The definition of serpentine channels follows the work of Liu etal. [42]. The
channel consists of a number of repeating modules that are periodic in nature. Studies
reported by Kalb and Seader [43] and Masliyah and Nandakumar [44] have shown that
the heat transfer enhancements can exceed the relative pressure-drop penalty by a
significant amount (by factors of 2 or more for water) for laminar flows with constant
axial heat flux and peripherally uniform temperature.
Serpentine channels enhance the heat transfer by augmenting single-phase
convective heat transfer in channels. The key idea is to periodically interrupt
hydrodynamic and thermal boundary layers. Periodic restart of thermal boundary layers
leads to higher heat transfer coefficients. Periodic restart of hydrodynamic boundary
layers also creates a series of entrance regions, and heat transfer coefficients in the entry
region are significantly higher than that in the fully developed region. Serpentine
minichannels have a larger heat transfer area for a given volume than conventional
straight channels. The series of right-angle turns also promotes mixing by impingement,
recirculation, and flow separation.Though higher pressure head is required to pump the
coolant fluid though the serpentine minichannels because higher frictional losses are
inevitably incurred in producing curved passages flow.
2.5 Overview of CPU Water Cooling System
Cooling hot computer components with various fluids has been in use since at
least as far back as the development of Cray-2 in 1982, using Fluorinert. Through the
20
1990s water cooling for home PCs slowly gained recognition amongst enthusiasts
(overclockers) as it allow quieter operation with improved processor speeds, it only
started to become noticeably more prevalent after the introduction of AMD's hot-
running Athlon processor in mid 2000. Apple's Power Mac G5 was the first mainstream
desktop computer to have water cooling as standard. Dell followed suit by shipping
their XPS computers with liquid cooling, using thermoelectric cooling to help cool the
liquid.
A CPU water cooling system consists of heat sink, reservoir, pump, radiator, and
fan. The fluid is closed and circulated within these components with each component
acting interdependently as shown in Figure 2.2.
Figure 2.2: Schematic Layout of a Computer Water Cooling System
(Source: Heat Transfer and Fluid Flow in Minichannels and Microchannels book)
21
2.5.1 The Heat Sink
22
23
Figure 2.3: Water Based Heat Sink
http://www.customthermoelectric.com/Water_blocks.html
Heat sink transfer and dissipate heat generated from the processor. Heat transfer
from the processor to the heat sink depends on the thermal convection by the water and
to a lesser extent by thermal conduction from processor. Normally high thermal
conductivity material such as copper is used for the water block. Copper is widely used
due to its availability and relatively low material cost. Annealed copper has a thermal
conductivity of 385 W/m-K. Even though silver has a better thermal conductivity of 419
W/m-K it is impractical in-terms of price-performance ratio. A sample of heat sink is
shown in Fig 2.3.
2.5.2 The Reservoir
Figure 2.4: Reservoir
24
http://www.addka.com/watercooling/watercoolingkits.html
The reservoir is the place to store the water. It is one of the important
components for the cooling system. Once equilibrium is achieved, there are a number of
useful observations available in the reservoir. It serves as the point of checking the
states of the coolant. The states involved are pressure, mass flow rate and temperature.
A sample of reservoir is shown in Fig 2.4.
2.5.3 The Pump
Figure 2.5: Pump
http://wizdforums.co.uk/showthread.php?p=131633
The pump provides the required pressure for the water to circulate around the
system. From the research work of developing new generation of liquid cooling system
by Sukhvinder, it is reviewed that centrifugal pump is the most suitable pump type as it
is well known for its reliability and long lifecycle. The pump has three factors that are
important to look: pump capacity, maximum lift and sound level. A sample of pump is
shown in Fig 2.5.
25
2.5.4 The Radiator
Figure 2.6: Radiator
http://www.clunk.org.uk/reviews/corsair-h50-cpu-cooler-review/Page-3.html
Radiator is the component of the liquid cooling system that is in contact with the
outer surrounding, typically the air which let the heat dissipated out the computer
casing. In a radiator, heat is transferred through air cooled fins usually with a computer
fan. The fan is mounted on the radiator and forced air blow removes the heat of the
coolant into the ambient. This heat is dissipated continuously while heated coolant is
flowing through. The radiator fin surface area determines the performance of the liquid
cooling system. A sample of radiator is shown in Fig 2.6.
26
2.5.5 The Fan
27
Figure 2.7: Fanhttp://www.clunk.org.uk/reviews/corsair-h50-cpu-cooler-review/Page-3.html
28
The fans are employed to further improve the overall cooling system. The fan is
mounted at the fins of the radiator. This will create a turbulent air flow to carry away the
heat effectively that dissipated by the radiator. A sample of fan is shown in Fig 2.7.
2.6 Summary
The evolution of electronic packaging technology which enable exponential increase of
transistors in chip packaging has posed greater need for more efficient cooling system.
Traditional and advanced cooling techniques have been studied to improve the
efficiency of the system and to fulfill future need of electronic industry. Water cooling
through serpentine mini-channels is studied to address the CPU cooling problem as it
has potential for high heat transfer performance due to high specific heat capacity of
water, higher heat flux dissipation in mini-channels and design of serpentine channels
which can further enhance the heat transfer by augmenting single-phase convective heat
transfer in channels. From recent studies in mini- channels detailed and summarized, the
following summary can be drawn as:
(i) In the liquid minichannels, the conventional physical and mathematical
models for fluid flow and heat transfer with no-slip boundary conditions are
still valid;
(ii) The friction factor and heat transfer correlations for conventional channels
can also be used in minichannels as long as their relative surface roughness
and relative channel wall thickness are not too high.
29
CHAPTER 3: ANALYSIS USING COMPUTATIONAL
FLUID DYNAMICS
3.1 Introduction
Modeling is a technique used by the designer to define a real situation for
generating quantitative solutions. Relevant approximations in the modeling are critical
for obtaining the best solution to the engineering problem. Optimization of the design
generally concentrates on selectively choosing the best nominal values of design
30
parameters that optimize performance reliability. In this study the parameters are
designed based on Design of Experiments and the analysis are done using
Computational Fluid Mechanics to find the best design parameters for serpentine mini-
channel heat sink.
3.1.1 Design of Experiments (DOE)
Statistically designed experiments are invaluable in reducing the variability in
the quality characteristics and determining the levels of controllable variables which
will optimize process performance. Designed experiment technique is adopted to
discover the key variables influencing the quality characteristics of the interest of this
study. Often, there will be significant breakthrough in process performance benefited
from using the design experiments.
A statistical design of experiment is the process of planning experiments so that
appropriate data will be collected, the minimum number of experiments will be
performed to acquire the necessary technical information and suitable statistical
methods will be used to analyze the collected data. There are two aspects to any
experimental design; the design of the experiment and the statistical analysis of the
collected data.
3.1.1.1 Orthogonal Array Method
For the experiment the L16(45) orthogonal array is chosen. Sixteen experiments are
needed to be carried out for the L16(45) orthogonal array of Taguchi’s method.
31
Figure 2.9: Notation of Orthogonal Array
The “L” notation as shown in Figure 2.9 indicates that the information is based
on the Latin square arrangement of factors. A Latin square arrangement is a square
matrix arrangement of factors with separable factors effects. Thus, the L notation
indicates that the information is orthogonal array information.
The number of rows indicates the number of experiments required when using
that orthogonal array. The number of columns indicates the number of factors that can
be studied in the orthogonal array. The number of levels indicates the number of factor
levels.
3.1.2 Computational Fluid Dynamics
The physical aspect of fluid flow is governed by three fundamental principles
which are conservation of mass, conservation of energy and Newton’s second law.
These fundamental principles can be expressed in term of mathematical equation. Fluid
dynamics is the science of determining a numerical solution to the governing equations
of fluid flow whilst advancing the solution through space or time to obtain a numerical
description of the complete flow field of interest. Computational Fluid Dynamics is the
L 16 ( 4 5 )Number of columns
Number of levels
Number of rows
Represent Latin squares
32
study of fluid dynamics using a computer. The mathematical equations that govern the
flow of fluids are Nonlinear Partial Differential Equations which is extremely difficult
to solve. Because of this those who studied fluid flow without access to modern
computers had to make simplifying assumptions in order to solve these equations.
Experimental fluid dynamics has played an important role in validating and delineating
the limits of various approximation to the governing equations. Traditionally this has
provided a cost effective alternative to full scale measurement. However in the design of
equipment that depends critically on the flow behaviour , full scale measurement as part
of design process is economically impractical.
The first problem is that with an experiment one can only takes measurements at
certain points in the flow field and does not know what is happening anywhere apart
from it, in the flow. Obviously one can use many measuring devices and take more
measurements, but this adds to the cost of the experiment, and usually the measuring
devices themselves will effect the flow, so any additional devices can actually make the
results inaccurate. In a computer simulation one solves for the entire flow field and thus
one has access to information about the flow field at every point in the flow.
The second problem are due to the prototyping process. Prototyping is a time
consuming and expensive process. If the model tested is not perfect, the model have to
be redesigned and retested till the requirement is met.
Computational fluid dynamic complements experimental and theoretical fluid
dynamics by providing an alternative cost effective means of simulating real flow .As
such it offers the means of testing theoretical advances for conditions unavailable on
experimental basis.The role of Computational fluid dynamics in engineering prediction
33
has become so strong that it may viewed as a new third dimension of fluid dynamics.
Computational fluid dynamics is a critical part of the research and development in
academia and industries such as aerospace, automotive, naval architecture, power plant
design, biotechnology, aquaculture, environmental engineering and many more.
The most fundamental consideration in CFD is how one treats a continuous fluid
in a discretized fashion on a computer. One way is to discretize the spatial domain into
small cells to form a volume mesh or grid, and then apply a suitable algorithm to solve
the equations of motion (Euler equations for inviscid, and Navier-Stokes equations
forviscid flow). In addition, such a mesh can be either irregular (for instance consisting
of triangles in 2D, or pyramidal solids in 3D) or regular; the distinguishing
characteristic of the former is that each cell must be stored separately in memory. If the
problem is highly dynamic and occupies a wide range of scales, the grid itself can be
dynamically modified in time, as in adaptive mesh refinement methods
The stability of the chosen discretization is generally established numerically
rather than analytically as with simple linear problems. Special care must also be taken
to ensure that the discretization handles discontinuous solutions gracefully. The Euler
equations and Navier-Stokes equations both admit shocks, and contact surfaces.
One of the discretization methods being used are the finite volume method. This is
the "classical" or standard approach used most often in commercial software and
research codes. The governing equations are solved on discrete control volumes. This
integral approach yields a method that is inherently conservative (i.e., quantities such as
density remain physically meaningful):
34
0FdAQdVt (2.1)
Where Q is the vector of conserved variables, F is the vector of fluxes (see Euler
equations or Navier-Stokes equations), V is the cell volume, and A is the cell surface
area.
It is possible to directly solve the Navier-Stokes equations for laminar flow cases
and for turbulent flows when all of the relevant length scales can be contained on the
grid .In general however, the range of length scales appropriate to the problem is larger
than even today's massively parallel computers can model. In these cases, turbulent flow
simulations require the introduction of a turbulence model. large eddy simulations and
the RANS formulation (Reynolds-Averaged Navier-Stokes equations), with the k-ε
model or the Reynolds stress model, are two techniques for dealing with these scales In
many instances, other equations (mostly convective-diffusion equations) are solved
simultaneously with the Navier-Stokes equations. These other equations can include
those described as species concentration, chemical reactions, heat transfer, etc. More
advanced codes allow the simulation of more complex cases involving multi-phase
flows (eg, liquid/gas, solid/gas, liquid/solid) or non-Newtonian fluids (such as blood).
3.1.2.1 FLUENT
FLUENT is the world's leading supplier of computational fluid dynamics
(CFD) software and services. FLUENT is a state-of-the-art computer program for
modeling fluid flow and heat transfer in complex geometries. FLUENT provides
35
complete mesh flexibility, solving problems with unstructured meshes that can be
generated by complex geometries with relative ease. Supported mesh types include 2D
triangular/quadrilateral, 3D tetrahedral/hexahedral/pyramid/wedge, and mixed
(hybrid) meshes. FLUENT also allows the own solution. FLUENT is written in the C
computer language and makes full use of the flexibility and power offered by the
language. Consequently, true dynamic memory allocation, efficient data structures,
and flexible solver control are all made possible.
3.2 Overview
Mini-channel serpentine heat sink is designed using Gambit. Heat sink mini-channel
design parameters are set based on heat sink surface area size of 20mm x 20mm and
thickness of 10mm. The study is constrained to laminar flow so the initial flow velocity
is calculated mathematically. Chosen design variables is then sampled according to
Design Experiment Method Orthogonal Array. Computational Fluid Dynamic software
of FLUENT is used to analyze the test samples. Design geometry analysis and inlet
velocity analysis is done to obtain the optimized model. The chronology of design,
relevant calculation, sampling and analysis are as detailed in Figure 3.1.
36
Figure 3.1: Research Methodology Flowchart
3.3 Heat Sink Design Specification
< Top View > < Isometric View >
CFD Analysis
CFD Analysis
Setting Heat Sink Design Parameters
Range
Setting Initial Velocity Parameters
Design Parameter Sampling using Orthogonal Array
Heat Sink Design
Design Geometry Analysis
Inlet Velocity Analysis
Nearly Optimized Model
37
38
Wc
< Front View >
39
Ww
Figure 3.2: Heat Sink Design Specification
40
Hb
Figure 3.2 shows a pictorial view of the suggested model. The total area heat
sink is 20 mm X 20mm with minichannel flow passage dimensions of Wc X Hc. The
wall separating the two channels has thickness of Ww. The bottom plate thickness is Hb.
The top cover is assumed to be bonded, glued, or clamped to provide closed channels
for liquid flow. The channel dimensions Wc and Hc, the channel wall thickness Ww, the
bottom plate thickness Hb, and the coolant flow velocity Uin are the parameters of
interest in designing a minichannel heat sink.
3.4 Design Parameters
In order to obtain better thermal performance and acceptable mild pressure drop,
it is important to find the heat sink design parameters. The channel width and the
channel aspect ratio (Hc /Wc) have significant effects on the performance of
minichannel heat sink. The bottom plate thickness has a very high influence on the
thermal performance as it conducts the heat flux from the chip. The channel width
varies from 0.5mm to 2mm in accordance to the higher order of mini-channels for
example, for ease of fabrication.
Narrow channels are studied as it is expected to result in lower wall temperature.
The heat convection from the walls to the fluid in the channel is faster because of the
high-aspect-ratio of the channels. Channel heights varies from 2 mm to 5 mm.
Table 3.1: Range of Design Parameters
Parameters Range
Bottom Plate Thickness, Hb (mm) 0.2 – 0.8
Channel Width, Wc (mm) 0.5 – 2
Channel Height, Hc (mm) 2.0 - 5.0
41
Channel Wall Thickness, Ww (mm) 0.5 - 2.0
3.4.1 Initial Velocity
Inlet velocity influence on thermal performance and pressure drop of a heat sink
are studied. The type of flow also depends on inlet velocity. In this research, the study
of flow is confined to laminar flow. The condition of flow is usually expressed using
Reynolds number. To decide that the range of initial velocity within the laminar range a
simple Excel programming based on Reynolds number are done. All the possible design
parameters which affect the Reynolds number are taken into account and the range
values of initial velocity within laminar flow are used.
3.4.1.1 Reynolds Number
The transition from laminar to turbulent flow depends on geometry, surface
roughness, flow velocity, surface temperature and type of fluid. Reynolds number is a
dimensionless quantity which are referred to express the condition of flow
(3.1)
Where, Re = Reynolds number; = Fluid density; V = Mean fluid velocity;
Dh = Hydraulic Diameter; µ = Dynamic viscosity.
42
Table 3.2: Input Values for Reynolds Number Calculation
3.4.1.2 Hydraulic Diameter
Hydraulic Diameter is used for the calculation of Reynold's Number. Hydraulic
Diameter for rectangular duct can be calculated by applying the formula shown below.
Dh = 4ab
2(a+b)
(3.2)
Where, Dh = Hydraulic Diameter; a = Length of Rectangular; b = Width of Rectangular.
Re 2300 Laminar flow
2300 Re 4000 Transitional Flow
Re 4000 Turbulent flow
43
3.4.1.3 Initial Velocity Range
Input values based on the properties of water at 40°C are taken into account to calculate
Reynolds Number. The input parameters are as presented in Table 3.2. Based on the
calculation in Table 3.3, initial velocity range is suggested from 0.1 m/s to 0.4 m/s to
maintain a laminar flow in the channel
Table 3.2: Input Values for Reynolds Number Calculation
3.5 Orthogonal Array Test Run
The effects of channel width Wc, channel height Hc, bottom plate thickness Hb and wall
thickness Ww are studied using Design Experiment Method Orthogonal Array. The
numerical results present about the influences of those parameters on the water pressure
drop and the total surface heat flux. In the analysis, orthogonal array method are
adapted based on the details that can be found in Montgomery [31] and Liu [32]. The
factors and their levels have been shown in Table 3.4 where factors A, B, C and D are
Input ValuesTemperature ,T 40°CDensity , 991.8 kg/m3
Dynamic Viscosity , 6.55E-04Average Velocity ,Vavg 0.1- 0.4 m/s
44
the channel height, the channel width, the channel wall thickness and the bottom plate
thickness, respectively. The arrangement of test using L16 (45) orthogonal array can be
seen in Table 3.5. Considering test parameters as shown in Table 3.4. Since there are
only four factors in the current analysis, the last column in Table 3.5 is blank. Sixteen
combinations referring to the test parameters are analyzed.
Table 3.4: Test Parameters Serpentine Minichannels
Table 3.5: Arrangement of Test Run using Orthogonal Array
Column No. 1 2 3 4 5
Factors A B C D
Blank
RunChannel Height
Hc (mm)
Channel Width
Wc (mm)
Wall ThicknessWw (mm)
Bottom Plate
ThicknessHb (mm)
1 1 2 3 2 /
2 3 4 1 2 /
3 2 4 3 3 /
Levels
FactorsChannel Height Hc(mm)
Channel Width
Wc(mm)
Wall Thickness Ww(mm)
Bottom Plate Thickness Hb(mm)
A B C D1 2 0.5 0.5 0.82 3 1 1 0.23 4 1.5 1.5 0.44 5 2 2 0.6
45
4 4 2 1 3 /
5 1 3 1 4 /
6 3 1 3 4 /
7 2 1 1 1 /
8 4 3 3 1 /
9 1 1 4 3 /
10 3 3 2 3 /
11 2 3 4 2 /
12 4 1 2 2 /
13 1 4 2 1 /
14 3 2 4 1 /
15 2 2 2 4 /
16 4 4 4 4 /
3.6. Assumptions
To analyze the thermal and flow characteristics of this model, the following
assumptions are made:
(i) The flow is three-dimensional, incompressible, laminar and in steady-state.
(ii) The effect of body force is neglected.
(iii) Fluid thermophysical properties are constant and heat dissipation
neglected
3.7 Modeling Using FLUENT
Computational Fluid Dynamic software of FLUENT is used to analyze the models. Two
main analysis were done to obtain the optimized model .The design geometry of the
model were analyzed based on the Design of Experiment Method Orthogonal Array.
Constant initial velocity of 0.1m/s is assumed throughout the analysis. The optimized
46
design geometry is then analyzed with a range of velocity from 0.1 m/s to 0.4 m/s to
obtain optimum initial velocity for the model.
3.8 Geometry of Models
3.8.1 Pre Processing
Pre Processing is related to the tasks in generating a flow model. It includes creating
geometry, meshing and applying boundary conditions.
3.8.1.1 Creating Geometry
Sixteen solid models are created by using GAMBIT and Orthogonal Array Tool is
used to create the systematic statistical way of testing. GAMBIT is a state of the art
preprocessor for engineering analysis with advanced geometry and meshing tools in
a powerful, flexible, tightly-integrated and easy to use interface. Thus, geometry of
mini channels is easily developed in Gambit.
47
Figure 3.3: Heat Sink Model Designed using Gambit
3.8.1.2 Meshing48
The models are then discretized using hex-sub-map meshing. Mesh refinement on the
models can be carried out by setting nodal interval of 0.2 nd 0.15.
Copper is chosen from the Fluent Library as the heat sink wall material and water as the
liquid flow material.
49
.
Figure 3.4: Model Meshed using Hex-Sub Map
50
3.8.1.3 Applying Boundary Conditions
The model is divided to four boundaries for the purpose of analysis which bottom plate,
top wall, inlet velocity and outflow .The wall are divided into upper wall and bottom
plate to accommodate bottom thickness analysis
3.8.1.4 Boundary Condition Specification
i) The thermal boundary conditions are specified as follows.
(a) The left and right of the wall surfaces are adiabatic
y = 0 , T/y = 0
(3.6)
y = Wc + Ww , T/y = 0
(3.7)
(b) At the bottom position the heat flux given :
z = 0 , -sT/y = qw
(3.8)
Heat flux of 100W/cm assumed to be produced by the chip
(c) The top surface is assumed to be adiabatic
y = Hb + Hc , -fT/z = 0
(3.9)
(d) At the inlet position, the inlet temperature of water is given to be constant
and outlet boundary is considered of local one way type:
x = 0 T = Tin (3.10)
ii) Constant velocity of 0.1m/s is applied.
51
Figure 3.5: Bottom Plate Boundary
52
Figure 3.6: Top Wall Boundary
53
Figure 3.7: Velocity Inlet and Outflow Boundary
54
3.8.2 Processing
3.8.2.1 Solver Specification
A stability-guaranteed second-order difference scheme (SGSD) is used to
analyze the convective terms while the others are approximated by centre-difference
approach. The SIMPLE solution algorithm is adopted to deal with the linkage between
pressure and velocities. Mesh convergence studies are carried out to obtain optimum
number of elements used in the models.
During the iterative solution process if the relative deviation between two
consecutive iterations is specified to be less than 10%, then the solution is considered to
be converged.
The optimized design geometry is than analyzed by applying a range of initial
velocity to find the optimum model. The optimum initial velocity were analyzed and
chosen from the Design of Experiment method based on relatively high heat transfer
rate and low pressure drop
3.8.3 Post Processing
This is to view the results solved CFD analysis. It includes the organization and
interpretation of the predicted flow data and the distribution of the expected results. The
data is validated by applying Grid Independence Study. The optimized design geometry
were analyzed and chosen from the Design of Experiment sample based on relatively
high heat transfer rate and low pressure drop.
3.9 Summary
55
Heat sink surface area size of 20mm x 20mm and thickness of 10mm are a chosen to
study the mini-channel serpentine flow. Five parameters which are bottom plate
thickness, channel width,channel height and channel thickness are designed based on
Design of Experiments method Orthogonal Array. As the study is constrained to
laminar flow so the initial flow velocity is calculated mathematically. Design geometry
analysis based on samples from orthogonal array method and inlet velocity analysis is
done to obtain the optimized model.
19.
CHAPTER 4: RESULTS AND DISCUSSION
4.1 Overview
The CFD analysis and convergence studies have been carried out on the samples
which was designed based on Orthogonal Array. The details of the test run variables can
be referred at Table 4.1. Pressure drop has to be calculated manually based on the CFD
analyzed data as the exact value is not given. Samples of the CFD analysis summary are
attached in Appendix. Effects of design geometry parameters and inlet velocity
56
parameters on total surface heat flux and pressure drop are discussed in detail to find the
optimum design. Other significant effect such as effect of number of channels and effect
of serpentine channels are also discussed.
Table 4.1: Design Geometry Orthogonal Array Test Run
Column No.
1 2 3 4
Factors A B C D
Run
Channel Height
Channel Width
Channel Wall
Thickness
Bottom Plate
Thickness
Hc (mm)
Wc (mm)
Ww (mm)
Hb (mm)
1 2 1 1.5 0.22 4 2 0.5 0.23 3 2 1.5 0.44 5 1 0.5 0.4
57
5 2 1.5 0.5 0.66 4 0.5 1.5 0.67 3 0.5 0.5 0.88 5 1.5 1.5 0.89 2 0.5 2 0.410 4 1.5 1 0.411 3 1.5 2 0.212 5 0.5 1 0.213 2 2 1 0.814 4 1 2 0.815 3 1 1 0.616 5 2 2 0.6
4.2.1 Results Convergence Study
In this analysis three different convergence studies are conducted according to
residual convergence, outflow pressure convergence and outflow heat flux convergence.
Convergence studies were done on all sixteen test runs which were selected based on
Orthogonal Array Test. The convergence results for Test Run 2 of Orthogonal Array
Test with meshing size 0.2 are as shown in Figures 4.1 - 4.3.
58
Residual Continuity
x velocityy velocity
z velocityenergy
Figure 4.1: Residual Convergence Study
59
Iteration
Re
sid
ual
s
Iteration
Integral(pascal)(m2)
Figure 4.2: Outflow Pressure Convergence Study
Figure 4.3: Outflow Heat Flux Convergence Study
4.2.2 Grid Independence Study
The validity of meshing is done using Grid Independence Study. The analyzed model
are remeshed with a smaller fraction of the applied meshing size and reanalyzed. During
the iterative solution process if the relative deviation between two consecutive iterations
is less than 10% the iteration is considered converged.
60
Iteration
Integral (k)(m2)
0
500
1000
1500
2000
2500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Orthogonal Test Run
Pre
ss
ure
Dro
p (
Pa
sc
al)
Meshing Size 0.15 Meshing Size 0.2
Figure 4.4: Pressure Drop Grid Independence Study
0
100
200
300
400
500
600
700
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Orthigonal Test Run
Su
rfa
ce H
ea
t Flu
x (W
/cm
2)
Meshing Size 0.15 Meshing Size 0.2
Figure 4.5: Total Surface Heat Flux Grid Independence Study
In this study the model were meshed using mesh size 0.2 and 0.15. For pressure drop
analysis the results are can be assumed converged in almost all the samples except for
sample 4,7 and 14. For total surface heat flux analysis the results can be assumed
converged in almost all the samples except for sample 4, 7 and 14. Sample 4 couldn’t be
meshed with meshing size 0.2 due to a very small channel size with non symmetrical 61
Orthogonal Test Run
dimensions because the wall thickness size is different from channel dimensions. It also
can’t be meshed with meshing size 0.1 due to its high computer resources requirement.
The validity of meshing reduces as the size of channel decreases and number of channel
increases due to the limitation in computer processing which explains the reason for non
convergence in other samples. However, as the range of results gained from
nonconverged models are much smaller compared to other samples, it is safe to assume
that none of this sample can be considered as a candidate for optimum model. The
results analyzed and studied are based on meshed size 0.15.
4.3 Pressure Drop Calculation
Pressure Drop is the important parameter and its results are evaluated .The
analyzed data doesn’t calculate Pressure Drop automatically .However based on the
analyzed data in Attachment I, Pressure Drop of the channel can be calculated with
simple formula as shown below: The calculated pressure drop data are than summarized
in Table 4.1
Pressure Drop = Pressure Inlet – Pressure Outlet (4.1)
62
Table 4.1: Pressure Drop Calculation
Orthogonal Array Test
Inlet Pressure (Pascal)
Outflow Pressure (Pascal)
Pressure Drop (Pascal)
1 101487.450 101119.520 367.9302 101419.710 101253.480 166.2303 101379.180 101252.330 126.8504 101542.310 101138.230 404.0805 101496.770 101188.650 308.1206 101788.620 100756.970 1031.6507 102361.830 100394.160 1967.6708 101403.350 101271.960 131.3909 101839.730 100933.410 906.32010 101410.250 101217.490 192.76011 101394.100 101231.710 162.39012 101845.670 100710.580 1135.09013 101389.540 101239.050 150.49014 101462.500 101231.810 230.69015 101539.480 101153.580 385.90016 101351.79 101278.42 73.370
4.4 Analysis of the Results
The orthogonal samples are analyzed and effects of design geometry parameters
with fixed inlet velocity 0.1m/s on total surface heat flux and pressure drop. The
optimized design geometry is than analyzed with inlet velocity 0.2m/s,0.3m/s and
0.4m/s to find the optimum model.
The pressure drop and total surface heat flux analysis results areas according to
the Orthogonal Array Test sample with meshing size 0.15 and inlet velocity 0.1m/s is as
shown in Table 4.2. Graph in Figure 4.6 and Figure 4.7 are plotted according to Table
4.2.
63
Table 4.3:Total Surface Heat Flux and Pressure Drop Analysis of Design Geometry
Orthogonal Test Run
Total Surface Heat Flux (W/cm2 )
Pressure Drop (Pascal)
1 333.994 367.9302 611.033 166.2303 423.956 126.8504 62.106 404.0805 240.982 308.1206 15.511 1031.6507 14.188 1967.6708 71.577 131.3909 43.761 906.32010 176.278 192.76011 475.368 162.39012 37.027 1135.09013 320.219 150.49014 40.140 230.69015 72.943 385.90016 170.050 73.370
0
100
200
300
400
500
600
700
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16Orthogonal Test Run
Su
rfa
ce H
ea
t Flu
x (W
/cm
2)
Figure 4.6: Total Surface Heat Flux Analysis
64
0
500
1000
1500
2000
2500
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Orthogonal Test Run
Pre
ssur
e D
rop
(Pas
cal)
Figure 4.7: Pressure Drop Analysis
4.5 Effect of Design Geometrical Parameters
Analysis of geometrical parameters of the channels include channel height,
channel width, channel wall thickness and bottom plate thickness using Orthogonal Test
Run approach. Inlet Velocity are set to be 0.1m/s during the entire Orthogonal Test Run.
The detailed results derived from CFD are presented in Table 4.3. Pressure loss varies
from 73.37 Pa (Sample 16) to 1967.67 Pa (Sample 7). Total surface heat flux varies
from 14.188W/cm2 (Sample 7) to 611.03W/cm2 (Sample 2).
65
4.5.1 Effect of Channel Height
0
400
800
1200
1600
2000
2 4 3 5 2 4 3 5 2 4 3 5 2 4 3 5
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Channel Height(mm) According to Orthogonal Test Run
Surface Heat Flux (W/cm2 ) Pressure Drop (Pascal)
Figure 4.8: Effect of Channel Height
The variations of pressure drop and total surface heat flux for different channel
height are shown in Fig. 4.8. Based on the observation in Figure 4.8, the channel height
influence on pressure drop and total surface heat flux can be assumed minimal. Sample
4, 8, 12 and 16 with channel height of 5 mm gives mixed range of pressure drop.
Sample 16 with channel height 5 mm provides the lowest pressure drop of 73.370
Pascal.
66
Channel Height(mm)
Test Run
The channel height dimension is assumed as a significant parameter as it ranges
from 1 mm to 5mm compared to other design geometry dimensions which only ranges
up to 1mm. At the beginning of research, channel height are expected to influence high
rate pressure drop. However the analysis proved otherwise .It is due to the significance
of other design geometry factor in influencing the data compared to channel height.
Assumption can be made that channel height factor weightage in influencing pressure
drop is low.
In the case of height influence on total surface heat flux, further study is needed
to prove the insignificancy .In this study the upper wall (which includes the height of
the channels) is assumed adiabatic thus giving a room to overlook the significance
of .channel height on total surface heat flux.
4.5.2 Effect of Channel Width
67
0
400
800
1200
1600
2000
1 2 2 1 1.5 0.5 0.5 1.5 0.5 1.5 1.5 0.5 2 1 1 2
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Channel Width (mm) According to Orthogonal Test Run
Surface Heat Flux (W/cm2 ) Pressure Drop (Pascal)
Figure 4.9: Effect of Channel Width
The variations of pressure drop and total surface heat flux with channel width
are shown in Fig. 4.9. Smaller channel width has a significant negative effect on both
total surface heat flux and pressure drop. Total surface heat flux from all the samples
with 0.05mm is 45 W/cm2 which is less than 7 % of the highest recorded value. A factor
that might be overlooked is the possibility boiling in the channels which might be the
cause for such a low total surface heat flux. Further study on heat sink temperature is
needed before any solid argument can be made.
All the channels with smaller channel width recorded highest pressure drop for
as expected .As channel dimension become smaller, the number channel occupying the
68
Channel Width(mm)
Test Run
space will be higher. The pressure drop will be magnified by both the resistance of the
wall and the length in smaller channel.
All the channels with larger dimension, recorded high total surface heat flux and
lower pressure drop. Assumption can be made the larger channels have higher heat
transfer rate and lower pressure not only due to the dimensions of the channel but also
because as the channel grows larger the number of channel decreases which thereby the
length of channels.
4.5.3 Effect of Channel Wall Thickness
69
0
400
800
1200
1600
2000
1.5 0.5 1.5 0.5 0.5 1.5 0.5 1.5 2 1 2 1 1 2 1 2
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Channel Wall Thickness (mm) According to Orthogonal Test Run
Surface Heat Flux (W/cm2 ) Pressure Drop (Pascal)
Figure 4.10: Effect of Channel Wall Thickness
The variations of pressure drop and total surface heat flux with channel wall
thickness are as shown in Fig. 4.10. Based on the observation in Figure 4.10 , the
channel wall thickness influence on total surface heat flux are insignificant. Sample 2, 4,
5 and 7 with channel wall thickness 0.5 mm gives mixed range of total surface heat flux.
. Even though it is safe to presume that 0.5 mm gives the best results however its design
geometry weightage is much lesser due to the significance of other design geometry
factor in influencing total surface heat flux. Based on Figure 4.10, the influence of wall
thickness within range of 0.5mm to 2.0 mm in this design on pressure drop can be
assumed insignificant. However further detailed study can be done on the effect wall
70
Channel Wall T.(mm)
Test Run
thickness at the right angle. It can be concluded the the analysis of channel wall
thickness is only valid if the fluid structure interaction analysis is done.
The heat transfer process from the bottom to the cooling water of the channel
includes the conduction through the channel wall and the convection of the side wall,
which is fixed at the given inlet velocity and the given side wall surface area. When the
channel wall thickness Ww is too narrow, the conductive thermal resistance
predominated and the increase in Ww reduces the conductive thermal resistance, hence,
the total thermal resistance. However, further increase in Ww leads to the significant
increase in the total heat transfer rate entering the computational unit for the fixed heat
flux condition.
As the wall thickness parameter is set from 0.5mm to 2mm according to the
channels dimension, the search for the turning point which results from the balancing
between the heat conduction of the two parts had not been focused in this research.
Even though it is safe to presume that 0.5 mm gives the optimum performance in this
study based on the range suggested, further study can be done by reducing the wall
thickness to find the turning point of wall thickness which will give the optimum
parameter for it.
4.5.4 Effect of Bottom Plate Thickness71
0
400
800
1200
1600
2000
0.2 0.2 0.4 0.4 0.6 0.6 0.8 0.8 0.4 0.4 0.2 0.2 0.8 0.8 0.6 0.6
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Bottom Plate Thickness According to Orthogonal Test Run
Surface Heat Flux (W/cm2 ) Pressure Drop (Pascal)
Figure 4.11: Effect of Bottom Plate Thickness
Table 4.4: Bottom Plate Thickness Analysis
Bottom Plate Thickness (mm)
Pressure Out (Pascal)
Pressure In (Pascal)
Pressure Drop (Pascal)
Total Surface Heat Flux (W/cm2 )
0.2 101254.71 101419.43 164.72 623.440.4 101254.71 101419.43 164.72 315.250.6 101254.71 101419.43 164.72 211.000.8 101254.71 101419.43 164.72 158.00
72
Bottom Plate T.(mm)Test Run
0
200
400
600
800
2 4 6 8
Bottom Plate Thickness (mm)
Pressure Drop (Pascal) Surface Haet Flux (W/cm2)
Figure 4.12: Verification on Effect of Bottom Plate Thickness
The variations of pressure drop and total surface heat flux with bottom plate
thickness are as shown in Fig. 4.10. Based on the observation on Figure 4.10, the
channel wall thickness influence on total surface heat flux can be assumed
significant .Sample 1, 2 and 11 which had thin bottom plate thickness of 0.2 mm had
shown high range of total surface heat flux and Sample 7,8 and 14 which had thin
bottom plate thickness of 0.8 mm had shown low range of total surface heat flux .The
exception for this observation , Sample 12 and 13 behavior can be rationalized as
channel width and number of channels effect
Theoretically bottom plate thickness has no effect on the pressure drop of
channels. An additional analysis is done to confirm the study and verify that bottom
plate thickness has no effect on the pressure drop .The Orthogonal Array Sample 2 is
analyzed with bottom plate thickness 0.2 mm, 0.4mm, 0.6mm and 0.8mm with inlet
velocity 1.0m/s as shown in Table 4.4 . The result is plotted in Figure 4.12.
From the Figure 4.12, it can be seen that the total surface heat flux decreases
with the increase of bottom plate thickness and it does not have any effect on the
pressure drop. .Even though copper has high conduction rate, increasing its thickness
does have a significant negative effect on the total surface heat flux.
73
The heat transfer rate entering into the bottom surface of Hb is transferred
through two ways: one way is the heat conduction from the bottom surface to the top
surface of Hb and the other is from the bottom surface to the two side walls of the
computational unit. The first part increases with the decrease of Hb, while the second
part decreases with the decrease of Hb. The turning point which results from the
balancing between the heat conduction of the two parts had not been focused in this
research.
Even though it is safe to presume that 0.2 mm gives the optimum performance in
this study based on the range 0.2mm to 0.8mm, further study can be done by reducing
the bottom plate thickness to find the turning point of bottom plate thickness which will
give the optimum parameter for the bottom plate thickness.
4.5.5 Results of Optimized Design Geometry
Based on the above discussion about the influences of parameters of interest on
the pressure drop and total surface heat flux, Sample 2 is chosen as nearly-optimized
design geometry parameters. Total surface heat flux of Sample 2 is the highest from all
samples at 611.033 Pascal with reasonable pressure drop of 166.23 Pascal.
4.6 Effect of Inlet Velocity
The pressure drop and total surface heat flux analysis results according to the
nearly optimized design geometry with meshing size 0.15 is Sample 2 as shown in
Table 4.5. The results are plotted according to Table 4.5 in Figure 4.13.
Table 4.5: Total Surface Heat Flux and Pressure Drop Analysis of Inlet Velocity
Inlet Pressure Out Pressure In Pressure Drop Total Surface Heat 74
Velocity(m/s) (Pascal) (Pascal) (Pascal) Flux (W/cm2 )0.1 101254.71 101419.43 164.72 623.440.2 101095.42 101623.49 528.07 621.230.3 100859.9 101930.02 1070.12 791.380.4 100592.63 102331.51 1738.88 749.78
0
500
1000
1500
2000
0.1 0.2 0.3 0.4
Inlet Valocity (m/s)
Pressure Drop (Pascal) Surface Haet Flux (W/cm2)
Figure 4.13: Effect of Inlet Velocity
Pressure loss varies from about 164.72 Pa at inlet velocity of 0.1m/s to about 1738.88
Pa at inlet velocity of 0.4m/s. Pressure drop increases with the increase of inlet velocity
while total surface heat flux shows an erratic relation with increase of inlet velocity.
However the relatively low gain of total surface heat flux with increase of inlet velocity
comes with high penalty of pressure drop. At the low inlet velocity of 0.1m/s, the total
surface heat flux is 623.44 W/cm2 and 164.72 Pascal respectively. By increasing the
velocity to 0.3m/s the total surface heat flux only increased by 10.7% while the pressure
dropped almost 560%.The inlet velocity 0.1m/s is suggested as the optimum inlet
velocity for the model
75
4.7 Other Significant Effects
Based on the analysis done, two other factors which influence the results are identified.
From the analysis the effect of number of channels and serpentine right angle had been
too significant to be ignored.
4.7.1 Effect of Number of Channels
0
400
800
1200
1600
2000
8 8 6 12 10 10 20 6 8 8 6 12 6 6 10 4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Number of Channels According to Orthogonal Test Run
Surface Heat Flux (W/cm2 ) Pressure Drop (Pascal)
Figure 4.14: Effect of Number of Channels
The variations of pressure drop and total surface heat flux with number of
channels are as shown in Fig. 4.14. As the design of heat sink is only limited by the size
(20mmX20mm), number of channels are proportional to channel width and channel
76
wall thickness .Sample 6,7, 9 and 12 shows pressure drop of more than 900 Pascal .
All these four samples with highest pressure drop have more than 8 channels .Pressure
drop can be assumed proportional with the number of channels as the length of channels
increases more than 20 mm with every increase in number of channels .The number of
channels seems to have an adverse effect on the total surface heat flux .Sample 7 which
has the highest number of channels shows the lowest total surface heat flux of 14.188
W/cm2.However a more detailed study shows that the effect is due to channel
dimensions rather than the length.
4.7.2 Effect of Serpentine Right Angle
77
Figure 4.15: Effect of Serpentine Right Angle on Total Surface Heat Flux
The variations of surface heat flux with serpentine right angle are as shown in
Fig. 4.15.From the figure it can be seen that the effect of serpentine channel right angle
78
is very significant. At the right angle the surface heat flux of 5790 W/cm 2 is
recorded .By periodically interrupting hydrodynamic and thermal boundary layers, it
causes periodic restart at every right angle. Periodic restart of hydrodynamic boundary
layers creates a series of entrance regions and heat transfer coefficients are significantly
higher than in the fully developed region as can be clearly seen in the figure. .The effect
of augmentation at the right angles of the channels seems less insignificant at the
smaller channels. Sample 7 which has the maximum of 18 right angles has recorded
surface heat flux of only 14.188 W/cm2.
79
Figure 4.16: Effect of Serpentine Right Angle on Pressure
The variations of pressure drop with serpentine right angle are as shown in
Fig.4.16. From the figure it can be seen that the serpentine channel right angle doesn’t
80
effect pressure drop significantly .The cause of pressure drop seen are due to the length
of the channel.
4.8 Summary
Various factors that influences heat sink performances had been studied and how these
factors interact with one another to have an optimized result had been discussed .From
the results and observation made, Sample 2 from Orthogonal Array Test Run is chosen
as the nearly optimized geometric design .Total surface heat flux of Sample 2 is the
highest at 611.033 W/cm2 and the pressure drop is within the lower range at 166.230
Pascal .The inlet velocity of 0.1m/s is suggested as the optimum inlet velocity for
laminar flow .The nearly optimized model parameters for 20mm X 20mm serpentine
minichannel heat sink for laminar flow are as suggested in Figure 4.17.
Table 4.6: Nearly Optimized Model Parameters
Channel Height (mm) 4Channel Width (mm) 2Channel Wall Thickness (mm) 0.5Bottom Plate Thickness (mm) 0.2Inlet Velocity (m/S) 0.1
81
CHAPTER 5: CONCLUSION AND RECOMMENDATION
5.1 Overview
In conclusion, the objectives of this research work had been achieved. Design
geometry parameters which effects the pressure drop and total surface heat flux were
identified as channel height, channel width, channel wall thickness and bottom plate
thickness. The design geometry parameters range for channel dimension and channel
wall thickness were set to the upper range of minichannel dimensions which are from
0.5 mm to 2 mm. The channel height is set within the range of 2 mm to 5 mm to
increase the heat transfer rate through high-aspect-ratio of the channels. The bottom
plate thickness is set at minimum dimension of 0.2 mm to 0.8mm. Then, the design
geometry parameters are analyzed using Computational Fluid Dynamics software
FLUENT by applying Experimental Design approach Orthogonal Array method. Initial
velocity of 0.1m/s is applied on all the sample. The results were weighted based on the
highest total surface heat flux at acceptable pressure drop. Based on the analysis,
Sample 2 is chosen as the nearly optimized design geometry. The total surface heat flux
82
of Sample 2 is the highest at 611.033 W/cm2 and the pressure drop is within the lower
range at 166.230 Pascal.
Sample 2 are then analyzed with range of inlet velocity using Computational
Fluid Dynamics Method to find the nearly optimized heat sink model. The inlet velocity
are set within range of 0.1m/s to 0.4m/s to maintain laminar flow in the channels. The
analysis results show that as the inlet velocity increases, a slight increment of total
surface heat flux is penalized with huge of pressure drop.
Based on the design geometry analysis and inlet velocity analysis, heat sink
model with channel width 2mm, channel height 4mm, channel wall thickness 0.5mm,
bottom plate thickness 0.2mm and inlet velocity of 0.1m/s were chosen as dimensions
of the nearly optimized model of serpentine minichannel heat sink with bottom
dimension 20mm x 20mm .The nearly optimized model recorded total surface heat flux
of 611.033 W/cm2 and the pressure drop of 166.230 Pascal
5.2 Conclusion
From the research work done it can concluded that the design geometry
parameters such as bottom thickness and channel dimension have a significant effect on
the results while channel wall thickness and channel height does not have a significant
effect on the results. Channel height of the nearly optimized model in the study can be
be reviewed by applying lower range of height from current 4 mm as channel height
effect is insignificant.
Inlet velocity has significant adverse effect on the results. Better results with
lower inlet velocity give certain indication that laminar flow will give a better results in
minichannel compared to turbulent flow.
83
The total surface heat flux is mainly influenced by the serpentine right angle as
shown in Figure 4.15 due to the augmentation of the channel and not due to the high
aspect ratio of minichannels as assumed.
In this study, the bottom dimensions were set at 20 mm x 20 mm with the
channels designed to optimize the space. This design aspect causes large number of
channels with smaller channel fitted in the area. As number of channels increase the
length of the channel also increases. Every extra channel increases the length by 20 mm.
Increase of length relates directly proportional to pressure drop. A very high pressure
drop in smaller channels shown in the results are due to the high number of smaller
dimension channels and not due to the channel dimensions or inlet velocity. The effect
high aspect ratio of smaller dimensional channels was compromised by the length of the
channels.
There is a huge probability of flow boiling in the smaller channels due to its
length, high aspect ratio of the area and high heat flux from the processor. The range of
bottom plate thickness need to be reviewed as it has a very significant influence on the
total surface heat flux results. To find the optimum bottom plate thickness, the turning
point which results from the balancing between the heat conduction of from the bottom
surface to the top surface from the bottom surface to the two side walls has to be
studied.
5.3 Recommendation
The current nearly optimized model can be further detailed to find the optimized
model for serpentine minichannel laminar flow by reviewing the bottom plate and
channel wall thickness range and by applying lower range of inlet velocity. Parametric 84
analysis can be done to further study the individual effect of the geometric dimension
parameters. The channels should be modeled and analyzed periodically to get a better
analytical results especially for lower dimension channels
Temperature study can be done on the current model to study the existence of
the flow boiling in the current model.
Experimental study can be done based on the nearly optimized model to verify
the results gained.
The serpentine minichannel heat sink model with bottom dimension 20 mm X
20 mm can be optimized by studying the effect of turbulent flow through increasing the
inlet velocity. An optimized model of serpentine minichannels heat sink with dimension
20 mm X 20 mm can be achieved by optimization of design geometry, inlet velocity
and flow type parameters.
Heat sink model with bottom dimension 20 mm X 20 mm can be analyzed by
applying straight rectangular minichannels design .The results can be compared to the
current research to study the effect of augmentation due to the right angles with the
effect of high aspect ratio in minichannels.The results can be further optimized by
applying Dean Vortices and radius at the bend curvatures.
Dielectric liquid can be used instead of water for the analysis. Though water has
higher specific heat and higher heat of vaporization, dielectric liquid has higher
dielectric constant which reduces the unwanted current.
85
86