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 rightangle 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 microchips has led to the increase in the microchip heat fluxes. The current method of passive 1

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.

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 2

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.

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 3

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.

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 4

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

e)

Conclusion and Recommendation This chapter summarizes the important finding of work and further studies of work are presented.

CHAPTER 2: LITERATURE REVIEW 2.1 Thermal Management in Industry

Moores 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 6

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].

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 7

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. 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 todays 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 8

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. 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 9

higher speed chips. Ineffective cooling could lead to overstress in electrical components, causing the computer to fail prematurely, typically at the spot where the 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 10

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 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 m 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 11

when designing a system of micro channels: pressure drop and flow uniformity across 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

12

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.

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 13

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 1980s Tuckerman and Pease [15] introduced the concept of microchannel 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 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. 14

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 100 m > Dh > 1 mm Dh > 6 mm

Microchannels Meso-channels Conventional Passages

1 > Dh > 6 mmCompact 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 1 D > 0.1 m 10 D > 1 m 200 D > 10 m 3 mm D > 200 m D 3 mm

Nanochannels Transitional Nanochannels Transitional Microchannels Microchannels Minichannels Conventional Channels

D: Channel Diameter. For non circular channels, the minimum channel dimension is used.

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 13mm flow passage dimensions. However, it is hard to implement it 15

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 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 16

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/microchannels 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 singlephase 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 = 0.1982.01mm) 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 investigated the single- and twophase 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 17

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 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 18

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 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.

19

Figure 2.2: Schematic Layout of a Computer Water Cooling System (Source: Heat Transfer and Fluid Flow in Minichannels and Microchannels book)

2.5.1 The Heat Sink

20

Figure 2.3: Water Based Heat Sink http://www.customthermoelectric.com/Water_blocks.html

21

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 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 22

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.

23

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.

24

2.5.5 The Fan

25

Figure 2.7: Fan

http://www.clunk.org.uk/reviews/corsair-h50-cpu-cooler-review/Page-3.html

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. 26

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 27

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 Taguchis method.

28

L 16 ( 4 5 )

Number of columns Number of levels Number of rows Represent Latin squares

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 Newtons 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 study of fluid dynamics using a computer. The mathematical equations that govern the 29

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 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 30

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):

31

 QdV + FdA = 0 t
(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- 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).

32

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 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 33

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.

Heat Sink Design

Setting Heat Sink Design Parameters Range

Setting Initial Velocity Parameters Design Parameter Sampling using Orthogonal Array

CFD Analysis Design Geometry Analysis CFD Analysis Inlet Velocity Analysis

Nearly Optimized Model

Figure 3.1: Research Methodology Flowchart

3.3 Heat Sink Design Specification

< Top View >

< Isometric View >

34

Wc

Ww

< Front View >

35

Hb

Figure 3.2: Heat Sink Design Specification

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 36

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 Bottom Plate Thickness, Hb (mm) Channel Width, Wc (mm) Channel Height, Hc (mm) Channel Wall Thickness, Ww (mm)

Range 0.2 0.8 0.5 2 2.0 - 5.0 0.5 - 2.0

37

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

Re =

VDh

(3.1) Where, Re = Reynolds number; = Fluid density; V = Mean fluid velocity; Dh = Hydraulic Diameter; = Dynamic viscosity.

Table 3.2: Input Values for Reynolds Number Calculation

Re 2300 2300 Re 4000 Re 4000

Laminar flow Transitional Flow Turbulent flow

38

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 = 2(a+b) (3.2) Where, Dh = Hydraulic Diameter; a = Length of Rectangular; b = Width of Rectangular. 4ab

3.4.1.3 Initial Velocity Range

Input values based on the properties of water at 40C 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

Input Values Temperature ,T 40C Density , 991.8 kg/m3 Dynamic Viscosity , 6.55E-04 Average Velocity ,Vavg 0.1- 0.4 m/s

39

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 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

40

Levels

1 2 3 4

Channel Height Hc(mm) A 2 3 4 5

Factors Channel Wall Width Thickness Wc(mm) Ww(mm) B C 0.5 0.5 1 1 1.5 1.5 2 2

Bottom Plate Thickness Hb(mm) D 0.8 0.2 0.4 0.6

Table 3.5: Arrangement of Test Run using Orthogonal Array

Column No. Factors

1 A Channel Height Hc (mm) 1 3 2 4 1 3

2 B Channel Width Wc (mm) 2 4 4 2 3 1

3 C Wall Thickness Ww (mm) 3 1 3 1 1 3

4 D Bottom Plate Thickness Hb (mm) 2 2 3 3 4 4

Blank

Run

1 2 3 4 5 6

/ / / / / / 41

7 8 9 10 11 12 13 14 15 16

2 4 1 3 2 4 1 3 2 4

1 3 1 3 3 1 4 2 2 4

1 3 4 2 4 2 2 4 2 4

1 1 3 3 2 2 1 1 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 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

42

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.

43

Figure 3.3: Heat Sink Model Designed using Gambit

3.8.1.2 Meshing
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. 44

.
Figure 3.4: Model Meshed using Hex-Sub Map

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

45

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 , (3.6) y = Wc + Ww , (3.7) (b) At the bottom position the heat flux given : z=0, (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 , (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) -fT/z = 0 -sT/y = qw T/y = 0 T/y = 0

ii) Constant velocity of 0.1m/s is applied.

46

Figure 3.5: Bottom Plate Boundary

47

Figure 3.6: Top Wall Boundary

48

Figure 3.7: Velocity Inlet and Outflow Boundary

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 49

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
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. 50

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 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.

51

Table 4.1: Design Geometry Orthogonal Array Test Run

Column No. Factors

1 A Channel Height Hc (mm) 2 4 3 5 2 4 3 5 2 4 3 5 2 4 3 5

2 B Channel Width Wc (mm) 1 2 2 1 1.5 0.5 0.5 1.5 0.5 1.5 1.5 0.5 2 1 1 2

3 C Channel Wall Thickness Ww (mm) 1.5 0.5 1.5 0.5 0.5 1.5 0.5 1.5 2 1 2 1 1 2 1 2

4 D Bottom Plate Thickness Hb (mm) 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 52

Run

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

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.

Residual Continuity x y velocity z velocity energy

ual

sid

Re
Iteration Figure 4.1: Residual Convergence Study

53

Integral (pascal)(m2)

Iteration

Figure 4.2: Outflow Pressure Convergence Study

54

Integral (k)(m2)

Iteration

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.

55

2500
Pressure Drop (Pascal)

2000 1500 1000 500 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Meshing Size 0.2 Orthogonal Test Run Meshing Size 0.15

Figure 4.4: Pressure Drop Grid Independence Study

Surface Heat Flux (W/cm2)

700 600 500 400 300 200 100 0 1 2 3 4 5 6

Orthogonal Test Run

10

11

12

13

14

15

16

Orthigonal Test Run 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 56

converged in almost all the samples except for sample 4, 7 and 14. Sample 4 couldnt be meshed with meshing size 0.2 due to a very small channel size with non symmetrical dimensions because the wall thickness size is different from channel dimensions. It also cant 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 doesnt 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)

57

Table 4.1: Pressure Drop Calculation

Orthogonal Array Test 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Inlet Pressure (Pascal) 101487.450 101419.710 101379.180 101542.310 101496.770 101788.620 102361.830 101403.350 101839.730 101410.250 101394.100 101845.670 101389.540 101462.500 101539.480 101351.79

Outflow Pressure (Pascal) 101119.520 101253.480 101252.330 101138.230 101188.650 100756.970 100394.160 101271.960 100933.410 101217.490 101231.710 100710.580 101239.050 101231.810 101153.580 101278.42

Pressure Drop (Pascal) 367.930 166.230 126.850 404.080 308.120 1031.650 1967.670 131.390 906.320 192.760 162.390 1135.090 150.490 230.690 385.900 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.

58

Table 4.3:Total Surface Heat Flux and Pressure Drop Analysis of Design Geometry

Orthogonal Test Run 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Total Surface Heat Flux (W/cm2 ) 333.994 611.033 423.956 62.106 240.982 15.511 14.188 71.577 43.761 176.278 475.368 37.027 320.219 40.140 72.943 170.050

Pressure Drop (Pascal) 367.930 166.230 126.850 404.080 308.120 1031.650 1967.670 131.390 906.320 192.760 162.390 1135.090 150.490 230.690 385.900 73.370

700 Surface Heat Flux (W/cm2) 600 500 400 300 200 100 0 1 2 3 4 5 6 7 8 9 10 Orthogonal Test Run 11 12 13 14 15 16

Figure 4.6: Total Surface Heat Flux Analysis

59

2500 Pressure Drop (Pascal) 2000 1500 1000 500 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Orthogonal Test Run

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).

60

4.5.1 Effect of Channel Height

2000

1600

1200

800

Channel 400

Height(mm) Test Run

0 2 1 4 2 3 3 5 4 2 5 4 6 3 7 5 8 2 9 4 10 3 11 5 12 2 13 4 14 3 15 5 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. 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 61

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

2000

1600

1200

800

Channel 400 Width(mm) Test Run

0 1 1 2 2 2 3 1 4 1.5 5 0.5 6 0.5 7 1.5 8 0.5 9 1.5 10 1.5 11 0.5 12 2 13 1 14 1 15 2 16

Channel Width (mm) According to Orthogonal Test Run Surface Heat Flux (W/cm2 ) Pressure Drop (Pascal)

62

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 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.

63

4.5.3 Effect of Channel Wall Thickness

2000

1600

1200

800

400
Channel Wall T.(mm) Test Run

1.5 1

0.5 2

1.5 3

0.5 4

0.5 5

1.5 6

0.5 7

1.5 8

2 9

1 10

2 11

1 12

1 13

2 14

1 15

2 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 64

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 Thickness

65

2000

1600

1200

800

400
Bottom Plate T.(mm) Test Run

0.2 1

0.2 2

0.4 3

0.4 4

0.6 5

0.6 6

0.8 7

0.8 8

0.4 9

0.4 10

0.2 11

0.2 12

0.8 13

0.8 14

0.6 15

0.6 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) 0.2 0.4 0.6 0.8

Pressure Out (Pascal)

101254.71 101254.71 101254.71 101254.71

Pressure In (Pascal)
101419.43 101419.43 101419.43 101419.43

Pressure Drop (Pascal)

164.72 164.72 164.72 164.72

Total Surface Heat Flux (W/cm2 )

623.44 315.25 211.00 158.00

66

800 600 400 200 0 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. 67

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 Velocity(m/s) 0.1 0.2 0.3

Pressure Out (Pascal) 101254.71 101095.42 100859.9

Pressure In (Pascal) 101419.43 101623.49 101930.02

Pressure Drop (Pascal) 164.72 528.07 1070.12

Total Surface Heat Flux (W/cm2 ) 623.44 621.23 791.38 68

0.4

100592.63

102331.51

1738.88

749.78

2000 1500 1000 500 0 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

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. 69

4.7.1 Effect of Number of Channels

2000

1600

1200

800

400

0 8 1 8 2 6 3 12 4 10 5 10 6 20 7 6 8 8 9 8 10 6 11 12 12 6 13 6 14 10 15 4 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 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 70

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

Figure 4.15: Effect of Serpentine Right Angle on Total Surface Heat Flux

71

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 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.

72

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 doesnt effect pressure drop significantly .The cause of pressure drop seen are due to the length of the channel.

4.8 Summary
73

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) Channel Width (mm) Channel Wall Thickness (mm) Bottom Plate Thickness (mm) Inlet Velocity (m/S)

4 2 0.5 0.2 0.1

74

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 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 75

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. 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 76

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 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 77

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.

78