COMPARISON OF NONWOVEN FIBERGLASS AND STAINLESS STEEL MICROFIBER MEDIA IN AEROSOL COALESCENCE FILTRATION

Transcription

1 COMPARISON OF NONWOVEN FIBERGLASS AND STAINLESS STEEL MICROFIBER MEDIA IN AEROSOL COALESCENCE FILTRATION A Dissertation Presented to The Graduate Faculty of The University of Akron In Partial Fullfillment of the Requirements for the Degree Doctor of Philosophy Gabriel Manzo August, 2015

2 COMPARISON OF NONWOVEN FIBERGLASS AND STAINLESS STEEL MICROFIBER MEDIA IN AEROSOL COALESCENCE FILTRATION Gabriel Manzo Dissertation Approved: Accepted: Advisor Dr. George G. Chase Department Chair Dr. Michael H. Cheung Committee Member Dr. Lingyun Liu Dean of College Dr. Rex Ramsier Committee Member Dr. Bi-min Zhang Newby Dean of Graduate School Dr. Chand Midha Committee Member Date Dr. Peter Gordon Committee Member Dr. Alex Povitsky ii

3 ABSTRACT Coalescing filters are used to remove small liquid droplets from air streams. They have numerous industrial applications including dehumidification, cabin air filtration, compressed air filtration, metal working, CCV, and agriculture. In compressed air systems, oils used for lubrication of compressor parts can aerosolize into the main air stream causing potential contamination concerns for downstream applications. In many systems, humid air can present problems to sensitive equipment and sensors. As the humid air cools, small water drops condense and can disrupt components that need to be kept dry. Fibrous nonwoven filter media are commonly used to coalesce small drops into larger drops for easier removal. The coalescing performance of a medium is dependent upon several parameters including permeability, porosity, and wettability. In many coalescing filters, glass fibers are used. In this work, the properties of steel fiber media are measured to see how these properties compare to glass fiber media. Steel fiber media has different permeability, porosity and wettability to oil and water than fiber glass media. These differences can impact coalescence performance. The impact of these differences in properties on coalescence filtration performance was evaluated in a coalescence test apparatus. The overall coalescence performance of the steel and glass nonwoven fiber media are compared using a filtration efficiency and filtration index. In many cases, the stainless steel media performed comparably to fiber glass media with efficiencies near 90%. Since stainless steel media had lower pressure drops than fiber glass media, its filtration index values were significantly higher. iii

4 Broader impact of this work is the use of stainless steel fiber media as an alternative to fiber glass media in applications where aerosol filtration is needed to protect the environment or sensitive equipment and sensors. iv

5 DEDICATION I would like to dedicate this work to my parents Mr. Gregory J. Manzo and Mrs. Mary Ellen Manzo and my grandparents the late Mr. Emmett Manzo, Mrs. Concetta Manzo, the late Mr. Anthony Russo and the late Mrs. Josephine Russo. v

6 ACKNOWLEDGEMENTS I would like to acknowledge the following people who helped me achieve this goal. I would like to express my gratitude and admiration towards Dr. George G. Chase for his invaluable guidance, support and encouragement throughout the course of my doctorate studies. Dr. Chase has been a great inspiration in my life in what it means to truly be an engineer and has helped me grow immensely as a professional in this field. I would like to thank my committee members; Dr. Lingyun Lui, Dr. Bi-min Zhang Newby, Dr. Peter Gordon, and Dr. Alex Povitsky for their valuable advice and guidance. I would like to thank Mr. Frank Pelc and Mr. Will Imes for their assistance in troubleshooting problems with the coalescence filtration test stand. Their assistance and support was greatly appreciated and their friendship invaluable. I would like to thank all my friends in the multiphase group for their help and support. All past and present multiphase group members that I have worked with over the years have become some of my best friends and helped encourage me when the work seemed overwhelming. I would like to acknowledge the Coalescence Filtration Nanofiber Consortium and their member companies, especially N.V. Bekaert S.A. for their financial support and guidance on this work. I would like to thank my parents for their constant support and unwavering belief in my abilities to complete this goal. They have pushed me to shoot for the stars and reminded me that nothing is impossible. I would also like to thank the rest of my family members including my brother and sister-in-law, Mr. Emmett and Shannon Manzo, my sister, Miss vi

7 Catherine Manzo, my grandmothers, Mrs. Concetta Manzo and the late Mrs. Josephine V. Russo, my aunts and uncles and cousins for their support of my endeavor. I am truly blessed to have such a loving and caring family. vii

8 TABLE OF CONTENTS LIST OF TABLES. xiv Page LIST OF FIGURES xviii CHAPTER I. INTRODUCTION Motivation Problem Statement Hypothesis Research Objectives Benefits of Current Work Dissertation Outline..6 II. LITERATURE REVIEW Overview of Filtration Types of Filtration Coalescence Filtration...9 viii

9 2.4 Aerosol Coalescence Filtration Sources of Liquid Aerosol Fibrous Aerosol Coalescing Filters Capture Mechanisms Saturation of Fibrous Filter Media 16 III. FABRICATION AND CHARACTERIZATION OF FILTER MEDIA Filter Media Fabrication Wet Laid Formation of Glass Fiber Media Media Characterization Porosity Measurement Techniques Permeability Measurement Techniques Wettability Measurement Techniques Nonwoven Structural Properties Results and Discussion Summary of Results...44 IV. AEROSOL COALESCENCE FILTRATION TESTING...45 x

10 4.1 Coalescence Experimental Setup and Performance Quantities Prefiltration Aerosol Generation Filter Holder Measurement Equipment Filtration Performance Characterization Aerosol Coalescence Filtration Testing Layered Media Compressed Layered Media Drainage Augmented Layered Media Summary of Results 73 V. SHRINKAGE OF FIBROUS FILTER MEDIA DUE TO INTERNAL CAPILLARY FORCES Introduction Filter Media and Testing Liquids Experimental Procedure Results and Discussion Summary of Results.92 xi

11 VI. SHELL BALANCE APPROACH TO MODELING EFFICIENCY Introduction and Motivation Single Fiber Theory Shell Balance Modeling Approach Single Fiber Efficiency Regression Analysis Summary of Results.115 VII. MULTIPHASE TRANSPORT PHENOMENA APPROACH TO MODELING EFFICIENCY Assumptions Multiphase Conservation Equations Modeling Approach First Model Regression Analysis Second Model Regression Analysis Summary of Results..143 VIII. CONCLUSIONS 144 IX. FUTURE WORK..149 xii

12 REFERENCES 151 APPENDICES.168 APPENDIX A: STATISTICAL TESTS IN MLR APPENDIX B: EFFECT OF MULTICOLLINEARITY IN MLR..173 APPENDIX C: PROBLEMS ENCOUNTERED WITH ESTIMATION OF FILTRATION EFFICIENCY USING PREVIOUS SINGLE FIBER THEORY 177 APPENDIX D: MULTIPHASE TRANPORT PHENOMENA THEORY.196 APPENDIX E: REFERENCES..205 xiii

13 LIST OF TABLES Table Page 3.1 Media Information Uncompressed porosity values Properties of Sullube-32 and its contact angle on representative 316 SS glass surfaces Physical properties and airflow permeability Wettability properties Pore size and airflow permeability Wettability properties Properties of the filter samples used in the shrinkage experiment. All filter media were disk shaped with diameters of 2.5 cm and stacked to thicknesses of 1cm Properties of liquids used in the shrinkage experiment Young s modulus, Eb, for the fiber media Logistic Regression with first parameter, R Logistic Regression two parameter (R,Ku) fit of data..110 xiv

14 6.3 Logistic Regression two parameter (R,Pe) Comparison of measured and predicted efficiency using Eq and ANOVA table for overall filter efficiency model using Eq and Regression based on one parameter (Ln αf) Two parameter regression (Lnαf and LnR) Three parameter regression (Lnαf, LnR and LnS) Four parameter regression (Lnαf, LnR, LnS and Ln Ku) Regression analysis with parameters Ln S and Ln Ku Regression analysis with parameters Lnαf, LnR, LnS and LnPe Regression analysis with parameters Lnαf, LnS and LnPe Regression analysis combing Pe and R into one term LnPe/R Regression analysis combing S, Pe and R into one term LnSPe/R LOF regression analysis Experimental efficiency vs. calculated efficiency using Eq ANOVA table for overall filter efficiency model using Eq Regression analysis of LnY vs Ln(1-ε), LnSPe/R LOF regression for LnY vs. Ln(1-ε) and LnSPe/R 139 xv

15 7.15 Experimental efficiency vs. calculated efficiency using Eq ANOVA table for overall filter efficiency model using Eq Regression analysis with modified Pe number C.1 Predicted efficiency using Eqs. C.1, C.4, C.6, and C C.2 Predicted efficiency using Eqs. C.1, C.5, C.6 and C C.3 Predicted efficiency using Eqs. C.1 and C C.4 Predicted efficiency using Eqs. C.1, C.11, and C C.5 Predicted overall filtration efficiency using Eq. C C.6 Predicted efficiency using Eqs. C.4, C.6, C.9 and C C.7 Predicted filter efficiency using Eqs. C.12 and C C.8 Predicted filter efficiency using Eqs. C.13 and C C.9 Predicted efficiency using Eqs. C.4, C.6, C.9 and C C.10 Predicted efficiency using Eqs. C.12 and C C.11 Predicted efficiency using Eqs. C.13 and C C.12 Predicted efficiency using Eqs. C. 12 and C C.13 Predicted filter efficiency using Eqs. C.13 and C C.14 Predicted efficiency using Eqs. C.4, C.6 and C.9 and C xvi

16 C.15 Predicted efficiency using Eq. C.12 and C C.16 Predicted efficiency using Eqs. C.13 and C D.1 Property values for the conservation equations xvii

17 LIST OF FIGURES Figure Page 2.1 Fibrous aerosol coalescing filter Single fiber capture mechanisms by (a) inertial impaction (b) direct interception (c) diffusion and (d) gravitational deposition Slurry mixing tank Sheet former for wet laying glass fiber media Fine wire mesh used for wet laying glass fiber media Fiber glass media made by custom wet laid process. The disks on the right are the raw product of the wet laid process and are cut to 6 cm diameter disks shown on the left Permeability testing set-up for compressed media Wettability and coalescence Porosities of media at different compressions Permeability of uncompressed and compressed media Total specific fiber length per unit volume of filter Total specific surface area per unit volume of filter media 41 xviii

18 3.11 Pore size Distance between fiber contact points Process diagram of the gas-liquid coalescence experiment Lab scale coalescence filtration setup a) Surge tank and b) prefiltration setup a) Laskin nozzle and b) mixing chamber a) Filter holder parts and b) assembly a) Pressure transducer, b) rotameter and c) HEPA filter SMPS and CPC Layered media are composed of five layers of the same media type a) SS, b) glass Steady-state E, dp, FI, S for layered uncompressed media and no drainage channels Schematic of mechanism to compress stainless steel fiber media in filter holder Steady-state E, dp, FI and S for compressed layered media layers of a) glass and b) SS augmented with c) wire mesh drainage channel a) layered media with no drainage channel b) layered media with drainage channel between each layer a) Steel b) polypropylene (microscopic picture) 62 xx

19 4.15 Steady-state E, dp, FI and S for drainage augmented layered media a) polypropylene b) Nylon c) PTFE Steady-state E, dp, FI and S for layered media augmented with different drainage materials Steady-state E, dp, FI and S for layered media augmented with same drainage material with different pore size Summary of all filter media tested with an efficiency exceeding 80% Sequence of steps in shrinkage experiments The shrinkage of the media thickness (SH) versus the volume fraction of the liquid ε L. Each plot is for a different liquid: (a) Sullube-32, (b) Diesel, and (c) Viscor The shrinkage of the media thickness (SH) versus the volume fraction of the liquid in the filter. The plots indicate the SH for glass fibers of sizes: (a) 2μm, and (b) 6μm Plot of maximum shrinkage versus the fiber diameter Strain vs. applied stress for filter media used to determine Young s modulus Maximum shrinkage values versus the Bond number Visual representation of single fiber efficiency Visual depiction of droplets captured by direct interception Visual depiction of droplets captured by impaction Visual depiction of droplets captured by diffusion.. 99 xxi

20 6.5 Gas containing particles flowing through a rectangular bed of fibers Gas containing particles flowing through a cylindrical bed of fibers xxii

21 CHAPTER I INTRODUCTION 1.1 Motivation Coalescing filters are used to remove small liquid droplets from air streams. They have numerous industrial applications including dehumidification, cabin air filtration, compressed air filtration, metal working, crankcase ventilation (CCV), and agriculture [1-4]. In the coalescence process an aerosol passes through a nonwoven microfiber filter medium to coalesce the droplets into larger drops. The inlet aerosol droplets are captured on the fibers of a medium, the collected droplets migrate through the fiber medium and coalesce to form enlarged drops. The enlarged drops are dragged through the medium by the gas flow and drain from the exit surface of the filter medium [5, 6]. The performances of coalescing filters are measured in terms of separation efficiencies and pressure drops. The separation efficiency is dependent on properties of the aerosol (e.g. composition, density, viscosity, drop size. It also depends on the filter properties including the fiber surface wetting properties, fiber size, fiber orientation, porosity, binder content, and filter bed length [7-13]. The gas flow rate is important as it controls the mechanism of droplet capture on the fibers, the inertia force of drop-drop collision and coalescence, and the drag force on the coalesced drops to cause the drops to migrate through the filter [10-14]. 1

22 Glass fiber media are commonly used as coalescing filter media. However, new nonwoven polymer, carbon, metal, and ceramic fiber materials are being developed. Nonwoven steel fiber media with comparable fiber diameters to glass fiber media have been produced and have unique properties that may improve coalescence performance [15,16]. For this work, new non-woven steel fiber media with comparable fiber diameters to glass fiber media have been tested. Steel fiber media have several differences in material properties from glass fiber media that may prove advantageous in coalescence performance and filter design. The steel fibers in the media evaluated in this work were sintered and hence did not have a glue or binder to hold the fibers in the media together. The glass fiber media often are fabricated with a binder to hold the fibers together and give the media its rigidity. 1.2 Problem Statement An aerosol is a suspension of solid particles or liquid droplets in air. Air is the bulk transportation medium for transmission of particulate contaminants. The control over airborne solid and liquid contaminants, hazardous biological agents, allergens and pollutants is a key issue in many industries like metal-cutting, automobile, semiconductor, food, pharmaceuticals and biotechnology. These fields require centralized air conditioning in production environment, clean gases and effluent/waste treatment. The particle size matter is determined by the process that generates the particles. The processes that generate liquid aerosols are include mechanical atomization, evaporationcondensation, and entrainment by gas flow in liquid-gas contactors and CCV. By these processes the typical aerosol droplets of interest have sizes ranging from about 0.01 µm to 50 µm [1-3,17-24] and represent a significant waste stream as well as a health hazard 2

23 to humans. Many of the fluids that are generated into aerosols by these processes affect the health of the people when they are exposed for long durations of time. Respiratory illnesses associated with the inhalation of the above mentioned aerosols include respiratory irritation, bronchitis, occupational asthma, and loss of lung function [1]. Several studies have also shown increases in cancer of the esophagus, stomach, pancreas, larynx, colon and rectum due to prolonged exposure to the above mentioned aerosols [4]. In addition to the health hazards, these aerosols can cause other problems as well. In compressed air systems, oils used for lubrication of compressor parts can aerosolize into the main air stream causing potential contamination concerns for downstream applications. Deposition of oil drops from the aerosol can disrupt flowmeters, pressure gauges and other hardware designed to handle dry air. The oil can also cause corrosion of valves, piping and tubing leading to safety issues. In many systems, humid air can present problems to sensitive equipment and sensors. As the humid air cools, small water drops condense and can disrupt components that need to be kept dry. Condensed water droplets can also cause corrosion issues as well. If these aerosols are not properly handled, they can lead to increased maintenance costs for the equipment. As one of the most common methods of separating and removing particles in micro and sub-micro size ranges, aerosol filtration by fibrous filters has found such diverse applications as in the protection of humans and delicate devices from exposure to hazardous fine particles. It has therefore been the object of intense research, both theoretical and experimental [25-70]. Glass fiber media are commonly used as coalescing filter media. However, new non-woven polymer, carbon, metal, and ceramic fiber 3

24 materials are being developed as health concerns of silica fibers are being more fully understood [71-73]. 1.3 Hypothesis Typically, glass fiber media has been used as media in coalescence filters. However, new non-woven steel fiber media with comparable fiber diameters to glass fiber media have been produced. Steel fiber media have several differences in material properties from glass fiber media that may prove advantageous in coalescence performance and filter design. The steel fibers in the media evaluated in this work were sintered and hence did not have a glue or binder to hold the fibers in the media together. The glass fiber media often are fabricated with a binder to hold the fibers together and give the media its rigidity. This leads to a difference in porosity and permeability. The permeability of the media affects the pressure drop across the media during filtration process. The porosity is related to the pore size and fiber spacing which affects droplet capture by the media. In some filter designs, the fibrous filter media can be compressed during filter fabrication. An example is when fiber sheets are wrapped onto a cylinder to form a spiral-wound filter cartridge. By adjusting the tension on the sheet during the wrapping different compressive forces can be applied to the media. Being more ductile, the steel fiber media can be compressed without breaking the fibers, to change properties of the media such as permeability and porosity. Stainless steel and glass also have different wettability characteristics towards liquids due to their different surface energies. The different wetting characteristics affect how droplets interact with the fibers once captured, coalescence in the fiber matrix and how they drain out of the media. Wettability of the fiber media can affect both the capture efficiency and pressure drop. 4

25 1.4 Research Objectives Following are the primary objectives of this research work Characterize permeability, porosity and wettability of nonwoven stainless steel and glass fiber filter media Conduct aerosol coalescence filtration tests using nonwoven stainless steel and glass fiber media Determine the effect of permeability, porosity and wettability on filtration performance Determine if nonwoven stainless steel fiber media is an acceptable alternative to nonwoven glass fiber media Use single fiber filtration theory to model efficiency of nonwoven fibrous media and determine important parameters 1.5 Benefits of Current Work Fibrous nonwoven coalescing filters are used to remove aerosolized droplets from air streams in many different applications. Traditionally, nonwoven fiber glass media have been used due to the availability of fiber glass diameters small enough to remove aerosol droplets less than 1 μm. New rules and regulations regarding the concentration of air particulates acceptable for safe operation conditions have driven the market to produce more efficient filters using finer diameter fibers. However, as the fiber diameter decreases, this leads to another concern for safe operating conditions for workers. Manufacturing of nonwoven fiber glass coalescing filters requires significantly handling of loose fibers by workers. Handling of these loose fibers can easily cause them to 5

26 become suspended in air. This can cause of number of health issues. While most industrial places require workers who handle these fibers to wear respirators and other personal protection equipment, most of the time the respirators that are worn are simple dust masks. As the need to make more efficient filters drives the need to use smaller diameter fibers, simple dust masks cannot block these fibers from being inhaled. Once inhaled, these smaller fibers also tend to stay inside the lungs instead of being exhaled. It is also well known than inhalation of silica particles can cause silicosis and other respiratory illnesses. Also, contact with fiber glass can cause significant skin irritation. If alternative nonwoven fibrous coalescing filters can be developed and these filters can meet the efficiency demands for reducing aerosol particulate concentrations, it will give manufacturers more choices of which fiber materials to use in making fibrous filters. While nonwoven fiber glass coalescing filters have been proven to meet the demands of protecting workers from inhaling aerosol particulates that can cause health issues as well as protecting downstream sensitive equipment, it would be even more beneficial to develop coalescing filters that use fiber materials that are less likely to be hazardous to the workers who manufacture them. 1.6 Dissertation Outline a) Chapter I outlines the motivation behind the work, problem statement and hypothesis of the work, research objectives and approach of the work, and benefits of the work b) Chapter II outlines background information on aerosol filtration and industrial applications of coalescence filtration c) Chapter III explains fabrication and characterization of filtration media 6

27 d) Chapter IV explains the experimental setup for aerosol coalescence filtration testing, how to characterize filtration performance, and results of aerosol coalescence filtration tests e) Chapter V explains determination of shrinkage of fibrous media due to internal capillary forces, maximum shrinkage and modeling f) Chapter VI is the modeling of efficiency of fibrous filters using single fiber theory from a shell balance approach. Data from experimental results will be used to data fit parameters g) Chapter VII is the modeling of efficiency of fibrous filters using single fiber theory from a multiphase transport approach. Data from experimental results will be used to data fit parameters h) Chapter VIII presents the conclusions in aerosol coalescence testing of nonwoven fibrous filter media i) Chapter XI presents the recommendations of future work in aerosol coalescence testing of nonwoven fibrous filter media 7

28 CHAPTER II LITERATURE REVIEW This chapter reviews basics of filtration, coalescence filtration, aerosol filtration, sources liquid aerosol generation and industrial applications of these types of filter media 2.1 Overview of Filtration Filtration is a mechanical process in which a dispersed phase is separated from a continous phase by allowing it to flow through a porous material. The dispersed phase can be a solid or liquid and the continuous phase can either be a liquid or gas. The filtration process can either remove unwanted impurities from the continuous phase or to recover expensive dispersed phase particulates from the continuous phase. When the continuous phase flows through the filter medium, the dispersed phase particulates are captured by the filter media. The filter medium is the core of any filtration process. It is defined as a permeable medium which allows the continuous phase to flow through while capturing the dispersed phase particles. The flow through the medium is due to a pressure gradient between the inlet and outlet surfaces. The efficiency of the process depends on the effectiveness of the filter in removing impurities. The selection of filter media is an 8

29 important aspect of filtration. Depending on the type of continuous and dispersed phases, filtration is divided into solid-gas, liquid-gas, solid-liquid, and liquid-liquid filtration. 2.2 Types of Filtration Filtration can be classified as surface or depth filtration. When the dispersed phase particle size is bigger than the pore size of the filter medium, the filtration occurs at the inlet surface and no particle can pass through the medium. This is called surface filtration. When the dispersed phase particle size is less than the pore size of the filter medium, the particles penetrate into the filter and get captured internally by several different mechanisms. 2.3 Coalescence Filtration Coalescence filtration is a specific type of depth filtration. Coalescence is the agglomeration and growth of small liquid droplets to form bigger drops. In coalescence filtration, dispersed liquid droplets are removed from a continuous phase. The dispersed liquid enter with the continuous phase, are captured internally by the filter due to different mechanisms, coalesce with other droplets and enlarge, and eventually drain out of the filter due to gravity once they reach a critical size. 2.4 Aerosol Coalescence Filtration Aerosol coalescence filtration is a specific type of coalescence filtration. An aerosol is a suspension of solid particles or liquids droplets in air. The control of airborne solid and liquid contaminants, biological agents, allergens and pollutants is a large concern in industries focusing on metal cutting, automobile, semiconductor, food processing, 9

30 pharmaceuticals, and biotechnology. These processes have strict rules regarding particulate matter in air for their centralized air systems, process gas, and waste effluent treatment systems. The processes that generate liquid aerosols are discussed in the next section. 2.5 Sources of Liquid Aerosol Liquid aerosols are generated from a variety of different processes of which some are listed below. Liquids is contact with high rotation rate equipment have enough mechanical energy transferred to shear them into small droplets [17]. They can also be generated from impaction and centrifugal forces [18]. Metal working fluids are used to lubricate components during metal cutting processes [1,19] and droplets from these lubricants are dispersed into the air during operation due to shearing action between moving surfaces. Average droplet size formed due to this type of process is about 5-8 μm [20,21]. Lubrication oils are used in compressed air systems and during the compression process, tiny droplets of oil are introduced in the air stream. Removing liquid aerosols from gas streams are important processes in industries such as petroleum refineries, natural gas industries, and nuclear power plants. At sufficiently high temperatures, a liquid can evaporate and spontaneously re-condense around locations of slightly lower temperatures which allow for growth of liquid nuclei [22]. The heat is generated during high temperature metal working processes and the oil droplets can condense on dust particles to form liquid droplets [23]. Dust particles from combustion processes are typically in the nm range but when they combine with 10

31 oil, water or solvent droplets, they agglomerate to form larger particles. The agglomerate particles may be broken down into smaller particles and released into the air but it is difficult to break down the particles any smaller than 0.5 μm [2]. Flowing gases in spray towers, cooling towers, and plate columns can entrain droplets from the liquid phase [21]. Compressed air is important in many industrial applications including pneumatic conveying, spray paint equipment, laboratory air, gas separator systems, aeration in pharmaceutical and chemical processes and air bearings for mechanical power transmission. Water droplets entrained in the air stream are μm while oil droplets entrained are μm [3]. These impurities in the air are harmful to workers, affect downstream processes and equipment. Filtration is used to improve process stability and reliability of equipment. Liquid aerosols are also generated in internal combustion engines and are released into the environment through crank case ventilation. The crank case encloses many of the moving parts of the engine that are lubricated by oil. Oil droplets are formed due to the shearing action of the moving parts inside the crankcase. Also, gases inside the combustion chamber leak out of the piston and enter the crankcase. Oil droplet size distribution is μm and soot particles are μm [24]. Regulations require the removal of these impurities to achieve overall emissions standards for the engines. The other applications in filtration in the automotive industry are tank venting, engine air intake, engine exhaust, cabin air, coolant system, gearbox oil and engine oil. The typical aerosol droplet sizes from these applications have a size range from 0.01 μm to 50 μm. While the larger drop sizes are easily removed by simple filter media, it is the 11

32 drop sizes at the smaller end of the spectrum that have been the focus of new filter media development. 2.6 Fibrous Aerosol Coalescing Filters Fibrous coalescing filters are a common type of filtration media used to remove small liquid droplets from air. In fibrous coalescence filtration, liquid droplets carried by the flowing gas are captured by the fibers in the filter medium. Ensuing droplets carried by the gas collide and coalesce with prior drops. Coalescence occurs when two more liquid drops collide with sufficient energy to overcome the surface energy [74]. The coalesced drops grow in size in the filter medium until drag force is enough to cause them to migrate through the filter to the outlet surface. At the outlet surface, the drops continue to coalesce until they are big enough that gravity causes them to drain out of the filter [75]. The critical size for these processes to occur depends on the local velocity, fiber size, interfacial tension, and contact angle. It has been observed that if the contact angle is of the drop on the fiber is low, drops do not break away from the fiber but is conducted along the length of the fiber [6,76-79]. An example of a fibrous aerosol coalescing filter is shown in Figure

33 Figure 2.1. Fibrous aerosol coalescing filter. Increase in the separation efficiency is usually accompanied by an increase in pressure drop [80]. When a critical mass of liquid is trapped in the filter, the pressure drop quickly rises and then stabilizes as an equilibrium between drainage rates and capture rate is established. After this point, the liquid mass inside the filter remains a constant value [81]. The separation efficiency depends on the drop size, liquid viscosity, gas velocity, gas pressure, gas temperature, structure of the filter, fiber diameter, packing density, fiber surface properties and filter thickness. The surface tension of the liquid has a crucial impact on the formation of droplets [82]. There is a list of filtration specific and application oriented properties important for the selection of filter media [83]. The filter performance is characterized by the combined performance of pressure drop and separation efficiency. Fibrous aerosol coalescing filters are very effective at removing liquid droplets down to 0.1 μm. High Efficiency Air Particulate (HEPA), is 13

34 made of nonwoven fiber glass filter media. According to the standards of most industries, it removes 99.97% of airborne particles of 0.3 μm. 2.7 Capture Mechanisms As droplets pass through the fibrous filter medium, they are capture by four main capture mechanisms. These mechanisms are called the single fiber capture mechanisms. The capture of droplets on a single fiber is considered to be dependent upon the local flow conditions around the fiber and when the filter properties are uniform, the capture of droplets on a single fiber is assumed to representative of the entire filter medium. Another assumption is that all the fibers in the filter medium are of the same diameter. This concept is widely used in filtration industry because of its consistent results with experiments. All of these mechanisms act simultaneously at any point and one mechanism may be predominant due to the size of the droplet [25,82]. Figure 2.2 shows the four main capture mechanisms. 14

35 Figure 2.2. Single fiber capture mechanisms (a) inertial impaction (b) direct interception (c) diffusion (d) gravitational deposition a) Inertial Impaction: When a droplet has a sufficient size, it may have enough inertia that as is it being transported by the gas streamline, its inertia will cause the droplet to deviate from the streamline as it bends around the fiber. If this causes the droplet to come within a distance of the drop s radius to the fiber, it will be captured. This mechanism is dominant for drop sizes larger than 1 μm. b) Direct Interception: Droplet sizes between μm follow the air streamline as it bends around the fiber. If this streamline comes within a distance of the drop s radius to the fiber, it will captured c) Diffusion: Droplets less than 0.3 μm have a tendency to follow Brownian motion due to collisions with air molecules. This Brownian motion is superimposed on 15

36 the streamline in which it is carried and increases the chances it will be captured on fibers. d) Gravitational Deposition: Droplet sizes between μm have enough mass that they drop from the air streamline due to gravitational forces. This vertical motion increases the likelihood of capture on a fiber. A comprehensive view of fibrous filtration and its theory and forces controlling the filtration process are given in several texts [25,84-86]. Liquid saturation and pressure drop across the filter media are important parameters which need to be measured and controlled to optimize filter operation life. Filter life and performance can be improved by lowering saturation and pressure drop. 2.8 Saturation of Fibrous Filter Media Liquid saturation is the volume fraction of the void space occupied by liquid. It is dependent on many of the above listed parameters and it directly impacts the pressure drop across the filter and local velocity within the filter. Saturation is usually measured by weighing the mass of a dry filter before operation and then weighing the mass of the wet filter after operation. The difference in mass is the amount of liquid holdup in the filter. Converting the mass of the liquid into a volume by using the liquid s density, the average saturation is defined as the ratio of liquid volume to void volume. In general, the greater the saturation the larger the pressure drop across the filter holding all other parameters constant. Local saturation has been investigated by models and experiments [87-89]. Continuum models show that saturation profiles affect the pressure drop by restricting flow and 16

37 increasing local velocity [89]. Experimental measurements show the local saturation in a filter of uniform properties is higher near the inlet and outlet surfaces and lower in the interior [87-91]. The increase in the local saturation at the exit is not fully understood but it is suspected that unbalanced capillary forces holding the liquid to the fibers creating resistance to flow of liquid down the exit surface of the filter. Authors [92-94] consider drainage and saturation in terms of the Washburn equation. At boundary of the filter, saturation builds up due to barrier effect: this barrier being unbalanced capillary forces pulling the drops back into the filter. This barrier creates a resistance to liquid flow and results in higher liquid saturation at the exit. 17

38 CHAPTER III FABRICATION AND CHARACTERIZATION OF FILTER MEDIA Characterization of filter media was one of the objects of this research work. This chapter focuses on fabrication of media, characterization techniques, and comparison of properties between different filter media 3.1 Filter Media Fabrication The 316 stainless steel (SS) fiber media (Bekaert, Belgium) with seven different fiber sizes in the range of μm were tested for coalescence performance. Two sizes of glass fiber media (Hollingsworth and Vose (H&V), USA), 2 and 6 μm, were tested. Table 3.1 lists the basic information for the different media. 18

39 Table 3.1. Media information. Supplier Fiber Material Nominal Fiber Diameter Mass/Area (g/m2) Medium Diameter Medium Thickness (μm) (cm) (mm) Bekaert SS Bekaert SS Bekaert SS Bekaert SS Bekaert SS Bekaert SS Bekaert SS Bekaert SS H&V Glass H&V Glass All of the SS fiber media and the 6 μm glass fiber media were supplied as sheets of commercial media having sheet thicknesses 0.2 ± 0.05 cm or 0.4 ± 0.05 cm. The SS media were fabricated of sintered steel wires of uniform intrinsic density and contained no chemical binders. The 6 micron glass fiber media sheets contained an unspecified small amount of binder for mechanical strength. 2 micron glass fiber media were made by mixing five grams of loose fibers in 6 L of water to form a uniform slurry, adding 5 ml of binder (Megasol S50 and starch, Wesbond Corporation) and using a custom 19

40 made wet laid process to form media of 0.2 ± 0.05 cm thickness as discussed in Section Wet Laid Formation of Glass Fiber Media 2 micron glass fibers with shortcut lengths of 0.25 inch are supplied in bulk from Hollingsworth and Vose. To construct a filter media, a slurry was made using the following recipe and the mixing apparatus shown in Figure 3.1 Start with 6L of filter water in a container Begin agitation at moderate speed Add 5 ml of HCl to the water Check that the ph is between 2.5 and 3. Weigh out 5g of glass fibers on a balance Slowly add the glass fibers to the water solution. Break off small clumps of fibers when adding to the solution for best dispersion. If clumps are too big, the fibers will not break apart and there will be significant clumping. Add 0.5 g of corn starch to slurry. The corn starch helps the binder adhere to the fibers. Mix slurry for 30 minutes. Add 5 ml of Megasol S50 (Wesbond Corp.) binder to the slurry. Continue to mix for another 30 minutes. 20

41 Figure 3.1. Slurry mixing tank. After the slurry has been prepared, it is then wet laid using a sheet former apparatus shown in Figure

42 Figure 3.2. Sheet former for wet laying glass fiber media. 22

43 The slurry is poured in the top of the sheet former. At the bottom of a top chamber a fine mesh that use to separate the glass fibers from the water and allows the formation of the filter media. This mesh is shown in Figure 3.3 Figure 3.3. Fine wire mesh used for wet laying of glass fiber media. After the slurry is poured in the top of the upper chamber of the sheet former, it is allowed to settle for a minute. Once the slurry has settled, the valve on the bottom chamber is opened, allowing the water to pass through the mesh in Figure 3.3 while the fibers form a uniform cake on the mesh. Once the cake has been formed and the water drained, the top chamber is opened and the cake removed. This cake is soft and soggy due to residual liquid in the cake. The cake is then removed very carefully from the wire mesh and placed in the oven for heating. The cake is heated at 100 C for 2 hours. The 23

44 heating is for two purposes: to dry the remaining water and to cure the binder. The curing of the binder is what gives the filter media its strength so it does not break apart during filtration testing. Glass fiber media made by this process are shown in Figure 3.4. Figure 3.4. Fiber glass media made by custom wet laid process. The disks on the right are the raw product of the wet laid process and are cut to 6 cm diameter disks shown on the left. 3.2 Media Characterization Permeabilities were measured using a Frazier Permeability Tester, porosities were measured using a custom made pycnometer, and wetting properties were measured using 24

45 a drop shape analyzer (DSA20 Easy Drop, Krüss, Hamburg, Germany). Characteristics of nonwoven fiber matrix for stainless steel fiber filter media were analyzed as well Porosity Measurement Techniques A mass method and a pycnometer method were used to determine the media porosities. The mass method is applicable for a medium of one material (ie, no binder) with known intrinsic fiber density. The pycnometer method, based on gas displacement and the ideal gas law, is applicable to media with multiple materials or materials of unknown relative mass quantities. Both methods are applied to the SS media and only the pycnometer method is applied to the glass fiber media. For the mass method the porosity is calculated from the mass of the medium and its bulk volume using the formula ε = V void = 1 M SS (3.1) V Filter ρ SS V Filter where ε is the porosity (void volume), MSS is the mass of a medium, ρss is the intrinsic density of the SS fibers and VFilter is the macroscopic volume of the filter medium. The intrinsic density of SS fibers were assumed to be 8 g/cm 3 [95] and also checked by measuring the weight of SS rod of known physical dimensions (density was calculated to be 7.72±0.06 g/cm 3 ) and by a water displacement test of the same rod (density was calculated to be 7.72±0.41 g/cm 3 ). A custom made pycnometer was used to measure the changes in pressures during the gas displacements. The changes in pressure corresponded to changes in chamber volumes calculated using the Ideal Gas Law and from which the porosity was calculated [96,97]. 25

46 When a medium is compressed its thickness decreases. The cross-sectional area of the filter, the mass of fibers, and the intrinsic density of the fibers are constants. Equation 3.1 can be manipulated to obtain (1 ε)l = constant (3.2) where L is the thickness of the medium. A compressive force was used to change the medium thickness L. By knowing the porosity and thickness of an uncompressed medium, then the average porosity of a medium compressed to a different thickness can be calculated by Eq We define ε a, ε L and ε f to be the volume fractions of the air, liquid and fibers within the media. These quantities are related through the definition of volume fraction and porosity by the expressions ε a + ε L + ε f = 1 (3.3) ε a + ε L = ε (3.4) As the thickness L changes, each of the volume fractions change as constrained by Eqs Permeability Measurement Techniques The permeabilities of the uncompressed media were measured using a Frazier Permeability Tester and calculated using Darcy s Law in Eq Q A = k P μl (3.5) 26

47 To measure the permeability of compressed media, a custom made filter holder was fabricated that compressed the sample while air flowed through the filter media at a measured rate and pressure drop. Figure 3.5 shows the schematic of the filter holder, flow meter, and pressure gauge. Media samples were tested at compression levels of 25%, 50% and 75% of its original thickness. The media samples were placed between two pieces of steel hexagonal mesh of appropriate thicknesses that compressed the samples uniformly as the holder was assembled and bolted together. The permeabilities of the hexagonal meshes were very high so the pressure drops were only due to the fiber media itself. Figure 3.5. Permeability testing set-up for compressed media Wettability Measurement Techniques The wettability of the surface of the fiber to the liquid phase is important in the coalescence of the droplets in the filter, and hence controls the overall performance of a 27

48 medium. With high wettability, the droplets adhere strongly to the fiber surface and form a thick film around the fiber. Droplets readily coalesce to form bigger droplets but the equilibrium saturation will be higher. With low wettability, droplets do not adhere to the fiber surface. Drops are less likely to be captured and this will reduce efficiency. Intermediate wettability is normally the desired range because drops will adhere to the fiber somewhat and can be captured but the saturation is lower which reduces the pressure drop [11]. The impact of wettability on droplet capture is shown in Figure 3.6. Figure 3.6. Wettability and coalescence. It was not possible to measure directly the contact angle Sullube-32 on the fibrous media themselves due to the morphology of the fibrous surfaces making it was impossible to get clear or consistent images to measure the contact angles. The individual fibers were too small in diameter and length to be practical for measuring contact angles. Hence, flat smooth surfaces of representative materials similar to the 28

49 fibers, were used to measure the contact angles. The liquid was the same as that used in the coalescence experiments, Sullube-32 (Diversified Air Systems, Ohio, USA). The contact angles were optically measured using a drop shape analyzer (DSA20 Easy Drop, Krüss, Hamburg, Germany). The contact angles on representative materials at best give an indication of the contact angles of the fiber materials in the media. The representative materials may not be true representations. The crystal structures in SS metal grains in the flat sheet may differ from that of the fibers due to differences in manufacturing. If we attempt to melt the SS fibers to create a flat surface, the heating of the material would alter the crystal structures and affect the measurement. The glass fibers are amorphous and can be melted. But measuring the contact angles on a flat surface of melted glass fibers would not account for the effects of the binders present in the filter media. Other methods such as capillary rise can give an indication of contact angles but rely in fitted parameters and also have large uncertainties. Because of the large uncertainties no further measurements of the contact angles on the representative materials were attempted Nonwoven Structural Properties In addition to bulk filter media properties such as porosity and permeability and fiber properties such as wettability, the nonwoven characteristics of the fiber matrix can directly impact filtration performance as well. The first critical parameter of nonwoven characteristics is the total specific fiber length LT and is expressed in mm/mm 3. It is also known as the basis fiber length. It is defined as 29

50 L T = i L i = 10x ig i (3.6) T i L Where xi is the mass fraction of the fiber in the medium, G is the basis weight in g/m 2 listed in Table 3.1 Ti is the linear density of the fiber in dtex, and L is the thickness of the medium in mm. The relationship for Ti in dtex to the fiber diameter is given by [98] 2 T i = 7855ρ f d fi (3.7) where df is measured in mm and ρf is measured in g/cm 3. Eqations 3.6 and 3.7 can only be used for stainless steel fiber media because glass fiber media has an unknown amount of binder which prohibits the determination of G and ρf. Also, since there is only one fiber diameter in the medium, there is only one ith component per medium and the subscript can be removed and Eqs. 3.6 and 3.7 reduce to L T = 10G TL (3.8) T = 7855ρ f d f 2 (3.9) A second parameter in characterizing the nonwoven characteristics of a fiber matrix is the total specific fiber surface area. The total specific surface area represents the total surface area of fibers per unit volume of filter medium and is obtained using the following relationship in mm 2 /mm 3 SS = i SS i = L ip i i (3.10) 10 3 where Li is the specific fiber length as described in Eq. 3.6 and Pi is the perimeter of the fiber in μm. The perimeter of the fiber can determined by 30

51 P i = πd fi (3.11) Eqs and 3.11 can only be used for stainless steel fiber media because glass fiber media has an unknown amount of binder which prohibits the determination of G and ρf. Also, since there is only one fiber diameter in the medium, there is only one ith component per medium and the subscript can be removed and Eqs and 3.11 reduce to SS = L TP 10 3 (3.12) P = πd f (3.13) The final parameter in characterizing the nonwoven fiber structure is the average pore diameter. This represents the size of air zones in microns inside the fibrous structure based on the entanglement of fibers and is based on the White model [99] d p = d f (1 α) α (3.14) where α is the solid volume fraction and df is the fiber diameter in microns. The solid volume fraction α is related to the porosity ε by α = 1 ε (3.15) Combining Eqs and 3.15 d p = d fε 1 ε (3.16) Different models have been developed in order to predict the compressional modulus which account for several fiber characteristics such as orientation, fiber to fiber contact 31

52 angle and crimp level. Various parameters were introduced to account for crowd factor, friction and slippage phenomena [99,100]. One of the first compressive studies based on microstructural approach of fibrous materials was made by Van Wyk [101] assuming that the governing mechanism was the bending behavior between fibers for a random distribution in the fiber matrix. Friction between fibers was assumed to be insignificant for the compressive behavior of the fiber structure [99]. They also developed a geometric approach to estimate the average distance between contact points inside a fibrous structure and the number of contact points between fibers. This approach was completed by Komori and Makishima [102,103] who were able to predict the structural characteristics such as mean number of contact points, pore size distribution, and fiber orientation distribution. Pan then derived his own approach with a modified theory of compressive analysis of general fiber assemblies which was criticized by Komori and Makishima through a modification of their own solution [103]. According to Komori and Makishima analysis, the mean distance between the contact points is defined as [102,103] b = V 2d f L V I (3.17) where df is the fiber diameter, LV is total fiber specific length within the filter volume V and I is the mean value of the sine of the angle between the connecting fibers. I is π/4 for a randomly oriented fibrous structure. LT defined in Eq. 3.8 is the total fiber specific length per unit volume of the filter so by inspection L V = L T V (3.18) Combining Eqs and

53 b = 2 πd f L T (3.19) According to Pan et al. [104], the number of fiber contact points in a filter volume V is n V = d fl V 2 I V (3.20) where again df is the fiber diameter, LV is total fiber specific length within the filter volume V and I is the mean value of the sine of the angle between the connecting fibers. I is π/4 for a randomly oriented fibrous structure. Combining Eqs and 3.20 n V = πd fl T 2 V 4 (3.21) 3.3 Results and Discussion The porosities of the uncompressed media are listed in Table 3.2. The measurements were made in triplicate and the +/- error indicated as one standard deviation of the three values. The pycnometer and mass basis methods gave consistent porosity values for the SS fiber media. 33

54 Table 3.2. Uncompressed porosity values. Fiber Material Porosity (ε) ( - ) Fiber Diameter (μm) Pycnometer Mass Basis Eq.(1) SS ± ±0.000 SS ± ±0.000 SS ± ±0.000 SS ± ±0.000 SS ± ±0.000 SS ± ±0.000 SS ± ±0.000 SS ± ±0.000 Glass ± Glass ± All of the SS fiber media had porosities of 99% using both methods. The glass fiber media had porosities of 96% as determined only by the pycnometer method. Since the SS media had the higher porosities, they were expected to hold more oil and take longer to reach steady saturation. This could allow SS media to last longer during aerosol filtration. The 2 μm glass fiber media permeability were about 3 times smaller than the same size SS fiber media. The permeability of the 6 μm glass fiber media were about 10 times smaller than the 6.5 μm SS media. These differences could be due to the differences in porosities and differences in fiber diameter distributions. From microscope inspections the SS fibers had a narrower size distribution than the glass fiber media even though the average sizes were about the same. 34

55 The porosities of the compressed media were calculated using Eq The compressions of the media were quantified by the compression ratio given by Compression = L o L L o (3.22) where L o is the uncompressed thickness and L is the compressed thickness. The porosities for compression ratios of 0%, 25%, 50% and 75% are plotted in Figure 3.6. Figure 3.7: Porosities of media at different compressions. 35

56 The 0%C media porosities were measured by the pycnometer for glass fiber media and calculated by Eq.3.1 for the SS media. The porosities of the compressed media were calculated using Eq.3.2. Figure 3.7 shows the porosities of the tested media reduced due to the reduction of media thickness by compression. When the fiber volume fractions doubled (50%C), for the stainless steel fiber media, the reductions in porosities were only about 1~2% due to the high uncompressed porosities of the SS samples. By comparison, the reductions in porosities of the glass fiber media were similarly only about 4% when the fiber volume fractions doubled. Permeability of uncompressed media was measured using a Frazier Permeability Tester. Permeability measurements of compressed media were made using the set up in Figure 3.5. Each measurement was made in triplicate and error bars represent one standard deviation. The permeability of uncompressed and compressed media are shown in Figure

57 Figure 3.8. Permeability of uncompressed and compressed media. Not unexpected, the uncompressed permeability increased with the fiber diameter. Figure 3 shows a nearly proportional relationship between the reduction in permeability and the amount of compression. When the media thicknesses were compressed by 25%, 50% and 75% of its original thickness, the permeability values were reduced by about 36-54%, 61-77% and 83-90%, respectively. The reductions in permeability were expected because the compressions caused reductions in the void volumes The properties of the Sullube-32 are listed in Table 3.3. The contact angles were measured for Sullube-32 on flat surfaces of representative stainless 316 and glass surfaces. The flat surface materials were not made of exactly the same materials as the 37

58 fibers and some variation may occur due to small differences in chemical compositions and surface smoothness. The contact angles were measured at least three times and the values reported in Table 3.3 are the averages of the multiple measurements. The +/- errors are one standard deviation of the multiple values. Table 3.3. Properties of Sullube-32 and its contact angle on representative SS 316 and glass surfaces. Fluid Density (kg/m 3 ) Viscosity (kg/m-s) Surface tension (N/m) Contact Angle (deg) Stainless 316 B-glass Sullube ± ± 2.75 It was determined that stainless 316 is more highly wetting of Sullube-32 than glass due to its lower contact angle. In fibrous filter media, how strongly the liquid drops adhere to the fiber surfaces determines the sizes of the coalesced drops leaving the media and pressures required for the gas to flow through the media loaded with liquid [11]. Low contact angles favor large coalesced drops and high liquid saturation. Large contact angles result in little if any coalescence and low pressure drops. Intermediate contact angles favor the best combined performances with highest FI values. The nonwoven characteristics of stainless steel fiber media were also calculated and compared for the different fiber diameters at different compression rations. The total specific fiber length per unit volume of filter is plotted in Figure

59 Figure 3.9. Total specific fiber length per unit volume of filter. From Eq. 3.9, it is seen that LT is not dependent on either the solidity or porosity of the fiber media. Essentially, Eqs. 3.8 and 3.18 say that the total length of fiber is constant but compressing the fiber media just increases the length of fiber per unit volume of the filter. In essence, this is another way of plotting the mass balance for the fibers in the filter media. However, from Figure 3.8, it is can definitely be determined that smaller fiber diameters have greater length of fibers available for droplet capture than do larger fiber diameters. Inspection of Table 3.1, Eqs. 3.8 and 3.9, for filters with the same GSM and same overall area, the smaller fiber diameter filter media have more fibers in the media than do media made with larger fiber diameters. This makes sense intuitively. For an individual fiber of two different diameters made of the same material with the same density, for them to have the same mass they must have the same volume. Since the volume a cylinder is dependent 39

60 on df 2 and L, to achieve the same volume, a cylinder with a smaller df must have a larger L than a cylinder with a larger df. Since fibrous media are made from shortcut fibers of the same length, by extension of the previous two points, to have the same mass of fibers, there must be a greater number of fibers in the media. Essentially this can be written mathematically as 1 4 n 1πd 2 f1 L f = 1 4 n 2πd 2 f2 L f (3.23) In Eq. 3.23, n is the number of fibers, df is the fiber diameter, and Lf is the length of the fiber and subscript 1 refers to the smaller diameter while 2 refers to the larger diameter. If the length of each of the fibers are the same, then Eq reduces to n 1 2 = d f2 n 2 (3.24) 2 d f1 From Eq. 3.24, if df2 is larger than df1, then n1 must be larger than n2. The total specific fiber surface area is plotted in Figure The total specific fiber surface area of a filter is useful when blending different fiber diameters together. 40

61 Figure Total specific surface area per unit volume of filter media. From Eq. 3.12, it is seen that SS is not dependent on either the solidity or porosity of the fiber media. Essentially, Eq say that the total surface area of fibers is constant but compressing the fiber media just increases the surface area of fiber per unit volume of the filter. Again, this is another way of plotting the mass balance for the fibers in the filter media. However, from Figure 3.9, it is can definitely be determined that smaller fiber diameters have greater fiber surface area available for droplet capture than do larger fiber diameters. As stated above, since the smaller fiber diameter media have a more fibers, they will also have more fiber surface area available for capturing droplets. 41

62 In Figure 3.11 and 3.12, the pore size and distance between fiber contact points are plotted. The pores size and distance between fiber contact points were calculated using from the porosity values measured using the techniques discussed in Section Figure Pore size. 42

63 Figure Distance between fiber contact points. From Eq. 3.16, the pore size is related to the porosity. From Figure 3.11, as the compression ratio increases, the pore size decreases. This makes sense intuitively because as the fibers are compressed, they will pack in more closely and touch together in more places decreasing the pore size. This is supported by the distance between contact points in shown in Figure Inspection of Eq shows that as LT increases, the distance between contact points should decrease. As stated previously, compressing the media does in fact increase the total length of fiber per unit volume of the filter media. Therefore, the data in Figures 3.11 and 3.12 are reasonable. This data also supports the intuitive fact that squeezing a fibrous material will cause the fibers to pack in more closely and decrease the pore size. The decrease in pore size due to compression also correlates well with the 43

64 decrease in permeability due to compression as shown in Figure 3.8. As the pore size decreases, it is more difficult for air to travel through the pore. This increase in resistance is reflected in a lower permeability as the permeability in Darcy s Law shown in Eq. 3.5 can be considered as the inverse of resistance to flow. 3.4 Summary of Results The permeability and porosity of stainless steel and fiber glass media were measured both using the mass balance and custom made pycnometer which uses a gas expansion method to determine void space. Using the continuity laws, the porosity for compressed stainless steel media was determined. Essentially, the mass balance says for a given fibrous media with a fixed mass of fibers, the volume of fibers is also constant. During compression, the ratio of volume of fibers to volume of filter increases, decreasing the ratio of volume of voids to volume of filter. The permeability of stainless steel and fiber glass media were measured in both uncompressed and compressed states. For compressed media, there is a decline in the permeability. This decline in permeability can be attributed to the decrease in pore size of the fibrous media due to compression. Compressing a fibrous media causes the fibers to pack in more densely, decreasing the distance between fiber to fiber contact points and increasing the resistance to air flow. 44

65 CHAPTER IV AEROSOL COALESCENCE FILTRATION TESTING Aerosol coalescence filtration testing of filter media is one of the objectives of this research work. This chapter focuses on the experimental setup for aerosol coalescence filtration testing, how to characterize filtration performance, and the results of filtration tests. 4.1 Coalescence Experimental Setup and Performance Quantities The aerosol coalescence experiments were performed using the filtration apparatus shown in the schematic in Figure 4.1. A photo of the lab scale setup of coalescence filtration and its major components are shown in Figures 4.2 and 4.3 respectively. The process is described in four sections in this chapter. Figure 4.1. Process diagram of the gas-liquid coalescence experiment [105]. 45

66 4.1.1 Prefiltration To ensure there is no contamination in the experimental filtration test stand, an extensive prefiltration system is used. Atmospheric air contains various impurities and water vapor. This atmospheric air supplies the compressed air system used in the test stand. The compressed air system also adds various contaminants into the air supply that need to be removed to ensure accurate testing results. Those impurities are removed before challenging the air to the coalescing filter medium. Since the compressed air system supplies high pressure air to the whole building, there can be fluctuations in supply pressure depending on the usage by other equipment in the building. A surge tank is used to dampen effects of fluctuations in house air pressure on the experiment. The compressed air initially passes through a drier to remove particulates and moisture, then through HN2L-10CD coarse filter, HN2L-6CD fine filter; DN4L-SG4 silicon gel desiccator and HN2L-AUD filter to remove micron and submicron size dust and dirt particles. All these filters are Parker Hannifin Corporation filters. The surge tank and prefiltration system are shown in Figure 4.3 Figure 4.2. Lab scale coalescence filtration setup. 46

67 a) b) Figure 4.3. a) Surge tank and b) prefiltration setup Aerosol Generation After removal of impurities and moisture, the purified pressurized air is then divided in to two streams. The main stream goes directly to the filter holder and the side stream goes through a Laskin nozzle to generate the aerosol. A Pressure drop of 25 psig needs to be maintained across the Laskin nozzle to generate droplets of average diameter of 300 nm. The aerosol is generated by dispersing fine droplets of compressor oil Sullube 32 which is manufactured by Dow Chemical Company and contains % of polypropylene glycol with a density kg/m 3. The Laskin nozzle generates the droplets by forcing high pressure air through a small diameter tube, which increases the gas velocity. The tube is submerged into the oil and the high velocity gas causes the oil to 47

68 aerosolize into a fine mist upon impact of the gas with the liquid. The oil droplets generated in the Laskin nozzle are mixed with the main air stream upstream of the coalescing filter assembly in the mixing chamber and the aerosol is challenged to the filter. The Laskin Nozzle and the mixing chamber are shown in Figure 4.4. Figure 4.4. a) Laskin nozzle and b) mixing chamber Filter Holder The filter holder assembly and its parts are shown in Figure 4.5. The filter holder is constructed of stainless steel and is equipped with a spacer; hence the filter samples of different thicknesses can be tested. A stainless steel wire mesh is placed behind the filter medium to protect the medium from deforming due to air flow as well as to protect the downstream equipment from fiber shedding if the filter does break apart. A threaded aluminum plug is fixed at the end to keep the filter medium in its position. The liquid collected in the filter holder can be drained using drainage ports at various locations in the filter holder. The filter holder assembly without a filter was tested to understand 48

69 particle loss and change in pressure drop. The filter holder without a filter indicated atmospheric pressure drop and similar upstream and downstream droplet size distribution. The loss in the droplets was not detectable [105]. Figure 4.5. a) Filter holder parts [105] and b) assembly Measurement Equipment An electronic pressure gauge is used to measure the pressure drop across the filter media. Pressure drop is continuously monitored and recorded every 30 minutes during the experiment. The air flow is measured by using a rotameter. To protect the needle valve inside the rotameter, the downstream air passes through the HEPA filter before entering the rotameter. Since the filters tested in the setup have an unknown efficiency, this precaution is taken to prevent damage to the rotameter and losing control over the flowrate. After the air passes through the rotameter, it is vented. The air flow rate is maintained 60 SCFH at to obtain a face velocity of 0.1 m/s. At this flowrate, all media 49

70 were tested with a Reynolds number less than unity (Re<1) which is the laminar flow regime. The Reynolds number is defined as Re = ρvd f μ(1 ε) (4.1) where Re is the Reynolds number, ρ is the density of air, v is the superficial velocity, µ is the viscosity of air and ε is the porosity of the filter medium. Figure 4.6. a) Pressure transducer, b) rotameter and c) HEPA filter. A Scanning Mobility Particle SizerTM (SMPS) (TSI Inc. Model No. 3080) is used to monitor the upstream and downstream droplet size distributions. From these distributions, the inlet and outlet concentrations with respect to the filter can be measured and determined. The SMPS is setup to complete one droplet size distribution measurement in 135 seconds and it can count up to 1 million droplets at the same time. To use the equipment below its threshold limit and not damage the equipment, the sample stream is diluted with air at a known dilution ratio using rotameters. The TSI 3080 has three major components known as Electrostatic Classifier, Differential Mobility Analyzer (DMA) and Condensation Particle Counter (CPC). The intake flow rate of the Electrostatic Classifier 50

71 is termed as sheath flow rate while the sample air flow rate is termed as the aerosol flow rate. The sheath flow rate is always 1/10 times the aerosol flow rate. The electrostatic Classifier normalizes the charge on the droplets, if any, by using a krypton-85 bipolar charger. The charged droplets then enter the DMA. The droplets are segregated according to their electrical mobility. The DMA passes one size of droplets at one time to the CPC depending on their electrical mobility. The CPC counts the number of droplets of a particular size that are sent from the DMA. The CPC uses N- Butanol to condense on the droplets to increase their size to a size that can be detected by a laser. The air stream from CPC is further vented. Readings for the inlet and outlet concentrations are measured every 30 minutes. The Aerosol Instrument Manager software supplied by TSI converts the drop size distribution into a concentration that is used for analysis of the efficiency. The software also gives droplet size distributions based on number, length, surface area, volume, and mass. The mass distribution is the one that is used to calculate the concentration used to determine filter efficiency. 51

72 Figure 4.7. SMPS and CPC Filtration Performance Characterization The coalescence performances were characterized by the capture efficiencies and the pressure drops. The capture efficiencies were calculated based on the measured total mass concentrations of the aerosol up and down stream of the filter test sample shown by 52

73 the Eq As mentioned previously, the concentrations were measured regularly during filtration experiments. E = 1 C outlet C inlet (4.2) The efficiency and the pressure drop performances of the media are combined into one measure of performance referred to as the Filtration Index [106], Figure of Merit [107,108] and Quality Factor [25] as defined by FI = ln(1 E) P (4.3) The saturation of a filter media, S, quantifies how much of the void space is occupied by liquid. It is defined as S = εl ε (4.4) In Eq. 4.4, ε L is the volume fraction of liquid in the filter as defined by ε L = V VL (4.5) Filter The volume of the liquid is related to the mass of the liquid by V L = ml (4.6) ρl The mass of the liquid is determined by measuring the filter before and after filtration testing. The difference in mass is the mass of the liquid held within the filter. Once the 53

74 mass of liquid is determined, Eqs can be used to determine the saturation. In applying Eqs , the porosity of the filter is measured by techniques discussed in Chapter 3. The volume of the filter is a directly measureable quantity. 4.2 Aerosol Coalescence Filtration Testing Aerosol coalescence filtration testing was performed on stainless steel and glass nonwoven fibrous filter media listed in Table 3.1. Different parameters such as porosity, permeability, fiber material and drainage augmentation were varied to study their impact on filtration performance Layered Media The media listed in Table 3.1 were stacked as layers shown in Figure 4.8 and tested for coalescence performance. The drop size distributions were measured upstream and downstream every 30 minutes using an SMPS. The steady state concentrations were used to calculate the steady state E. The steady state pressure drops and E were used to calculate the steady state FI. The experiments were run in triplicate and the average steady state E, dp, FI and S values are plotted in Figure 4.9 for the various media samples. The error bars are one standard deviation of the averaged values. Figure 4.8. Layered media are composed of five layers of the same media type: a) SS, b) glass. 54

75 Figure 4.9. Steady-state E, dp, FI, S for layered uncompressed media and no drainage channels. For this configuration of layered, uncompressed media with no drainage channels, 2 and 4 µm media performed very poorly, especially the 4 µm with efficiency below 20%. This is more than likely due to the thicknesses of the 2 and 4 micron SS layered media decreasing and depth change in the media can affect the filter performance. There are many reasons why shrinkage can occur in woven and nonwoven fabrics [ ] and the cause of this shrinkage phenomena in this particular case was investigated in a separate set of experiments. The results of these experiments are covered in Chapter 5. Since the smaller fiber diameters allow for greater capture of drops, even though both 2 and 4 µm media undergo depth shrinkage due to capillary forces, 2µm media had a higher capture efficiency than 4µm. Since the calculation of saturation is dependent upon knowing the volume of the filter, the saturation values were calculated using the original thickness. Since the actual thickness was smaller but hard to measure accurately, the 55

76 listed saturation in Figure 4.9 is not accurate. The capillary shrinkage effects were not observed with the other media in these experiments. The 2 μm glass, 5 μm SS and 6.5 μm SS media had the highest average efficiencies of about 85%. The SS media had significantly lower pressure drops resulting in much higher FI than the 2 μm glass media. Saturation of the media were relatively the same in the range of about 30 to 40%. From this, conclude that the lower pressure drop of SS media is due to its much higher permeability than glass fiber media due to the absence of chemical binder. In the next tier, 6 micron glass, 8 micron SS and 12 micron SS had efficiencies between 70%-80% and again the SS media had a lower pressure drop Compressed Layered Media Since SS media is made from ductile fibers, the media can be mechanically compressed to change its porosity and permeability before saturation from coalescence filtration. Compression ratio of 50% and 75% were tested for coalescence performance. With the current filter holder design, it is very difficult to test 25% compression. Compression of layered media changes the fiber matrix. As the fiber matrix compresses in on itself, both the porosity and permeability decrease. While the decrease in permeability may be a concern since it will increase the pressure drop, the reduction in porosity and pore size could help to more efficiently capture and remove drops from the air stream. To investigate the effect of compression, the 2, 4, 6 and 8 micron fiber diameter SS media were tested for coalescence filtration performance at two compression levels: 50% and 75%. To compress the media, the locking ring shown in Figure 4.5 was tightened 56

77 further down on the spacer ring to compress the media. To get the correct compression ratio, first the lock ring was tightened until it just touched the spacer ring. At this point, a caliper was used to measure the distance from the back of the locking ring to the back edge of the filter holder on the left side of Fig After this distance was measured, the locking ring was tightened further, moving the spacer ring inwards to compress the media. After tightening the locking ring one turn, a caliper was used to measure how far inwards the locking ring moved and therefore how much the thickness of the media decreased. This was done until the locking ring moved the correct distance to get proper compression ratio. This distance was determined from the original thickness of the layered media and the definition of compression ratio in Eq To assure that force applied to the spacer ring by the locking ring was applied uniformly to the fiber media, a large pore wire mesh was placed between the filter media and the spacer ring. A schematic diagram of how this process looks like is shown in Figure

78 Figure Schematic of mechanism to compress stainless steel fiber media in filter holder. Again, the steady state concentrations were used to calculate the steady state E. The steady state pressure drops and E were used to calculate the steady state FI. The experiments were run in triplicate and the average steady state E, dp, FI and S values are plotted in Figure 4.11 for the various media samples. The error bars are one standard deviation of the averaged values. 58

79 Figure Steady-state E, dp, FI and S for compressed layered media. As expected, compressed media have a higher steady-state pressure drop because of lower permeability due to compressing the fiber matrix. The impact of compression of pressure drop is twofold. The first is that compression of fibrous media decreases the pore size of the filter increasing the permeability of the dry filter media as discussed in Chapter 3. The second is the decrease in pore size also affects the ability of captured liquid to drain from the media. The smaller the pore size, the stronger the capillary forces inside the filter media which make it more difficult for liquid to drain due to gravity. This increases the saturation of the media. Since the presence of the liquid in the void spaces 59

80 restricts the air flow, this also causes an increase in pressure drop. These two combined are why all media compressed to 75% have the highest pressure drop. Also from Figure 4.11, for a constant fiber diameter, the efficiency always increases when the filter media is compressed. Also for a constant fiber diameter, the efficiency is the same within the standard error of the experiments whether they are compressed 50% or 75%. For the 2 and 4 micron stainless steel fiber media, the saturation is nearly 100% at both 50% and 75% compression. For 6.5 and 8 micron stainless steel fiber media, the saturation of the media increases with increasing compression ratio. From this, it can be concluded that the efficiency is not dependent upon the saturation for the media tested. Since the efficiency is independent of the compression ratio but the pressure drop is highly dependent on the compression ratio, the best performance is achieved at 50% compression. At this compression ratio, the increase in efficiency due to compression offsets the increase in pressure drop due to compression resulting in higher FI. In most applications, it is acceptable to trade a slight increase in pressure drop for an increase in filter efficiency Drainage Augmented Layered Media Layered media can also be augmented with drainage channels in between each layer as illustrated in Figure

81 Figure layers of a) glass and b) SS augmented with c) wire mesh drainage channel. In coalescing filters, the captured liquid must move through the filter and down the exit surface before it collects at the bottom of the filter. The presence of this liquid reduces the space available for gas flow and thus causes an increase in pressure drop. Furthermore, filters with horizontal flow often have a small part at the bottom of the media that is completely saturated through which no gas can flow. This restriction causes the gas to move through a reduced area with a greater face velocity can decrease the efficiency and increase the pressure drop. The presence of drainage channels provide greater pore openings to allow faster drainage and can reduce some of the saturation [122]. Drainage channels allow pathways for coalesced liquid to rapidly flow out of the filter without compromising the capture efficiency. This decreases the saturation of the liquid in the fiber media, reducing the pressure drop and increasing the FI as illustrated in Figure 4.13 [122, 123]. 61

82 Figure a) layered media with no drainage channel b) layered media with drainage channel between each layer. Initially, two types of drainage materials were tested. One was a hexagonal large 7.3 mm pore steel wire mesh and the second was a woven polypropylene (PP) mat with pore size of 1000 μm. They are shown in Figure Figure a) Steel b) polypropylene (microscopic picture) Physical properties such as pore size and air flow permeability are listed in Table 4.1 and wettability properties are shown in Table

83 Table 4.1. Physical properties and air flow permeability. Material Property SS 316 PP [123] Fiber Size (μm) Pore Size (μm) Thickeness (μm) Permeability (m 2 ) N/A 3.13E-09 Table 4.2. Wettability properties. Sullube-32 Contact Angle (deg) Material Flat Surface Fibrous Surface SS N/A PP [123] From Tables 4.1 and 4.2, the steel drainage channel has larger pore size opening and negligible air flow resistance. The polypropylene drainage channel has some air flow resistance due to smaller pores but it has a lower wettability towards the oil being captured by the fibrous media. The contact angle on the actual wire mesh surface could not be measured due to its large pore opening size. The steady state concentrations were used to calculate the steady state E. The steady state pressure drops and E were used to calculate the steady state FI. The experiments were run in triplicate and the average steady state E, dp, FI and S values are plotted in Figure 4.15 for the various media samples. The error bars are one standard deviation of the averaged values. 63

84 Figure Steady-state E, dp, FI and S for drainage augmented layered media. Augmentation of layered media with large pore wire mesh and polypropylene drainage channels improved the efficiencies and the filtration indexes compared to layered media for all filter media tested. For 2 micron SS media, comparisons of the drainage augmented media against the layered media is not really a valid comparison for reasons mentioned previously. However, comparing the 2 micron SS media augmented with steel wire mesh drainage channels versus the PP drainage channels, the filter augmented with the PP drainage channels had a higher efficiency and saturation and pressure drop. This indicates that changing the wettability of the drainage material did not affect the drainage 64

85 rate. The increase in performance with the lower wettability drainage material may possibly be attributed to the fact that the PP material is more of a resistance to capillary forces between layers than the steel mesh. For 2 micron glass, the addition of drainage layers increased both the efficiency and the saturation. This indicates that adding drainage materials did not significantly affect the drainage rate. In addition, using a lower wettability drainage material did not improve drainage performance as the efficiency and saturation was higher when using PP drainage materials versus the wire mesh drainage channels. For 6 micron glass, when adding wire mesh drainage materials, the drainage rate did not change significantly from layered media because both the efficiency increased and the saturation increased. When adding the PP drainage materials, the drainage rate was significantly increased because the efficiency was much higher but the saturation the same as when using wire mesh drainage materials. For 6.5 micron SS media, the addition of drainage materials greatly improved the drainage rate. For both drainage materials tested, the efficiency increased while the saturation decreased. The wettability of the drainage material didn t significantly affect the drainage rate as the 6.5 micron SS media augmented with PP drainage materials had slightly higher efficiency and saturation than wire mesh augmented 6.5 micron SS media. The media augmented with drainage channels did not shrink. The strength of the capillary force is dependent on the pore size and the large pore sizes of the wire mesh reduced the capillary forces between the layers and reduced the observable shrinkage of the media. Further investigation was done on augmenting layered media with different drainage materials. Patel et al investigated the effect of drainage material on performance and found that using fiber materials that were lipophillic and drainage channel material that 65

86 was lipophobic yielded the best performance. However, that work was limited to monolayer media fabricated with only one fiber material type [122,123].The work here with drainage materials was conflicting in the relationship between fiber wettability and drainage material. However, two parameters were changed: pore size and wettability. Both of these parameters can affect the drainage rate. For a better comparison, the pore size should be held constant and the wettability varied. When different materials are added to fibrous media to facilitate drainage, the wettability of both the drainage material and the fibrous filter media can impact how coalescence occurs inside the fiber matrix and how they drain out of the media. If the wettability gradient between the drainage material the filter media are low, coalescence and drainage may not significantly change but the saturation may increase which can impact performance. If the wettability gradient is too large, liquid may not want to leave the higher wettability fibrous material which will hinder drops entering drainage material, creating localized zones of high saturation which can negatively impact performance. In addition to the above, for fibers with the same wettability, smaller fiber diameters will be more likely to hold captured liquid than larger fiber diameters. Therefore, the wettability gradients between fiber material and drainage material may also depend on fiber size in the fibrous filter. By trying different fibrous filters and drainage materials, these hypotheses can be tested. Three more types of drainage materials are available for testing. They have have the same pore size and are made of three different materials: Nylon, Polypropylene and PTFE. Microscope images are shown in Figure

87 Figure a) Polypropylene b) Nylon c) PTFE. Physical properties such as pore size and air flow permeability are listed in Table 4.3. Table 4.3. Pore size and air flow permeability. Material [123] Property Nylon PP PTFE Fiber Size (µm) Avg. Pore Size (µm) Thickness (µm) Air Permeability (m 2 ) 3.39E E E-10 Wettability properties are shown in Table 4.4. Table 4.4. Wettability properties. Sullube-32 Contact Angle (deg) Material Flat Surface Fibrous Surface Nylon [123] 4 6 Polypropylene [123] PTFE [123] The goal of this work is to test two types of fibrous filter media, stainless steel and glass, with three different types of drainage materials, nylon, PP, and teflon. Stainless steel is 67

88 more wetting towards oil than glass. Nylon has a high wettability with oil, polypropylene is partially wetting with oil, and teflon is non-wetting with oil. There will also be different wettability gradients between stainless steel and the three drainage materials compared to glass and the three drainage materials. By testing each fiber type with each drainage material, impact of fiber, drainage and interaction between fiber and drainage wettability can be tested. The steady state concentrations were used to calculate the steady state E. The steady state pressure drops and E were used to calculate the steady state FI. The experiments were run in triplicate and the average steady state E, dp, FI and S values are plotted in Figure 4.17 for the various media samples. The error bars are one standard deviation of the averaged values. The media was layered with three different drainage materials in the same exact way as shown in Figure

89 Figure Steady-state E, dp, FI and S for layered media augmented with different drainage materials. For 2 micron SS fiber media, there is no clear indication of the effect of wettability of the drainage material. For 2 micron glass, the drainage rate is almost bell shaped with respect to the wettability of the drainage material. Nylon and Teflon have the best drainage rates and they have the highest and lowest wettability. Polypropylene has an intermediate wettability between the two and it has the lowest drainage rate. These results are interesting. For 6 micron glass, lower wettability equals a better drainage rate. The same is true for 6.5 micron SS. 69

90 For both stainless steel media types tested, the best efficiency was obtained using polypropylene drainage materials. Taking into account the average pressure drops and the standard deviations, there was no difference in steady-state pressure drop when using different drainage materials with stainless steel media. For the 2 micron glass media, taking into account the averages and standard deviations, the steady-state efficiency and pressure drop did not vary with variation of drainage material used. For 6 micron glass media, the highest efficiency was achieved using polypropylene and Teflon drainage materials. Similar to the stainless steel media, pressure drop did vary with variation of drainage materials. Overall, 2 micron stainless steel media with polypropylene drainage materials and 2 micron glass media with all three drainage materials had the best efficiency. At this same efficiency, the 2 micron SS media had a pressure drop about 10 times smaller. This is mainly due to the permeability of the 2 micron SS media being much less than that of the 2 micron glass. This permeability difference is due mainly to how the media are constructed. As stated previously, SS media do not contain binder and this lack of binder greatly reduces the inhibition of air flow through the porous media. Looking at the standard deviations of the experiments for each combination, most of this current work is in line with previous coalescence filtration of these types of filter media. For 2 micron SS media with nylon drainage channels, the standard deviation is very high. However, this due to the fact that nylon is highly wetting towards oil and it was previously determined that for SS fiber media with diameters less than 4 micron, performance is negatively impacted due to shrinkage of the filter phenomena due to capillary forces inside the media caused by captured liquid. Since nylon is highly wetting towards oil, this same shrinkage phenomena was observed in this case and therefore the 70

91 results are consistent with those previously. For 2 micron glass fiber media with all three drainage materials, the efficiency results were consistent but the standard deviation on the pressure drop was high. This is most likely due to the process by which the media are made. While all the other fibrous are made commercially by a standardized industrial process, the 2 micron glass media were made in house by a custom made process. This process is not nearly as consistent and there is some amount of variation in the media made batch to batch. The main sources of variation in the process is in the slurry formation, custom wet laying procedure and binder curing. If the slurry is not completely uniform, fibers will clump and not wet lay uniformly, which also impacts binder adhesion and curing. This will impact permeability variation which is the main driver of pressure drop variation. At a constant drainage material wettability, the impact of pore size of the drainage material used on drainage can be compared. For this comparison, the data is presented in Figure In this Figure, the steady-state E, dp, FI, and S are presented for filter media augmented with 1000 micron polypropylene and 500 micron polypropylene drainage materials. 71

92 Figure Steady-state E, dp, FI and S for layered media augmented with same drainage material with different pore size. In Figure 4.18, no matter what type of fibrous media is tested, the ones augmented with the smaller pore size drainage material have a higher saturation compared to those augmented with the larger pore size. This is a direct indication that the drainage is better when using a larger pore size on the drainage material. For fibrous media with a fiber diameter of 2 microns, this drainage rate does not seem to have an impact on the filtration performance. The efficiency and pressure drop do not vary with the variation in drainage rate. For fibrous media with a fiber diameter of 6 microns, the filtration performance is negatively impacted when the drainage rate decreases. At lower drainage rates, the efficiency decreases and the pressure drop either changes slightly or increases slightly 72

93 which both are negative in terms of filtration performance. This suggests that teflon and nylon drainage channels with 1000 micron pore size may offer better performance than that shown in Figure Summary of Results Stainless steel and glass fiber media were tested for aerosol coalescence filtration performance in several different configurations. The most basic configuration was a simple layering or stacking of media to a specified length. Stainless steel samples containing fibers whose diameters ranged from 2-22 μm were tested as well as fiber glass samples with fiber diameters of 2 and 6 micron. In some cases, stainless steel media performed as well as glass fiber media and in other cases they did not. The reasoning for this will be discussed in Chapter 5. Stainless steel media also operated at a slightly higher saturation due to its greater wettability with the testing fluid. Despite its higher saturation, when the efficiencies were the same, stainless steel media had a lower pressure drop because its dry permeability was much lower than glass fiber media due to stainless steel media not containing any chemical binder. Taking advantage of stainless steel s ductility, four different stainless steel fiber media were compressed to two different thicknesses and tested for filtration performance. Across the samples tested, compression improved the capture efficiency. Compressing the sample too far, however, negated the increase in efficiency with too much of an increase in pressure drop, causing a decline in FI. Layered media was also augmented with woven drainage materials in between the layers of fibrous material. For a more direct comparison between stainless steel fiber media and glass fiber media, only stainless steel fiber media with fiber diameters the same as the glass fiber media were tested. For drainage materials, woven 73

94 materials of different pore sizes, materials and wettability were chosen. For glass fiber media, lower wettability drainage materials yielded optimum performance. For stainless steel fiber media, intermediate wettability drainage materials yielded the best performance. For drainage materials with the same wettability, the larger pore size yielded the best performance. Figure 4.19 summarizes all the media types and configurations that had a filtration efficiency exceeding 80%. In many filtration applications, the requirements for filtration are more demanding requiring an efficiency exceeding 90%. Several filters tested in this work reached this plateau and in some cases with a minimal pressure drop. These results are promising but more research into coalescence filtration using these materials still needs to be done. Figure Summary of all filter media tested with an efficiency exceeding 80%. 74

95 It has been demonstrated that stainless steel fiber media can achieve the same filtration efficiency as fiber glass media in many instances. In all cases, stainless steel fiber media has a lower pressure drop due its much higher permeability. Overall, the best performance tested was a stainless steel fiber media: when 6.5 micron stainless steel is compressed 50%, its filtration efficiency is 94% with only a 0.07 kpa pressure drop yielding a FI of 40 kpa

96 CHAPTER V SHRINKAGE OF FIBROUS FILTER MEDIA DUE TO INTERNAL CAPILLARY FORCES Aerosol coalescence filtration testing on layered filter media revealed that for small fiber diameter stainless steel filtration media, saturation by oil causes the media to shrink considerably from its original thickness. This chapter reviews experimental method to determine shrinkage, reasons for the shrinkage, and characterization of the phenomena. 5.1 Introduction Many media made of fine fibers are fabricated using binders to hold the fiber structure together and provide rigidity [12, ]. The binders can reduce media porosity and alter surface properties of the fibers which can adversely affect filter performance. Typically, when glass or polymeric fiber media are wet laid to form non-woven filter media, chemical binders are applied to hold the fiber structures together to prevent fiber shedding during filtration operation. Often the binders provide enough rigidity to offset internal capillary forces and shrinkages due to capillary forces are not normally observed or considered in filter performance [ ]. However, stainless steel fiber media do not apply chemical binders but apply heat to partially melt or soften the surfaces of the fibers 76

97 to cause the fibers to fuse at their points of contact.. This fusing mechanism is enough to prevent fiber shedding but the media structures are soft and compressive enough thatcapillary forces may cause significant changes to the macroscopic thickness when wetted by captured liquids. Shrinkage can occur in woven and nonwoven fabrics and filter media by several mechanisms including thermal [109, 110] and chemical reactions [111], and through polymer chain relaxation in polymer fibers due to interaction with solvents [112]. Shrinkage of fabrics of several different materials has been reported, including cotton fabrics [113], cotton-polyester blended fabrics [114], silk fabrics [115], wool fabrics [116] and polymer fibrous mats [117]. In regard to fibrous filter media, shrinkage can happen during the manufacturing processes such as sintering [118] or shrinkage can occur when the filter is put into service. For those filter media used in high temperature applications, the heat-shrinkage of the media can significantly change the filter properties [119]. Moreover, shrinkage may also occur due to surface wetting properties and capillary forces during coalescence filtration applications. Mullins [120] noted that some low packing density filter media became thinner during wetting processes. At the microscale Mullins et al [133] observed movement of fibers and fibers being pulled together due to capillary forces between drops and fibers, Hatt [121] concluded that swelling and shrinkage of storm water filter media was the most probably cause for the variation of the infiltration capacity of the filters during wet and dry periods. As far as the authors are aware, no systematic study has been performed to quantify the effects of shrinkage of filter media due to capillary forces. 77

98 Shrinkage is affected by small scale structures in the fibrous media and interactions between the fiber materials and the liquid droplets. The work here only considers the effects of fiber size of media made of layers of nonwoven stainless steel and glass fibers and their interaction with several liquids. 5.2 Filter Media and Testing Liquids The filter samples used in the shrinkage experiments include stainless steel (SS) fibers (with average fiber diameters of 1.5μm, 2μm, 4μm, 5μm, 6.5μm, 8μm, 12μm or 22μm) and glass (Glass) fibers (with average fiber diameters of 2μm or 6μm). The dry and uncompressed porosity of the filter media are summarized in Table 5.1. Table 5.1. Properties of the filter samples used in the shrinkage experiment. All filter media were disk shaped with diameters of 2.5 cm and stacked to dry thicknesses of 1.0 cm. Material Fiber Diameter ε SS SS SS SS SS SS SS SS Glass Glass All of the SS fiber media and the 6 μm glass fiber media were supplied as sheets of commercial media having sheet thicknesses 0.2 ± 0.05 cm or 0.4 ± 0.05 cm. The SS 78

99 media were fabricated of sintered steel wires of uniform intrinsic density and contained no chemical binders. The 6 micron glass fiber media sheets contained an unspecified small amount of binder for mechanical strength. 2 micron glass fiber media were made by mixing five grams of loose fibers supplied in 6 L of water to form a uniform slurry, adding 5 ml of binder (Megasol S50 and starch, Wesbond Corporation) and using a custom made wet laid process described in Chapter 3 to form media of 0.2 ± 0.05 cm thickness. Porosities were determined using methods described in Chapter 3. Three organic liquids and deionized water were used as the liquid phases. The organic liquids included a transmission fluid Sullube-32 (Diversified Air Systems, Ohio, USA), ultra-low sulfur diesel (purchased locally from a commercial supplier, winter blend, used as supplied) and Viscor 1487 (Rock Valley Oil and Chemical Co., Il, USA). The four liquids had a narrow variation in densities. The viscosities of the water, diesel, and Viscor 1487 were similar but the viscosity of the Sullube 32 was much larger. The three organic liquids had similar surface tensions but the surface tension of the water was much larger. Contact angle measurements were made using a DSA20 Easy Drop (Kruss Inc.) on flat plates of 316 stainless steel and glass slides as representative of the material types. Table 5.2 summarizes the properties of the liquids used and their contact angles on the flat surfaces. The material compositions of the flat plates and glass slides may not be exactly identical to the material compositions of the fibers, but they give an indication of the flat surface contact angles of those types of the materials. 79

100 Table 5.2 Properties of liquids used in the shrinkage experiment Density (kg/m 3 ) Viscosity (kg/m-s) Surface tension (N/m) Contact Angle (deg) on flat surfaces of representative material samples Stainless 316 Glass slide Sullube / / Diesel / /- 2.1 Viscor / / Water / /- 2.1 The 316 stainless steel was more hydrophobic and lipophyllic than the glass surface. The low contact angles of the oils on the stainless steel had significant variations resulting in relatively large standard deviations. Measurements of three or more drops were averaged and the +/- values are one standard deviation. 5.3 Experimental Procedure In coalescence filtration, aerosols of fine droplets challenge the filter media. Normally measurements of the media thicknesses can only be made at the end of the experiments when the media are removed from the holder and the thicknesses can be observed. The act of removing the media from the holder deforms the media and interferes with quantified measurements of the thickness. To minimize the error caused by the handling of the media at the end of the filtration experiments, a more direct approach was used to measure the change in thickness due to the presence of the liquid. As indicated in Figure 5.1, liquid drops were applied uniformly over the top surfaces of each medium using a syringe one drop at a time. The diameters of the droplets were about 2mm and a caliper was used to measure the thicknesses of the media as increments of 0.2 ml liquid were applied. This is a modified 80

101 ASTM D method to measure thickness of nonwoven fabrics. ASTM D requires a scale of readability of inch with an accuracy of inch. The calipers used meet these requirements having a scale readability of inch with an accuracy of inch. However, ASTM-D5736 also requires specific dimensions for the thickness testing gauge and means of measuring and controlling the applied pressure when making thickness measurements which the calipers used in this experiment do not meet these requirements. For total compliance with ASTM D , a specific testing thickness gauge must be purchased from Certain-Teed Corporation designed by Spartan Engineering Co. [134]. For the purposes of this work, this modified version of ASTM D was appropriate to give accurate enough readings for thickness measurements. While the pressure applied can have an impact on the thickness measurement of high porosity fabrics, care was taken to eliminate pressure as a variable in the thickness measurement. When making measurements with the calipers, the measurement was taken when the edge of the caliper just touched the surface of the fabric and the pressure applied by the caliper was negligible. A total of 5 ml of liquid were applied on each sample to displace air out of the media by capillary action. Some air bubbles may have been trapped within the media. In the last step of the experiment, the wetted media were submerged into the same liquid as shown in Figure 5.1 to observe expansion as the liquid displaced most of the remaining entrapped air bubbles and the capillary forces were reduced, allowing the media to elastically expand back towards their initial dry thickness. Water did not penetrate into the SS media when wetted drop-wise by a syringe but water did penetrate the glass fiber media and the results for measurements of the glass fiber media for all four liquids are reported. 81

102 Figure 5.1. Sequence of steps in shrinkage experiments. 5.4 Results and Discussion The reduction in thickness of a medium is caused by a compressive deformation of the fiber matrix and corresponds to a reduction of the pore volume of the medium. The average porosities of the wetted and shrunk media were calculated via mass continuity balance discussed in Chapter 3. (1 ε)l = constant (3.2) ε a + ε L + ε f = 1 (3.3) ε a + ε L = ε (3.4) The changes in L due shrinkage by capillary forces causes each of the volume fractions to change as constrained by Eqs Shrinkage (SH) due to capillary forces is defined the same way as compression in Eq. 3.6 SH = Compression = L o L L o (5.1) The media thicknesses (L) were measured for every 0.2ml of liquid added to the media. The liquid volume fraction of the medium is used as defined by 82

103 ε L = VL AL (5.2) where A is the area of the filter medium and L is the thickness of the medium measured with the amount of liquid in the medium. The SH and ε L were calculated and plotted in Figure 5.2 for the stainless steel samples. Figure 5.2. The shrinkage of the media thickness (SH) versus the volume fraction of the liquid ε L. Each plot is for a different liquid: (a) Sullube-32, (b) Diesel, and (c) Viscor Data points are the averages of three replicate measurements. 83

104 As can be seen from the plots in Figure 5.2, the shrinkage phenomena occurred in all of the tested samples when the media were wetted drop-wise by all of the oils. The shrinkage SH increased from 0 to a maximum value when ε L approached 90%. Past this point, the shrinkage began to decrease as the media approached complete saturation. When the drop-wise wetted media were completely submerged in the liquid and completely saturated, as indicated in the final step in Figure 4, most of the air bubbles were displaced from the media and the capillary forces reduced to near zero, causing the media to expand back to approximately the original starting thickness of the dry media. The trend in the shrinkage for the SS media showed smaller diameter fibers had greater shrinkage than the media of larger fibers. The glass fibers had the opposite performance where the larger 6 μm fiber media had the greater shrinkage than the 2 μm fiber media. This may be due to the way the glass fiber media were fabricated. The 6 μm fiber media were formed as layers of thin sheets of media, similar to the SS media. The 2 μm fiber media were fabricated as a thick monolayer. The monolayer media had a small amount of Megasol binder to hold the media together, which may have given the media a different surface property than the binder material used to hold the 6 μm fiber sheets together. The deionized water could not be used to test the stainless steel fiber media because the water did not penetrate into the media. However, water did penetrate into the glass fiber media. Figure 5.3 shows plots comparing shrinkage by the oils and by water for the 2 μm and 6 μm glass fiber media, respectively. 84

105 Figure 5.3. The shrinkage of the media thickness (SH) versus the volume fraction of the liquid in the filter. The plots indicate the SH for glass fibers of sizes: (a) 2μm, and (b) 6μm. Data points are the averages of three replicate measurements. Figure 5.3 shows the shrinkage by water was less than the shrinkage that occurred with the oils. The capillary forces are a function of the surface tension and the contact angle. From the capillary rise equation [135] capillary forces are directly proportional to the surface tension and cosine of the contact angle. In Table 2 the surface tension of water is greater than that for the oils and the typical contact angle for water about the same for the oils on a flat glass surface. Hence we would expect the capillary forces and shrinkage to be similar. However, the glass fiber media contained a small amount of chemical binder used to hold the fiber structure intact. This binder material was hydrophobic and changed the surface properties of the fibers slightly by lowering its wettability. This lead to smaller capillary forces and hence less shrinkage. 85

106 It was also observed for the 2 μm glass fibers that as the liquid volume fraction exceeded about 0.9 the shrinkage value became negative, indicating the media expanded greater than the original dry thickness. It is possible the original dry media had inherent compression in the fiber structure that was released when the media was saturated with water but it is more likely the water may have caused some swelling of the saturated media and thus resulting in a negative SH. The interactions between fiber swelling and capillary forces causing shrinkage are complicated and the negative SH may require further research. All of the media and liquids performed almost the same and have the same shrinkage trend versus fiber size, with exception of the 2 micron glass fibers discussed below. All of the oils have similar densities and surface tensions, and similar wettability with 316 stainless steel. The viscosity of the Sullube 32 was about 100 times higher than the other liquids hence the small variation between the liquids indicates the viscosity had a minor effect, if any, on the shrinkage. The trend in the shrinkage for the oils was significant for fibers less than about 5 microns. To compare how the shrinkages vary between the different experiments, Figure 5.4 shows a plot of the maximum SH values from the data in Figures 5.2 and 5.3 plotted versus the fiber diameter. The maximum shrinkages of the 6 μm glass fiber with the oils fit well with the oil data and the steel fibers but the 2 μm glass fiber media shrinkages were much lower. The 6 μm glass fiber media and the SS fiber media were similarly assembled as layers but the 2 μm glass fiber media was a monolayer structure. This suggests the structure may have contributed to the performance. Many filters are fabricated as composite layers. Larger pore openings or voids can occur at the interface 86

107 between the layers than occur between fibers within the interior of a given layer. These larger pores allow more space for the media to shrink. The monolayer structure does not have theses layers which may limit the extent of shrinkage that occurs. Another consideration is the 2 μm glass fiber media had a binder material added to the fibers that modified the wetting properties of the fibers and added rigidity to the monolayer that resisted the shrinkage. Figure 5.4. Plot of maximum shrinkage versus the fiber diameter. Data points are the averages of three replicate measurements. A more general way of relating the maximum shrinkage to the media properties is through a plot of the shrinkage versus the Bond number (Bo). The Bond number is the ratio of an applied force to the surface tension [136]. 87

108 Typically, the Bo number defined as the ratio of gravitational body force to the capillary force. In this work the gravitational force is replaced with a compressive force applied to the surfaces of the media. The Bond number is modified here as Bo = E bd f γcosθ (5.3) where, Eb is Young s modulus, defined as the slope of the stress versus strain relation [137]. For small stresses, the differential change in strain is related to the differential change in length by dε = dl L (5.4) Which, when holding the denominator L constant, integrates to ε = L L (5.5) However, when deformation takes place in a series of increments, as in the experiments conducted here, the strain path must be taken into account. Integration of Eq.5.4 allowing L to vary yields ε = ln L L o (5.6) The bulk modulus for each material was determined using a simple experiment. Each material was stacked in layers to form disks 2 cm in diameter and thickness of 1 cm. Different weights (forces) were applied to the stack and the change in thickness was measured for each applied weight. The weights were divided by the area of the disks to 88

109 calculate the applied stress. The change in the thicknesses of the stacked layers were measured using calipers. The applied stresses were compressive and negative by definition. The thicknesses of the disks decreased due to the compression hence the strains calculated by Eq. 5.6 also were negative. The stress-strain plots are shown in Figure 5.5. Except for the 6 micron glass fiber media, the data fit reasonably well to the lines passing through the origin. The reason is that the 6 micron glass samples do not have a smooth surface so the force distribution is not uniform. Despite the fit not being the greatest, the slope of the line that minimizes the error is used to estimate the bulk modulus. The slopes of the lines are the values for Eb and are listed in Table 5.3. The trend is that the Young s moduli increase with the fiber diameter for like materials. The moduli for the glass fiber media are larger than for the stainless steel media due to the inherent stiffness of the fiber materials and the binders used in the glass fiber media. The values generally decreased as the fiber diameter increased, for the SS fibers as expected. However, the Eb value of the 2 μm glass fiber media was significantly larger than the 6 μm glass fiber media. This is most likely due to the binder and the less controlled process of forming the 2 μm glass fiber media. 89

110 Figure 5.5. Strain vs. applied stress for filter media used to determine Young s modulus. Data points are the averages of three replicate measurements. Table 5.3. Young s modulus, Eb, for the fiber media Material Fiber Diameter Eb (N/m 2 ) SS ± 10 SS ± 29 SS ± 8 SS ± 44 SS ± 57 SS ± 46 SS ± 44 Glass ± 799 Glass ±

111 Using the Eb values from Table 5.3, the Bo numbers for each media type were calculated. The maximum shrinkage versus Bo number is plotted in Figure 5.6. Figure 5.6. Maximum shrinkage values versus the Bond number. The data for the stainless steel fiber media are shown as circular symbols without distinction between fiber sizes. The data for the glass fibers are distinguished between fiber sizes by the symbol shape. Data points are the averages of three replicate measurements. In Figures 5.4 and 5.6 the organic liquids performed with a similar shrinkage trend versus fiber size and Bond number. The organic liquids have similar densities, surface tensions, and contact angles with 316 stainless steel. The viscosity of the Sullube 32 was about 100 times higher than the other liquids hence the small variation between the liquids indicates the viscosity had a minor effect, if any, on the shrinkage. 91

112 The general trend shows that the shrinkage is largest for small Bo and smallest for large Bo. This trend was expected because Bo represents the ratio of the strength of the material to resist compression divided by the force due to capillaries to cause the media to shrink. For the SS fibers the Bo generally decreased as the fiber diameter increased, as expected, because the strength of the fiber media decreases as the fiber diameter decreased. Hence the data in Figure 9 do not need to distinguish between the fiber sizes for the steel fiber media. The shrinkage data for the glass fiber media did not follow the pattern observed for the stainless steel fiber, hence the fiber size of the glass fibers in the media are indicated in the figure. The Bo numbers were larger for the 2 μm glass fiber media than the 6 μm glass fiber media for each liquid due to the larger values of the product (Eb df). In general, the shrinkage was less for larger Bo compared to smaller Bo for each fiber material type. Due to the binder effects and fiber structure effects on the glass fiber media, no general conclusions can be made for the glass fiber media. The data for the water and the glass fiber media seem to fit the trend of the SS fiber data. The 2 micron glass fiber data appear to fall on the same path if the data for the SS fibers plateau for Bo greater than about 0.5. The SH data for the 6 micron glass fiber media deviate significantly from the SS fiber data. 5.5 Summary of Results The experiments in this work showed layered fibrous media shrank in thickness due to capillary forces when liquid droplets entered the media. Small fiber diameter media 92

113 shrank more than large fiber diameter media. The amount of shrinkage depended on the amount of liquid in the media and on the Young s modulus of the media. The latter was related to a modified Bond number. The viscosity of the liquid did not appear to be a significant factor. Understanding and quantifying this phenomena is important for wet filter applications because the filtration efficiency is dependent upon the thickness of the filter as will be discussed in the next two chapters. As demonstrated by results of the filtration experiments discussed in the last chapter, change in depth of the media by liquid saturation can have a negative impact on the filtration efficiency. The effect of capillary forces acting inside a filter media must be taken into account when designing filtration media. 93

114 CHAPTER VI SHELL BALANCE APPROACH TO MODELING EFFICIENCY 6.1 Introduction and Motivation Various experimental characterization techniques are available to measure filtration efficiency and/or the transport/deposition behavior of aerosols in fibrous filters. A typical system usually includes an aerosol generator, a filtration set and a particle detector. These types of filtration testing setups require a great deal of capital expenditure, maintenance, and cleaning. Each filtration testing experiment can require a great amount of time for a filter to reach steady-state to determine efficiency performance. Therefore, a single filtration testing experiment can become time-consuming and costly. It would be beneficial for a researcher to be able to predict the efficiency of an aerosol coalescence filter from the properties of a dry filter. The theory for efficiency of fibrous filters assumes that the filter contains only fibers of the same diameter. Therefore, only efficiency data from layered and compressed layered stainless steel fiber media will be used for modeling. Glass fiber media used in this work contain a chemical binder which cannot be accounted for in the simple models that would be ideal for a researcher to use. 94

115 6.2 Single Fiber Theory The ability of fibrous filters to collect particles is usually expressed in terms of filtration efficiency, E, the fraction of entering particles that are retained by the filter: E = 1 C out C in (4.2) where Cout and Cin are the outlet and inlet aerosol particle concentration, respectively. Single fiber efficiency, η, is the fraction of upstream particles that collects on a fiber. This process is shown in Figure 6.1 All particles in the region bounded by the upper and lower lines that extend to the left of the of the fiber cross section are eligible for collection because they are upstream of the fiber. The fraction of these particles that collects on the fiber as gas passes the fiber is η. Particles can collect on fibers by several mechanisms. Several of these are discussed below. Figure 6.1. Visual representation of single fiber efficiency [26]. 95