<%BANNER%>

Enhancing Heat Transport through Oscillatory Flows

HIDE
 Title Page
 Dedication
 Acknowledgement
 Table of Contents
 List of Tables
 List of Figures
 List of symbols
 Abstract
 Introduction
 Oscillatory flow along a flat...
 Mathematical models for oscillatory...
 Oscillatory flow in an off centered...
 The experiment and results from...
 Conclusions and future researc...
 Appendix
 References
 Biographical sketch
 

PAGE 1

1 ENHANCING HEAT TRANSPORT THROUGH OSCILLATORY FLOWS By JONATHAN VOLK A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

2 Copyright 2006 By Jonathan Volk

PAGE 3

3 To my grandfather, John Wolsko (1911-2006)

PAGE 4

4 ACKNOWLEDGMENTS I would first like to thank my advisor, Dr Ranga Narayanan, for his guidance during my graduate study here at the Universi ty of Florida. I appreciate all the time he has spent with me in the classroom and in the laboratory. I would also like to thank my doctoral comm ittee members, Dr. Oscar Crisalle, Dr. Alain Kassab, Dr. James Klausner, and Dr. Aaron Thomas, for taking the time to evaluate my research. I am grateful for the NASA GSRP Fellowship (grant number NNG04GO12H), which helped fund this research. I am very appreciative of Alan Stottlemyer and Nicolas Munoz, for the time they spent assembling our experiment and obtaining our results. Ken Reed and TMR engineering did a wonderf ul job machining our experiment; we were very fortunate to have them work with us. I am thankful for all the t eachers I have had since the very beginning of my schooling, I have learned so much from all of them. I would like to thank my family for all their s upport in my life. My mother and father are the greatest parents in the world, and they have shown me how to be not only a good student, but a good person. Lastly, I want to thank my grandfather, who c onstantly stressed the importance of education. He has definitely influenced my life a great deal.

PAGE 5

5 TABLE OF CONTENTS ACKNOWLEDGEMENTS 4 LIST OF TABLES .... 7 LIST OF FIGURES .. 8 LIST OF SYMBOLS .... 10 ABSTRACT 14 CHAPTER 1 INTRODUCTION 16 1.1 Physics of Mass Transport 16 1.2 Application of Mass Transport .. 19 1.3 Physics of Heat Transport .. 19 1.4 Applications of Heat Transport .................. 20 1.5 Prior Research in Oscillatory Flows .. 21 1.6 Project Objectives .. 25 2 OSCILLATORY FLOW ALONG A FLAT PLATE ... 28 2.1 Model for Oscillatory Flow Along a Flat Plate .. 28 2.2 Analysis of Flat Plate Model .. 32 2.3 Discussion Note 1: Os cillating Input in a CSTR 35 2.4 Discussion Note 2: Rod Heated Th rough Oscillatory Temperature Gradient .... 37 3 MATHEMATICAL MODELS FOR OSCILLATORY FLOW IN A TUBE ... 42 3.1 Oscillatory Flow in a Cylindrical Tube .. 42 3.1.1 Pressure Driven Flow .. 42 3.1.2 Piston Driven Flow .. 48 3.1.3 Boundary Driven Flow 51 3.2 Oscillatory Flow in an Annulus .. 53 3.2.1 Pressure Driven Flow .. 53 3.2.2 Piston Driven Flow .. 55 3.2.3 Boundary Driven Flow 55 3.3 Comparison of Annular and Cylindrical Geometries 56 3.4 Comparison of Two Compartment Flow versus One Compartment Flow 61 3.4.1 Heat Transfer ... 61 3.4.2 Mass Transfer and Separation of Species 66

PAGE 6

6 4 OSCILLATORY FLOW IN AN OFF CENTERED ANNULUS 76 4.1 Modeling an Off Centered Annulus .. 76 4.2 Analysis of Off Centered Model 83 5 THE EXPERIMENT AND RESULTS FROM THE EXPERIMENTATION 87 5.1 Experimental Set-Up 87 5.2 Experimental Method 89 5.3 Error Analysis .... 92 5.4 Note on Viscous Dissipation 96 5.5 Experimental Results and Discussion 97 5.5.1 Varying Amplitude with Water .. 97 5.5.2 Varying Amplitude with Glycerol-Water Mixture 99 5.5.3 Varying Frequency 101 5.5.4 Varying Tube Radius .. 105 5.5 Summary of Results ... 107 5.6 Potential Improvements on this Work ... 107 6 CONCLUSIONS .. 115 6.1 Conclusions .... 115 6.2 Future Research in Oscillatory Flows 118 APPENDIX CONSTANTS IN THE SOLUTIONS TO THE OFF CENTERED ANNULUS PROBLEM ........................................................................................... 120 LIST OF REFERENCES .. 123 BIOGRAPHICAL SKETCH 125

PAGE 7

7 LIST OF TABLES Table Page 5.1 Physical Properties of Wate r at Different Temperatures ... 114 5.2 Physical Properties of a Glycerol-Wat er Mixture at Different Temperatures ... 114

PAGE 8

8 LIST OF FIGURES Figure Page 1.1 Physics of mass transfer using oscillatory flows ... 26 1.2 Physics of species separati on using oscillatory flows 26 1.3 Application to species sepa ration in the environment ... 27 2.1 Design for oscillating flat plate problem ... 40 2.2 Heat flux at various Prandtl numbers as a function of Womersley number .. 40 2.3 CSTR with oscillating input .. 41 2.4 Rod with oscillatory heat source on one end 41 3.1 Oscillations driven by pressure drop .. 68 3.2 Oscillations driven by a piston ... 68 3.3 Oscillations driven by moving boundary ... 68 3.4 Oscillations in an annular geometry .. 69 3.5 Heat transport in annular ve rsus cylindrical geometry with the outer radius and pressure drop held fixed 69 3.6 Heat transport in annular vers us cylindrical geometry with the outer radius and flow rate held fixed 70 3.7 Heat transport in annular ve rsus cylindrical geometry wi th the cross sectional area and pressure drop held fixed .. 70 3.8 Heat transport in annular ve rsus cylindrical geometry wi th the cross sectional area and flow rate held fixed 71 3.9 Output/Input ratio with fixed pres sure drop and cross sectional area 71 3.10 Diagrams for one and two compartment models respectively ... 72 3.11 External piston oscillating one compar tment model and two compartment model ... 72

PAGE 9

9 3.12 Heat transport in the one vs. two comp artment model at various frequencies at small 73 3.13 Heat transport in the one vs. two comp artment model at various frequencies at medium 73 3.14 Heat transport in the one vs. two comp artment model at various frequencies at large 74 3.15 Separation ratio for various frequencies for the one compartment model and two compartment model at small ... 74 3.16 Separation ratio for various frequencies for the one compartment model and two compartment model at medium ... 75 3.17 Separation ratio for various frequencies for the one compartment model and two compartment model at large 75 4.1 Difference between centered annulus and off-centered annulus ... 85 4.2 Net change in transport due to off cen tered inner rod at various frequencies ... 85 4.3 Thin and thick regions that result fr om an annulus that is off-centered 86 5.1 Experimental set-up .. 109 5.2a Schematic diagram of heat source 110 5.2b Schematic diagram of heat sink 110 5.3 Heat transport increase as a result of piston amplitude increase for low Prandtl number fluid ... 111 5.4 Heat transport increase as a result of piston amplitude increase for high Prandtl number fluid ... 111 5.5 Heat transport increase as a result of oscillation frequency increase for high Prandtl number fluid ... 112 5.6 Large tube with separate insert, insert pl aced in large tube creating an air filled gap .. 112 5.7 Heat transport increase as a result of osci llation frequency increase for high Prandtl number fluid ... 113 5.8 Experimental ratio compared to th eoretical ratio for changing radius .. 113

PAGE 10

10 LIST OF SYMBOLS A Cross sectional area of cha nnel. Applies to Chapter 1. A Oscillation amplitude, cm. A* Scaled amplitude. Applies to Chapter 2. A Constants in solutions to perturba tion equations. Applies to Chapter 4. b Width where 99% of heat transfer occurs in oscilla ting plate problem. B Constants in solutions to perturba tion equations. Applies to Chapter 4. c Coefficient for outer radius of annulus. C Concentration of CSTR. Applies to Discussion Note 1 in Chapter 2. C Constants in solutions to ve locity and temperature equations. Cin Inlet concentration of CS TR. Applies to Discussion Note 1 in Chapter 2. C* Scaled concentration of CSTR. Applies to Discu ssion Note 1 in Chapter 2. CV Specific heat, J/(g*C). d Channel width. Applies to Chapter 1. D Molecular diffusion coefficient, cm^2/sec. g Gravitational constant. 1 i Imaginary number. J Heat flux, W/cm^2. JCOND Conductive heat flux, W/cm^2. JCONV Convective heat flux, W/cm^2. k Effective diffusion coefficient. Applies to Chapter 1. k Thermal conductivity, W/(cm*C).

PAGE 11

11 K Temperature drop per unit length. Applies to Chapter 2. K Input amplitude of driving force. App lies to Discussion Notes 1 & 2 in Chapter 2. L Length of pipe, cm. P Pressure, g/(cm*sec^2). Pr Prandtl number, dimensionless. Q Volumetric flow rate for one compartment model. Applies to Chapter 3.4 Q Volumetric flow rate for two compar tment model. Applies to Chapter 3.4. QCOND Conductive heat transfer, W. QCONV Convective heat transfer, W. QTOT Total heat transfer, W. r Radial coordinate R Radial distance, cm. S Molecular diffusivity fact or. Applies to Chapter 1. S Ratio of flow rates for one and two co mpartment model. Applies to Chapter 3.4. t Time, sec. T Temperature, C. T* Scaled temperature. Applies to Discussion Note 2 in Chapter 2. uo Maximum velocity at the ax is. Applies to Chapter 1. V Volume of the channel. Applies to Chapter 1. V Velocity, cm/sec. W Scaled time. Applies to Discussion Note 1 in Chapter 2. 2L W Scaled time. Applies to Discussion Note 2 in Chapter 2.

PAGE 12

12 2 A Wo Womersley number, dimensionl ess. Applies to Chapter 2. 2R Wo Womersley number, dimensionl ess. Applies to Chapter 3. x Coordinate in the x-direction. x* Scaled length. Applies to Discussion Note 2 in Chapter 2. X Grouping used to simplify perturbati on equations. Applies to Chapter 4. y Coordinate in the y-direction. z Coordinate in the z-direction. Z Grouping used to relate pressure dr op to piston amplitude in annulus. Greek Symbols i Viscous length grouping to assi st in expressing solutions, cm. i Conjugate of i Diffusive length grouping to assist in expressing solutions, cm. Size of solid gap in tw o compartment model, cm. Distance inner rod is displaced, cm. Applies to Chapter 4. Angle measurement, radians. Thermal diffusivity, cm^2/sec. 2 W Dimensionless parameter. Applies to Discussion Note 2 in Chapter 2. Dynamic viscosity, g/(cm*sec).

PAGE 13

13 Kinematic viscosity, cm^2/sec. Constants in solutions to perturbati on equations. Applies to Chapter 4. Density, g/cm^3. Residence time, sec. Frequency, Hz. Superscripts ^, ~ Complex conjugates. Denotes complex conjugate (overbar)

PAGE 14

14 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ENHANCING HEAT TRANSPORT THROUGH OSCILLATORY FLOWS By Jonathan Volk December 2006 Chair: Ranga Narayanan Major Department: Chemical Engineering This research project is about the use of osc illatory flows to enhance heat transport. The physical idea behind this is that if a heat source is connected to a heat sink via a fluid and the fluid is oscillated, the convectiv e motion will bring about sharp sp ikes in the velocity profile which in turn will enhance the heat transport over pure conduction due to both radial and axial gradients. This research has many applica tions including removing heat from outer space modules, reactors, closed cabins, etc. Many parameters can be varied to maximize the convective heat transport. These parameters include pulse amplitude and frequency, pipe radius and length, as well as the tran sporting fluid and the geometry of the transport zone. This problem shows from a theoretical standpoi nt how various pipe geometries affect the heat transfer. For example, an annular geometry is studied and it is seen that an annulus is an optimal geometry compared to a cylinder, when th e flow rates through both are held fixed. This happens because the annular geometry contains an extra boundary compared to an open cylinder and the spike in the velocity is greater than the open cylindrical geometry. The comparison was taken one step further using an annulus and a two compartment model containing an inner cylinder and an outer annulus. The transport in the two compartment model proved to be higher in the case of fixed flow rates, due to the additional boundary. Lastly, an off-centered annulus

PAGE 15

15 was studied, and it was shown that the net enhancement is positive for low frequencies and negative for high frequencies compared to a centered annulus. An experiment was designed to test th e various parameters listed above. The experimental results that have been obtain ed are in good qualitative agreement with the theoretical calculations. The experiment has show n that the heat transport is proportional to the square of the oscillation amplitude. The study also demonstrated that heat transfer increases with rising frequency. In summary, it has been show n through the experiment and the theory that oscillatory flows increase the heat transport of a system by several orders of magnitude compared to ordinary conduction.

PAGE 16

16 CHAPTER 1 INTRODUCTION This study is concerned with th e investigation of heat transf er in enclosed systems. The goal is to find ways to enhance the transport of he at from a heat source to a heat sink using fluid contained connectors. These methods also need to be efficient and many times faster than molecular conduction. The application of thes e findings has a place in NASAs long-term manned and un-manned missions. In a space vehicle, heat can be generated due to equipment, such as an on-board reactor. The thermal contro l of the environment is of utmost importance for the maintenance of crew habitat, or even for the sa fe operation of instruments. This heat is best conducted to external radiator sails that can then emit the heat to outer space. Getting the heat from the source to the heat sinks could be done via heat pipes and metal conductors, but the mass of such equipment is high. It is important to keep the mass on board minimal due to the immense weight during lift off and acceleration maneuvers. Th is then calls for a novel way to transport heat from high temperature source s in enclosed environments to sinks using fluid connectors. One such way is to use the principle of oscillatory flows to enhance heat transport. This involves the use of connecting pipes that contain a fluid wh ich is in perpetual motion, so that the heat can be conveyed fast. To understand how the heat ca n be transported by simply moving fluid back and forth, a companion problem will first be pr esented, which involves the mass transport of species. 1.1 Physics of Mass Transport Imagine two tanks connected by a pipe with one tank containing a species while the other tank contains a carrier. If a connecting pipe betw een the tanks were left open, a concentration

PAGE 17

17 gradient would be formed between the two tanks in the direction of the axis of the pipe (see Figure 1.1a). This gradient would ultimately decrease and equilibrium would be achieved. Now suppose one wants to quicken the rate at which the species move down the pipe. One way for this to occur is thr ough the use of oscillatory flows. When a fluid is oscillated in a pipe through some driving force, it acquires a pa rabolic profile far away from the entrance and exit regions (Figure 1.1b). This curved shape gene rates a larger area for mass transfer to take place, mostly in the radial dir ection. The species will have a net movement down the pipe and towards the boundary. These radial concentration gradients play a large role in quickening the rate of transport. In ordinary molecular diffu sion, there exist only axial concentration gradients for the species to move. However, once the oscill ations begin, the particles have more ways to move toward the end of the pipe (radially and axia lly). Since an oscillato ry flow means that the fluid must return to its original position, a reve rse parabola occurs (Figure 1.1c). In the forward part of the cycle, there was a net movement of species toward the boundary. Now since there is a higher local concentration of species at the boundary compared to the core, there will be a movement of species back toward the core of th e pipe. Although there is no net flow in this problem, the species will work its way down the pipe in a zig-zag fashion as the pulsations occur (Figure 1.1d). It can be shown that these os cillations can rapidly imp rove the rate at which the species is transported down the tube. Now take the case where there are two species with a carrier fluid in one tank and pure carrier in the other, as shown in Figure 1.2. Suppose it is desired to separate the tw o species. If the system is left without any induced flow, the sp ecies will diffuse down the pi pe at a rate that is dependent on their diffusion coefficients. Take for example two species whose diffusion coefficients are quite different from each othe r, helium and carbon dioxide. Carbon dioxide is

PAGE 18

18 the larger molecule so it will ha ve the slower diffusion coefficient. This means under the case of pure diffusion, the smaller helium molecules wi ll move down the pipe quicker and can be separated out first (see Figure 1.2a ). The carbon dioxide would then follow later. However, if oscillations are imposed, one will notice that something different occurs (see Figure 1.2b). As mentioned above, the pulsations will push the species toward the boundary of the tube. The species most likely to be pushed to the boundary is the faster diffuser, the helium. The carbon dioxide will remain behind in the center of the tube When looking at the ve locity profile in the tube, one will notice that the flui d speed is faster in the center compared to the boundary. This means that the carbon dioxide will actually move faster down the tube sin ce it is in the quicker flow region. The helium can be thought of as getting stuck near th e slow moving boundary region. A separation is achieved with the slow er diffuser being transported faster. This separation will also occur much faster compared to the case where only pure diffusion happens, since the oscillations force the two sp ecies down the pipe at faster rate. In order for a separation to be maximized, se veral parameters must be optimized. In all chemical engineering problems, there exist time scales. This one is no different. In fact, multiple time constants arise that must be note d. These time constants involve the oscillation frequency, the kinematic viscosity of the carrier fluid, as well the diffusion coefficients of the species. For example, the oscillation period must be at a rate that is comparable to the time constant due to the kinematic viscosity of the flui d, so that a parabolic prof ile can exist. If the fluid has a low kinematic viscosit y, the frequency must not be too high, or the inertial force from the pulse would overpower the viscous effects at the walls, causing the velocity profile to be much flatter, and reducing the con centration gradient in the radial direction. In fact, this suggests the existence of a dimensionless group which is a ratio of the tw o time constants. The analysis

PAGE 19

19 will show the natural occurrence of such a gr oup, the Womersley number. Another example is that the diffusion time constant for each specie s must be different. Otherwise, both species would move down the pipe at the same rate, and a distinct separa tion would not occur. Lastly, the viscous time scale must be greater than the di ffusive time scale, since the viscous effects play a big role in producing the radial concentration gradients. This is because a more viscous fluid can generate a sharper parabolic flow profile due to the higher amount of friction at the tube walls. A longer and sharper parabola produces mo re radial concentration gradients which will increase the transport in the tube compared to ordinary diffusion. 1.2 Application of Mass Transport A basic application for enhanced species transport is shown in Figure 1.3. In this example, it is important for human systems to have clean air to breathe. When humans generate carbon dioxide as they exhale, it becomes mixed in with air. Th is mechanism could assist in removing the carbon dioxide from the air so that it is pure when it is regenerated by plant photosynthesis. Imagine if there is a carbon dioxide and air mixtur e in the left hand tank, while the right hand tank consists of only water vapor. When the osci llations are induced, the slower diffusing carbon dioxide molecules will reach the right hand tank faster, and are able to be separated out first, leaving the air free from high levels of carbon dioxide. 1.3 Physics of Heat Transport The same idea of pulsatile flows can be applie d to transporting heat between a source and a sink. Take the case of a fluid in a channel. Su ppose at one end, the flui d is in a hot tank, while at the other end, it is in a cold sink. Heat w ill be transported from the source to the sink by conduction if the fluid remains stagnant; this heat will flow down th e channel at a rate proportional to the thermal diffusivity of the fluid. If the fluid is oscillated back and forth, radial

PAGE 20

20 temperature gradients will form (see Figure 1.1b) and the heat movement would be enhanced down the pipe through the zig-zag movement discu ssed earlier. The oscill atory movement of the fluid transports the heat through convection, and when the appropria te parameters are chosen for the system, meaning the type of fluid and th e geometry, the rate of convection can be significantly higher than the rate of conduction. Like in th e mass transfer analog, several time scales exist. For the heat transport problem, th e time scales will involve the thermal diffusivity, along with the kinematic viscosity and oscillation frequency as before. The first time scale involves the thermal di ffusivity of the fluid medium. This rate determines how effectively heat is transported through conduction. Just like in the mass transfer example, the viscous time scale (determined by the kinematic viscosity of th e fluid) needs to be much larger than the diffusive time scale, so th at heat can be transported in multiple directions through convection via the radial temperature gradie nts, as opposed to only axially, which occurs when the diffusive time scale dominates. The time constant for the osci llation frequency must again be similar to the viscous time scale in order to ensure th e formation of a sharp, parabolic spike in the velocity profile. 1.4 Applications of Heat Transport In the beginning of this chapter, the applic ation of oscillatory flows for use in enclosed space capsules was described. One of the main advantages of this method was the limited mass required. The mass can be kept minimal since there is no net flow. The fluid to fill the channel is all that is needed, and can reused throughout. This mechanism can be run with simple fluids such as water and air. No volatile or hazardous chemicals are needed, which reduces the number of potential safety risks on boar d. Compare this to a heat pi pe, where vaporizing fluids are sometimes used, and the pulsatile flow m odel seems fit for use in space capsules.

PAGE 21

21 There are many relevant uses on Earth for this model as well. Many industrial processes and appliances generate excess heat that need s to be removed. These include exothermic reactions such as in polypropyl ene and polybutylene production, as we ll as in fuel cells. This would be useful for heat removal from hazardous s ubstances such as radioactive fluids. One of the main goals of this mechanism is to remove heat quickly and efficiently so that it can be useful in many different types of processes. 1.5 Prior Research in Oscillatory Flows The roots of this problem can be traced back to Taylor [13] and his work in dispersion. He showed through his work that a soluble subs tance in a long and narrow tube spreads out due to two effects: molecular diffusion and velocity gr adients in the radial di rection. He developed a model for dispersion that contai ns a diffusion coefficient whic h is now known as the Taylor diffusion coefficient. This effective diffusion value is D u R ko 2 2~ (1.1) Here R is the radius of the tube, uo is the maximum veloc ity at the axis, and D is the molecular diffusion coefficient. The important thing to notice is that this effective diffusion coefficient is inversely proportional to the molecula r diffusion coefficient. This parallels the idea that at appropriate oscillating frequencies, th ere will be a reverse in the separation, where the species with the smaller molecular diffusion coef ficient will have the higher effective diffusion coefficient and separate out quick er. It can also be seen from (1.1) that the best separation achievable when the fluid is oscillated is propo rtional to the molecular diffusion coefficient of the small molecule divided by the molecular di ffusion coefficient of the large molecule. Aris [1] later used the method of moments to see if Taylors resu lt could be reproduced. Taylor had to restrict certain pa rameters for his conditions to be valid, such as the dimensions of

PAGE 22

22 the tube and the velocity of th e fluid through the pipe. Aris me thod eliminated those constraints by modeling the solute based on its moments in th e direction of the flow. Through this method, Aris was able to obtain the Taylor diffusion coef ficient. He also demonstrated how dispersion can be altered by changing the geometry of the cross section. He was able to show that there is less dispersion in an elliptical tube compared to a circular tube of the same area. This will be important in the modeling chapter when convection in various geometries is discussed. Bowden [2] also applied Taylors idea to show that this pr ocess occurs in large scales as well. In the sea for example, there is horizontal mixing due to a shearing current. This shearing current is an alternating, or osci llating, flow. He was able to show th rough experimental observations how the effective coefficient of horizontal diffusion is inversely proportional to the vertical eddy diffusion coefficient. Harris and Goren [4] developed a model for oscillatory mass transfer in an open tube using an oscillating piston to drive the flow. Th ey were able to show through this model that the enhanced species transport is proportional to th e square of the piston amplitude. Their model was also developed to show how much the os cillations increase the mass transport over pure molecular diffusion. They were able to show how their model produces Aris result for the Taylor diffusion coefficient when the frequenc y of oscillations and tube radius are small compared to the kinematic viscosity and the molecular diffusion coefficient (low Womersley numbers). They also conducted an experiment to test their model us ing HCl. The results were in good agreement with their theory as they showed that the mass transport rate of HCl can be increased by as much as sixty times the rate of pure diffusion under the proper tuning conditions. A similar model to Harris and Gorens is used in a later chapter to develop the heat transfer model for piston driven flow.

PAGE 23

23 Watson [17] later studied how mass flux increased in a circular pipe and a two dimensional channel when the fluid was oscillate d. He concluded that the effective diffusivity was increased from the molecular diffusivity by a factor of S where 2) ( ~ Ad V S (1.2) Here is the frequency of the oscillations, V is the volume of the channel, A is the cross sectional area of the channel and d is the width of the channel. An important point to note in this result is that the enhanced transport is proportional to the square root of the frequency. This will be seen in the derivation of the heat transfer model. Kurzweg did extensive work in this field with Jaeger [7,8] and Zhao[9]. They did studies in both mass and heat transfer applications. Kurzweg and Jaeger performed experiments to verify Watsons relation for the effective diffusi on coefficient under oscillatory conditions. They were able to obtain a good data correlation with Watsons theory which shows that the transport increases with the square root of the frequency. In this study, they also found that there exists an optimal frequency for a particular system (fluid medium and dimensions of tube), at which maximum transport will take place. Their results show that as frequency is increased starting from rest, there will be an increase in transpor t until it reaches the optimal tuning frequency, after which point it will decay. They also presented a study in high frequency oscillations in which the longitudinal dispersion coefficient was proporti onal to the square of the spike amplitude (as in Harris and Goren), and to the first power of frequency. They determined that the dispersion under oscillatory conditions can be more than twice as much compared to steady turbulent flow. Kurzweg and Zhao [9] developed an experiment in heat transfer to enhance the thermal diffusivity of a fluid. They connected a hot a nd cold reservoir with a capillary bundle containing several small tubes. The oscillations were imposed by shaking a flexible membrane on each

PAGE 24

24 tank. They were able to enhance the therma l transport in water by almost 18,000 times its molecular thermal diffusivity. It was also show n in their theoretical model that oscillating a liquid metal such as lithium can achieve a heat fl ux that is two orders of magnitude greater than those obtained in liquid metal heat pipes. Others have studied the capi llary and tube configurations such as Inaba et al. [5] and Peattie and Budwig [11]. They found that the tube or capillary geometry does affect the overall heat transport, and needs to be considered when designing a model. In se parate studies, the input power required to oscillate a flui d in a heat transfer configura tion has been examined by Nishio et al. [10] and Furukawa [3]. Nishio examined the optimum ratio of h eat transport to input power and found that the ratio can be increased when there is a phase shift between the oscillations in adjacent tubes within the cap illary bundle. Furukawa developed algebraic expressions for several output parameters as a function of the Womersley number, including input power, and the number of capillary tubes needed to achieve a desired transport rate. Thomas and Narayanan [15,16] were able to compare the differences between the various types of driving forces used to produce oscillations. They dete rmined that pressure driven oscillations produced a greater convective ma ss transport compared to the case where the moving outer wall of the tube produced the pulsa tions. It was also f ound that the ratio of convective transport to input power was iden tical in both cases except for extremely low frequencies where the boundary driven flow had a higher ratio. They also studied cross over frequencies for separations in pulsatile flow, meani ng frequencies at which the two species switch in being the faster and slower transporter. It was determined that as many as three cross over frequencies can arise in certain binary ga seous systems. Thomas [14] also conducted several experiments to show that these cross over frequencies existe d. He used different binary

PAGE 25

25 gaseous mixtures and was able to induce th e pulsations through a moving boundary. The separations were measured through the use of a ga s chromatograph. In addition to finding the cross over frequencies, he also validated Kurzweg et al. that an optimal frequency existed, at which the separation would be at a maximum. All of this prior work has assisted in formulating the objectives for this project, where further prog ress can be made in usi ng oscillatory flows to enhance heat transport. 1.6 Project Objectives This project has both theore tical and experimental compone nts. From a theoretical standpoint, the oscillatory flow model is derive d after showing some si mpler problems which will give the reader a better understanding of the physics. In certain cases, previous theoretical results will be verified. The m odel is presented in three different ways since the oscillations can be driven in various methods. The model is al so discussed using variou s geometries which may or may not enhance the transport. These different geometries include a centered annulus, an off centered annulus, and a two comp artment model involving both a cy linder and an annulus. The objective here is to determine what geometry pr oduces the optimal transport from a theoretical perspective, as well as to see if that result changes depending on what the input parameters and conditions are. An experimental set-up was constructed to test the basic cylindrical model for heat transport. The testing was conducted by varying ke y input parameters such as the amplitude of the spike, the frequency of the os cillations, the dimensions of the t ube, and the type of fluid. The experimental results were compared to the theore tical model to see if they are in good agreement with each other. Also, in the case in which th e experimental results deviate from the model, possible reasons are discussed. This experiment showed that heat can be transported more

PAGE 26

26 effectively through the use of oscillatory flows, a nd can therefore be applied to removing heat in the situations described earlier, whether it is on a large or a small scale. Figure 1.1: Physics of mass transf er using oscillatory flows; (a) pure molecular diffusion in tube, (b) forward pulse in os cillatory flow with molecules movi ng toward the boundary, (c) reverse pulse with molecules moving toward the core, (d ) end result with molecules moving down at a faster rate compared to pure diffusion. Figure 1.2: Physics of species separation using oscill atory flows; (a) molecular diffusion with smaller molecules moving down tube faster (b) osci llatory flow with larger molecules remaining in the faster core region, resu lting in faster transport comp ared to the smaller molecules

PAGE 27

27 Figure 1.3: Application to specie s separation in the environment

PAGE 28

28 CHAPTER 2 OSCILLATORY FLOW ALONG A FLAT PLATE In order to understand the physic s of the oscillating flow in a pipe, a simple case will first be discussed involving an os cillating plate in an infinite medium of fluid. The oscillatory nature of the plate motion is mathematically expressed as a cosine wave function. Figure 2.1 illustrates this design. 2.1 Model for Oscillatory Flow Along a Flat Plate The oscillating plate is assumed to m ove in the horizontal direction. A represents the amplitude of the oscillating stroke, and is the frequency of the plate motion. The velocity profile in this case depends only on y . The Navier-Stokes equation in the y-direction becomes 2 2y V t V (2.1) The boundary conditions are given by no slip at the plate ( y =0), meaning the fluid will have the same velocity as the oscillating plate (1/2 A cos( t )), and that at an infinite distance away from the plate, the velocity of the fluid will approach zero. The simplest way to solve this equation is to express the velocity field as a sum of two exponentials, that is t i t ie V e V V ~ ^ (2.2) Since one term is a conjugate of the other, the algebraic manipulations are simplified and the problem can be solved usi ng either term, meaning that 2 ^ 2 ^y V V i (2.3) The solution to this equation can be found to be

PAGE 29

29 y i y ie C e C V 2 1 ^ (2.4) It can easily be seen from the boundary condition far away from the wall that C1 must be equal to zero since the velocity must be zer o. The boundary conditi on at the wall can be rewritten using the exponential relati onship for a cosine wave, which is 2 cos i ie e (2.5) Therefore the new boundary condition at the wall will be A y V 4 1 ) 0 (^ (2.6) C2 can now be solved for, and the result is y ie A V 2 ) 1 ( ^4 (2.7) The entire expression for velocity can be f ound by taking twice the real part of (2.7). Therefore ) 2 cos( 2 12y t e A Vy (2.8) This method of solving the differential equation a nd the result is useful in the next step of this problem. The velocity field is used in findi ng the temperature profile. As in the equation of motion, only the temperature change in the y-direction will be obtained. Hence 2 2y T V L T t T (2.9) The temperature field can be rewritten as a sum of complex expone ntials much like the velocity. Therefore t i t ie T e T T ~ ^ (2.10)

PAGE 30

30 The temperature equation now becomes 2 ^ 2 ^ ^y T V L T T i (2.11) The T/L term is the temperature change in the horizontal direction and can be assumed to be a constant in this case ( K ), while is the thermal diffusivity. The upper boundary condition is similar to the velocity field, where at an infi nite distance from the plate, the heat flux tends to zero. The lower condition would be that there is no heat flux through the wall ( dT/dy =0). The solution becomes y i y i y ie i KA e C e C T 2 ) 1 ( 2 ) 1 ( 4 2 ) 1 ( 3 ^) 1 1 ( 4 (2.12) Here C3 is zero due to the far-field boundary condition. The wall boundary condition can be used to solve for C4 yielding ) )( 1 1 ( 42 ) 1 ( 2 ) 1 ( ^ y i y ie e iKA T (2.13) Equation (2.13) can be used to solve for the actual temperature profile in (2.10) giving )] 2 sin( ) 2 sin( [ ) 1 ( 22 2y t e y t e KA Ty y w (2.14) The last quantity that needs to be found is the total heat flux. This can be expressed as the sum of the diffusive flux and the convective flux. The flux that is found is averaged over time and space to give b v bdtdy T V b C dtdy Kk b J0 2 0 0 2 0) ( 2 ) ( 2 1 (2.15)

PAGE 31

31 In this expression, k represents the thermal conductivity of the fluid, is the density of the fluid, Cv is the specific heat of the fluid, and b represents a certain distance from the oscillating plate. It is determined so that 99% of the heat transfer takes place within a region of width b . When (2.2) and (2.10) are inserted into (2 .15), and the time integral is taken, only the cross terms will remain since the other terms integrat e out to zero. This will be explained later in further detail. The equation for flux becomes ] ) ( Re[ 20 ~ ^dy T V b C Kk Jb v (2.16) Using the results found in (2.7) and (2.13), performing the integration and rearrangement produces an expression for the heat flux. It is )} 2 2 ( )] 2 2 cos( ) 2 2 ( ) 2 2 sin( ) 2 2 [( { ) 1 ( 8) 2 2 ( 2 2 b b b b e KA C Kk Jb V (2.17) The flux can also be expressed in dimensionless variables. This will be beneficial in analyzing the results and understand ing the role that these different time scales play in the physics of the problem. Define the Womersley and Prandtl numbers to be 2 A Wo (2.18) and

PAGE 32

32 Pr (2.19) The amplitude can be made dimensionless as well. Thus b A A* (2.20) The conductive heat flux is Kk JCOND (2.21) Equation (2.17) can now be expressed as a ra tio of convective flux to conductive flux. This ratio is 1 )} Pr ( )] Pr cos( ) Pr ( ) Pr sin( ) Pr [( { ) Pr 1 ( Pr 82 / 1 * 2 / 1 2 / 1 * 2 / 1 2 / 1 / ) Pr ( 2 2 / 1 ** 2 / 1 Wo Wo A Wo A Wo Wo Wo A Wo A Wo Wo Wo e A J JA Wo Wo COND CONV (2.22) 2.2 Analysis of Flat Plate Model A simple analysis can be performed on (2.22) to see what the flux should be for extreme values of the dimensionless terms. In the simple case where the plate is not oscillating, the ratio equals 1, which makes physical sense as the only transport that occurs is through conduction. Another case to examine is when the Prandtl num ber becomes much greater than 1. Taking the limit as Pr goes to infinity produces a result that is ) 1 8 (* Wo A J JCOND CONV (2.23) This approximation can be useful in estim ating the flux for fluids with high Prandtl numbers (such as oils). A physical interpretation for this case is to look at the problem when the kinematic viscosity is much larger than the ther mal diffusivity (the definition for a high Prandtl

PAGE 33

33 number). Looking at the equation for the ve locity field (2.8) and taking the limit when tends to infinity, the result is the same as the velocity of the plate. This means that every particle of fluid moves at the same velocity as the plate. In th is case there wont be so me fluid elements moving with the forward stroke and others on the backwa rd stroke; instead they will all move together. This promotes a higher rate of heat flux due to convection as opposed to conduction. This is what would be expected since the thermal diffusi on coefficient is extremely low compared to the kinematic viscosity. It can also be noted from this equation that the ratio is proportional to the square of the amplitude (observe that the amplitude appears in the Womersley number as well). The frequency of the oscillations, the amplitude of the stroke, and the physical properties of the fluid are all input variables in this case, while the convective to conductive ratio is the output result. The equation for an infinite Prandtl nu mber fluid (2.23) can be shown to be accurate by looking at Figure 2.2, which uses (2.22) to show heat flux for various Prandtl numbers. It can be seen that for the higher Prandtl number fluid ( Pr =500), the heat flux appro aches the graph for Pr tending to infinity. This same line is the plot of (2.23). It also makes sense since for high Pr fluids, the heat flux increases line arly with the Womersley number fr om (2.23). It is obvious that a linear plot results in Figure 2.2. Therefore, it can be concluded that (2.23) is an accurate representation of (2.22) for an infi nite Prandtl number. Although it can be seen in Figure 2.2 that the convective heat flux increases with increasing Prandtl number, it should be pointed ou t that it only increases for a given Womersley number, and that kinematic visc osity appears in the denominator of the Womersley number. When the kinematic viscosity (or Prandtl numbe r) becomes high, the c onvective transport only increases if there is an increase in the amplitude or frequency of the plat e to keep the Womersley number constant. This makes phys ical sense since a fluid with a hi gh viscosity is very sluggish.

PAGE 34

34 Although the fluid elements are more likely to move with the velocity of the plate as stated above, there is more energy required by the system to shear a very viscous fluid back and forth. This extra energy would come from increasing the size and/or frequency of the oscillations. In fact, there will be a decrease in the transport if only the kinematic viscosity is varied and nothing else. This becomes an issue during experimentation, and is discussed more in Chapter 5. Lastly, the flux value can be found when the Prandtl number approaches 1. This will help in determining a result for those fluids w ith low Prandtl numbers such as gases. Taking (2.22) as a limit when Pr approaches 1, we get ) 1 ) ) 2 ( ( 32 (* 2 2 ** Wo Wo A Wo e A J JA Wo COND CONV (2.24) This equation shows that for a low Prand tl number fluid, the c onvective to conductive ratio will be lower than the case for a very high Prandtl number fluid simply by the fact that the ratio in (2.24) roughly s cales by a factor of e-2/32 compared to 1/8 in (2.23). This would make sense since a lower Prandtl number fluid can mean it has a high thermal diffusivity; such as in the case of air. Not only will this drive up th e conductive heat transfer, but since gases are lighter and have a lower viscosity than liquids, they will feel less of th e shearing effect brought about by a moving plate. This means that the mo lecules will be less likely to move with the speed of the plate, and instead move at a rate pr oportional to the thermal diffusivity of the gas. This means less transport via convection. For both of these reas ons, the convective to conductive ratio will decrease. Although simple from a physical standpoint, the mathematics of the problem can be somewhat complicated. It is im portant to understand not only the physics of the problem, but the mathematical techniques as well, since they will be applied in more detail to the scenario involving oscillating flow in a pi pe. Before that point is re ached, two other example problems

PAGE 35

35 will be presented to show how the length and time scales relate to the physics of the problem. These discussion notes are very basic chemical engineering problems, however a lot can be obtained from them. A key idea to take from these problems is to see how an input function (in these cases an input oscillation) affects the be havior of a system. Also, it is important to understand how these input functions affect the system response depending on the size and type of the input signal, as well as th e size of the system. The first problem presented will involve a simple CSTR with an oscillating input. It can be easily solved from a mathematical standpoint and clearly shows the relationship between input signals and system response. The second problem is a little more complicated mathematically but still shows the same relationship. This problem is one where a solid rod is heated at one end in an oscillatory manner. Since the first example relates to mass transfer, and the second relates to heat transfer, these problems should help the reader conclude that these systems behave in a similar way, regardless of the transport medium. 2.3 Discussion Note 1: Os cillating Input in a CSTR Figure 2.3 shows a constant stirred tank reacto r that is receiving i nput concentration as a sinusoidal oscillation. Writing the mass balance for the CSTR, we get dt t dC t C t Cin) ( ) ( ) ( (2.25) where C is the concentration in the tank, and is the residence time which is equivalent to the volume of the tank divided by the volumetric flow rate. K is a concentration amount that represents the magnitude at which mass enters the tank. The inlet concentration is expressed as a sinusoidal wave to give dt t dC t C t K ) ( ) ( ) sin( (2.26)

PAGE 36

36 The initial condition for this problem is that the concentration in the tank is zero when time is equal to zero ( t =0). The problem can be scaled where t (2.27) W (2.28) and K C C* (2.29) The scaled form of (2.26) is written d dC W C* *sin (2.30) The differential equation (2.30) can be easily solved using an integr ating factor to obtain We C W W W W C 1 2 2 2 *) 1 1 ( cos ) 1 1 ( sin (2.31) C1 is a constant that can be found base d on the scaled init ial condition that C *=0 at =0. The final expression for C can be found after some rearranging. It is 2 *1 cos sinW We W CW (2.32) This equation describes the response of the sy stem to the input parameters. First take the case of a very small frequency, which will cause W to become very small. If the limit of (2.32) is taken as W tends to zero, the result will be that the concentration tends to sin which is the behavior of the input wave. This makes sense if the inertia is small enough; there will be little resistance in the tank to change the response in the tank. If W were increased, it can be seen from (2.32) that the response of the tank will behave as a combination of sine and cosine waves.

PAGE 37

37 Now that the inertia has become large, there will be more of a resistance from the tank, causing a time lag. Note that the volume of the tank appears in the constant W When W is small, the volume of the tank can be considered small. This means that there is less room for the input signal to break up into sine and cosine pa rts, which is the case for small W For large W the volume of the tank can be large, so the signal can become different by th e time it permeates throughout the tank. Being able to examine the output response of a system based on setting the input parameters is important for understanding the beha vior the system. This will be critical when the oscillating flow in a pipe model is presented in the next chapter. One more example will be shown to further discuss these concepts. 2.4 Discussion Note 2: Rod Heated Through Oscillatory Te mperature Gradient A rod is heated at one end via an oscillatory heat source and kept constant at the other end in Figure 2.4. The one-dimensional energy equation is established to be 2 2dx T d dt dT (2.33) Here represents the thermal diffusivity of the rod material. The problem can be scaled in a way similar to the previous example where t (2.34) L x x / (2.35) K T T/ (2.36) and 2L W (2.37)

PAGE 38

38 Again W can be considered to be the input, alth ough the characteristic length now is the length of the rod. A similar term will arise in the oscillatory flow model in a tube. The energy equation can be written in terms of comp lex exponentials as in (2.10), that is i ie T e T T ~ ^ ^ (2.38) Scaling the problem, we get 2 * ~ 2 ~dx T d T W i (2.39) The solution for this equation can now be found to be ) sin( ) cos(2 1 ~W i C W i C T (2.40) The boundary conditions can be found in Figure 2.4; however the condition at the right hand end of the rod can be rewritten 2 ) 1 (* ~i x T (2.41) since i e ei i2 sin (2.42) The time independent part of the temperature field is now ) sin( 2 ) sin(* ~W i x W i i T (2.43) The complete temperature field can now be found. It is ] ) sin( ) sin( Re[* ie W i x W i i T (2.44) After taking the real pa rt of (2.44), and some tedious algebra, the final result of the temperature field is completed. Hence

PAGE 39

39 } cos )] cosh( ) sin( ) cos( ) sinh( ) cos( ) sinh( ) cosh( ) [sin( sin )] cos( ) sinh( ) cos( ) sinh( ) cosh( ) sin( ) cosh( ) {[sin( )] cos( ) [sinh( )] cosh( ) [sin( 12 2 x x x x x x x x T (2.45) where 2 W (2.46) When W is taken to be small, it can be seen from (2.45) that cos terms will disappear, leaving the temperature response in the rod to behave as sin just like the input. Obviously, as W is increased, the temperature field will behave as a mixture of sine and cosine terms. Again, this is similar to the prev ious discussion note. Since W contains the characteristic length L a small value of W can mean a small rod length. The s horter length keeps the input sine wave from breaking up all the way to the end of the rod. Again, a larger characteristic length will cause a break up in the response to the inertial force, resultin g in a combination of response terms. Although these problems appear si mple, the length and time s cales that are generated in them can tell you a lot about the physics in the problems. Hope fully these two examples have clarified the concept of input signa l relating to system behavior. It should also be noted that the oscillating plate problem presented earlier in this chapter behaves in the exact same way. The Womersley number in the problem is quite similar to W in the CSTR and solid rod examples. When the Womersley number is very small, the system behaves in the same way as the input signal, as cos( t ). When the Womersley number grows, the system will behave differently, thus there is a time lag that arises. When the osc illatory flow model is presented, one may notice

PAGE 40

40 similarities with these examples. There are many ways to impose the oscillations in the model, and depending on what the input signals are, the output response from the other parameters will behave in a way that is consistent with these prior examples. When the signal is small, the system will respond proportionally to the signal. A larger signal (which will occur in the oscillatory flow model) will cause a mixture of ou tput responses from the system. Now that the reader has a better understanding of these basic problems, the oscillatory flow model can be presented. Figure 2.1: Design for oscill ating flat plate problem Figure 2.2: Heat flux at va rious Prandtl numbers as a function of Womersley number

PAGE 41

41 Figure 2.3: CSTR with oscillating input Figure 2.4: Rod with oscill atory heat source on one end

PAGE 42

42 CHAPTER 3 MATHEMATICAL MODELS FOR OSCILLATORY FLOW IN A TUBE Consider a model of heat transfer enhancement by oscillatory flow where the fluid resides in a cylindrical tube c onnected between a heat source and a heat sink. In order to oscillate the fluid in the tube, one of many methods may be used. The first is to oscillate the fluid by imposing an oscillatory pressure gradient across the ends of the tube. The second way is to push the fluid back and forth with an extern al mechanical piston. A third method is to move the wall of the tube back and forth, whence the shearing of the fluid can also cause the oscillations. The models for these three methods are first discussed for a circular cylindrical geometry. 3.1 Oscillatory Flow in a Cylindrical Tube 3.1.1 Pressure Driven Flow To model this case, write the Na vier Stokes equations. Therefore g P V V V dt dV 2) ( (3.1) Several assumptions will be made to simplif y this momentum balan ce. Fully developed flow is assumed in the axial (z) direction only. This means that the pressure gradient is only assumed to be in the axial direction and gravitati onal effects are assumed to be negligible. This reduces (3.1) for a cylindrical geometry, hence z P r V r r V t Vz z z ) 1 (2 2 (3.2) This equation can be rewritten as P r V r r V t Vz z z 1 12 2 (3.3)

PAGE 43

43 where z P P 1 (3.4) To solve for the velocity profile, it is conveni ent to break the velocity field into a sum of time complex exponentials from the start, that is t i z t i z ze V e V V ~ ^ (3.5) Likewise, P becomes t i t ie P e P P ~ ^ (3.6) Substituting (3.5) and (3.6) into (3.3) yields ^ ^ 2 ^ 2 ^1 P r V r r V V iz z z (3.7) The common exponential terms are factored out. In addition, since th e two terms in (3.5) and (3.6) are complex conjugates of each other resp ectively, only one of the two terms needs to be solved. In the case of an oscillating pressure drop being an input, the value must be real; therefore ~ ^ P P (3.8) The solution for the velocity field is 2 ^ 0 2 0 1 ^) ( ) ( P r Y C r J C Vz (3.9) where i (3.10)

PAGE 44

44 When (3.10) is multiplied by the characteristic length, which in this case is the radius of the tube, one can obtain a dimensionless term known as the Womersley number. This happens when iWo R i R 2 (3.11) This group is similar to those developed in the Discussion Notes in Chapter 2. Note here that the length scale is the radius of the tube, which is different from the case in Discussion Note 2. In that example, the inertial term was scal ed with the axial direct ion, as the heat through conduction moved down the rod. In this model, the critical length scale is now the tube width. This is the case because the inertial effect comes from moving the molecules in the radial direction. The tube width determines the magn itude of radial temperature gradients that can arise, and these gradients are what enhance the heat transport through convection. In addition to that, the tube length is assumed to be much great er than the tube width. This means that the effects of the gradients will be more significant over the tube width compared to the effects of the axial gradients over the tube length. This is why the radius of the tube is the important length scale for this problem. The boundary conditions arise from observing that there is zero velocity at the wall ( r = R ) and a finite velocity at th e center point of the tube ( r = 0). It can be seen in (3.9) that C2 is equal to zero from the second boundary conditi on since the Y-Bessel function is unbounded at r = 0. C1 can be found by appl ying the condition at R The expression for the velocity field is now found to be ] ) ( ) ( 1 [ ) (0 0 2 ^ ^R J r J P r Vz (3.12)

PAGE 45

45 The solution for zV~ is simply the complex conjugate ofzV^. This result is then used to solve the energy equation, which is ) 1 (2 2r T r r T z T V t Tz (3.13) Here, is the thermal diffusivity. The axial temperature gradient (dT/dz) is assumed to be constant since entrance and exit effects are ne glected in this model. The viscous dissipation term will be neglected as well. The same method can be applied to express the temperature field as a sum of complex exponentials, hence t i t ie T e T T ~ ^ (3.14) Using the corresponding term in th e velocity field, (3.13) is now ) 1 (^ 2 ^ 2 ^ ^r T r r T L T V T iz (3.15) The solution is ) ( ) ( ) (0 6 5 0 4 0 3 ^r J C C r Y C r J C T (3.16) where i (3.17) C3 and C4 are constants that can be found from the boundary conditions. C5 and C6 are constants from the particular solution. Observe that the boundary conditions derive from the fact that the temperature is finite at the center of the tube (r = 0) and that there is no heat flux through the walls of the tube (dT/dr = 0). Once the constants are found, the temperature field solution is now complete. It is

PAGE 46

46 ] P r 1 ) ( ) ( ) ( ) ( ) ( ) ( Pr 1 [ ) (1 0 1 0 1 0 0 1 2 ^ ^ R J R J r J R J R J r J L T Pi r T (3.18) where Pr (3.19) The Prandtl number (Pr) is important because it compares the two time scales used in this problem, the viscous time scale and the thermal diffu sive time scale. The next task is to find the total heat flux which is equal to the conductive h eat flux plus the convective heat flux. This is T V C L T k Jz v (3.20) Here is the density and Cv is the specific heat. In order to evaluate the total heat transfer rate, (3.20) must be integrated over th e cross sectional area and time averaged over one cycle to get 2 00) ( 2 2 rdrdt T V C L T k QR z v TOT (3.21) The conductive term is a simple expression a nd can be easily evaluate d. Inserting this into (3.21) we obtain 2 00 22 2R z v TOTTrdrdt V C R L T k Q (3.22) When substituting (3.5) and (3.6) for the velo city and temperature fields respectively, we get 2 00 ~ ^ ~ ^) )( ( 2 2R t i t i t i z t i z v CONVrdrdt e T e T e V e V C Q (3.23)

PAGE 47

47 When the time integral is evaluated, only the cross terms will remain since the other two terms integrate out to zero. Therefore 2 0 2 0 2 20 dt e dt et i t i (3.24) so R z z v CONVrdr T V T V C Q0 ^ ~ ~ ^) ( 2 (3.25) Since these terms are complex conjugates of one another, they can be combined and twice the real part of the remaining term can be taken to get the result for the convective heat transfer, which is ] ) ( 2 Re[ 20 ^ ~R z v CONVrdr T V C Q (3.26) Evaluating the integral and combining it w ith the conductive part results in the final expression for the total heat transfer, and after some algebraic manipulation we get ] 1 ) ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( Re( ) 1 (Pr Pr 4 [0 4 1 1 0 0 3 0 1 1 2 2 2 2 ^ R J R J J R J R J J R J R J R P Q QCOND TOT (3.27) where 2R L T k QCOND (3.28) and (3.29)

PAGE 48

48 In order to show the enhancement of heat tr ansfer due to the oscillations, (3.27) can be expressed as a ratio of convective heat tran sport to conductive heat transport to yield ) ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( Re( ) 1 (Pr Pr 40 4 1 1 0 0 3 0 1 1 2 2 2 2 ^R J R J J R J R J J R J R J R P Q QCOND CONV (3.30) A key thing to note in (3.30) is how the probl em scales with the square of the pressure drop. This is because the velocity and temperatur e fields scale with the pressure drop. Since the convective heat transport is base d on the product of the velocity a nd temperature fields, it should come as no surprise that the convective heat trans port is proportional to th e square of the driving force. Another term in (3.30) that is important is the group where the Prandtl number appears. This term gets smaller as the Prandtl number grow s. This is consistent with what was discussed in Chapter 2. A large Prandtl number can imply a very viscous fluid. More energy is required to drive a more viscous fluid; therefore, less heat will be transported c onvectively if the input energy is held constant. One ma y wonder what happens when the Pr andtl number is less than 1, which is usually the case for gases. This term would become negative under those circumstances. The convective to conductive ra tio will still be positive since the Bessels functions terms will account for the sign discrepanc y. When the thermal diffusivity is greater than the kinematic viscosity, the real part of the Bessels functions group will be negative, keeping the overall ratio positive. All of these re lationships will still exist even when the driving force is altered in the next case, where an oscillating piston is used. 3.1.2 Piston Driven Flow

PAGE 49

49 If a mechanical piston is used to drive the oscillations in stead of an imposed pressure drop, the model does not change very much. Supp ose the position of the piston behaves as a sine wave, that is t A xsin 2 1 (3.31) Then the speed of the piston would be t A dt dx cos 2 1 (3.32) The momentum balance of course will not chan ge. However, in order to get (3.12) in terms of piston amplitude, another relationship n eeds to be used. Since the integral of the velocity profile over the cross sectional area is equal to the volumetric flow rate, we get t A rdr VR z cos 2 1 20 (3.33) Also (3.33) can be re-written as A rdr VR z4 1 20 ^ (3.34) where we have used 2 cost i t ie e t (3.35) Using (3.34), the pressure term can now be found in terms of amplitude to yield ] ) ( ) ( 8 4 [0 1 2 ^R J R R J A P (3.36) The velocity field is now in terms of the i nput amplitude as (3.36) is substituted into (3.12) to give

PAGE 50

50 ] ) ( ) ( 8 4 ) ( ) ( 1 [ ) (0 1 0 0 ^R RJ R J R J r J A r Vz (3.37) The rest of the derivation is exactly the sa me as before with the exception of this one substitution. The temperature field is now given by ] ) ) ( ) ( 8 4 )( Pr 1 ( ) ( ) ( ) ( ) ( ) ( ) ( Pr 1 [ ) (0 1 1 0 1 0 1 0 0 1 ^R RJ R J R J R J r J R J R J r J L T C Ai r Tv (3.38) The total heat transfer in terms of input amplitude is } 1 ] 8 ) ( ) ( 4 ][ 8 ) ( ) ( 4 [ ] ) ( ) ( 4 ) ( ) ( 4 [ ) 1 (Pr Pr {2 2 1 0 2 2 1 0 1 0 1 0 2 2 2 2 R R RJ R J R R RJ R J R RJ R J R RJ R J R A Q QCOND TOT (3.39) Expressed as a convective to conductive ratio, we get } ] 8 ) ( ) ( 4 ][ 8 ) ( ) ( 4 [ ] ) ( ) ( 4 ) ( ) ( 4 [ ) 1 (Pr Pr {2 2 1 0 2 2 1 0 1 0 1 0 2 2 2 2R R RJ R J R R RJ R J R RJ R J R RJ R J R A Q QCOND CONV (3.40) This is the result obtained by Harris and Goren [4] in their model for pulsating flow, the only difference being the dimensi onless Prandtl number for heat transfer replaces the Schmidt number (Sc) for mass transfer in their model. Again the driving force (amplitude) appears to the second power in the final heat tr ansfer equation. This occurs for the same reasons as in the

PAGE 51

51 pressure driven flow case. The last method, wh ere the boundary is oscill ating, will simplify the initial equations somewhat, but a sim ilar solution will still be obtained. 3.1.3 Boundary Driven Flow Suppose now that the oscillations are gene rated through the outer wall moving back and forth in the axial direction with a velocity of 1/2* A cos( t ). The momentum balance will be the same as (3.3), however there will not be a driving force term from the pressure drop as the motion is driven by the boundaries. The solution is solved in the same way as before, except there is only a homogeneous solution, hence ) ( ) (0 2 0 1 ^r Y C r J C Vz (3.41) The boundary condition at the center of the tube is the same as before, meaning C2 is equal to zero, and using (3.35), the velocity at the wall is equal to A R r Vz4 1 ) (^ (3.42) C1 can now be solved for and the velocity profile is ) ( 4 ) ( ) (0 0 ^R J r J A r Vz (3.43) The temperature profile can be solved as before. The boundary conditions remain the same as in the other cases; this means that we have finite temperature at the center and no heat flux through the wall. Therefore ] ) ( ) ( ) ( ) ( ) ( ) ( [ ) Pr 1 ( 4 ) (0 1 0 1 0 0 1 ^R J R J r J R J R J r J L T Ai r T (3.44) The total heat transfer is determined in the same way as before to get

PAGE 52

52 } 1 ] ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( [ ) Pr 1 ( 16 {0 0 1 2 2 0 1 1 0 1 2 R J R J R J R J R J R J R J R A Q QDIFF TOT (3.45) and the ratio which is ] ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( [ ) Pr 1 ( 160 0 1 2 2 0 1 1 0 1 2R J R J R J R J R J R J R J R A Q QCOND CONV (3.46) Although these three methods of oscillating have a different driving force, the one thing that appears in all three models is that the overal l heat transfer is proporti onal to the square of the driving force. For example, if the amplitude is doubled, then the heat transfer will be quadrupled. This relationship was tested experimentally and will be discussed in Chapter 5. The relationship of input signal to output re sponse can be seen in each of these cases as well. It was shown in the Discussion Notes in Chapter 2 that the output response depends on the size of the input signal and the si ze of the system. In looking at the final expression for heat transfer in each of these exam ples, it is evident that the system response is a combination of terms (Bessels functions) unlike the input signal s. For example, when the wall oscillates as a cosine function, the velocity a nd temperature profiles will behave as a combination of sine and cosine functions (expressed as Bessels functions fo r a cylindrical geometry). Just like in the last chapter, the break up of the input signal will occur when the Womersley number reaches a certain limit. The specific value is not relevant he re, but it is important to see that the oscillatory flow model has a simple input signal, and increasi ng that signal will resu lt in a different output response, no matter what th e means of oscillation are. The open cylinder is a common configuration for a connecting tube. However, one might imagine that potential improvements can be ma de by increasing the number of boundaries, which would in turn increase the amount of radial te mperature gradients, by ge nerating more velocity

PAGE 53

53 spikes. A simple configuration that comes to mind th at meets this criterion is that of an annular tube case to which we now turn. 3.2 Oscillatory Flow in an Annulus Notice from Figure 3.4 that the inner bounda ry generates an additional spike when viewing this cross section. The method for the annular flow model is exactly the same as the cylindrical flow model with the exception of the inner boundary condition. Although this is the only difference, this will make the result much mo re complicated. The three variations presented above will also be shown here. 3.2.1 Pressure Driven Flow The solution for the momentum equation is id entical to (3.9). However, with the new boundary condition at the inner wall, C2 is no longer zero. The boundary condition at the inner wall ( r = R ) and the outer wall ( r = cR c >1) are now the same, that is the velocity equals zero. The solved velocity field is now } ) ( ) ( ) ( ) ( )] ( ) ( )[ ( ) ( ) ( ) ( ) ( )] ( ) ( )[ ( 1 { ) (0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 ^ ^R Y cR J cR Y R J cR J R J r Y R Y cR J cR Y R J R Y cR Y r J P r Vz (3.47) As in the case of the velocity field, th e inner and outer wall boundary conditions on temperature are the same, meaning there is no heat flux through either wall (therefore dT/dr = 0 at these walls). The solution for the temperature field is similar to (3.16 ), with an additional particular solution that is proportional to the Y-Bessels function. Therefore ) ( ) ( ) ( ) (0 5 0 4 3 0 2 0 1 ^r Y C r J C C r Y C r J C T (3.48) C3, C4, and C5 are found from solving for the particular solution, while C1 and C2 are determined from the boundary conditions all yielding

PAGE 54

54 } )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( [ { ) (1 1 1 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 2 2 ^ 1cR Y R J R Y cR J cR Y R Y R Y cR Y R Y cR J cR Y R J cR J R J cR Y R J R Y cR J cR Y R J R Y cR J R Y cR J cR Y R J R Y cR Y L T P C (3.49) } )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( [ { ) (1 1 1 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 2 2 ^ 2cR Y R J R Y cR J cR Y R J R Y cR J R Y cR J cR Y R J cR J R J cR Y R J R Y cR J cR J R J R J cR J R Y cR J cR Y R J R Y cR Y L T P C (3.50) L T P C 2 2 ^ 3 (3.51) )] ( ) ( ) ( ) ( [ )] ( ) ( [ ) (0 0 0 0 0 0 2 2 2 ^ 4R Y cR J cR Y R J R Y cR Y L T P C (3.52) and )] ( ) ( ) ( ) ( [ )] ( ) ( [ ) (0 0 0 0 0 0 2 2 2 ^ 5R Y cR J cR Y R J cR J R J L T P C (3.53) The expression for the total heat transfer is equal to cR R z TOTrdr T V c R L T k Q^ ~ 2 22 Re 2 ) 1 ( (3.54) zV~ is the conjugate of zV^. Since the expression is now quite complicated, MAPLE is used to further evaluate (3.54). The next case for annular flow, the osci llating piston problem, is

PAGE 55

55 quite similar to the oscillating pressure drop case The same method to relate the pressure drop model to the piston model for a cylinder will be us ed now, with a change in the integration limits to account for the geometry difference. 3.2.2 Piston Driven Flow The solution for the piston driven problem is the same as the pressure driven problem with the exception of one substitution. By insert ing (3.47) into (3.34) and solving for the scaled pressure, we get Z c R A P 8 ) 1 (2 2 2 ^ (3.55) where } )] ( ) ( ) ( ) ( [ )] ( ) ( )][ ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( )][ ( ) ( [ 2 ) 1 ( {0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 1 2 2R Y cR J cR Y R J cR J R J R Y cR cY R R Y cR J cR Y R J R Y cR Y R J cR cJ R c R Z (3.56) (3.47) through (3.53) can now be found in terms of piston amplitude and the final result can be determined using (3.54). 3.2.3 Boundary Driven Flow For the problem where the outer wall is oscillating, the bound ary condition for the momentum equation will be the same as (3.42). The velocity will be zero at the inner wall. Using these to solve for the velocity profile, we get )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ 4 ) (0 0 0 0 0 0 0 0 ^cR J R Y R J cR Y r J R Y r Y R J A r Vz (3.57) There is no heat flux through either wall, so the temperature solution is ) ( ) ( ) ( ) ( ) (0 4 0 3 0 2 0 1 ^r Y C r J C r Y C r J C r T (3.58) where

PAGE 56

56 )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ )] ( ) ( ) ( ) ( [ {1 1 1 1 1 1 1 1 4 1 1 1 1 1 1 1 1 3 1cR Y R J R Y cR J cR Y R Y cR Y R Y C cR Y R J R Y cR J R Y cR J cR Y R J C Co (3.59) ) ( ) ( ) ( ) (1 1 3 1 4 1 1 2R Y R J C R Y C R J C C (3.60) ) ( ) ( ) ( ) ( ) ( ) Pr 1 ( 40 0 0 0 0 1 3cR J R Y R J cR Y R Y L Ti A C (3.61) and ) ( ) ( ) ( ) ( ) ( ) Pr 1 ( 40 0 0 0 0 1 4cR J R Y R J cR Y R J L Ti A C (3.62) The overall heat transfer can be found by solv ing equation (3.54). In summary, the input parameters for this model are the amplitude and frequency of the driving force (either the pressure drop, the piston, or moving wall), the le ngth and radius of the connecting tube, the fluid used, the geometry of the tube (cylindrical or annular), and the axial temperature drop. Once all these parameters are inserted in to the appropriate model, the conductive, convect ive, and total heat transport can be calculated. Now that th e cylindrical and annular models have been presented, it is important to determine which geometry is more optimal, and under what conditions. 3.3 Comparison of Annular and Cylindrical Geometries Four cases are discussed to determine whet her the cylindrical or annular geometry is preferable for enhanced heat transfer. In the fi rst case, the heat trans port in an open tube is compared to one in an annular geometry whose ou ter radius is held equa l to the radius of the open tube. As the inner rods radi us is decreased, one expectedly recovers the results for an open tube. The comparison between th e two geometries is made hol ding the pressure drop per unit

PAGE 57

57 length fixed. In a companion calcu lation, the volumetric flow rate in a half cycle is held fixed instead of the pressure drop. While an idea regardi ng the effect of the walls can be obtained, the comparison between the two geometri es is not entirely fair becaus e the cross-sectional areas of the two geometries are not kept equal. Thus two other calculations th at mirror the first two cases are examined with the only difference being that the cross-sectiona l area of the cylinder and annulus are now held fixed. The results are telling. The Prandtl number used in the calculations corresponds to water and is set to a value of 7. The physical properties of water were used when evaluating the dens ity, viscosity, thermal conductivity, and the specific heat. The other parameter values held constant in all the calculations are the following: tube length of 10 cm, a frequency of 0.1 Hz, and an axial temperature difference of 10 deg C. These values ar e of little concern as the qualitative nature is what is important. Figure 3.5 depicts the ratio of the transport in the annulus to the transport in the cylinder while the inner radius of the annul us is varied. A value of 1 cm for the outer radius was used in the first two sets of calcu lations. It is observed that as th e inner radius is decreased in the annulus, the values of the heat transport come clos e to the open tube resu lt. One is led to the conclusion that the open tube cylindrical geometry is a more optimum design compared to an annulus. The reason for this conclusion is that th e available region for heat transfer decreases as the inner radius increases. This means that for a fixed pressure drop, as the size of the inner rod increases, the resistance to flow increases, and the sharpness in velocity spikes decrease, leading to a lower heat transport. The transport decreas es from the open tube even when the radius of the inner rod is small. This is expected as the resistance to flow increases dramatically when a thin rod is introduced leading to a decrease in he at transfer. Likewise, as the inner rod diameter

PAGE 58

58 gets larger, the annular geometry asymptotica lly becomes a thin gap, and the heat transport slowly approaches zero. In the first case, the pressure drop was held fixed, but if the flow displacement in one half-cycle was held fixed, and the comparison was made between the two geometries, something different may occur and that is indeed the case. As there is now a smaller cross sectional area in the annular region than the open cylinder, the pre ssure drop will have to increase in the annulus so that the flow rates can be the same in both case s. The results of the ca lculations are depicted in Figure 3.6. The calculations depicted in the figure now show that the heat transport is enhanced with an annular geometry compared to an open tube provided that the flow rate per half-cycle is assumed equal in both geometries. The flow rate was calculated by integr ating the velocity over the flow area for one half-cycle to give 02cR R zrdrdt V Flowrate (3.63) The pressure drop (which is contained in the so lution for the velocity field) in the annulus was then changed so that the flow rates woul d be equal. On calcul ating the pressure drop increase in the annulus as the inner radius increa ses, it is evident that this is a reason why the transport increases in the annular case. Another contributing f actor to the increase in the pressure drop is due to the resistance of the i nner rod. The wall friction can lower the flow rate, so the pressure drop must increase to overcome th at resistance and make the flow rate the same as the cylinder. For the cases where the inner ra dius of the annulus is high, the transport in the annulus becomes roughly 15 times higher than the corresponding cylindrical case. The pressure drop is related to the amplitude of the spike, meaning that a higher pressure drop in the tube will

PAGE 59

59 result in a greater amplitude of oscillation, a higher velocity spike, and hen ce, greater transport. This is why the annulus is the more optimal method in the case of fixed flow rates. In the first two cases, the area changes as th e radius of the inner rod was changed. This raises the question as to whethe r the comparison between a cylindr ical and an annular geometry is a fair one. To make a fair comparison, consider the case wher e the cross-sectional areas and the pressure drop are held equal for both annular and open tubes. This means that as the inner radius of the annular geometry changes, the cross-sectional area of the companion open tube must also change so that both c onfigurations will have the same amount of area available for heat transport. The results for this comparison are gr aphically depicted in Fi gure 3.7 for a definite frequency. The calculations show that the heat transport is agai n greater in the open cylinder, just like the first case. It a ppears to have an advantage over th e annular case by 1-2 orders of magnitude for smaller cross-sectional areas. Again, if the flow rate is held fixed, for the same cross-sectional area, we see that the heat transfer favor s the annular configuration. The issue with the case of fixed flow rates, when the areas were not held equal to each other, was that the lower cross-sectional area in the annulus drives up th e pressure drop needed in order to have the flow the same as the cylindric al case, and consequently leads to larger spikes in the flow profile. However, in the last set of calculations the cross-sectional areas are the same for the open tube and annulus, and therefore this f actor is now not an issue. The question then is: why does the annular geometry still give a higher tr ansport? The reason is that in the last case, as before, the pressure drop increases because the inner rod offers resistance on account of wall friction. This factor continues to remain importa nt even where the crosssectional area is fixed and made equal. One will notice in looking at Fi gure 3.8, that the ratio at high inner radii is still large (about 8 times), although not as large as be fore (15 times). This makes physical sense

PAGE 60

60 since the cross-sectional area discrepancy is now eliminated in Figure 3.8, leaving only the inner rod inhibitor to play a role in driving up the pressure drop. From the calculations it appears that the ope n tube geometry is best when the pressure drop is held fixed, but the annular geometry is better when the volumetric flow rate is held fixed. The ultimate message is that the optimal case de pends on what one holds fixed, pressure drop or flow rate. Another factor in determining which geometry is best is to see what system gives a higher heat transport output per unit of mechanical power inpu t. The power only takes into account the amount needed to cause the oscillator y fluid movement. The power needed to move an oscillating piston of a given ma ss for example is not incorporated as this is subject to the materials used. The mechanical power is calcul ated using the following relationship, which is 2 0) ( 2 2cR R zrdrdt pressure V Power (3.64) For the case of a cylinder, the lower limit of integration in r will be zero. In order to obtain the pressure in terms of P from (3.64), it must be converted to units of pressure. Therefore 2 02 2cR R zLrdrdt P V Power (3.65) (3.5) and (3.6) can be in serted into (3.65) giving 2 0 ~ ^ ~ ^) )( ( 2 2Lrdrdt e P e P e V e V Powert i cR R t i t i z t i z (3.66) The integration with respect to time can be easily done (see ( 3.24)) leaving a final expression for power which is

PAGE 61

61 Lrdr P V PowercR R z ~ ^2 Re 2 (3.67) To demonstrate a comparison we will first ta ke the case of fixed cross sectional area. The case where the pressure drops are held fi xed produce Figure 3.9. There is no need to consider the case of fixed flow rates because the results are identical to Figure 3.9. These sets of calculations show that the ratio of transport/input power is higher for the open tube case. Again, the annular case results approach the cylindrical values when the inner rod gets small. The reasoning for the input power be ing greater for the annular ca se is because there are two boundaries in the annulus compared to just one for the open tube. The extra boundary means there is more viscous effects. This means more energy is required to oscillate the fluid since it now must overcome the viscous effects at two wa lls as opposed to just one. This must be considered when determining the geometry for the most efficient heat tran sport. If the amount of input power is a concern, then that ne eds to be factored into the decision. In summary, four methods have been worked out to compare the heat transport in an open cylinder versus an annulus. Although the resu lts can differ depending on the method of choice, the results of each method make physical sense. It can be concluded from this study that the annular model would be optimal if the system is ab le to handle a large pr essure drop, and if input power is not a concern. If not, then the open cylinder would be the geometry of choice. 3.4 Comparison of Two Compartment Flow versus One Compartment Flow 3.4.1 Heat Transfer It is possible to make another comparison involving cylindrical and annular geometries. Ranger [12] proposed a model for steady Poiseuille flow through concentric pipes with various areas of cross sections. He determined that it was possible to have two different geometric

PAGE 62

62 models which have the same cross section and th e same flow rate, although each model contains a different number of compartments. For exam ple, the one compartment model consists of a large annulus, while the two compartment model ha s an annulus and a cylinder. Note that the outer radii of both models ( R4) are set equal to each other. In order to make the cross sectional area of both models equal, a constr aint needs to be used. It is 2 2 2 3 2R R R (3.68) In the case of the one compartment model, the flow rate is ] ) ln( 2 1 ln ) ( ][ [ 82 2 2 3 4 2 2 2 3 2 4 2 2 2 3 2 4 2 2 2 3 2 4 'R R R R R R R R R R R R P Q (3.69) Note that (3.68) was substituted for R where appropriate. For the two compartment model, the flow rate is ] ) / ln( ) ( [ 83 4 2 2 3 2 4 4 3 4 4 4 2R R R R R R R P Q (3.70) The derivation for (3.68) through (3.70) wa s done by Ranger for Poiseulle flow and can be found in [12]. A ratio can now be es tablished rela ting the two flow rates as 'Q Q S (3.71) When S=1, the flow rates become equal. This means that for a given value of R4, R3 and R2 can be adjusted so that the flow rates are equa l. Pressure drop and viscosity will not play a role since they factor out of (3.71). There exists a value of R2 where S=1 for a set R3 value (again assuming R4 is held fixed). In order to examin e the behavior of the one compartment system versus the two compartment system, certain relationships need to be developed for use in the oscillatory flow model. Fo r this study, assume that the pist on driven model is used in both

PAGE 63

63 the one and two compartment models. It is im portant for this problem that the pistons are external to the channels. See Figure 3.11 for a visual depiction. The shaded areas represent the flow regions. The external pistons in both cases are assumed to oscillate with the exact same amplit ude and frequency, and they both have the same cross sectional area since the outer radius (R4) of the one and two compartment models are assumed to be the same. This means that the volumetric flow rate going into the one compartment system will be equal to that go ing into the two compartment system. The only issue that now arises is the break-up of th e flow into each section of the two compartment model. Since the models were derived earlier for the cylinder and the annulus, the amplitude needs to be known for each case so that the trans port can be found in both sections separately. The transport in each section is simply added t ogether to obtain the total transport for the two compartment model. The oscillation amplitude and frequency of the external piston is assumed to be set. The volume displacement equation is used to obtain the oscillation amplitude in the annular channel of the one compartment model. Thus ) (2 2 4 1 2 4R R A R Aa ex (3.72) Aex is the amplitude of the external piston and A1 a is the amplitude inside the annulus (the on each side of the equation cancel). A1 a can be solved for since ev erything else in (3.72) is known. The transport in the annulus for the one compartment model can now be obtained by using the equations for piston driven flow in an a nnulus described earlier. In order to obtain the amplitude in the cylinder and annulus for the two compartment system, two relationships are needed to solve for the two unknown amplitudes. The first says that the total flow rate is equal to the sum of the flow rates in th e cylinder and the annulus. Therefore

PAGE 64

64 2 2 2 2 3 2 4 2 2 4) (R A R R A R Ac a ex (3.73) A2 c and A2 a are the oscillation amplitudes inside the cylinder and annulus of the two compartment model respectively. Since the volu metric flow rates are being used in this equation, note that the oscillat ion frequency and the constant factor out of each term in (3.73). The second relationship involves th e resulting pressure drops gene rated by the moving piston in each compartment. Since the piston moves exte rnally, the resulting pressure drop in each compartment must be the same. This makes physical sense since there will be only one pressure drop from the beginning of the pipe to the end of the pipe. This means that the right hand sides of (3.36) and (3.55) can be se t equal to one another, yielding ] ) ( ) ( 8 4 [ 8 ) (2 0 2 2 1 2 2 3 2 4 2R J R R J A Z R R Ac a (3.74) The expression for Z can be found in (3.56) where the outer radius is equal to R4 and the inner radius is equal to R3 instead of cR and R, respectively. There are now two equations for these two unknown amplitudes (A2 a and A2 c), meaning they can now be solved. Comparisons were made at various frequencies for se veral different values of a parameter This is the width of the annular solid (no flow) region in the two compartm ent model (equal to R3 minus R2). Since the cross sectional areas are the same for each model fo r the flow regions, the regions where there is no flow is also identical between the one and two compartment systems, because R4 is the same for both cases. This means a lower corresponds to more available room for flow. The problem was solved using a numerical solver, with water as the fluid medium, a pipe length of 50 cm, a temperature differ ence of 10 C, an outer radius (R4) of 1 cm, and an external piston length of 8 cm. The results are given as a ra tio of the heat transfer in the one compartment

PAGE 65

65 model to the heat transfer in the two compartment model for various sizes of (0.1, 0.3, and 0.5 cm). In all three instances, the heat transport in the two compartment model is greater than the one compartment model. This makes sense by un derstanding what is going on with the physics of the problem. In the two compartment model, th ere are a total of three boundaries in the cross section (the outer edge of the annulus, the inner edge of the a nnulus, and the outer edge of the cylinder). There are only two boundaries in the one compartmen t model (the outer and inner edge of the annulus). Keep in mind that the total flow rates must be the same through each model. Since there is no flow at the boundaries, and very little flow right ne xt to the boundaries (the parabolic spike in the veloc ity profile approaches zero near the walls), there will be an extra region of very little flow in the two compartment model sin ce it has one more boundary. This will cause a local decrease in the flow rate in th ese regions. To account for the extra region of slow flow, the amplitude of the spike must grow proportionately, so as to keep the overall flow rates the same for each model. It has been s hown earlier in this chapter that increasing the amplitude in the flow profile drives up the heat tr ansport by the square of the amplitude increase. This is why the two compartment model is mo re favorable, since the extra boundary will produce larger spikes in the flow profile. The trend for the curves in Fi gures 3.12, 3.13, and 3.14 is for the one compartment model to approach the transport of th e two compartment model, then tail off as frequency increases. This is due to the sharp spike forming in the wide annular region of the one compartment model as frequency grows, then the profile begins to fl atten and the transport be gins to plateau off as the frequency becomes large. The cylindrical region will maintain a sharper velocity profile for

PAGE 66

66 a longer period of time since there is no inhibiting r od in the middle. This will cause the ratio to decrease at high frequency. As the value increases, it can be seen that th e maximum ratio approaches 1, which is when the transport in both models are equal, but it never reaches 1. Note that the maximum points on the curves seem to appr oach 1 asymptotically, so when becomes 1 cm, the models will have the same transport. This makes sense because if the solid (no flow) region encompasses the entire area enclosed by the outer wall, both models will have the same amount of heat transport going thr ough it, which is zero. A comparison has been made between two geom etries to determine which transports heat better through it, which in this case is the two compartment model. These models can also be studied to figure out which provi des a better separation between two species. 3.4.2 Mass Transfer and Separation of Species The same models can be compared to determine which transports mass more efficiently. Since the mathematics of the models for heat and mass transport are virtually the same, the obvious conclusion can be reached that the two co mpartment model is better since it transports heat at a higher rate, which is indeed true. Let us take this one step farther to see how effectively each model can separate out a mixture of two species in a carrier fluid. This problem was first presented in the Ch apter 1. For this study, helium and carbon dioxide will be separated in a nitrogen carrier. Starting from rest, the helium molecules will separate out faster than the car bon dioxide since helium has a larger diffusion coefficient. When the oscillation frequency becomes high enough, the carbon dioxide will emerge first as the small helium molecules get trapped near the boundaries in the slow flow regions. Eventually, there will be an optimal frequency for which the separati on ratio is at a maximum. This is consistent

PAGE 67

67 with the results of Kurzweg and Jaeger [8]. Th e same parameters were used for this calculation as in the heat transfer calcul ation discussed above, with the ex ception of the fluid medium. The results are again presented for different values of The separation ratio in the figures represents the mass flow rate of carbon dioxi de divided by the mass flow rate of helium. Note in Figure 3.16 that the dotted line at the e nd of the curve represents an es timate due to inaccuracy of the machine performing the calculation. Since th is region occurs well after the maximum is achieved, it is not significant. The maximum separation ratio that is achieved in the one and two compartment models is about the same in all three cases. It can also be seen how the size of the peaks grow when increases, from about 2.5 for small to about 3.8 for larger This is because the amount of solid (no flow) space grows with Since the external piston amplitude remains the same, the amplitudes of the spikes in the tubes must grow to satisfy (3.72) and (3.73). Even though the maximum ratios are similar in each case, one big difference is at what frequency the maximums occur. In each case, the highest ratio occurs at a much lower frequency in the two compartment model co mpared to the one compartment model. Roughly 3 to 4 times the optimal frequency of the two compartment syst em is needed to achieve the optimal frequency for the one compartment system. The reasoning for this is very similar to the heat transfer analog. Since there are more boundaries in the two compartment model, there are more slow flow regions. This means that th e helium molecules can move quick ly into the slow part of the flow near the walls, since there is more availabl e space. The separation will occur faster since the helium molecules become quickly trapped, whic h means the larger carbon dioxide molecules will have more room in the faster moving core region, and therefore transport quickly down the tube.

PAGE 68

68 Achieving the separation at a lower frequenc y is beneficial since it limits the amount of power needed to drive the system. The input powe r needed to oscillate the fluid increases with increasing frequency. Therefore the same sepa ration can be produced with either model, but much less energy is being used in the two comp artment system. The conclusion can be reached that the two compartment model is the more op timal geometry, whether heat or mass is being transported, or if a separation is desired. One la st consideration will be made in the next chapter regarding the geometry of the sy stem, that being what happens wh en the inner rod of an annulus is moved to an off centered position. Figure 3.1: Oscillations driven by pressure drop Figure 3.2: Oscillatio ns driven by a piston

PAGE 69

69 Figure 3.3: Oscillations driven by moving boundary Figure 3.4: Oscillations in an annular geometry Figure 3.5: Heat transport in a nnular versus cylindrical geometry with the outer radius and pressure drop held fixed

PAGE 70

70 Figure 3.6: Heat transport in annul ar versus cylindrical geometry wi th the outer radius and flow rate held fixed Figure 3.7: Heat transport in a nnular versus cylindrical geometry with the cross sectional area and pressure drop held fixed

PAGE 71

71 Figure 3.8: Heat transport in a nnular versus cylindrical geometry with the cross sectional area and flow rate held fixed Figure 3.9: Output/Input Ratio with Fixed Pressure Drop and Cross Sectional Area

PAGE 72

72 Figure 3.10: Diagrams for one and two compartm ent models respectively (flow regions shaded) Figure 3.11: External piston oscillating one compartment model (top) and two compartment model (bottom).

PAGE 73

73 Figure 3.12: Heat transport in the one vs. two compartment m odel at various frequencies at small Figure 3.13: Heat transport in the one vs. two compartment m odel at various frequencies at medium

PAGE 74

74 Figure 3.14: Heat transport in the one vs. two compartment model at various frequencies at large Figure 3.15: Separation ratio for various freque ncies for the one compartment model (dashed line) and two compartment model (solid line) at small

PAGE 75

75 Figure 3.16: Separation ratio for various freque ncies for the one compartment model (dashed line) and two compartment model (solid line) at medium Figure 3.17: Separation ratio for various freque ncies for the one compartment model (dashed line) and two compartment model (solid line) at large

PAGE 76

76 CHAPTER 4 OSCILLATORY FLOW IN AN OFF CENTERED ANNULUS In the previous chapter we show ed that oscillatory flow in an annular geometry leads to an enhancement of transport over an open geometry under certain conditions. Another case to examine is when the inner cylinder in the annular geometry is moved to a slightly off-centered position (see Figure 4.1). A model of this problem will first be presented, followed by a calculation to see what option is be tter, and under what circumstances. 4.1 Modeling an Off Centered Annulus Let the deviation in distance of the inner rod from the center be small and be denoted as To get the relation between the radial distance in the distorted geometry with the radius in the reference geometry, it is seen that 2 0 2 2] [R Y X (4.1) Converting this into polar coordinates we get 2 2 0 2cos 2 R R R (4.2) R can be expressed in terms of to get the mapping from the reference to the current configurations. To get a better understandi ng of the mapping equations, see Johns and Narayanan [6]. Thus ) ( 2 1 ) ( ) (0 2 2 0 1 0 R R R R (4.3) Here R1 and R2 are determined from th eir definitions yielding 0 1cos ) 0 ( d dR R (4.4) and

PAGE 77

77 0 0 2 2 2 2sin ) 0 (R d R d R (4.5) These results will be useful in obtaining th e boundary conditions at the displaced inner surface. For the distorted geomet ry, let the driving force come from the inner and outer walls oscillating in the axial direction, both movi ng in phase, and whose speed is expressed as 1/2*A cos( t). Using the mapping of the inner boundary the velocity and temperature profiles can be obtained at the various orders. The expression for the lowest order (i.e., (0)) is obtained in a manner as done earlier. Sin ce the zeroeth order probl em is for the case of the centered inner rod, this simply becomes the boundary driven problem in an annulus shown in Chapter 3, with the exception that the boundary condition at the i nner wall is now equal to the condition at the outer wall. The velocity field is now (the solved constants are listed in the Appendix ) ) ( ) (0 0 0 0 0 ^r J B r Y A Vv v Z (4.6) Note that the same method is used here to express velocity as a sum of time complex exponentials. Solving the first order problem will be different from the base case, since velocity will become a function of in addition to the radial positi on. The first order equation is 0 1 12 0 1 ^ 2 2 0 1 ^ 2 0 1 ^ 0 2 1 ^ 2 d V d r V dr V d r dr V dZ Z Z o Z (4.7) At first order, the boundary condition at the ou ter wall yields the firs t order velocity field to be zero. This is true becau se there is no displacement at th e outer surface, so even though the outer wall is moving at fixed amplitude, there are no first order terms, implying that the velocity is zero at the outer bounda ry to first order or any order higher than zero. For the inner boundary, the surface is displaced, so a mapping must be used yielding

PAGE 78

78 ... )] ( ) ( ) ( ) ( ) ( ) ( 2 ) ( [ 2 1 )] ( ) ( ) ( [ ) (0 0 0 0 2 0 2 0 0 2 0 2 1 0 0 0 1 0 1 0 0 2 2 0 0 0 0 1 0 0 1 0 0 0 R dr dV R R dr V d R R dr dV R R V R dr dV R R V R V VZ Z Z Z Z Z Z Z (4.8) In order to obtain the boundary conditions at the inner wall for each order greater than zero, set the terms of the desi red order in (4.8) equal to ze ro and obtain a condition for the velocity at the wall for that particular order. Note that these boundary conditions stay the same when the velocity is expressed as a sum of e xponentials since there are no time dependent terms in the boundary conditions. At first order, the i nner wall boundary condition is 0 00 0 ^ 0 0 1 ^ 1 0 0 1 ^cos ) (R Z R Z Zdr V d dr V d R R V (4.9) The dependence is seen to be proportional to cos in order to satisfy differential equation (4.7) and the inner wa ll boundary condition (4.9). The solution would then be cos )) ( ) ( (0 1 1 0 1 1 1 ^r J B r Y A Vv v Z (4.10) The differential equation for O( 2) is the same as (4.7 ) with the exception of Vz 2 replacing Vz 1 in the notation. The second order boundary condition is 0 0 00 0 ^ 0 0 2 2 0 0 ^ 2 0 2 0 1 ^ 0 0 0 2 ^sin cos cos 2 ) (R Z R Z R Z Zdr V d R dr V d dr V d R V (4.11) The complete result for the second order velocity field is 2 cos )] ( ) ( [ )] ( ) ( [0 0 0 2 0 2 2 ^r DJ r CY r J B r Y A Vv v Z (4.12) This form of the solution makes sense sin ce all three terms in (4.11) can be broken up into dependent and independent terms through the use of the double angle formulae. For this

PAGE 79

79 problem, only the independent terms are taken into account. The reasoning for this will be explained later. This completes the calculation of the velocity field in the problem. These expressions are used to solve the temperature field equations. The method for solving the temperature field equations is the same as before. Each order in of the temperature profile is solved using the corresponding velocity profile (i .e., the first order temperature field uses the first order velocity solution). The temperature is broken up into a sum of two exponential terms just like the velocity field case. The boundary co nditions for the base case is that there is no flux through the walls ( dT/dr =0). For the base case, the temper ature field is identical to (3.56). Therefore ) ( ) ( ) ( ) (0 0 0 0 0 0 0 0 0 ^r J r Y r J B r Y A Tj y T T (4.13) The first order problem for the temperature field has the same form as (3.15) plus a dependent term. It is 1 ^ 2 0 1 ^ 2 2 0 1 ^ 0 1 ^ 0 2 0 1 ^ 21 1ZV L T d T d r T dr T d r dr T d (4.14) In a manner used for treating the boundary cond itions for the velocity field, we have the temperature field mapping for the inner rod, which is ... )] ( ) ( ) ( ) ( ) ( ) ( 2 ) ( [ 2 1 )] ( ) ( ) ( [ ) (0 2 0 0 2 0 2 0 3 0 0 3 0 2 1 0 0 2 0 1 2 0 1 0 0 0 2 2 0 2 0 0 2 0 1 0 0 0 1 0 0 0 0 0 R dr T d R R dr T d R R dr T d R R dr dT R dr T d R R dr dT R dr dT dr dT (4.15)

PAGE 80

80 The flux is equal to zero at the outer wall and is zero for all subsequent orders. The solution to this first order problem has a homogenous and particular solution multiplied by a cos term, just like the velocity field. It is cos )] ( ) ( ) ( ) ( [1 1 1 1 1 1 1 1 1 ^r J r Y r J B r Y A Tj y T T (4.16) As in the case of the velocity field, the boundary condition for the temperature field at the inner wall is found by taking terms of the appropriate order in (4.15), and solving for the respective order. For the second order problem, the differential equation for temperature is the same form as (4.14) with T2 and Vz 2 replacing T1 and Vz 1, respectively. As in the second order velocity problem, the solution is broken up into two parts to satisfy the dependent and independent boundary condition, and again, only the independent part of the solution will be useful. We get ) ( ) ( ) ( ) (0 2 0 2 0 2 0 2 2 ^r J r Y r J B r Y A Tj y T T (4.17) The cosine and sine functions can be broken up as in the second order velocity problem and only the independent terms are needed. Now that the velocity and temperature e quations have been determined up to second order in the heat transport can now be obtained. The expression fo r the transport is the same as before, which is 2 0 ^ ~0Re 2R R Z vrdrd T V C Q (4.18) where ) (^ ~ Z ZV conjugate V (4.19)

PAGE 81

81 Note that the time dependence has already been factored out leaving only the cross terms (cf. Chapter 3). In the case of a perturbed surface, the transport ra te Q will have to be expressed as a perturbation series as well, so ...... 2 12 2 1 0 Q Q Q Q (4.20) where d dQ Q 1 (4.21) and 2 2 2d Q d Q (4.22) Since Q is expressed as an integral, the derivative of Q can be determined using Leibnitzs rule for differentiating integrals. Note that the veloc ity and temperature terms can be expressed as derivatives, which will simplify the algebra somewhat. Therefore .. 2 12 1 0 Z Z Z ZV V V V (4.23) where d dV VZ Z1 (4.24) and 2 2 2d V d VZ Z (4.25) The same expansion can be done for the temp erature. Using Lei bnitzs rule to find Q1, it is found that ( cos ] [1 ~ 0 ^ 1 ^ 0 ~ 1 rdrd V T T V C QZ Z vRd T VZ)^ 0 0 ~ (4.26)

PAGE 82

82 Looking at (4.26), it can be concluded that Q1 is equal to zero. Both terms in the double integral contain a cos term from the first order e xpression and integration over 2 makes the entire integral equal to zero. The same thi ng occurs in the single integral with the cos term, making that integral equal to zero as well. The expression for Q2 can be found, and after some cancellations, the result is d R dr V d R T R R rdrd T V T V T V C QZ Z Z Z v) ( ) ( ] 2 [0 0 0 ~ 0 0 ^ 2 1 0 0 ^ 2 ~ 1 ^ 1 ~ 2 ^ 0 ~ 2 (4.27) From this expression it can be understood why the dependent terms were neglected in the second order velocity and temperature expres sions. If the expressions for the perturbed velocity and temperature field were in serted into (4.27) including all the dependent parts, several terms would result that either contained a cos or a cos 2 term. Since the integration is carried out over 2 radians in the direction, all of these terms w ill integrate out to zero. This is why the dependent terms were discar ded in the second order pert urbation, since they would just contribute to the terms that would eventual ly integrate out to zero. A final expression can now be obtained for Q2 after further evaluating the term in the single integral. It is )} ( 2 ) ( ] [ Re{ 40 0 0 ~ 0 0 ^ 0 ^ 2 ~ 1 ^ 1 ~ 2 ^ 0 ~ 20 0R dr V d R T rdr T V T V T V C QZ R R Z Z Z v (4.28) Since Q1 is equal to zero, (4.28) must be added to Q0 to get the final equation for the total heat transport in the off-centered annular geometry to give

PAGE 83

83 0 0] [ Re{ 40 ^ 0 ~ R R Z vrdr T V C Q )] ( 2 ) ( ] [ [ 2 10 0 0 ~ 0 0 ^ 0 ^ 2 ~ 1 ^ 1 ~ 2 ^ 0 ~ 20 0R dr V d R T rdr T V T V T VZ R R Z Z Z } (4.29) 4.2 Analysis of Off Centered Model One of the important results that need to be obtained from these equations is a solution for Q2 at various frequencies. This will determin e if the off centered annulus is a beneficial geometry, or a detrimental one, since Q2 is equal to the change in transport due to the perturbation. The results for a hypothetical fl uid with a Prandtl number of 10, a kinematic viscosity of 0.1 cm2/s, and amplitude of 10 cm are presente d in Figure 4.2. Like the comparison between the cylinder and the annulus presented in Chapter 3, the qualitative nature is what is important here. MAPLE was used to solve (4.29) at various fr equencies. It is seen here that the change in transport is positive for small frequencies, and quickly becomes negative for higher frequencies. Although this has been shown from a mathematical standpoint, some explanation needs to be given why this makes physical sense. When the annulus is positioned in the cente r, the only gradients that exist during the oscillations are in the radial di rection. If the annulus were moved to an off centered position, there will be gradients forming in the direction in addition to the ra dial direction. In order to visualize this better, imagine that the off-cente red position of the annulus creates a thin region and a thick region (see Figure 4.3). The thin region consists of radi al lengths less than R0, and the thick region contains the area of flui d that has radial le ngths greater than R0. The total cross sectional area of the thick region tu rns out to be greater than the thin region. As stated before, an azimuthal gradient is formed. The heat will then flow from the th in region to the thick region.

PAGE 84

84 Since the thick region is the freer flowing region, it will be easier for the heat to transport down the tube, leading to a positive change. This ar gument was proven in Chapter 3 when the heat transport decreases when the region of flow become s more constricted. This was the case of an inner rod being stuck inside an open tube to form an annulus. The positive gain in transport from the thick region is higher than the loss from the thin region, causing the ov erall net change to be positive for low frequencies. When the frequency increases, somethi ng different happens. Since the walls are oscillating, the velocity near the walls will be higher than in the center of the tube. This is similar to the oscillating flat plate problem described in Chapter 2, when the velocity is at its highest at the wall and decreases as one moves away from it. This is especially true when the frequency is high, because there will be a sharper velocity gradient going from the wall to the core. Since the thick region has more available area farther away from the walls, the heat is essentially moving from the fast to the slow part of the flow (from the thin region to the thick region). This means that heat will take a longer period time to move down the tube. This phenomenon will become more and more dominant as the frequency is increased. That is why the net transport becomes negati ve, since the effect of fast moving walls dominates over the effect of the free flowing region, and continues to be more and more negative as frequency is increased. Now that the models have been presented with various geometric comparisons, experiments can be designed to test the accuracy of these models.

PAGE 85

85 Figure 4.1: Difference between centered annulus (clear inner rod) and off-centered annulus (shaded inner rod) Figure 4.2: Net change in transport due to off centered inner rod at various frequencies

PAGE 86

86 Figure 4.3: Thin and thick regions that re sult from an annulus that is off-centered

PAGE 87

87 CHAPTER 5 THE EXPERIMENT AND RESULTS FROM THE EXPERIMENTATION The theory presented thus far shows how the enhancement of heat transfer depends on frequency of oscillation, the amplitude of the priming motion, and the geometry. An experimental apparatus was designed to test the validity of the theoretical model. This experiment was meant to verify the case of pist on driven flow in a cylindrical tube. The apparatus is described below as well as the method that was used to obtain the data. The reader will learn about the difficulties that arose when designing the testing procedure, as well as the logic behind the alterations that were made to improve the performance of the experiment. The possible experimental errors are also discusse d, and finally the resu lts are given with the discussion as to why the data agrees or disagrees with certa in aspects of the theory. 5.1 Experimental Set-Up The experimental design that was used in this study is pictured in Fi gure 5.1. A sketch is shown in Figure 5.2a and 5.2b. The top cylindric al tank was used as the heat source, while the bottom tank was the heat sink, with a connect ing tube between them. The top tank was approximately 6 L in volume, and the bottom tank had a volume of 4.5 L. The length of the pipe in the figure was 30 cm, although that length c ould be varied. The fluid medium filled the bottom tank, the connecting pipe, and most of the upper tank. The orientation of the heat source above the heat sink was optimal since a lighter (warmer) fluid was on top of a heavier (colder) fluid. This minimized the effect of natural c onvection arising from grav itational effects. The oscillations were imposed by a m echanical piston that was attached to a bull wheel which was rotated by a power drill. The am plitude of the oscillations was changed by adjusting the position of the hinge on the bull wheel, while the freque ncy was changed by varying the speed of the

PAGE 88

88 power drill. It was important to use a materi al for the piston that could handle a reasonable amount of torque (from the drill) In the early expl oratory stage of experimentation, several pistons broke because of this issue. After a wh ile, the acrylic piston was exchanged for a piston made of PVC (polyvinyl chloride). This piston had a higher re sistance to torque, and when properly lubricated with oil, it was able to last the entire experimentation process. The material selected for the tanks and the c onnecting pipe was plexiglass. Th is was chosen since its thermal conductivity is low (0.18 W/m.K) compared to th e fluids used during experimentation. These fluids were water (conductivity of 0.60 W/m. K) and a 60% glycerol -water mixture (0.38 W/m.K). It was important to have a tank made of material that had a low conductivity since the model assumed that no heat was lost through the walls of the tank. A material with a higher thermal conductivity could transport heat out of th e pipe in the radial direction, which would provide a source of error. An important part of running this experiment effectively wa s to control the temperature of the heat source and the heat sink. This kept the axial temperature difference constant throughout the experiment and consistent with the model. This was accomplished with the assistance of controlled temperature baths that circulated constant te mperature water through both the heat source a nd the heat sink. In the heat source, in add ition to having the controlled temperature bath circulate the water in the tank through a copper coil, a heater was placed in the tank that was controlled by a PID controller. A mixing rod attached to a moto r was also present in the heat source which assisted in keeping the temperature uniform in th e bath. Copper was chosen as the coil material due to its high thermal conductivity. When th e set point for the heater in the tank was established, the heater turned on until it reached the set point sin ce the set point was greater than

PAGE 89

89 room temperature. However, as there was an overshoot as the system was trying to reach steady state, the temperature bath had a set point which was slightly below the set point of the heater. This assisted in removing the excess heat that the heater emitted as it tried to maintain constant temperature in the upper tank. This was n ecessary since heat emitted by the heater in overshooting the set point in theory really had nowhere to es cape. Since the diameter of the connecting tube was extremely small compared to the tank, excess heat would have taken an inordinate amount time to move itself down the t ube and out of the tank. This cooling loop expedited the process. In the lower tank there was also a cooling loop with a temperature bath; however, this loop took up a very large amount of the volume of the tank. Larger amounts of coils were needed in the sink since in addition to keeping th e sink at constant temper ature; there was heat that was transported down the tube that needed to be removed on ce the oscillations began. Now that the apparatus has been described, we discu ss the method that was developed to gather data. 5.2 Experimental Method When both tanks reached steady state, the heat er in the heat source was on at a certain rate to maintain the set point temperature. The co ntroller measured this rate and it was recorded. This was important to do since this rate took into account the h eat loss through the coils as well as losses to the ambient. Although the assumption of no heat loss through the walls of the tanks and walls of the pipe was made, no amount of insu lation could have made this assumption totally accurate. Therefore heat was still lost th rough these walls, although the heat source and connecting pipe were covered with multiple layers of foam insulation. This rate of heat loss needed to be established before the oscillations be gan, since it had to be factored out of the total rate. Another way that the heat leakages from the top tank were minimized was by the set point

PAGE 90

90 of the heat source. The set point used for these experiments was always between 24 and 26 degrees Celsius, which were slightly above r oom temperature. This made the temperature gradient between the hot tank and the ambient rath er small, which aided in reducing heat loss out of the heat source. The set point of the bo ttom tank was slightly below room temperature, between 16 and 18 degrees Celsius. Again, this minimized the temperature gradient between the heat sink and the surroundings. Although the axia l temperature difference between the source and the sink was only between 6 and 10 deg C for the experimental testing, the theory still showed that a significant amount of heat transfer would occur. When the oscillations began for a given amp litude and frequency, heat was transported quickly down the pipe. This drove the temperat ure of the top tank down, since heat was being lost to the sink. In order to compensate for th is, additional heat was adde d to the source to drive the temperature back up to the set point. The heat that was put back into the top tank to maintain the set point was equal to the heat that was trans ported down the tube to the heat sink. Again, to get the actual rate of heat that was transported do wn the tube, one must subt ract the rate of heat loss to the ambient and coils, which were determin ed before the oscillations started. In the bottom tank, the temperature tried to increase as heat entered it from the pipe. This heat was removed via the copper coils so that a constant te mperature could still be maintained in the heat sink. In both tanks, the coils remo ved the heat, and this heated wa ter went into a chiller, where it was cooled down. This chilled water then was se nt back to the temperature bath, where it was used to export constant temperature water back to the tanks. This cyclic process occurred throughout the experiment. In order to check the accuracy of the expe rimental results against the theory, several different parameters were tested. One theoretical conclusion that was desired to be tested was to

PAGE 91

91 see if the oscillations improved the heat transport by multiple orders of magnitude as predicted by the model. However, in order to do that, th e pure conductive case needed to be evaluated as well. Unfortunately, the time required for the linear temperature profile to form in the tube via conduction was very high when water was used as the transporting fluid. Since the model assumed that this linear temperat ure profile existed, this must form first before the conductive heat transfer is determined. The time constant for the conductive case scaled as the pipe length squared over the thermal diffusivity. Remember that there were no osci llations in the conductive case, so the heat transport occurre d strictly in the axial direction, making the characteristic length the length of the tube and not its radius. Even with a relatively small pipe length of 30 cm in this experiment, it would still take on the order of one week for the steady linear profile to form. It would be possible for this time to decrease if the pipe length was really small. However, the model also neglected the entrance and exit eff ects as discussed before, meaning that the oscillation amplitude used would have to decrease substantially as well. This small oscillation amplitude would have enhanced the transport in a very minimal way, and would not have been very useful. Another possible way to avoid this problem would be to use a fluid with a much higher thermal diffusivity, since this would have lowered the amount of time needed for the steady state profile to form down the tube. Air would have been a possible substi tute since its thermal diffusivity is over 100 times that of water. Howe ver, a problem arose when air was put into the tank. Mixing air in the tank proved to be very di fficult as it was observed that there were severe temperature fluctuations in the heat source. Since the model assumed a well mixed heat source, another mixing method had to be used instead of th e stirring rod. A few small fans were inserted into the tank to help keep the air constantly circ ulating. However, it turned out that heat was

PAGE 92

92 being generated from the fan motors, which in th e end would skew the results since not all the heat being added to the tank would have come fr om the heater, some would have come from the fan motors. Since only the output of the heater got measured, th is would have become a large source of error. In the end, it was determined that air wasnt a feasible test fluid, and water became the fluid of choice again. It was then decided to comp are different convective cases since the time needed for a convective steady state to form was on the or der of a few minutes under the influence of oscillations. Different parameters in the apparatu s were varied to see if indeed the experimental results followed the theoretical model. Before that occurred, the sources of error in the experiment needed to be evaluated to determ ine how much they woul d affect the results. 5.3 Error Analysis In the experimental set-up, there were several sources of error that arose. The first issue was related to the physical properties of the fluid. There are several that come into play when calculating the heat transported through the system. These incl ude the thermal conductivity, the kinematic viscosity, and the density of the fluid. These propertie s are functions of temperature, so their values varied when the temperature ch anged. The property values obtained for both water and the 60% glycerol-water mixture were those at room temperature (20 C). Since the set point of the upper tank was a few degrees higher than room temperature (24-26 C), and the set point of the lower tank was a few degrees belo w room temperature (16-18 C), it was assumed that the average temperature in the tank was roughly that of the r oom temperature. However, if the upper temperature limits of th e tank were used, the average temp erature in the pipe would be 22 C instead of 20 C. It was also possible for th e room temperature to have been greater than 20 C, which might have influenced the average temp erature in the pipe. However the pipe was well

PAGE 93

93 insulated and it was unlikely that the room temp erature excursion would have had much of an effect. Therefore the physical properties of wate r and glycerol needed to be examined between 20 and 22 C. A range of values was obta ined (see Tables 5.1 and 5.2). These data show that even with a two degree temperature difference, there is a difference in the Prandtl number of the fluid. The contribu ting factor to this is the kinematic viscosity, which appears in the Prandtl number. It can be seen that for water ther e can be as much as a 2.3% deviation from the average Prandtl numbe r and a corresponding deviation of 3.5% for the glycerol-water mixture. Other errors arose from deviations in th e dimensions of the experiment from design values. The pipe had a total design length of 30 cm. It was divided into two pieces, each having a tolerance of 0.0127 cm. The to lerance in the radius of the pipe was also 0.0127 cm. Two different radii were used in the experiment, whos e sizes were 3/16 in. and 1/8 in. The error then in the pipe length was 0.085%, while the error in radius was 2.68% for the wide tube, and 4.00% for the narrow tube. The amplitude measurement was taken by de termining the length of the piston hinge from the center of the bull wheel. This was e quivalent to the actual length the piston was moving. The maximum possible amplitude that the piston could have generated was 7.14 cm, and minimum amplitude of 2.73 cm. Note that this wasnt the actual amplitude of the flow in the pipe since the tube in which the piston oscillated had a different radius then that in the actual pipe. The amplitude in the pipe can be easily determined by using the following formula which is that of equal volume displacement. It is 2 2) ( ) (pipe pipe piston pistonR A R A (5.1)

PAGE 94

94 whence the amplitude in the pipe is calculated. The pipe amplitude not only varied when the piston amplitude was changed, but also with the ra dius of the pipe, as di fferent sizes were used during experimentation. The piston amplitude was measured using calipers. Although the calipers were quite accurate, an eyeball measurement needed to be taken to place the caliper ends with the center of the hinge a nd the center of the bull wheel. Th e error in the measurement was assumed to be about 0.5 mm. This means that for the smallest piston amplitude, the maximum error that could exist is 1.83%. A larger pi ston amplitude would produce a smaller percentage error. The temperature drop between the heat s ource and heat sink was dependent on the accuracy of the controllers in the respective tank s. The controller produced very good results as the temperature range was always within 0.05 C of the set point in both tanks at steady state. Temperature measurements were taken every 5 seconds, meaning that 240 measurements were taken over the 20 minute trial period to obtain an average temperat ure reading. It was unlikely that the error in the average temperature was an ything over 0.01 C for each tank. The error in an 8 C temperature gradient (the typical temperature difference) would have been about 1/4 %. Another source of error in the experiment o ccurred when measuring the frequency. Since the bull wheel was attached to a drill with a moto r, there would be surges or decreases in the voltage to the motor. This w ould cause a change in the frequency measurement for short periods of time. When this occurred, the drill was adju sted quickly to bring th e frequency back to the original set point. It was dete rmined that the maximum change that was brought about in the frequency by these voltage changes was no more than 0.2 Hz. This change usually lasted no more than a minute before it wa s noticed and rectified. The am ount of time for a trial period was 20 minutes, meaning that the difference in th e average frequency was 0.01 Hz per voltage

PAGE 95

95 change. The number of voltage ch anges usually never exceeded 3 in a trial period, which made the total deviation in the average frequency 0.03 Hz at most. The frequencies used during the entire experimentation process ranged between 0.9 and 1.7 Hz. The maximum error for small frequencies would have been roughly 3. 3% and 1.8 % for large frequencies. So far, several input parameters have been discussed in regard to the error in their measurements. These include kinematic viscos ity, thermal diffusivity, thermal conductivity, temperature gradient, pipe length, pipe width, oscillation amplitude and oscillation frequency. Note that all these parameters appear in (3. 40), which was the model equation for piston driven flow in a cylinder. Since the overall heat transpor t was proportional to all of these parameters in different ways, the maximum and mi nimum values that arose from errors were substituted into (3.40) for each term to get the maximum a nd minimum possible value for the overall heat transport. There was another source of error in the experiment that doesnt appear in (3.40). When the actual amount of heat transported in the experiment wa s determined, the time that the heater was on was counted by a timer in the comput er that controls the he ater. This amount of time had an error associated with it. The tim er was rounded off to the nearest hundredth of a second. When a start time and stop time were determined, there was a rounding error associated with each time in addition to the start and stop time for when the heat loss to the ambient was calculated. Since there was a ma ximum of 0.005 s of rounding error with each timing, a total of 0.02 s in error could result when 4 timings are done. Considering the heater was on anywhere from 3 to 100 s depending on the trial, the error as sociated with this could range from 0.02% to 0.67%. In the case where a ratio of two trials was taken, 8 timings were done, and the maximum total error doubled to 0.04 s, re sulting in a percentage error ranging from 0. 04% to 1.33%. This additional error was added into the maximum and minimum values for the overall heat transfer

PAGE 96

96 to obtain the total experimental error. This tota l error is expressed on a trial to trial basis with error bars shown on the plot s in the results section. 5.4 Note on Viscous Dissipation A term that was neglected in the energy bala nce earlier was that of viscous dissipation due to movement of the fluid in the tube. If th is term were to appear in the energy balance, (3.13) would be rewritten as 2 2 2) ( ) 1 ( r V C r T r r T z T V t Tz v z (5.2) We can scale (5.2) to obtain 2 * * 2 * 2 * *) ( ) 1 ( ) Pr( r V Br r T r r T V L A dt dT Woz z (5.3) Where the Brinkman number is defined as T k A Br 2 2 (5.4) Typical experimental values were used to determine the Brinkman number. When water was considered as the fluid medium, this dimensi onless group turned out to be five orders of magnitude lower than the first order conductive term and four orders of magnitude lower for the glycerol-water mixture. The dimensionless groups of the time dependent and convective terms were even greater than the conductive term. Although from a scaling perspective, we have reasoning to drop the viscous dissipation term, we need to evaluate the entire term to see how much heat is really lost due to the viscos ity. Therefore, a calculation was performed by integrating the viscous dissipation term over the cross sectional area and comparing it to the output heat transport. It turned out the heat transport was over 500 times the heat lost to viscous dissipation for the case of water as a fluid, a nd over 50 times the heat lost in the case of a

PAGE 97

97 glycerol-water mixture. From these results, it can be concluded that the viscous dissipation term is insignificant, and this why it wa s neglected in the energy balance. 5.5 Experimental Results and Discussion 5.5.1 Varying Amplitude with Water The first series of experiments were meant to verify the relationship that the convective heat transferred is proportional to the square of the piston amplitude. Initially, it was determined that the time needed to reach the conductive steady state for water was roughly 120 hours for the 30 cm pipe length. Through a theoretical calculat ion it was figured that the oscillations increase the overall transport by a factor of about 600. A rough estimate was made to determine the amount of time needed for the convective steady state to obtain. It was roughly 12 minutes. Once the oscillations were in effect for that length of time, it was assumed that the velocity profile in the tube was fully developed and heat was being transported at a constant rate. Initially, the amount of heat input to th e top tank over a 20 minute period was measured without oscillations in order to determine the amount of heat lost to the ambient. Once that measurement was completed the oscillations were started. The steady state profile was developed and another 20 minute measuring period was conduc ted for an input piston amplitude. In the next experiment, the amplitude was changed and there was a wait time until the steady state formed at the new amplitude. The heat tran sported was again measured over a span of 20 minutes. It should be noted here that the freque ncy was set between 1-1.5 Hz for every trial. Although there was a variation in frequency over a ll the trials, each indivi dual trial had the same frequency for each of the two amplitudes tested. Figure 5.3 shows the experimental results co mpared to a theoretical curve at various amplitude ratios. The amplitudes chosen for these trials could not exceed 40% of the pipe length

PAGE 98

98 since a larger amplitude would have induced en trance and exit effectseffects that were neglected in the model. This limited the amount of possible amplitude ratios available, which is why the ratios shown in the figure are between 1.1 and 1.6. Observe from the figure that the data points are quite close to the theoretical curve for amplitude ratios between 1.2 and 1.5. For th e extremely low and extremely high ratios, the experimental results were slightly higher than the model prediction. The reasoning for the difference at the high ratios could possibly be du e to the entrance and ex it effects. Although a 40% limit was placed on the amplitude size, this is just a rough guess as to when the entrance and exit effects become relevant. When an oscillation pulse is generated, the temperature gradients ought to exist only in the pipe and the model assumes that these gradient s are not felt in either tank. However, when a spike becomes large, temperature surges can occu r up through the entire length of the pipe and into the tank. Since the heater and controller were both in th e top tank, a sudden temperature drop in the tank would have obviously caused more heat to be added to the tank to keep the temperature at the set point. Now the model only takes into account heat tr ansported between the ends of the pipe. Because a larger portion of the tank was now cooler, not just at the spot where the pipe was connected, a greater amount of heat would have to be added to cool the tank down. These are considered the entrance and exit effects because the transport occurs outside the ends of the pipe in addition to inside the pipe. This excess heat increases the total amount of time the heater is on during an experimental run. This overshoot causes the heat transport ratio to become larger than expected (since the transport ratio is given with the higher amplitude resu lt in the numerator) and this is consistent with what the data shows for high amplitude ratios. If higher amplitude ratios

PAGE 99

99 were used than the ones shown in Figure 5.3 one would likely see the error become even greater, as the spikes generated would become ever larger affecting the tanks more and more. The discrepancy at lower amp litude ratios would not be for the same reason as the high end ratios, since the spikes were not as large. It is likely that the error in the data points between the amplitude ratio of 1.1 and 1.2 are due to some additional heat leakages in the experiment. These heat leakages will be disc ussed in a later secti on in this chapter. Looking at the points in the middle of the data set (ratios between 1.2 and 1. 3), it can be seen that these points are in very good agreement with the theory. 5.5.2 Varying Amplitude with Glycerol-Water Mixture The same experimental method discussed in the previous section wa s applied to a fluid with a higher Prandtl number, a 60% glycerol-4 0% water mixture which has a Prandtl number around 85. The idea behind running this part of th e experiment is to see if the model is a good predictor for a more viscous fluid and also to see if a higher Prandtl number fluid would behave worse than a low Prandtl number fluid for varyin g amplitudes. This could possibly be so as a higher Prandtl number fluid usually decreases the convective transpor t in the pipe. The decrease occurs because a higher Prandtl number fluid usually is more viscous (the glycerol water-mixture is roughly 10 times the kinematic viscosity of wa ter). The term usually is used here since thermal diffusivity can affect the Prandtl number as well. Two fluids can have different Prandtl numbers even when they have the same kinematic viscosity due to differing thermal diffusivity values. It has been shown from the model in Ch apter 2 that the heat transport decreases with decreasing Womersley number. Since the kinematic viscosity appears in the denominator of the definition of the Womersley number (see Chapter 2), a more viscous fluid will decrease that value, assuming the oscillation frequency and tube radius are kept constant Although ratios in

PAGE 100

100 heat transported are being considered here, this could affect the accuracy of the experiment. For example, suppose at a certain amplitude, the heat per unit time transported down a water filled pipe is 2 W. If the amplitude were doubled, th e transport is predicted to become 8 W according to the model. Now suppose at the same original amplitude the heat tran sported in a glycerolwater pipe is 1 W. If twice the amplitude wa s used, the theoretical transport would become 4 times that, or 4 W. For the water example, the rate of heat transport increases by 6 W when the amplitude is doubled. In the glycerol water ca se, the increase is only 3 W. Suppose in both cases the experiment produced an increase in tran sport that was 1 W higher than expected. The percent error would be 16.67% fo r the water case (6 W +/1 W) versus 33.33% for the glycerolwater mixture (3 W +/1 W). From this example, it could be concluded th at the glycerol-water mixture is more sensitive from an percent accura cy standpoint. Using this fluid could prove to be a better test to see how accurate the experiment really is. Figure 5.4 shows the results for the glycerol-water mixture. The results here show better agreement with the theoretical prediction compared to the trials with pure water. All the data points are within the experime ntal error bars. It should be noted though that the very high am plitude ratios were not taken in to account in this experiment due to the larger errors obtained in the last experiment at those respective values. If large amplitude ratios were tested with the glycerol-water mixture, th e errors associated with them would be even higher as a more viscous fluid generates sharpe r velocity spikes, causing even more temperature gradients to form in the tanks when the amplitudes are large. That is why only ratios below 1.4 were tested. This data set is a positive sign, as good agreement with the theory was obtained for a fluid medium that was less conducive to transporting heat, meaning that it

PAGE 101

101 was more difficult to measure, due to smaller changes in transport when the amplitude was varied. In these first two data sets, only the accu racy of ratios was examined. Although the results were quite good, the actual heat transported in the experime nt was not determined. If the amplitude is increased by 10% from a first case to a second case, the experiment has shown that it can produce about a 20% increase in heat transport. However, does that 20% mean an increase of 1 W to 1.2 W, or 100 to 120 W? There is clearl y a big difference between 0.2 and 20 W. It is important to determine the actual amount of heat transported in the tube to see how completely accurate the experiment turns out to be. This will be examined in the next section as the frequency is varied. 5.5.3 Varying Frequency The frequency of the oscillations was altered to see how it affects the overall heat transport. The same glycerol-water mixture wa s used as the fluid medium during this set of experiments. The set-up of the trials was simila r to the previous experime ntal runs in that a 20 minute period run was conducted with no oscillations to determine the heat lo st to the ambient. Then the heat transfer over a 20 minute period with oscillations was measured for various frequencies. Again, enough time was allotted for the oscillations to re ach steady state before each trial commenced. Frequency values ranged from 0.9 to 1.7 Hz. This was an appropriate frequency range for measurement since the oscilla tions were produced by a drill turning a bull wheel. Frequencies lower than these were difficu lt to keep constant since the minimum speed of the drill produced frequencies around 0.9 Hz. Highe r frequency values were not used in order to keep the amount of torque on the bull wheel from the drill low. A safety concern was that the piston could snap if it were oscillating too fast (this occurred during preliminary trials). Also,

PAGE 102

102 there was an upper limit for freque ncy since turbulent flow needed to be avoided in the pipe, since the model assumed laminar flow. The resu lts for the experimental trials for varying frequency are shown in figure 5.5. It is clear from this data that there is a noticeable deviation from the model prediction. For the larger frequencies, there is roughly an entire order of magnitude difference between the experimental results and the theoretical model. This discrepancy didnt occur in the first two experiments as only the ratio of heat transport was compared to the theoretical predictions. It appears as though the deviation got larger as the frequency was increased. One reason for a higher experimental heat tran sport was due to heat losses in the system while the oscillations were occurring. Heat lo sses to the ambient were measured before the experimental trials are conducted. However this really just take s into account the heat losses from the heat source to the ambient, since the heater and controller we re all in the top of the tank. When there were no oscillations, th e controller only rec ognized heat that wa s being lost through the tank walls, since the surface ar ea of the tank was drastically hi gher than the surface area of the region where the tank and pipe were conne cted. Also, the location of the heater and controller was much closer to the walls of the tank compared to the pipe (see Figure 5.2a). All of this means that when the rate of heat loss to the ambient was determined, only the heat loss out of the tank was included. There was still the potentia l for heat leakages through the pipe walls while the oscillations were occurring. As stat ed before, although the pipe was insulated, no amount of insulation could totally prevent heat from being lost through the pipe walls. In addition to that, the fluid medium and pipe material which were discussed above assisted in limiting the amount of heat leakages through the walls, but again, it was not a complete fix to the problem. Also, as the radius of the tube could be varied, different inserts were machined.

PAGE 103

103 These inserts fit into a larger size tube which is de picted in Figure 5.6. When the inserts were fit into the larger tube, there were air gaps that fo rmed between the wall of the insert and the wall of the outer tube. The heat could possibly move th rough the inner wall and in to the air gaps, which is not part of the fluid channel. This could cont ribute to part of the heat loss; however, since the movement of the fluid is periodic some of the heat may return back to the fluid in the reverse stroke. The entrance and exit effects could be anothe r reason for the experimental heat transfer being high compared to the theory. The amplitude used for all of these frequency measurements was rather high. This was needed since the theo retical heat transfer w ould have been quite low for a small amplitude case, and it would have be en difficult to obtain a measurement from the experiment to compare with the theory for such a small amount. Since the amplitude in the experimentation was high, the larger spike in the channel might have caused the entrance and exit effects to contribute in overshooting th e theoretical heat transport as well. It should be noted that Thomas [14] faced th e same issue in his mass transfer experiment, where the experimental mass transfer was much hi gher than the model prediction. However, in his case, he didnt have leakages of mass through th e walls, unlike this heat transfer experiment. He suggested that secondary flows may have form ed in his channel, causing a deviation in the parabolic profile. This could also be a problem in the heat transfer experiment, especially when the oscillation frequency grows. All of these issues could help explain why the experimental rate of heat transfer was much higher than expected. Although there is a disagreemen t in the experimental results with the theory in terms of total heat transported down the pipe, it is still important to see how much increasing the frequency of the oscillations improve the heat trans port, and if it is consistent with the theory. In

PAGE 104

104 order to do that, a normalized heat transport was developed. This was done by dividing the theoretical transport by some value at every frequency in the range to get a normalized theoretical curve. The value that was chosen for the normalization was the experimental heat transfer value at a mid-range frequency. It wa s selected because it is allows for a comparison with the model in the cases where the frequenc y is increased or decreased. Figure 5.7 shows these results. It can be seen that the change in the heat transport is much more drastic than the model predicts. As frequency increases, there is a much larger increase in the transport. Again, this is likely due to the reasons discussed earlier. It should also be no ted that only 4 frequencies were used. In an ideal situation, ma ny more frequency values would be chosen, however the nature of the oscillating mechanism prevented this from ha ppening. Considering a drill was being used to drive the oscillations, it was diffi cult to keep a constant speed going for every possible frequency within the available range. The drill setting wa s kept constant by placing a clamp on the trigger mechanism and loosening or tightening it to vary the speed. The frequencies selected here for the experiment were those frequenc ies that the drill could produce at a fairly constant rate. It was very difficult to tighten the clamp by even th e slightest amount so as to only increase the frequency by say 0.05 Hz, as usually any minor ch ange to the clamp brought about no change in the drill speed. After several minor tweaks, ther e would eventually be a larger surge in the speed. This increase roughly correlated to a freq uency increase of 0.2 Hz. That is why there is roughly a 0.2 Hz difference between the chosen frequencies. Ideally, a mechanism should be designed to generate perfectly constant frequencies at any setting over a large range. Several other trials were conducted to see if these results could be improved. However, the regular and normalized plots were very simila r to Figures 5.5 and 5.7 for subsequent runs.

PAGE 105

105 The fluid medium was also changed from the gl ycerol-water mixture back to pure water. Unfortunately, there was no difference in the outcome. This experimental test did not produce the desi red results except for the fact that it was demonstrated experimentally that heat transpor t does increase when the oscillation frequency is increased. It was shown from this trial that measuring the actual heat transported in the system (not a ratio) can be qui te difficult due to heat losses that can be very tough to eliminate. 5.5.4 Varying Tube Radius The final experimental test involved cha nging the pipe radius. In these trials, two different radial sizes were used 3/16 in. and 1/8 in. Again, th e heat lost to the ambient was calculated before the 20 minute trials were conduct ed, one at each radius length. A ratio between the two trials was developed and compared to the model. In this set of trials, amplitude was adjusted so that the theoretical heat transpor t was roughly the same be tween the two cases. Looking at (5.1), the amplitude of the pulsation in the pipe must increase when the tube radius is decreased for the equation to still be satisfied. Adjusting the amplitude is necessary in order for the total flow rates to be roughly th e same. The point of this set of experimental trials was not to see how radius affects the transp ort, but to see if the system can transport the same amount of heat when the radius is changed by altering the other parameters, namely the amplitude (and in some cases the frequency, although it was used minimally since past experimental results showed the accuracy was not very good compared to the amplitude variation). This test could be considered to be a culmination of the previous tria ls to see if multiple parameters can be adjusted to produce a desired result. Figure 5.8 shows the experimental ratio versus the theoretical ratio. The solid line has a slope of 1, which represents the case when th e experiment agrees with the theory. Since it was desired to have the total am ount of heat transfer be equal between the two

PAGE 106

106 cases according to the theoretical model, most of the ratios that were chosen were very close to 1. The results were reasonably satisfactory as some points were in good agreement with the theory and some were slightly outside the error bars. Keeping in mind all of the qualitative errors that were discussed involving the amplitude and frequency, these results are consistent with what was obtained in previous trials. Ev en though the frequency was not changed, the air gaps were still present in the tube (see Fig. 5.6). This combined with the entrance and exit effects generated from large amplitudes is a likely reason for the deviation in some of the data points from the model. Another possible contributing fact or to the error was the dist urbance that occurred when changing the pipe inserts. When amplitude need ed to be changed, the hinge on the bull wheel was adjusted. When the frequency needed to be changed, the clamp on the drill was adjusted. In the case where the radius size was switched, seve ral parts of the experiment needed to be disassembled. This included draining the top tank and the connecting pipe, shutting off the mixer, the cooling loop, the temperature bath, an d the controller. Once the insert was finally switched and the experiment was reassembled, seve ral minutes had elapsed. Although it is hard to pinpoint a single error arisi ng from this (except for maybe a change in room temperature during the transition period), a lot of experimental parts were switched ou t or adjusted. There was also a period of several minutes where the system had to be shut off. All of this could have contributed to the likelihood that the system didnt reach the exact same equilibrated state once it was started up again. This could have led to a different response from th e controller perhaps or the cooling loop. This is something that is ha rd to measure quantitati vely, but these potential

PAGE 107

107 small disturbances combined with the other so urces of error mentione d above could explain some of the data points lyi ng outside of the error bars. 5.6 Summary of Results During the experimentation process, several of the input parameters were varied to see how they affected the heat tran sport from the heat source to th e heat sink. These parameters included the type of fluid (with drastically di fferent Prandtl numbers), the amplitude of the piston, the frequency of the oscillations, and the radi us of the tube. Several plots were developed that show the performance of the experiment compar ed to the theoretical model. Initially, one of the important characteristics of the data was that all the points obeyed th e theoretical trend. When amplitude was increased, the heat transport increased. When frequency was increased, the heat transport increased, etc. How accurately the data followed the model varied from trial to trial. In some instances, the resu lts were quite close to the model, in others the data points laid just outside the error bars, and in a couple trials (mostly in the frequency variation) there was quite a large discrepancy between the theory a nd the data points. Whenever there was some deviation from the model, whether small or large, possible explanations were given as to why the errors occurred. These results are a good beginn ing in showing how much oscillatory flows can improve heat transport. 5.7 Potential Improvements on this Work There are many ways to change the experiment to improve the accuracy of the results. Others who follow up on this work might consider making the following changes: Design a better way to adjust frequency so that there are more values for which the frequency can remain stable. One way coul d be to use a constant power supply to prevent voltage surges as well as using a moto r with a shaft connecting to the bull wheel.

PAGE 108

108 The speed of the motor could be adjusted w ith a voltage regulator as opposed to the current method with a clamp on a drill. This c ould be taken one step further and have the frequency controlled by a computer program, much like the controller in the heat source which monitors the temperature. Develop ways to minimize the heat leakages in the pipe. This could include designing inserts that eliminate the air gaps or finding a material to make the tube and inserts out of which have a significantly lower thermal conductivity than the fluid medium (the fluid medium could also be changed to promote a bigger discrepancy). Expand the range for certain variables. A llow for a greater range of amplitudes and frequencies to be tested. Improve the mixing me chanism so that lighter fluids (i.e., gases) can be temperature controlled. The following improvements can be made whic h wont necessarily improve the accuracy of the current results, but will help show why osc illatory flows can increase the heat transport in a system: Develop a way to measure the convective to c onductive ratio, through either adjusting the parameters or redesigning the experiment to reduce the amount of time needed for the conductive steady state to exist in the pipe. Redesign the experiment where the transport re gion is a capillary bundl e. Using tubes of very small radii will drastica lly increase the convective to conductive ratio. This is how Kurzweg and Zhao [9] were able to generate such a la rge ratio (~18000x).

PAGE 109

109 Test the other two methods of imposing osci llations. Allow for a pressure valve to generate a pressure drop across the pipe, or design the experiment which would have a moving boundary to impose the oscillations. Allow the experiment to consist of various geom etries in the transport tube. Let the tube be either a centered or off centered annular region, or a two compartment region with a cylinder and annulus. It would be of interest to see if experiment al results could be generated which would suppor t the theoretical argument s made earlier comparing different geometries. Figure 5.1: Experimental set-up

PAGE 110

110 Figure 5.2a: Schematic diagram of heat source Figure 5.2b: Schematic diagram of heat sink

PAGE 111

111 Figure 5.3: Heat transport increase as a resu lt of piston amplitude increase for low Prandtl number fluid (water). Theoretical curve (solid li ne) is shown with experimental results (circles) with error bars Figure 5.4: Heat transport increase as a resu lt of piston amplitude increase for high Prandtl number fluid (glycerol-water mixture). Theoretical curve (solid line) is shown with experimental results (circles) with error bars.

PAGE 112

112 Figure 5.5: Heat transport increase as a result of oscillation frequency increase for high Prandtl number fluid (glycerol-water mixture). Theoretical curve (solid line) is shown with experimental results (circles) with error bars. Figure 5.6: Large tube with separate insert (left), insert placed in large tube (right) creating an air filled gap (shaded regions)

PAGE 113

113 Figure 5.7: Heat transport increase as a result of oscillation frequency increase for high Prandtl number fluid (glycerol-water mixture). Normalized theoretical curve (solid line) is shown with experimental results (circles) with error bars. Figure 5.8: Experimental ratio compared to th eoretical ratio for changing radius. Solid line represents agreement between experiment and th eory. Experimental data points (circles) are shown with error bars.

PAGE 114

114 Table 5.1: Physical Properties of Water at Different Temperatures Physical Property Water @ 20 C Water @ 22 C Density (g/cm3) 1.0 1.0 Thermal Conductivity (W/cm/C) 0.0059 0.0059 Kinematic Viscosity (cm2/s) 0.0100 0.0096 Specific Heat (J/g/C) 4.2 4.2 Prandtl Number (Pr) 7.1 6.8 Table 5.2: Physical Properties of a Glycer ol-Water mixture at Different Temperatures Physical Property Gly.-Water @ 20 C Gly.-Water @ 22 C Density (g/cm3) 1.2 1.2 Thermal Conductivity (W/cm/C) 0.0038 0.0038 Kinematic Viscosity (cm2/s) 0.094 0.087 Specific Heat (J/g/C) 3.1 3.1 Prandtl Number (Pr) 89 83

PAGE 115

115 CHAPTER 6 CONCLUSIONS AND FUTURE RESEARCH 6.1 Conclusions Several examples were presen ted to show how oscillations can improve heat transport in a pipe. In these examples, various parameters were examined to determine what conditions produce the best transport. These parameters included the geometry of the channel, the dimensions of the channel, the type of fl uid medium, and the oscillation frequency and amplitude. In all these examples, a physical expl anation was provided to bolster the argument of why the system behaved as it did. The oscillatory flow model was presented in three different wa ys, using a pulsating pressure drop, an external oscillating piston, and an os cillating moving boundary, for both a cylindrical and an annular geometry. For all thes e scenarios, it was found that the heat transfer was proportional to the square of the driving force. It was also shown that the heat transfer decreases with increasing Prand tl number. In each case, the input signal of the oscillation mechanism was a single periodic function, and th e response of the system was a combination of periodic functions with a phase lag. This was due to the size of the si gnal and the size of the system. A comparison was made between the cylin der and the annulus to determine which geometry is more optimal for heat transfer. Depending on what parameters were held fixed, different conclusions could be re ached. In the case where the pressure drop was held fixed, the cylinder provided a higher rate of transport. When the flow rate through the tubes was held fixed, the annulus provided the better results. Th ese trends held whether the outer radius of the cylinder and annulus were the same or whether the cross sectional areas were held the same. The

PAGE 116

116 inner rod of the annulus imposed a resistance to flow which drove the transport down when the pressure drop was held fixed. For the case of fixed flow rates, the re sistance in the annulus lowered its flow rate, meaning the pressure drop had to increase to keep the flow rate the same as in a cylindrical tube. This drove the transport in the annulus up for the case of fixed flow rates. It was also determined that the cylinder required less power to drive the fluid back and forth compared to the annulus, since substantial ener gy must be exerted by th e fluid to overcome the viscous effects of two walls in the annulus compared to just one for the cylinder. This geometric study was taken one step furt her where an annulus wa s compared to a two compartment model where the geometry consisted of an inner cylinder and an outer annulus. The conclusion was reached that the two compartm ent model always produ ced a higher rate of heat transport assuming the cross sectional area s of both models were held fixed. The comparison was also made to determine which model produced a better separation between two species. Although the effective separation was si milar between the two models in terms of magnitude, the maximum separation occurred at a much lower frequency for the two compartment model versus the one compartment m odel. The lower frequency correlates to a lower amount of energy needed to produce the op timal separation. The two compartment model produced the optimal transport in all cases since th e flow rates were held fixed in both models. As in the comparison of the annulus versus the cylinder, there is an extra boundary in the two compartment model, which acts as a resistance to flow, so to account for that, the amplitude of the spikes in the two compartment channels must increase to keep the flow rate identical to the one compartment model. Since the two compartmen t model has larger spikes, the transport will be higher in that model.

PAGE 117

117 The last effect of geometry to be stud ied was that of an of f centered annulus. A perturbation study was presented using the scenario where both the inner and outer walls were oscillating in phase. This showed how the heat transfer changed when the inner rod of the annulus was moved to an off centered position. The off centered model provided a positive change in heat transport for very low frequenc ies, but quickly decrease d as the frequency was raised. This was due to the thick region being larger than the thin region. When the walls oscillated fast enough, there was more fluid in th e thick region that was farther away from the fast moving walls, compared to the thin region. This produced the net de crease in the transport at higher frequencies. An experiment was designed in which multiple parameters were varied to compare its performance to the model. These included the oscillation amplitude, the fluid medium (water and a glyercol-water mixture were interchanged), the oscillation fr equency, and the tube radius. For most of the data points that were taken through all these e xperiments, there was a reasonable agreement with the theoretical model within the erro r bars. This was especially true for when the amplitude, fluid medium, and tube radius were varied. The experimental results obeyed the relationship of the heat transpor t being proportional to the square of the piston amplitude, which was predicted by the model, for different Pran dtl number fluids. Fo r the case of varying frequency, the agreement was not very good with the theory. This was because the total heat transported was measured as the frequency was cha nged. In the other experiments, a comparison of two cases was made. The experimental accu racy was found by determining a ratio for the heat transported in the two cases, and comparing th at to a theoretical ratio. Any error in the actual heat transfer would have been virtually eliminated since it would have been evident in both cases. For the trials where frequency was vari ed, the error in the total heat transport could

PAGE 118

118 be attributed to heat leakages at different sp ots in the tube which were difficult to measure quantitatively. Overall, the expe rimental results were encouragi ng as the oscillatory flow model provided a good comparison with the data points. 6.2 Future Research in Oscillatory Flows The oscillatory flow model can be put to us e in many areas of research today. The separation of species example has applications beyond those me ntioned in the introductory section. The first involves the detection of danger ous chemicals. This is an era when there is a deep concern regarding biological and chemical weapons. It is cons equently important to be able to detect harmful species. This project can of course be employed in the separation of species, which can be a valuable tool. This method coul d serve as a helpful step of concentrating the harmful species before an unknown sample containing the species enters a de tector. It is obvious that it is easier for a detector to find a hazardous chemical when there are fewer substances in a sample. By separating out certain species of a sample through this pulsating mechanism, samples would be made simpler for the detector. Another application woul d be the use of this project to separate viruses from samples. Sinc e new and dangerous viruse s threaten society all the time, it would be beneficial to create a mechan ism that could separate out harmful particles. In addition, the pulsating mechanism can be us ed for spin-off technologies such as DNA and protein separation. Other creative uses for this technology could in clude electromagnetic fields, in which an imposed field is oscillated to move charges in a particul ar direction. Also, gas molecules which are entrapped in porous medi a could be removed through oscillatory flows. This could occur by means of an oscillating pressu re drop on either side of the matrix, where the force exerted on the molecules could assist in moving them out of the porous matrix.

PAGE 119

119 From a heat transfer standpoint, there was an emphasis put on this method as a way to transport heat out of enclosed space capsules. Although the heat transfer experiment that was conducted follows the theoretical model well, it ha s yet to be seen whether this experiment would perform well under zero gr avity conditions. One might conc lude that this experiment would produce better results at ze ro gravity since natural convecti on would be totally eliminated. It is also important to see if this technology would work on a la rger scale, since the volume of outer space modules are several or ders of magnitude greater than the volume of the experiment conducted in the laboratory. Hopefully NASA w ill look into testing this method on their space habitats and use oscilla tory flows as a way to improve heat removal in their applications.

PAGE 120

120 APPENDIX CONSTANTS IN THE SOLUTIONS TO TH E OFF CENTERED ANNULUS PROBLEM ) ( ) ( ) ( ) ( )] ( ) ( [ 4 10 0 0 0 0 0 0 0 0 0 0 0 0cR J R Y R J cR Y cR J R J A Av (A.1) ) ( ) ( ) ( ) ( )] ( ) ( [ 4 10 0 0 0 0 0 0 0 0 0 0 0 0cR J R Y R J cR Y R Y cR Y A Bv (A.2) ) ( ) ( ) ( ) ( ) (0 1 0 1 0 1 0 1 0 0 0 1 10cR Y R J R Y cR J dr dV cR J AR v (A.3) ) ( ) ( ) ( ) ( ) (0 1 0 1 0 1 0 1 0 0 0 1 10cR Y R J R Y cR J dr dV cR Y BR v (A.4) ) ( ) ( ) ( ) ( ) 2 1 2 1 )( (0 0 0 0 0 0 0 0 0 0 ^ 0 2 0 0 ^ 2 0 1 ^ 0 0 20 0 0R Y cR J cR Y R J dr V d R dr V d dr V d cR J AR Z R Z R Z v (A.5) ) ( ) ( ) ( ) ( ) 2 1 2 1 )( (0 0 0 0 0 0 0 0 0 0 ^ 0 2 0 0 ^ 2 0 1 ^ 0 0 20 0 0R Y cR J cR Y R J dr V d R dr V d dr V d cR Y BR Z R Z R Z v (A.6) L T iAgv gy) Pr 1 (1 (A.7) L T iBgv gj) Pr 1 (1 (A.8) Where g is the corresponding order of the constant (0,1,2, etc.)

PAGE 121

121 ) ( ) ( ) ( ) (0 1 0 1 0 0 1 0 0 1 0 0R Y R J R Y R J B Aj y T T (A.9) ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( (0 1 0 1 0 1 0 1 0 1 0 1 0 0 1 0 1 0 0 1 0 1 0 1 0 1 0 1 0 1 0 0 1 0 1 0 0cR Y R J R Y cR J R Y cR J R Y cR Y cR Y R J R Y cR J cR Y R J cR Y R Y Bj y j y T (A.10) 0 0 1 0 0 1/ ) ( ) ( R R Y R Y R (A.11) 0 0 1 0 0 2/ ) ( ) ( R R J R J R (A.12) 0 0 1 0 0 3/ ) ( ) ( R R Y R Y R (A.13) 0 0 1 0 0 4/ ) ( ) ( R R J R J R (A.14) ) ( ) ( ) 2 (0 0 1 2 2 0 1r Y R r Y R R (A.15) ) ( ) ( ) 2 (0 0 1 2 2 0 2r J R r J R R (A.16) ) ( ) ( ) 2 (0 0 1 2 2 0 3r Y R r Y R R (A.17) ) ( ) ( ) 2 (0 0 1 2 2 0 4r Y R r J R R (A.18) 1 4 1 3 1 2 1 1 j y c TB A (A.19) 2 1 1 2 1 1 4 1 1 1 3 1 4 1 3 1 1 2 1 1 2 4 0 3 0 2 0 1 0 1 1) / / ( ) (c R c R c R c c R c R R c c R c R R R R B R A c Bj y j y j y T T T (A.20) 0 0 1 0 0 1/ ) ( ) (cR R Y cR Y c (A.21) 0 0 1 0 0 2/ ) ( ) (cR cR J cR J c (A.22)

PAGE 122

122 0 0 1 0 0 3/ ) ( ) (cR cR Y cR Y c (A.23) 0 0 1 0 0 4/ ) ( ) (cR cR J cR J c (A.24) ) ( ) ( ) ( ) (0 1 0 1 2 0 1 2 0 1 2 2cR Y cR J cR Y cR J B Aj y T T (A.25) )) ( ) ( ( ) ( ) ( ) ( ) ( ( )) ( ) ( ) ( ) ( ( ) (0 1 2 0 1 2 0 1 0 1 0 1 2 0 1 2 0 1 0 1 0 1 0 1 0 1 2cR J cR Y R Y R Y R J R Y X cR Y R J R Y cR J cR Y Bj y j y T (A.26) ) ( 2 1 ) ( 2 1 ] [4 0 3 0 2 0 1 0 0 4 0 3 0 2 0 1 0 4 1 3 1 2 1 1 1R R R B R A R R R R B R A R R R B R A Xj y T T j y T T j y T T (A.27)

PAGE 123

123 LIST OF REFERENCES [1] R. Aris, On the dispersion of a solute in a fluid flowing through a tube, Proc. Roy. Soc. A235 (1956) 67-77. [2] K.F. Bowden, Horizontal mixing in the s ea due to shearing current Journal of Fluid Mechanics, 21 (2) (1965) 83-95. [3] M. Furukawa, Design formulas for oscillat ory heat transport in open-ended tubes. Journal of Heat Transf er, 125 (6) (2003) 1183-1186. [4] H.B. Harris, S.L. Goren, Axial diffusion in a cylinder with pulsed flow, Chemical Engineering Science, 22 (1967) 1571-1576. [5] T. Inaba, M. Tahara, K. Saitoh, Longitudinal heat transfer in oscill atory flow in pipe bundles of various cross sections, JSME International Journa l, 43 (3) (2000) 460-467. [6] L.E. Johns, R. Narayanan, Interfacial In stability, Springer-Verlag, New York, 2002, pp. 16-58. [7] U.H. Kurzweg, M.J. Jeager, Determination of the longitudinal disp ersion coefficient in flows subjected to high-frequency osci llations, Physics of Fluids, 26 (6) 13801382. [8] U.H. Kurzweg, M.J. Jeager, Tuning effect in enhanced gas dispersion under oscillatory conditions, Physics of Fl uids, 29 (4) (1986) 1324-1325. [9] U.H. Kurzweg, L. Zhao, Heat transfer by high-frequency oscillations: A new hydrodynamic technique for achieving la rge effective thermal conductivities, Physics of Fluids, 27(11) (1984) 2624-2627. [10] S. Nishio, X.H. Shi, W.M. Zhang, Oscill ation-induced heat transport heat transport characteristics along liquid columns of oscillation-controlled heat transport tubes, International J ournal of Heat and Mass Tr ansfer, 38 (13) (1995) 2457-2470. [11] R.A. Peattie, R. Budwig, Heat transfer in laminar, oscillatory flow in cylindrical and conical tubes, Internatio nal Journal of Heat and Ma ss Transfer 32 (5) (1989) 923-934. [12] K.B. Ranger, Research note on the steady Poiseuille flow through pipes with multiple connected cross sections, Physic s of Fluids, 6 (6) (1994) 2224-2226.

PAGE 124

124 [13] G.I. Taylor, Dispersion of soluble matter in solvent flowing slow ly through a tube, Proc. Roy. Soc. A29 (1953) 186-203. [14] A.M. Thomas, Mass Transfer and Separati on of Species in Oscillating Flows, PhD Thesis, University of Florida, Gainesville, Florida, 2001. [15] A.M. Thomas, R. Narayanan, Physics of oscillatory flow and its effect on the mass transfer and separation of species, Physics of Fluids, 13 (4) (2001) 859-866. [16] A.M. Thomas, R. Narayanan, A comparis on between the enhanced mass transfer in boundary and pressure driven oscillatory flow, International J ournal of Heat and Mass Transfer, 45 (2002) 4057-4062. [17] E.J. Watson, Diffusion in os cillatory pipe flow, Journal of Fluid Mechanics, 133 (1983) 233-244.

PAGE 125

125 BIOGRAPHICAL SKETCH The author was born on June 3, 1980, in Pitt sburgh, Pennsylvania. His family moved to Northern Virginia in 1986, and he graduated fr om Centreville High School in Clifton, Virginia, in 1998. He attended the University of Notre Dame and graduated with a B.S. in chemical engineering in 2002. He will graduate with his Ph.D. in chemical engineering from the University of Florida in December 2006.


Permanent Link: http://ufdc.ufl.edu/UFE0017120/00001

Material Information

Title: Enhancing Heat Transport through Oscillatory Flows
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0017120:00001

Permanent Link: http://ufdc.ufl.edu/UFE0017120/00001

Material Information

Title: Enhancing Heat Transport through Oscillatory Flows
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0017120:00001


This item has the following downloads:


Table of Contents
    Title Page
        Page 1
        Page 2
    Dedication
        Page 3
    Acknowledgement
        Page 4
    Table of Contents
        Page 5
        Page 6
    List of Tables
        Page 7
    List of Figures
        Page 8
        Page 9
    List of symbols
        Page 10
        Page 11
        Page 12
        Page 13
    Abstract
        Page 14
        Page 15
    Introduction
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
    Oscillatory flow along a flat plate
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
    Mathematical models for oscillatory flow in a tube
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
    Oscillatory flow in an off centered annulus
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
    The experiment and results from the experimentation
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
    Conclusions and future research
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
    Appendix
        Page 120
        Page 121
        Page 122
    References
        Page 123
        Page 124
    Biographical sketch
        Page 125
Full Text






ENHANCING HEAT TRANSPORT THROUGH OSCILLATORY FLOWS


By

JONATHAN VOLK
















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2006































Copyright 2006

By

Jonathan Volk




























To my grandfather, John Wolsko (1911-2006)











ACKNOWLEDGMENTS

I would first like to thank my advisor, Dr. Ranga Narayanan, for his guidance during my

graduate study here at the University of Florida. I appreciate all the time he has spent with me in

the classroom and in the laboratory.

I would also like to thank my doctoral committee members, Dr. Oscar Crisalle, Dr. Alain

Kassab, Dr. James Klausner, and Dr. Aaron Thomas, for taking the time to evaluate my research.

I am grateful for the NASA GSRP Fellowship (grant number NNG04GO 12H), which

helped fund this research.

I am very appreciative of Alan Stottlemyer and Nicolas Munoz, for the time they spent

assembling our experiment and obtaining our results.

Ken Reed and TMR engineering did a wonderful job machining our experiment; we were

very fortunate to have them work with us.

I am thankful for all the teachers I have had since the very beginning of my schooling, I

have learned so much from all of them.

I would like to thank my family for all their support in my life. My mother and father are

the greatest parents in the world, and they have shown me how to be not only a good student, but

a good person. Lastly, I want to thank my grandfather, who constantly stressed the importance

of education. He has definitely influenced my life a great deal.












TABLE OF CONTENTS

A CK N OW LED GEM EN T S...... ... .................................................. ... 4

LIST OF TABLES ................... ............... ........................... ........ 7

LIST OF FIGURES................... ................ .................. ...... ........ 8

LIST OF SYMBOLS ................... ............... ......................... ........ 10

ABSTRACT .............................................................................. ....... 14

CHAPTER

1 INTRODUCTION................... ................ .................... ......... 16

1.1 Physics of Mass Transport................................... ..................... 16
1.2 Application of Mass Transport .................................................... 19
1.3 Physics of H eat Transport.................. .............. .............. ...... .... 19
1.4 A applications of H eat Transport ............................................... .......... 20
1.5 Prior Research in Oscillatory Flows ......... ........... ..................... 21
1.6 Project Objectives................... .......................................... .... 25

2 OSCILLATORY FLOW ALONG A FLAT PLATE............................... 28

2.1 Model for Oscillatory Flow Along a Flat Plate........... .................... 28
2.2 Analysis of Flat Plate Model ........... ........ .............................. 32
2.3 Discussion Note 1: Oscillating Input in a CSTR.......... .................... 35
2.4 Discussion Note 2: Rod Heated Through Oscillatory Temperature Gradient.... 37

3 MATHEMATICAL MODELS FOR OSCILLATORY FLOW IN A TUBE......... 42

3.1 O scillatory Flow in a Cylindrical Tube............................................... 42
3.1.1 Pressure D riven Flow .................. ............. ..... ............ 42
3.1.2 Piston D riven Flow .................. ............... ........ .. .......... 48
3.1.3 Boundary Driven Flow.................. ............. ................ 51
3.2 Oscillatory Flow in an Annulus...................................................... 53
3.2.1 Pressure D riven Flow .................. ............... ...... .. ......... 53
3.2.2 Piston D riven Flow ............... ................... ........ .. ......... 55
3.2.3 Boundary Driven Flow................................................ 55
3.3 Comparison of Annular and Cylindrical Geometries............ ............... 56
3.4 Comparison of Two Compartment Flow versus One Compartment Flow....... 61
3.4.1 Heat Transfer................... ... ................ ....................... 61
3.4.2 Mass Transfer and Separation of Species............... ........... 66









4 OSCILLATORY FLOW IN AN OFF CENTERED ANNULUS.................... 76

4.1 M odeling an Off Centered Annulus.................. ........................... 76
4.2 Analysis of Off Centered Model........... ......... ... ......... ........ 83

5 THE EXPERIMENT AND RESULTS FROM THE EXPERIMENTATION...... 87

5.1 E xperim mental Set-U p ........................................................ ... ...87
5.2 Experim ental M ethod....................................................... ........ 89
5.3 Error Analysis ................... .. ................. ................ ... .... .. 92
5.4 N ote on Viscous Dissipation................. ............................. ........ 96
5.5 Experimental Results and Discussion ....................................... 97
5.5.1 Varying Amplitude with W ater................ .. ........ ......... 97
5.5.2 Varying Amplitude with Glycerol-Water Mixture.................... 99
5.5.3 V arying Frequency.................. ................ ......... .......... 101
5.5.4 V arying Tube R adius.................. ............... ...................... 105
5.5 Summary of Results................... ....... ........... ........................ .. 107
5.6 Potential Improvements on this W ork................................... ......... 107

6 CON CLU SION S................... .............................................. .. 115

6.1 C onclusions.................................... ......... ......... 115
6.2 Future Research in Oscillatory Flows.................. ............... ......... 118

APPENDIX

CONSTANTS IN THE SOLUTIONS TO THE OFF CENTERED
ANNULUS PROBLEM.... .............................. ........ 120


LIST OF REFEREN CES...................................... ................... ..... .. 123

BIOGRAPHICAL SKETCH .................. ............... .............................. 125












LIST OF TABLES


Table Page

5.1 Physical Properties of Water at Different Temperatures................... ......... 114

5.2 Physical Properties of a Glycerol-Water Mixture at Different Temperatures......... 114












LIST OF FIGURES


Figure Page

1.1 Physics of mass transfer using oscillatory flows..................................... 26

1.2 Physics of species separation using oscillatory flows............... ............... 26

1.3 Application to species separation in the environment ................................. 27

2.1 Design for oscillating flat plate problem......................... ....... 40

2.2 Heat flux at various Prandtl numbers as a function of Womersley number........... 40

2.3 C STR w ith oscillating input.................. .................. ..................... .... 41

2.4 Rod with oscillatory heat source on one end.............................................. 41

3.1 Oscillations driven by pressure drop.................... ........ ... ....... ......... 68

3.2 Oscillations driven by a piston.................. ................. ............ ... ...... 68

3.3 Oscillations driven by moving boundary..... ................................... 68

3.4 Oscillations in an annular geometry................................. ............... 69

3.5 Heat transport in annular versus cylindrical geometry with the outer radius and
pressure drop held fixed................... .................... ........................ 69

3.6 Heat transport in annular versus cylindrical geometry with the outer radius and flow
rate held fixed...................................... ................... ....... ........ 70

3.7 Heat transport in annular versus cylindrical geometry with the cross sectional area
and pressure drop held fixed............................................................... 70

3.8 Heat transport in annular versus cylindrical geometry with the cross sectional area
and flow rate held fixed................... .................... ................ ...... 71

3.9 Output/Input ratio with fixed pressure drop and cross sectional area................. 71

3.10 Diagrams for one and two compartment models respectively ......................... 72

3.11 External piston oscillating one compartment model and two compartment model... 72









3.12 Heat transport in the one vs. two compartment model at various frequencies
at sm a ll 6 ...................................... .................................... 7 3

3.13 Heat transport in the one vs. two compartment model at various frequencies
at m edium 6 ................................ .... .................. ........ ....... 73

3.14 Heat transport in the one vs. two compartment model at various frequencies
at large 6 ................... ................. ................... ......... ........ 74

3.15 Separation ratio for various frequencies for the one compartment model and two
compartment model at small 6.................. ................................ ......... 74

3.16 Separation ratio for various frequencies for the one compartment model and two
compartment model at medium 6............................................. ........... 75

3.17 Separation ratio for various frequencies for the one compartment model and two
com partm ent m odel at large 6.................. ............. .............. .......... 75

4.1 Difference between centered annulus and off-centered annulus.................... 85

4.2 Net change in transport due to off centered inner rod at various frequencies........ 85

4.3 Thin and thick regions that result from an annulus that is off-centered .............. 86

5.1 Experimental set-up......................................................... .........109

5.2a Schematic diagram of heat source..................................................... 110

5.2b Schem atic diagram of heat sink ............ ........... ... ...... ........ ........ .. 110

5.3 Heat transport increase as a result of piston amplitude increase for low Prandtl
number fluid................... ........ ... ........... .............. ... 111

5.4 Heat transport increase as a result of piston amplitude increase for high Prandtl
number fluid................... ........ ... ........... .............. ... 111

5.5 Heat transport increase as a result of oscillation frequency increase for high Prandtl
number fluid................... ......... ... ........... ............. ... 112

5.6 Large tube with separate insert, insert placed in large tube creating an air filled
gap .......................... ...... ........... ......... ................. 112

5.7 Heat transport increase as a result of oscillation frequency increase for high Prandtl
number fluid................... ......... ... ........... ............. ... 113

5.8 Experimental ratio compared to theoretical ratio for changing radius ............... 113











LIST OF SYMBOLS

A Cross sectional area of channel. Applies to Chapter 1.

A Oscillation amplitude, cm.

A Scaled amplitude. Applies to Chapter 2.

A Constants in solutions to perturbation equations. Applies to Chapter 4.

b Width where 99% of heat transfer occurs in oscillating plate problem.

B Constants in solutions to perturbation equations. Applies to Chapter 4.

c Coefficient for outer radius of annulus.

C Concentration of CSTR. Applies to Discussion Note 1 in Chapter 2.

C Constants in solutions to velocity and temperature equations.

Cm Inlet concentration of CSTR. Applies to Discussion Note 1 in Chapter 2.

C* Scaled concentration of CSTR. Applies to Discussion Note 1 in Chapter 2.

Cv Specific heat, J/(g*C).

d Channel width. Applies to Chapter 1.

D Molecular diffusion coefficient, cmA2/sec.

g Gravitational constant.

i = Imaginary number.

J Heat flux, W/cmA2.

JcoND Conductive heat flux, W/cmA2.

JcoNV Convective heat flux, W/cmA2.

k Effective diffusion coefficient. Applies to Chapter 1.

k Thermal conductivity, W/(cm*C).









K Temperature drop per unit length. Applies to Chapter 2.

K Input amplitude of driving force. Applies to Discussion Notes 1 & 2 in Chapter 2.

L Length of pipe, cm.

P Pressure, g/(cm*sec^2).

Pr Prandtl number, dimensionless.

Q Volumetric flow rate for one compartment model. Applies to Chapter 3.4

Q' Volumetric flow rate for two compartment model. Applies to Chapter 3.4.

QCOND Conductive heat transfer, W.

QCONV Convective heat transfer, W.

QroT Total heat transfer, W.

r Radial coordinate

R Radial distance, cm.

S Molecular diffusivity factor. Applies to Chapter 1.

S Ratio of flow rates for one and two compartment model. Applies to Chapter 3.4.

t Time, sec.

T Temperature, C.

T* Scaled temperature. Applies to Discussion Note 2 in Chapter 2.

Uo Maximum velocity at the axis. Applies to Chapter 1.

V Volume of the channel. Applies to Chapter 1.

V Velocity, cm/sec.

W = cor Scaled time. Applies to Discussion Note 1 in Chapter 2.


W = L Scaled time. Applies to Discussion Note 2 in Chapter 2.
K










Wo = A F



Wo =oR
V


Womersley number, dimensionless. Applies to Chapter 2.



Womersley number, dimensionless. Applies to Chapter 3.


Coordinate in the x-direction.

Scaled length. Applies to Discussion Note 2 in Chapter 2.

Grouping used to simplify perturbation equations. Applies to Chapter 4.

Coordinate in the y-direction.

Coordinate in the z-direction.

Grouping used to relate pressure drop to piston amplitude in annulus.


Greek Symbols


V v


a'= )
V=



V K


A w
2


Viscous length grouping to assist in expressing solutions, cm.



Conjugate of a.



Diffusive length grouping to assist in expressing solutions, cm.


Size of solid gap in two compartment model, cm.

Distance inner rod is displaced, cm. Applies to Chapter 4.

Angle measurement, radians.

Thermal diffusivity, cmA2/sec.


Dimensionless parameter. Applies to Discussion Note 2 in Chapter 2.


Dynamic viscosity, g/(cm*sec).









v Kinematic viscosity, cmA2/sec.

r Constants in solutions to perturbation equations. Applies to Chapter 4.

p Density, g/cmA3.

r Residence time, sec.

co Frequency, Hz.


Superscripts

A, ~ Complex conjugates.

Denotes complex conjugate (overbar)









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

ENHANCING HEAT TRANSPORT THROUGH OSCILLATORY FLOWS

By

Jonathan Volk

December 2006

Chair: Ranga Narayanan
Major Department: Chemical Engineering

This research project is about the use of oscillatory flows to enhance heat transport. The

physical idea behind this is that if a heat source is connected to a heat sink via a fluid and the

fluid is oscillated, the convective motion will bring about sharp spikes in the velocity profile

which in turn will enhance the heat transport over pure conduction due to both radial and axial

gradients. This research has many applications including removing heat from outer space

modules, reactors, closed cabins, etc. Many parameters can be varied to maximize the

convective heat transport. These parameters include pulse amplitude and frequency, pipe radius

and length, as well as the transporting fluid and the geometry of the transport zone.

This problem shows from a theoretical standpoint how various pipe geometries affect the

heat transfer. For example, an annular geometry is studied and it is seen that an annulus is an

optimal geometry compared to a cylinder, when the flow rates through both are held fixed. This

happens because the annular geometry contains an extra boundary compared to an open cylinder

and the spike in the velocity is greater than the open cylindrical geometry. The comparison was

taken one step further using an annulus and a two compartment model containing an inner

cylinder and an outer annulus. The transport in the two compartment model proved to be higher

in the case of fixed flow rates, due to the additional boundary. Lastly, an off-centered annulus









was studied, and it was shown that the net enhancement is positive for low frequencies and

negative for high frequencies compared to a centered annulus.

An experiment was designed to test the various parameters listed above. The

experimental results that have been obtained are in good qualitative agreement with the

theoretical calculations. The experiment has shown that the heat transport is proportional to the

square of the oscillation amplitude. The study also demonstrated that heat transfer increases with

rising frequency. In summary, it has been shown through the experiment and the theory that

oscillatory flows increase the heat transport of a system by several orders of magnitude

compared to ordinary conduction.











CHAPTER 1
INTRODUCTION

This study is concerned with the investigation of heat transfer in enclosed systems. The

goal is to find ways to enhance the transport of heat from a heat source to a heat sink using fluid

contained connectors. These methods also need to be efficient and many times faster than

molecular conduction. The application of these findings has a place in NASA's long-term

manned and un-manned missions. In a space vehicle, heat can be generated due to equipment,

such as an on-board reactor. The thermal control of the environment is of utmost importance for

the maintenance of crew habitat, or even for the safe operation of instruments. This heat is best

conducted to external radiator sails that can then emit the heat to outer space. Getting the heat

from the source to the heat sinks could be done via heat pipes and metal conductors, but the mass

of such equipment is high. It is important to keep the mass on board minimal due to the immense

weight during lift off and acceleration maneuvers. This then calls for a novel way to transport

heat from high temperature sources in enclosed environments to sinks using fluid connectors.

One such way is to use the principle of oscillatory flows to enhance heat transport. This involves

the use of connecting pipes that contain a fluid which is in perpetual motion, so that the heat can

be conveyed fast. To understand how the heat can be transported by simply moving fluid back

and forth, a companion problem will first be presented, which involves the mass transport of

species.

1.1 Physics of Mass Transport

Imagine two tanks connected by a pipe with one tank containing a species while the other

tank contains a carrier. If a connecting pipe between the tanks were left open, a concentration









gradient would be formed between the two tanks in the direction of the axis of the pipe (see

Figure 1. la). This gradient would ultimately decrease and equilibrium would be achieved.

Now suppose one wants to quicken the rate at which the species move down the pipe.

One way for this to occur is through the use of oscillatory flows. When a fluid is oscillated in a

pipe through some driving force, it acquires a parabolic profile far away from the entrance and

exit regions (Figure 1.lb). This curved shape generates a larger area for mass transfer to take

place, mostly in the radial direction. The species will have a net movement down the pipe and

towards the boundary. These radial concentration gradients play a large role in quickening the

rate of transport. In ordinary molecular diffusion, there exist only axial concentration gradients

for the species to move. However, once the oscillations begin, the particles have more ways to

move toward the end of the pipe (radially and axially). Since an oscillatory flow means that the

fluid must return to its original position, a reverse parabola occurs (Figure 1. 1c). In the forward

part of the cycle, there was a net movement of species toward the boundary. Now since there is

a higher local concentration of species at the boundary compared to the core, there will be a

movement of species back toward the core of the pipe. Although there is no net flow in this

problem, the species will work its way down the pipe in a "zig-zag" fashion as the pulsations

occur (Figure 1.1 d). It can be shown that these oscillations can rapidly improve the rate at which

the species is transported down the tube.

Now take the case where there are two species with a carrier fluid in one tank and pure

carrier in the other, as shown in Figure 1.2. Suppose it is desired to separate the two species. If

the system is left without any induced flow, the species will diffuse down the pipe at a rate that is

dependent on their diffusion coefficients. Take for example two species whose diffusion

coefficients are quite different from each other, helium and carbon dioxide. Carbon dioxide is









the larger molecule so it will have the slower diffusion coefficient. This means under the case of

pure diffusion, the smaller helium molecules will move down the pipe quicker and can be

separated out first (see Figure 1.2a). The carbon dioxide would then follow later. However, if

oscillations are imposed, one will notice that something different occurs (see Figure 1.2b). As

mentioned above, the pulsations will push the species toward the boundary of the tube. The

species most likely to be pushed to the boundary is the faster diffuser, the helium. The carbon

dioxide will remain behind in the center of the tube. When looking at the velocity profile in the

tube, one will notice that the fluid speed is faster in the center compared to the boundary. This

means that the carbon dioxide will actually move faster down the tube since it is in the quicker

flow region. The helium can be thought of as getting "stuck" near the slow moving boundary

region. A separation is achieved with the slower diffuser being transported faster. This

separation will also occur much faster compared to the case where only pure diffusion happens,

since the oscillations force the two species down the pipe at faster rate.

In order for a separation to be maximized, several parameters must be optimized. In all

chemical engineering problems, there exist time scales. This one is no different. In fact,

multiple time constants arise that must be noted. These time constants involve the oscillation

frequency, the kinematic viscosity of the carrier fluid, as well the diffusion coefficients of the

species. For example, the oscillation period must be at a rate that is comparable to the time

constant due to the kinematic viscosity of the fluid, so that a parabolic profile can exist. If the

fluid has a low kinematic viscosity, the frequency must not be too high, or the inertial force from

the pulse would overpower the viscous effects at the walls, causing the velocity profile to be

much flatter, and reducing the concentration gradient in the radial direction. In fact, this suggests

the existence of a dimensionless group which is a ratio of the two time constants. The analysis









will show the natural occurrence of such a group, the Womersley number. Another example is

that the diffusion time constant for each species must be different. Otherwise, both species

would move down the pipe at the same rate, and a distinct separation would not occur. Lastly,

the viscous time scale must be greater than the diffusive time scale, since the viscous effects play

a big role in producing the radial concentration gradients. This is because a more viscous fluid

can generate a sharper parabolic flow profile due to the higher amount of friction at the tube

walls. A longer and sharper parabola produces more radial concentration gradients which will

increase the transport in the tube compared to ordinary diffusion.

1.2 Application of Mass Transport

A basic application for enhanced species transport is shown in Figure 1.3. In this

example, it is important for human systems to have clean air to breathe. When humans generate

carbon dioxide as they exhale, it becomes mixed in with air. This mechanism could assist in

removing the carbon dioxide from the air so that it is pure when it is regenerated by plant

photosynthesis. Imagine if there is a carbon dioxide and air mixture in the left hand tank, while

the right hand tank consists of only water vapor. When the oscillations are induced, the slower

diffusing carbon dioxide molecules will reach the right hand tank faster, and are able to be

separated out first, leaving the air free from high levels of carbon dioxide.

1.3 Physics of Heat Transport

The same idea of pulsatile flows can be applied to transporting heat between a source and

a sink. Take the case of a fluid in a channel. Suppose at one end, the fluid is in a hot tank, while

at the other end, it is in a cold sink. Heat will be transported from the source to the sink by

conduction if the fluid remains stagnant; this heat will flow down the channel at a rate

proportional to the thermal diffusivity of the fluid. If the fluid is oscillated back and forth, radial









temperature gradients will form (see Figure 1.lb) and the heat movement would be enhanced

down the pipe through the zig-zag movement discussed earlier. The oscillatory movement of the

fluid transports the heat through convection, and when the appropriate parameters are chosen for

the system, meaning the type of fluid and the geometry, the rate of convection can be

significantly higher than the rate of conduction. Like in the mass transfer analog, several time

scales exist. For the heat transport problem, the time scales will involve the thermal diffusivity,

along with the kinematic viscosity and oscillation frequency as before.

The first time scale involves the thermal diffusivity of the fluid medium. This rate

determines how effectively heat is transported through conduction. Just like in the mass transfer

example, the viscous time scale (determined by the kinematic viscosity of the fluid) needs to be

much larger than the diffusive time scale, so that heat can be transported in multiple directions

through convection via the radial temperature gradients, as opposed to only axially, which occurs

when the diffusive time scale dominates. The time constant for the oscillation frequency must

again be similar to the viscous time scale in order to ensure the formation of a sharp, parabolic

spike in the velocity profile.

1.4 Applications of Heat Transport

In the beginning of this chapter, the application of oscillatory flows for use in enclosed

space capsules was described. One of the main advantages of this method was the limited mass

required. The mass can be kept minimal since there is no net flow. The fluid to fill the channel

is all that is needed, and can reused throughout. This mechanism can be run with simple fluids

such as water and air. No volatile or hazardous chemicals are needed, which reduces the number

of potential safety risks on board. Compare this to a heat pipe, where vaporizing fluids are

sometimes used, and the pulsatile flow model seems fit for use in space capsules.









There are many relevant uses on Earth for this model as well. Many industrial processes

and appliances generate excess heat that needs to be removed. These include exothermic

reactions such as in polypropylene and polybutylene production, as well as in fuel cells. This

would be useful for heat removal from hazardous substances such as radioactive fluids. One of

the main goals of this mechanism is to remove heat quickly and efficiently so that it can be

useful in many different types of processes.

1.5 Prior Research in Oscillatory Flows

The roots of this problem can be traced back to Taylor [13] and his work in dispersion.

He showed through his work that a soluble substance in a long and narrow tube spreads out due

to two effects: molecular diffusion and velocity gradients in the radial direction. He developed a

model for dispersion that contains a diffusion coefficient which is now known as the Taylor

diffusion coefficient. This "effective" diffusion value is

R2U2
k (1.1)
D

Here R is the radius of the tube, Uo is the maximum velocity at the axis, and D is the

molecular diffusion coefficient. The important thing to notice is that this effective diffusion

coefficient is inversely proportional to the molecular diffusion coefficient. This parallels the idea

that at appropriate oscillating frequencies, there will be a reverse in the separation, where the

species with the smaller molecular diffusion coefficient will have the higher effective diffusion

coefficient and separate out quicker. It can also be seen from (1.1) that the best separation

achievable when the fluid is oscillated is proportional to the molecular diffusion coefficient of

the small molecule divided by the molecular diffusion coefficient of the large molecule.

Aris [1] later used the method of moments to see if Taylor's result could be reproduced.

Taylor had to restrict certain parameters for his conditions to be valid, such as the dimensions of









the tube and the velocity of the fluid through the pipe. Aris' method eliminated those constraints

by modeling the solute based on its moments in the direction of the flow. Through this method,

Aris was able to obtain the Taylor diffusion coefficient. He also demonstrated how dispersion

can be altered by changing the geometry of the cross section. He was able to show that there is

less dispersion in an elliptical tube compared to a circular tube of the same area. This will be

important in the modeling chapter when convection in various geometries is discussed. Bowden

[2] also applied Taylor's idea to show that this process occurs in large scales as well. In the sea

for example, there is horizontal mixing due to a shearing current. This shearing current is an

alternating, or oscillating, flow. He was able to show through experimental observations how the

effective coefficient of horizontal diffusion is inversely proportional to the vertical eddy

diffusion coefficient.

Harris and Goren [4] developed a model for oscillatory mass transfer in an open tube

using an oscillating piston to drive the flow. They were able to show through this model that the

enhanced species transport is proportional to the square of the piston amplitude. Their model

was also developed to show how much the oscillations increase the mass transport over pure

molecular diffusion. They were able to show how their model produces Aris' result for the

Taylor diffusion coefficient when the frequency of oscillations and tube radius are small

compared to the kinematic viscosity and the molecular diffusion coefficient (low Womersley

numbers). They also conducted an experiment to test their model using HC1. The results were in

good agreement with their theory as they showed that the mass transport rate of HC1 can be

increased by as much as sixty times the rate of pure diffusion under the proper tuning conditions.

A similar model to Harris and Goren's is used in a later chapter to develop the heat transfer

model for piston driven flow.









Watson [17] later studied how mass flux increased in a circular pipe and a two

dimensional channel when the fluid was oscillated. He concluded that the effective diffusivity

was increased from the molecular diffusivity by a factor of S where


S -( >)2 (1.2)
Ad

Here co is the frequency of the oscillations, Vis the volume of the channel, A is the cross

sectional area of the channel and dis the width of the channel. An important point to note in this

result is that the enhanced transport is proportional to the square root of the frequency. This will

be seen in the derivation of the heat transfer model.

Kurzweg did extensive work in this field with Jaeger [7,8] and Zhao[9]. They did studies

in both mass and heat transfer applications. Kurzweg and Jaeger performed experiments to

verify Watson's relation for the effective diffusion coefficient under oscillatory conditions. They

were able to obtain a good data correlation with Watson's theory which shows that the transport

increases with the square root of the frequency. In this study, they also found that there exists an

optimal frequency for a particular system (fluid medium and dimensions of tube), at which

maximum transport will take place. Their results show that as frequency is increased starting

from rest, there will be an increase in transport until it reaches the optimal tuning frequency, after

which point it will decay. They also presented a study in high frequency oscillations in which

the longitudinal dispersion coefficient was proportional to the square of the spike amplitude (as

in Harris and Goren), and to the first power of frequency. They determined that the dispersion

under oscillatory conditions can be more than twice as much compared to steady turbulent flow.

Kurzweg and Zhao [9] developed an experiment in heat transfer to enhance the thermal

diffusivity of a fluid. They connected a hot and cold reservoir with a capillary bundle containing

several small tubes. The oscillations were imposed by shaking a flexible membrane on each









tank. They were able to enhance the thermal transport in water by almost 18,000 times its

molecular thermal diffusivity. It was also shown in their theoretical model that oscillating a

liquid metal such as lithium can achieve a heat flux that is two orders of magnitude greater than

those obtained in liquid metal heat pipes.

Others have studied the capillary and tube configurations such as Inaba et al. [5] and

Peattie and Budwig [11]. They found that the tube or capillary geometry does affect the overall

heat transport, and needs to be considered when designing a model. In separate studies, the input

power required to oscillate a fluid in a heat transfer configuration has been examined by Nishio

et al. [10] and Furukawa [3]. Nishio examined the optimum ratio of heat transport to input

power and found that the ratio can be increased when there is a phase shift between the

oscillations in adjacent tubes within the capillary bundle. Furukawa developed algebraic

expressions for several output parameters as a function of the Womersley number, including

input power, and the number of capillary tubes needed to achieve a desired transport rate.

Thomas and Narayanan [15,16] were able to compare the differences between the various

types of driving forces used to produce oscillations. They determined that pressure driven

oscillations produced a greater convective mass transport compared to the case where the

moving outer wall of the tube produced the pulsations. It was also found that the ratio of

convective transport to input power was identical in both cases except for extremely low

frequencies where the boundary driven flow had a higher ratio. They also studied "cross over"

frequencies for separations in pulsatile flow, meaning frequencies at which the two species

switch in being the faster and slower transporter. It was determined that as many as three cross

over frequencies can arise in certain binary gaseous systems. Thomas [14] also conducted

several experiments to show that these cross over frequencies existed. He used different binary









gaseous mixtures and was able to induce the pulsations through a moving boundary. The

separations were measured through the use of a gas chromatograph. In addition to finding the

cross over frequencies, he also validated Kurzweg et al. that an optimal frequency existed, at

which the separation would be at a maximum. All of this prior work has assisted in formulating

the objectives for this project, where further progress can be made in using oscillatory flows to

enhance heat transport.

1.6 Project Objectives

This project has both theoretical and experimental components. From a theoretical

standpoint, the oscillatory flow model is derived after showing some simpler problems which

will give the reader a better understanding of the physics. In certain cases, previous theoretical

results will be verified. The model is presented in three different ways since the oscillations can

be driven in various methods. The model is also discussed using various geometries which may

or may not enhance the transport. These different geometries include a centered annulus, an off

centered annulus, and a two compartment model involving both a cylinder and an annulus. The

objective here is to determine what geometry produces the optimal transport from a theoretical

perspective, as well as to see if that result changes depending on what the input parameters and

conditions are.

An experimental set-up was constructed to test the basic cylindrical model for heat

transport. The testing was conducted by varying key input parameters such as the amplitude of

the spike, the frequency of the oscillations, the dimensions of the tube, and the type of fluid. The

experimental results were compared to the theoretical model to see if they are in good agreement

with each other. Also, in the case in which the experimental results deviate from the model,

possible reasons are discussed. This experiment showed that heat can be transported more










effectively through the use of oscillatory flows, and can therefore be applied to removing heat in

the situations described earlier, whether it is on a large or a small scale.


rr
L Sp o .I .I. I








r'

m~ Ii
C ~-''i ^

-- '


Figure 1.1: Physics of mass transfer using oscillatory flows; (a) pure molecular diffusion in tube,
(b) forward pulse in oscillatory flow with molecules moving toward the boundary, (c) reverse
pulse with molecules moving toward the core, (d) end result with molecules moving down at a
faster rate compared to pure diffusion.




OL-

Species m a Carrier






Specie-- Carrier





Figure 1.2: Physics of species separation using oscillatory flows; (a) molecular diffusion with
smaller molecules moving down tube faster (b) oscillatory flow with larger molecules remaining
in the faster core region, resulting in faster transport compared to the smaller molecules



























Figure 1.3: Application to species separation in the environment











CHAPTER 2
OSCILLATORY FLOW ALONG A FLAT PLATE

In order to understand the physics of the oscillating flow in a pipe, a simple case will first

be discussed involving an oscillating plate in an infinite medium of fluid. The oscillatory nature

of the plate motion is mathematically expressed as a cosine wave function. Figure 2.1 illustrates

this design.

2.1 Model for Oscillatory Flow Along a Flat Plate

The oscillating plate is assumed to move in the horizontal direction. "A" represents the

amplitude of the oscillating stroke, and co is the frequency of the plate motion. The velocity

profile in this case depends only on "y." The Navier-Stokes equation in the y-direction becomes

OV 02V
v-=- (2.1)
at y 2

The boundary conditions are given by no slip at the plate (y=0), meaning the fluid will

have the same velocity as the oscillating plate (1/2 Aocos(ot)), and that at an infinite distance

away from the plate, the velocity of the fluid will approach zero. The simplest way to solve this

equation is to express the velocity field as a sum of two exponentials, that is


V =Ve'0st +Veot (2.2)

Since one term is a conjugate of the other, the algebraic manipulations are simplified and

the problem can be solved using either term, meaning that

i o) V 02 V
(2.3)
v y 2

The solution to this equation can be found to be










V = Ce + C2e (2.4)

It can easily be seen from the boundary condition far away from the wall that C1 must be

equal to zero since the velocity must be zero. The boundary condition at the wall can be

rewritten using the exponential relationship for a cosine wave, which is

el0 +e
cosO = (2.5)
2

Therefore the new boundary condition at the wall will be

1
V(y =)= -Ac (2.6)
4

C2 can now be solved for, and the result is

A Ao (1O+~v
V = --e (2.7)
4

The entire expression for velocity can be found by taking twice the real part of (2.7).

Therefore


V=-Ao)e cos(Ot y) (2.8)
2 2v

This method of solving the differential equation and the result is useful in the next step of

this problem. The velocity field is used in finding the temperature profile. As in the equation of

motion, only the temperature change in the y-direction will be obtained. Hence

dT AT 2T
-+ V = c (2.9)
at L ay2

The temperature field can be rewritten as a sum of complex exponentials much like the

velocity. Therefore


T Te'a+t Te-'ot (2.10)









The temperature equation now becomes


is AT ^ 02 T
T+ V AT =2T (2.11)
K LK y2

The AT/L term is the temperature change in the horizontal direction and can be assumed

to be a constant in this case (K), while K is the thermal diffusivity. The upper boundary condition

is similar to the velocity field, where at an infinite distance from the plate, the heat flux tends to

zero. The lower condition would be that there is no heat flux through the wall (dT/dy=O). The

solution becomes

(1+) ~v (1+) KA2,Y 1 (1+) oy
T =Ce 2+C 4e (-- )e 2(2.12)
4i 1-K
v

Here C3 is zero due to the far-field boundary condition. The wall boundary condition can

be used to solve for C4 yielding

iKA 1 K (1+, (2.13)
T=- ( )( -e ) (2.13)
4 1-/ Vv


Equation (2.13) can be used to solve for the actual temperature profile in (2.10) giving


T =[ e sin(ot y) e sin(t y)] (2.14)
2(1_ K v 2yK )2v
2(1- )
V

The last quantity that needs to be found is the total heat flux. This can be expressed as

the sum of the diffusive flux and the convective flux. The flux that is found is averaged over

time and space to give

2ir 2ir

J = f (-Kk)dtdy +pC ) f (VT)dtdy (2.15)
0b20 b 20o









In this expression, k represents the thermal conductivity of the fluid, p is the density of

the fluid, C, is the specific heat of the fluid, and b represents a certain distance from the

oscillating plate. It is determined so that 99% of the heat transfer takes place within a region of

width 'b'. When (2.2) and (2.10) are inserted into (2.15), and the time integral is taken, only the

cross terms will remain since the other terms integrate out to zero. This will be explained later in

further detail. The equation for flux becomes

S b _
J=-Kk+2Re[ v (VT)dy] (2.16)
b

Using the results found in (2.7) and (2.13), performing the integration and rearrangement

produces an expression for the heat flux. It is


pCLKA2 K-V
J Kk e [( + ))
1 K 2K v
8( 2 )
K V

sin( b b) + ) cos( b b)] + (2.17)
2K 2v 2K 2v 2K 2v






The flux can also be expressed in dimensionless variables. This will be beneficial in

analyzing the results and understanding the role that these different time scales play in the

physics of the problem.

Define the Womersley and Prandtl numbers to be


Wo= A (2.18)

and2v
and










Pr = (2.19)


The amplitude can be made dimensionless as well. Thus

A A (2.20)
b

The conductive heat flux is

JOND = -Kk (2.21)

Equation (2.17) can now be expressed as a ratio of convective flux to conductive flux.

This ratio is

JCONV A {e _(WPr1/Z+Wo) /A
JCOND 8Pr1 2(l-Pr 2)
WoPr /2 Wo
[(WoPr' +Wo)sin( *)+(WoPrl -Wo)* (2.22)
A A
WoPr1 2 Wo
cos( + (Wo WoPr )}+ 1
A A

2.2 Analysis of Flat Plate Model

A simple analysis can be performed on (2.22) to see what the flux should be for extreme

values of the dimensionless terms. In the simple case where the plate is not oscillating, the ratio

equals 1, which makes physical sense as the only transport that occurs is through conduction.

Another case to examine is when the Prandtl number becomes much greater than 1. Taking the

limit as Pr goes to infinity produces a result that is

JCON A*Wo
C (A*W + 1) (2.23)
J 8
JCOND 8

This approximation can be useful in estimating the flux for fluids with high Prandtl

numbers (such as oils). A physical interpretation for this case is to look at the problem when the

kinematic viscosity is much larger than the thermal diffusivity (the definition for a high Prandtl









number). Looking at the equation for the velocity field (2.8) and taking the limit when v tends to

infinity, the result is the same as the velocity of the plate. This means that every particle of fluid

moves at the same velocity as the plate. In this case there won't be some fluid elements moving

with the forward stroke and others on the backward stroke; instead they will all move together.

This promotes a higher rate of heat flux due to convection as opposed to conduction. This is

what would be expected since the thermal diffusion coefficient is extremely low compared to the

kinematic viscosity. It can also be noted from this equation that the ratio is proportional to the

square of the amplitude (observe that the amplitude appears in the Womersley number as well).

The frequency of the oscillations, the amplitude of the stroke, and the physical properties of the

fluid are all input variables in this case, while the convective to conductive ratio is the output

result. The equation for an infinite Prandtl number fluid (2.23) can be shown to be accurate by

looking at Figure 2.2, which uses (2.22) to show heat flux for various Prandtl numbers. It can be

seen that for the higher Prandtl number fluid (Pr=500), the heat flux approaches the graph for Pr

tending to infinity. This same line is the plot of (2.23). It also makes sense since for high Pr

fluids, the heat flux increases linearly with the Womersley number from (2.23). It is obvious that

a linear plot results in Figure 2.2. Therefore, it can be concluded that (2.23) is an accurate

representation of (2.22) for an infinite Prandtl number.

Although it can be seen in Figure 2.2 that the convective heat flux increases with

increasing Prandtl number, it should be pointed out that it only increases for a given Womersley

number, and that kinematic viscosity appears in the denominator of the Womersley number.

When the kinematic viscosity (or Prandtl number) becomes high, the convective transport only

increases if there is an increase in the amplitude or frequency of the plate to keep the Womersley

number constant. This makes physical sense since a fluid with a high viscosity is very sluggish.









Although the fluid elements are more likely to move with the velocity of the plate as stated

above, there is more energy required by the system to shear a very viscous fluid back and forth.

This extra energy would come from increasing the size and/or frequency of the oscillations. In

fact, there will be a decrease in the transport if only the kinematic viscosity is varied and nothing

else. This becomes an issue during experimentation, and is discussed more in Chapter 5.

Lastly, the flux value can be found when the Prandtl number approaches 1. This will

help in determining a result for those fluids with low Prandtl numbers such as gases. Taking

(2.22) as a limit when Pr approaches 1, we get

J A* 2Wo 2W02
NV = (e A ( + Wo)-Wo)+1) (2.24)
JCOND 32 A

This equation shows that for a low Prandtl number fluid, the convective to conductive

ratio will be lower than the case for a very high Prandtl number fluid simply by the fact that the

ratio in (2.24) roughly scales by a factor of e-2/32 compared to 1/8 in (2.23). This would make

sense since a lower Prandtl number fluid can mean it has a high thermal diffusivity; such as in

the case of air. Not only will this drive up the conductive heat transfer, but since gases are

lighter and have a lower viscosity than liquids, they will feel less of the shearing effect brought

about by a moving plate. This means that the molecules will be less likely to move with the

speed of the plate, and instead move at a rate proportional to the thermal diffusivity of the gas.

This means less transport via convection. For both of these reasons, the convective to

conductive ratio will decrease.

Although simple from a physical standpoint, the mathematics of the problem can be

somewhat complicated. It is important to understand not only the physics of the problem, but the

mathematical techniques as well, since they will be applied in more detail to the scenario

involving oscillating flow in a pipe. Before that point is reached, two other example problems









will be presented to show how the length and time scales relate to the physics of the problem.

These discussion notes are very basic chemical engineering problems, however a lot can be

obtained from them. A key idea to take from these problems is to see how an input function (in

these cases an input oscillation) affects the behavior of a system. Also, it is important to

understand how these input functions affect the system response depending on the size and type

of the input signal, as well as the size of the system. The first problem presented will involve a

simple CSTR with an oscillating input. It can be easily solved from a mathematical standpoint

and clearly shows the relationship between input signals and system response. The second

problem is a little more complicated mathematically, but still shows the same relationship. This

problem is one where a solid rod is heated at one end in an oscillatory manner. Since the first

example relates to mass transfer, and the second relates to heat transfer, these problems should

help the reader conclude that these systems behave in a similar way, regardless of the transport

medium.

2.3 Discussion Note 1: Oscillating Input in a CSTR

Figure 2.3 shows a constant stirred tank reactor that is receiving input concentration as a

sinusoidal oscillation. Writing the mass balance for the CSTR, we get

dC(t)
C (t) C(t) = (2.25)
dt

where C is the concentration in the tank, and r is the residence time which is equivalent to the

volume of the tank divided by the volumetric flow rate. K is a concentration amount that

represents the magnitude at which mass enters the tank. The inlet concentration is expressed as

a sinusoidal wave to give

dC(t)
K sin(cot) C(t) = r (2.26)
dt









The initial condition for this problem is that the concentration in the tank is zero when

time is equal to zero (t=0). The problem can be scaled where

0 = ct (2.27)

W = O)r (2.28)

and


C= (2.29)
K

The scaled form of (2.26) is written

dC*
sin C = W (2.30)
dO

The differential equation (2.30) can be easily solved using an integrating factor to obtain

C = sin 0 cos0 0
CI + CIe w (2.31)
1 1
W( + 1) W( + 1)
W W

C1 is a constant that can be found based on the scaled initial condition that C*=0 at 0=0.

The final expression for C* can be found after some rearranging. It is


SsinO-WcosO+We w
C* =(2.32)
1+W2

This equation describes the response of the system to the input parameters. First take the

case of a very small frequency, which will cause Wto become very small. If the limit of (2.32)

is taken as Wtends to zero, the result will be that the concentration tends to sin 0, which is the

behavior of the input wave. This makes sense if the inertia is small enough; there will be little

resistance in the tank to change the response in the tank. If Wwere increased, it can be seen

from (2.32) that the response of the tank will behave as a combination of sine and cosine waves.









Now that the inertia has become large, there will be more of a resistance from the tank, causing a

time lag.

Note that the volume of the tank appears in the constant W. When Wis small, the volume

of the tank can be considered small. This means that there is less room for the input signal to

break up into sine and cosine parts, which is the case for small W. For large W, the volume of

the tank can be large, so the signal can become different by the time it permeates throughout the

tank. Being able to examine the output response of a system based on setting the input

parameters is important for understanding the behavior the system. This will be critical when the

oscillating flow in a pipe model is presented in the next chapter. One more example will be

shown to further discuss these concepts.

2.4 Discussion Note 2: Rod Heated Through Oscillatory Temperature Gradient

A rod is heated at one end via an oscillatory heat source and kept constant at the other

end in Figure 2.4. The one-dimensional energy equation is established to be

dT d2T
=TK T (2.33)
dt dx2

Here K represents the thermal diffusivity of the rod material. The problem can be scaled

in a way similar to the previous example where

0 = ot (2.34)

x*= x/L (2.35)

T*= T/K (2.36)

and

L2(o
W (2.37)
IC









Again W can be considered to be the input, although the characteristic length now is the

length of the rod. A similar term will arise in the oscillatory flow model in a tube. The energy

equation can be written in terms of complex exponentials as in (2.10), that is

T= Te' +Te "' (2.38)

Scaling the problem, we get

2 T
-iOWT = (2.39)


The solution for this equation can now be found to be


T = C, cos(\iOW) + C2 sin( iW) (2.40)

The boundary conditions can be found in Figure 2.4; however the condition at the right

hand end of the rod can be rewritten


T (x = 1) (2.41)
2

since

e10 -e0
sin 0 = (2.42)
2i

The time independent part of the temperature field is now

i sin(-\Wx)
T 2 sin((2.43)
2sin( iW)

The complete temperature field can now be found. It is

isin( Wx)
T*= Re[ ei ] (2.44)
sin( 1TW )

After taking the real part of (2.44), and some tedious algebra, the final result of the

temperature field is completed. Hence









1 1
T* = *
[sin(A) cosh(A)]2 + [sinh(A) cos(A)]2
{ [sin(Ax) cosh(Ax) sin(A) cosh(A) +
sinh(Ax) cos(2x) sinh(A) cos(A)] sin 0 + (2.45)
[sin(ix) cosh(ix) sinh(A) cos(A) -
sinh(Ax) cos(2x) sin(A) cosh(A)] cos 0}

where


A = W(2.46)
F2

When Wis taken to be small, it can be seen from (2.45) that cos 0 terms will disappear,

leaving the temperature response in the rod to behave as sin 0, just like the input. Obviously, as

Wis increased, the temperature field will behave as a mixture of sine and cosine terms. Again,

this is similar to the previous discussion note. Since W contains the characteristic length L, a

small value of W can mean a small rod length. The shorter length keeps the input sine wave

from breaking up all the way to the end of the rod. Again, a larger characteristic length will

cause a break up in the response to the inertial force, resulting in a combination of response

terms.

Although these problems appear simple, the length and time scales that are generated in

them can tell you a lot about the physics in the problems. Hopefully these two examples have

clarified the concept of input signal relating to system behavior. It should also be noted that the

oscillating plate problem presented earlier in this chapter behaves in the exact same way. The

Womersley number in the problem is quite similar to Win the CSTR and solid rod examples.

When the Womersley number is very small, the system behaves in the same way as the input

signal, as cos(cot). When the Womersley number grows, the system will behave differently, thus

there is a time lag that arises. When the oscillatory flow model is presented, one may notice













similarities with these examples. There are many ways to impose the oscillations in the model,


and depending on what the input signals are, the output response from the other parameters will


behave in a way that is consistent with these prior examples. When the signal is small, the


system will respond proportionally to the signal. A larger signal (which will occur in the


oscillatory flow model) will cause a mixture of output responses from the system. Now that the


reader has a better understanding of these basic problems, the oscillatory flow model can be


presented.


Infinite Fluid


V = 1/2*A co cos (ct)


Figure 2.1: Design for oscillating flat plate problem


Heat Flux as a Function of Womersley Number for Various Prandtl Numbers
r r i "I 1 I T

- Pr=lO
Pr=500
o Pr->Infinity


/J"
SW-
12 ^"


.sY'
0' ,.~

-7/"

'-


'-


,r


0 10 20 30 40 50 60 70 80 90 100
Wo


Figure 2.2: Heat flux at various Prandtl numbers as a function of Womersley number


100 -


X
U,

IT

S60
'a
r
C
0)


J
















K*sin(wt)


Figure 2.3: CSTR with oscillating input


T=Ksn(wt)


Figure 2.4: Rod with oscillatory heat source on one end











CHAPTER 3
MATHEMATICAL MODELS FOR OSCILLATORY FLOW IN A TUBE

Consider a model of heat transfer enhancement by oscillatory flow where the fluid

resides in a cylindrical tube connected between a heat source and a heat sink. In order to

oscillate the fluid in the tube, one of many methods may be used. The first is to oscillate the

fluid by imposing an oscillatory pressure gradient across the ends of the tube. The second way is

to "push" the fluid back and forth with an external mechanical piston. A third method is to move

the wall of the tube back and forth, whence the shearing of the fluid can also cause the

oscillations. The models for these three methods are first discussed for a circular cylindrical

geometry.

3.1 Oscillatory Flow in a Cylindrical Tube

3.1.1 Pressure Driven Flow

To model this case, write the Navier Stokes equations. Therefore

dV
p(- + V. VV) = ,/V2 VP pg (3.1)
dt

Several assumptions will be made to simplify this momentum balance. Fully developed

flow is assumed in the axial (z) direction only. This means that the pressure gradient is only

assumed to be in the axial direction and gravitational effects are assumed to be negligible. This

reduces (3.1) for a cylindrical geometry, hence

8V 2VF 1 OV OP
p- =(v2 + -)-- (3.2)
at Or r Or dz

This equation can be rewritten as

1 OVZ a 2 1 V
+ v- AP (3.3)
v at r2 r Or









where


1 8P
AP = (3.4)
u dz

To solve for the velocity profile, it is convenient to break the velocity field into a sum of

time complex exponentials from the start, that is


VZ = Vz e't1 + V e'" (3.5)

Likewise, AP becomes


AP = P e"t + P e-t (3.6)

Substituting (3.5) and (3.6) into (3.3) yields


o) 82V 1 8Vz
Vz +- P (3.7)
v r r Or

The common exponential terms are factored out. In addition, since the two terms in (3.5)

and (3.6) are complex conjugates of each other respectively, only one of the two terms needs to

be solved.

In the case of an oscillating pressure drop being an input, the value must be real;

therefore


P=P (3.8)

The solution for the velocity field is


P
V = C1Jo (ar)+C2Y (atr)+ 2 (3.9)


where


a== -(3.10)
V v









When (3.10) is multiplied by the characteristic length, which in this case is the radius of

the tube, one can obtain a dimensionless term known as the Womersley number. This happens

when


aR = -i iWo (3.11)
V v

This group is similar to those developed in the Discussion Notes in Chapter 2. Note here

that the length scale is the radius of the tube, which is different from the case in Discussion Note

2. In that example, the inertial term was scaled with the axial direction, as the heat through

conduction moved down the rod. In this model, the critical length scale is now the tube width.

This is the case because the inertial effect comes from moving the molecules in the radial

direction. The tube width determines the magnitude of radial temperature gradients that can

arise, and these gradients are what enhance the heat transport through convection. In addition to

that, the tube length is assumed to be much greater than the tube width. This means that the

effects of the gradients will be more significant over the tube width compared to the effects of

the axial gradients over the tube length. This is why the radius of the tube is the important length

scale for this problem.

The boundary conditions arise from observing that there is zero velocity at the wall (r =

R) and a finite velocity at the center point of the tube (r = 0). It can be seen in (3.9) that C2 is

equal to zero from the second boundary condition since the Y-Bessel function is unbounded at r

= 0. C1 can be found by applying the condition at R. The expression for the velocity field is now

found to be


V (r) = [1 J(a)12)
a J,(aR) (3.12)










The solution for Vz is simply the complex conjugate ofVz This result is then used to

solve the energy equation, which is

OT T 2T 1 OT
-+ z = K(- ) (3.13)
at rz Or r

Here, K is the thermal diffusivity. The axial temperature gradient (dT/dz) is assumed to

be constant since entrance and exit effects are neglected in this model. The viscous dissipation

term will be neglected as well. The same method can be applied to express the temperature field

as a sum of complex exponentials, hence


T = Te 't +Te'Ct (3.14)

Using the corresponding term in the velocity field, (3.13) is now


iow Vz AlT 2T IT
T+ = ( +--) (3.15)
K K L Or r r

The solution is


T = C3J(r) +CY4C (8r) +5 + C, J (ar) (3.16)

where


/7 = (3.17)
V K

C3 and C4 are constants that can be found from the boundary conditions. C5 and C6 are

constants from the particular solution. Observe that the boundary conditions derive from the fact

that the temperature is finite at the center of the tube (r = 0) and that there is no heat flux through

the walls of the tube (dT/dr = 0). Once the constants are found, the temperature field solution is

now complete. It is










1 Pr1 J (ar) aJ (R)Jo (Fr)
Pi AT Jo (aR) + j (fR)Jo (aR)
T(r) = [ ( (3.18)
a o) L 1-Pr

where


Pr = (3.19)
K

The Prandtl number (Pr) is important because it compares the two time scales used in this

problem, the viscous time scale and the thermal diffusive time scale. The next task is to find the

total heat flux which is equal to the conductive heat flux plus the convective heat flux. This is


J = -k + pCVT (3.20)
L

Here p is the density and Cv is the specific heat. In order to evaluate the total heat

transfer rate, (3.20) must be integrated over the cross sectional area and time averaged over one

cycle to get

27r
SR AT
0Qo = f2(-k L + pC, VT)rdrdt (3.21)


The conductive term is a simple expression and can be easily evaluated. Inserting this

into (3.21) we obtain

2,7
ATk 2 + C9 Or
Qo =k iR2 + pC 2zV, Trdrdt (3.22)
L 2r 0 0

When substituting (3.5) and (3.6) for the velocity and temperature fields respectively, we

get

27T
) R
QcoNV -pC, 2K e'(V + V, e-'")(T et +T e-et)rdrdt (3.23)
2;r 0 0










When the time integral is evaluated, only the "cross terms" will remain since the other

two terms integrate out to zero. Therefore

2,2 2,2

fe2Ztdt = e 2'tdt = 0 (3.24)
0 0

so
SO

R
QCONV = C, 2IK(V T+Vz T)rdr (3.25)
0

Since these terms are complex conjugates of one another, they can be combined and

twice the real part of the remaining term can be taken to get the result for the convective heat

transfer, which is

R
QcN = 2Re[pC, J22r(V, T)rdr] (3.26)
0

Evaluating the integral and combining it with the conductive part results in the final

expression for the total heat transfer, and after some algebraic manipulation we get

2
4P Pr2 ( 'J (aR)J (a'R)Jo (8)
rTOT CONDO L [ Re(I\
QO2R (Pr2 1) ea o3 J0 (R)J0(a'R)J1 (f,) (3.27)

a' J (a'R)
) + 1]
a4J0(a'R)

where


QCOND = -k A R2 (3.28)
L

and


a'=a (3.29)









In order to show the enhancement of heat transfer due to the oscillations, (3.27) can be

expressed as a ratio of convective heat transport to conductive heat transport to yield

S2
QcoNv 4P Pr2 e( 'J1 (aR)J (a'R)Jo (f) a' J( ('R)
QCOND C2R (Pr2_ ) fla3Jo aR)Jo(a'R)JI(f) a4('R) (3.30)

A key thing to note in (3.30) is how the problem scales with the square of the pressure

drop. This is because the velocity and temperature fields scale with the pressure drop. Since the

convective heat transport is based on the product of the velocity and temperature fields, it should

come as no surprise that the convective heat transport is proportional to the square of the driving

force.

Another term in (3.30) that is important is the group where the Prandtl number appears.

This term gets smaller as the Prandtl number grows. This is consistent with what was discussed

in Chapter 2. A large Prandtl number can imply a very viscous fluid. More energy is required to

drive a more viscous fluid; therefore, less heat will be transported convectively if the input

energy is held constant. One may wonder what happens when the Prandtl number is less than 1,

which is usually the case for gases. This term would become negative under those

circumstances. The convective to conductive ratio will still be positive since the Bessel's

functions terms will account for the sign discrepancy. When the thermal diffusivity is greater

than the kinematic viscosity, the real part of the Bessel's functions group will be negative,

keeping the overall ratio positive. All of these relationships will still exist even when the driving

force is altered in the next case, where an oscillating piston is used.

3.1.2 Piston Driven Flow









If a mechanical piston is used to drive the oscillations instead of an imposed pressure

drop, the model does not change very much. Suppose the position of the piston behaves as a sine

wave, that is


x = -Asin t (3.31)
2

Then the speed of the piston would be

dx 1
-- -A cos oit (3.32)
dt 2

The momentum balance of course will not change. However, in order to get (3.12) in

terms of piston amplitude, another relationship needs to be used. Since the integral of the

velocity profile over the cross sectional area is equal to the volumetric flow rate, we get


2Vrdr= 2-Am cost (3.33)
f 2
0

Also (3.33) can be re-written as

R1
S27rVrdr = Ao (3.34)
0 4

where we have used

e"at +e -t
cos ot = (3.35)
2

Using (3.34), the pressure term can now be found in terms of amplitude to yield

S A oa2
P = A (3.36)
[ 8J1 (aR)
RoaJ (aR)

The velocity field is now in terms of the input amplitude as (3.36) is substituted into

(3.12) to give









1 Jo(ar)
Jo (aR)
V. (r)= Ao[ ](a (3.37)
S8J1 (aR)
aRJo (aR)

The rest of the derivation is exactly the same as before with the exception of this one

substitution. The temperature field is now given by



1 -Pr-1 Jo() oQ (aR)Jo (Fr)
(r) AipC AT J (aR) PJi(fR)Jo (aR)38)
L 8J1 (aR)
L (1 Pr 1)(4 ( )
aRJO (aR)

The total heat transfer in terms of input amplitude is

A2 Pr2
QTOT = QCOND f CP2_
RT n (Pr -1)
4J (fiR) 4J(aR) (339)
fRJI (PR) aRJI (aR) +1
4Jo(aR) 8 4Jo(a'R) 8
aRJ (aR) a'R2 a'RJ (a'R) a'2R2

Expressed as a convective to conductive ratio, we get

4Jo (fR) 4J (aR)
QcoNv A Pr f lRJ (R ) aRJ ) (aR)
QCOND R2 (Pr2-1) 4Jo(aR) 8 4Jo(a'R) 8
aRJ (aR) aR2 a'RJ (a'R) a'2 R2

This is the result obtained by Harris and Goren [4] in their model for pulsating flow, the

only difference being the dimensionless Prandtl number for heat transfer replaces the Schmidt

number (Sc) for mass transfer in their model. Again the driving force (amplitude) appears to the

second power in the final heat transfer equation. This occurs for the same reasons as in the









pressure driven flow case. The last method, where the boundary is oscillating, will simplify the

initial equations somewhat, but a similar solution will still be obtained.

3.1.3 Boundary Driven Flow

Suppose now that the oscillations are generated through the outer wall moving back and

forth in the axial direction with a velocity of 1/2*Acocos(ct). The momentum balance will be the

same as (3.3), however there will not be a driving force term from the pressure drop as the

motion is driven by the boundaries. The solution is solved in the same way as before, except

there is only a homogeneous solution, hence

Vz = C Jo (ar) + C2Y0 (ar) (3.41)

The boundary condition at the center of the tube is the same as before, meaning C2 is

equal to zero, and using (3.35), the velocity at the wall is equal to

Vz(r = R)= -Ac (3.42)
4

C1 can now be solved for and the velocity profile is

A coJ0 (ar)
V (r) =4 (3.43)
4J, (aR)

The temperature profile can be solved as before. The boundary conditions remain the

same as in the other cases; this means that we have finite temperature at the center and no heat

flux through the wall. Therefore

AiAT J (ar) J (aR)J (r)(3.44)
T(r) = [ ] (3.44)
4L(1-Pr1) Jo(aR) 8Pj ,(R)J(oR)

The total heat transfer is determined in the same way as before to get









A2, *
QToT = QDIFF f
16R(1- Pr) (345)
a' J (,R)J1 (a'R)- P,1 (fR)J (a'R)
(92 a2 )J1 (R)J, (a'R)Jo (aR)

and the ratio which is

QCONV A2a 'Jo (08R)JI (a'R) PI, (R)Jo (a'R)
QCOND 16R(- Pr-1) ( -2 _,2 )J1(R)J (a'R)J (aR)

Although these three methods of oscillating have a different driving force, the one thing

that appears in all three models is that the overall heat transfer is proportional to the square of the

driving force. For example, if the amplitude is doubled, then the heat transfer will be

quadrupled. This relationship was tested experimentally and will be discussed in Chapter 5.

The relationship of input signal to output response can be seen in each of these cases as

well. It was shown in the Discussion Notes in Chapter 2 that the output response depends on the

size of the input signal and the size of the system. In looking at the final expression for heat

transfer in each of these examples, it is evident that the system response is a combination of

terms (Bessel's functions) unlike the input signals. For example, when the wall oscillates as a

cosine function, the velocity and temperature profiles will behave as a combination of sine and

cosine functions (expressed as Bessel's functions for a cylindrical geometry). Just like in the last

chapter, the break up of the input signal will occur when the Womersley number reaches a

certain limit. The specific value is not relevant here, but it is important to see that the oscillatory

flow model has a simple input signal, and increasing that signal will result in a different output

response, no matter what the means of oscillation are.

The open cylinder is a common configuration for a connecting tube. However, one might

imagine that potential improvements can be made by increasing the number of boundaries, which

would in turn increase the amount of radial temperature gradients, by generating more velocity









spikes. A simple configuration that comes to mind that meets this criterion is that of an annular

tube case to which we now turn.

3.2 Oscillatory Flow in an Annulus

Notice from Figure 3.4 that the inner boundary generates an additional spike when

viewing this cross section. The method for the annular flow model is exactly the same as the

cylindrical flow model with the exception of the inner boundary condition. Although this is the

only difference, this will make the result much more complicated. The three variations presented

above will also be shown here.

3.2.1 Pressure Driven Flow

The solution for the momentum equation is identical to (3.9). However, with the new

boundary condition at the inner wall, C2 is no longer zero. The boundary condition at the inner

wall (r = R) and the outer wall (r = cR, c>l) are now the same, that is the velocity equals zero.

The solved velocity field is now


(r J0 (ar)[YJ (acR)- Yo (aR)]
a Jo (aR)Y (acR) J (acR)Y (aR) (3.47)
Y (ar)[J (aR)- J (acR)]
J, (aR)Y (acR) J0 (acR)Y, (aK)

As in the case of the velocity field, the inner and outer wall boundary conditions on

temperature are the same, meaning there is no heat flux through either wall (therefore dT/dr = 0

at these walls). The solution for the temperature field is similar to (3.16), with an additional

particular solution that is proportional to the Y-Bessel's function. Therefore

T = C, Jo (f) +C2Y (fr) + C3 + CJ (ar) + CY (ar) (3.48)

C3, C4, and C5 are found from solving for the particular solution, while C1 and C2 are

determined from the boundary conditions all yielding









PAT
C
aPL(f32 -
[J, (acR)Y, (fR)
[J, (PcR)Y, (fpR)
[J (aR)
[Jo (aR)Yo (acR)
[Y1 (acR)Y1 (fpR)
[J, (fpcR)Y, (fPR)


[Yo (acR)
a2 [J (caR)Yo (acR)-
- J, (aR)Y, (tcR)]
- J, (fR)Y, (ticR)]
-J (acR)]
- J (acR)Yo (aR)]
- Y, (aR)Y, (tcR)]
- J, (fR)Y, (PcR)]


- Y (aR)]
Jo (acR)Yo (aR)]


(3.49)


PAT [Y (acR)
2 ,KL(2 2) [Jo (aR)Yo (acR)

[J, (ipcR)J, (aR) J, (fR)J, (acR)]
[J, (fcR)Y, (fR) J, (fR)Y, (PcR)]
[Jo (aR) Jo (acR)]
[J0 (aR)Yo (acR) Jo (acR)Yo (aR)]
[J, (PcR)Y, (aR) J, (iR)Y, (acR)]
[J, (PcR)Y, (fR) J, (fR)Y, (PcR)]


PAT
C3 a2
3 2p2


-PAT [Y (acR)
C4 aL2 a) [ (a)
4 2 (2 2) [J0 (aR)Y (acR)-


- J (acR)Yo (aR)]


(3.50)


-PAT [J (aR)
C5 a2 (2 a) [j( )(a
a2"C( 2 2 -_ ) [J0 (aR)Yo (acR)-


Jo(acR)]
-J (acR)Y (aR)]


The expression for the total heat transfer is equal to

AT 2R ~
Qror = -k A (c2 -1)+2Re 27Vz Trdr
L R


(3.53)


(3.54)


V is the conjugate of V Since the expression is now quite complicated, MAPLE is

used to further evaluate (3.54). The next case for annular flow, the oscillating piston problem, is


-Y (aR)]
Jo (acR)Yo (aR)]


(3.51)



(3.52)









quite similar to the oscillating pressure drop case. The same method to relate the pressure drop

model to the piston model for a cylinder will be used now, with a change in the integration limits

to account for the geometry difference.

3.2.2 Piston Driven Flow

The solution for the piston driven problem is the same as the pressure driven problem

with the exception of one substitution. By inserting (3.47) into (3.34) and solving for the scaled

pressure, we get

SAco)cR2 (c2 -1)
P = (3.55)
8Z

where

R 2 (c2 -1) R[cJ1((acR)- J,(aR)][Yo(acR)- Y(aR)]
Z=
2 a[Jo (aR)Y, (acR) Jo (acR)Y, (aR)] (3.56)
R[cY, (acR) Y, (aR)][J0 (aR) J0 (acR)]
a[J0 (aR)Yo (acR) Jo (acR)Y, (aR)]

(3.47) through (3.53) can now be found in terms of piston amplitude, and the final result

can be determined using (3.54).

3.2.3 Boundary Driven Flow

For the problem where the outer wall is oscillating, the boundary condition for the

momentum equation will be the same as (3.42). The velocity will be zero at the inner wall.

Using these to solve for the velocity profile, we get

S Am [Jo (aR)Y(ar)- Y (aR)J (ar)]
V= (r) = (3.57)
4 [YO (acR)Jo (aR) Y (aR)Jo (acR)]

There is no heat flux through either wall, so the temperature solution is


T(r) = CI r) J C2Y C () + CJ,0 (oa) + C4Y (or) (3.58)

where









a C[J1 (aR)Y, (icR) J (acR)Y, (/fR)
c, -
P [J, (/3cR)Y, (/R) J, (/R)Y, (3cR)]
(3.59)
C4 [1 (aR)Y (PcR) Y, (aRo)r, (PcR)]
[J, (3cR)Y, (fR) J, (R)Y, (PcR)]

C, p (J R) aC4Y (R) aC J1 ) (3.60)
pY, (R)

AA Yo (aR) (3.61)
C= (3.61)
4L(1- Pr 1) Yo (acR)Jo (aR) Yo (aR)J, (acR)

and

AATi Jo (aR)
C4 = (3.62)
4L(1- Pr 1) Yo (acR)Jo (aR) Yo (aR)Jo (acR)

The overall heat transfer can be found by solving equation (3.54). In summary, the input

parameters for this model are the amplitude and frequency of the driving force (either the

pressure drop, the piston, or moving wall), the length and radius of the connecting tube, the fluid

used, the geometry of the tube (cylindrical or annular), and the axial temperature drop. Once all

these parameters are inserted into the appropriate model, the conductive, convective, and total

heat transport can be calculated. Now that the cylindrical and annular models have been

presented, it is important to determine which geometry is more optimal, and under what

conditions.

3.3 Comparison of Annular and Cylindrical Geometries

Four cases are discussed to determine whether the cylindrical or annular geometry is

preferable for enhanced heat transfer. In the first case, the heat transport in an open tube is

compared to one in an annular geometry whose outer radius is held equal to the radius of the

open tube. As the inner rod's radius is decreased, one expectedly recovers the results for an open

tube. The comparison between the two geometries is made holding the pressure drop per unit









length fixed. In a companion calculation, the volumetric flow rate in a half cycle is held fixed

instead of the pressure drop. While an idea regarding the effect of the walls can be obtained, the

comparison between the two geometries is not entirely fair because the cross-sectional areas of

the two geometries are not kept equal. Thus two other calculations that mirror the first two

cases are examined with the only difference being that the cross-sectional area of the cylinder

and annulus are now held fixed. The results are telling.

The Prandtl number used in the calculations corresponds to water and is set to a value of

7. The physical properties of water were used when evaluating the density, viscosity, thermal

conductivity, and the specific heat. The other parameter values held constant in all the

calculations are the following: tube length of 10 cm, a frequency of 0.1 Hz, and an axial

temperature difference of 10 deg C. These values are of little concern as the qualitative nature is

what is important.

Figure 3.5 depicts the ratio of the transport in the annulus to the transport in the cylinder

while the inner radius of the annulus is varied. A value of 1 cm for the outer radius was used in

the first two sets of calculations. It is observed that as the inner radius is decreased in the

annulus, the values of the heat transport come close to the open tube result. One is led to the

conclusion that the open tube cylindrical geometry is a more optimum design compared to an

annulus. The reason for this conclusion is that the available region for heat transfer decreases as

the inner radius increases. This means that for a fixed pressure drop, as the size of the inner rod

increases, the resistance to flow increases, and the sharpness in velocity spikes decrease, leading

to a lower heat transport. The transport decreases from the open tube even when the radius of

the inner rod is small. This is expected as the resistance to flow increases dramatically when a

thin rod is introduced leading to a decrease in heat transfer. Likewise, as the inner rod diameter









gets larger, the annular geometry asymptotically becomes a thin gap, and the heat transport

slowly approaches zero.

In the first case, the pressure drop was held fixed, but if the flow displacement in one

half-cycle was held fixed, and the comparison was made between the two geometries, something

different may occur and that is indeed the case. As there is now a smaller cross sectional area in

the annular region than the open cylinder, the pressure drop will have to increase in the annulus

so that the flow rates can be the same in both cases. The results of the calculations are depicted

in Figure 3.6.

The calculations depicted in the figure now show that the heat transport is enhanced with

an annular geometry compared to an open tube provided that the flow rate per half-cycle is

assumed equal in both geometries. The flow rate was calculated by integrating the velocity over

the flow area for one half-cycle to give


SocR
Flowrate = f f 2zV rdrdt (3.63)
OR

The pressure drop (which is contained in the solution for the velocity field) in the annulus

was then changed so that the flow rates would be equal. On calculating the pressure drop

increase in the annulus as the inner radius increases, it is evident that this is a reason why the

transport increases in the annular case. Another contributing factor to the increase in the

pressure drop is due to the resistance of the inner rod. The wall friction can lower the flow rate,

so the pressure drop must increase to overcome that resistance and make the flow rate the same

as the cylinder. For the cases where the inner radius of the annulus is high, the transport in the

annulus becomes roughly 15 times higher than the corresponding cylindrical case. The pressure

drop is related to the amplitude of the spike, meaning that a higher pressure drop in the tube will









result in a greater amplitude of oscillation, a higher velocity spike, and hence, greater transport.

This is why the annulus is the more optimal method in the case of fixed flow rates.

In the first two cases, the area changes as the radius of the inner rod was changed. This

raises the question as to whether the comparison between a cylindrical and an annular geometry

is a fair one. To make a fair comparison, consider the case where the cross-sectional areas and

the pressure drop are held equal for both annular and open tubes. This means that as the inner

radius of the annular geometry changes, the cross-sectional area of the companion open tube

must also change so that both configurations will have the same amount of area available for heat

transport. The results for this comparison are graphically depicted in Figure 3.7 for a definite

frequency. The calculations show that the heat transport is again greater in the open cylinder,

just like the first case. It appears to have an advantage over the annular case by 1-2 orders of

magnitude for smaller cross-sectional areas. Again, if the flow rate is held fixed, for the same

cross-sectional area, we see that the heat transfer favors the annular configuration.

The issue with the case of fixed flow rates, when the areas were not held equal to each

other, was that the lower cross-sectional area in the annulus drives up the pressure drop needed

in order to have the flow the same as the cylindrical case, and consequently leads to larger spikes

in the flow profile. However, in the last set of calculations the cross-sectional areas are the same

for the open tube and annulus, and therefore this factor is now not an issue. The question then is:

why does the annular geometry still give a higher transport? The reason is that in the last case,

as before, the pressure drop increases because the inner rod offers resistance on account of wall

friction. This factor continues to remain important even where the cross-sectional area is fixed

and made equal. One will notice in looking at Figure 3.8, that the ratio at high inner radii is still

large (about 8 times), although not as large as before (15 times). This makes physical sense










since the cross-sectional area discrepancy is now eliminated in Figure 3.8, leaving only the inner

rod inhibitor to play a role in driving up the pressure drop.

From the calculations it appears that the open tube geometry is best when the pressure

drop is held fixed, but the annular geometry is better when the volumetric flow rate is held fixed.

The ultimate message is that the optimal case depends on what one holds fixed, pressure drop or

flow rate.

Another factor in determining which geometry is best is to see what system gives a

higher heat transport output per unit of mechanical power input. The power only takes into

account the amount needed to cause the oscillatory fluid movement. The power needed to move

an oscillating piston of a given mass for example is not incorporated as this is subject to the

materials used. The mechanical power is calculated using the following relationship, which is

2ff
) cR
Power =- 27iV, (pressure)rdrdt (3.64)
2 0 R

For the case of a cylinder, the lower limit of integration in r will be zero. In order to

obtain the pressure in terms of P from (3.64), it must be converted to units of pressure. Therefore

27f
) cR
Power = Z- f 2V PupLrdrdt (3.65)
20 R

(3.5) and (3.6) can be inserted into (3.65) giving

27;
) cR
Power = 27(Vz e' + V zee -'")(Pet +Pe P -t)jLrdrdt (3.66)
27 0 R

The integration with respect to time can be easily done (see (3.24)) leaving a final

expression for power which is









cR
Power = 2Re f 2trV P pLrdr (3.67)
R

To demonstrate a comparison we will first take the case of fixed cross sectional area.

The case where the pressure drops are held fixed produce Figure 3.9. There is no need to

consider the case of fixed flow rates because the results are identical to Figure 3.9. These sets of

calculations show that the ratio of transport/input power is higher for the open tube case. Again,

the annular case results approach the cylindrical values when the inner rod gets small. The

reasoning for the input power being greater for the annular case is because there are two

boundaries in the annulus compared to just one for the open tube. The extra boundary means

there is more viscous effects. This means more energy is required to oscillate the fluid since it

now must overcome the viscous effects at two walls as opposed to just one. This must be

considered when determining the geometry for the most efficient heat transport. If the amount of

input power is a concern, then that needs to be factored into the decision.

In summary, four methods have been worked out to compare the heat transport in an open

cylinder versus an annulus. Although the results can differ depending on the method of choice,

the results of each method make physical sense. It can be concluded from this study that the

annular model would be optimal if the system is able to handle a large pressure drop, and if input

power is not a concern. If not, then the open cylinder would be the geometry of choice.

3.4 Comparison of Two Compartment Flow versus One Compartment Flow

3.4.1 Heat Transfer

It is possible to make another comparison involving cylindrical and annular geometries.

Ranger [12] proposed a model for steady Poiseuille flow through concentric pipes with various

areas of cross sections. He determined that it was possible to have two different geometric









models which have the same cross section and the same flow rate, although each model contains

a different number of "compartments." For example, the one compartment model consists of a

large annulus, while the two compartment model has an annulus and a cylinder. Note that the

outer radii of both models (R4) are set equal to each other. In order to make the cross sectional

area of both models equal, a constraint needs to be used. It is

R2 =R2-R (3.68)

In the case of the one compartment model, the flow rate is

)iA-P (R2 R2 + R 2)
Q' = [R4 -R3 +Rf ][R2 +R -R ( -4 +R ] (3.69)
8U InnR4 I- ln(R2 R2)
2

Note that (3.68) was substituted for R where appropriate. For the two compartment

model, the flow rate is

P (R2 R2 )2
Q! = [R4 + R4 R4 (R4 R ] (3.70)
8U Iln(R4/ R3)

The derivation for (3.68) through (3.70) was done by Ranger for Poiseulle flow and can

be found in [12]. A ratio can now be established relating the two flow rates as

S=- (3.71)
Q

When S=1, the flow rates become equal. This means that for a given value of R4, R3 and

R2 can be adjusted so that the flow rates are equal. Pressure drop and viscosity will not play a

role since they factor out of (3.71). There exists a value of R2 where S=1 for a set R3 value

(again assuming R4 is held fixed). In order to examine the behavior of the one compartment

system versus the two compartment system, certain relationships need to be developed for use in

the oscillatory flow model. For this study, assume that the piston driven model is used in both









the one and two compartment models. It is important for this problem that the pistons are

external to the channels. See Figure 3.11 for a visual depiction.

The shaded areas represent the flow regions. The external pistons in both cases are

assumed to oscillate with the exact same amplitude and frequency, and they both have the same

cross sectional area since the outer radius (R4) of the one and two compartment models are

assumed to be the same. This means that the volumetric flow rate going into the one

compartment system will be equal to that going into the two compartment system. The only

issue that now arises is the "break-up" of the flow into each section of the two compartment

model. Since the models were derived earlier for the cylinder and the annulus, the amplitude

needs to be known for each case so that the transport can be found in both sections separately.

The transport in each section is simply added together to obtain the total transport for the two

compartment model.

The oscillation amplitude and frequency of the external piston is assumed to be set. The

volume displacement equation is used to obtain the oscillation amplitude in the annular channel

of the one compartment model. Thus

AexR = A,,(R2 -R2) (3.72)

Ae is the amplitude of the external piston and Ala is the amplitude inside the annulus (the

7t on each side of the equation cancel). Ala can be solved for since everything else in (3.72) is

known. The transport in the annulus for the one compartment model can now be obtained by

using the equations for piston driven flow in an annulus described earlier. In order to obtain the

amplitude in the cylinder and annulus for the two compartment system, two relationships are

needed to solve for the two unknown amplitudes. The first says that the total flow rate is equal

to the sum of the flow rates in the cylinder and the annulus. Therefore









AexR2 = A (R2 R ) + A2R2 (3.73)

A2, and A2a are the oscillation amplitudes inside the cylinder and annulus of the two

compartment model respectively. Since the volumetric flow rates are being used in this

equation, note that the oscillation frequency and the constant t factor out of each term in (3.73).

The second relationship involves the resulting pressure drops generated by the moving piston in

each compartment. Since the piston moves externally, the resulting pressure drop in each

compartment must be the same. This makes physical sense since there will be only one pressure

drop from the beginning of the pipe to the end of the pipe. This means that the right hand sides

of (3.36) and (3.55) can be set equal to one another, yielding

A2(R- R) A2c(3.74)
-=4 3 (3.74)
8Z [ 8J1 (R2)
[4 ]
R,2J0 (aR2)

The expression for Z can be found in (3.56) where the outer radius is equal to R4 and the

inner radius is equal to R3 instead of cR and R, respectively. There are now two equations for

these two unknown amplitudes (A2a and A2c), meaning they can now be solved. Comparisons

were made at various frequencies for several different values of a parameter 6. This is the width

of the annular solid (no flow) region in the two compartment model (equal to R3 minus R2).

Since the cross sectional areas are the same for each model for the flow regions, the regions

where there is no flow is also identical between the one and two compartment systems, because

R4 is the same for both cases. This means a lower 6 corresponds to more available room for

flow. The problem was solved using a numerical solver, with water as the fluid medium, a pipe

length of 50 cm, a temperature difference of 10 C, an outer radius (R4) of 1 cm, and an external

piston length of 8 cm. The results are given as a ratio of the heat transfer in the one compartment









model to the heat transfer in the two compartment model for various sizes of 6 (0.1, 0.3, and 0.5

cm).

In all three instances, the heat transport in the two compartment model is greater than the

one compartment model. This makes sense by understanding what is going on with the physics

of the problem. In the two compartment model, there are a total of three boundaries in the cross

section (the outer edge of the annulus, the inner edge of the annulus, and the outer edge of the

cylinder). There are only two boundaries in the one compartment model (the outer and inner

edge of the annulus). Keep in mind that the total flow rates must be the same through each

model. Since there is no flow at the boundaries, and very little flow right next to the boundaries

(the parabolic spike in the velocity profile approaches zero near the walls), there will be an extra

region of very little flow in the two compartment model since it has one more boundary. This

will cause a local decrease in the flow rate in these regions. To account for the extra region of

slow flow, the amplitude of the spike must grow proportionately, so as to keep the overall flow

rates the same for each model. It has been shown earlier in this chapter that increasing the

amplitude in the flow profile drives up the heat transport by the square of the amplitude increase.

This is why the two compartment model is more favorable, since the extra boundary will produce

larger spikes in the flow profile.

The trend for the curves in Figures 3.12, 3.13, and 3.14 is for the one compartment model

to approach the transport of the two compartment model, then tail off as frequency increases.

This is due to the sharp spike forming in the wide annular region of the one compartment model

as frequency grows, then the profile begins to flatten and the transport begins to plateau off as

the frequency becomes large. The cylindrical region will maintain a sharper velocity profile for









a longer period of time since there is no inhibiting rod in the middle. This will cause the ratio to

decrease at high frequency.

As the 6 value increases, it can be seen that the maximum ratio approaches 1, which is

when the transport in both models are equal, but it never reaches 1. Note that the maximum

points on the curves seem to approach 1 asymptotically, so when 6 becomes 1 cm, the models

will have the same transport. This makes sense because if the solid (no flow) region

encompasses the entire area enclosed by the outer wall, both models will have the same amount

of heat transport going through it, which is zero.

A comparison has been made between two geometries to determine which transports heat

better through it, which in this case is the two compartment model. These models can also be

studied to figure out which provides a better separation between two species.

3.4.2 Mass Transfer and Separation of Species

The same models can be compared to determine which transports mass more efficiently.

Since the mathematics of the models for heat and mass transport are virtually the same, the

obvious conclusion can be reached that the two compartment model is better since it transports

heat at a higher rate, which is indeed true. Let us take this one step farther to see how effectively

each model can separate out a mixture of two species in a carrier fluid.

This problem was first presented in the Chapter 1. For this study, helium and carbon

dioxide will be separated in a nitrogen carrier. Starting from rest, the helium molecules will

separate out faster than the carbon dioxide since helium has a larger diffusion coefficient. When

the oscillation frequency becomes high enough, the carbon dioxide will emerge first as the small

helium molecules get trapped near the boundaries in the slow flow regions. Eventually, there

will be an optimal frequency for which the separation ratio is at a maximum. This is consistent









with the results of Kurzweg and Jaeger [8]. The same parameters were used for this calculation

as in the heat transfer calculation discussed above, with the exception of the fluid medium. The

results are again presented for different values of 6. The separation ratio in the figures represents

the mass flow rate of carbon dioxide divided by the mass flow rate of helium. Note in Figure

3.16 that the dotted line at the end of the curve represents an estimate due to inaccuracy of the

machine performing the calculation. Since this region occurs well after the maximum is

achieved, it is not significant.

The maximum separation ratio that is achieved in the one and two compartment models is

about the same in all three cases. It can also be seen how the size of the peaks grow when 6

increases, from about 2.5 for small 6, to about 3.8 for larger 6. This is because the amount of

solid (no flow) space grows with 6. Since the external piston amplitude remains the same, the

amplitudes of the spikes in the tubes must grow to satisfy (3.72) and (3.73).

Even though the maximum ratios are similar in each case, one big difference is at what

frequency the maximums occur. In each case, the highest ratio occurs at a much lower frequency

in the two compartment model compared to the one compartment model. Roughly 3 to 4 times

the optimal frequency of the two compartment system is needed to achieve the optimal frequency

for the one compartment system. The reasoning for this is very similar to the heat transfer

analog. Since there are more boundaries in the two compartment model, there are more slow

flow regions. This means that the helium molecules can move quickly into the slow part of the

flow near the walls, since there is more available space. The separation will occur faster since the

helium molecules become quickly trapped, which means the larger carbon dioxide molecules

will have more room in the faster moving core region, and therefore transport quickly down the

tube.









Achieving the separation at a lower frequency is beneficial since it limits the amount of

power needed to drive the system. The input power needed to oscillate the fluid increases with

increasing frequency. Therefore the same separation can be produced with either model, but

much less energy is being used in the two compartment system. The conclusion can be reached

that the two compartment model is the more optimal geometry, whether heat or mass is being

transported, or if a separation is desired. One last consideration will be made in the next chapter

regarding the geometry of the system, that being what happens when the inner rod of an annulus

is moved to an off centered position.


Pressure Drop


Hot I Cold






Figure 3.1: Oscillations driven by pressure drop




Hot Cold






Figure 3.2: Oscillations driven by a piston
















Hot ) Cold


Figure 3.3: Oscillations driven by moving boundary


Hot Cold


Figure 3.4: Oscillations in an annular geometry






0.9

0.8 -

0.7

S0.6

S0.5

S0.4



0.2

0 .1 "

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Inner Radius of Annulus


Figure 3.5: Heat transport in annular versus cylindrical geometry with the outer radius and
pressure drop held fixed


















10






0-






0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Inner Radius of Annulus


Figure 3.6: Heat transport in annular versus cylindrical geometry with the outer radius and flow
rate held fixed


U
O 0.05 01 0.15 02 0.25 03 0.35 0.4 0.45 0.5
Inner Radius of Annulus


Figure 3.7: Heat transport in annular versus cylindrical geometry with the cross sectional area
and pressure drop held fixed



















/ -




4 4


1
-0




1

0 0.05 0.1 0.15 02 0.25 0.3 0.35 0.4 0.45 0.5
Inner Radius of Annulus



Figure 3.8: Heat transport in annular versus cylindrical geometry with the cross sectional area
and flow rate held fixed


0.9



0.7
c





06
0


0.4

0.3

0.2

o0.1

0 L L
0 0.05 0.1 0.15 0.2 0.25 0.3
Inner Radius of Annulus


0.35 0.4 0.45 0.5


Figure 3.9: Output/Input Ratio with Fixed Pressure Drop and Cross Sectional Area











Figure 3.10: Diagrams for one and two compartment models respectively (flow regions shaded)

I_



Figure 3.11: External piston oscillating one compartment model (top) and two compartment
model (bottom).













1

0.9


0.8 -
E

S0.7 -
E

0
C 0.6



r 0.4


0.3


0.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 18 2
Frequency (Hz)

Figure 3.12: Heat transport in the one vs. two compartment model at various frequencies at
small 6




0.9

0.8 -

U 0.7

E
S0.6

0.5 -

0. 04

0.3

0.2

0 .1 1 1 1 1 1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Frequency (Hz)

Figure 3.13: Heat transport in the one vs. two compartment model at various frequencies at
medium 6

















0.9

0.8
E
0.7

E
o 0.6

r
| 0.6

0.4


0.3

0.2

0.1
0


Figure 3.
large 6


3


0 2 4 6 8 10 12
Frequency (Hz)

Figure 3.15: Separation ratio for various frequencies for the one compartment model (dashed
line) and two compartment model (solid line) at small 6


0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Frequency (Hz)

14: Heat transport in the one vs. two compartment model at various frequencies at


CD
S2
rN
O
o
1.5
o7
C5
(0
- 1
U3


0.5































0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Figure 3.16: Separation ratio for various frequencies for the one compartment model (dashed
line) and two compartment model (solid line) at medium 6


0 5 10 15
Frequency (Hz)


20 25 30


Figure 3.17: Separation ratio for various frequencies for the one compartment model (dashed
line) and two compartment model (solid line) at large 6











CHAPTER 4
OSCILLATORY FLOW IN AN OFF CENTERED ANNULUS

In the previous chapter we showed that oscillatory flow in an annular geometry leads to

an enhancement of transport over an open geometry under certain conditions. Another case to

examine is when the inner cylinder in the annular geometry is moved to a slightly off-centered

position (see Figure 4.1). A model of this problem will first be presented, followed by a

calculation to see what option is better, and under what circumstances.

4.1 Modeling an Off Centered Annulus

Let the deviation in distance of the inner rod from the center be small and be denoted as

e. To get the relation between the radial distance in the distorted geometry with the radius in the

reference geometry, it is seen that

[X -]2 +2 =R2 (4.1)

Converting this into polar coordinates we get

R2 -2Rcos9 = R -2 (4.2)

R can be expressed in terms of F to get the mapping from the reference to the current

configurations. To get a better understanding of the mapping equations, see Johns and

Narayanan [6]. Thus


R(O, e) = Ro + ER, (O)+-2R2 (0) (4.3)
2

Here R1 and R2 are determined from their definitions yielding

dR
R, = ( = 0) = cos 0 (4.4)
and
and










R= 0) sin2 0 (4.5)
de2 Ro

These results will be useful in obtaining the boundary conditions at the displaced inner

surface. For the distorted geometry, let the driving force come from the inner and outer walls

oscillating in the axial direction, both moving in phase, and whose speed is expressed as

1/2*Acocos(ot). Using the mapping of the inner boundary, the velocity and temperature profiles

can be obtained at the various orders. The expression for the lowest order (i.e., e (0)) is obtained

in a manner as done earlier. Since the zeroeth order problem is for the case of the centered inner

rod, this simply becomes the boundary driven problem in an annulus shown in Chapter 3, with

the exception that the boundary condition at the inner wall is now equal to the condition at the

outer wall. The velocity field is now (the solved constants are listed in the Appendix)


Vzo = AoVYo (ar) + BoJo (ar) (4.6)

Note that the same method is used here to express velocity as a sum of time complex

exponentials. Solving the first order problem will be different from the base case, since velocity

will become a function of 0, in addition to the radial position. The first order equation is


d 2 Vz 1 dVz1 2 1 d2 Vzi
+-- +a2 VVzi+- -- =0 (4.7)
dr2 ro dro ro2 d ()

At first order, the boundary condition at the outer wall yields the first order velocity field

to be zero. This is true because there is no displacement at the outer surface, so even though the

outer wall is moving at fixed amplitude, there are no first order F terms, implying that the

velocity is zero at the outer boundary to first order or any order higher than zero. For the inner

boundary, the surface is displaced, so a mapping must be used yielding









dV 1
Vz = Vzo(Ro,O0)+ [Vz (Ro,00)+R1(00) dz0 (Ro)]+ -2[VZ2(Ro,00)
dro 2
(4.8)
dV d 2 dV
+2R,(00) z' (Ro,00)+R,2 0) (RO)+R2 (0 (RO)]
dro dro dro

In order to obtain the boundary conditions at the inner wall for each order greater than

zero, set the terms of the desired order in (4.8) equal to zero and obtain a condition for the

velocity at the wall for that particular order. Note that these boundary conditions stay the same

when the velocity is expressed as a sum of exponentials since there are no time dependent terms

in the boundary conditions.

At first order, the inner wall boundary condition is


dVzi dVzo
Vz (Ro,00)= -R, = -cos0 (4.9)
dro dro
R0 Ro

The 0 dependence is seen to be proportional to cos 0 in order to satisfy differential

equation (4.7) and the inner wall boundary condition (4.9). The solution would then be


Vz = (ArY (ar,)+B J1 (a0r))cos (4.10)

The differential equation for 0(e2) is the same as (4.7) with the exception of Vz2 replacing

Vz1 in the notation. The second order boundary condition is


dVzl d2 Vzo sin2 0 dVzo
Vz2(R, 0)= -2cos 0 r -cos2 0 n2 V(4.11)
drY dro R, dry
R0 Ro Ro

The complete result for the second order velocity field is


Vz2 = [A2v Y(ar) + B2J, (ar)] + [CY0 (ar)+ DJo (ar)] cos 20 (4.12)

This form of the solution makes sense since all three terms in (4.11) can be broken up

into 0 dependent and 0 independent terms through the use of the double angle formulae. For this









problem, only the 0 independent terms are taken into account. The reasoning for this will be

explained later. This completes the calculation of the velocity field in the problem. These

expressions are used to solve the temperature field equations. The method for solving the

temperature field equations is the same as before. Each order in e of the temperature profile is

solved using the corresponding velocity profile (i.e., the first order temperature field uses the

first order velocity solution). The temperature is broken up into a sum of two exponential terms

just like the velocity field case. The boundary conditions for the base case is that there is no flux

through the walls (dT/dr=O). For the base case, the temperature field is identical to (3.56).

Therefore

To = Ar YO (fr) + Br OJ (fr) + yY,0 (ar) + J0J,0 (ar) (4.13)

The first order problem for the temperature field has the same form as (3.15) plus a 0

dependent term. It is

d2 T 1 dT1 1 d2T1 AT .
S+---+ Ti+ = -Vzi (4.14)
dr ro dro r2 dO LK

In a manner used for treating the boundary conditions for the velocity field, we have the

temperature field mapping for the inner rod, which is

dTT d ,O dT, d2T
S= (RO,00)+E [ 1 (Ro,00)+R1 (00) (Ro)]+
dro dro dro dr(
1 dT, d2T
-E [ (Ro,00)+2R, (00) 2 (RO,00)+ (4.15)
2 dro dro

R2(0) dajr (Ro)+ R (0)-) (R)]+...
dro dro









The flux is equal to zero at the outer wall and is zero for all subsequent orders. The

solution to this first order problem has a homogenous and particular solution multiplied by a cos

0 term, just like the velocity field. It is


T1 = [AI ,Y (fr) + B, J1 (fr) + YY + (ar) + 1, J1 (ar)] cos (4.16)

As in the case of the velocity field, the boundary condition for the temperature field at the

inner wall is found by taking terms of the appropriate a order in (4.15), and solving for the

respective order.

For the second order problem, the differential equation for temperature is the same form

as (4.14) with T2 and Vz2 replacing T1 and Vz1, respectively. As in the second order velocity

problem, the solution is broken up into two parts to satisfy the 0 dependent and 0 independent

boundary condition, and again, only the 0 independent part of the solution will be useful. We get


T2 = A2T Y (fr) + B2T J (GF) + 2,Y0 (ar) + 2JJ0 (ar) (4.17)

The cosine and sine functions can be broken up as in the second order velocity problem

and only the 0 independent terms are needed.

Now that the velocity and temperature equations have been determined up to second

order in e, the heat transport can now be obtained. The expression for the transport is the same

as before, which is

2HKRo
Q = 2pC, Re J J Vz TrdrdO (4.18)
OR

where


Vz = conjugate(Vz ) (4.19)









Note that the time dependence has already been factored out leaving only the cross terms

(cf. Chapter 3). In the case of a perturbed surface, the transport rate Q will have to be expressed

as a perturbation series as well, so


Q = Qo +Q, +-2Q2 +...... (4.20)
2

where

dQ
Q, =- (4.21)
de

and

d2Q
Q2 (4.22)
de2

Since Q is expressed as an integral, the derivative of Q can be determined using

Leibnitz's rule for differentiating integrals. Note that the velocity and temperature terms can be

expressed as derivatives, which will simplify the algebra somewhat. Therefore


Vz = Vo +EZ + Z2+.. (4.23)
2

where

dV
Vz (4.24)
de

and

dz2V
VZ2 -- (4.25)
dc2

The same expansion can be done for the temperature. Using Leibnitz's rule to find Qi, it

is found that


Q, = PC, Vz T0+To Vz i]rdrdO- cosO( Vzo To)RdO (4.26)









Looking at (4.26), it can be concluded that Qi is equal to zero. Both terms in the double

integral contain a cos 0 term from the first order expression and integration over 271 makes the

entire integral equal to zero. The same thing occurs in the single integral with the cos 0 term,

making that integral equal to zero as well.

The expression for Q2 can be found, and after some cancellations, the result is


Q2 = C,, [fV z T2+2Vzi T+Vz2 To]rdrd +
(4.27)
f RoR,2 To (Ro) (Ro)dO
dro

From this expression it can be understood why the 0 dependent terms were neglected in

the second order velocity and temperature expressions. If the expressions for the perturbed

velocity and temperature field were inserted into (4.27) including all the 0 dependent parts,

several terms would result that either contained a cos 0, or a cos 20 term. Since the integration is

carried out over 27n radians in the 0 direction, all of these terms will integrate out to zero. This is

why the 0 dependent terms were discarded in the second order perturbation, since they would

just contribute to the terms that would eventually integrate out to zero. A final expression can

now be obtained for Q2 after further evaluating the term in the single integral. It is

.. .. .. To (Ro) dVzo

Ro
Q2 = 4cpC, Re{J [Vo Tz+Vz T+Vz2 To]rdr +(O(RO)V )}(42
o2 dr0 (4.28)


Since Qi is equal to zero, (4.28) must be added to Qo to get the final equation for the total

heat transport in the off-centered annular geometry to give










Q = 4rpC, Re{ J [Vz To ]rdr +
Ro


2 [VTz+VlTl+VzTr. To(Ro) dVzo
-E2[ [VzoT2+VziT+VZ2T]rdr+ o (R0)]} (4.29)
2 2 dr
Ro 0

4.2 Analysis of Off Centered Model

One of the important results that need to be obtained from these equations is a solution

for Q2 at various frequencies. This will determine if the off centered annulus is a beneficial

geometry, or a detrimental one, since Q2 is equal to the change in transport due to the

perturbation. The results for a hypothetical fluid with a Prandtl number of 10, a kinematic

viscosity of 0.1 cm2/s, and amplitude of 10 cm are presented in Figure 4.2. Like the comparison

between the cylinder and the annulus presented in Chapter 3, the qualitative nature is what is

important here. MAPLE was used to solve (4.29) at various frequencies. It is seen here that the

change in transport is positive for small frequencies, and quickly becomes negative for higher

frequencies. Although this has been shown from a mathematical standpoint, some explanation

needs to be given why this makes physical sense.

When the annulus is positioned in the center, the only gradients that exist during the

oscillations are in the radial direction. If the annulus were moved to an off centered position,

there will be gradients forming in the 0 direction in addition to the radial direction. In order to

visualize this better, imagine that the off-centered position of the annulus creates a "thin" region

and a "thick" region (see Figure 4.3). The thin region consists of radial lengths less than R0, and

the thick region contains the area of fluid that has radial lengths greater than Ro. The total cross

sectional area of the thick region turns out to be greater than the thin region. As stated before, an

azimuthal gradient is formed. The heat will then flow from the thin region to the thick region.









Since the thick region is the freer flowing region, it will be easier for the heat to transport down

the tube, leading to a positive change. This argument was proven in Chapter 3 when the heat

transport decreases when the region of flow becomes more constricted. This was the case of an

inner rod being stuck inside an open tube to form an annulus. The positive gain in transport from

the thick region is higher than the loss from the thin region, causing the overall net change to be

positive for low frequencies.

When the frequency increases, something different happens. Since the walls are

oscillating, the velocity near the walls will be higher than in the center of the tube. This is

similar to the oscillating flat plate problem described in Chapter 2, when the velocity is at its

highest at the wall and decreases as one moves away from it. This is especially true when the

frequency is high, because there will be a sharper velocity gradient going from the wall to the

core. Since the thick region has more available area farther away from the walls, the heat is

essentially moving from the fast to the slow part of the flow (from the thin region to the thick

region). This means that heat will take a longer period time to move down the tube. This

phenomenon will become more and more dominant as the frequency is increased. That is why

the net transport becomes negative, since the effect of fast moving walls dominates over the

effect of the free flowing region, and continues to be more and more negative as frequency is

increased. Now that the models have been presented with various geometric comparisons,

experiments can be designed to test the accuracy of these models.





























Figure 4.1: Difference between centered annulus (clear inner rod) and off-centered annulus
(shaded inner rod)


500


0


-500


t -1000
0
CL

g -1500

a)

-C
0-200

-2500


-3000


-3500 L
0


0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (Hz)


Figure 4.2: Net change in transport due to off centered inner rod at various frequencies















R

centered annulus off centered annulus


Figure 4.3: Thin and thick regions that result from an annulus that is off-centered











CHAPTER 5
THE EXPERIMENT AND RESULTS FROM THE EXPERIMENTATION

The theory presented thus far shows how the enhancement of heat transfer depends on

frequency of oscillation, the amplitude of the priming motion, and the geometry. An

experimental apparatus was designed to test the validity of the theoretical model. This

experiment was meant to verify the case of piston driven flow in a cylindrical tube. The

apparatus is described below as well as the method that was used to obtain the data. The reader

will learn about the difficulties that arose when designing the testing procedure, as well as the

logic behind the alterations that were made to improve the performance of the experiment. The

possible experimental errors are also discussed, and finally the results are given with the

discussion as to why the data agrees or disagrees with certain aspects of the theory.

5.1 Experimental Set-Up

The experimental design that was used in this study is pictured in Figure 5.1. A sketch is

shown in Figure 5.2a and 5.2b. The top cylindrical tank was used as the heat source, while the

bottom tank was the heat sink, with a connecting tube between them. The top tank was

approximately 6 L in volume, and the bottom tank had a volume of 4.5 L. The length of the pipe

in the figure was 30 cm, although that length could be varied. The fluid medium filled the

bottom tank, the connecting pipe, and most of the upper tank. The orientation of the heat source

above the heat sink was optimal since a lighter (warmer) fluid was on top of a heavier (colder)

fluid. This minimized the effect of natural convection arising from gravitational effects. The

oscillations were imposed by a mechanical piston that was attached to a bull wheel which was

rotated by a power drill. The amplitude of the oscillations was changed by adjusting the position

of the hinge on the bull wheel, while the frequency was changed by varying the speed of the









power drill. It was important to use a material for the piston that could handle a reasonable

amount of torque (from the drill). In the early exploratory stage of experimentation, several

pistons broke because of this issue. After a while, the acrylic piston was exchanged for a piston

made of PVC (polyvinyl chloride). This piston had a higher resistance to torque, and when

properly lubricated with oil, it was able to last the entire experimentation process. The material

selected for the tanks and the connecting pipe was plexiglass. This was chosen since its thermal

conductivity is low (0.18 W/m.K) compared to the fluids used during experimentation. These

fluids were water (conductivity of 0.60 W/m.K) and a 60% glycerol-water mixture (0.38

W/m.K). It was important to have a tank made of material that had a low conductivity since the

model assumed that no heat was lost through the walls of the tank. A material with a higher

thermal conductivity could transport heat out of the pipe in the radial direction, which would

provide a source of error.

An important part of running this experiment effectively was to control the temperature

of the heat source and the heat sink. This kept the axial temperature difference constant

throughout the experiment and consistent with the model. This was accomplished with the

assistance of controlled temperature baths that circulated constant temperature water through

both the heat source and the heat sink.

In the heat source, in addition to having the controlled temperature bath circulate the

water in the tank through a copper coil, a heater was placed in the tank that was controlled by a

PID controller. A mixing rod attached to a motor was also present in the heat source which

assisted in keeping the temperature uniform in the bath. Copper was chosen as the coil material

due to its high thermal conductivity. When the set point for the heater in the tank was

established, the heater turned on until it reached the set point since the set point was greater than









room temperature. However, as there was an overshoot as the system was trying to reach steady

state, the temperature bath had a set point which was slightly below the set point of the heater.

This assisted in removing the excess heat that the heater emitted as it tried to maintain constant

temperature in the upper tank. This was necessary since heat emitted by the heater in

overshooting the set point in theory really had nowhere to escape. Since the diameter of the

connecting tube was extremely small compared to the tank, excess heat would have taken an

inordinate amount time to move itself down the tube and out of the tank. This "cooling loop"

expedited the process.

In the lower tank there was also a cooling loop with a temperature bath; however, this

loop took up a very large amount of the volume of the tank. Larger amounts of coils were

needed in the sink since in addition to keeping the sink at constant temperature; there was heat

that was transported down the tube that needed to be removed once the oscillations began. Now

that the apparatus has been described, we discuss the method that was developed to gather data.

5.2 Experimental Method

When both tanks reached steady state, the heater in the heat source was on at a certain

rate to maintain the set point temperature. The controller measured this rate and it was recorded.

This was important to do since this rate took into account the heat loss through the coils as well

as losses to the ambient. Although the assumption of no heat loss through the walls of the tanks

and walls of the pipe was made, no amount of insulation could have made this assumption totally

accurate. Therefore heat was still lost through these walls, although the heat source and

connecting pipe were covered with multiple layers of foam insulation. This rate of heat loss

needed to be established before the oscillations began, since it had to be factored out of the total

rate. Another way that the heat leakages from the top tank were minimized was by the set point









of the heat source. The set point used for these experiments was always between 24 and 26

degrees Celsius, which were slightly above room temperature. This made the temperature

gradient between the hot tank and the ambient rather small, which aided in reducing heat loss out

of the heat source. The set point of the bottom tank was slightly below room temperature,

between 16 and 18 degrees Celsius. Again, this minimized the temperature gradient between the

heat sink and the surroundings. Although the axial temperature difference between the source

and the sink was only between 6 and 10 deg C for the experimental testing, the theory still

showed that a significant amount of heat transfer would occur.

When the oscillations began for a given amplitude and frequency, heat was transported

quickly down the pipe. This drove the temperature of the top tank down, since heat was being

lost to the sink. In order to compensate for this, additional heat was added to the source to drive

the temperature back up to the set point. The heat that was put back into the top tank to maintain

the set point was equal to the heat that was transported down the tube to the heat sink. Again, to

get the actual rate of heat that was transported down the tube, one must subtract the rate of heat

loss to the ambient and coils, which were determined before the oscillations started. In the

bottom tank, the temperature tried to increase as heat entered it from the pipe. This heat was

removed via the copper coils so that a constant temperature could still be maintained in the heat

sink. In both tanks, the coils removed the heat, and this heated water went into a chiller, where it

was cooled down. This chilled water then was sent back to the temperature bath, where it was

used to export constant temperature water back to the tanks. This cyclic process occurred

throughout the experiment.

In order to check the accuracy of the experimental results against the theory, several

different parameters were tested. One theoretical conclusion that was desired to be tested was to









see if the oscillations improved the heat transport by multiple orders of magnitude as predicted

by the model. However, in order to do that, the pure conductive case needed to be evaluated as

well. Unfortunately, the time required for the linear temperature profile to form in the tube via

conduction was very high when water was used as the transporting fluid. Since the model

assumed that this linear temperature profile existed, this must form first before the conductive

heat transfer is determined. The time constant for the conductive case scaled as the pipe length

squared over the thermal diffusivity. Remember that there were no oscillations in the conductive

case, so the heat transport occurred strictly in the axial direction, making the characteristic length

the length of the tube and not its radius. Even with a relatively small pipe length of 30 cm in this

experiment, it would still take on the order of one week for the steady linear profile to form. It

would be possible for this time to decrease if the pipe length was really small. However, the

model also neglected the entrance and exit effects as discussed before, meaning that the

oscillation amplitude used would have to decrease substantially as well. This small oscillation

amplitude would have enhanced the transport in a very minimal way, and would not have been

very useful.

Another possible way to avoid this problem would be to use a fluid with a much higher

thermal diffusivity, since this would have lowered the amount of time needed for the steady state

profile to form down the tube. Air would have been a possible substitute since its thermal

diffusivity is over 100 times that of water. However, a problem arose when air was put into the

tank. Mixing air in the tank proved to be very difficult as it was observed that there were severe

temperature fluctuations in the heat source. Since the model assumed a well mixed heat source,

another mixing method had to be used instead of the stirring rod. A few small fans were inserted

into the tank to help keep the air constantly circulating. However, it turned out that heat was









being generated from the fan motors, which in the end would skew the results since not all the

heat being added to the tank would have come from the heater, some would have come from the

fan motors. Since only the output of the heater got measured, this would have become a large

source of error. In the end, it was determined that air wasn't a feasible test fluid, and water

became the fluid of choice again.

It was then decided to compare different convective cases since the time needed for a

convective steady state to form was on the order of a few minutes under the influence of

oscillations. Different parameters in the apparatus were varied to see if indeed the experimental

results followed the theoretical model. Before that occurred, the sources of error in the

experiment needed to be evaluated to determine how much they would affect the results.

5.3 Error Analysis

In the experimental set-up, there were several sources of error that arose. The first issue

was related to the physical properties of the fluid. There are several that come into play when

calculating the heat transported through the system. These include the thermal conductivity, the

kinematic viscosity, and the density of the fluid. These properties are functions of temperature,

so their values varied when the temperature changed. The property values obtained for both

water and the 60% glycerol-water mixture were those at room temperature (20 C). Since the set

point of the upper tank was a few degrees higher than room temperature (24-26 C), and the set

point of the lower tank was a few degrees below room temperature (16-18 C), it was assumed

that the average temperature in the tank was roughly that of the room temperature. However, if

the upper temperature limits of the tank were used, the average temperature in the pipe would be

22 C instead of 20 C. It was also possible for the room temperature to have been greater than 20

C, which might have influenced the average temperature in the pipe. However the pipe was well









insulated and it was unlikely that the room temperature excursion would have had much of an

effect. Therefore the physical properties of water and glycerol needed to be examined between

20 and 22 C. A range of values was obtained (see Tables 5.1 and 5.2).

These data show that even with a two degree temperature difference, there is a difference

in the Prandtl number of the fluid. The contributing factor to this is the kinematic viscosity,

which appears in the Prandtl number. It can be seen that for water there can be as much as a

2.3% deviation from the average Prandtl number and a corresponding deviation of 3.5% for the

glycerol-water mixture.

Other errors arose from deviations in the dimensions of the experiment from design

values. The pipe had a total design length of 30 cm. It was divided into two pieces, each having

a tolerance of 0.0127 cm. The tolerance in the radius of the pipe was also 0.0127 cm. Two

different radii were used in the experiment, whose sizes were 3/16 in. and 1/8 in. The error then

in the pipe length was 0.085%, while the error in radius was 2.68% for the wide tube, and 4.00%

for the narrow tube.

The amplitude measurement was taken by determining the length of the piston hinge

from the center of the bull wheel. This was equivalent to the actual length the piston was

moving. The maximum possible amplitude that the piston could have generated was 7.14 cm,

and minimum amplitude of 2.73 cm. Note that this wasn't the actual amplitude of the flow in the

pipe since the tube in which the piston oscillated had a different radius then that in the actual

pipe. The amplitude in the pipe can be easily determined by using the following formula which

is that of equal volume displacement. It is

Apiston (Rpston ) = ApIpe (Rpipe )2 (5.1)









whence the amplitude in the pipe is calculated. The pipe amplitude not only varied when the

piston amplitude was changed, but also with the radius of the pipe, as different sizes were used

during experimentation. The piston amplitude was measured using calipers. Although the

calipers were quite accurate, an eyeball measurement needed to be taken to place the caliper ends

with the center of the hinge and the center of the bull wheel. The error in the measurement was

assumed to be about 0.5 mm. This means that for the smallest piston amplitude, the maximum

error that could exist is 1.83%. A larger piston amplitude would produce a smaller percentage

error.

The temperature drop between the heat source and heat sink was dependent on the

accuracy of the controllers in the respective tanks. The controller produced very good results as

the temperature range was always within 0.05 C of the set point in both tanks at steady state.

Temperature measurements were taken every 5 seconds, meaning that 240 measurements were

taken over the 20 minute trial period to obtain an average temperature reading. It was unlikely

that the error in the average temperature was anything over 0.01 C for each tank. The error in an

8 C temperature gradient (the typical temperature difference) would have been about 1/4 %.

Another source of error in the experiment occurred when measuring the frequency. Since

the bull wheel was attached to a drill with a motor, there would be surges or decreases in the

voltage to the motor. This would cause a change in the frequency measurement for short periods

of time. When this occurred, the drill was adjusted quickly to bring the frequency back to the

original set point. It was determined that the maximum change that was brought about in the

frequency by these voltage changes was no more than 0.2 Hz. This change usually lasted no

more than a minute before it was noticed and rectified. The amount of time for a trial period was

20 minutes, meaning that the difference in the average frequency was 0.01 Hz per voltage









change. The number of voltage changes usually never exceeded 3 in a trial period, which made

the total deviation in the average frequency 0.03 Hz at most. The frequencies used during the

entire experimentation process ranged between 0.9 and 1.7 Hz. The maximum error for small

frequencies would have been roughly 3.3% and 1.8 % for large frequencies.

So far, several input parameters have been discussed in regard to the error in their

measurements. These include kinematic viscosity, thermal diffusivity, thermal conductivity,

temperature gradient, pipe length, pipe width, oscillation amplitude, and oscillation frequency.

Note that all these parameters appear in (3.40), which was the model equation for piston driven

flow in a cylinder. Since the overall heat transport was proportional to all of these parameters in

different ways, the maximum and minimum values that arose from errors were substituted into

(3.40) for each term to get the maximum and minimum possible value for the overall heat

transport. There was another source of error in the experiment that doesn't appear in (3.40).

When the actual amount of heat transported in the experiment was determined, the time that the

heater was on was counted by a timer in the computer that controls the heater. This amount of

time had an error associated with it. The timer was rounded off to the nearest hundredth of a

second. When a start time and stop time were determined, there was a rounding error associated

with each time in addition to the start and stop time for when the heat loss to the ambient was

calculated. Since there was a maximum of 0.005 s of rounding error with each timing, a total of

0.02 s in error could result when 4 timings are done. Considering the heater was on anywhere

from 3 to 100 s depending on the trial, the error associated with this could range from 0.02% to

0.67%. In the case where a ratio of two trials was taken, 8 timings were done, and the maximum

total error doubled to 0.04 s, resulting in a percentage error ranging from 0.04% to 1.33%. This

additional error was added into the maximum and minimum values for the overall heat transfer









to obtain the total experimental error. This total error is expressed on a trial to trial basis with

error bars shown on the plots in the results section.

5.4 Note on Viscous Dissipation

A term that was neglected in the energy balance earlier was that of viscous dissipation

due to movement of the fluid in the tube. If this term were to appear in the energy balance,

(3.13) would be rewritten as

OT OT 02T 1 OT ap _V2
-+ = ) +,-- ) (+ ) (5.2)
at az r2 r Or pC, Or

We can scale (5.2) to obtain

drT A 2T 1 T* V*2
WoPr( T + Vz)= ( 2+ )+Br( )2 (5.3)
dt L Or *2 r r *r

Where the Brinkman number is defined as


Br =uA (5.4)
kAT

Typical experimental values were used to determine the Brinkman number. When water

was considered as the fluid medium, this dimensionless group turned out to be five orders of

magnitude lower than the first order conductive term and four orders of magnitude lower for the

glycerol-water mixture. The dimensionless groups of the time dependent and convective terms

were even greater than the conductive term. Although from a scaling perspective, we have

reasoning to drop the viscous dissipation term, we need to evaluate the entire term to see how

much heat is really lost due to the viscosity. Therefore, a calculation was performed by

integrating the viscous dissipation term over the cross sectional area and comparing it to the

output heat transport. It turned out the heat transport was over 500 times the heat lost to viscous

dissipation for the case of water as a fluid, and over 50 times the heat lost in the case of a









glycerol-water mixture. From these results, it can be concluded that the viscous dissipation term

is insignificant, and this why it was neglected in the energy balance.

5.5 Experimental Results and Discussion

5.5.1 Varying Amplitude with Water

The first series of experiments were meant to verify the relationship that the convective

heat transferred is proportional to the square of the piston amplitude. Initially, it was determined

that the time needed to reach the conductive steady state for water was roughly 120 hours for the

30 cm pipe length. Through a theoretical calculation it was figured that the oscillations increase

the overall transport by a factor of about 600. A rough estimate was made to determine the

amount of time needed for the convective steady state to obtain. It was roughly 12 minutes.

Once the oscillations were in effect for that length of time, it was assumed that the velocity

profile in the tube was fully developed and heat was being transported at a constant rate.

Initially, the amount of heat input to the top tank over a 20 minute period was measured

without oscillations in order to determine the amount of heat lost to the ambient. Once that

measurement was completed the oscillations were started. The steady state profile was developed

and another 20 minute measuring period was conducted for an input piston amplitude. In the

next experiment, the amplitude was changed and there was a wait time until the steady state

formed at the new amplitude. The heat transported was again measured over a span of 20

minutes. It should be noted here that the frequency was set between 1-1.5 Hz for every trial.

Although there was a variation in frequency over all the trials, each individual trial had the same

frequency for each of the two amplitudes tested.

Figure 5.3 shows the experimental results compared to a theoretical curve at various

amplitude ratios. The amplitudes chosen for these trials could not exceed 40% of the pipe length









since a larger amplitude would have induced entrance and exit effects- effects that were

neglected in the model. This limited the amount of possible amplitude ratios available, which is

why the ratios shown in the figure are between 1.1 and 1.6.

Observe from the figure that the data points are quite close to the theoretical curve for

amplitude ratios between 1.2 and 1.5. For the extremely low and extremely high ratios, the

experimental results were slightly higher than the model prediction. The reasoning for the

difference at the high ratios could possibly be due to the entrance and exit effects. Although a

40% limit was placed on the amplitude size, this is just a rough guess as to when the entrance

and exit effects become relevant.

When an oscillation pulse is generated, the temperature gradients ought to exist only in

the pipe and the model assumes that these gradients are not felt in either tank. However, when a

spike becomes large, temperature surges can occur up through the entire length of the pipe and

into the tank. Since the heater and controller were both in the top tank, a sudden temperature

drop in the tank would have obviously caused more heat to be added to the tank to keep the

temperature at the set point.

Now the model only takes into account heat transported between the ends of the pipe.

Because a larger portion of the tank was now cooler, not just at the spot where the pipe was

connected, a greater amount of heat would have to be added to cool the tank down. These are

considered the entrance and exit effects because the transport occurs outside the ends of the pipe

in addition to inside the pipe. This excess heat increases the total amount of time the heater is on

during an experimental run. This overshoot causes the heat transport ratio to become larger than

expected (since the transport ratio is given with the higher amplitude result in the numerator) and

this is consistent with what the data shows for high amplitude ratios. If higher amplitude ratios









were used than the ones shown in Figure 5.3 one would likely see the error become even greater,

as the spikes generated would become ever larger, affecting the tanks more and more.

The discrepancy at lower amplitude ratios would not be for the same reason as the high

end ratios, since the spikes were not as large. It is likely that the error in the data points between

the amplitude ratio of 1.1 and 1.2 are due to some additional heat leakages in the experiment.

These heat leakages will be discussed in a later section in this chapter. Looking at the points in

the middle of the data set (ratios between 1.2 and 1.3), it can be seen that these points are in very

good agreement with the theory.

5.5.2 Varying Amplitude with Glycerol-Water Mixture

The same experimental method discussed in the previous section was applied to a fluid

with a higher Prandtl number, a 60% glycerol-40% water mixture which has a Prandtl number

around 85. The idea behind running this part of the experiment is to see if the model is a good

predictor for a more viscous fluid and also to see if a higher Prandtl number fluid would behave

worse than a low Prandtl number fluid for varying amplitudes. This could possibly be so as a

higher Prandtl number fluid usually decreases the convective transport in the pipe. The decrease

occurs because a higher Prandtl number fluid usually is more viscous (the glycerol water-mixture

is roughly 10 times the kinematic viscosity of water). The term "usually" is used here since

thermal diffusivity can affect the Prandtl number as well. Two fluids can have different Prandtl

numbers even when they have the same kinematic viscosity due to differing thermal diffusivity

values. It has been shown from the model in Chapter 2 that the heat transport decreases with

decreasing Womersley number. Since the kinematic viscosity appears in the denominator of the

definition of the Womersley number (see Chapter 2), a more viscous fluid will decrease that

value, assuming the oscillation frequency and tube radius are kept constant. Although ratios in









heat transported are being considered here, this could affect the accuracy of the experiment. For

example, suppose at a certain amplitude, the heat per unit time transported down a water filled

pipe is 2 W. If the amplitude were doubled, the transport is predicted to become 8 W according

to the model. Now suppose at the same original amplitude the heat transported in a glycerol-

water pipe is 1 W. If twice the amplitude was used, the theoretical transport would become 4

times that, or 4 W. For the water example, the rate of heat transport increases by 6 W when the

amplitude is doubled. In the glycerol water case, the increase is only 3 W. Suppose in both

cases the experiment produced an increase in transport that was 1 W higher than expected. The

percent error would be 16.67% for the water case (6 W +/- 1 W) versus 33.33% for the glycerol-

water mixture (3 W +/- 1 W). From this example, it could be concluded that the glycerol-water

mixture is more sensitive from an percent accuracy standpoint. Using this fluid could prove to

be a better test to see how accurate the experiment really is. Figure 5.4 shows the results for the

glycerol-water mixture.

The results here show better agreement with the theoretical prediction compared to the

trials with pure water. All the data points are within the experimental error bars. It should be

noted though that the very high amplitude ratios were not taken into account in this experiment

due to the larger errors obtained in the last experiment at those respective values. If large

amplitude ratios were tested with the glycerol-water mixture, the errors associated with them

would be even higher as a more viscous fluid generates sharper velocity spikes, causing even

more temperature gradients to form in the tanks when the amplitudes are large. That is why only

ratios below 1.4 were tested. This data set is a positive sign, as good agreement with the theory

was obtained for a fluid medium that was less conducive to transporting heat, meaning that it