A COMPUTER PROGRAM FOR FILTER MEDIA DESIGN OPTIMIZATION

A Thesis

Presented to

The Graduate Faculty of The University of Akron

In Partial Fulfillment

of the Requirements for the Degree

Master of Science

Sailaja Dharmanolla

August, 2007
 A COMPUTER PROGRAM FOR FILTER MEDIA DESIGN OPTIMIZATION

Sailaja Dharmanolla

Thesis

Approved: Accepted:

_______________________ _______________________
Advisor Department Chair
Dr. George G. Chase Dr. Lu-Kwang Ju

_______________________ _______________________
Committee Member Dean of the Graduate School
Dr. Helen K. Qammar Dr. George R. Newkome

_______________________ _______________________
Committee Member Date
Dr. Bi-min Zhang Newby

ii
 ABSTRACT

In depth filtration, mixtures of nanofibers and microfibers provide efficient and

effective filters for capture of micron and submicron sized particles. Experimental

approaches are time consuming and expensive to design filters. There is a need for

reliable computational models to analyze and evaluate the filter performance. The

objective of this research is to present a computational approach for designing mixed

fiber filter media for depth filtration. The goal is to find an optimum solution, given

specified ranges for a set of design parameters: thickness of the media, diameter of

microfiber, diameter of nanofiber, surface area ratio of nanofiber to microfiber, and mass

of microfiber. The idea is to develop with a software tool that may reduce the number of

experiments.

This program applies a Genetic Algorithm to search for an optimum filter media

design based on Quality Factor which quantifies the filter performance. A user friendly

computer program is developed that provides inputs, outputs and controls to design a

filter media. This program provides a starting point for design of filter media for

particular applications.

Experiments were performed to validate the modeling and experimental data for

comparison. The results show about 25% error in quality factor between computer model

and experiments. More experiments are needed in future work for model validation.
iii
 ACKNOWLEDGMENTS

I would like to express my sincere gratitude towards Dr. George Chase for

extending support and encouragement throughout the course of this degree. I would like

to thank Dr. Helen Qammar and Dr. Bi-min Zhang Newby for serving in my committee. I

am grateful to the Department of Chemical Engineering for providing financial assistance

towards this degree. I thank the Coalescence Filtration Nanomaterials Consortium

members for their research funding and for providing useful insights on this research.

To my loving husband George Atluri who supported and encouraged me in my

research. Thanks to my parents, sisters and brother-in-law for being there when I needed.

Thanks to Multiphase group and friends for all the support.

iv
 TABLE OF CONTENTS

Page

LIST OF TABLES ...........................................................................................................viii

LIST OF FIGURES............................................................................................................ix

I. INTRODUCTION................................................................................................... 1

1.1 Background information ................................................................................... 1

1.2 Research Objectives .......................................................................................... 2

1.3 Technological importance ................................................................................. 3

1.4 Thesis Outline ................................................................................................... 3

II. LITERATURE REVIEW........................................................................................ 5

2.1 Methods of filtration.......................................................................................... 5

2.2 Fibrous Filters ................................................................................................... 6

2.3 Computer models ............................................................................................ 11

2.4 Optimization Techniques ............................................................................... 12

2.5 Incorporation of nanofiber in microfiber ........................................................ 15

III. MODELING AND RESULTS.............................................................................. 17

v
 3.1 Assumptions .................................................................................................... 17

3.2 Particle Capture Mechanism ........................................................................... 18

3.3 Conservation Equations................................................................................... 20

3.4 Model Calculations ......................................................................................... 27

IV. COMPUTER PROGRAM .................................................................................... 31

4.1 Approach ......................................................................................................... 31

4.2 Fitness Function .............................................................................................. 32

4.3 Convergence.................................................................................................... 32

4.4 Input and Outputs ............................................................................................ 34

4.5 Features of program ........................................................................................ 35

4.6 Results ............................................................................................................. 36

V. FILTER MEDIA PREPARATION AND CHARACTERIZATION ................... 42

5.1 Filter Media Preparation.................................................................................. 42

5.2 Filter Media Characterization.......................................................................... 44

VI. EXPERIMENTAL RESULTS AND DISCUSSIONS ......................................... 48

6.1 Coalescence Filtration ..................................................................................... 48

6.2 Results and Conclusion ................................................................................... 50

VII. CONCLUSIONS AND FUTURE WORK ........................................................... 57

vi
 7.1 Conclusions ..................................................................................................... 57

7.2 Future work ..................................................................................................... 59

NOMENCLATURE.......................................................................................................... 60

BIBLIOGRAPHY ............................................................................................................. 62

APPENDICES…………………………………………………………………………...65

APPENDIX A. FORTRAN PROGRAM................................................................. 66

APPENDIX B. VISUAL BASIC PROGRAM ........................................................ 77

vii
 LIST OF TABLES

Table Page

4.1 Results from software program for several particle diameters…………………..36

4.2 Results from software program with different parameter ranges ……………….37

4.3 Results from software program showing the optimum value for area ratio……..37

5.1 Average porosity, permeability and hardness of the filters……………...............47

6.1 Comparison of optimum quality factor from experiment and the computer
program…………………………………………………………………………..54

6.2 Comparison of quality factor from experiment and the optimum from the
FiL2RO computer program……………………………………………………...55

viii
 LIST OF FIGURES

Figure Page

2.1 Methods of filtration ……………………………………………………………...5

2.2 Common air contaminants and their Relative Sizes [5]......................................... .7

2.3 Schematic showing electrospinning set up………………………………………15

3.1 Schematic of Diffusion Deposition………………………………………………..18

3.2 Schematic of Interception Mechanism…………………………………...............19

3.3 Schematic of Inertial Impaction mechanism …………………………………….20

3.4 Relative Quality Factor for different nanofiber sizes and area ratios for
filters of 3 micron microfibers and challenged with 150 nm particles[13].……...29

3.5 Effect of parameters on quality factor…………………………………………...30

4.1 Flowchart for applying a Genetic Algorithm to optimize filter design………….31

4.2 Screenshot of program FiL2RO………………………………………………….34

4.3 Thickness of media vs. Quality Factor and Pressure drop……………………….39

4.4 Mass of microfiber vs. Quality Factor and Pressure drop……………………….40

5.1 SEM image of electrospun nylon nanofiber on a 2 micron glass fiber…………..43

5.2 Vacuum mold assembly………………………………………………………….43

5.3 Pycnometer set up in the laboratory……………………………………...............45

5.4 Frazier’s Permeability set up in the laboratory ..................................................... 46

6.1 Coalescence Filtration set up in the laboratory ..................................................... 48

ix
6.2 Experimental quality factor versus Anf/Amf........................................................... 51

6.3 Quality factor versus Anf/Amf from modeling ....................................................... 51

6.4 Pressure Drop versus Anf/Amf…………………………………………………..52

6.5 Concentration profile versus particle size………………………………………..52

x
 CHAPTER I

INTRODUCTION

1.1 Background information

In many industrial sectors fibrous filters are used to remove undesirable particles

from gas streams. These filters are used in nuclear, pharmaceuticals, chemical industry,

power generation, food processing and beverage industry, electronics, biotechnology,

mining and the pulp and paper industries [1].

Micron and submicron sized particles in air are commonly known to be harmful

to human health and to the environment. One method to remove these particles from

aerosols is by non woven microfiber filters. The performance of a fibrous filter media is

dependent upon the size of particles to be filtered, the structure of the filter media, such

as fiber diameter, the packing density, filter thickness, filtration velocity, and the

viscosity, pressure and temperature of the gas. The addition of nanofibers to the

microfiber media improves the filter performance but design correlations for the relative

amounts of nanofibers to add to the fiber mixture are not commonly available.

In this thesis a computer program is developed that searches for the best values,

within specified ranges, of five design parameters of filter media that optimizes the filter

1
performance. The filter performance is defined by the quality factor which is given by

R.C.Brown[2] which accounts for capture efficiency and pressure drop. The higher the

quality factor the better the filter performance.

The computer program is named FiL2RO (pronounced “fill.too.rho”) which is an

acronym for “Filter Media Local Mechanism 2 Fiber Optimization.” The governing

equations solved by the program are based on volume averaged continuum theory species

and momentum balances. Local single fiber mechanisms for particle capture and drag

force are used to model the interphase terms in the equations. The program assumes the

particles to be captured are monodispersed and the filter is clean (ie. no particles are

loaded into the medium). The program allows the user to select two fiber sizes for which

the fibers are mixed together to form the medium. The porosity of the medium is

calculated for the optimum design and is large (greater than 0.9) making the local single

fiber mechanism assumptions reasonable.

1.2 Research Objectives

1. Develop a computational model to predict the filter performance for depth

filtration.

2. Selection of design parameters of filter media for optimization.

3. Develop a software program to find optimum parameter values of mixed fiber

filter media design using the Genetic Algorithm search technique.

4. Selection of fitness function and convergence criterion.

5. Interpret and analyze the results from the software program.

2
 6. Construct and test filters to measure quality factor with varying amounts of

nanofibers in glass fiber media.

7. Compare model calculations with experimental data.

8. Compare optimum filter performance with optimization program.

1.3 Technological importance

This work shows that existing correlations can be applied with the volume

averaged continuum equations to predict performance trends in filter media. The

computer program can reduce the number of laboratory trial runs of making and testing

filter media. The program can help in studying the trends or improvements in making

filter media that would be costly to find using traditional or experimental methods. We

can predict the effects of filtration properties of non-woven filter media on filter

performance and its efficiency. This work can be extended to quantify the efficiency of

filters for layers of media with different structures.

1.4 Thesis Outline

The thesis is structured into six chapters. The second chapter contains literature

review explaining about the aerosol filters, fibrous filters, characteristics of fibrous filter,

simulation models, introduction to search technique Genetic Algorithm and importance

of nanofiber in filtration. The third chapter explains about the model background, particle

capture mechanism, assumptions, and the equations used to develop filter design

optimization program and model results. Chapter 4 explains about the software program

written in FORTRAN and Visual Basic user interface, inputs and output parameters,

3
features and limitations of the program. Chapter 5 describes the media preparation and

characterization of media. Sixth chapter explain about the coalescence experiments and

the results. Chapter 7 has conclusion and future work followed by appendices.

4
 CHAPTER II

LITERATURE REVIEW

2.1 Methods of filtration

The filtration of fine particles and contaminants present in gases that originate

from process industries, automotive, healthcare, aerospace is paramount to health and

environment [3].

Flow Flow

Dirt loads throughout

Dirt loads at surface the depth

Surface Filtration Depth Filtration

Figure 2.1 Methods of filtration

The particles can be removed by filter media in two methods namely Surface and

Depth filtration, as shown in Figure 2.1. In this review, the focus is on depth filtration.

Surface filtration is common in filtration of gases with high particle concentrations.

Depth filtration is most commonly applied to low concentrations of particles in gases to

reduce particle concentrations to very low levels.

5
1. Surface Filtration

Surface filtration traps contaminants larger than the pore size on the top surface of

the media forming a filter cake [4]. Contaminants smaller than the pore size may pass

through the filter. Mechanisms such as bridging may capture some of the particles

smaller than the pore size on the filter surface. These filters are generally polymeric films

approximately 120 µm thick with a narrow pore size distribution. Examples are

membrane filters and cloths dryer filter.

2. Depth Filtration

The process of depth filtration allows particles to penetrate into the filter media

and get captured throughout the depth of the filter media and on the surface of the media.

Depth filters are composed of random mats of metallic, polymeric or inorganic materials.

These filters have broad pore size distribution. Examples are fibrous filters and furnace

filters.

2.2 Fibrous Filters

As evident from Figure 2.2, contaminant particles range from 0.1 nm to 100 µm

and the most penetrating particle range is between 100 and 500 nm. Fibrous filters are

commonly used medium for particles in size ranging from about 0.2 to 500 microns.

They are composed of randomly oriented fibers, which may or may not be glued together

with a binder. They can be characterized as having low resistance to air flow, low

pressure drop and exceptional particle capture capabilities [2]. Fibrous media are widely

used in disposable filters, due to relatively low cost with acceptable performance.

6
 Scanning electron microscope Optical microscope Eye

Smog

Fumes Dusts

Spores

Tobacco smoke

Bacteria

Human hair
Viruses
Molecules Pollen

0.0001 0.001 0.01 0.1 1 10 100

Particle size (micrometers)

Figure 2.2 Common air contaminants and their relative sizes [5]

2.2.1 Factors effecting particle capture

1. Physical properties of Filter media

Physical parameters such as fiber sizes, filter thickness, porosity (void volume), and

solid volume fraction have marked effects on the performance of fibrous filters [3].

The effecting parameters are

d i - Fiber diameter

L - Filter thickness

α - Projected area of fiber per unit filter area and length of filter

l f - Fiber length per unit filter surface area

7
 S - Packing density (solid volume fraction)

ε - Porosity (void volume fraction)

Some relationship among these parameters are

ε = 1−α 2.1

S = πd i2 l f / L 2.2

α = 4 S /(πd i ) 2.3

2. Particle characteristics

a. Size distribution

b. Phase solid/ liquid

3. Gas conditions

a. Flow rate

The flow per unit area depends largely on the particle size and distribution

of the particulates, pressure drop, total porosity and filter area. Generally

flow can be maximized by adding filter surface area.

b. Temperature

c. Pressure

d. Viscosity

The viscosity of the material passing through the medium can greatly affect

the flow rate, differential pressure and energy cost associated with filtration.

8
2.2.2 Filter efficiency measurement [6]

1. Pressure Drop

Pressure drop is the difference between the upstream pressure and the

downstream pressure. The pressure drop is the most important design feature in the

filtration industry and a significant contributor to a systems energy cost. Pressure drop is

primarily influenced by the properties of filter media, gas velocity and viscosity.

2. Filter Efficiency Rating

Efficiency can be measured in many ways which includes

a. Beta Ratio

Beta ratio is an efficiency measurement of a filter is expressed as ratio of

number of particles of a given size in upstream per unit of volume to the number

of particles in downstream of that same size per unit volume. It is expressed as

percentage

b. Gravimetric rating

Gravimetric rating is the amount of particles measured as weight in a

predetermined volume removed by a filter. It is determined by weighing the

solution passing from the upstream to the downstream side.

c. Microbial rating: H.I.M.A. test

Microbial rating is the measure of a filter’s ability to sterilize liquids, usually

through use of 0.2 micron rated microporous membranes. This liquid challenge

procedure is defined as the logarithmic of the ratio of microorganisms upstream

to those found downstream and expressed as efficiency as a log reduction value

(LRV).

9
3. Dirt Holding Capacity

The amount of dirt a filter can capture and hold with an acceptable pressure drop

is defined as the dirt holding capacity. It is determined by the size of the particulate

matter retained as influenced by the size of the pores.

4. Filter rating

Filter rating used to refer to the particle size capture capability of a filter.

5. Useful Life

The useful life of a filter is determined at of the point when particles

(contaminants) cause an adverse flow rate, low efficiency, and high pressure drop.

Fibrous filter performance depends on many parameters of filter media. Fiber

diameter, fiber cross section and media thickness are the big drivers that effect filter

performance. The smaller the fiber diameter, the greater the capture efficiency and the

smaller the fiber spacing, the greater the filter efficiency. The larger the cross section, the

greater the capture capability [3]. Pressure drop across a filter varies with it thickness, air

velocity, the fiber radius and coefficient of viscosity of the air [2]. Media thickness and

basis weight influence filter porosity. Fiber diameter and fiber size distribution, pore size

influence aerosol flow in the media which play an important role in filter efficiency [4].

The traditional way of making and testing of filters is time consuming and

expensive. This is due to highly complex relationship between physical, chemical and

geometric properties of individual fibers, physical and chemical properties of gas, and

size, mass and shape of dirt particles. Some work has been done to mathematically model

the behavior of flow through filter media. The following section reviews selected work.

10
2.3 Computer models

Several theories on air filtration are described by R.C.Brown [2]. He reviews the

single fiber efficiency theory wherein several particle capture mechanism are combined

in the capture of a particle. Rosner et al [9] describe the theory of particle on cylindrical

surfaces. In the late 1950s, Happel [10,11] and Kuwabara [12] formulated the multifiber

filter models. It described the influence of neighboring fibers and employed artificial

boundary conditions. Brown [2], in the late 1980’s solved the Navier Stokes equation and

further advanced the multifiber filtration theory. Liu and Wang (1996) have recently

improved the multifiber model to greater degree [13].

Fraunhoffer University had developed many simulation software models; one of

them is software simulation of flow through oil filters [14]. It provides evaluation of

pressure drop, flow rate ratio before manufacturing a prototype, optimal design of the ribs

(supporting the filter medium) on the basis of flow computations and evaluation of

uniform loading of the filter medium. The mathematic model, used for filtration

simulation, showed excellent efficiency with respect to the analysis of the investigated

filtration process. It shows detailed information about velocity and pressure distribution

in the filter thus assisting engineers in the design of more efficient filters.

A non woven media model together with the simulation of fluid flow and particle

tracking and particle deposition provides deep insights into the filtration processes in

complex filter media [15]. Non woven geometries, pressure distribution, local flow

velocities and media clogging, filter efficiencies and filter lifetime including pressure

drop evolution and filter efficiency over time are visualized and evaluated in this model.

The Navier Stokes equation is used to describe the motion of fluid and the flow is
11
computed using an Eulerian model (applying equations from continuum mechanics). The

motion and deposition of spherical particles of various sizes is computed by a Lagrangian

formulation for the particle motion (i.e, force balances are applied to each particle to

model particle-particle and particle-fiber interactions).

Srinivasan [16] investigated the effects of incorporating nanofibers in glass

microfiber filter media. In her model volume averaged multiphase transport theory [17]

was used to aid the set up the governing equations used to solve the species balance to

predict the trend in quality factor of filter media upon addition of nanofibers. The model

results show the advantages of adding nanofibers to microfiber filter media [18,19].

Numerical models can significantly reduce the experimental effort of producing

new material prototypes [15]. These models have potential to provide valuable insights

on how to design new filter media for modern filtration applications as needs arise for

separating and filtering particles of different sizes at different operating conditions.

2.4 Optimization Techniques [20]

There are different search optimization techniques like calculus of variation,

linear programming, non linear programming and dynamic programming. Calculus of

variation is meant only for linear equations. Dynamic programming has difficulty in

dimensionality. All these techniques are initial point dependent and it cannot analyze the

wide spread solutions. It is not always practical to find global optima using these

techniques on large scale problems. Enhancement of these techniques to search for

12
global optima greatly increases computational time with no guarantees that the optimal

solutions will be found.

As a compromise to make computational efforts more manageable, other search

techniques have been developed such as Evolutionary Algorithms. A Genetic Algorithm

(GA) is one such evolutionary algorithm and is used in this work.

Evolutionary algorithms are global optimization techniques for solving large scale

problems that have many local optima. However they require high CPU times, and they

are very poor in terms of convergence performance. Evolutionary algorithms tend to not

get trapped on local minima and can often find global optimal solutions. There are other

search techniques like simulated annealing which is a probabilistic search algorithm, tabu

search which avoid discrete searches getting stuck in local minima and alpha beta

pruning which can be used to search good moves in zero sum games. GAs and Neural

Networks have been found to offer advantages over conventional methods, for those

involving non linear and complex mathematical relations [21].

GAs were invented by John Holland and his student in 1975. The GA introduces

the principle of evolution and genetics into search among possible solutions to a given

problem. The idea is to simulate the process in natural systems. This is done by an

iterative procedure that consists of fixed population of possible solutions; each solution is

called a “chromosome.” Chromosomes are stored in the computer in binary form. The

binary chromosome is converted to decimal values for the purpose of calculations.

Calculated values for each chromosome are compared for relative “fitness” as a solution

to a set of optimization criteria. A new population is created using three genetic

13
operators’ reproduction, mutation and crossover. This population in turn is then evaluated

for fitness and the process is repeated until a programmed stopping criterion is reached.

2.4.1 GA characteristics

1. Easy to program and understand.

2. Domain independent search method.

3. Do not get stuck on local optima.

4. Support multiobjective optimization.

5. Always has an answer, the answer improves with greater computation time.

6. Easy to search in a large search space.

Disadvantages

1. They are very slow.

2. They give better solution but not the absolute best solution.

GAs have been applied in many applications and research areas such as numerical

problems, knapsack problems, traveling salesman problem and sequence scheduling,

euler path, finding shape of protein molecules, designing neural networks, and functions

of creating images. GAs has been successfully applied for optimal design of shell and

tube heat exchangers [22]. A case study has been made for the examination of

performance of GAs. It concludes that combinatorial algorithms such as GAs provide

significant improvement in the optimal design compared to other traditional designs. The

GA application for determining the global minimum heat exchanger cost is significantly

faster and has an advantage over other methods in obtaining multiple solutions of same

quality, which provide flexibility to the designer [22].

14
2.5 Incorporation of nanofiber in microfiber

Nanofibers are small sized fibers of diameters less than a micron. They have large

surface area per unit mass. Nanofibers can be formed in low density, large surface area to

mass, highly porous structures that are appropriate for various industrial, consumer, and

military filtration applications [23]. Polymeric nanofibers have been used in commercial

air filtration over the last twenty years. Nanofiber filter media have enabled new levels of

filtration performance in several diverse applications with broad range of environments

and contaminants [24]. These filter media improve filter life and dust holding capacity.

Nanofibers are used in high efficiency filter media, protective clothing material, drug

release membranes, nanotube materials, chemical catalytic materials, bio-transplant

materials, and in hydrogen storage tank for fuel cell, etc. Polymeric nanofibers can be

produced using the electrospinning process [27].

Electrospinning

Electrospinning is a process to make nanofibers with fiber diameters in the range

of about 10 nm to several hundred nanometers from polymer solution through

electrostatic forces. As shown in Figure 2.3, when a voltage is applied to the polymer

solution, the charged solution overcomes the surface tension and form fibers due to

electrostatic repulsion forces.

15
 Collector
Polymer solution

Jet
Fibers

High voltage
Supply

Figure 2.3 Schematic showing electrospinning set up

Hajra et.al showed that addition of nanofibers to microfibers filter medium

significantly improves capture efficiency of sub micron particles [25]. Overall filter

performance can be improved by adding nanofibers to the microfiber filter media for the

most penetrating aerosol particle sizes [26]. Smaller fibers leads to higher pressure drop,

but interception and diffusion efficiencies increases at a faster rate than the pressure drop.

The nanofibers contribute to the pressure drop as well as to the capture efficiency. Thus

in capturing of submicron and micron particles, better filter efficiency can be achieved at

the same pressure drop [27]. Therefore due to dependence of filter efficiency and

pressure drop on fiber sizes, small fiber sizes of 0.2 to 0.3 micron diameter are highly

recommended for filtration application. Filter media can be designed to optimize the

particle removal without creating excessive pressure drop.

16
 CHAPTER III

MODELING AND RESULTS

The filter is modeled where an aerosol is assumed to penetrate the media via

depth filtration. Some assumptions have to be considered to simplify the model. The

theoretical equations are complicated and have many more unknowns than equations. To

make the set of equations more tractable for the optimization calculations to be

completed within a reasonable timeframe, a number of assumptions are applied.

3.1 Assumptions

1. The fibers in the filter medium are rigid and stationary

2. The medium is incompressible.

3. The process is isothermal.

4. The flow is constant and one-dimensional

5. The solid particles are captured on the fibers (not on other particles).

6. The filter is clean (no accumulated particles).

7. In the momentum balance the pressure drop and drag terms are assumed to

dominate, the other mechanisms are neglected.

17
3.2 Particle Capture Mechanism

Fibrous filters remove submicron and micron particles through different

mechanisms which include diffusional deposition, direct interception, inertial impaction

and gravitational deposition.

1. Diffusional depostition

In this mechanism the particles in the range of 0.001 to 0.2 µm range collide with

fibers and are captured due to random brownian motion. This random motion occurs

when particles collide with gas molecules, thereby moving completely independent of the

bulk air stream as gaseous molecules in flowing air. The particles adhere to the fiber

surface via intermolecular forces. Diffusion efficiency is a function of the flow field and

the Peclet number Pe.

d iV
Pe = 3.1
D

Where, d i is fiber diameter, V is fluid velocity and D is Particle diffusion coefficient

Flow streamline

Cross
section
Particle of fiber

Brownian
motion

Figure 3.1 Schematic of Diffusion Deposition

18
2. Direct Interception

The particles in the range of 0.1 to 1 µm size are captured due to this mechanism.

Interception occurs when the fiber diameter is smaller that the particle diameter, the

particle is brought within one particle radius of the fiber as it follows the flow streamline

and captured. Interception efficiency is a function of the flow field and the size ratio NR.

dp
NR = 3.2
di

where d p is the particle diameter.

Flow streamline

Cross
Particle section
of fiber

Figure 3.2 Schematic of Interception Mechanism

3. Inertial impaction

Particles of size 1µm and larger are removed by this method. In this the particles

are captured due to their mass and they develop momentum to deviate from the air flow

stream line when the stream line bends to flow around a fiber. The inertial efficiency is a

function of stokes number St. St is defined as

ρ p Cd p 2V
St = 3.3
18ηd i

19
Where, ρ p is particle density, C is Slip correction factor and η is Gas viscosity

Flow streamline
Cross
section
of fiber

Figure 3.3 Schematic of Inertial Impaction mechanism

4. Gravitational deposition

Particles of greater than 10µm size at relatively low velocities can be captured by

this mechanism.

 ρgd i2 
EG =   3.4
 18µV 

Where, EG is single fiber capture efficiency due to gravitational deposition, g is

acceleration due to gravity, ρ is fluid density, d i is fiber diameter

3.3 Conservation Equations

Multiphase transport theory has been employed to aid the modeling. Volume

averaging theory provides the general transport equations for the aerosol flow through the

filter medium [2].

Gas phase:

• Oil Species Balance:

+
(
∂ (ερ A ) ∂ ερ A v z )
+ ∇ ⋅ ε j A − ε rA + E A + G A + I A = 0 3.5
∂t ∂z

20
 • Momentum Balance (z component):

∂(ερν )
+ ∇ ⋅ (ερνν ) + ε ∇P + ∇ ⋅ (ε τ ) + F − ερ g + ( E + G ) = 0 3.6
∂t

Here, I is the interphase mass transfer, E represent convective transport across the

interphase between the phases, G is generation at the interface due to heterogeneous

reaction, τ is the shear stress, rA rate of mass generation due to homogenous reaction,

j A is the species mass flux vector. The gas phase species balance is used to determine the

outlet particle concentration. The gas phase momentum balance is used to determine the

pressure drop across the media. With the above stated assumptions the balances reduce

to:

Oil Species Balance

d (εwρv z )
+ IA = 0
dz 3.7

Here ε is the void volume fraction (porosity) of the filter, w is the particle mass fraction

in the gas, ρ is the gas phase density, and v z is the gas phase velocity in the z direction.

The species balance is solved by introducing a constitutive equation for the interphase

mass transfer term I A related to the filter coefficient, α , by

I A = αwρεv z 3.8

A relation for calculating the filter coefficient, α , as it is dependent on the fiber size and

amount of each fiber is derived in a later section.

21
Momentum Balance (z component)

dP
ε + FZ = 0 3.9
dz

Here P is the local pressure in the gas phase. To solve the momentum balance a

constitutive relation for the drag force term, Fz must be introduced.

3.3.1 Calculation of outlet concentration

To determine the outlet concentration Eq.3.7 is rearranged to obtain

dw − I A
= 3.10
dz ερv z

Substituting 3.8 into 3.10 and canceling terms gives

dw
= −αw 3.11
dz

On integrating Eq 3.11 from 0 to z gives a general expression for the concentration of

particles in the gas stream as a function of depth within the filter.

w( z ) = wIN exp(−αz ) 3.12

where wIN is the inlet mass fraction. The outlet mass fraction in the gas stream is

obtained from Eq.3.11 as

wOUT = wIN exp(−αL ) 3.13

22
Filter Coefficient (α)

The filter coefficient α is related to the single fiber capture efficiency of fibers of

diameters d i as given by

n
α = ∑
i =1
E iα i 3.14

where E i is the overall single fiber efficiency taking into account mechanisms of direct

interception, inertial impaction, Brownian diffusion, and gravity, given by Brown [2].

The overall efficiency E i is defined as

(1 − E i ) = (1 − E Ri )( 1 − E Ii )( 1 − E Di )( 1 − E Gi ) 3.15

where (1 − E Ri ) is the fraction of particles that escape from capture by direct

interception. Similarly the fraction of particles escaping from the other mechanisms are

given inertial impaction (1 − E Ii ) , diffusion (1 − E Di ) , and gravity (1 − E Gi ) .

Correlations for these mechanisms are available in [2] and [7].

1  −1  εi  εi 3
E Ri = 2(1 + N R ) ln(1 + N R ) − (1 + N R )(1 − ε i ) + (1 + N R ) 1 −  − (1 + N R ) 
2 Ku   2 2 

3.16

where ε i is the volume fraction of fibers of diameter d i in the medium. The correlations

account for slip-flow effects when the Knudsen numbers are greater than 0.01.

E Ri =
{(1 + N )−1 − (1 + N R ) + 2(1 + 1.996 Kn )(1 + N R ) ln(1 + N R )}
R
3.17
(2(− 0.75 − 0.5 ln(ε i )) + 1.996 Kn(− .0.5 − ln(ε i )))

23
 The Knudsen number is defined as

2λ
Kn = 3.18
di

where λ is the mean free path of the molecules in the gas stream.

 1 ε i 
2

Ku = − ln(ε i ) − 0.75 + ε i − 3.19
 2 4 


The single fiber efficiency for capture by diffusional deposition is given as [2]

(
Ε D i = 2.27 Pe − 2 / 3 1 + 0.39 KnΡe −1 / 3ζ ' −1 / 3
) 3.20

In reality, the multiple mechanisms act together for capture of sub-micronic

particles along with the aerodynamic slip affect on direct interception and brownian

diffusion. R.C. Brown has given the slip affect as [2],

−0.5
E RD = 1.24ζ ' Pe −0.5 N R2 3
'
3.21

Overall single fiber efficiency, E RD is given by,

E RD = E R + E D + E ' RD 3.22

The single fiber efficiency is defined as the ratio of the number of particles

actually removed relative the number that would be removed if particle capture was

proportional to the projected area of the fibers. The projected area is used to define α i as

4ε i
αi = 3.23
πd i

24
3.3.2 Calculation of Drag Force and Pressure drop

The total force acting on all fibers, FTOT, is given by

FTOT=A∆P 3.24

We define the force per unit length of fiber as f. For a filter of multiple fiber diameters

the pressure drop is related to the force per unit length by

n
A∆P = ∑ Li f i 3.25
i =1

where Li is the total length of fibers of diameter d i within the filter.

Integration of Eq.3.9, assuming uniform properties over the filter thickness, the

pressure drop is related to the drag force by

FZ L
∆P = 3.26
ε

where Fz is the total drag force and L is the thickness of media.

Combining Equations 3.25 and 3.26 we get the expression for the drag force to be

ε ∑ Li f i
FZ = 3.27
AL

The volume fraction of the filter occupied by the fibers of diameter di is defined as

Vi
εi = 3.28
AL

25
where Vi is the fiber volume, A is the filter inlet area and L is the filter depth. The

volume occupied by the fibers of diameter d i is given by

πd i 2 Li
Vi = 3.29
4

Hence, combining Eqs.3.28 and 3.29 the length of the ith fiber is given by

4 ALε i
Li = 3.30
πd i 2

Substituting Equation 3.30 in 3.27, we get the relation for the local drag force to be

4ε εi
FZ = ∑( fi ) 3.31
π di
2

If the medium has only one fiber size, then through use of Eqs.3.31 and 3.26 existing

correlations relating pressure drop to flow rate can be used to determine expressions for

the functions f i . For example, for fibers of diameters in the slip flow

range 0.01 < K n < 0.25 , given by Brown [2] the pressure drop relation is

16 µευ Z (1 + 1.996 Kn )πd i ε i

2

∆P = 3.32
 εi
2
εi 
2

d i − 0.5 ln(ε i ) − 0.75 + ε i − + 1.996 Kn( −0.5 ln(ε i ) − .025 +

2

 4 4 

and the expression for fi is

4 µπευ Z (1 + 1.996 Kn )
fi = 3.33
 εi
2
εi 
2

 − 0 . 5 ln( ε i ) − 0 . 75 + ε i − + 1 . 996 Kn ( − 0 . 5 ln( ε i ) − . 025 + 

 4 4 

Other correlations are used for other ranges of Knudsen number.

26
3.3.3 Quality Factor ( QF )

The Quality Factor, QF , is a measure of the particle capture efficiency relative to

the energy (ie., pressure drop) required for the separation. The QF is defined by the

expression [2]

w 
− ln out 
QF =  win  3.34
∆p

Combining Eqs. 3.13, 3.26, 3.31, and 3.34 gives the working equation for

calculating the QF ,

α
QF = 3.35
4 εi
∑( fi )
π di
2

For comparison between filter media we define the relative quality factor, RQF , as

QF
RQF = 3.36
QFCTRL

where the QFCTRL is the QF for a control filter medium against which other media area

compared.

3.4 Model Calculations

The model is easily programmed into a spreadsheet or in programming software

such as C or FORTRAN. As an example, the filter is modeled at room temperature,

atmospheric pressure, with air face velocity of 2.1 m/s. The filter is disk shaped 1 cm

27
thick and 6 cm in diameter. The filter is challenged with a concentration of 500 ppm

mass fraction of 150 nm spherical particles. The filter is made of 3 grams of 2 micron

diameter microfibers having a density of 2500 kg/m3. Varying amounts of nanofibers of

different diameters are added to the filter. The filter porosity is determined from the

volumes of the fibers present in the filter.

The amount of nanofibers in the filter is calculated as the ratio of the external

surface area of the nanofibers divided by the external surface area of the microfibers in

the filter medium. An area ratio of zero means no nanofibers are present in the filter.

The QF for filter media with zero nanofibers are used as the control value, QFCTRL
.

Figure 3.4 shows the RQF curves for three different nanofiber sizes. The RQF

increases rapidly with small additions of nanofiber, reaches a maximum at around an area

ratio of about 1 or 2, and then declines at a slower rate. The rapid increase in RQF is

due to the significant increase in surface area available for capture of particles. The

RQF passes through a maximum and declines because as more nanofiber is added to the

filter the pressure drop increases more rapidly than the capture efficiency.

28
 150 nm Particles and 3 um Microfibers
Variation in Nanofiber
12
100 nm Nanofibers
10

Relative Quality Factor

8

6
300 nm
4
500 nm
2

0
0 2 4 6 8 10 12
Area Ratio

Figure 3.4 Relative Quality Factor for different nanofiber sizes and area ratios for filters of 3

micron microfibers and challenged with 150 nm particles [16]

Sensitivity analysis of the filter

Sensitivity analysis for the filter is conducted to study the variation in quality

factor. The base case is taken at temperature of 210C, 20 Psi pressure, face velocity of 2.1

m/sec, particle diameter of 150nm, 2 micron microfiber diameter, 200nm nanofiber

diameter, 1 cm thickness of media, 3gms of microfiber and area ratio of nanofiber to

microfiber as 0.1. The quality factor calculated for the base case is 0.135 (1/Kpa). Effect

of diameter of nanofiber, diameter of microfiber, particle diameter, thickness of media,

area ratio, mass of microfiber, temperature and face velocity on quality factor is shown in

Figure 3.5. Figure 3.5 shows that diameter of nanofiber, particle diameter and face

velocity have dominant effects on the quality factor. The variations in quality factor due

to fiber amounts, fiber sizes and other design parameters suggest that a computer program

29
may be written to determine optimum design values where the filter has maximum

performance.

10

8
Nanofiber diameter
7
area ratio
6 microfiber diameter
QF/QFbase

thickness of media
5 mass of microfiber
particle diameter
4
face velocity
temperature
3

0
0 1 2 3 4 5 6
Parameter/Parameterbase

Figure 3.5 Effect of parameters on quality factor

30
 CHAPTER IV

COMPUTER PROGRAM

4.1 Approach

To optimize the filter design we apply a Genetic Algorithm (GA). GAs are robust

in selecting optimum size and amounts of nanofiber and microfiber and other design

parameters for composite filter media. GAs are used in search and optimization, such as

finding the maximum of a function over some domain space [28]. A flow chart

describing the major steps in applying the GA to filter design is shown in Figure 4.1.

Initialize
Population

Iterations
Evaluate
Population
Fitness (QF)

Stopping criteria
satisfied?
No
Yes

Genetic Stop
“Offspring”
Operations Iterations

Figure 4.1 Flowchart for applying a Genetic Algorithm to optimize filter design
31
 A random number generator is used to initialize a population of solutions. Each

solution consists of the values of several variables that are being optimized. In our

program, the variables being optimized are the thickness of the filter medium, the

diameter of the micro fibers, the diameter of the nanofibers, the area ratio of the

nanofibers to microfibers, and the mass of the microfiber (in a 6 cm diameter disk). A

range is given for each variable over which the search is conducted. Each solution in the

population is evaluated by calculating the QF (fitness function) for that solution.

4.2 Fitness Function

The fitness function links the Genetic Algorithm to the problem to be solved. The

optimization problem can be used to maximize or minimize a function. Our program is

based on maximizing the function “Quality Factor” for efficient filter performance. The

best fit solution in the final generation is the one that maximizes the quality factor. The

program will calculate quality factor for each set of solutions until a stopping criterion is

reached. The fitness function evaluates each solution to decide whether it will contribute

to the next generation of solutions. Then, through operations analogous to gene transfer

in sexual reproduction, the algorithm creates a new population of candidate solutions.

4.3 Convergence

Convergence is the final problem to stop the execution of the algorithm. A

stopping criterion is checked to see if it is satisfied. In our program the user specifies the

maximum number of populations to be evaluated. When this maximum is reached the

program stops. A second stopping criterion is also applied in which the change of the

32
maximum QF to the number of generations between changes is monitored and if this

change rate becomes smaller than a user specified value then the program stops. When

the stopping criteria is not satisfied the program proceeds to generate a new population.

Convergence of less than 10-8 is recommended for a better solution. Each new population

is generated by three genetic operators.

The new population is generated by copying existing solutions into the new

population. The number of copies of each solution are weighted to make more copies of

the best solutions and fewer copies of the worst solutions. This is analogous to the

“survival of the fittest” from evolutionary theory. The solutions are stored in the

computer in binary form that makes it possible to apply cross-over in which bits between

two parents from the previous generation are exchanged to create a new offspring

solution. Also, a small number of the new solutions are randomly selected for mutation,

in which one of the bits in the solution is randomly selected and changed from 0 to 1 or 1

to 0. The cross-over and mutation creates solutions in the new population different from

the previous population that enables the search to extend to other parts of the solution

space.

33
 Figure 4.2 Screenshot of program FiL2RO

4.4 Input and Outputs

Figure 4.2 shows screenshot of the program FiL2RO. The input parameters are

operating conditions such as pressure, temperature, air face velocity (inlet velocity of air

entering into the filter), diameter of filter media, particle size which needs to be captured

and density of microfiber. The ranges for the five physical parameters of filter media

namely thickness of media, diameter of microfiber, diameter of nanofiber, area ratio of

nanofiber to microfiber, mass of microfiber that needs to be optimized are entered.

Finally the genetic algorithm parameters such as population is fixed at 1000, but the

number of generations, mutation rate and the convergence value can be set by the user.

Depending on the convergence tolerance, the program stops and shows the result. If

convergence tolerance is not achieved, the number of generations can be increased and

the program can be re-run. A convergence tolerance of less than 10-8 is recommended to

get better solutions.

34
 The output shows the optimum design parameters of filter media with generation

at which the program got converged and quality factor. It also calculates the porosity,

pressure drop, and permeability of the best medium at program convergence. The

generation vs. convergence and quality factor vs. selected design parameters can be

plotted after the program has converged.

4.5 Features of program

1. Interactive presentation of data to enable user input and interpretation of the

calculated data.

2. Warns the user if input data are incorrectly entered or if data are missing.

3. Conversion of units according to your choice.

4. Output data in graphical and tabular format.

5. If the convergence is not achieved it warns the user to increase the iterations or

increase the convergence.

6. User can view input parameters and graphical interpretation of final result.

7. Results can be saved into a file to be viewed later.

8. All data required for filter design are shown.

Limitations

1. Particles are spherical in shape.

2. Surface properties are not considered.

3. Uniform distribution of fibers in media.

4. Gas phase is incompressible (small pressure drop)

5. Steady state momentum (after loading )

35
 6. GA is iterative in nature hence time consuming.

7. Diffusion and interception mechanism are dominant only in sub micron sized

particles.

4.6 Results

The computer program is written in Fortran and Visual basic to allow the user to

input operating conditions such as the size of the particle to be captured, gas temperature,

flow rate, etc. It is not practical to try to test each of these conditions to report on the

optimum design parameters for the filters. However we have made several general

observations that we report here.

Table 4.1 Results from software program for several particle diameters

Diameter of particle 150nm 500nm 1000nm

Parameters Search Optimum Optimum Optimum

Range
Thickness of media (m) 0.01-0.02 0.014 0.0172 0.013

Diameter of microfiber (microns) 1-10 1.03 1.07 1.02

Diameter of nanofiber (nm) 10 -1000 10 10 10

Area ratio 0.1-10 9.76 9.9 9.85

Mass of microfiber (kg) 0.002-0.01 0.00252 0.00275 0.0022

Convergence 3.40E-09 3.64E-09 1.94E-09

Quality factor (per Kpa) 5.26E-03 1.78E-02 3.65E-02

Porosity 0.972 0.975 0.973

Pressure drop (Kpa) 1.45E+04 1.55E+04 1.29E+04

2
Permeability (m ) 3.94E-11 4.50E-11 3.96E-11

Generation 35256 137865 134957

36
 Table 4.2 Results from software program with different parameter ranges

Diameter of particle 150nm 500nm 1000nm

Parameters Search Optimum Optimum Optimum

Range
Thickness of media (m) 0.01-0.05 0.048 0.0491 0.0488

Diameter of microfiber (microns) 1-10 9.98 9.73 9.99

Diameter of nanofiber (nm) 200-1000 200 200 200.04

Area ratio 0.1-10 1.81 1.91 1.9

Mass of microfiber (kg) 0.001-0.008 0.001 0.001 0.001

Convergence 4.97E-10 2.82E-09 4.41E-11

Quality factor (per Kpa) 4.40E-04 1.69E-03 3.97E-03

Porosity 0.997 0.997 0.997

Pressure drop (Kpa) 2.33E+01 2.54E+01 2.44E+01

2
Permeability (m ) 8.50E-08 7.82E-08 8.10E-08

Generation 34554 18296 190420

Table 4.3 Results from software program showing the optimum value for area ratio

Parameters Search Range Optimum

Thickness of media (m) 0.01 0.01

Diameter of microfiber (microns) 2 2

Diameter of nanofiber (nm) 200-1000 200

Area ratio 0.1-10 2.02

Mass of microfiber (kg) 0.003 0.001

Convergence 7.56E-9

Quality factor (per Pa) 3.21E-04

porosity 0.949

Pressure drop (Kpa) 9.55E+02

Permeability (m2) 4.24E-10

37
 Tables 4.1, 4.2 and 4.3 show sample results from the software program. The

calculations assume conditions of room temperature, atmospheric pressure, with air face

velocity of 2.13 m/sec, the diameter of filter disk is 6cm, and the density of the micro

fibers is 2500 kg/m3. The program was run for three different particle sizes of 150nm,

500nm, 1000nm. The two tables compare results for different search ranges for the

media thickness, nanofiber diameter, and mass of microfiber in the filters. The tables

show the characteristics of media and number of iterations the program takes for the

convergence criterion to reach less than 10-8. The program took about 30 minutes to give

optimal solutions.

We observe from the data in the tables that when we take minimum value for

nanofiber diameter as 10nm or less that the optimum values for thickness of media, and

mass of microfiber are between the extremes specified in the search range limits, the

microfiber diameter goes to the minimum value, and the external surface area ratio of

nanofiber to microfiber goes to the maximum value. The pressure drop for this filter

design is high. In the Table 4.1 when we take minimum value for nanofiber diameter as

200 the optimum values for thickness of media and the diameter of microfiber goes to

maximum in the search range limit, the optimum values for nanofiber diameter and the

mass of microfiber go to the minimum in the range limits, and the surface area ratio of

nanofiber to microfiber falls between the range limits. For this filter design the pressure

drop is lower compared to data obtained in Table 4.2 and the quality factor is lower when

compared to data in Table 4.1. Table 4.3 shows that optimum value for area ratio is 2.02

for 150nm particle diameter for a range of 0.1 to 10 and diameter of nanofiber as 200nm

for a range of 200-1000nm, fixing the other variables.

38
 Thickness of m edia vs Quality Factor and Pressure drop

0.00465 Quality factor

Quality factor 630

620
0.0046
610

Quality Factor (per Kpa)

Pressure Drop (Kpa)

0.00455 600

590
0.0045
580

0.00445 570
Pressure drop
Pressure drop 560
0.0044
Optimum
is 2.79 cm Optimum 550
is 2.79 cm
0.00435 540
0 1 2 3 4 5 6 7
Thickness of m edia (cm )

Figure 4.3 Thickness of media vs. Quality Factor and Pressure drop

Figure 4.3 shows the variation of quality factor and pressure drop with thickness

of media. The optimum solution for thickness of media is calculated from the program.

The data plotted in this figure are for microfiber diameters of 2 microns, mass of

microfiber is 0.002 kg and the area ratio of nanofiber to microfiber is 1. The range for the

thickness of media is 0.01-0.05 m and the diameter range of the nanofiber is 10-1000 nm.

The optimal solution for the thickness of media is 2.79 cm and the optimum nanofiber

diameter is 10 nm.

39
 Mass of microfiber vs Quality Factor and Pressure drop

0.0051 12000
Quality factor
Quality factor
0.00508 10000

Q uality Factor (per KPa)

0.00506

Pressure drop (Kpa)

Pressure drop 8000
Pressure drop
0.00504
6000
0.00502

4000
0.005
Optimum in Optimum in
2000
0.00498 Quality FactorQuality
is Factor is
about 3 g about 3 g
0.00496 0
0 2 4 6 8 10
Mass of microfiber (gm)

Figure 4.4 Mass of Microfiber vs. Quality Factor and Pressure drop

Figure 4.4 shows the variation of quality factor and pressure drop with mass of

microfiber. The optimum mass of microfiber in a 6 cm diameter disk filter 0.02 m thick is

plotted in this figure. In the program calculations the diameter of microfiber is 2 microns

and the area ratio of nanofiber to microfiber is fixed at 5. The search range of mass of

microfiber is 0.001-0.008 kg and search range for the diameter of nanofibers is 10-1000

nm. The optimal mass of microfber obtained is 0.003 kg and the optimum nanofiber

diameter is 10 nm.

We can conclude from the data in Tables 4.1, 4.2 and 4.3 that smaller fibers

contribute to higher pressure drop but also contribute to greater capture efficiency. The

quality factor increases as long as the capture efficiency increases faster than the pressure

drop. The plots in Figures 4.3 and 4.4 show that there are optimum values for thickness
40
of media and mass of microfibers. Sometimes the optimums occur within the specified

search range and sometimes the optimum occurs at either the upper or lower limits of the

search ranges. When the optimum occurs at the limits this suggests that the search ranges

should be expanded, but that that will be a topic for further study.

Overall, there are observable trends in the optimum design parameters. However,

because there are so many design parameters and it is not easy ways to represent all of

the optimum design conditions such as on a single plot or chart. A computer search code

such as described here is a valuable tool for helping design engineers to determine

optimum designs.

41
 CHAPTER V

FILTER MEDIA PREPARATION AND CHARACTERIZATION

5.1 Filter Media Preparation

Slurry preparation

A measured quantity of glass fibers having diameters of 2 microns are dispersed

in water in a slurry tank. About 3ml of carboset 560 binder is added to bind the fibers

together. Carboset 560 (supplied by B.F.goodrich) is an acrylic polymer dispersed in

water; milky white in appearance. Acid is added to maintain pH in the range of 2.5-3.0

for uniform mixing of fiber. The solution is stirred for 24 hrs. Then the slurry is vacuum

molded to form filters of 6cm diameter. There are several methods for mixing nanofiber

and microfiber to make filter media of mixed fiber. The polymer solution can be directly

electrospun into the prepared microfiber slurry. After a certain amount of nanofiber is

collected, the slurry is continuously mixed with air bubbles. Another method of adding

nanofiber is to chop up the required amount of nanofiber and then make slurry of mixed

fiber of microfiber and chopped nanofiber and vacuum molded to the make the filter

medium. Microfiber and nanofiber composition are B-glass fiber of diameter 2 microns

and nylon nanofiber of diameter 200nm which is electrospun at the rate of 2µl/min and

voltage of 20KV. Figure 5.1 shows the SEM image of electrospun nylon nanofiber.

Filter media containing different amounts of nylon nanofiber and 3 gms of

microfiber are prepared using vacuum molding and tested using the coalescence filtration
42
apparatus. The amount of nanofibers added to the filter medium is based on the ratio of

surface area of the nanofibers to microfibers, Anf/Amf. Filters are prepared with the

Anf/Amf = 0.0, 0.05, 0.4, 0.8 1.2, 3.6. The amount of nanofibers is varied depending on the

desired value of Anf/Amf.

Figure 5.1 SEM image of electrospun nylon nanofiber on a 2 micron glass fiber

Vacuum pump

Mixing tank

Plexiglas disks

Steel mesh screen

Collecting tank

Figure 5.2 Vacuum mold assembly

43
The vacuum mold consists of collecting tank and a vacuum pump which is capable of

generating pressure of about 100mm to 500mm mercury. The prepared slurry is poured

into the mixing tank. The mixing tank is connected to the mold which is connected to a

vacuum hose. The filter cake is formed over the steel mesh screen in the mold. With

vacuum pressure all the acidic solution is collected in the tank and discarded and a

cylindrical shaped filter of 6cm diameter is formed. The wet filter media is removed from

the mold and heated in an oven at about 120-1500C to ensure proper curing of the binder.

5.2 Filter Media Characterization

Fiber parameters, such as fiber type, fiber fineness, cross-sectional shape, and

filter parameters, such as mass per unit area, fiber packing fraction, thickness, and surface

characteristics, play a major part in determining the filter media structure and related

properties, in particular pore size distribution, and air permeability. Porosity, permeability

and hardness will provide the information how the media will perform.

Porosity

Porosity of Filter media is measured using pycnometer shown in Figure 5.3.

Porosity is void fraction present in the filter medium. Porosity of fibrous filters consisting

of a mat of fine fibers arranged in such a way that most are perpendicular to the direction

of air flow have the porosity from 70% to more than 99%. Flow rate is directly

proportional to the porosity of the media; higher porosity means higher flow rate, for a

given pore size and thickness of filter medium.

44
 Figure 5.3 Pycnometer set up in the laboratory

Frazier permeability test

Permeability is the ability of medium to allow the passage of liquid though the

pores of medium. The permeability of a medium is defined by Darcy’s equation

describing flow through a porous layer. Higher the permeability lesser will be the

resistance to airflow. Permeability is measured using the Frazier test in Figure 5.4.

Q L
K= µ 5.1
A ∆P

where,

K=Permeability coefficient

∆P= Pressure drop

L=Thickness of media

Q=Volumetric flow rate

45
µ= viscosity

A=Area

Figure 5.4 Frazier’s Permeability set up in the laboratory

Hardness

A durometer is used to measure the hardness of the filter medium defined as the

material's resistance to permanent deformation.

Table 5.1 shows the characteristics of filter media for different amount of

nanofibers. Nanofibers increase the volume fraction of the fibers and reduce the void

spaces thus decreasing the porosity. Nanofibers have tighter pore structure which reduces

the permeability of the media, but have lower hardness.

46
 Table 5.1 Average porosity, permeability and hardness of the filters

Area ratio Porosity Permeability (k *10^11, m2 ) Hardness

0 0.9 6.3 25

0.38 0.85 4.2 23

0.78 0.84 3.29 22

1.17 0.81 2.8 18

2.02 0.8 2.5 16

3.6 0.78 1.03 14

47
 CHAPTER VI

EXPERIMENTAL RESULTS AND DISCUSSIONS

6.1 Coalescence Filtration

A coalescence filtration apparatus shown in Figure 6.1 is used to determine the

pressure drop and efficiency of a media. The main components of the coalescence

filtration apparatus are: pressurized air supply, Laskin nozzle for generating oil droplets,

filter holder, and Scanning Mobility Particle Sizer (not shown in figure).

Figure 6.1 Coalescence Filtration set up in the laboratory

48
 The cylindrical test media of 6cm diameter and thickness of 1 cm is inserted in a

filter holder and the steel mesh is placed to give support to the filter at higher pressure

drops. The upstream section comprises various elements to generate the aerosol. The

Laskin nozzle is used to generate oil droplets in order to challenge the filter medium and

the filtered compressed air at a flow rate of 0.0035 m3/sec. The oil used in the Laskin

nozzle is propylene glycol. Moisture in the compressed air stream is removed using a

dryer. The exit stream from the dryer splits into two streams, one stream passes into the

Laskin nozzle for generating oil droplets under pressure. The other stream passes

through a pressure regulator on its way to the filter sample. A pressure regulator is used

to attain the desired pressure drop across the Laskin nozzle. The exit air stream from

Laskin nozzle carries oil droplets in the form of an aerosol. A differential pressure

transducer (Omega PX 7771) is used to measure the pressure at the upstream and

downstream. A photometer (Air Techniques TDA-2G) is used to verify the steady state.

The Photometer is used to measure the concentration, once it is calibrated. A Scanning

Mobility Particle Sizer, SMPS (TSI instruments, Model 3080) is used to measure the

particle size distribution

The aerosol was sampled upstream and downstream of the media. The particle

size distribution was measured by SMPS. The data from SMPS are analyzed for

measuring the inlet and outlet concentration of the aerosol.

A High Efficiency Particulate Air, HEPA filter at the downstream captures the oil

and prevents the oil from entering into the rotameter (Omega, series FL 86). The flow

rate to the rotameter is kept constant for all filter experiments.

49
6.2 Results and Conclusion

Performance of filters from coalescence experiments:

Filters with various area ratio of nanofiber to microfiber were analyzed. The

experiment runs until a steady state is reached in pressure drop and outlet concentration.

Pressure drop, saturation and quality factor at steady state are compared. Quality factor is

calculated taking the inlet and outlet concentration data which are obtained from SMPS

and pressure drop using pressure transducer at steady state. Figures 6.2 and 6.3 show

quality factors from experiments and from the model as a function of the area ratio

analogous to Figure 3.4. The quality factor increases with addition of nanofibers and

reaches an optimum value and then decreases.

Figure 6.4 shows pressure drop trend with addition of nanofiber.. The pressure

drop increases on adding nanofibers which makes quality factor to decrease. Figure 6.5

shows the SMPS data for particle size distribution. The plot shows the diameter of

particle versus droplet count per cubic centimeter of air measured for outlet and inlet

concentration for area ratio of 3.6 (0.45 gm of nanofiber).

50
 Quality factor vs area ratio

0.29
0.28
0.27
0.26
0.25

QF (1/Kpa ) 0.24
0.23
0.22
0.21
0.2
0 1 2 3 4 5 6
Anf/Amf

Figure 6.2 Experimental quality factor versus Anf/Amf

Quality factor vs. area ratio

0.5

0.4
Q F (1/Kpa)

0.3

0.2

0.1
0 1 2 3 4 5 6
Anf/Am f

Figure 6.3 Quality factor versus Anf/Amf from modeling

51
 Pressure drop vs Area ratio

20
18
16

Pressure drop (Kpa)

14
12
10
8
6
4
2
0
0 1 2 3 4 5 6
Anf/Am f

Figure 6.4 Pressure Drop versus Anf/Amf

Concentration vs. Particle diameter

12000

10000
Concentration (num ber/cm 3)

8000

Cout
6000
Cin

4000

2000

0
19.5
24.1
30
37.2
46.1
57.3
71
88.2
109
136
168
209
259
322
400
496
615
764

Particle diameter (nm)

Figure 6.5 Concentration profile versus particle size

52
Comparison of model with experiment

Table 6.1 shows a set of experimental and computer data. An experiment was

conducted on filter media with specific properties including an area ratio of 0.78. The

same filter characteristics were entered into the computer program to calculate the quality

factor. Table 6.2 shows the QF from experiments as 0.24 and QF from computer program

as 0.297

Table 6.2 shows the optimum quality factor from experiments and the computer

program. In the computer program the search range is given for diameter of nanofiber

and area ratio and the optimum area ratio comes out to be 2.02 which is shown in Table

4.3. Figure 6.2 shows the optimum QF value as 0.27 KPa-1, whereas the optimum QF

calculated by FiL2RO is 0.321 KPa-1. These experiments were conducted on filter media

having properties as close to the predicted properties as could be constructed in our

laboratory using the vacuum molding process. This shows the predicted optimum filter

design is reasonably close to the experimental optimum.

The difference in the values between the experiments and the computer

calculations may be due to certain assumptions taken in modeling which cannot be

controlled during experiments. The fiber size and particle size are not uniform during

experiments. Hence the averages of three sets for experiments are reported in Table 6.1

and Table 6.2 for comparison with the computed results. However, some experimental

conditions are not accounted for in the computer model. The real filter media have binder,

which can account for part of the lower porosity (fiber structure and packing also affects

the porosity). The model does not account for liquid saturation, which further reduces the

available space for gas flow. The computer model results do not exactly match the

53
experimental results but the results are of the same order of magnitude and show similar

trends within about 25% error.

Table 6.1 Comparison of quality factor from experiment and the computer program

Parameters Experimental data Computer Program

Thickness of media (m) 0.01±0.009 0.01

Diameter of microfiber (microns) 2±3 2

Diameter of nanofiber (nm) 200±150 200

Area ratio 0.78 0.78

Mass of microfiber (kg) 0.003 0.003

Porosity 0.84±0.01 0.95

Saturation 0.30±0.02 0

Quality factor (per KPa) 0.24±0.008 0.297

54
 Table 6.2 Comparison of quality factor from experiment and the optimum

from the FIL2RO computer program

Parameters Experimental data Computer Program

Thickness of media (m) 0.0098±0.001 0.01

Diameter of microfiber (microns) 2±3 2

Diameter of nanofiber (nm) 200±150 200

Area ratio 2.01 2.02

Mass of microfiber (kg) 0.003 0.003

Porosity 0.80±0.02 0.95

Saturation 0.35±0.01 0

Quality factor (per KPa) 0.27±0.01 0.321

In the coalescence filtration, saturation across the filter is estimated and that

saturation value is incorporated in the model using the volume averaging equations.

Saturation across the filter is estimated using the following equation [29]

(Mass final − Mass initial )

Sa = 6.1
Vcake (1 − ε i )ρ oil

Mass initial - Mass of the filter before experiment.

Mass final -Mass of the filter after experiment.

Vcake - Volume of the filter.

ρ oil - Density of oil, i.e., polypropylene glycol = 983 kg/m3

In our modeling we have assumed as clean filter which means that saturation is 0.

The range of saturation varied from 0.20 to 0.47 during the coalescence filter experiments.
55
The filter with more amount of nanofiber has higher saturation. At surface area ratios

greater than 2 the pressure drop increases faster than the capture efficiency and decreases

the QF.

After running a few experiments; the results are encouraging and suggest that the

error is reasonable. However a more extensive validation is needed in future work.

56
 CHAPTER VII

CONCLUSIONS AND FUTURE WORK

7.1 Conclusions

This work has 8 objectives. The results for each objective are summarized here.

1. Develop a computational model to predict filter performance for mixed fiber

media for depth filtration.

I developed a model using volume averaged continuum equations and

multiphase transport theory taking into accounts all the assumptions (Section 3.3).

The model calculates single fiber efficiency, pressure drop and the quality factor

of the filter media for mixed fiber media. The model shows that quality factor

depends on fiber size, fiber quantities and other design parameters of media

(Figure 3.5).

2. Selection of design parameters of filter media for optimization.

The basic requirement of filter media is high efficiency and low pressure drop.

The computer model finds the best value of five physical parameters thickness of

media, diameter of microfiber, diameter of nanofiber, area ratio of naofiber to

microfiber and mass of microfiber that optimizes capture efficiency and pressure

drop.

57
3. Develop a software program to find optimum parameter values of mixed fiber

filter media design using the Genetic Algorithm search technique.

I developed a computer program FiL2RO that determines optimum design

parameter values of the five physical parameters listed in 2. Genetic Algorithm

coding and all the modeling equations are written in FORTRAN (Appendix A).

The Visual Basic is used for user interface so that the user can input parameters

and visualize and interpret the results (Figure 4.2). GA is explained in detail in

chapter 4.

4. Selection of fitness function and convergence criterion.

I have chosen Quality Factor as fitness function (Section 4.2). Performance of

filter is quantified by Quality factor. Higher the quality factor, the better the filter

performs. A convergence criterion is based on maximum number of

population and slope of the threshold value of maximum quality factor to the

number of generation (Section 4.3).

5. Interpret and analyze the results from the software program.

The program results are shown in chapter 4. The FiL2RO program calculates the

optimal solution based on a given range for the five parameters.

6. Construct and test filters with varying amounts of nanofibers in glass fiber media.

I prepared filters using different amount of nanofiber and the filter media were

tested using coalescence experiments to find quality factor and pressure drop. The

optimum performances of filters were obtained for surface area ratio of about 2

(Figure 6.2).

58
 7. Compare model calculations with experimental data (Figures 6.2 and 6.3). Figures

6.2 and 6.3 show similar trend with optimum quality factor at area ratio of 2.

8. Compare optimum filter performance with optimization program (Table 6.1).

Coalescence filtration experiments were done to verify the computer modeling

results. The results show about 25% error in quality factor between computer and

experiment (Table 6.2)

A computer code has been developed that optimizes filter medium design. This

program gives physical properties (fiber diameter, amount of fibers and thickness of

media) that are expected to give clean media with optimum performance. This design can

be used as the starting basis for filter companies to design media with nanofiber for

specific applications.

7.2 Future work

The following are some suggestions for the future work:

1. Add particle size and fiber size distributions to the model to get better

agreement between the model and experiments.

2. More thorough validation of model to experiment.

3. Model can be extended for multilayered media.

4. Particle loading or saturation can be incorporated in the model.

5. Refinement of the model limiting assumptions like considering effects of

binder which accounts for porosity.

6. Model Unsteady State Coalescence stage.

7. Influence of other filter characteristics like loading on filter life.

59
 NOMENCLATURE

A Area of filter

di , fiber diameter

Ei Single fiber efficiency

Fz Drag force

FTOT Total force acting on a filter

fi Drag force per unit length of fiber of diameter di

I Mass flux of particles captured from gas phase onto fibers

Kn Knudsen number

L Filter thickness

Li Total length of fibers of d i within elemental volume of the filter medium

∆P Pressure drop across filter

QF Quality Factor

RQF Relative quality factor (QF of the medium/QF of medium of only microfibers)

60
vz Intrinsic gas phase velocity in the z direction

w Mass fraction of particles in the gas phase

win , wout Inlet and outlet mass fractions of particles in gas phase

α Overall Filter Coefficient for the filter medium

αi Filter Coefficient based on projected area of fibers

ε Porosity of the media

εi Volume fraction of the fibers with size i

λ Mean free path

µ Viscosity of gas phase

ρ Gas phase density

D Particle diffusion coefficient

E Convective transport across the interpahse between the phases

G Generation at the interphase due to heterogenous reaction

I Interphase mass transfer term

Sa Saturation across the filter

61
 BIBLIOGRAPHY

1. Bemer D., S. Calle, “Evolution of the Efficiency and Pressure Drop of a Filter
Media with Loading”, Journal of Aerosol Science and Technology, 2000.

2. Brown R.C., “Air filtration”, Pergamon, New York, 1993.

3. Spurny K.R., “Advances in Aerosol Filtration”, CRC Press LLC, 1998.

4. Purchas D.B., “Handbook of Filter Media”, 1st edition, Elsevier Advanced

Technology, Oxford, 1996.

5. Hinds W.C., “Aerosol technology: properties, behavior, and measurement of

airborne particles”, John Wiley, New York, 1982.

6. Johnson P.R., “A survey of test methods in fluid filtration “, Gulf Publishing

Company, Houston, 1995.

7. http://www.natick.army.mil/soldier/jocotas/ColPro_Papers/Larzelere.pdf, accessed on
May 15, 2007.

8. Jaroszczyk T., S. Fallon, J. Wake, T.J. Ptak ,“Media needs for automotive cabin
air treatment “, Nelson Industries, Inc. Stoughton, WI 53589-0600, Filtration
Conference Proceedings, Baltimore, Maryland, 1994.

9. Rosner D.E., P. Tandon, A.G Konstantopulous, “Local size distributions of

particles deposited by internal impaction on cylindrical target in dust-loaden
streams”, J.Aerosol science, 26, 1257-1279, 1995.

10. Happel J., H. Brenner, “Low Reynolds number hydrodynamics”, Kluwer

Academic Publishers, Dordrecht, 1991.

11. Happel J., “Viscous flow relative to arrays of cylinders”, American Institute of
Chemical Engineers J., 5,174-177, 1959.

12. Kuwabara S., “The forces experienced by randomly distributed parallel circular
cylinders of spheres in a viscous flow at small Reynolds number”, J.Physc. Soc.
Jpn., 14, 527-532, 1959.

62
13. Liu Z.G., P.K. Wang, “Numerical investigation of viscous flow fields around
multifiber filter”, Aerosol Sci. Technol, 25, 375-391, 1966.

14. Iliev O., V. Laptev, “Software for flow simulation in oil filters “, Computing and
Visualization in Science, Springer 6( 2-3), 139-146, 2004.

15. Wiegmann A., S. Rief., A. Latz,“Virtual Material Design and Air Filtration
Simulation Techniques inside GeoDict and Filter Dict “, Proceedings American
Filtration and Seperation Society annual conference, Atlanta, April 2005.

16. Srinivasan P., “Nanofiber incorporated filter media”, Master’s thesis, The
University of Akron, Akron, Ohio, 2005.

17. Chase G.G., “Transport Phenomena in Porous Media”, in Fluid Flow Handbook,
J.Saleh. ed., McGraw Hill, New York, 2002.

18. Chase G.G., P. Srinivasan, S. Andan, “Do Nanofibers Improve Filter

Performance? “, World Oil, 227(10), 95-98, 2006.

19. Vasudevan G., C.G. Shin, B. Raber, A. Suthar, G.G. Chase, “Modeling the start-
up stage of Coalescence Filtration “, FPSJ, 14(3), 169-176, 2002.

20. Michalewicz Z., “Genetic Algorithms + Data Structures = Evolution Programs”,

3rd edition, Springer, New York, 1999.

21. Izadifar M., M.Z. Jahromi, “Application of genetic algorithm for optimization of
vegetable oil hydrogenation process, Journal of food engineering 78,1-8, 2007.

22. Selbas R., O. Kizilkan, M. Reppich “A new design approach for shell-and-tube
heat exchangers using genetic algorithms from economic point of view”,
Chemical Engineering and Processing 45, 268-275, 2006.

23. Kalayaci V., M. Ouyang., K. Graham, “Polymer Nanofibers in high efficiency

filtration applications “, Filtration, 6(4), 2006.

24. Chase G.G., D.H. Reneker, “Nanofibers in Filter Media “, FPSJ, 16(2), 105-117,
2004.

25. Hajra M. G., K. Mehta, G.G. Chase, “Effects of humidity, temperature, and
nanofibers on drop coalescence in glass fiber media”, Separation and Purification
Technology, 30(1), 79-88, 2003.

26. Podgorski A., A. Bałazy, L. Gradon, “Application of nanofibers to improve the

filtration efficiency of the most penetrating aerosol particles in fibrous filters “,
Chemical Engineering Science, 61, 6804- 6815, 2006.

63
27. Grafe T.H., K.M. Graham, “Nanofiber webs from Electrospinning “, Proceedings
of non woven in Filtration - Fifth International Conference, Stuttgart, Germany,
March 2003.

28. Goldberg D.E., “Genetic Algorithms in Search, Optimization and Machine

Learning”, Addison-Wesley, New York, 1989.

29. Mehta K., “Effect of humidity, temperature and binder content in oil-droplet
coalescence in glass fiber filter media with and without polymer nanofibers”,
Master’s thesis, The University of Akron, Akron, Ohio, 2000.

64
APPENDICES

65
 APPENDIX A

FORTRAN PROGRAM

Genetic algorithm code and modeling equations are written in fortran program.

This program is designed to find the optimum design parameters of filter media
IMPLICIT REAL (A-H, O-Z)
IMPLICIT INTEGER (I-N)
!
PARAMETER (NPOP=1000, NVAR=5,NBIT=22)
! NPOP = population size of each generation
! NVAR = number of variables to be fitted by the GA
! NBIT = number of logic bits that represent each variable
! The parameter statement sets the values for these quantities. These
! quantities cannot be changed when the program is running (such as
! by adding lines to the menu). When changing these parameters they must
! be changed in all subroutines where the parameter statement occurs.
!

LOGICAL CHROM(NPOP,NVAR*NBIT), OLD(NPOP,NVAR*NBIT)

COMMON CHROM,OLD,N(NPOP),G(NPOP),F(NPOP),VAR(NVAR)
COMMON VOPT(NVAR),FOPT,NGEN,MGEN,PMUT,RAND,IFITTYPE
COMMON PABS,T,VO,D,FMDEN,DP,ZMAX,ZMIN,DMFMAX,DMFMIN,DNFMAX,DNFMI
N,ANFAMFMAX, ANFAMFMIN,THICKNESSCHK,FMAX,ANFAMFMIN, FMMASSMAX,
FMMASSMIN,THICKNESSCHK,DIAMFCHK,DIANFCHK, ARRATIOCHK,MASSMFCHK

! MAIN PROGRAM
CALL INIT ! Initialize variables
CALL CALC
CALL RESULTS ! Calculate generations
END
!
! INITIALIZE VARIABLES
SUBROUTINE INIT
IMPLICIT REAL (A-H, O-Z)
IMPLICIT INTEGER (I-N)
!
PARAMETER (NPOP=1000,NVAR=5,NBIT=22)
!
LOGICAL CHROM(NPOP,NVAR*NBIT), OLD(NPOP,NVAR*NBIT)
COMMON CHROM,OLD,N(NPOP),G(NPOP),F(NPOP),VAR(NVAR)
COMMON VOPT(NVAR),FOPT,NGEN,MGEN,PMUT,RAND,IFITTYPE

66
 COMMON PABS,T,VO,D,FMDEN,DP,ZMAX,ZMIN,DMFMAX,DMFMIN,DNFMAX,DNFMI
N,ANFAMFMAX, ANFAMFMIN,THICKNESSCHK,FMAX,ANFAMFMIN, FMMASSMAX,
FMMASSMIN,THICKNESSCHK,DIAMFCHK,DIANFCHK, ARRATIOCHK,MASSMFCHK
!REAL Z,D,PABS,T,VO,FMMASS,FMDEN
DIMENSION IVALUE(8)
CHARACTER CH1,CH2,CH3

MGEN = 0 ! COUNTER TO TRACK TOTAL NUMBER OF GENERATIONS

IFITTYPE = 1 ! FITNESS TYPE.
! TYPE = 1: THE LARGER THE FITNESS FUNCTION THE BETTER THE FIT
! TYPE = 2: THE SMALLER THE FITNESS FUNCTION THE BETTER THE FIT
!
! SET GLOBAL FITTNESS VALUE
! WRITE(*,*)'INPUT VARIABLES INIT'
IF(IFITTYPE.EQ.1)THEN
FOPT=1.E-10
ELSE
FOPT=1.E12
ENDIF
!
! Randomize the seed generator
CALL DATE_AND_TIME(CH1,CH2,CH3,IVALUE)
DO I = 1, IVALUE(8) ! CALL RANDOM NUMBER IN LOOP TO VARY
THE STARTING VALUE
CALL RANDOM_NUMBER(RAND)
ENDDO
!
! SET CHROMOSOMES FOR INITIAL POPULATION
DO I=1,NPOP
DO L=1,NVAR*NBIT
CALL RANDOM_NUMBER(RAND) ! RANDOM NUMBER
GENERATOR, RETURNS
IF(RAND.GT.0.5)THEN ! A NUMBER 0<= RAND <= 1
CHROM(I,L)=.TRUE.
ELSE
CHROM(I,L)=.FALSE.
ENDIF
ENDDO
ENDDO
! SAVING THE OUTPUT INTO GA.DAT FILE

OPEN(UNIT=7,FILE='C:\temp\ga.txt',STATUS='UNKNOWN')

! THIS WILL OPEN AN EXTERNAL FILE SAMPLE.TXT IN WHICH THE USER INPUTS ARE
STORED (VARIABLES ARE SAVED INTO TEXT FILE VISUAL BASIC PROGRAM.
! THIS OPEN STATEMENT WILL READ THE VARIABLES AND GOTO LOOP TERMINATES
WHEN END OF FILE IS REACHED.

!OPENING AN EXTERNAL FILE INPUTVAR.TXT WHERE THE DATA IS STORED FROM VISUAL
BASIC FILE

OPEN(UNIT=1,file="c:\vbprograms\inputvar.txt", STATUS="OLD")

67
!READING ALL THE VARIABLES FROM INPUTVAR.TXT

8
READ((1,*,END=12)PABS,T,VO,D,FMDEN,DP,THICKNESSCHK,ZMIN,ZMAX,DIAMFCHK,DMFM
IN,DMFMAX,DIANFCHK,DNFMIN,DNFMAX,ARRATIOCHK,ANFAMFMIN,ANFAMFMAX,MASS
MFCHK,FMMASSMIN,FMMASSMAX,NGEN,PMUT

GO TO 8

12 PRINT *

!
RETURN
END

!
! CALCULATE THE FITNESS OF ALL CHROMOSOMES OVER MULTIPLE GENERATIONS
SUBROUTINE CALC
IMPLICIT REAL (A-H, O-Z)
IMPLICIT INTEGER (I-N)
!
PARAMETER (NPOP=1000,NVAR=5,NBIT=22)
!
LOGICAL CHROM(NPOP,NVAR*NBIT), OLD(NPOP,NVAR*NBIT)
COMMON CHROM,OLD,N(NPOP),G(NPOP),F(NPOP),VAR(NVAR)
COMMON VOPT(NVAR),FOPT,NGEN,MGEN,PMUT,RAND,IFITTYPE
COMMON PABS,T,VO,D,FMDEN,DP,ZMAX,ZMIN,DMFMAX,DMFMIN,DNFMAX,DNFMI
N,ANFAMFMAX, ANFAMFMIN,THICKNESSCHK,FMAX,ANFAMFMIN, FMMASSMAX,
FMMASSMIN,THICKNESSCHK,DIAMFCHK,DIANFCHK, ARRATIOCHK,MASSMFCHK

DIMENSION VMAX(NVAR),VMIN(NVAR)
!REAL Z,D,PABS,T,VO,FMMASS,FMDEN

MOPT=0
RPOP=NPOP ! USE REAL VALUE OF NPOP IN CALCULATIONS
DO K=1,NGEN
MGEN=MGEN+1
WRITE(*,*)' GENERATION =',MGEN
!
! EVALUATE THE FITNESS OF EACH CHROMOSOME
FMAX=0.0
FMIN=100000000000.0 ! SET INITIAL FMIN TO SOME ARBITRARY
LARGE NUMBER
DO I=1,NPOP
! I = INDEX TO EACH CHROMOSOME
CALL DECODE(I) ! DECODE PARAMETERS FROM
BINARY TO DECIMAL
!
F(I)=FUNC() ! FUNC IS USER DEFINED FUNCTION TO BE
FITTED, F>=0
IF(F(I).GT.FMAX)THEN
IMAX=I ! STORE MAX VALUES
FMAX=F(I)
68
 DO J=1,NVAR
VMAX(J)=VAR(J)

ENDDO
ELSE IF(F(I).LT.FMIN)THEN
IMIN=I ! STORE MIN VALUES
FMIN=F(I)
DO J=1,NVAR
VMIN(J)=VAR(J)

ENDDO
ENDIF
ENDDO
!
! SAVE BEST STRING TO FILE
IF (IFITTYPE.EQ.1)THEN

IF(FMAX.GT.FOPT)THEN !FOPT IS THE OPTIMUM FOR ALL

GENERATIONS

CONVRG=(FMAX-FOPT)/FOPT/(MGEN-MOPT)
MOPT=MGEN !CONVERGENCE SHOULD GO
TOWARDS ZERO (10^-8)
FOPT=FMAX
DO J=1,NVAR
VOPT(J)=VMAX(J)
ENDDO
WRITE(7,*)MGEN,CONVRG,FOPT,VOPT
ENDIF
ELSE
IF(FMIN.LT.FOPT)THEN
CONVRG=(FOPT-FMIN)/FOPT/(MGEN-MOPT)
MOPT=MGEN
FOPT=FMIN
DO J=1,NVAR
VOPT(J)=VMIN(J)
ENDDO
WRITE(7,*)MGEN,CONVRG,FOPT,VOPT
ENDIF
ENDIF
!
! DETERMINE THE REPLICATION FRACTIONS FOR THE NEXT GENERATION
! See end of subroutine for definitions of fitness and replication parameters
FSUM=0.0
DO I=1,NPOP
FSUM=FSUM+F(I)
ENDDO
HSUM=(FSUM-NPOP*FMIN)/(FMAX-FMIN)
DO I=1,NPOP
HI=(F(I)-FMIN)/(FMAX-FMIN)
IF(IFITTYPE.EQ.1)THEN
G(I)=HI/HSUM ! LARGEST F IS BEST
ELSE
G(I)=(1.0-HI)/(RPOP-HSUM) ! SMALLEST F IS BEST
ENDIF
69
 N(I)=G(I)*NPOP+0.5 ! CALCULATE THE NUMBER
OF REPLICATIONS
! OF THE Ith CHROMOSOME
! ADD 0.5 FOR ROUNDUP
ENDDO
!
! FORM THE NEW POPULATION ARRAY
! TEMPORARILY SAVE THE PREVIOUS ARRAY
DO I=1,NPOP
DO L=1,NVAR*NBIT
OLD(I,L)=CHROM(I,L)
ENDDO
ENDDO
!
! COPY CHROMOSOME INTO NEW POLULATION BASED ON THE FITNESS
ISTOP=0
ISTART=0
DO I=1,NPOP
IF(N(I).GT.0.AND.ISTOP.LT.NPOP)THEN
ISTART=ISTOP+1
ISTOP=ISTOP+N(I)
IF(ISTOP.GT.NPOP)ISTOP=NPOP
DO J=ISTART,ISTOP
DO L=1,NVAR*NBIT
CHROM(J,L)=OLD(I,L)
ENDDO
ENDDO
ENDIF
ENDDO
!
! CROSSOVER: EXCHANGE CHROMOSOME BITS BETWEEN CHOMOSOMES FOR
! EXPANDED SEARCH SPACE

DO I=1,NPOP
! EACH CHROMOSOME HAS A RANDOM CHANCE OF CROSSING WITH ANOTHER
I1=I ! FIRST PARENT
!
CALL RANDOM_NUMBER(RAND)
I2=NPOP*RAND+0.5 ! SECOND PARENT
IF(I2.LT.1)I2=1
IF(I2.GT.NPOP)I2=NPOP
!
! RANDOMLY SELECT CROSSING POINT
CALL RANDOM_NUMBER(RAND)
LC=NVAR*NBIT*RAND+0.5
IF(LC.LT.1)LC=1
IF(LC.GT.NVAR*NBIT)LC=NVAR*NBIT
!
DO L=LC,NVAR*NBIT
CHROM(I1,L)=OLD(I2,L)
ENDDO
ENDDO
!
! MUTATION: ALLOW A SMALL FRACTION OF THE BITS TO MUTATE
!
70
 DO I=1,NPOP
!DETERMINE IF A MUTATION OCCURS
CALL RANDOM_NUMBER(RAND)
IF(RAND.LT.PMUT)THEN!MUTATION OCCURS WHEN TRUE
!SELECT BIT
CALL RANDOM_NUMBER(RAND)
LBIT=NVAR*NBIT*RAND+0.5
IF(LBIT.LT.1)LBIT=1
IF(LBIT.GT.NVAR*NBIT)LBIT=NVAR*NBIT
!
! DETERMINE MUTATION
CALL RANDOM_NUMBER(RAND)
IF(CHROM(I,LBIT))THEN
CHROM(I,LBIT)=.FALSE.
ELSE
CHROM(I,LBIT)=.TRUE.
ENDIF
ENDIF
ENDDO
!
ENDDO
RETURN
END
!
! DECODE FROM BINARY LOGICAL STRING TO BASE 10 REAL NUMBERS
SUBROUTINE DECODE(I)
!
IMPLICIT REAL (A-H, O-Z)
IMPLICIT INTEGER (I-N)
!
PARAMETER (NPOP=1000,NVAR=5,NBIT=22)
!
LOGICAL CHROM(NPOP,NVAR*NBIT), OLD(NPOP,NVAR*NBIT)
COMMON CHROM,OLD,N(NPOP),G(NPOP),F(NPOP),VAR(NVAR)
COMMON VOPT(NVAR),FOPT,NGEN,MGEN,PMUT,RAND,IFITTYPE
COMMON PABS,T,VO,D,FMDEN,DP,ZMAX,ZMIN,DMFMAX,DMFMIN,DNFMAX,DNFMI
N,ANFAMFMAX, ANFAMFMIN,THICKNESSCHK,FMAX,ANFAMFMIN, FMMASSMAX,
FMMASSMIN,THICKNESSCHK,DIAMFCHK,DIANFCHK, ARRATIOCHK,MASSMFCHK

DO J=1,NVAR
VAR(J)=0.0
ISTART=(J-1)*NBIT+1
ISTOP=J*NBIT
DO L = ISTART,ISTOP
IF(CHROM(I,L))THEN
VAR(J)=VAR(J)+2.0**(L-(J-1)*NBIT-1)
ENDIF
ENDDO
VAR(J)=VAR(J)/(2.0**NBIT-1.0)
ENDDO
RETURN
END
!
! WRITE RESULTS TO THE SCREEN SO THE USER MAY VIEW THEM
71
! THIS ROUTINE NEEDS TO BE MODIFIED FOR EACH APPLICATION
SUBROUTINE RESULTS
IMPLICIT REAL (A-H, O-Z)
IMPLICIT INTEGER (I-N)
!
PARAMETER (NPOP=1000,NVAR=5,NBIT=22)
!
LOGICAL CHROM(NPOP,NVAR*NBIT), OLD(NPOP,NVAR*NBIT)
COMMON CHROM,OLD,N(NPOP),G(NPOP),F(NPOP),VAR(NVAR)
COMMON VOPT(NVAR),FOPT,NGEN,MGEN,PMUT,RAND,IFITTYPE
COMMON PABS,T,VO,D,FMDEN,DP,ZMAX,ZMIN,DMFMAX,DMFMIN,DNFMAX,DNFMI
N,ANFAMFMAX, ANFAMFMIN,THICKNESSCHK,FMAX,ANFAMFMIN, FMMASSMAX,
FMMASSMIN,THICKNESSCHK,DIAMFCHK,DIANFCHK, ARRATIOCHK,MASSMFCHK

DIMENSION V(NVAR)
!REAL Z,D,PABS,T,VO,FMMASS,FMDEN

REWIND(7)
WRITE(*,*)'MGEN CONVERGENCE FOPT VARIABLES'
DO I=1,MGEN
READ(7,*,END=10)K,CONV,F1,V
WRITE(*,*)K,CONV,F1,V
10 CONTINUE
ENDDO

RETURN
END
!
FUNCTION FUNC()
IMPLICIT REAL (A-H, O-Z)
IMPLICIT INTEGER (I-N)

PARAMETER (NPOP=1000,NVAR=5,NBIT=22)
!
LOGICAL CHROM(NPOP,NVAR*NBIT), OLD(NPOP,NVAR*NBIT)
COMMON CHROM,OLD,N(NPOP),G(NPOP),F(NPOP),VAR(NVAR)
COMMON VOPT(NVAR),FOPT,NGEN,MGEN,PMUT,RAND,IFITTYPE
COMMON PABS,T,VO,D,FMDEN,DP,ZMAX,ZMIN,DMFMAX,DMFMIN,DNFMAX,DNFMI
N,ANFAMFMAX, ANFAMFMIN,THICKNESSCHK,FMAX,ANFAMFMIN, FMMASSMAX,
FMMASSMIN,THICKNESSCHK,DIAMFCHK,DIANFCHK, ARRATIOCHK,MASSMFCHK

! VARIABLES VAR ARE LIMITED TO A DIMENSIONLESS RANGE OF 0 TO 1

! THE ACTUAL VALUES USED IN THE FUNCTION ARE SCALED TO THE SEARCH RANGE

! MEDIA SIZE

IF (THICKNESSCHK.EQ.1) THEN
Z=(ZMAX-ZMIN)*VAR(3)+ZMIN

ELSE

Z=ZMIN
VAR(3)=0
END IF
72
PI = 3.141592654
V = PI*D**2/4*Z ! MEDIUM VOLUME, M3

! MICROFIBERS (MF)
IF (DIAMFCHK.EQ.1) THEN

DMF=(DMFMAX-DMFMIN)*VAR(2)+DMFMIN
ELSE

DMF=DMFMIN
VAR(2)=0
END IF

DMF = DMF/1000000.0 ! DIAMETER MF, M

IF(MASSMFCHK.EQ.1)THEN

FMMASS=(FMMASSMAX-MMASSMIN)*VAR(5)+FMMASSMIN
ELSE
FMMASS=FMMASSMIN
VAR(5)=0

END IF

VMF = FMMASS/FMDEN ! VOLUME OF MF, M3

ZMF = VMF*4.0/PI/DMF/DMF! LENGTH OF MF, M
AMF = PI*DMF*ZMF ! EXTERNAL AREA OF MF, M2
EMF = VMF/V ! MF VOLUME FRACTION

! NANOFIBERS (NF)

IF (ARRATIOCHK.EQ.1)THEN

ANFAMF=(ANFAMFMAX-ANFAMFMIN)*VAR(4)+ANFAMFMIN
! AREA RATIO OF NF TO MF
ELSE
ANFAMF=ANFAMFMIN
VAR(4)=0

END IF

ANF = ANFAMF*AMF ! EXTERNAL AREA NF, M2

IF(DIANFCHK.EQ.1)THEN

DNF=(DNFMAX-DNFMIN)*VAR(1)+DNFMIN

ELSE
DNF=DNFMIN
VAR(1)=0
END IF

DNF = DNF/1.0E9 ! DIAMETER OF NF, M

ZNF = ANF/PI/DNF ! LENGTH OF NF, M
VNF = PI/4.0*DNF**2*ZNF ! VOLUME OF NF, M3
ENF = VNF/V ! NF VOLUME FRACTION
73
 !
! MEDIA VALUES
ES = ENF+EMF ! FIBER VOLUME FRACTION
E = 1.0 - ES ! FILTER POROSITY
S = 0.1 ! LIQUID PHASE VOLUME FRACTION
!
! OPERATING CONDITIONS (ASSUMES AIR FOR GAS)
VGAS = VO/(1.0-S)/E ! PORE AVG GAS VELOCITY, M/S
ZLAMBDA=0.231548E-9*T*101.3/PABS ! GAS MEAN FREE PATH, M
GMW = 29.0 ! GAS MOLECULAR WT, KG/KGMOL
GR = 0.08205 ! GAS LAW CONST, ATM*M3/K/KGMOL
GDEN = PABS*GMW/101.34/T/GR ! GAS DENSITY, KG/M3
GVIS = 0.019E-3 ! GAS VISCOSITY, KG/M/S
!
! PARTICLES
! DP = 0.15 ! PARTICLE DIAMETER, MICRONS
DP1 = DP/1.0E6 ! PARTICLE DIAMETER, M
BOLTK = 1.380660E-23 ! BOLTZMANN CONST, JOULES/K
DIFF = BOLTK*T/GVIS/3/PI/DP1 ! PART DIFFUSIV KG/M/S
! BSL2, PG 529, EQ 17.4-3

! ALPHA CALCULATIONS
! ASSUMES DIFFUSION AND INTERCEPTION DOMINATE
! REF R.C. BROWN, AIR FILTRATION, PERGAMON,1993
!
! MICROFIBERS
ALPHAMF = 4.0*EMF/PI/DMF ! IDEAL ALPHA MF
ZETA = -(0.75+0.5*LOG(EMF)) ! HYDRO FACTOR
! BROWN PG 44
ZKN = 2.0*ZLAMBDA/DMF ! KNUDSEN NO.
IF(ZKN.LT.0.01)THEN ! CONTINUUM FLOW
EDRMF = 4.0/ZETA/(DMF**2)*(3.0*DIFF*ZETA*(DMF**2)*PI/8.0/VGAS&
+(DP1/2.0)**3)**(2.0/3.0) ! BROWN EQ 4.63
!
ELSE ! SLIP FLOW
ZPE = VGAS*DMF/DIFF ! PECLET NO.
ZETAP = -0.5*LOG(EMF)-0.52+0.64*EMF+1.43*(1.0-EMF)*ZKN
ED = 2.7*ZPE**(-2.0/3.0)&
*(1.0+0.39*(ZETAP*ZPE)**(-1.0/3.0)*ZKN)
R=1+DP1/DMF
ER=0.5/ZETAP*(1./R-R+2.*R*LOG(R)+2.86*ZKN*(1.+R)*(R-1)/R)
EPRIME=1.24*(R-1.0)**(2./3.)/(ZETAP*ZPE)**0.5
EDRMF = ED + ER + EPRIME ! BROWN EQS 4.64 - 4.68
END IF
!
! NANOFIBERS
ALPHANF = 4.0*ENF/PI/DNF ! IDEAL ALPHA MF
ZETA = -(0.75+0.5*LOG(ENF)) ! HYDRO FACTOR
! BROWN PG 44
ZKN = 2.0*ZLAMBDA/DNF ! KNUDSEN NO.
IF(ZKN.LT.0.01)THEN ! CONTINUUM FLOW
EDRNF = 4.0/ZETA/DNF**2*(3.0*DIFF*ZETA*DNF**2*PI/8.0/VGAS&
+(DP1/2.0)**3)**(2.0/3.0) ! BROWN EQ 4.63
!
ELSE ! SLIP FLOW
74
 ZPE = VGAS*DNF/DIFF ! PECLET NO.
ZETAP = -0.5*LOG(ENF)-0.52+0.64*ENF+1.43*(1.0-ENF)*ZKN
ED = 2.7*ZPE**(-2.0/3.0)&
*(1.0+0.39*(ZETAP*ZPE)**(-1.0/3.0)*ZKN)
R=1+DP1/DNF
ER=0.5/ZETAP*(1./R-R+2.*R*LOG(R)+2.86*ZKN*(1.+R)*(R-1)/R)
EPRIME=1.24*(R-1.0)**(2./3.)/(ZETAP*ZPE)**0.5
EDRNF = ED + ER + EPRIME ! BROWN EQS 4.64 - 4.68
ENDIF
!
! ALPHA
ALPHA = EDRNF*ALPHANF + EDRMF*ALPHAMF

! FI CALCULATIONS (CHASE NOTES PP 79-81)

!
! MICROFIBERS
ZKN = 2.0*ZLAMBDA/DMF
IF(ZKN.LT.0.01)THEN ! CONTINUUM
FIMF=4.0*GVIS*PI*VGAS/(-0.5*LOG(EMF)-0.738+EMF)
!
ELSE IF(ZKN.LT.0.25)THEN ! SLIP
FIMF=4.0*GVIS*PI*VGAS*(1.0+1.996*ZKN)&
/(-0.5*LOG(EMF)-0.75+EMF-0.25*EMF**2+1.996*ZKN&
*(-0.5*LOG(EMF)-0.25*(1.0-EMF**2)))
!
ELSE IF(ZKN.GT.10) THEN ! MOLECULAR
FIMF=7.194*GVIS*VGAS/ZKN
!
ELSE ! INTERMEDIATE
FI25=4.0*GVIS*PI*VGAS*(1.0+1.996*0.25)&
/(-0.5*LOG(EMF)-0.75+EMF-0.25*EMF**2+1.996*0.25&
*(-0.5*LOG(EMF)-0.25*(1.0-EMF**2)))
FI10=7.194*GVIS*VGAS/10
FIMF=(FI10-FI25)/(10.-0.25)*(ZKN-0.25)+FI25
END IF
!
! NANOFIBERS
ZKN = 2.0*ZLAMBDA/DNF
IF(ZKN.LT.0.01)THEN ! CONTINUUM
FINF=4.0*GVIS*PI*VGAS/(-0.5*LOG(ENF)-0.738+ENF)
!
ELSE IF(ZKN.LT.0.25)THEN ! SLIP
FINF=4.0*GVIS*PI*VGAS*(1.0+1.996*ZKN)&
/(-0.5*LOG(ENF)-0.75+ENF-0.25*ENF**2+1.996*ZKN&
*(-0.5*LOG(ENF)-0.25*(1.0-ENF**2)))
!
ELSE IF(ZKN.GT.10) THEN ! MOLECULAR
FINF=7.194*GVIS*VGAS/ZKN
!
ELSE ! INTERMEDIATE
FI25=4.0*GVIS*PI*VGAS*(1.0+1.996*0.25)&
/(-0.5*LOG(ENF)-0.75+ENF-0.25*ENF**2+1.996*0.25&
*(-0.5*LOG(ENF)-0.25*(1.0-ENF**2)))
FI10=7.194*GVIS*VGAS/10
FINF= (FI10-FI25)/(10.-0.25)*(ZKN-0.25)+FI25
75
 END IF
!
PGRAD=4.0/PI*(EMF*FIMF/DMF**2+ENF*FINF/DNF**2)
! QF=ALPHA/PGRAD
FUNC=QF
RETURN
END

76
 APPENDIX B

VISUAL BASIC PROGRAM

The user interface is written in visual basic program. The input from the user is sent to
fortran program and the calculation is done in the fortran executable file and the results is
returned back in visual basic interface.

PRIVATE SUB START_CLICK()

'THESE ARE THE OPTIMUM PARAMETERS
'GETTING THE VALUE FROM TEXT

THICKNESS1 = THICKNESS.TEXT

'RETURNS AN INTEGER SPECIFYING THE STARTING POSITION OF OCCURENCE OF STRING '-

' WITHIN ANOTHER
SPACEPOSZ = INSTR(THICKNESS1, "-")

'MID FUNCTION WILL EXTRACT THE SUBSTRING THE STRING FROM POSITION 1 TO THE
STRING POSITION WITHSPACEPOZ-1
ZMIN1 = MID(THICKNESS1, 1, SPACEPOSZ - 1)

'THIS WILL EXTRACT THE STRING FROM SPACEPOSZ+1 TO END OF THE STRING AND
STORE IT IN VARIABLE

ZMAX1 = MID(THICKNESS1, SPACEPOSZ + 1)

DIAMF1 = DIAMF.TEXT
SPACEPOSMF = INSTR(DIAMF1, "-")
DMFMIN1 = MID(DIAMF1, 1, SPACEPOSMF - 1)
DMFMAX1 = MID(DIAMF1, SPACEPOSMF + 1)
DIANF1 = DIANF.TEXT
SPACEPOSNF = INSTR(DIANF1, "-")
DNFMIN1 = MID(DIANF1, 1, SPACEPOSNF - 1)
DNFMAX1 = MID(DIANF1, SPACEPOSNF + 1)
AREARATIO1 = AREARATIO.TEXT
SPACEPOSAR = INSTR(AREARATIO1, "-")
ARMIN1 = MID(AREARATIO1, 1, SPACEPOSAR - 1)
ARMAX1 = MID(AREARATIO1, SPACEPOSAR + 1)
MASSMF1 = MASSMF.TEXT
SPACEPOSMMF = INSTR(MASSMF1, "-")
MMFMIN1 = MID(MASSMF1, 1, SPACEPOSMMF - 1)
MMFMAX1 = MID(MASSMF1, SPACEPOSMMF + 1)

TEXT2.TEXT = ""
77
TEXT7.TEXT = " "
MSCHART1.VISIBLE = FALSE
VIEWRESULTS.ENABLED = FALSE
SAVERESULTS.ENABLED = FALSE
VIEWPLOTS.ENABLED = FALSE

' UNIT CONVERSION

IF COMBO1(0).TEXT = "PSI" THEN
PRESSURE1 = VAL(PRESSURE(0).TEXT) / 0.145
ELSE
PRESSURE1 = VAL(PRESSURE(0).TEXT)
END IF

IF COMBO2(0).TEXT = "C" THEN

TEMPERATURE1 = VAL(TEMPERATURE(0).TEXT) + 273.15
ELSE
TEMPERATURE1 = VAL(TEMPERATURE(0).TEXT)
END IF

IF COMBO3(0).TEXT = "CM/SEC" THEN

FACEVELOCITY1 = VAL(FACEVELOCITY(0).TEXT) / 100
ELSE
FACEVELOCITY1 = VAL(FACEVELOCITY(0).TEXT)
END IF

IF COMBO4(0).TEXT = "CM" THEN

DIAFILTER1 = VAL(DIAFILTER(0).TEXT) / 100
ELSE
DIAFILTER1 = VAL(DIAFILTER(0).TEXT)
END IF

IF COMBO5(0).TEXT = "G/CM3" THEN

DENSITY1 = VAL(DENSITY(0).TEXT) * 1000
ELSE
DENSITY1 = VAL(DENSITY(0).TEXT)
END IF

IF COMBO9.TEXT = "M" THEN

DIAP1 = VAL(DIAP(1).TEXT) / 0.00001
ELSE
DIAP1 = VAL(DIAP(1).TEXT)
END IF

THICKNESSCHK1 = THICKNESSCHK.VALUE
ARRATIOCHK1 = ARRATIOCHK.VALUE
DIANFCHK1 = DIANFCHK.VALUE
DIAMFCHK1 = DIAMFCHK.VALUE
MASSMFCHK1 = MASSMFCHK.VALUE
IF THICKNESSCHK.VALUE = 0 THEN
TEXT1.VISIBLE = TRUE
THICKNESS.VISIBLE = FALSE
IF COMBO6.TEXT = "CM" THEN
ZMIN1 = VAL(TEXT1.TEXT) / 100
ELSE
ZMIN1 = VAL(TEXT1.TEXT)
78
 END IF
ZMAX1 = 0

ELSE
THICKNESS.VISIBLE = TRUE
TEXT1.VISIBLE = FALSE
END IF

IF ARRATIOCHK.VALUE = 0 THEN
TEXT4.VISIBLE = TRUE
AREARATIO.VISIBLE = FALSE
ARMIN1 = TEXT4.TEXT
ARMAX1 = 0
ELSE
AREARATIO.VISIBLE = TRUE
TEXT4.VISIBLE = FALSE
END IF
IF DIAMFCHK.VALUE = 0 THEN
TEXT6.VISIBLE = TRUE
DIAMF.VISIBLE = FALSE
IF COMBO7.TEXT = "M" THEN
DMFMIN1 = VAL(TEXT6.TEXT) / 0.00001
ELSE
DMFMIN1 = VAL(TEXT6.TEXT)
END IF
DMFMAX1 = 0
ELSE
DIAMF.VISIBLE = TRUE
TEXT6.VISIBLE = FALSE
END IF
IF DIANFCHK.VALUE = 0 THEN
TEXT3.VISIBLE = TRUE
DIANF.VISIBLE = FALSE
IF COMBO8.TEXT = "M" THEN
DNFMIN1 = VAL(TEXT3.TEXT) / 0.00000001
ELSE
DNFMIN1 = VAL(TEXT3.TEXT)
END IF
DNFMAX1 = 0
ELSE
DIANF.VISIBLE = TRUE
TEXT3.VISIBLE = FALSE
END IF

IF MASSMFCHK.VALUE = 0 THEN
TEXT5.VISIBLE = TRUE
MASSMF.VISIBLE = FALSE
IF COMBO10.TEXT = "G" THEN
MMFMIN1 = TEXT5.TEXT / 1000
ELSE
MMFMIN1 = TEXT5.TEXT
END IF
MMFMAX1 = 0
ELSE
MASSMF.VISIBLE = TRUE
79
 TEXT5.VISIBLE = FALSE
END IF

'GENETIC ALGORITHM PARAMETERS

POPULATION1 = POPULATION.TEXT
GENERATION1 = GENERATION.TEXT
MUTATION1 = MUTATION.TEXT
CONVERGENCE1 = CONVERGENCE.TEXT

IF THICKNESSCHK1 = 0 THEN
Z = ZMIN1
ELSE
Z = ZMIN1
Z = ZMAX1
END IF

IF DIAMFCHK1 = 0 THEN

DMF = DMFMIN1
ELSE
DMF = DMFMIN1
DMF = DMFMAX1
END IF

IF MASSMFCHK1 = 0 THEN
FMMASS = MMFMIN1
ELSE
FMMASS = MMFMIN1
FMMASS = MMFMAX1
END IF

IF ARRATIOCHK1 = 0 THEN
ANFAMF = ARMIN1
ELSE
ANFAMF = ARMIN1
ANFAMF = ARMAX1
END IF
IF DIANFCHK1 = 0 THEN
DNF = DNFMIN1
ELSE
DNF = DNFMIN1
DNF = DNFMAX1
END IF
DP = DIAP1
D = DIAFILTER1
FMDEN = DENSITY1

PI = 3.141592654
VO = FACEVELOCITY1
V = PI * D ^ 2 / 4 * Z ' MEDIUM VOLUME M3

' MICROFIBERS (MF)

DMF = DMF / 1000000 ' DIAMETER MF, M
80
 VMF = FMMASS / FMDEN ' VOLUME OF MF, M3
ZMF = VMF * 4 / PI / DMF / DMF ' LENGTH OF MF, M
AMF = PI * DMF * ZMF ' EXTERNAL AREA OF MF, M2
EMF = VMF / V ' MF VOLUME FRACTION
ANF = ANFAMF * AMF ' EXTERNAL AREA NF, M2
DNF = DNF / 1000000000 ' DIAMETER OF NF, M
ZNF = ANF / PI / DNF ' LENGTH OF NF, M
VNF = PI / 4 * DNF ^ 2 * ZNF ' VOLUME OF NF, M3
ENF = VNF / V ' NF VOLUME FRACTION
'
' MEDIA VALUES
ES = ENF + EMF ' FIBER VOLUME FRACTION
E = 1 - ES ' FILTER POROSITY

IF E < 0 THEN

MSGBOX " CHECK THE PARAMETERS : POROSITY " & FORMAT(E, "#.00E-00")
PRESSURE(0).SETFOCUS

'OPENING A TEXT FILE TO STORE THE INPUT DATA

ELSE

OPEN "C:\INPUTVAR.TXT" FOR OUTPUT AS #2

PRINT #2, PRESSURE1, TEMPERATURE1, FACEVELOCITY1, DIAFILTER1, DENSITY1, DIAP1,

THICKNESSCHK1, ZMIN1, ZMAX1, DIAMFCHK1, DMFMIN1, DMFMAX1, DIANFCHK1,
DNFMIN1, DNFMAX1, ARRATIOCHK1, ARMIN1, ARMAX1, MASSMFCHK1, MMFMIN1,
MMFMAX1, GENERATION1, MUTATION1, CONVERGENCE1
CLOSE #2

'FORTRAN EXECUTABLE FILE

W$ = "C:\GAVBCONV.EXE"
'USED SHELL FUNCTION TO RUN THE FORTRAN EXECUTABLE FILE

VIEWRESULTS.ENABLED = FALSE
LABEL11.CAPTION = "PROGRAM IS RUNNING"
CALL SHELLANDWAIT(W$, 8000)
LABEL11.CAPTION = "PROGRAM IS STOPPED"
VIEWRESULTS.ENABLED = TRUE

END IF
END SUB

OPTION EXPLICIT
'

PRIVATE DECLARE FUNCTION OPENPROCESS LIB "KERNEL32" _ (BYVAL

DWDESIREDACCESS AS LONG, BYVAL BINHERITHANDLE AS LONG, _
BYVAL DWPROCESSID AS LONG) AS LONG

PRIVATE DECLARE FUNCTION GETEXITCODEPROCESS LIB "KERNEL32" _

(BYVAL HPROCESS AS LONG, LPEXITCODE AS LONG) AS LONG
81
PRIVATE CONST STATUS_PENDING = &H103&
PRIVATE CONST PROCESS_QUERY_INFORMATION = &H400
'DIM VALUE1 AS STRING
'
PUBLIC FUNCTION SHELLANDWAIT(EXEFULLPATH AS STRING, _
OPTIONAL TIMEOUTVALUE AS LONG = 0) AS BOOLEAN

DIM LINST AS LONG

DIM LSTART AS LONG
DIM LTIMETOQUIT AS LONG
DIM SEXENAME AS STRING
DIM LPROCESSID AS LONG
DIM LEXITCODE AS LONG
DIM BPASTMIDNIGHT AS BOOLEAN

DIM I AS INTEGER
ON ERROR GOTO ERRORHANDLER

LSTART = CLNG(TIMER)
SEXENAME = EXEFULLPATH

'DEAL WITH TIMEOUT BEING RESET AT MIDNIGHT

IF TIMEOUTVALUE > 0 THEN
IF LSTART + TIMEOUTVALUE < 86400 THEN
LTIMETOQUIT = LSTART + TIMEOUTVALUE

ELSE
LTIMETOQUIT = (LSTART - 86400) + TIMEOUTVALUE
BPASTMIDNIGHT = TRUE
' MSGBOX "PROGRAM IS RUNNING"
END IF
END IF

LINST = SHELL(SEXENAME, VBMINIMIZEDNOFOCUS)

' LINST = SHELL(SEXENAME, VBMAXIMIZEDFOCUS)

LPROCESSID = OPENPROCESS(PROCESS_QUERY_INFORMATION, FALSE, LINST)

DO
CALL GETEXITCODEPROCESS(LPROCESSID, LEXITCODE)
DOEVENTS
IF TIMEOUTVALUE AND TIMER > LTIMETOQUIT THEN
IF BPASTMIDNIGHT THEN
IF TIMER < LSTART THEN EXIT DO
ELSE
EXIT DO
END IF
END IF

LOOP WHILE LEXITCODE = STATUS_PENDING

MSGBOX "COMPLETED EXECUTING THE MODEL"

SHELLANDWAIT = TRUE
82
ERRORHANDLER:
SHELLANDWAIT = FALSE
EXIT FUNCTION
END FUNCTION

PRIVATE SUB VIEWPLOTS_CLICK()

DIM SFILE AS STRING, SMSG AS STRING
DIM VAL1() AS SINGLE, VAL2() AS VARIANT, VAL3() AS SINGLE, VAL4() AS SINGLE
DIM VAL5() AS SINGLE, VAL6() AS VARIANT, VAL7() AS SINGLE, VAL8() AS SINGLE
DIM FIRST AS SINGLE
DIM IIDX AS LONG
DIM CONV AS SINGLE
DIM K AS INTEGER
'DIM DATATHK(0 TO 10) AS STRING

DIM ARRDATA(100, 1 TO 2)
DIM I AS INTEGER, COUNT AS INTEGER
DIM MAXN1, MIN1
DIM MAXN2, MIN2
COMBO14.CLEAR
COMBO15.CLEAR
MSCHART1.VISIBLE = TRUE

'OPENING AN EXTERNAL FILE WHERE THE OUTPUT IS STORED.

'THE CALCULATION IS DONE IN FORTRAN FILE AND THE OUTPUT IS STORED IN GA.DAT
FILE.

SFILE = "C:\GA.TXT"

'READ ALL THE VARIABLES UNTIL END OF FILE IS REACHED

IIDX = 0
OPEN SFILE FOR INPUT AS #1
DO WHILE NOT EOF(1)
REDIM PRESERVE VAL1(IIDX)
REDIM PRESERVE VAL2(IIDX)
REDIM PRESERVE VAL3(IIDX)
REDIM PRESERVE VAL4(IIDX)
REDIM PRESERVE VAL5(IIDX)
REDIM PRESERVE VAL6(IIDX)
REDIM PRESERVE VAL7(IIDX)
REDIM PRESERVE VAL8(IIDX)

INPUT #1, VAL1(IIDX), VAL2(IIDX), VAL3(IIDX), VAL4(IIDX), VAL5(IIDX),

VAL6(IIDX), VAL7(IIDX), VAL8(IIDX)
IIDX = IIDX + 1

LOOP

CLOSE
FOR IIDX = LBOUND(VAL1) TO UBOUND(VAL1)
CONV = VAL2(IIDX)

83
 VAL6(IIDX) = (ZMAX1 - ZMIN1) * VAL6(IIDX) + ZMIN1
VAL5(IIDX) = (DMFMAX1 - DMFMIN1) * VAL5(IIDX) + DMFMIN1
VAL4(IIDX) = (DNFMAX1 - DNFMIN1) * VAL4(IIDX) + DNFMIN1
VAL7(IIDX) = (ARMAX1 - ARMIN1) * VAL7(IIDX) + ARMIN1
VAL8(IIDX) = (MMFMAX1 - MMFMIN1) * VAL8(IIDX) + MMFMIN1

SMSG = SMSG & _ VAL1(IIDX) & " " & VAL2(IIDX) & _ " " & VAL3(IIDX) & " "&
VAL4(IIDX) & " " & VAL5(IIDX) & " " & VAL6(IIDX) & " " & VAL7(IIDX) & "
" & VAL8(IIDX) & " " & VBCRLF
COMBO14.ADDITEM (VAL1(IIDX))
COMBO15.ADDITEM (VAL2(IIDX))
NEXT IIDX
COUNT = COMBO14.LISTCOUNT
MSCHART1.REFRESH
FOR I = 2 TO COUNT

ARRDATA(I, 1) = COMBO14.LIST(I - 1)
ARRDATA(I, 2) = COMBO15.LIST(I - 1)

NEXT I

MAXN1 = COMBO15.LIST(1)
MAXN2 = COMBO14.LIST(1)
FOR I = 2 TO COUNT
IF MAXN1 < VAL(COMBO15.LIST(I)) THEN
MAXN1 = COMBO15.LIST(I)
END IF

IF MAXN2 < VAL(COMBO14.LIST(I)) THEN

MAXN2 = COMBO14.LIST(I)
END IF
NEXT I

MSCHART1.VISIBLE = TRUE

MSCHART1.PLOT.AXIS(VTCHAXISIDX).AXISTITLE = "GENERATION"
MSCHART1.PLOT.AXIS(VTCHAXISIDY).AXISTITLE = "CONVERGENCE"
WITH MSCHART1.PLOT.AXIS(VTCHAXISIDY).VALUESCALE
.AUTO = FALSE
.MINIMUM = 0.000000001
.MAXIMUM = MAXN1

MSCHART1.PLOT.AXIS(VTCHAXISIDY).VALUESCALE.MAJORDIVISION = 0.001
MSCHART1.PLOT.AXIS(VTCHAXISIDY).VALUESCALE.MINORDIVISION = 0.0002
MSCHART1.PLOT.AXIS(VTCHAXISIDY).AXISSCALE.HIDE = FALSE

END WITH
WITH MSCHART1.PLOT.AXIS(VTCHAXISIDX).VALUESCALE
.AUTO = FALSE
.MINIMUM = 1
.MAXIMUM = MAXN2
MSCHART1.PLOT.AXIS(VTCHAXISIDX).VALUESCALE.MAJORDIVISION = 2
MSCHART1.PLOT.AXIS(VTCHAXISIDX).VALUESCALE.MINORDIVISION = 0.4
84
 MSCHART1.PLOT.AXIS(VTCHAXISIDX).AXISSCALE.HIDE = FALSE
END WITH
MSCHART1.CHARTDATA = ARRDATA
MSCHART1.PLOT.UNIFORMAXIS = FALSE
IF (THICKNESSCHK.VALUE = 0) THEN
MSGBOX "THICKNESS ENABLED"
COMBO13.REMOVEITEM (1)
END IF
END SUB

85