"Applied Science, Faculty of"@en .
"Mechanical Engineering, Department of"@en .
"DSpace"@en .
"UBCV"@en .
"Sylvester, Alexander"@en .
"2018-04-30T00:00:00"@en .
"2017"@en .
"Master of Applied Science - MASc"@en .
"University of British Columbia"@en .
"Energy Recovery Ventilators (ERVs) save energy by recovering sensible and latent heat from exhaust ventilation streams. They consist of compact channels in a cross- or counter-flow arrangement separated by moisture-permeable membranes. This work employs computational fluid dynamics with experimental validation to study angled rib mixing features to enhance heat and mass transport in membrane-based ERVs. The model simulates air-to-air heat and mass exchange between two rectangular ducts in a counter-flow arrangement. Local results from these simulations are used in a mathematical model to predict the effectiveness of a cross-flow ERV. The flow is steady and laminar. The channels are modeled with and without ribs for various channel aspect ratios and flow rates. Periodic inlet/outlet conditions are assumed for all cases to highlight the effect of the ribs. Ribs are either on the membrane surface or the opposite wall, however ribs on the membrane surface are the focus of this work due to their superior performance. \r\n\r\nResults show that the ribs increase the channel Sherwood and Nusselt numbers by a larger fraction than the corresponding increase in friction factor. The improvement is larger for sensible recovery than latent recovery because the membrane\u00E2\u0080\u0099s mass-transfer resistance is higher than its thermal resistance. For a typical commercial grade, counter-flow ERV, total effectiveness can be improved by over 10% for an equal pressure drop by adding ribs and slightly increasing the channel height. \r\n\r\nMass-transfer and pressure drop results from the simulations were validated experimentally for channels with and without ribs using a commercially available membrane (dPoint Technologies). Ribs were either formed into the membrane surface or made as extensions of the channel wall. The experimental and simulation results agree with one another within the experimental uncertainty of the test apparatus and variability of the membrane permeability.\r\n\r\nTypically, ERV performance targets (recovery effectiveness, pressure drop) are met by varying the channel dimensions or number of channels. The work presented here indicates that angled ribs could be used instead, which would not require altering the ERV footprint or using extra materials."@en .
"https://circle.library.ubc.ca/rest/handle/2429/63366?expand=metadata"@en .
" RIB-ROUGHENED HEAT AND MASS TRANSFER ENHANCEMENT IN MEMBRANE-BASED ENERGY RECOVERY VENTILATORS by Alexander Sylvester B.Sc., McGill University, 2014 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Mechanical Engineering) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) October 2017 \u00C2\u00A9 Alexander Sylvester, 2017 ii Abstract Energy Recovery Ventilators (ERVs) save energy by recovering sensible and latent heat from exhaust ventilation streams. They consist of compact channels in a cross- or counter-flow arrangement separated by moisture-permeable membranes. This work employs computational fluid dynamics with experimental validation to study angled rib mixing features to enhance heat and mass transport in membrane-based ERVs. The model simulates air-to-air heat and mass exchange between two rectangular ducts in a counter-flow arrangement. Local results from these simulations are used in a mathematical model to predict the effectiveness of a cross-flow ERV. The flow is steady and laminar. The channels are modeled with and without ribs for various channel aspect ratios and flow rates. Periodic inlet/outlet conditions are assumed for all cases to highlight the effect of the ribs. Ribs are either on the membrane surface or the opposite wall, however ribs on the membrane surface are the focus of this work due to their superior performance. Results show that the ribs increase the channel Sherwood and Nusselt numbers by a larger fraction than the corresponding increase in friction factor. The improvement is larger for sensible recovery than latent recovery because the membrane\u00E2\u0080\u0099s mass-transfer resistance is higher than its thermal resistance. For a typical commercial grade, counter-flow ERV, total effectiveness can be improved by over 10% for an equal pressure drop by adding ribs and slightly increasing the channel height. Mass-transfer and pressure drop results from the simulations were validated experimentally for channels with and without ribs using a commercially available membrane (dPoint Technologies). iii Ribs were either formed into the membrane surface or made as extensions of the channel wall. The experimental and simulation results agree with one another within the experimental uncertainty of the test apparatus and variability of the membrane permeability. Typically, ERV performance targets (recovery effectiveness, pressure drop) are met by varying the channel dimensions or number of channels. The work presented here indicates that angled ribs could be used instead, which would not require altering the ERV footprint or using extra materials. iv Lay Summary Energy Recovery Ventilators (ERVs) are heat exchangers used in building ventilation systems. They are used to move heat and moisture between the outgoing and incoming building air streams and save energy. This is especially beneficial in hot, humid climates where air conditioners typically consume a lot of energy cooling and dehumidifying air. ERVs consist of an array of channels separated by moisture-permeable membranes. This study investigates a way of improving the performance of ERVs by adding angled rib features to the channels. Most of the work was done with numerical simulations, with subsequent validation experiments. Results show that the ribs increase the heat and moisture recovery, but also increase the fan power required to push air through the ERV. However, the overall impact is a net benefit. In fact, total effectiveness can be improved by over 10% for equal fan power by adding ribs and increasing the channel height. v Preface The general research objective of geometric enhancement was identified by David Kadylak and Ryan Huizing of dPoint technologies and Professors Sheldon Green and Steven Rogak of the University of British Columbia. I subsequently identified specific geometries to study, setup and ran appropriate simulations, designed and executed experimental validation, and analyzed the resulting data. Simulations were performed using ANSYS Fluent\u00C2\u00AE. The experimental testing apparatus described in Section 2.2.1 was designed and built in conjunction with PhD candidate Amin Engarnevis. The preliminary work for this project was completed by MEng Student Rajitha Nakandala with supervision by Amin Engarnevis. It involved using ANSYS Fluent\u00C2\u00AE to study mass transfer between two smooth triangular ducts in a counter-flow arrangement. The Matlab model used to predict the cross-flow core performance with angled ribs in Section 3.2.2 was written by Amin Engarnevis and Steven Rogak. The results described in Chapter 3 (except section 3.2.2) will be submitted for publication in the International Journal of Heat and Mass Transfer as \u00E2\u0080\u009CCFD Analysis of Rib-Roughened Channels in Membrane-Based Energy Recovery Ventilators\u00E2\u0080\u009D. The results of section 3.2.2 will be submitted for publication in the journal Energy and Buildings as \u00E2\u0080\u009CConjugate heat and mass transfer modeling in membrane-based enthalpy exchangers: performance enhancement via rib-roughened channels\u00E2\u0080\u009D. vi Table of Contents Abstract .......................................................................................................................................... ii\tLay Summary ............................................................................................................................... iv\tPreface .............................................................................................................................................v\tTable of Contents ......................................................................................................................... vi\tList of Tables ................................................................................................................................ ix\tList of Figures .................................................................................................................................x\tList of Symbols ........................................................................................................................... xiii\tList of Abbreviations ................................................................................................................. xiv\tAcknowledgements ......................................................................................................................xv\tChapter 1: Introduction ................................................................................................................1\t1.1\t Energy Recovery in Air Conditioning ............................................................................ 1\t1.1.1\t Building Ventilation ................................................................................................ 1\t1.1.2\t Energy Recovery Ventilators .................................................................................. 2\t1.1.3\t Other Applications .................................................................................................. 5\t1.2\t Theory ............................................................................................................................. 6\t1.2.1\t Heat and Mass Transport ........................................................................................ 7\t1.2.2\t Membrane Transport ............................................................................................... 8\t1.2.3\t Performance Metrics ............................................................................................. 10\t1.2.4\t Periodic Boundary Conditions .............................................................................. 13\t1.2.5\t Geometric Enhancement ....................................................................................... 15\t1.3\t Current Study Objectives .............................................................................................. 19\tChapter 2: Methods .....................................................................................................................20\tvii 2.1\t Simulation ..................................................................................................................... 20\t2.1.1\t Problem Description ............................................................................................. 20\t2.1.2\t Simulation Parameters .......................................................................................... 22\t2.1.3\t Membrane Modelling Approach ........................................................................... 22\t2.1.4\t Convergence Criteria ............................................................................................ 23\t2.1.5\t Meshing and Mesh Validation .............................................................................. 23\t2.1.6\t Post Processing ..................................................................................................... 27\t2.2\t Experimental ................................................................................................................. 28\t2.2.1\t Water Vapour Transport Station ........................................................................... 28\t2.2.2\t Testing Module ..................................................................................................... 29\t2.2.3\t Membrane Forming .............................................................................................. 31\tChapter 3: Results and Discussion .............................................................................................33\t3.1\t Simulations ................................................................................................................... 33\t3.1.1\t Membrane Ribs ..................................................................................................... 33\t3.1.2\t Ribs on Walls ........................................................................................................ 41\t3.2\t Effectiveness Predictions .............................................................................................. 43\t3.2.1\t Counter-Flow Channel .......................................................................................... 43\t3.2.2\t Cross-Flow ............................................................................................................ 45\t3.3\t Experimental Validation ............................................................................................... 51\tConclusions ...................................................................................................................................57\tReferences .....................................................................................................................................59\tAppendix A \u00E2\u0080\u0093 User Defined Functions .......................................................................................61\tA.1\t Membrane transport, Nusselt number, Sherwood number, and Friction Factor ....... 61\tviii A.2\t Periodic Boundary Conditions .................................................................................. 70\tA.3\t Rectangular Channel Fully Developed Velocity Profile .......................................... 79\t ix List of Tables Table 1.1 Previous studies investigating heat transfer enhancement in ducts with angled ribs ... 17\tTable 2.1 Dimensions of the housing slot ..................................................................................... 31\tTable 3.2 Measured dimensions of the flow channel inserts ........................................................ 55\tTable 3.3 Air properties assumed at 50\u00C2\u00B0C and 1atm .................................................................... 55\t x List of Figures Figure 1.1 - Typical thermal comfort zone on a psychrometric chart ........................................... 1\tFigure 1.2 \u00E2\u0080\u0093 Function of an ERV .................................................................................................... 3\tFigure 1.3 \u00E2\u0080\u0093 Composite polymer membrane construction . ............................................................ 4\tFigure 1.4 \u00E2\u0080\u0093 ERV Construction. Cross-flow (left) and counter-flow (right). ................................. 5\tFigure 1.5 - dPoint fuel cell humidifier .......................................................................................... 6\tFigure 1.6 \u00E2\u0080\u0093 Membrane permeation/separation modes . ................................................................. 9\tFigure 1.7 - Hydrodynamically developing velocity profile ........................................................ 13\tFigure 1.8 - Resistance in series model ........................................................................................ 16\tFigure 1.9 - Common enhancement geometries ........................................................................... 17\tFigure 2.1 \u00E2\u0080\u0093 Schematic of the simulation domain ........................................................................ 21\tFigure 2.2 \u00E2\u0080\u0093 Isometric view of the mesh, mesh spacing has been doubled in each dimension for clarity ............................................................................................................................................ 25\tFigure 2.3 \u00E2\u0080\u0093 Various views of the mesh. Mesh spacing has been doubled in each dimension for clarity ............................................................................................................................................ 26\tFigure 2.4 \u00E2\u0080\u0093 Mesh sensitivity study .............................................................................................. 27\tFigure 2.5 \u00E2\u0080\u0093 Experimental test station .......................................................................................... 29\tFigure 2.6 \u00E2\u0080\u0093 Aluminum housing of testing module ...................................................................... 30\tFigure 2.7 \u00E2\u0080\u0093 Flow configuration in the membrane housing module ............................................. 31\tFigure 2.8 - Membrane forming template. Dimension are in inches. ........................................... 32\tFigure 3.1 - Transverse velocity fields 1, 5, and 8 step widths downstream of the rib for the middle of the three flow rates. (a) AR = 0.6, Re = 390. (b) AR = 1, Re = 468. (c) AR = 1.4, Re = 585................................................................................................................................................. 35\txi Figure 3.2 - Transverse velocity fields 1, 5, and 8 step widths downstream of the rib for AR = 1. (a) Re =703. (b) Re = 468. (c) Re = 234. ...................................................................................... 36\tFigure 3.3 \u00E2\u0080\u0093 Speed contour plot for the AR = 1, Re = 468 case ................................................... 37\tFigure 3.4 \u00E2\u0080\u0093 Contour plots of humidity (top) and temperature (bottom) on the membrane surface for the AR = 1, Re = 468 case. Looking down to on the feed side. Red (blue) indicates highest (lowest) humidity and temperature. .............................................................................................. 37\tFigure 3.5 \u00E2\u0080\u0093 Local performance data along the channel for AR = 1 and Re = 468. ..................... 38\tFigure 3.6 \u00E2\u0080\u0093 Performance data of the ribbed and smooth channels for three different flow rates. (a) Nusselt number. (b) Sherwood number. (c) Friction factor times Reynolds number. (d) Relative performance of ribbed and smooth channels for the AR=1 case. ................................... 41\tFigure 3.7 - Diagram of ribs on walls ........................................................................................... 42\tFigure 3.8 \u00E2\u0080\u0093 Performance of smooth, membrane ribs, and wall ribs ............................................ 42\tFigure 3.9 \u00E2\u0080\u0093 Performance data of a 1m long counter-flow channel .............................................. 43\tFigure 3.10 - Recovered power and fan power for AHRI standard summer operating conditions........................................................................................................................................................ 45\tFigure 3.11 - Cross-flow ERV configuration ............................................................................... 46\tFigure 3.12\u00E2\u0080\u0093 Cross flow core simulation domain ......................................................................... 47\tFigure 3.13\u00E2\u0080\u0093 Cross flow local Nu and Sh values for various values of Re .................................. 48\tFigure 3.14. - Variation of the effectiveness of enthalpy exchangers with smooth and ribbed channels for three different operating flowrates.. ......................................................................... 50\tFigure 3.15. - Total recovered power of enthalpy exchangers with smooth and ribbed channels for three different operating flowrates in AHRI Summer (cooling) operating conditions. .......... 50\tFigure 3.16 \u00E2\u0080\u00933D-Printed Flow channel insert.. ............................................................................. 52\txii Figure 3.17- Rib features formed into the membrane ................................................................... 53\tFigure 3.18 - Rectangular channel cross-section and dimensions ................................................ 54\tFigure 3.19 \u00E2\u0080\u0093 Experimental performance of smooth and ribbed channels with corresponding simulation predictions. .................................................................................................................. 56\t xiii List of Symbols A = Area c = Species mass concentration D = Mass diffusivity Dh = 4A/P = Hydraulic diameter f = Friction factor H = Channel height J\u00E2\u0080\u009D = Mass flux per unit area l = membrane thickness Nu = Nusselt number P = membrane permeability q\u00E2\u0080\u009D = Heat flux per unit area Re = Reynolds number Sh = Sherwood number T = Temperature W = Channel width Y = Humidity mass fraction \u00CE\u00BB = Thermal conductivity Subscripts 1,2,3,4 = sweep intlet, sweep outlet, feed inlet, feed outlet b = bulk fluid property w = wall property f = feed stream s = sweep stream xiv List of Abbreviations ERV \u00E2\u0080\u0093 Energy Recovery Ventilator HRV \u00E2\u0080\u0093 Heat Recovery Ventilator UDF \u00E2\u0080\u0093 User Defined Function WRR \u00E2\u0080\u0093 Water Recovery Ratio AR \u00E2\u0080\u0093 Aspect Ratio xv Acknowledgements Studying on this project to further my education has been a truly fulfilling experience. Not only did I grow as an engineer and researcher, but also as a person. I would like to thank my supervisors, Dr. Steve Rogak and Dr. Sheldon Green for supervising my studies and providing deep and insightful guidance throughout my time at UBC. I was fortunate enough to work collaboratively with dPoint Technologies during my studies. They are an exciting and innovative company and were collectively a pleasure to work with. In particular, I would like to thank David Kadylak and Dr. Ryan Huizing for their expert guidance and supervision during my thesis and a summer internship I spent at dPoint. My fellow graduate students, Amin Engarnevis and Sarah Cosby, were wonderful to work with and made my workdays so much more pleasant. Talking things through with them always lead to solutions to our problems. I am happy to call them my friends. Last but not least, I would like to thank my family. My parents and brother have always been incredibly supportive of my endeavors and even though we are usually thousands of kilometers apart, our skype conversations never failed to bring a smile to my face. I have also been incredibly lucky to have my loving partner, Kay Penney, by my side for the past 2 years. She moved across the continent, and into a new country, so we could stay together during my study and none of this would have been possible without her. 1 Chapter 1: Introduction 1.1 Energy Recovery in Air Conditioning 1.1.1 Building Ventilation Air conditioning accounts for roughly 50% of the energy consumption in the building sector in North America [1], [2]. It is required in order to maintain occupant comfort and health in buildings, which is increasingly important as people spend more of their time indoors. Figure 1.1 shows the thermal comfort zone on a psychrometric chart for winter (1.0 clo) and summer (0.5 clo) clothing levels [3]. It is the goal of the air conditioning system to maintain conditions within the comfort zone. This is complicated by the need for fresh air supply to maintain air quality. Indeed, minimum flow rates are required by modern standards [4]. Therefore, outdoor air must be continually introduced into the building and conditioned before it is exposed to occupants, making for a tradeoff between energy expenditure, thermal comfort, and air quality. Figure 1.1 - Typical thermal comfort zone on a psychrometric chart [3] 2 A significant component of air conditioning is controlling the temperature and humidity. This is especially important in hot and humid climates where it accounts for 20-40% of the air conditioning load, and even more when 100% fresh air is required (eg. Hospitals) [5, pp. 1\u00E2\u0080\u00932]. Humidity can be particularly energy-intensive to control due to the high heat of vaporization of water vapour. Further, the method often used to control humidity, in which air is dehumidified by cooling it to below its dew point and the re-heating it to comfortable levels, is particularly energy intensive [5, pp. 5\u00E2\u0080\u00936]. Because of the importance of adequate air conditioning, its high energy cost, and the need to continually bring in air from the outdoors, it is desirable to recover energy from exhaust air that has already been conditioned. One popular technology for this is the energy recovery ventilator (ERV). 1.1.2 Energy Recovery Ventilators Energy recovery ventilators (ERVs) provide a way to recover sensible and latent heat from exhaust air in ventilation systems. This reduces energy demand since intake air is partially conditioned by exhaust air. They differ from heat recovery ventilators (HRVs) which recover only sensible heat. ERVs work by passing exhaust and intake air through adjacent flow passages separated by a thermally conductive and porous material. This material must strike a balance between permeability to heat/moisture and impermeability to air-born contaminants which would otherwise be recirculated back into the building, shown in Figure 1.2. 3 Figure 1.2 \u00E2\u0080\u0093 Function of an ERV Particularly promising materials for ERVs are polymeric membranes with high water-vapour permeability and selectivity of water-vapour over contaminants, such as those used by dPoint Technologies [6]. They are anisotropic membranes made up of a porous substrate for structural support (~100\u00C2\u00B5m) and a dense selective coating (0.1-10\u00C2\u00B5m), shown in Figure 1.3 [7]. The coating is made sufficiently thin to maintain high water-vapour permeability. The overall thickness of the membrane is small enough so that its sensible heat transfer resistance is negligible. Overall, the membrane water vapour transfer resistance is several orders of magnitude larger than the thermal resistance, making the behavior of the two processes different [8, pp. 12\u00E2\u0080\u009314]. 4 Figure 1.3 \u00E2\u0080\u0093 Composite polymer membrane construction [7]. ERVs are commonly made as parallel plate exchangers in either a cross- or counter-flow arrangement [8, pp. 93\u00E2\u0080\u0093120]. They are made by stacking multiple layers of membrane material separated by spacers. Both cross- and counter-flow arrangements are shown in Figure 1.4. Cross-flow cores are easier to fabricate, but have lower effectiveness [9, pp. 336\u00E2\u0080\u0093347]. When selecting a core for an application, there is a tradeoff between footprint, cost, and effectiveness. Traditionally, recovery effectiveness targets are met by varying the channel height, width and length. However, this requires more material and increases the ERV footprint. Another method of improving effectiveness is by using more permeable material, however highly-permeable and selective materials are typically costly. 5 Figure 1.4 \u00E2\u0080\u0093 ERV Construction. Cross-flow (left) and counter-flow (right). An alternative method of increasing performance is through geometric mixing features within the channels themselves. As air moves along a smooth channel, the temperature and moisture concentration fields become less homogeneous, which reduces the effectiveness of the exchanger [8, pp. 120\u00E2\u0080\u0093123]. Certain variations to the flow passage shape can promote mixing and reduce the heterogeneity of the temperature and moisture distributions. This will be discussed in more detail in section 1.2.5. 1.1.3 Other Applications While ERVs are the main focus of this work, heat and mass exchangers have other important applications which could also benefit from the ideas discussed here. One other application is for PEM fuel cells. Fuel cells require a controlled level of moisture at the inlet even though the fuel cell reactions produce water [10]. Membrane-based fuel cell humidifiers offer a method of recovering the produced moisture and using it for inlet humidification. They are similar to ERVs, though the temperature and humidity is typically higher and the footprint must be smaller [11]. A commercial cross-flow fuel cell humidifier is shown in Figure 1.5. Cross-Flow\t Counter-Flow\t6 Figure 1.5 - dPoint fuel cell humidifier Other applications include gas separation with specialized membranes that are selective to a particular species of interest. These are typically large-scale industrial processes used to obtain a pure gas from a mixture. Examples include separating nitrogen from air, hydrogen from ammonia plant purge-gas streams, carbon dioxide from methane in natural gas, and dehydration of air [12, pp. 325\u00E2\u0080\u0093326]. 1.2 Theory In order to maintain generality in the current study, the fluid in the simulations is considered Newtonian with constant properties. This means that properties such as density and thermal conductivity do not change with pressure, temperature, or humidity. 7 1.2.1 Heat and Mass Transport Under steady state conditions, the governing equations of the fluid are [9] Mass\t% \u00E2\u008B\u0085 ' = ) \t(1)\tMomentum\t3' \u00E2\u008B\u0085 %' = %4' \u00E2\u0088\u0092 %6 \t(2)\tEnergy\t(Assuming\tviscous\tdissipation\tis\tnegligible)\t3D6' \u00E2\u008B\u0085 %E = F%4E \t(3)\tHumidity\t' \u00E2\u008B\u0085 %I = J%4I \t(4)\tWhere \u00CF\u0081 is the density, cp is the specific heat, \u00CE\u00BB is the thermal conductivity, Y = c/\u00CF\u0081 is the mass fraction of water vapour (kg water/ kg mixture), and D is the diffusivity of water vapour in air. The behavior of the flow can be described by the Reynolds Number, which is the ratio of inertial to viscous forces. LM = 3'JNO \t(5)\tWhere \u00C2\u00B5 is the fluid viscosity and Dh = 4A/P is the hydraulic diameter of the flow passage, where A and P are the cross-section area and perimeter, respectively. The thermal behavior is 8 described by the Prandtl Number, which is the ratio of momentum diffusivity to thermal diffusivity QR = OD6F \t(6)\tSimilarly, the Schmidt Number is the ratio of momentum diffusivity to mass diffusivity TD = O3J \t(7)\tFor this study, Pr = 0.7 and Sc = 0.6, which are typical values for air. The fact that Pr>Sc means that thermal energy diffuses through the fluid more quickly than mass (humidity). 1.2.2 Membrane Transport As mentioned in section 1.1.2, anisotropic polymer membranes are effective ERV materials possessing high water vapour permeability and selectivity over contaminants. As shown in Figure 1.3, they are made up of a porous substrate, used for structural support, with dense coating on one side that provides selectivity of water vapour over contaminants. Transport theory in these two media has been described by Baker [12, pp. 15\u00E2\u0080\u009392]. The porous substrate has permanent pores through which fluid flows driven by a pressure gradient through the material, known as the pore-flow model. Some molecules are filtered out due to their size, but this effect is not significant in ERV membranes. The dense coating, on the other hand, has no well-defined pores. Instead, species dissolve into the membrane material and diffuse through it driven by a concentration gradient, known as the solution-diffusion model. The inability of some molecules to dissolve into the dense coating is the source of membrane selectivity in ERVs. 9 Pore-flow model: molecules flow through permeant pores. Some are filtered out due to their size. Solution-diffusion model: molecules dissolve into the membrane material. Some are separated out due to an inability to dissolve. Figure 1.6 \u00E2\u0080\u0093 Membrane permeation/separation modes [12, p. 16]. The flux of species i according to the pore-flow model is VW = \u00E2\u0088\u0092X\u00E2\u0080\u00B2DW Z6Z[ \t(8)\tWhere ci is the concentration of species i, and K\u00E2\u0080\u0099 is a constant depending on the membrane material and fluid. The flux for the solution-diffusion model is VW = \u00E2\u0088\u0092JW ZDWZ[ \t(9)\tWhere Di is the diffusion coefficient, which describes the ability of the species to adsorb to the membrane surface, diffuse through it, and desorb from the other side. For composite membranes, a similar relation is used V\"W = \u00E2\u0088\u0092QZDWZ[ = \u00E2\u0088\u0092Q DW(_) \u00E2\u0088\u0092 DW(`)a \t(10)\tWhere (s) and (f) refer to the sweep and feed sides of the membrane, respectively, l is the membrane thickness, and P is the membrane permeability, which is typically measured experimentally and depends on the membrane material, temperature, and humidity. 10 Sensible heat moves through the membrane according to the conduction equation c\" = \u00E2\u0088\u0092Fd ZEZ[ = \u00E2\u0088\u0092Fd E(_) \u00E2\u0088\u0092 E(`)a \t(11)\tThe membrane is more permeable to heat than to mass, this can be quantitatively seen through the ratios of membrane flux to air-side diffusion to given by e = Q/aJ/g\t \t(12)\th = Fd/aF/g \t(13)\tThese values depend on the membrane material but, typically, e~1, so the membrane mass transfer resistance is the same order of magnitude as the air-side diffusion resistance. On the hand, \u00CF\u0083~100, so the membrane heat transfer resistance is negligible compared to air-side diffusion [6], [13], [14]. 1.2.3 Performance Metrics Heat transfer performance can be characterized by the Nusselt number, which is the ratio of convective to conductive heat transfer klm = JNNF \t \t(14)\tWhere Dh is the hydraulic diameter of the channel and h is the heat transfer coefficient, defined for the air-side convection transfer as N = noo(Ep \u00E2\u0088\u0092 Eq)\t \t(15)\t11 Where q\u00E2\u0080\u0099\u00E2\u0080\u0099 and Tw are averaged over the periphery of the membrane. One can also define the overall heat transfer coefficient between the two channels as g = noo(Eq(`) \u00E2\u0088\u0092 Eq(_))\t \t(16)\tIn this form the heat transfer coefficient accounts for the air-side resistance of both channels as well as the membrane resistance. One can analogously define the local Sherwood number, which gives the ratio of convective to diffusive species (e.g. moisture) transfer \tTNm = JNrJ \t \t(17)\t\tr = Voo(Dp \u00E2\u0088\u0092 Dq)\t \t(18)\t\tX = Voo(Dq(`) \u00E2\u0088\u0092 Dq(_))\t \t(19)\tThe average Nusselt and Sherwood numbers are defined as kl = st klmZmt) \t \t(20)\tTN = st TNmZmt) \t \t(21)\tWhere the integrals are taken over the length of the channel. The overall recovery efficiency of the exchanger is given by uE = Ev(_) \u00E2\u0088\u0092 EW(_)EW(`) \u00E2\u0088\u0092 EW(_)\t \t(22)\tuD = Iv(_) \u00E2\u0088\u0092 IW(_)IW(`) \u00E2\u0088\u0092 IW(_)\t \t(23)\t12 Because the fluid has (by assumption) constant properties with equal flow rates on each side of the counter-flow exchanger, the bulk temperature and humidity difference is constant along the channel. This gives simplified equations for the LMTD method [9] n = gw\u00E2\u0088\u0086E\t \t(24)\tV = Xw3\u00E2\u0088\u0086I\t \t(25)\tWhere A is the smooth-channel membrane area in a periodic section. Combining these with the definitions of efficiency, it can be shown that uE = ss + dDgwE\t \t(26)\tuD = ss + d3XwE\t \t(27)\tWhere AT is the total smooth-channel membrane area over the entire channel length. The heat and mass transfer coefficients can be calculated in a single periodic section of the duct and used with the above equations to predict the effectiveness for the entire duct. The benefit of increasing heat and mass transfer must be weighed against a consequential increase in frictional losses. This can be expressed through the friction factor, defined locally as \t`m = sJN Z6/Zm3l \t \t(28)\tand globally as ` = st `mZmt) \t \t(29)\tThe power consumed to force air through the duct is then 13 Q`z{ = c\u00E2\u0088\u00866u`z{ = c(`tJN3l)u`z{\t \t(30)\tWhere Q is the volumetric flow rate through the duct and \u00CE\u00B7fan < 1 is the fan efficiency. 1.2.4 Periodic Boundary Conditions The hydrodynamically developing region is an important aspect of ERV performance. It is characterized by the section of channel downstream of the entrance where the flow fields change from homogeneous to a \u00E2\u0080\u0098developed\u00E2\u0080\u0099 distribution defined by the channel boundary conditions. Figure 1.7 shows a developing velocity profile in a 2D duct, in which the velocity field changes from uniform to a parabola due to the no-slip (v = 0) boundary condition at the walls. Similar developing profiles exist for the temperature and humidity fields. This entrance region is characterized by high values of Nu, Sh, and f values that decrease asymptotically to lower developed values. This region can take a up a large fraction of the channel length and account for a significant portion of the ERV performance [8, pp. 120\u00E2\u0080\u0093124]. Figure 1.7 - Hydrodynamically developing velocity profile Unfortunately, the developing region is so long that it can incur considerable computational expense to simulate. For this reason, and in order to highlight the effects of the mixing features, the channel entrance region is ignored in the current study. This means that the effectiveness predictions are lower than they would be in reality. However, because the developing region 14 would increase the effectiveness of both ribbed and smooth channels, the differences in performance between the two would be roughly the same. Future work could be done to investigate the magnitude of this effect in different circumstances. The entrance region is ignored by assuming periodic boundary conditions between the inlet and outlet. This is typically only valid when the boundary conditions are uniform temperature (concentration) or uniform wall heat (mass) flux. While this case is neither, the total heat (mass) flux is the same for each periodic section, so the analysis of Patankar et al. can be applied [15]. They found the velocity field and the dimensionless quantities | = E \u00E2\u0088\u0092 EpEq \u00E2\u0088\u0092 Ep \t(31)\t} = I \u00E2\u0088\u0092 IpIq \u00E2\u0088\u0092 Ip \t(32)\trepeat between the inlet and outlet. Where the wall values are averages over the cross-section periphery on the membrane. This gives the inlet conditions as E(W) = Ep(W) + (EqW \u00E2\u0088\u0092 Ep(W)) \tE(v) \u00E2\u0088\u0092 Ep(v)Eq(v) \u00E2\u0088\u0092 Ep(v) \t(33)\tI(W) = Ip(W) + (IqW \u00E2\u0088\u0092 Ip(W)) \tI(v) \u00E2\u0088\u0092 Ip(v)Iq(v) \u00E2\u0088\u0092 Ip(v) \t(34)\t'(W) = '(v) \t(35)\tWhere ~\u007F\u00C3\u0084 and \u00C3\u0085\u007F\u00C3\u0084 and the mass flow rate are simulation inputs and the other terms come from the previous simulation iteration. Each simulation was initialized with uniform velocity, 15 temperature, and humidity profiles, and then the above equations were used after each iteration until simulation convergence. 1.2.5 Geometric Enhancement As discussed in section 1.1.2, geometric features that promote mixing can be an attractive method of increasing ERV performance without requiring additional materials or increase the core footprint. Others have demonstrated significant performance boosts through such variations. Zhong et al. found that adding air deflectors and spreader plates to a cross-flow ERV can improve sensible and latent effectiveness by up to 17.4% and 7.8%, respectively [16]. Koester et al. investigated the performance of an ERV with netting installed between membrane layers. They found the new configuration had higher effectiveness and pressure drop, which overall could result in significant energy savings depending on the cost of energy and climate conditions [17]. While these studies demonstrate that certain mixing features can improve overall core performance, they did not investigate the local performance within the individual channels. A goal of the present study was to investigate how mixing features can alter local flow fields improve individual channel performance. Recovery effectiveness between a pair of channels can be described in terms of a resistance in series model. Within each channel there is an air-side resistance due to the time it takes for heat/mass to diffuse to the membrane surface. The membrane itself also has a resistance as the heat/mass must diffuse through the membrane. Finally, there is another air-side resistance in the opposite channel as the heat/mass must diffuse away from the membrane surface and move along the channel. 16 Figure 1.8 - Resistance in series model These resistances add together to oppose the flux from one channel to another, much the same was as electrical resistors oppose the flow of current in a circuit. It is the work of material scientists to try and create the most permeable membrane (ie. lowest resistance) without sacrificing selectivity of the membrane. This work, on the other hand, aims to improve performance by lowering the air-side resistance. This is done by mixing the flow in such a way to reduce the distance for diffusion towards the membrane. Many different mixing features have been investigated to enhance sensible heat transfer in ducts. Figure 1.9 shows some of the most popular features: delta winglets [18], [19], wire-coil inserts [20], and angled ribs. A summary of studies investigating angled rib features is in Table 1.1. lH HRa Rm Ra17 Delta winglets Coil insert Angled rib Figure 1.9 - Common enhancement geometries Table 1.1 Previous studies investigating heat transfer enhancement in ducts with angled ribs Study\t Regime\t Description\tZheng\tet\tal.\t(2016)\t-\tNumerical\tinvestigations\tof\tthe\tthermal-hydraulic\tperformance\tin\ta\trib-grooved\theat\texchanger\ttube\tbased\ton\tentropy\tgeneration\tanalysis\t\tTurbulent\t Studied\ta\tcircular\ttube\twith\tdiscrete\tribs\tcovering\tthe\tinside\tsurface\tand\tconstant\twall\ttemperature.\tNusselt\tnumber\tincreased\tby\ta\tfactor\t1.58-2.46\twhile\tfriction\tfactor\tincreased\tby\t1.82-5.03.\tThis\tresulted\tin\tan\toverall\tincrease\tin\tthermal\tperformance\tof\t1.19-1.68.\t18 Olsson\tand\tSund\u00C3\u00A9n\t(1997)\t-\t\tExperimental\tstudy\tof\tflow\tand\theat\ttransfer\tin\trib-roughened\trectangular\tchannels\t[21]\tLaminar\tand\tTurbulent\tExperimental\tstudy\tof\tribs\ton\topposite\tsides\tof\ta\trectangular\tchannel\tat\tconstant\twall\ttemperature.\tLooked\tat\tparallel\tribs,\tcross\tribs,\tcross\tV-ribs,\tparallel\tV-ribs,\tand\tmultiple\tV-ribs.\tThey\tfound\tall\tcases\tincreased\tthe\theat\ttransfer\tcoefficient\tfor\ta\tgiven\tpumping\tpower\tper\theat\ttransfer\tarea.\tTanda\tand\tAbram\t(2009)\t-\tForced\tconvection\tand\theat\ttransfer\tin\tchannels\twith\trib\tturbulators\tinclined\tat\t45\tdeg\t[22]\tTurbulent\t Experimental\tstudy\tof\tangled\tribs\tthrough\ta\trectangular\tchannel\twith\tassumed\tconstant\twall\theat\tflux.\tNusselt\tnumber\tincrease\tby\tup\tto\t50%\tfor\tfixed\tpumping\tpower.\tKim\tet\tal.\t(2009)\t-\tOptimal\tdesign\tof\tangled\trib\tturbulators\tin\ta\tcooling\tchannel\t[23]\tTurbulent\t Numerical\tstudy\tof\ta\tsquare\tchannel\twith\tribs\ton\topposite\twalls\tand\tconstant\twall\ttemperature.\tImproved\tthermal\tperformance\tby\tup\tto\t60%\tby\toptimizing\trib\tangle\tof\tattack\tand\theight/spacing\tratio\tNonino\tand\tComini\t(2002)\t\u00E2\u0080\u0093\tConvective\theat\ttransfer\tin\tribbed,\tsquare\tchannels\t[24]\tLaminar\t Numerical\tstudy\tof\ta\tsquare\tchannel\twith\tangled\tribs\tand\tuniform\twall\ttemperature\ton\tsurfaces.\tAngled\tribs\tincrease\tNusselt\tnumber\tby\tup\tto\t20%\tmore\tthan\tthe\tcorresponding\tincrease\tin\tfriction\tfactor\tSaha\t(2010)\t-\tThermal\tand\tfriction\tcharacteristics\tof\tlaminar\tflow\tthrough\trectangular\tand\tsquare\tducts\twith\ttransvers\tribs\tand\twire\tcoil\tinserts\t[20]\tLaminar\t Experimental\tstudy\tcombining\tangled\tribs\tand\tcoil\tinserts\tin\trectangular\tand\tcircular\tchannels.\tFound\tthat,\tfor\tequal\tpumping\tpower,\tthe\tcombination\tof\tribs\tand\twires\tcan\tincrease\theat\trecovery\tby\tup\tto\t50%\tcompared\tto\tribs\tor\twires\talone.\t\t These studies all observe a significant heat transfer enhancement with ribs and, provided the various design parameters are tuned correctly, the increased heat transfer outweighs the corresponding increase in frictional pressure drop. This makes angled ribs an attractive feature for ERVs. However, the data from these studies cannot be applied to ERVs directly because, for one, the analysis was restricted to either constant wall temperature or constant wall heat flux boundary conditions. The boundary conditions in an ERV are neither constant wall temperature 19 (concentration) nor constant wall heat (mass) flux. Rather, they are defined by the coupling between adjacent flow passages through the membrane. Another issue with applying this data to ERVs is that mass transfer was not investigated. Because the fundamental equations of heat and mass transfer are very similar, it is often possible to convert heat transfer data to mass transfer data with the empirical Chilton-Colburn j-factor analogy, which states N3D6 (QR)s/\u00C3\u0087 = r(TD)s/\u00C3\u0087 \t(36)\tUnfortunately, in recirculating flows, such as those introduced by internal channel ribs, the Chilton-Colburn j-factor analogy breaks down [8, p. 16], [9, pp. 538\u00E2\u0080\u0093539]. Therefore, a full counter-flow exchanger model considering both heat and mass transfer is required. 1.3 Current Study Objectives This study investigates full conjugate heat and mass transfer in an ERV with angled ribs in the channels. The model geometry consists of two rectangular ducts in a counter-flow arrangement, separated by a porous polymer membrane. In order to compare the effect of ribs to simply changing the channel dimensions or flow rate, channels are considered for various aspect ratios and flow rates, with and without ribs. The ribs are placed on the membrane surface and are themselves permeable to heat and moisture. 20 Chapter 2: Methods Instead of investigating the performance of a full ERV directly, this study investigates the performance of a relatively short pair of channels in a counter-flow arrangement. This is done to reduce computational and experimental complexity. The results from this short section are then used to estimate the performance of full length channels as well as a cross-flow ERV. Validation experiments are performed to investigate the mass transfer and pressure drop predictions of the simulations. 2.1 Simulation 2.1.1 Problem Description The geometry of the counter-flow ERV model is shown in Figure 2.1. Two square ducts are in a counter-flow arrangement separated by a polymer membrane that is permeable to water vapour. All other walls are adiabatic. Feed air (building exhaust) enters at point 1 and leaves through point 2. Sweep air (building intake) enters point 3 and leaves through point 4. The inlets at 1 and 3 are modeled as mass-flow inlets (wherein the inlet pressure is varied to give a desired mass flow) and outlets 2 and 4 are pressure outlets at a gauge pressure of zero. Periodic boundary conditions are applied between the inlet and outlet of each duct as described in Section 1.2.4. The user define function (UDF) for implementing the periodic boundaries is included in Appendix A.2. The channel width is constant for all cases. The height is varied to give an aspect ratio (height/width) of 0.6, 1, or 1.4. Simulations are performed with and without ribs with ribs 21 spaced 16 channel widths apart. The ribs themselves are 1/3 channel height tall and 1/3 of the channel width wide. The fluid is considered Newtonian with constant properties (Pr = 0.7 and Sc = 0.6). Three different flow rates are considered, yielding a Reynolds number in the range of 200\u00E2\u0089\u00A4Re\u00E2\u0089\u00A4900, which is typical for a commercial-scale ERV. This range of Re was studied for a similar geometry [24] and found to be in the laminar regime, so laminar flow is assumed for all cases in this study. (Planform View) (Elevation View) Figure 2.1 \u00E2\u0080\u0093 Schematic of the simulation domain 22 2.1.2 Simulation Parameters Simulations were run on the finite-volume solver, Fluent (version 16.1). Meshing was done on Ansys Meshing (version 16.1). The flow was solved in the 3d laminar, pressure-based solver with a 2nd order upwind discretization scheme. SIMPLE pressure-velocity coupling was used with a least squares cell-based gradient calculation. 2.1.3 Membrane Modelling Approach Fluent has built in functionally to implement equation (11) for heat transfer across the membrane, however, there is no such functionally to directly implement equation (10) for mass transfer. Instead, a custom User-Defined Function (UDF) must be written to implement equation (10). Among the many uses of UDFs is they allow the user to script custom profiles and source terms over mesh elements during the simulation [25]. Previous studies [26]\u00E2\u0080\u0093[28] have used UDFs to define custom sink/source terms on the mesh elements adjacent to the membrane surface of the form ZdWZ\u00C3\u0089 = \u00E2\u0088\u0092Q Dp\u00C3\u0091\u00E2\u0088\u0086(_) \u00E2\u0088\u0092 Dp\u00C3\u0091\u00E2\u0088\u0086(`)a \u00C3\u0096DMaawDMaa \t(37)\tWhere c\u00C3\u009C\u00C3\u0091\u00C3\u00A1 is the humidity in the cell adjacent to the membrane surface and Vcell and Acell are the volume and area of the cell on which the source term is implemented. The drawback of this method is that when attempting to implement it in this study, it was found to depend heavily on the size of mesh elements adjacent to the membrane surface. Instead, an alternate method was used in which the membrane surface concentration was varied in order to get a desired concentration gradient. This method has been used in at least one other 23 study [29] and can be described as follows. Mass-flux through the membrane into the sweep stream can be described by equation (10) and the air-side flux into the sweep stream is given by \tV\" = \u00E2\u0088\u0092J%D _ = \u00E2\u0088\u0092J(Dp_ \u00E2\u0088\u0092 Dp\u00C3\u0091\u00E2\u0088\u0086_ )/\u00E2\u0088\u0086\u00C3\u00A0 \t(38)\tWhere D is the diffusivity of water-vapour in air. Rearranging the above equation gives the membrane surface concentration on the sweep side as Dp_ = Dp\u00C3\u0091\u00E2\u0088\u0086_ + Q/aJ/\u00E2\u0088\u0086\u00C3\u00A0 (Dp_ \u00E2\u0088\u0092 Dp` ) \t(39)\tAfter each simulation iteration, the membrane surface concentration was varied according to equation (39) to produce the membrane flux predicted by equation (10). The terms on the right hand side of equation (39) were the data from the previous iteration. The UDF for implementing this is included in Appendix A.1. 2.1.4 Convergence Criteria Simulations were considered converged after all residuals dropped below 1e-6 and the sweep side outlet humidity mass fraction, outlet temperature, and inlet pressure remained unchanged to the 5th significant figure for 25 iterations. For smooth channels this was achieved within 1000 iterations. For ribbed channels up to 3000 iterations were required. 2.1.5 Meshing and Mesh Validation Meshing was performed in Ansys Meshing (v16.1). The mesh is shown in and Figure 2.3. It is a pure hexahedral structured mesh with a higher concentration of cells near the walls, membrane and ribs. In order to maintain a structured mesh with refinement near the ribs, block meshing was used with separate blocks around the ribs. Non-conformal interfaces were used at the interfaces 24 between the rib blocks and the rest of the mesh. The mesh for the smooth channel (no ribs) was a grid of 28\u00C3\u009728\u00C3\u0097180 (H\u00C3\u0097W\u00C3\u0097L) mesh elements with finer elements near the walls. With this mesh size a typical simulation run would take roughly six hours to complete. Richardson extrapolation was used to find an error estimate on the final result as described in [30]. In short, the numerical solution data is assumed to have a series expansion in the grid spacing, h ` = ` M[zD\u00C3\u0089 + \u00C3\u00A2sN + \u00C3\u00A24N4 + \u00E2\u008B\u00AF \t(40)\tThe error estimator then takes the form \u00C3\u00A3 = `4 \u00E2\u0088\u0092 `ss \u00E2\u0088\u0092 R6 \t(41)\tWhere r = h2/h1 > 1, p is the order of accuracy of the algorithm, and fi is the solution obtained for from using grid spacing hi. Usually the value of p is not known a priori, however it can be calculated by using three successive mesh refinements with equal values of r. The order of convergence is then 6 = a{ `\u00C3\u0087 \u00E2\u0088\u0092 `4`4 \u00E2\u0088\u0092 `s /a{(R) \t(42)\t This method was applied to the current work, with three different meshes where the grid spacing was successively reduced by 20% (ie. r = 1.25). The resulting variation in Nu, Sh, and fRe is shown in Figure 2.4. Equations (41) and (42) were used to find the error estimator for each value. The error estimator is usually multiplied by a safety factor, which, for high quality structured 25 meshes such as this one, can be taken as 1.25. This gave a final error estimate of 0.5% for Sh, 1% for Nu, and 2% for f. As a further validation, the smooth channel simulations were run with the membrane replaced with a constant temperature boundary condition. This gave Nu = 2.414 and fRe = 14.168, which agree within 1% of the values of Nu = 2.437 and fRe = 14.227 reported by [31]. Figure 2.2 \u00E2\u0080\u0093 Isometric view of the mesh, mesh spacing has been doubled in each dimension for clarity 26 (top) (side) (bottom) (left end) (right end) (side - detail) Figure 2.3 \u00E2\u0080\u0093 Various views of the mesh. Mesh spacing has been doubled in each dimension for clarity27 Figure 2.4 \u00E2\u0080\u0093 Mesh sensitivity study 2.1.6 Post Processing Post-processing was required to extract local air-side Nu, Sh, and f values. This was accomplished with a UDF, which is included in Appendix A.1. In brief, the simulation domain was divided up into small, equal-sized, sections along the length of the channel (20 for Nu and Sh, 60 for f). Average bulk and wall values were calculated on each section and used to calculate the local coefficients according to equations (15), (18) and (28). The average air-side coefficients for the entire channel were then calculated by averaging the coefficients in all the bins. 28 2.2 Experimental The experimental portion of this work was used to validate the results from the simulations. A test station was designed and built to measure water vapour flux through the membrane under isothermal conditions in a module that closely resembled the simulation domain in Figure 2.1. Tests were done for various channel aspect ratios with and without ribs and compared to the predictions of the simulation data. 2.2.1 Water Vapour Transport Station A diagram of the station used for testing in shown in Figure 2.5. Air enters the system after passing through a silica gel dehumidifier. One stream is at a controlled humidity (feed) and one the other stays dry (sweep). They are passed over opposite sides of a membrane inside a test module. Analog voltage humidity sensors (Measurement Specialties, HTM2500) take measurements at RH1, RH2, and RH3 while a more accurate chilled mirror sensor (Edgetech DewMaster) measures humidity at RH4. Mass-flow controllers (Alicat MSC-10SLPM) at MFC-1 and MFC-2 dictate the flow rate of each airstream. All testing takes place in an oven which maintains isothermal conditions. The pressure drop along the sweep side of the module is measured between pressure taps with a digital pressure transducer (Setra Model 267). Air is humidified using an automotive fuel cell humidifier (dPoint Technologies PX1). Water is circulated on one side of the humidifier and air passes through another side. Water moves to the airstream through a permeable membrane. Air leaves the humidifier fully saturated provided the water is at a high enough temperature, which is insured by preheating the water in a reservoir. 29 Figure 2.5 \u00E2\u0080\u0093 Experimental test station The recovery effectiveness of the module is !\" = \t %& '( \u00E2\u0088\u0092 '*%%+, '- \u00E2\u0088\u0092 '* \t \t(43)\tHere 23,5,6,7 are the humidity ratios at the feed inlet, feed outlet, sweep inlet, and sweep outlet, respectively and 89,: is the mass flow rate of dry air into the feed and sw eep streams, respectively and where 8;<= is the lower of the two. 2.2.2 Testing Module The membrane flow module is shown in Figure 2.6. It consists of four aluminum plates fastened together with bolts. Membrane material is placed between the two middle plates. Gaskets between each set of plates to create a seal. Feed and sweep streams enter from opposite sides and 30 opposite ends of the module to create a counter-flow arrangement as shown in Figure 2.7. Pressure taps on the sweep side of the module allow measuring the pressure drop through the passage. Table 2.1 gives the dimensions of the slots in the aluminum housing. The slot height represents the height, including gasket material, after the module has been fastened together (which compresses the gaskets slightly). Knowing these dimensions, it was possible to 3D print flow channel inserts to replicate the geometries studied in the simulations. The flow passages were created with a Polyjet 3D printer (Stratasys Object 30) using VeroBlue printing material. The inserts are discussed in section 3.3. Figure 2.6 \u00E2\u0080\u0093 Aluminum housing of testing module 31 Figure 2.7 \u00E2\u0080\u0093 Flow configuration in the membrane housing module Table 2.1 Dimensions of the housing slot Parameter Value (mm) Housing slot length 300 Housing slot width 10 Housing slot height 6.1 2.2.3 Membrane Forming The most direct method of validating the simulations in this work is to form rib features into the membrane surface. A two-part aluminum mould was used to emboss rib features into membrane samples. The mould pattern is shown in Figure 2.8. A pressure of 300 psi over the surface of the mould for 10 seconds at room temperature was adequate to create permanent features. 32 Figure 2.8 - Membrane forming template. Dimension are in inches. 33 Chapter 3: Results and Discussion 3.1 Simulations 3.1.1 Membrane Ribs Figure 3.1 shows the transverse velocity vectors 1, 5, and 8 step widths downstream of the rib for three different aspect ratios. These transverse flow patterns move the flow from the top of the channel towards the membrane and reduce the temperature (humidity) heterogeneity, resulting in a higher membrane flux. The recirculation pattern starts out closest to the membrane and moves upward towards the channel centre as the flow moves farther from the rib. As noted by [24], if the top of the channel is also transferring heat (mass), it is better to stagger the ribs on the top and bottom of the channel as this tends to move fluid from the centre of the channel to the sides. The main vortex is largest and most intense for the AR = 1; though, it occupies the largest fraction of the channel cross-section for AR = 0.6. Overall, the vortex appears least stable for the AR = 1.4 case, indicating that ribs are not as beneficial for aspect ratios higher than 1. There is a smaller, secondary vortex on the opposite size of the main one (far right corner) that may also aid the channel performance. Figure 3.2 shows how the qualitative behavior of the flow changes with Re. It shows the same types of velocity vectors as Figure 3.1, but with AR = 1 for all cases and three different values of Re. All three cases display the same main vortex a pattern downstream of the rib, however the vortex intensity increases with increasing Re. 34 Figure 3.3 shows the fluid speed from the elevation perspective (see Figure 2.1) for the AR = 1, Re = 468 case. The fluid reaches its maximum speed in the constriction created by the rib. There is a stagnant zone directly upstream of the rib (which is generally poor for heat and mass transfer) and a recirculating zone downstream of the rib. This highlights the importance of picking the correct rib geometry for the channel and flow rate of interest. In this particular case, the recirculation regions are an order of magnitude longer than the stagnant regions, suggesting that overall the effectiveness will increase. If the rib spacing were to be reduced to the point where the rib spacing was similar to the rib height (or width), the stagnant regions could become large enough that the channel effectiveness would actually decrease below the smooth channel value. Figure 3.4 shows contour plots of humidity and temperature on the membrane surface looking down on the feed side. There are large changes in both variables downstream of the rib. The large mass-transfer resistance of the membrane (relative to heat transfer resistance) can be seen in these plots as the humidity is highly asymmetric between the left and the right compared to the temperature. Note the periodic boundaries in both Figure 3.3 and Figure 3.4. 35 (a) (b) (c) Figure 3.1 - Transverse velocity fields 1, 5, and 8 step widths downstream of the rib for the middle of the three flow rates. (a) AR = 0.6, Re = 390. (b) AR = 1, Re = 468. (c) AR = 1.4, Re = 585. 36 (a) (b) (c) Figure 3.2 - Transverse velocity fields 1, 5, and 8 step widths downstream of the rib for AR = 1. (a) Re =703. (b) Re = 468. (c) Re = 234. 37 Figure 3.3 \u00E2\u0080\u0093 Speed contour plot for the AR = 1, Re = 468 case Figure 3.4 \u00E2\u0080\u0093 Contour plots of humidity (top) and temperature (bottom) on the membrane surface for the AR = 1, Re = 468 case. Looking down to on the feed side. Red (blue) indicates highest (lowest) humidity and temperature. The performance of the ribbed channels can be understood in terms of an increase in heat and mass transfer compared to the corresponding increase in pressure drop. These can be interpreted non-dimensionally in terms of Nu, Sh, and fRe. These performance data are presented locally for the feed side channel in Figure 3.5 for AR = 1 and Re = 468 (Figure 3.1(b) case). The friction value (top) mostly has the same value for the smooth case, except in the immediate vicinity of the ribs. The rib at \u00C2\u00BC of the channel length protrudes into the main flow and has the largest effect on fRe, while the second rib protrudes into the opposite channel and has a lesser effect. fRe has a shear component, caused by the shear stress tangential to the main flow, and a normal stress, caused by the components of the rib surface perpendicular to the main flow. A large part of the additional pressure drop in the ribbed channels is caused by this normal component. A notable feature of the friction plot is the 38 negative spikes directly downstream of the rib. These are caused by the flow reversing direction in the rib wake. The shear component of the friction factor also increases due to the flow speeding up in the constriction created by the rib. The local Nusselt and Sherwood numbers (bottom) behave differently that the friction coefficient. Both values stay above the smooth channel value over the entire channel length, with the largest increase in the 50% of the channel downstream of the first, protruding rib. There is then a dip in performance as the flow passes over the negative rib, which contains a pocket of stagnant air. Nu and Sh are increased by the flow recirculation shown in Figure 3.1 and, to a lesser extent, by the added membrane surface area on the rib. Nu tends to be higher than Sh due to the higher value of Pr than Sc (ie. the thermal diffusivity is larger than the mass diffusivity). Figure 3.5 \u00E2\u0080\u0093 Local performance data along the channel for AR = 1 and Re = 468. 39 Figure 3.6 shows the overall performance data for all cases studied. For smooth channels, Nu and Sh are independent of Re and closely approximate the case of constant wall heat flux (Nu, Sh = 2.712 for AR = 1 [31]). Sh and Nu are equal for smooth channels of the same aspect ratio, which is a consequence of the heat and mass transfer analogy being fully satisfied. As the aspect ratio decreases, Nu and Sh increase because the fluid is constrained to be closer to the membrane and the distance required for diffusion to the membrane is shorter. On the other hand, fRe is at a minimum for AR = 1 and increases with either an increase or decrease in aspect ratio. The recirculating flows in the ribbed channels break the heat and mass transfer analogy and Sh, Nu and fRe all increase with Re. Nu and Sh follow a similar pattern as each other but Nu is always slightly higher because Pr is higher than Sc. Ribbed channels with AR = 1 tend to produce the highest Nu and Sh. This case is likely most favorable because it allows the flow to circulate more freely around the channel axis (i.e., vortices of aspect ratio one are more stable than higher aspect ratio vortical structures) as shown in Figure 3.1. fRe follows a different pattern and consistently increases with AR, partially because the rib height also increases with AR. When comparing the ribbed channels to the smooth channels of the same aspect ratio, the increase in Sh and Nu is larger than the corresponding increase in fRe. The relationship is most favorable in the AR = 1 ribbed case, shown in Figure 3.6(d). In the range studied, increasing Re tends to increase the benefit gained from ribs. 40 The benefits from ribs shown here could be improved by optimizing the geometric parameters of the ribs. Namely, the rib height, width, spacing, and angle, and shape. For a specific ERV use-case, the optimum values would depend on several factors, such as the flow rate, footprint, and pressure drop requirements. Given that each simulation takes roughly six hours to complete, and using five values for each parameter, an optimization study is estimated to take one month to complete, allowing for time to manually mesh certain configurations. Once the effect of these parameters is better understood, any future optimization studies would be significantly shorter. 41 (a) (b) (c) (d) Figure 3.6 \u00E2\u0080\u0093 Performance data of the ribbed and smooth channels for three different flow rates. (a) Nusselt number. (b) Sherwood number. (c) Friction factor times Reynolds number. (d) Relative performance of ribbed and smooth channels for the AR=1 case. 3.1.2 Ribs on Walls Placing ribs on the top and bottom walls of the modules could be a simpler way of achieving improvement. This geometry is obtained by moving the rib from the membrane surface to the opposite wall, shown in Figure 3.7. The ribs are adiabatic so do not increase the membrane surface area compared to the smooth case, but the recirculating flow pattern is similar. 42 Figure 3.7 - Diagram of ribs on walls Figure 3.8 \u00E2\u0080\u0093 Performance of smooth, membrane ribs, and wall ribs For lower aspect ratio (height), wall ribs perform similarly to membrane ribs because, as shown in Figure 3.1(a), the downstream vortex convers most of the whole channel. For higher aspect ratios (heights) the vortex is concentrated on the same side as the rib, as shown in Figure 3.1(c). 43 It is better for performance to have recirculation closer to the membrane surface so membrane ribs perform better in these cases. As expected, the pressure drop for wall ribs and membrane ribs is identical, because the flow obstruction is the same in both cases. Because membrane ribs incur the same pressure drop but improved recovery, they are the focus of this work. 3.2 Effectiveness Predictions 3.2.1 Counter-Flow Channel Figure 3.9 \u00E2\u0080\u0093 Performance data of a 1m long counter-flow channel The data in Figure 3.6 can be used with equations (24) and (25) to give the effectiveness of a counter-flow channel of any given length. Figure 3.9 shows these data for a 1m long counter-flow channel with a 2.5mm width and a 1 LPM flow rate (the middle of the three flow rates 44 considered). This is typical for channels in a commercial-scale ERV. The data are presented for channels with and without ribs for various aspect ratios. For a given pressure drop, the ribbed channel always has a higher recovery effectiveness. This means that the recovery can be increased for an equal pressure drop by adding ribs and increasing the channel height. The sensible recovery in Figure 3.9 is consistently higher than the latent recovery. This is primarily because the membrane heat transfer resistance is lower than its mass transfer resistance but also because the air-side value of k increases more than h with ribs. This means that the rib enhancement is more apparent for heat transfer. Figure 3.10 shows the recovered power for a single channel in AHRI summer conditions (Feed stream: 35.0\u00C2\u00B0C dry bulb and 25.6\u00C2\u00B0C wet bulb, Sweep stream: 23.9\u00C2\u00B0C dry bulb and 17.2\u00C2\u00B0C wet bulb) [32]. The fan power was calculated with a fan efficiency of 0.56. It is notable that the recovered power is 2-3 orders of magnitude larger than the fan power, which is indicative of the overall energy effectiveness of the ERV. 45 Figure 3.10 - Recovered power and fan power for AHRI standard summer operating conditions. 3.2.2 Cross-Flow Cross-flow cores are the most popular arrangement due to their relative ease of construction, even though they are not as effective as a counter-flow arrangement. A typical cross flow arrangement is shown in Figure 3.11. It is constructed by stacking layers in alternate orthogonal directions separated my membrane material. Each layer is supported by a spacer that separates the flow into multiple parallel channels. The end result is that there is no heat or mass transfer on the side walls of each channel, but there is transfer through the top and bottom membrane surfaces. 46 Figure 3.11 - Cross-flow ERV configuration The boundary conditions in a cross-flow arrangement are different than the counter-flow arrangement presented in Section 3.1.1. For one, there is flux through both the top and bottom of the channel (the other walls remain adiabatic). The top and bottom walls can be approximated as constant wall temperature (humidity). To see why, consider that the sweep side channels are all identical and so will all have the same temperature (humidity) distribution for any given position along the y-axis (neglecting edge effects). Now consider a channel on the feed side which flows along the x-axis at a constant value of y. Because the channels above and below are all sweep-side channels at a set value of y, they will all have the same temperature (humidity) distribution, making these fields approximately constant on the membrane surface. The same argument can be made for any channel on the sweep side. The simulation presented thus far can be modified to consider this case by considering only one of the two channels and making the top and bottom 47 walls (including ribs) a constant wall temperature (humidity) boundary condition. Note that in this configuration every second layer of membrane has ribbed features while the others are flat. Figure 3.12\u00E2\u0080\u0093 Cross flow core simulation domain Because the channel shape and fluid properties are the same as the counter-flow arrangement, the local values of f are the same as that case. On the other hand, the local air-side Nu and Sh values are different and are shown in Figure 3.13. These values behave qualitatively similar to the counter-flow case, in that there is a positive spike downstream of the membrane and then a gradual return to the smooth-channel values. Also as in the counter-flow case, there is an overall increase in Nu and Sh with increasing Re (which corresponds to increasing flow rate in this case). However, because the smooth-channel performance is higher in this arrangement \u00E2\u0080\u0093 due to both the top and bottom walls transferring heat and mass \u00E2\u0080\u0093 the performance increase from ribs is not as high as it is for the counter-flow case. 48 Figure 3.13\u00E2\u0080\u0093 Cross flow local Nu and Sh values for various values of Re To find the predicted performance of a cross-flow core, the local values of Nu, Sh, and fRe were used as inputs to a MATLAB model which has been described in [33]. For this section of the work the core geometry was made up of 60 layers with 64 channels per layer. Each layer was of 160mmx160mm. The results are shown in Figure 3.14. As was case with the counter-flow predictions, the ribbed channels consistently have higher performance than the smooth channels. The solid lines give the performance of AR = 1 channels for three different flow rates. The dashed lines show the change in smooth channel performance as the channel height is decreased for a constant flow rate. For a given flow rate and pressure drop, the ribbed channel always has a higher effectiveness than the smooth channel. In the range of Re studied, the sensible improvement is between 1.1% to 6.7% and the latent improvement is between 0.6% to 49 2.7%. The sensible effectiveness improves more than the latent effectiveness, largely because the membrane mass transfer resistance is so high that the improvement in air-side resistance is less significant. Another reason for this is that Nu increases more than Sh with ribs, as shown in Figure 3.13. A final notable feature of these plots is that as the flow rate (and Re) increases, the overall effectiveness decreases for both smooth and ribbed cases because the residence time of an air parcel in the channel decreases. Despite this, the difference in effectiveness between ribbed and smooth channels increases because of the larger Nu and Sh in the ribbed channels. Physically, the larger Re increases the mixing in the ribbed channels and partially compensates for the decreased residence time. By assuming AHRI Summer operating conditions (Feed stream: 35.0\u00C2\u00B0C dry bulb and 25.6\u00C2\u00B0C wet bulb, Sweep stream: 23.9\u00C2\u00B0C dry bulb and 17.2\u00C2\u00B0C wet bulb [32]), the sensible and latent effectiveness can be converted to total recovered power, shown in Figure 3.15. Here the recovered power increases with increasing flow rate because, even though the effectiveness is decreasing, the total available heat moving through the ERV is larger. Again, it can be seen that ribbed channels can significantly increase recovery for an equal pressure drop compared to smooth channels. For the Re = 703 case, the ribbed core recovers nearly 100W more than the smooth channel with an equal pressure drop. 50 (a) (b) Figure 3.14. - Variation of the effectiveness of enthalpy exchangers with smooth and ribbed channels for three different operating flowrates. (a) Sensible effectiveness. (b) Latent effectiveness. Dashed lines indicate effectiveness variation for smooth channels with variable AR of 1.4, 1.2, 1, and 0.8 at constant flowrate.\t\tFigure 3.15. - Total recovered power of enthalpy exchangers with smooth and ribbed channels for three different operating flowrates in AHRI Summer (cooling) operating conditions. Dashed lines indicate effectiveness variation for smooth channels with variable AR of 1.4, 1.2, 1, and 0.8 from left to right.\t 234\t468\t703\t0.8\t1\t1.2\t1.4\t20%\t30%\t40%\t50%\t60%\t70%\t80%\t0\t 20\t 40\t 60\t 80\t 100\t 120\tSensible\tEffec,veness\t(%)\t\u00CE\u0094P\t(Pa)\tRibbed\tSmooth\t-\tVariable\tQ\tSmooth\t-\tVariable\tAR\t234\t468\t703\t0.8\t1\t1.2\t1.4\t20%\t30%\t40%\t50%\t60%\t70%\t80%\t0\t 20\t 40\t 60\t 80\t 100\t 120\tLatent\tEffec*veness\t(%)\t\u00CE\u0094P\t(Pa)\tRibbed\tSmooth\t-\tVariable\tQ\tSmooth\t-\t\tVariable\tAR\t234\t468\t703\t200\t400\t600\t800\t1000\t0\t 20\t 40\t 60\t 80\t 100\t 120\tRecovered\tPower\t(W)\t\u00CE\u0094P\t(Pa)\tRibbed\tSmooth\t-\tVariable\tQ\tSmooth\t-\tVariable\tAR\t51 3.3 Experimental Validation The CFD model was validated against experiments conducted in the test stand and counter-flow module described in section 2.2. Tests were performed isothermally at 50\u00C2\u00B0C with the feed stream at 50%RH (40.4g/kg) and the sweep stream kept dry (<0.3%RH, 0.2g/kg). Tests were run with a flow of 2.15 SLPM on the feed and sweep side. Because the sweep inlet was dehumidified (x3 = 0) and the flow rates were equal on each stream, the recovery effectiveness simplifies to !\" = \t%&%'\t \t(44)\tThe flow passages were 3D printed to replicate the configurations from the simulations, that is, smooth channels and ribbed channels of various aspect ratios. Each insert consists of a pair of parallel channels so that the overall flow rate could be increased to a nominal level for the mass flow controllers (1-3 SLPM) and to increase the amount of active membrane area. Two methods were considered for creating the ribbed features. In the first, the ribs were 3D printed as extensions of the channels. This method has the advantage of being able to create rectangular angled rib features with a high degree of accuracy, however the ribs cannot be permeable extensions of the membrane as they were considered in the simulations. Instead, they are impermeable ribs on top of the membrane. Figure 3.16 shows an example of one of these ribbed channels. The channel shown is for the AR = 1 case and has 2.5mmx2.5mm channel dimensions. Another method of creating the rib features is to form them into the membrane surface, as was considered in the simulations. Because a rectangular rib feature would tear the membrane on 52 forming, a triangular rib feature was selected instead. Also, the ribs were at an angle of 30 degrees with respect to the main flow direction as this geometry left less space at the rib edges for the flow to go around the rib. These modifications were added to the simulation so they could be compared directly to the experiments. Ribs were formed with a mould as described in section 2.2.3. The final formed rib features are shown in Figure 3.17. Figure 3.16 \u00E2\u0080\u00933D-Printed Flow channel insert. Case shown is 2.5mmx2.5mm channel with plastic ribs near the membrane surface. Isometric view (top) and end view (bottom). 53 Figure 3.17- Rib features formed into the membrane Each channel insert was 289mm long with a 62mm entrance length on each leaving 165mm length for active transfer across the membrane. The entrance regions were covered in impermeable tape so that no humidity transport could occur across the membrane. They were required to create a known velocity distribution at the start of the active membrane area. The length, Le, required for a fully developed velocity distribution in the laminar regime is given by [9, p. 180] \"+/-. = \t/. /1213+ \t(45)\tThe velocity profile for a rectangular channel has the analytical solution [34] 5 = \t\u00E2\u0088\u0092 '78'9:;< '=< (\u00E2\u0088\u0092')(=>')/: ' \u00E2\u0088\u0092 8?@.(=;A:9 )8?@.(=;B:9 ) 8?@ =;%:9C=D',<,\u00E2\u0080\u00A6 \t(46)\t54 5H = \t\u00E2\u0088\u0092 8'9:< ' \u00E2\u0088\u0092 'I:;1 9B '=1 J9=. =;B:9C=D',<,\u00E2\u0080\u00A6 \t(47)\tWhere a and b are \u00C2\u00BD the channel width and height (AR = 2b/2a), and the coordinate system origin is at the cross-section centre, as shown in Figure 3.18. The UDF for implementing this profile is included in Appendix A.3. Figure 3.18 - Rectangular channel cross-section and dimensions Two channel aspect ratios were considered for experiments, 2.5mmx2.5mm (AR=1) and 1.5mmx2.5mm (AR = 0.6). However, because the 3D printer used (Stratasys Object 30) had a reported accuracy of 0.1mm, the channel cross-section dimensions were measured with gauge blocks to find the actual channel dimensions. It was possible to measure the dimension to within 0.01mm this way. In order to make the simulations and experiments comparable, the simulations geometry was matched to the dimensions in Table 3.1. The fluid properties, listed in Table 3.3, were also used to match the test conditions of 50\u00C2\u00B0C and atmospheric pressure. The velocity profile at the inlets 55 was assumed to be fully developed and given by equation (46). The humidity, on the other hand, was considered to be well mixed so the profile was uniform. The flow rate on each side was 2.15 SLPM, the feed inlet humidity was 40.4 g/kg and the sweep inlet humidity was 0 g/kg (the same as the experiment conditions). Table 3.2 Measured dimensions of the flow channel inserts Channel Width (mm) Height (mm) AR = 1 2.38 2.43 AR = 1, Ribbed 2.31 2.41 AR = 0.6 2.28 1.43 Table 3.3 Air properties assumed at 50\u00C2\u00B0C and 1atm Property Value Density, \u00CF\u0081 1.086 kg/m3 [9] Viscosity, \u00C2\u00B5 1.962 \u00C3\u008D10-5 kg/m/s [9] Mass Diffusivity, D 2.90\u00C3\u008D10-5 m2/s [35] Figure 3.19 shows the experimental results and corresponding simulations predictions. Results for smooth channels, a channel with 3D printed ribs, and a channel with formed membrane ribs are shown. Error bars on the experimental data represents the 95% confidence interval based on the sample standard deviation. Error bars on the WRR of the simulations data represents the uncertainty on the membrane permeability used. Quality control data from dPoint technologies indicates that the reported permeability of a given roll of membrane samples has an uncertainty 56 of +/- 6%. The experimental and simulation data agree within this uncertainty. As expected the ribbed channels outperform the smooth channels. Another feature is that because the membrane ribs are permeable, the membrane ribs outperform the printed ribs. Note, however, that this effect has been exaggerated by the dimensions of the channels, listed in Table 3.2. In particular, the printed AR = 1, ribbed channel has a smaller width than its smooth AR = 1 counterpart, which tends to decrease the WRR performance. Therefore, the experimental data should be taken mainly as an indicator of the simulation accuracy. Figure 3.19 \u00E2\u0080\u0093 Experimental performance of smooth and ribbed channels with corresponding simulation predictions. 57 Conclusions Angled rib features can be an effective method for increasing the effectiveness of membrane-based energy recovery ventilators. This study showed that adding ribs to a channel can increase the heat and mass transfer coefficients enough to outweigh the corresponding increase in friction factor. The magnitude of the heat transfer enhancement is consistent with previous studies. Furthermore, for realistic values for the membrane mass transfer resistance, the latent heat effectiveness of the ERV is also substantially improved by the addition of ribs. For smooth channels, Nu, Sh, and fRe are independent of Reynolds number and equal to the constant-wall-heat-flux case. For ribbed channels, they are strong functions of Re. The influence of channel aspect ratio on effectiveness was also studied. An aspect ratio of 1 generally gave the most improvement with ribs, as that aspect ratio yielded the most robust vortical structure downstream of the rib. Adding ribs and slightly increasing the channel height can increase recovery effectiveness by over 10% for an equal pressure drop. Even with constant channel dimensions, the ribbed channel is energetically preferable to a smooth channel because the increase in recovered energy is 2-3 orders of magnitude larger than the corresponding increase in fan power (during typical summer conditions). That is, under realistic conditions, the energy benefit of the ribs far outweighs the energy penalty due to increased friction. Experiments were conducted to validate the simulations. Ribs were either 3D printed and rested on the membrane surface (and so were impermeable) or formed into the membrane surface. Both configurations were found to be consistent with experimental predictions. 58 Traditionally, ERVs meet their effectiveness requirements by using more permeable membrane materials, decreasing the channel dimensions, or increasing the channel length. The data reported here show that ribs embossed on the membrane surface could be used instead, yielding a better enhancement/pressure relation than simply changing the channel dimensions. Also, embossing the membrane to create ribs would not require additional materials and would not entail changing the footprint of the core, which are both advantages relative to changing the channel dimensions. Future work in this area should focus on testing rib features in a full-sized ERV core. The current study showed that the benefit of ribs increases with Re, in the range studied, so a larger Re would make validation simpler. The rib geometry (height, width, angle, and spacing) could be optimized for the channel dimensions and Reynolds number chosen in order to maximize the sensible and latent improvement. In addition to increasing the net benefit of ribs, this would make the improvements easier to measure. 59 References [1] Natural Resources Canada, \u00E2\u0080\u009CEnergy Use Data Handbook Tables,\u00E2\u0080\u009D 15-Dec-2004. [Online]. Available: http://oee.nrcan.gc.ca/corporate/statistics/neud/dpa/menus/trends/handbook/tables.cfm. [Accessed: 02-Oct-2015]. [2] \u00E2\u0080\u009CBuildings Energy Data Book,\u00E2\u0080\u009D US Department of Energy, 2011, p. 2. [3] ASHRAE, \u00E2\u0080\u009CASHRAE handbook fundamentals,\u00E2\u0080\u009D 2009, p. 9.12. [4] ASHRAE, \u00E2\u0080\u009CASHRAE Standard 62.1-2016 - Ventilation for Acceptable Indoor Air Quality.\u00E2\u0080\u009D 2016. [5] L.-Z. Zhang, Total heat recovery heat and moisture recovery from ventilation air. New York: Nova Science Publishers, 2008. [6] R. Huizing, W. M\u00C3\u00A9rida, and F. Ko, \u00E2\u0080\u009CImpregnated electrospun nanofibrous membranes for water vapour transport applications,\u00E2\u0080\u009D J. Membr. Sci., vol. 461, pp. 146\u00E2\u0080\u0093160, Jul. 2014. [7] A. Engarnevis, \u00E2\u0080\u009CAmin Engarnevis PhD Proposal - The Influence of Humidity, Temperature, and Particulate Fouling on Permeation Properties of Dense Polymer Membranes.\u00E2\u0080\u009D 18-May-2015. [8] L.-Z. Zhang, Conjugate Heat and Mass Transfer in Heat Mass Exchanger Ducts. Academic Press, 2013. [9] J. R. Welty, Ed., Fundamentals of momentum, heat, and mass transfer, 5th ed. Danver, MA: Wiley, 2008. [10] J. Larminie and A. Dicks, \u00E2\u0080\u009CFuel cell systems explained,\u00E2\u0080\u009D 2nd ed., Chichester, West Sussex: J. Wiley, 2003, pp. 80\u00E2\u0080\u009390. [11] R. Huizing, M. Fowler, W. M\u00C3\u00A9rida, and J. Dean, \u00E2\u0080\u009CDesign methodology for membrane-based plate-and-frame fuel cell humidifiers,\u00E2\u0080\u009D J. Power Sources, vol. 180, no. 1, pp. 265\u00E2\u0080\u0093275, May 2008. [12] R. W. Baker, Membrane Technology and Applications. Academic Press, 2012. [13] L.-Z. Zhang, \u00E2\u0080\u009CHeat and mass transfer in a quasi-counter flow membrane-based total heat exchanger,\u00E2\u0080\u009D Int. J. Heat Mass Transf., vol. 53, no. 23\u00E2\u0080\u009324, pp. 5478\u00E2\u0080\u00935486, Nov. 2010. [14] S. Metz, W. Vandeven, J. Potreck, M. Mulder, and M. Wessling, \u00E2\u0080\u009CTransport of water vapor and inert gas mixtures through highly selective and highly permeable polymer membranes,\u00E2\u0080\u009D J. Membr. Sci., vol. 251, no. 1\u00E2\u0080\u00932, pp. 29\u00E2\u0080\u009341, Apr. 2005. [15] S. V. Patankar, C. H. Liu, and E. M. Sparrow, \u00E2\u0080\u009CFully developed flow and heat transfer in ducts having streamwise-periodic variations of cross-sectional area,\u00E2\u0080\u009D ASME J Heat Transf., vol. 99, no. 2, pp. 180\u00E2\u0080\u0093186, 1977. [16] L. Y. Qiu Zhong, \u00E2\u0080\u009CAn optimized crossflow plate-fin membrane-based total heat exchanger,\u00E2\u0080\u009D Energy Build., vol. 86, pp. 550\u00E2\u0080\u0093556, 2015. [17] S. Koester, A. Klasen, J. L\u00C3\u00B6lsberg, and M. Wessling, \u00E2\u0080\u009CSpacer enhanced heat and mass transfer in membrane-based enthalpy exchangers,\u00E2\u0080\u009D J. Membr. Sci., vol. 520, pp. 566\u00E2\u0080\u0093573, Dec. 2016. [18] M. Fiebig, T. G\u00C3\u00BCntermann, and N. K. Mitra, \u00E2\u0080\u009CNumerical Analysis of Heat Transfer and Flow Loss in a Parallel Plate Heat Exchanger Element With Longitudinal Vortex Generators as Fins,\u00E2\u0080\u009D J. Heat Transf., vol. 117, no. 4, pp. 1064\u00E2\u0080\u00931067, Nov. 1995. 60 [19] P. Deb, G. Biswas, and N. K. Mitra, \u00E2\u0080\u009CHeat transfer and flow structure in laminar and turbulent flows in a rectangular channel with longitudinal vortices,\u00E2\u0080\u009D Int. J. Heat Mass Transf., vol. 38, no. 13, pp. 2427\u00E2\u0080\u00932444, 1995. [20] S. K. Saha, \u00E2\u0080\u009CThermal and friction characteristics of laminar flow through rectangular and square ducts with transverse ribs and wire coil inserts,\u00E2\u0080\u009D Exp. Therm. Fluid Sci., vol. 34, no. 1, pp. 63\u00E2\u0080\u009372, Jan. 2010. [21] C.-O. Olsson and B. Sunden, \u00E2\u0080\u009CExperimental study of flow and heat transfer in rib-roughened rectangular channels,\u00E2\u0080\u009D Exp. Therm. Fluid Sci., vol. 16, pp. 349\u00E2\u0080\u0093365, 1998. [22] G. Tanda and R. Abram, \u00E2\u0080\u009CForced Convection Heat Transfer in Channels With Rib Turbulators Inclined at 45 deg,\u00E2\u0080\u009D J. Turbomach., vol. 131, no. 2, p. 21012, 2009. [23] K. M. Kim, H. Lee, B. S. Kim, S. Shin, D. H. Lee, and H. H. Cho, \u00E2\u0080\u009COptimal design of angled rib turbulators in a cooling channel,\u00E2\u0080\u009D Heat Mass Transf., vol. 45, no. 12, pp. 1617\u00E2\u0080\u00931625, Oct. 2009. [24] C. Nonino and G. Comini, \u00E2\u0080\u009CConvective heat transfer in ribbed square channels,\u00E2\u0080\u009D Int. J. Numer. Methods Heat Fluid Flow, vol. 12, no. 5, pp. 610\u00E2\u0080\u0093628, Aug. 2002. [25] \u00E2\u0080\u009CFLUENT 6.3 UDF Manual.\u00E2\u0080\u009D [Online]. Available: https://www.sharcnet.ca/Software/Fluent6/html/udf/main_pre.htm. [Accessed: 30-May-2017]. [26] M. Staudacher, M. Harasek, T. Brinkmann, W. Hilgendorff, and A. Friedl, \u00E2\u0080\u009CCFD-simulation of mass transfer effects in gas and vapour permeation modules,\u00E2\u0080\u009D Desalination, vol. 146, no. 1, pp. 237\u00E2\u0080\u0093241, Sep. 2002. [27] M. Coroneo, G. Montante, J. Catalano, and A. Paglianti, \u00E2\u0080\u009CModelling the effect of operating conditions on hydrodynamics and mass transfer in a Pd\u00E2\u0080\u0093Ag membrane module for H2 purification,\u00E2\u0080\u009D J. Membr. Sci., vol. 343, no. 1\u00E2\u0080\u00932, pp. 34\u00E2\u0080\u009341, Nov. 2009. [28] Y. Benguerba, J. Amer, and B. Ernst, \u00E2\u0080\u009CCFD modeling of the H2/N2 separation with a nickel/\u00CE\u00B1-alumina microporous membrane,\u00E2\u0080\u009D Chem. Eng. Sci., vol. 123, pp. 527\u00E2\u0080\u0093535, Feb. 2015. [29] R. Al-Waked, M. S. Nasif, G. Morrison, and M. Behnia, \u00E2\u0080\u009CCFD simulation of air to air enthalpy heat exchanger,\u00E2\u0080\u009D Energy Convers. Manag., vol. 74, pp. 377\u00E2\u0080\u0093385, Oct. 2013. [30] P. J. Roache, \u00E2\u0080\u009CQuantification of uncertainty in computational fluid dynamics,\u00E2\u0080\u009D Annu. Rev. Fluid Mech., vol. 29, no. 1, pp. 123\u00E2\u0080\u0093160, 1997. [31] A. L. London, Ed., \u00E2\u0080\u009CAdvances in HEAT TRANSFER A2 - Shah, R.K.,\u00E2\u0080\u009D in Laminar Flow Forced Convection in Ducts, Academic Press, 1978, p. ii. [32] AHRI, \u00E2\u0080\u009C2013 Standard for Performance Rating of Airto-Air Exchangers for Energy Recovery Ventilation Equipment.\u00E2\u0080\u009D 2013. [33] A. Engarnevis, \u00E2\u0080\u009CConjugate heat and mass transfer modeling in membrane-based enthalpy exchangers,\u00E2\u0080\u009D Unublished Manuscr., 2017. [34] R. K. Shah and A. L. London, Laminar Flow Forced Convection in Ducts: A Source Book for Compact Heat Exchanger Analytical Data. Academic Press, 2014. [35] G. Nellis and S. Klein, Heat Transfer. Cambridge University Press, 2009. 61 Appendix A \u00E2\u0080\u0093 User Defined Functions The functions below are compiled within fluent and used to run the simulation and post-process the results A.1 Membrane transport, Nusselt number, Sherwood number, and Friction Factor #include \"udf.h\" #define del 0.000105 /*thickness of membrane in m*/ #define R 8.314 /*universal gas constant J/K/mol*/ #define L 0.160 //length in m #define Dia 0.00143//hydraulic diameter in m #define n 40 //divide channel into this many bins #define nm 400 //n for cf calculation #define k_air 0.0262 //thermal conductivity of air in w/m/K (assuming 30C) #define D_air_h20 2.6e-5 //from fundamental of heat and mass transfer book #define Tw_i 303 //initial wall temperature (for constant wall temperature cases) #define T_film 298.15 //temperature to convert concentration to partial pressure #define WALL_ID 3 //ID of the dry wall surface #define WALL_ID_WET 5 //ID of the wet wall surface #define DRY_ID 7 //cell zone id of dry #define WET_ID 6 //cell zone id of wet #define WALL_ID_2 9//adiabatic dry wall #define inD \"real-1\" real D; /*prepares cells for implementation of the source term (so only wall cells are affected)*/ DEFINE_INIT(initilize,d) { face_t f; cell_t c, c0, c1; real NV_VEC(farea); real area; Thread *t, *t0, *t1, *ct; d = Get_Domain(1); /* Initialize Cells */ /* this loops over all cells and lets the UDM = 0 */ thread_loop_c(ct, d) { begin_c_loop(c,ct) { C_UDMI(c,ct,0) = 0; } end_c_loop(c,ct) } /* Loop over all faces on wall */ t = Lookup_Thread(d, WALL_ID); 62 begin_f_loop(f,t) { /* c0 and t0 identify the adjacent dry cell, c1 and t1 identifies the corresponding humid cell */ c0 = F_C0(f, t); c1 = F_C1(f,t); t0 = THREAD_T0(t); t1 = THREAD_T1(t); /* this loops over all cells adjacent to wall and lets the UDM = 1.0 for dry, and -1 for humid*/ C_UDMI(c0, t0, 0) = 1; C_UDMI(c1, t1, 0) = -1; C_UDMI(c0, t0, 3) = 1; C_UDMI(c1, t1, 3) = -1; C_UDMI(c0, t0, 1) = c1; /*sets partner for dry cell*/ C_UDMI(c1, t1, 1) = c0; /*sets partner for humid cell*/ C_UDMI(c0, t0, 4) = F_YI(f,t,0); //initializes the stored wall concentration for cases when we have a uniform wall concentration C_UDMI(c1, t1, 4) = F_YI(f,t,0); C_UDMI(c1, t1, 5) = Tw_i;// temporarily hard coding in the wall temp C_UDMI(c0, t0, 5) = Tw_i;// temporarily hard coding in the wall temp /* this is still in loop, and stores the face area of each cell on the membrane. They are assumed to be the same for each side */ F_AREA(farea,f,t); area = NV_MAG(farea); C_UDMI(c0, t0, 2) = area; /*sets face area for dry cell*/ C_UDMI(c1, t1, 2) = area; /*sets face area humid cell*/ } end_f_loop(f, t) } /*******************************************************************/ /* UDF for specifying h20 mass source */ /* */ /*******************************************************************/ DEFINE_SOURCE(m_source, c, t, dS, eqn) { Domain *d; real area, vol, source = 0; Thread *t1, *tfd, *tfh; cell_t c1; D = RP_Get_Input_Parameter(inD); /*permiability of membrane in mol-m/m^2/s/pa, from paramaeters list */ d = Get_Domain(1); tfd = Lookup_Thread(d, WALL_ID); /* gives the faces thread of membrane-shadow*/ tfh = Lookup_Thread(d, WALL_ID_WET); /* gives the faces thread of membrane*/ area = C_UDMI(c, t, 2); /* area of cell face on membrane*/ 63 vol = C_VOLUME(c, t); /* volume of cell*/ if (C_UDMI(c, t, 0) == 1) /*dry cell*/ { t1 = THREAD_T0(tfh); /*identifies the humid stream cell thread using the faces thread of membrane*/ } else if (C_UDMI(c, t, 0) == -1) /*humid cell*/ { t1 = THREAD_T0(tfd); /*identifies the dry stream cell thread using the faces thread of membrane shadow*/ } else { source = 0; dS[eqn] = 0; return source; } c1 = C_UDMI(c, t, 1); /*identifies the partner of the cell, taken from results of initilization*/ source = D*R*T_film*(C_R(c1, t1)*C_YI(c1, t1, 0) - C_R(c, t)*C_YI(c, t, 0))*area / (vol*del); dS[eqn] = -D*R*C_R(c, t)*T_film*area/(vol*del); return source; } /*******************************************************************/ /* UDF for specifying an h20 fraction on membrane */ /* */ /*******************************************************************/ DEFINE_PROFILE(membrane_conc, tf0, i) { cell_t c0, c1; Thread *t0, *t1, *tf1; real x0[ND_ND], xf0[ND_ND], A[ND_ND]; /* this will hold the position vector */ real y, faceYI; face_t f, f0, f1; float dy, dr; D = RP_Get_Input_Parameter(inD); begin_f_loop(f0, tf0) { /* c0 and t0 identify the adjacent cell, c1 and t1 identify the opposite cell */ c0 = F_C0(f0, tf0); t0 = THREAD_T0(tf0); c1 = F_C1(f0, tf0); t1 = THREAD_T1(tf0); f1 = F_SHADOW(f0, tf0); 64 tf1 = THREAD_SHADOW(tf0); C_CENTROID(x0, c0, t0); F_CENTROID(xf0, f0, tf0); dy = sqrt(pow(x0[0] - xf0[0], 2) + pow(x0[1] - xf0[1], 2) + pow(x0[2] - xf0[2], 2)); //distance between wall and cenn center if (N_ITER>0) { //only do this after initialization faceYI = C_YI(c0, t0, 0) - (D*R / (D_air_h20*del))*T_film*(F_YI(f0, tf0, 0) - F_YI(f1, tf1, 0)*C_R(c1, t1) / C_R(c0, t0))*dy; F_PROFILE(f0, tf0, i) = faceYI; //set wall concentration to create the desired gradient C_UDMI(c0, t0, 4) = faceYI; C_UDMI(c0, t0, 5) = F_T(f0,tf0); } else { F_PROFILE(f0, tf0, i) = 0; } } end_f_loop(f0, tf0) } /*******************************************************************/ /* UDF for getting Sherwood number */ /* */ /*******************************************************************/ DEFINE_OUTPUT_PARAMETER(Sh, m, parlist) { Domain *d; //declare domain variable, since it is not passed by the function Thread *t, *tw, *tc; cell_t c; face_t f; real x[ND_ND], x0[ND_ND], area[ND_ND]; //for position and area int k, i; real dy; int num[n] = { 0 }; //hold the number of cells in each bin real T_sum[n] = { 0 }, T_norm[n] = { 0 }, T_b[n] = { 0 }; real H_sum[n] = { 0 }, H_norm[n] = { 0 }, H_b[n] = { 0 }; int numf[n] = { 0 }; //hold the number of faces in each bin real Tf_sum[n] = { 0 }, Tf_norm[n] = { 0 }, T_w[n] = { 0 }; real Hf_sum[n] = { 0 }, Hf_norm[n] = { 0 }, H_w[n] = { 0 }; real Qf_sum[n] = { 0 }, Qf_norm[n] = { 0 }, Q_w[n] = { 0 }; real Jf_sum[n] = { 0 }, Jf_norm[n] = { 0 }, J_w[n] = { 0 }; real Nu[n] = { 0 }, Sh[n] = { 0 }; //nusselt number at bin n real Nu_av = 0, Sh_av = 0; d = Get_Domain(1); real a = parlist[0], b = parlist[1]; real Dh = 4*a*b / (a + b); //compute the hydraulic diameter. a is 1/2 the hight, b is 1/2 the width //bin cells by z position and calculate bulk values 65 tc = Lookup_Thread(d, DRY_ID); begin_c_loop(c, tc) { C_CENTROID(x, c, tc); //get the position of the cell C_UDMI(c, tc, 0) = floor(x[2] * n / L); //set memory of cell to its bin number k = C_UDMI(c, tc, 0); //get the bin number in integer form if (k < n && k >= 0) { T_sum[k] += fabs(C_W(c, tc))*C_VOLUME(c, tc)*C_R(c, tc)*C_T(c, tc); //integrate to get bulk temperature T_norm[k] += fabs(C_W(c, tc))*C_VOLUME(c, tc)*C_R(c, tc); //normalization factor H_sum[k] += fabs(C_W(c, tc))*C_VOLUME(c, tc)*C_R(c, tc)*C_R(c, tc)*C_YI(c, tc, 0); //integrate to get bulk concentration H_norm[k] += fabs(C_W(c, tc))*C_VOLUME(c, tc)*C_R(c, tc); num[k]++; //increment the count for that bin } } end_c_loop(c, tc) //bin faces by z position tw = Lookup_Thread(d, WALL_ID); begin_f_loop(f, tw) { c = F_C0(f, tw); //get cell index of face neighbor tc = THREAD_T0(tw); F_CENTROID(x, f, tw); //get the position of the face C_CENTROID(x0, c, tc); //get the position of the neighbor cell F_AREA(area, f, tw); //get face area F_UDMI(f, tw, 1) = floor(x[2] * n / L); //set memory of face to its bin number k = F_UDMI(f, tw, 1); //get the bin number in integer form dy = sqrt(pow(x[0] - x0[0], 2) + pow(x[1] - x0[1], 2) + pow(x[2] - x0[2], 2)); //distance between wall and cell center Tf_sum[k] += NV_MAG(area)*F_T(f, tw); //integrate the wall temperature Tf_norm[k] += NV_MAG(area); //normalization factor, area including rib Hf_sum[k] += NV_MAG(area)*C_R(c, tc)*F_YI(f, tw, 0); //integrate to get wall h20 concentration Hf_norm[k] += NV_MAG(area); //normalization factor, area including rib Qf_sum[k] += NV_MAG(area)*k_air*(F_T(f, tw) - C_T(c, tc)) / dy;//BOUNDARY_HEAT_FLUX(f,tw); Qf_norm[k] += fabs(area[1]); //normalization factor, area not including rib Jf_sum[k] += NV_MAG(area)*D_air_h20*C_R(c, tc)*(F_YI(f, tw, 0) - C_YI(c, tc, 0)) / dy; //boundary H20 flux Jf_norm[k] += fabs(area[1]); //normalization factor, area not including rib numf[k]++; //increment the count for that bin } end_f_loop(f, tw) //get the bulk temperature for each bin 66 for (int i = 0; i < n; i++) { T_b[i] = T_sum[i] / T_norm[i]; T_w[i] = Tf_sum[i] / Tf_norm[i]; Q_w[i] = Qf_sum[i] / Qf_norm[i]; H_b[i] = H_sum[i] / H_norm[i]; H_w[i] = Hf_sum[i] / Hf_norm[i]; J_w[i] = Jf_sum[i] / Jf_norm[i]; Nu[i] = Dh*Q_w[i] / (k_air*(T_w[i] - T_b[i])); //calculate nusselt number in each bin Sh[i] = Dh*J_w[i] / (D_air_h20*(H_w[i] - H_b[i])); //calculate sherwood number in each bin Message(\"Bin %d: Nu = %g, Sh = %g , H_b = %g, H_w = %g, J_w = %g \n\", i, Nu[i], Sh[i], H_b[i], H_w[i], J_w[i]); } for (int i = n/4; i <= 3*n/4-1; i++) { Nu_av += Nu[i] / (n/2); //average over the middle of tube Sh_av += Sh[i] / (n/2); } //Message(\"Average Nu at end: %g, Average Sh at end: %g \n\", Nu_av, Sh_av); return Sh_av; } /*******************************************************************/ /* UDF for getting Nusselt number */ /* */ /*******************************************************************/ DEFINE_OUTPUT_PARAMETER(Nu, m, parlist) { Domain *d; //declare domain variable, since it is not passed by the function Thread *t, *tw, *tc; cell_t c; face_t f; real x[ND_ND], x0[ND_ND], area[ND_ND]; //for position and area int k, i; real dy; int num[n] = { 0 }; //hold the number of cells in each bin real T_sum[n] = { 0 }, T_norm[n] = { 0 }, T_b[n] = { 0 }; real H_sum[n] = { 0 }, H_norm[n] = { 0 }, H_b[n] = { 0 }; int numf[n] = { 0 }; //hold the number of faces in each bin real Tf_sum[n] = { 0 }, Tf_norm[n] = { 0 }, T_w[n] = { 0 }; real Hf_sum[n] = { 0 }, Hf_norm[n] = { 0 }, H_w[n] = { 0 }; real Qf_sum[n] = { 0 }, Qf_norm[n] = { 0 }, Q_w[n] = { 0 }; real Jf_sum[n] = { 0 }, Jf_norm[n] = { 0 }, J_w[n] = { 0 }; real Nu[n] = { 0 }, Sh[n] = { 0 }; //nusselt number at bin n real Nu_av = 0, Sh_av = 0; d = Get_Domain(1); real a = parlist[0], b = parlist[1]; 67 real Dh = 4 * a*b / (a + b); //compute the hydraulic diameter. a is 1/2 the hight, b is 1/2 the width //bin cells by z position tc = Lookup_Thread(d, DRY_ID); begin_c_loop(c, tc) { C_CENTROID(x, c, tc); //get the position of the cell C_UDMI(c, tc, 0) = floor(x[2] * n / L); //set memory of cell to its bin number k = C_UDMI(c, tc, 0); //get the bin number in integer form if (k < n && k >= 0) { T_sum[k] += fabs(C_W(c, tc))*C_VOLUME(c, tc)*C_R(c, tc)*C_T(c, tc); //integrate to get bulk temperature T_norm[k] += fabs(C_W(c, tc))*C_VOLUME(c, tc)*C_R(c, tc); //normalization factor H_sum[k] += fabs(C_W(c, tc))*C_VOLUME(c, tc)*C_R(c, tc)*C_R(c, tc)*C_YI(c, tc, 0); //integrate to get bulk concentration H_norm[k] += fabs(C_W(c, tc))*C_VOLUME(c, tc)*C_R(c, tc); num[k]++; //increment the count for that bin } } end_c_loop(c, tc) //bin faces by z position tw = Lookup_Thread(d, WALL_ID); begin_f_loop(f, tw) { c = F_C0(f, tw); //get cell index of face neighbor tc = THREAD_T0(tw); F_CENTROID(x, f, tw); //get the position of the face C_CENTROID(x0, c, tc); //get the position of the neighbor cell F_AREA(area, f, tw); //get face area F_UDMI(f, tw, 1) = floor(x[2] * n / L); //set memory of face to its bin number k = F_UDMI(f, tw, 1); //get the bin number in integer form dy = sqrt(pow(x[0] - x0[0], 2) + pow(x[1] - x0[1], 2) + pow(x[2] - x0[2], 2)); //distance between wall and cell center Tf_sum[k] += NV_MAG(area)*F_T(f, tw); //integrate the wall temperature Tf_norm[k] += NV_MAG(area); //normalization factor, area including rib Hf_sum[k] += NV_MAG(area)*C_R(c, tc)*F_YI(f, tw, 0); //integrate to get wall h20 concentration Hf_norm[k] += NV_MAG(area); //normalization factor, area including rib Qf_sum[k] += NV_MAG(area)*k_air*(F_T(f, tw) - C_T(c, tc)) / dy;//BOUNDARY_HEAT_FLUX(f,tw); Qf_norm[k] += fabs(area[1]); //normalization factor, area not including rib Jf_sum[k] += NV_MAG(area)*D_air_h20*C_R(c, tc)*(F_YI(f, tw, 0) - C_YI(c, tc, 0)) / dy; //boundary H20 flux Jf_norm[k] += fabs(area[1]); //normalization factor, area not including rib numf[k]++; //increment the count for that bin 68 } end_f_loop(f, tw) //get the bulk temperature for each bin for (int i = 0; i < n; i++) { T_b[i] = T_sum[i] / T_norm[i]; T_w[i] = Tf_sum[i] / Tf_norm[i]; Q_w[i] = Qf_sum[i] / Qf_norm[i]; H_b[i] = H_sum[i] / H_norm[i]; H_w[i] = Hf_sum[i] / Hf_norm[i]; J_w[i] = Jf_sum[i] / Jf_norm[i]; Nu[i] = Dh*Q_w[i] / (k_air*(T_w[i] - T_b[i])); //calculate nusselt number in each bin Sh[i] = Dh*J_w[i] / (D_air_h20*(H_w[i] - H_b[i])); //calculate sherwood number in each bin Message(\"Bin %d: Nu = %g, Sh = %g , T_b = %g, T_w = %g, Q_w = %g \n\", i, Nu[i], Sh[i], T_b[i], T_w[i], Q_w[i]); } for (int i = n / 4; i <= 3 * n / 4 - 1; i++) { Nu_av += Nu[i] / (n/2); //average over the end of tube Sh_av += Sh[i] / (n/2); } Message(\"Average Nu in middle: %g, Average Sh in middle: %g \n\", Nu_av, Sh_av); return Nu_av; } /*******************************************************************/ /* UDF for getting wall stress (for friction factor) */ /* */ /*******************************************************************/ DEFINE_OUTPUT_PARAMETER(cf, m, parlist) { Domain *d; //declare domain variable, since it is not passed by the function Thread *t, *tw, *tc; cell_t c; face_t f; real wallshear[ND_ND], x[ND_ND], x0[ND_ND], area[ND_ND]; //for position, area, and wall shear int k, i; real F_w[nm] = { 0 }; real Tau[nm] = { 0 }; real Tau_av = 0; real F_tot = 0; int walls[2] = { WALL_ID, WALL_ID_2 }; d = Get_Domain(1); real a = parlist[0], b = parlist[1]; 69 real Dh = 4 * a*b / (a + b); //compute the hydraulic diameter. a is 1/2 the hight, b is 1/2 the width real A_part = L * 4 * (a + b)/nm; //area of each bin //bin faces by z position for (int i = 0; i < 2; i++) { tw = Lookup_Thread(d, walls[i]); begin_f_loop(f, tw) { F_AREA(area, f, tw); //get face area F_CENTROID(x, f, tw); //get the position of the face k = floor(x[2] * nm / L); //set memory of face to its bin number NV_V(wallshear, =, C_STORAGE_R_NV(f, tw, SV_WALL_SHEAR)); //get wall shear stress F_w[k] += F_P(f, tw)*area[2] - wallshear[2]; //total force in z-direction is the sum of shear and pressure components } end_f_loop(f, tw) } //get the bulk temperature for each bin for (int i = 0; i < nm; i++) { Tau[i] = F_w[i] / A_part; //total wall stress in each bin F_tot += F_w[i]; Message(\"Bin %d: Tau = %g\n\", i, Tau[i]); } for (int i = nm / 4; i <= 3 * nm / 4 - 1; i++) { Tau_av += Tau[i] / (nm / 2); //average over the end of tube } Message(\"Average Tau in middle: %g, F_tot: %g \n\", Tau_av, F_tot); return Tau_av; } 70 A.2 Periodic Boundary Conditions #include \"udf.h\" #define OUT_ID_DRY 11 //ID of the dry wall surface #define OUT_ID_WET 12 //ID of the wet wall surface #define IN_ID_DRY 13 //cell zone id of dry #define IN_ID_WET 10 //cell zone id of wet #define N 1000 #define Rho 1.184 //density of dry air #define T_wet 308.15//T of dry air #define T_dry 297.0388889//T of wet air #define X_wet 0.01677 //this is actually Y, the humidity mass fraction #define X_dry 0.00751 //this is actually Y, the humidity mass fraction #define ina \"real-2\" //1/2 width of channel cross-section #define inb \"real-3\" //1/2 height of channel cross-section #define N_START 5 //number of iterations to wait before starting periodic DEFINE_PROFILE(periodic_velocity, tf0, i) { Domain *d; Thread *tf1, *t; cell_t c; real x0[ND_ND], x1[ND_ND], vel[N][N]; face_t f0, f1; int xn, yn; real a, b; a = RP_Get_Input_Parameter(ina)*2; //get channel dimensions from parameter list b = RP_Get_Input_Parameter(inb)*2; d = Get_Domain(1); if (tf0 == Lookup_Thread(d, IN_ID_DRY)) //match up dry inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_DRY); } else if (tf0 == Lookup_Thread(d, IN_ID_WET)) //match up wet inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_WET); } begin_f_loop(f1, tf1) //store the velocity field of the onlet in the array { if (N_ITER > 0) { //only do this after first iteration F_CENTROID(x1, f1, tf1); xn = floor(fabs(x1[0] * N / a)); yn = floor(fabs(x1[1] * N / b)); vel[xn][yn] = pow(pow(F_U(f1, tf1), 2)+ pow(F_V(f1, tf1), 2)+ pow(F_W(f1, tf1), 2), 0.5); } } end_f_loop(f1,tf1) begin_f_loop(f0, tf0) //apply the velocity field of the outlet to the inlet { 71 F_CENTROID(x0, f0, tf0); xn = floor(fabs(x0[0] * N / a)); yn = floor(fabs(x0[1] * N / b)); if (N_ITER>N_START) { F_PROFILE(f0, tf0, i) = fabs(vel[xn][yn]); } else { F_PROFILE(f0, tf0, i) = 1; } } end_f_loop(f0, tf0) } DEFINE_PROFILE(periodic_u, tf0, i) { Domain *d; Thread *tf1, *t; cell_t c; real x0[ND_ND], x1[ND_ND], vel[N][N]; face_t f0, f1; int xn, yn; real a, b; a = RP_Get_Input_Parameter(ina) * 2; //get channel dimensions from parameter list b = RP_Get_Input_Parameter(inb) * 2; d = Get_Domain(1); if (tf0 == Lookup_Thread(d, IN_ID_DRY)) //match up dry inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_DRY); } else if (tf0 == Lookup_Thread(d, IN_ID_WET)) //match up wet inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_WET); } begin_f_loop(f1, tf1) //store the velocity field of the onlet in the array { if (N_ITER > 0) { //only do this after first iteration F_CENTROID(x1, f1, tf1); xn = floor(fabs(x1[0] * N / a)); yn = floor(fabs(x1[1] * N / b)); vel[xn][yn] = F_U(f1, tf1); } } end_f_loop(f1, tf1) begin_f_loop(f0, tf0) //apply the velocity field of the outlet to the inlet { F_CENTROID(x0, f0, tf0); xn = floor(fabs(x0[0] * N / a)); yn = floor(fabs(x0[1] * N / b)); if (N_ITER>N_START) { 72 F_PROFILE(f0, tf0, i) = vel[xn][yn]; } else { F_PROFILE(f0, tf0, i) = 0; } } end_f_loop(f0, tf0) } DEFINE_PROFILE(periodic_v, tf0, i) { Domain *d; Thread *tf1, *t; cell_t c; real x0[ND_ND], x1[ND_ND], vel[N][N]; face_t f0, f1; int xn, yn; real a, b; a = RP_Get_Input_Parameter(ina) * 2; //get channel dimensions from parameter list b = RP_Get_Input_Parameter(inb) * 2; d = Get_Domain(1); if (tf0 == Lookup_Thread(d, IN_ID_DRY)) //match up dry inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_DRY); } else if (tf0 == Lookup_Thread(d, IN_ID_WET)) //match up wet inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_WET); } begin_f_loop(f1, tf1) //store the velocity field of the inlet in the array { if (N_ITER > 0) { //only do this after first iteration F_CENTROID(x1, f1, tf1); xn = floor(fabs(x1[0] * N / a)); yn = floor(fabs(x1[1] * N / b)); vel[xn][yn] = F_V(f1, tf1); } } end_f_loop(f1, tf1) begin_f_loop(f0, tf0) //apply the velocity field of the outlet to the inlet { F_CENTROID(x0, f0, tf0); xn = floor(fabs(x0[0] * N / a)); yn = floor(fabs(x0[1] * N / b)); if (N_ITER>N_START) { F_PROFILE(f0, tf0, i) = vel[xn][yn]; } else { F_PROFILE(f0, tf0, i) = 0; } 73 } end_f_loop(f0, tf0) } DEFINE_PROFILE(periodic_w, tf0, i) { Domain *d; Thread *tf1, *t; cell_t c; real x0[ND_ND], x1[ND_ND], vel[N][N]; face_t f0, f1; int xn, yn; real a, b, k=1; a = RP_Get_Input_Parameter(ina) * 2; //get channel dimensions from parameter list b = RP_Get_Input_Parameter(inb) * 2; d = Get_Domain(1); if (tf0 == Lookup_Thread(d, IN_ID_DRY)) //match up dry inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_DRY); } else if (tf0 == Lookup_Thread(d, IN_ID_WET)) //match up wet inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_WET); k = -1; } begin_f_loop(f1, tf1) //store the velocity field of the inlet in the array { if (N_ITER > 0) { //only do this after first iteration F_CENTROID(x1, f1, tf1); xn = floor(fabs(x1[0] * N / a)); yn = floor(fabs(x1[1] * N / b)); vel[xn][yn] = F_W(f1, tf1); } } end_f_loop(f1, tf1) begin_f_loop(f0, tf0) //apply the velocity field of the outlet to the inlet { F_CENTROID(x0, f0, tf0); xn = floor(fabs(x0[0] * N / a)); yn = floor(fabs(x0[1] * N / b)); if (vel[xn][yn] == 0) { F_PROFILE(f0, tf0, i) = k; } else if (N_ITER>N_START) { F_PROFILE(f0, tf0, i) = vel[xn][yn]; } else { F_PROFILE(f0, tf0, i) = k; } } 74 end_f_loop(f0, tf0) } DEFINE_PROFILE(periodic_species, tf0, i) { Domain *d; Thread *tf1, *t; cell_t c; real x0[ND_ND], x1[ND_ND], area[ND_ND], Y[N][N]; face_t f0, f1; int xn, yn; real a, b; real H_sum = 0, H_norm = 0, H_b0 = 0, X_b0 = 0.005, H_b1 = 0; real Hf_sum = 0, Hf_norm = 0, H_w0 = 0, H_w1 = 0; a = RP_Get_Input_Parameter(ina) * 2; //get channel dimensions from parameter list b = RP_Get_Input_Parameter(inb) * 2; d = Get_Domain(1); if (N_ITER>N_START) { if (tf0 == Lookup_Thread(d, IN_ID_DRY)) //match up dry inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_DRY); X_b0 = X_dry; } else if (tf0 == Lookup_Thread(d, IN_ID_WET)) //match up wet inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_WET); X_b0 = X_wet; } //calculate and store the species field of the outlet begin_f_loop(f1, tf1) { if (N_ITER > 0) { //only do this after first iteration F_CENTROID(x1, f1, tf1); //get the position of the cell xn = floor(fabs(x1[0] * N / a)); //x bin yn = floor(fabs(x1[1] * N / b)); //y bin F_AREA(area, f1, tf1); Y[xn][yn] = F_YI(f1, tf1, 0)*F_R(f1, tf1); //store the species concentraion on the outlet H_sum += fabs(F_W(f1, tf1))*NV_MAG(area)*F_R(f1, tf1)*F_R(f1, tf1)*F_YI(f1, tf1, 0); //integrate to get bulk concentration H_norm += fabs(F_W(f1, tf1))*NV_MAG(area)*F_R(f1, tf1); //normalization factor for bulk concentration c = F_C0(f1, tf1); //get cell adjacent to face t = THREAD_T0(tf1); // get thread for above cell 75 if (fabs(C_UDMI(c,t,3)) == 1) //check if the cell is adjacent to the membrane (+1 for dry, -1 for humid) { Hf_sum += C_VOLUME(c,t)*C_R(c, t)*C_UDMI(c,t,4); //integrate to get wall h20 concentration. the wall mass fraction has been stored in UDMI 4 by the fick's law macro Hf_norm += C_VOLUME(c, t); //normalization factor } } } end_f_loop(f1, tf1) H_b1 = H_sum / H_norm; //compute bulk and wall average at the outlet H_w1 = Hf_sum / Hf_norm; H_sum = 0; //reset the running averages H_norm = 0; Hf_sum = 0; Hf_norm = 0; //calculate inlet bulk and wall fields begin_f_loop(f0, tf0) { F_CENTROID(x0, f0, tf0); //get the position of the cell F_AREA(area, f0, tf0); H_sum += fabs(F_W(f0, tf0))*NV_MAG(area)*F_R(f0, tf0)*F_R(f0, tf0)*F_YI(f0, tf0, 0); //integrate to get bulk concentration H_norm += fabs(F_W(f0, tf0))*NV_MAG(area)*F_R(f0, tf0); //normalization factor for bulk concentration c = F_C0(f0, tf0); //get cell adjacent to face t = THREAD_T0(tf0); // get thread for above cell if (fabs(C_UDMI(c, t, 3)) == 1) //check if the cell is adjacent to the membrane (+1 for dry, -1 for humid) { Hf_sum += C_VOLUME(c, t)*C_R(c, t)*C_UDMI(c, t, 4); //integrate to get wall h20 concentration. the wall mass fraction has been stored in UDMI 4 by the fick's law macro Hf_norm += C_VOLUME(c, t); //normalization factor } } end_f_loop(f0, tf0) //compute bulk and wall concentrations at the inlet H_w0 = Hf_sum / Hf_norm; H_b0 = X_b0*Rho; //calculate bulk inlet concentration using inlet humidity ratio and dry air density //apply the field of the outlet to the inlet begin_f_loop(f0, tf0) { F_CENTROID(x0, f0, tf0); xn = floor(fabs(x0[0] * N / a)); yn = floor(fabs(x0[1] * N / b)); 76 F_PROFILE(f0, tf0, i) = ((Y[xn][yn] - H_w1) / (H_b1 - H_w1)*(H_b0 - H_w0) + H_w0) / F_R(f0, tf0); } end_f_loop(f0, tf0) } else { begin_f_loop(f0, tf0) { F_PROFILE(f0, tf0, i) = X_b0; } end_f_loop(f0, tf0) } //Message(\"Hb = %g, Hw = %g, Hb0 = %g, Hw0 = %g \n\", H_b1, H_w1, H_b0, H_w0); } DEFINE_PROFILE(periodic_temperature, tf0, i) { Domain *d; Thread *tf1, *t; cell_t c; real x0[ND_ND], x1[ND_ND], area[ND_ND], Y[N][N]; face_t f0, f1; int xn, yn; real a, b; real T_sum = 0, T_norm = 0, T_norm_1 = 0, T_b0 = 300, T_b1 = 0, T_b_inlet = 0; real Tf_sum = 0, Tf_norm = 0, T_w0 = 0, T_w1 = 0; a = RP_Get_Input_Parameter(ina) * 2; //get channel dimensions from parameter list b = RP_Get_Input_Parameter(inb) * 2; d = Get_Domain(1); if (N_ITER>N_START*3) { if (tf0 == Lookup_Thread(d, IN_ID_DRY)) //match up dry inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_DRY); T_b0 = T_dry; } else if (tf0 == Lookup_Thread(d, IN_ID_WET)) //match up wet inlet and outlet { tf1 = Lookup_Thread(d, OUT_ID_WET); T_b0 = T_wet; } //calculate and store the temperature field of the outlet begin_f_loop(f1, tf1) { 77 F_CENTROID(x1, f1, tf1); //get the position of the cell xn = floor(fabs(x1[0] * N / a)); //x bin yn = floor(fabs(x1[1] * N / b)); //y bin F_AREA(area, f1, tf1); T_sum += fabs(F_W(f1, tf1))*NV_MAG(area)*F_R(f1, tf1)*F_T(f1, tf1); //integrate to get bulk temperature T_norm_1 += fabs(F_W(f1, tf1))*NV_MAG(area)*F_R(f1, tf1); //normalization factor for bulk concentration Y[xn][yn] = F_T(f1, tf1); //store the temperature of the outlet c = F_C0(f1, tf1); //get cell adjacent to face t = THREAD_T0(tf1); // get thread for cell adjacenet to face if (fabs(C_UDMI(c, t, 3)) == 1) //check if the cell is adjacent to the membrane (+1 for dry, -1 for humid) { Tf_sum += C_VOLUME(c, t)*C_UDMI(c, t, 5); //integrate to get wall temperature. Tf_norm += C_VOLUME(c, t); //normalization factor } } end_f_loop(f1, tf1) T_b1 = T_sum / T_norm_1; //compute bulk and wall average at the outlet T_w1 = Tf_sum / Tf_norm; T_sum = 0; //reset the running averages Tf_sum = 0; Tf_norm = 0; //calculate inlet bulk and wall fields begin_f_loop(f0, tf0) { F_CENTROID(x0, f0, tf0); //get the position of the cell F_AREA(area, f0, tf0); T_sum += fabs(F_W(f0, tf0))*NV_MAG(area)*F_R(f0, tf0)*F_T(f0, tf0); //integrate to get bulk concentration T_norm += fabs(F_W(f0, tf0))*NV_MAG(area)*F_R(f0, tf0); //normalization factor for bulk concentration c = F_C0(f0, tf0); //get cell adjacent to face t = THREAD_T0(tf0); // get thread for above cell if (fabs(C_UDMI(c, t, 3)) == 1) //check if the cell is adjacent to the membrane (+1 for dry, -1 for humid) { Tf_sum += C_VOLUME(c, t)*C_UDMI(c, t, 5); //integrate to get wall temperature. Tf_norm += C_VOLUME(c, t); //normalization factor } } end_f_loop(f0, tf0) 78 //compute bulk and wall temperatures at the inlet T_w0 = Tf_sum / Tf_norm; T_b_inlet = T_sum / T_norm; //apply the field of the outlet to the inlet begin_f_loop(f0, tf0) { F_CENTROID(x0, f0, tf0); xn = floor(fabs(x0[0] * N / a)); yn = floor(fabs(x0[1] * N / b)); F_PROFILE(f0, tf0, i) = ((Y[xn][yn] - T_w1) / (T_b1 - T_w1)*(T_b0 - T_w0) + T_w0); //define inlet profile attempting to take into account changes in density } end_f_loop(f0, tf0) } else { begin_f_loop(f0, tf0) { F_PROFILE(f0, tf0, i) = T_b0; } end_f_loop(f0, tf0) } //Message(\"Tb = %g, Tw = %g, Tb0 = %g, Tw0 = %g \n\", T_b1, T_w1, T_b_inlet, T_w0); } 79 A.3 Rectangular Channel Fully Developed Velocity Profile #include \"udf.h\" #define pi 3.1415926535 DEFINE_PROFILE(square_profile, t, i) { real r[ND_ND]; //holds the position of the face cells real y, x; //coordinates of the face cell face_t f; real a = RP_Get_Input_Parameter(\"real-2\"); //half the inlet width from parameter list real b = RP_Get_Input_Parameter(\"real-3\"); //half the inlet heigh from parameter list begin_f_loop(f, t) { F_CENTROID(r, f, t); //get face position y = r[1]; //y coordinate x = r[0]; //x coordinate real prof = 0; //initialise the profile vale for the face int m = 1; if (y < 0) { m = -1; //this modifier allows for use in teh dry zone which is below which is shifted by 2b } for (int n = 1; n < 100; n = n + 2) { prof += 1/pow(n,3)*pow(-1, (n - 1) / 2)*(1 - cosh(n * pi*(y - m*b) / 2 / a) / cosh(n * pi*b / 2 / a))*cos(n * pi*(x-a) / 2 / a); } F_PROFILE(f, t, i) = prof; } end_f_loop(f,t) } "@en .
"Thesis/Dissertation"@en .
"2017-11"@en .
"10.14288/1.0357187"@en .
"eng"@en .
"Mechanical Engineering"@en .
"Vancouver : University of British Columbia Library"@en .
"University of British Columbia"@en .
"Attribution-NonCommercial-NoDerivatives 4.0 International"@* .
"http://creativecommons.org/licenses/by-nc-nd/4.0/"@* .
"Graduate"@en .
"Rib-roughened heat and mass transfer enhancement in membrane-based energy recovery ventilators"@en .
"Text"@en .
"http://hdl.handle.net/2429/63366"@en .