Liquid Permeability of Ceramic Foam Filters

Liquid Permeability of Ceramic Foam
Filters
Kexu Zhang
Light Metals Production

Submission date: July 2012
Supervisor:
Ragnhild Aune, IMTE
Norwegian University of Science and Technology

Department of Materials Science and Engineering
Liquid Permeability of Ceramic

Foam Filters
Kexu Zhang
Supervisor: Prof. Ragnhild E. Aune
Co-Supervisor: Mark W. Kennedy
Norwegian University of Science and Technology

Department of Materials Science and Technology
2012
Declaration of Authorship
Jeg erklrer herved at arbeidet har blitt utfrt selvstendig og i samsvar med Reglement for
Sivilarkitekt- og Sivilingenireksamen ved NTNU.
I hereby declare that this work has been carried out independently and in compliance with the
examination regulations of the Norwegian University of Science and Technology, NTNU.
Trondheim, July 2012
--------------------------------------------Kexu Zhang
Acknowledgements
I modestly extend my sincere appreciation to the following people for their psychological and
technical support during the writing of this report.
Prof. Ragnhild Elizabeth Aune is the main supervisor for this thesis. Special thanks for
organizing and realizing this thesis work, for guidance and good discussions.
The help of Mark William Kennedy for the introduction and consolidation of my knowledge
about fluid dynamics, designing and building the experiment set-up and support from the
initial to the final level is greatly appreciated. As an addition, he supported me with
COMSOL-Simulations and background knowledge about 2-D simulations. Mr. Kennedy has
carried out the numerical simulations and analytical modelling of the experimental results. It
is due to his enthusiasm that this work has come this far.
Thanks to Robert Fritzsch, for the support with the tortuosity experiments, and introducing me
to the Scanning Electron Microscope (SEM).
The experiments were performed together with Kurt Sandaunet from SINTEF and his
valuable contribution in the execution of the experiments is acknowledged.
Deepest gratitude is also due to Egil Torsetnes at NTNU, Trondheim, Norway, for helping
with the design, construction and prompt repair of the experimental apparatus.
Trondheim, July 2012
--------------------------------------------Kexu Zhang
Abstract
This project is in support of the PhD project: Removal of Inclusions from Liquid Aluminium
using Electromagnetically Modified Filtration. The purpose of the present project was to
measure the tortuosity and permeability of ~50 mm thick 30, 40, 50 and 80 pores per inch (ppi)
commercial alumina Ceramic Foam Filters (CFFs). Measurements have been taken of the cell
(pore), the window and strut sizes, as well as the porosity, tortuosity and liquid permeability.
Water velocities from ~0.015-0.77 m/s have been used to derive both the first order (Darcy)
and the second order (Non-Darcy) terms for use with the Forchheimer equation. Experiments
were made using 49 mm diameter straight through and 101 mm diameter expanding flow
field designs. Experimental data are compared with simulation results made using COMSOL
4.2a 2D axial symmetric Finite Element Modelling (FEM). Permeability results are
correlated using directly measurable parameters. Development of improved wall sealing (49
mm) and elimination of wall effects (101 mm), has lead to a high level of agreement between
experimental, analytical and FEM methods (0-7% on predicted pressure drop) for both
types of experiments. The liquid permeability experiments were also used to determine the
variability of permeability between different filters of the same ppi, and permeability variation
within the same CFF for changes position. Tortuosity has been determined by two inductive
methods, one using cold solidified samples at 60 kHz and the other using liquid metal at 50 Hz,
giving comparable results.
Contents
1. Introduction.....................................................................................................................................................................1
2. Theoretical Background .............................................................................................................................................2
2.1 CFF ...............................................................................................................................................................................2
2.2 Permeability .............................................................................................................................................................5
2.3 Fluid Mechanics ................................................................................................................................................... 10
2.3.1 Viscosity ......................................................................................................................................................... 10
2.3.2 Reynolds Number ....................................................................................................................................... 10
2.3.3 Entrance Length .......................................................................................................................................... 12
2.4 Tortuosity .............................................................................................................................................................. 14
3. Experimental Apparatus and Procedure .......................................................................................................... 19
3.1 Optical Microscopy ............................................................................................................................................. 19
3.1.1 Optical Microscopy - Theory .................................................................................................................. 19
3.1.2 Optical Microscopy - Experimental Procedure .............................................................................. 20
3.2. Optical Scanning ................................................................................................................................................. 20
3.3. Energy-Dispersive X-ray Spectroscopy .................................................................................................... 21
3.3.1 Energy-Dispersive X-ray Spectroscopy- Theory ........................................................................... 21
3.3.2 Energy-Dispersive X-ray Spectroscopy - Sample Preparation ................................................ 21
3.4 Pressure Drop Experiments ........................................................................................................................... 21
3.5 Tortuosity Experiments ................................................................................................................................... 26
4. Results and Discussion............................................................................................................................................. 28
4.1 CFF Porosities....................................................................................................................................................... 28
4.2 Cell, Window & Strut Diameter ..................................................................................................................... 30
4.3 Pressure Drop Results ...................................................................................................................................... 32
4.3.1 Improvement of Procedures .................................................................................................................. 34
4.3.2 Required Inlet Length to Fully Develop the Flow Regime ......................................................... 38
4.3.3 Permeability of Different Types of Filters. ....................................................................................... 40
4.3.4 Correlation of Pressure Drop ................................................................................................................ 45
4.3.5 Permeability of Different Filters with the Same ppi. ................................................................... 47
4.3.6 Permeability for Different Samples taken from the Same Filter. .......................................... 51
4.3.7 Permeability Changing with Fluid Flow Direction ....................................................................... 53

4.4 Filter Tortuosity .................................................................................................................................................. 54
5. CFD Modelling.............................................................................................................................................................. 57
5.1 Introduction .......................................................................................................................................................... 57
5.2 Theory ..................................................................................................................................................................... 58
5.2.1 Materials ........................................................................................................................................................ 58
5.2.2 Physics ............................................................................................................................................................ 58
5.3 FEM Model Details .............................................................................................................................................. 63
5.3.1 49 mm Filter Finite Element Modelling ............................................................................................ 63
5.3.2 101mm Filter Finite Element Modelling ........................................................................................... 66
5.4 Comparison of CFD Results with Experimental Data........................................................................... 69
6. Conclusions ................................................................................................................................................................... 72
7. Further Work ............................................................................................................................................................... 74
References .......................................................................................................................................................................... 75
Appendix ............................................................................................................................................................................. 80
Chapter 1
Introduction
Ceramic Foam Filters (CFFs) have been used commercially in the foundry industry for more
than four decades for the production of premium quality aluminium castings for use in such
applications as in the aerospace industry.[ 1 ] In recent years, CFFs have become more
important in the physical purification of metals, for the ferrous and nonferrous industries. The
nonferrous industry uses the largest amount of this filtering media, especially the aluminium
casting industry. In 1992, Aubrey and Dore[2] reported that eight million metric tons of
aluminium was filtered using CFFs, and this means that about 50 percent of all the aluminium
produced in the world in that year, was purified through this filtering media.
High-temperature alloys also use CFFs. Sutton[3] announced results on the filtration of
super-alloys in 1985. In 1993, Garing and Cummings reported the successful filtration of
carbon and alloy steels using CFFs.[4] Porous ceramics are now widely applied as materials
for filtration of fluid in high-temperature applications.[5,6,7]
It is widely accepted that based on their structural properties, CFFs can remove exogenous
and indigenous inclusions in the melts,[8] and moreover, have low flow resistance and high
filtration efficiency for particles over 10-20 microns.[9] The structure of the CFF creates a
unique, tortuous path for the fluid to flow, which captures inclusions and allows clean,
smooth-flowing metal to exit into the mould cavity.[10] CFFs are generally accepted as the best
filters for casting. Their main advantages are: high filtration efficiency, turbulence reduction,
refractoriness, and erosion resistance, for the most demanding casting applications.[11]
Permeability is an important parameter for the characterization of CFFs, since it is required to
predict the flow rate obtainable under a given pressure drop or to be able to predict the
pressure drop necessary to achieve a specific flow rate. The relationship between these
quantities can be expressed as a function of the fluid flow, and medium properties and is
obtained by fitting the experimental data with permeability equations. [12,13]
Chapter 2
Theoretical Background
2.1 CFF
Cellular structures can be found in natural materials, such like wood, bone, coral, etc.,
significantly, nature has optimized certain mechanical properties, for example: stiffness,
strength and mass, in an efficient manner.[14] About 50 years ago, the development of
synthetic materials manufacturing methods inspired investigation into the structure and
properties of cellular materials.[15,16] For ceramic foam, their good thermal resistance and high
porosity characteristics led to use as filters in the purification of liquid metals, which are still
the largest application today. Ceramic foam is also used for catalytic combustion, burner
enhancers, soot filters for diesel engine exhausts, catalyst supports, and biomedical
devices.[17,18,19,20]
It is widely accepted that based on their structural properties, CFFs can remove exogenous
and indigenous inclusions in the melts,[21] and moreover, to have low flow resistance and high
filtration efficiency.[9] The Figure.1 shows a ceramic foam filter and bowl for use in the
removal of solid inclusions from liquid metal. The filters are integrated into filter boxes and
applied to the casting process directly in front of the casting unit.
Figure.1 CFF and filter bowl.[22]
The type of filtration occurring in a CFF can be divided into deep bed, cake or mixed
filtration. The efficiency is dependent on the pores ppi (pores per inch) of the filter material,
as shown in Figure.2[23] and also on the size distribution of the particulates to be removed. The
2
ppi refers to the plastic material used to make the ceramic foam and is found from the average
number of pore boundaries encountered per inch.[24] More recently the CFFs are graded into
different ppi categories according to air permeability.[25] In Figure.2, the line between each
column of filtration efficiency ranges marks the mean efficiency.
Figure.2 Relative performance of different inline filtration systems measured by the N15 LiMCA.
The size ranges of typical commercial CFFs are 10 to 80 ppi (see Figure.3). 20-40 ppi are the
most commonly used,[26] for industrial cast houses to reach both the desired casting rate and
achieve minimally acceptable filtration efficiency. The 10 to 20 ppi filters have poor particle
retention, and are often used for filtration of entrapped bulk, surface borne oxides and other
large particulates generated during melting, holding and transport. High pore density filters
with 60-80 ppi are only useful for quality sensitive applications, such as surface critical
extrusion and sheet products.[27]
a)
c)
b)
d)
Figure.3 Sivex CFFs a) 30 ppi, b) 40 ppi, c) 50 ppi, d) 80 ppi.
The structure of the CFF creates a unique, tortuous path for the fluid to flow, which captures
inclusions and allows clean, smooth-flowing metal to exit into the mould cavity.[27] The
important filter parameters are effective porosity (i.e., the porosity that effectively contributes
to the fluid flow), tortuosity, specific surface area and pore diameter.[28] The filtration process
is also dependent on: the alloy type, grain refiner, casting rate, metal temperature, etc.[29]
Grain refiner added before the filters has a particularly negative impact on filtration efficiency.
CFFs are generally accepted as the best filters for casting based on cost, ease of use and
acceptable performance characteristics. Their main advantages are: high filtration efficiency,
turbulence reduction, refractoriness, and erosion resistance, for the most demanding casting
applications. Deep bed particle filters have generally better filtration performance, but are
more difficult and costly to operate.
The CFFs used for this research are commercial high alumina filters produced by SIVEX.
The chemical composition of the CFF is mainly alumina (the exact composition is proprietary
commercial information), which was phosphate bonded, as indicated in Figure.4. Table I
shows the detailed composition of the alumina CFFs produced by Jiangxi Jintai Special
Material Co., Ltd, which are taken as indicative of the alumina CFF material used in this
study. It should be noted that the carbon in Figure.4 can be ignored, since it originates from
the conductive carbon coating added to the sample before analysis. The porosity of the filter is
around 85 to 90%. The ceramic particle density is 3.480.02 g/cm3, based on private
communication with Norsk Hydro.
4
Figure.4 Electron Probe Micro Analysis(EPMA) of CFFs.
Table I Chemical composition for Ceramic Foam Filters (CFF) made from alumina.[30]
* Others are assumed to be mainly Phosphate (P2O5) binders
2.2 Permeability
Permeability is an important parameter for the characterization of CFFs, since it is required to
predict the flow rate obtainable under a given pressure drop or to be able to predict the
pressure drop necessary to achieve a specific flow rate. The relationship between these
quantities can be expressed as a function of the fluid flow, and medium properties and is
obtained by fitting the experimental data with permeability equations. [31,32]
There are two main equations used in the literature to represent the permeability of CFFs,
Darcys law and Forchheimers equation.
Darcys law[33] is shown as follows:
q
k1
(1)
Where q is the flux or discharge per unit area, with units of length per time, [m/s], k1 is the
Darcian permeability constant [m2], is the fluid dynamic viscosity [Pas] and P is the
pressure gradient vector [Pa/m], P is pressure drop [Pa].
The Darcys law only accurately describes the pressure flow relationship at low fluid
velocities, and becomes increasingly inaccurate at high velocities (i.e. turbulent Reynolds
numbers).[34] At high velocity, Forchheimers equation has been reliably employed in the
literature to predict the pressure flow relationship of ceramic filters in a broader velocity
range.[35] For an incompressible fluid and a rigid, homogeneous (i.e. isotropic) ceramic filter,
Forchheimers equation can be represented as follows[36]:
P
Vs Vs2
L k1
k2
(2)
Where P is pressure drop [Pa], L is the filter thickness [m], is the fluid dynamic viscosity
[Pas], is the fluid density [kg/m3], the constants k1 and k2 are called the Darcian and
non-Darcian permeability coefficients respectively and Vs is the superficial fluid velocity
[m/s].
The term P / L is a normalised pressure drop per unit length, i.e. the pressure gradient (see
Equation (3)). This assumes a pressure gradient in only one axis, which is essentially true
for the straight through design and not strictly true for the expanding flow field design,
which experiences gradients in both the z and r-axes.
P
P
L
(3)
The term Vs / k1 represents the contribution to flow resistance due to friction between fluid
layers and the pore walls (i.e. the viscous loss term). The term Vs / k 2 represents the
contributions of inertia and turbulence.[32, 37]
One publication[34] reported that computation of k1 and k2 using Forchheimers equation
generally worsens when less data are included in the velocity curve, particularly at low flow
velocities. The Darcian permeability constant k1 then varies more than the non-Darcian
permeability constant k2. This is likely due to the effect of the high pressure gradients at
higher velocities dominating the output of the regression techniques implemented by the
previous investigators.
In 1952, Ergun[38] proposed expressions to describe k1 and k2 as follows:
k1
k2
3d p2
150(1 ) 2
(4)
3d p
(5)
1.75(1 )
Where is the fractional porosity [unitless], and dp is the mean particle diameter of the foam
filter [m].
Forchheimers equation can be modified using the terms of Ergun:
P
(1 )2
(1 ) 2
150
Vs 1.75 3
V
3 2
L
dp
dp s
(6)
Although dp is not ambiguous as a parameter, Erguns equation was derived for packed beds
of solids and there is no precise way to assign a value to this variable for a porous filter media.
The major challenge to the application of these equations to a porous media is therefore to
define an equivalent mean particle diameter. It is possible to apply the Ergun formula using
alternately: the cell (dc), window (dw) or strut (ds) diameters. These diameters are indicated in
Figure.5. Several attempts have been presented in the literature trying to replace the particle
size and possibly the most obvious trial is the use of the pore or cell diameter (d c)[9,39,40,41],
which is usually determined by examining enlarged photographs of cross-sections of foam
samples.
Several investigators[41] have indicated that the use of Erguns equation yields errors in the
prediction of permeability of ceramic foams as high as 50%, while the introduction of window
size obtained by image analysis into the same equations, seems to give more reasonable
results to assess the permeability of ceramic filters.
Figure.5 Definition of CFF: cell, window and strut diameters.[42]
Ergun defined the equivalent particle diameter of a non-spherical solid (dp) as the diameter
of the sphere having the same outer specific surface area per unit solid volume (Sv) of the
actual material in question (internal porosity, and small projections or cavities were
ignored)[38]:
dp
6
Sv
(7)
In Equation (7) the nomenclature of Ergun is maintained. There is some confusion in recent
literature, where Sv is used to represent the surface area of solid per unit bed volume (i.e. SB).
Equation (6) can be rewritten using Equation (7) as:
S 2 (1 )2
S (1 ) 2
P
4.17 v
Vs 0.292 v 3
Vs
3
L
dp
(8)
Richardson[24] explored the relationship between Sv and dw for porous ceramics and suggested
applying the hydraulic diameter (dh) concept. They equated the hydraulic diameter to the
measured window diameter:
dw dh 4
wetted area
wetted perimeter
(9)
Assuming all the pores have the same hydrodynamic diameter, a simple geometric analysis
yields:
8
Sv
4
d w (1 )
(10)
Substituting Equation (10) into Equation (8) yields:
Vs
Vs2
P
66.7 2 1.17 2
L
dw
dw
(11)
Recently Dietrich[43 ,44 ] proposed the following equation after correlating 2500 separate
experimental values from 20 authors:
V
V 2
P
110 2s 1.45 2 s
L
dh
dh
(12)
The Equation (11) and (12) are obviously similar. If one assumes that the hydraulic diameter
in Equation (12) is equal to the window diameter, as both empirical constants in Equation (12)
are larger than those in Equation (11), it will yield higher estimated pressure drops for any
velocity. Equation (12) has recently been independently shown to give excellent results using
the optically determined hydraulic diameter, i.e. the equivalent circular window diameter,
(dw).[45]
Other published correlations are all present in Table II
Table II Pervious published correlations.[46]
Reference
Pressure drop on foams correlation proposed
J.T. Richardsom et al.
Sv (1 ) 2
P Sv2 (1 )2
Vs
Vs
L
3
3
9.73 102 d w0.743 (1 )0.0982

J.F. Liu et al.
E.A. Moreira et al.
T.J. Lu et al.
P 22(1 )
0.22(1 ) 2
Vs
Vs
L
3 Dp2
3 Dp
Sv
; ;
;
;
12.979[1 0.971(1 )0.5 ]

d w (1 ; )0.5
3.68 102 d w0.7523 (1 )0.07158
D p 1.5d w
Dp=equivalent spherical diameter
P
(1 )2
(1 )
1.275 109 3 0.05 Vs 1.89 104 3 0.25 Vs2
L
dw
dw
0.008(d c / d s )
d (V /1 d s / dc )
0.15
P [0.044
]Re
Re s s
(d c / d s 1)0.431.13( ds / dc )
L. Tadrist et al.
P
(1 )2
(1 )
3 2 Vs 3
Vs2
L
ds
ds
M.D.M Innocentini et al.
P
(1 ) 2
(1 ) 2
150
Vs 1.75 3
V
L
3d p2
dp s
M. Lacroix et al.
P
(1 ) 2
(1 ) 2
150
Vs 1.75 3
V
L
3d p2
dp s
100 865,0.65 2.6
110
V 1.45 2 Vs2
L
dh s
dh
d p 1.5
d p 1.5d s
dh 4
1
d
c
Sv
Sv 2.87
1
(1 ) 0.25
ds dw
2.3 Fluid Mechanics

2.3.1 Viscosity
Viscosity is that property of a fluid by virtue of which it offers resistance to shear. [47]
Newtons law (Equation (13)) gives the relationship between shear stress and viscosity.
du
dy
(13)
Where is the shear stress [Pa], is viscosity [Pas], and du/dy is the velocity gradient [s-1].
For a given velocity gradient, the shear stress is directly proportional to the viscosity. In
everyday terms (and for fluids only), viscosity can be considered thickness or internal
friction. Thus, water is thin, having a lower viscosity, while honey is thick, having a
higher viscosity. Put simply, the less viscous the fluid is, the greater its ease of movement
(fluidity).[48]
The viscosity of a liquid decreases with temperature, but the viscosity of a gas increases with
temperature. This can be explained by examining the causes of viscosity. The fluids
resistance to shear depends on fluids rate of transfer of molecular momentum and its
cohesion. Liquid has cohesive forces much larger than a gas, since liquid molecules are much
more closely spaced than in a gas. Therefore, cohesion will be the predominant cause of
viscosity in a liquid, so viscosity will decreases with temperature like cohesion, and also be
independent of pressure (except at very high pressure).[49]
Viscosity coefficients can be defined in two ways: dynamic and kinematic viscosity. Dynamic
viscosity , is also called the absolute viscosity, and the typical unit is Pas. Kinematic
viscosity v, is the ratio of viscosity to mass density[47], and the typical unit is [m2/s]:
v
(14)
2.3.2 Reynolds Number

In fluid mechanics, the Reynolds number (Re) is a dimensionless number that gives a measure
of the ratio of inertial forces to viscous forces and consequently quantifies the relative
importance of these two types of forces for given flow conditions.[50] The concept was
10
introduced by George Gabriel Stokes in 1851[51], but the Reynolds number is named after
Osborne Reynolds, who popularized its use in 1883.[52]
Since the Reynolds number expresses the ratio of inertial forces to viscous forces, so:
Re
InertiaForce
ViscousForce
(15)
dV
dx
Re
2
dV
2
dx
Re
(16)
VL VL
(17)
Where Re is Reynolds number, V is the mean velocity of the fluid relative to an object like a
pipe, a particle or a pore [m/s], L is a characteristic linear dimension, (e.g. pipe, particle or
pore diameter) [m], is the dynamic viscosity of the fluid [Pas or Ns/m or kg/(ms)], v is
the kinematic viscosity [m/s] (see Equation (14)), and is the density of the fluid [kg/m].
For flow in a pipe or tube, the Reynolds number is generally defined as[53]:
Re
Vd h Vd h Qd h
v
vA
(18)
Where dh is the hydraulic diameter of the pipe [m], Q is the volumetric flow rate [m3/s], and A
is the pipe cross-sectional area [m].
Reynolds numbers are used frequently to characterize different flow regimes, such as laminar
or turbulent flow. Figure.6 shows the area of different flow regimes. At low Reynolds numbers,
Re<2300, it is laminar flow, viscous forces are dominant, and characters are smooth and
constant fluid motion. Turbulent flow occurs at high Reynolds numbers, e.g. Re>4000,
inertial forces are dominant, tend to produce chaotic eddies, vortices and other flow
instabilities.[47,54] When 2300 < Re < 4000, the flow is called "transitional". Transitional flow
is a mixture of laminar and turbulent flow, with turbulence in the centre of the pipe, and
laminar flow near the edges. Each of these flows behaves in different manners in terms of
their frictional energy loss while flowing, and have different equations that predict their
behaviour.[55] Figure.7 shows the diagrammatic sketch of laminar and turbulent flow.
11
Figure.6 Moody diagram.[56]
Figure.7Diagrammatic sketch of laminar flow and turbulent flow.[57]
2.3.3 Entrance Length

A fluid requires some length to develop the velocity profile after entering the pipe or after
passing through components such as: bends, valves, pumps, turbines or similar.[58] At a finite
distance from the entrance, the boundary layers merge and the in-viscid core disappears. The
tube flow is then entirely viscous, and the axial velocity adjusts slightly further until at x=le
(i.e. the length to fully develop the velocity profile), it no longer changes with x (typically 99%
approach is assumed), and the velocity profile is fully developed and constant, the wall shear
is constant, and the pressure drops linearly with x, for either laminar or turbulent flow.[59]
12
These details are shown in Figure.8.
Figure.8Diagrammatic sketch of developing velocity profiles and pressure changes in the entrance of
a duct flow. [53]
The entrance length can be expressed with the dimensionless entrance lengths number
expressed as[58]:
EL le / d
(19)
Where EL is entrance length number, le is the length to fully develop the velocity profile [m]
and d is the pipe diameter [m].
Dimensional analysis shows that the Reynolds number is the only parameter affecting
entrance length[59].
13
le f (d ,V , , )
(20)
le
vd
g(
)
d
(21)
Substitution of Equation (17) in Equation (21) finally gives,

le
g (Re)
d
(22)
The entrance length number correlation with the Reynolds number for laminar flow is[60]:
ELla minar 0.03 Repipe
(23)
For turbulent flow one commonly accepted correlation is[61]:

ELturbulent 1.3590 Re1/4
(24)
2.4 Tortuosity
The tortuosity is typically described as the ratio between the real length of the fluids path and
the actual geometrical length of the sample,[62] as shown in Figure.9.
14
Figure.9 Determination of tortuosity: fluid path and geometrical length.[63]

Figure.10 illustrates porous structures with different flow tortuosities: if both structures have
the same volumetric flow rate and porosity, then the residence time of fluid is also identical in
both cases. Since the residence time is same and the flow through the left structure passes a
significantly longer path, so the structure on the left side has a higher velocity compare to the
structure on the right side and all else being equal will exhibit a higher pressure gradient for
the same bulk flow.[63] Depending mainly on the pores geometry and not on the contribution
of diffusion mechanisms, tortuosity of a low porosity matrix tends to be high and vice versa.
The porous solids with high connectivity, should have low tortuosities and vice versa.[64]
15
Figure.10 Porous structures with different flow toruosities.[63]
Scheidegger[ 65 ] suggested that the tortuosity in sponge structures can be obtained by

measuring the electrical resistivity of the structure completely saturated with a conductive
solution.
Thus tortuosity can be determined by electromagnetic induction experiments using alloys of
known electrical conductivity (). The actual resistance of a filter element filled with metal,
as shown in Figure.11 can be compared to the known resistance of an equivalent path length of
metal either liquid or solid and the tortuosity determined. The conductivity is related to the
path length, are and resistance by[66]:
l
aRm
(25)
Where m is the conductivity of the metal [-1m-1] at the measurement temperature, l is the
length of the conducting path [m], a is the area of the conducting path [m2].
16
Figure.11 Representative SEM micrographs showing a full 50 mm thickness vertical profile through of
a well primed 50 ppi commercial alumina ceramic foam filter.
Inside of the filter, the available conducting area is reduced by the presence of non-conducting
obstructions (e.g. trapped gas or solid and filter media), and the conducting path length is
increased due to the tortuosity. Assuming that the filter media is the only significant
obstruction, the reduced apparent electrical conductivity can be estimated as follows[66]:
f
(26)
Where f is the apparent electrical conductivity of the metal impregnated filter, the unit is
[-1m-1].
Rearranging Equation (26):
m
f
(27)
17
An average temperature was used to estimate the liquid metal conductivity within the filter
elements, starting with literature conductivity data for ultra pure metal, and correcting for the
actual measured room temperature conductivity of the clean metal after experimentation:[67]
24.77 108
IACSm293 K
(1 0.000571 [Tm - 933.2]) 65
(28)
Where IACS means International Annealed Copper Standard, IACSm293K is the average room
temperature conductivity of the solidified metal used during the experiment [% IACS] and Tm
is the temperature of the liquid alloy under experimental conditions [K].
18
Chapter 3
Experimental Apparatus and Procedure
3.1 Optical Microscopy
3.1.1 Optical Microscopy - Theory
Initial investigations were carried out using optical microscopy, but the large magnification
and small sample size made it necessary to switch to using an optical scanner. The optical
microscope, also called the "light microscope", uses visible light and a set of lenses to enlarge
the images of small samples. Optical microscopy is based on the principle of the compound
microscope, which dates back to the 17th century. The basic optical microscope is very simple.
Modern designs can greatly improve its properties, such like resolution and sample contrast.
Optical microscopy uses visible light, and samples are directly observed visually by the user.
These advantages make it easy to use and popular.
Modern developments allow observation of a sample via a computer by equipping the
microscope with a digital camera. The microscope can also be controlled partly by computer.
The images captured by the digital camera allow greater analysis, such as measurements of
distances and areas. The Figure.12 shows modern digital microscope.
Figure.12 Modern digital microscope-LEICA MEF4M.
19
3.1.2 Optical Microscopy - Experimental Procedure

The ceramic foam filters are brittle and fragile, so the grains of foam filter are easily broken
off. Sample specimens obtained from the filters are therefore irregularly shaped, as well as
porous. So it is necessary to mount the samples before analysis.
Firstly the filters are sectioned by cutting several small pieces from the surface to produce the
samples. Samples are then mounted, using Epofix, which is a cold-resin based on two fluid
epoxy components, a resin (bisphenol-a-diglycidylether) and a hardener (triethylenetetramine).
Because of Epofixs low viscosity, it will penetrate into all the pores of the samples.
Afterwards it is necessary to sand the surface of the specimens using progressively finer SiC
sandpapers such as #s: 320, 400 and 600.
After preparation, the samples can be observed using the optical microscope.
3.2. Optical Scanning

The new procedure of diameter measurements used in this study: scanned filters by Epson
Perfection V330 scanner (as shown in Figure. 13), and then measured the diameters manually
with the help of Imagic ImageAcess easyLab 7 analysis software. The pictures resolution is
1200 dpi (dots per inch) for 30 ppi, 40 ppi and 50 ppi CFFs, since 80 ppi CFF has much
smaller cell, window and strut size, a higher resolution 2400 dpi was used. The pictures got by
scanner are shown in Figure.3.
Figure. 13 Epson Perfection V330 scanner.[68]
There is no sample preparation for this method, as the CFF can be placed directly onto the
scanner. So this new procedure was both easier and faster to use. Compared with optical
microscopy, more measurements could be taken in the same amount of time, and a higher
level of statistical significance achieved for the obtained averages.
20
3.3. Energy-Dispersive X-ray Spectroscopy

3.3.1 Energy-Dispersive X-ray Spectroscopy- Theory
Energy-dispersive X-ray spectroscopy is often abbreviated as EDS. It is an analytical
technique used for elemental analysis. The working principle is that every element has a
unique atomic structure allowing X-rays that are characteristic of an elements atomic
structure to be identified uniquely from one another. An EDS system is commonly found on
most Scanning Electron Microscopes (SEM). The operating principle of the SEM is that it
scans a sample with a high-energy beam of electrons in a raster scan pattern, and the electrons
will interact with the atoms that causing the sample to produce signals containing information
about surface topography, composition and electrical conductivity.[69]
The elemental analysis obtained from the EDS is not very accurate. There are many factors
affecting the accuracy of EDS. Light elements such as: H, He and Li can not be detected by
X-ray spectroscopy, because they do not have enough electrons to produce characteristic
X-rays. Many elements produce similar or identical signal peaks, for example Ti and V, Mn
and Fe, making them difficult to distinguish.
3.3.2 Energy-Dispersive X-ray Spectroscopy - Sample Preparation

The samples must be cut to the correct size to fit in the specimen chamber and must be
mounted on specimen stub. Samples also need to be electrically conductive, or at least
electrically conductive on the surface. Non-conductive specimens tend to charge when
scanned by the electron beam, especially in secondary electron imaging mode, which will
causes scanning faults.[69] The CFF samples are embedded in Epofix, such that they are
non-conductive. It is therefore necessary to coat them with an ultra-thin electrically
conducting material. There are a lot of conductive materials that can be deposited on the
samples by low-vacuum sputter coating, such as: gold, gold/palladium alloy, platinum,
osmium, iridium, tungsten, chromium or graphite. [25] The author used graphite as the coating
material. After the described procedure, the specimens can be used for EDS microanalysis.
3.4 Pressure Drop Experiments

The liquid permeability of 50 mm thick commercial foam filters with: 30, 40, 50 and 80 ppi,
were measured using water as the working fluid, in a temperature range from 5-8 oC, typical
density is 999.9 kg/m3, viscosity is 1.377510-3 PaS. Mass flow rates from about 0.05 to 2
kg/s were used. Eight to ten different velocities were used to measure pressure drop for each
filter.
21
Water is a very good analogue for liquid aluminium as it has a very similar dynamic viscosity
to both aluminium and aluminium alloy A356 at normal casting temperatures (e.g.
700-720oC). A356 is a common foundry alloy and was the base alloy for the PhD. thesis to
which this work relates. The common composition of A356 is 0.015 wt% Cu, 7.2 wt% Si,
0.38 wt% Mg, 0.120 wt% Fe, 0.032 wt% Ti and 0.0029 wt% B.[70] The dynamic viscosity of
these three materials are presented in Table III.
Table III Dynamic viscosity of water, pure aluminium and A356 alloy.[71,
o
Temperature ( C)
Dynamic Viscosity (PaS)
Density (kg/m3)
2
Kinematic Viscosity (m /s)
Water
Pure Aluminium
A356
7
1.38E-03
710
1.25E-03
710
1.03E-03
1000
2386
2340
1.38E-06
5.25E-07
4.41E-07
72, 73]
The experimental set-up shown in Figure.14 was designed by Mark W. Kennedy and the filter
housing was constructed by Egil Torsetnes. Figure.15 shows the close up of the filter housing.
The pressure transducer was an AEP DF-2, 0-1 Bar measuring range, 4-20 mA output. The
pressure was shown by means of electric current, and the transducer was factory calibrated
and certified to an error of 0.04% of reading, over the full scale from 0-1 Bar, using a 6 point
calibration, represented the greatest uncertainty in the experimental and defining the
experimental uncertainty limits. The transducer was powered using a MANSON, 0-30 V, 2.5
A, DC power supply. The electric current measurements were taken using a FLUKE 26 III,
True RMS Multimeter. During the experiments the current produced by the transducer at zero
liquid flow velocity was determined manually using a FLUKE 26 III, True RMS Multimeter
(Fluke, USA) to a precision of 0.001 mA (6.25 Pa), using the lowest available current scale.
Current during the flow measuring periods were computer data logged at 100 mS intervals by
conversion to a 0-5 V signal, with a resolution of 0.001 V or 0.004 mA (i.e. 25 Pa resolution).
At higher than 4 mA, no bias could be detected between the manual and automated current
readings, at the available 0.01 mA resolution (the FLUKE switched to a lower resolution at
greater than 4.099 mA).
The pressure was produced using a Jula 1000 W submersible pump and with a maximum
pressure of 0.8 Bar. Fluid velocity was controlled by the use of a ball valve. Flow rate was
computed by the ratio of weight gain in a 100 L tank vs. time, with the help of a computerized
data logger (DATASCAN 7220, Analogue Input Measurement Processor), which scanned the
scale weight at 100 ms intervals. The scale used had a 4-20 mA output over a range of 0-100
kg, with a resolution of 10 g. The data logging software used was DAS-16, version 1.0.
22
Figure.14 Photograph of experimental set-up constructed using DN 50 mm x 1.8 mm plastic pipe.
Figure.15 Close up of filter housing.
The water temperature was determined using 1 mm diameter Type K, Inconel sheathed
thermocouples and a FLUKE 51 II thermocouple reader. Thermocouple readings were also
data logged.
These experiments were a continuation of the project Liquid Permeability of Ceramic Foam
Filters, completed as part of course number TMT 5500. The initial procedure which was
used previously was to directly put the large diameter 101 mm CFFs into the filter housing
without any measures to avoid liquid bypassing at the outer periphery (gaskets were used on
23
top and bottom). The experiment procedures were sequentially improved during the current
set of experiments on two occasions. The modified procedure (Method 1), used silicon grease
evenly daubed onto the side surface of the CFFs. The final procedure (Method 2), used paper
(cellulose) placed onto the side surface of CFF, which had silicon grease on it. Figure.16
shows these different procedures. Swelling of the cellulose fibres on contact with water
provided a negligibly permeable seal along the whole length of the side of the filter media.
a)
b)
c)
d)
Figure.16
e)
a) 49 mm(diameter) CFF sample. b) 49 mm CFF, Method 1. c) 49 mm CFF, Method 2.

d) 49 mm CFF in the filter housing, using Method 2. e) 101 mm CFF in the filter housing,
using Method 2.
There are two different designs of filter housings, one is for the experiment of filters with
~101 mm diameter, and the details are shown in Figure.17. A smaller filter housing was
designed for testing the ~49mm diameter filters, as shown in Figure.18.
24
Drawn to scale
Differential Pressure Transducer
20 cm long,
copper impulse lines,
4 mm ID
102 mm dia.
2 mm thick
O-ring
Smooth
Plexiglas
Pipe
49.8 mm ID
60 mm OD
Pressure taps
4 mm dia. hole
30, 40, 50 or 80
PPI commercial
ceramic foam
filter ~101 mm
diameter
Plexiglas housing
~0.5 mm grease
impregnated cellulose
2 mm thick rubber gasket

top and bottom of filter
with 49 mm dia. hole
Figure.17 Filter housing for ~101 mm diameter CFFs.
Drawn to scale
Differential Pressure Transducer
102 mm dia.
2 mm thick
O-Ring
Smooth
Plexiglas
Pipe
49.8 mm ID
60 mm OD
20 cm long,
copper impulse lines,
4 mm ID
Pressure taps
4 mm dia. hole
50 mm dia. inlet
and 48 mm dia.
outlet to hold
filter in place.
30, 50 or 80
PPI commercial
ceramic foam
filter ~49 mm dia.
Rubber
O-Ring
Seal
Plexiglas
housing
~0.5 mm thick
silicone grease
impregnated
cellulose
Figure.18 Filter housing for ~49mm diameter CFFs.
25
3.5 Tortuosity Experiments

As previously mentioned, the tortuosity of the CFFs was obtained by measuring the electrical
resistivity when CFFs completely filled with a conductive medium. Aluminium was used in
the experimental program.
The CFF was contained for the experiment in a type of filter bowl built using two Bimex
400 fibre tubes, each 150 mm high and with an internal diameter about 100 mm (4). The CFF
was mounted in the top of the lower refractory tube with Fibrefrax cement. Fibrefrax
mouldable cement was also used to glue the two fibre tubes together. The dimensions of the
filters have been standardized to 50 mm thick and were sanded to 100 mm diameter (see
[22]
Figure.19 ).
Figure.19 30 ppi Sivex CFFs (left side) and Bimex 400 insulated crucibles (right side).
Three filters with same ppi and 50 mm thickness were stacked on top of each other in the tube.
The last mounting step was to place the 300 mm tube on a dense alumina fibre plate, as shown
in Figure.20. The mounted crucibles were than dried for 12 hours at 60oC in a ventilated oven.
26
Figure.20 Mounted crucible.
The effective electrical conductivities of the filters were also determined from the same liquid
metal electromagnetic induction experiments, using procedures described in detail
elsewhere.[74,75,76] A schematic of the apparatus used is shown in Figure 21 a) and a photograph
is shown in Figure 21 b). The power induced in a tight stack of three 50 mm thick, and ~100
mm diameter filters of 30, 40, 50 or 80 ppi was determined electrically at a known
temperature (and therefore metal electrical conductivity), while filled with commercially
pure aluminium alloys with initial electrical conductivity from 61-62% IACS. Temperatures
were logged every 100 ms by type K thermocouples located under and over the 150 mm stack
of filters.
Two Bimex 400 fiber

risers, ~102 mm ID,
300 mm high total.
Two cemented
together and to the
base plate with
Fibrefrax cement
Ceramic Foam Filter

30, 40, 50 or 80 PPI,
total of 150mm thick
~100 mm diameter,
cemented with Fibrefrax.
1mm
thick mica
insulation
Type K Inconel sheathed

thermocouples
COMSOL simulation
showing the magnetic
field distribution within
the filter and metal
Double layer coil constructed of
6 mm x 1 mm thick tubing,
double insulated with glass
fiber sleeves,
140 mm avg. dia., 126 mm ID,
112 mm height, 31 turns total.
Re-usable
casting sand
a)
Alumina ceramic
plate, 20 mm thick
b)
Figure.21 Schematic of the filter tortuosity apparatus (a) and photograph (b), showing a 2 layer, 31 turn
(total) induction coil, operated at 371-734 A, using line frequency 50 Hz AC power.
27
Chapter 4
Results and Discussion
4.1 CFF Porosities
There are two types of porosity in CFFs: effective porosity, also called open porosity, it refers
to the fraction of the total volume in which fluid flow is effectively taking place; ineffective
porosity, also called closed porosity, that refers to the fraction of the total volume in which
fluids or gases are present but in which fluid flow cannot effectively take place and includes
the closed pores.[77]
It should be noted that the total and open porosity are of very similar magnitude in this study,
as shown in Figure.11, areas of closed porosity can be seen in black, some of which are
highlighted by dotted circles as examples, visual inspection confirms nearly all macroscopic
porosity is filled with metal. Previous literature[78,79] have described the morphological
characterization of CFFs in detail, by using mercury at up to 4000 Bar and found that the
difference between the total and the open porosity is <5% of the measured valued, even when
metal is forced at very higher pressure into the microscope pores between individual ceramic
particles. The convention of Dietrich has therefore been followed in this work, i.e. the
equality between total and open porosity.
The total porosity was computed through Equation (29):
Porosity
(TheoreticalWeight ActualWeight )
TheoreticalWeight
(29)
The composition of the CFFs was measured by EDS, and it was found that it is mainly
alumina, as indicated in Figure.4. The maximum theoretical density of alumina, e.g. corundum
is 3.9-4.0 g/cm3 could be used to compute the weight of a given volume of CFF and hence a
total porosity estimated. A more precise value of 3.48 (average of 3 readings) 0.02 g/cm3
determined by helium pycnometery, for the exact type of ceramic used by Sivex in the
construction of their CFF filters, was actually used in the present study, based on private
communication with Norsk Hydro.
For the 30, 50 and 80 ppi filters, porosity measurements were taken for complete
commercially sized filter elements (23, 20 and 23 square). The porosity and other relevant
28
information are listed in Table IV.
Table IV Porosity and other relevant information

ppi
30 ppi
30 ppi
40 ppi
50 ppi
50 ppi
80 ppi
80 ppi
80 ppi
80 ppi
Filter Number
I
II
19
I
II
I
II
III
IV
Volume (cm3) A.Weight (kg) T.Weight (kg)

16163
6.21
56.57
16132
6.18
56.46
393
0.157
0.137
12023
5.84
42.08
12082
5.62
42.29
16134
7.86
56.47
16022
7.61
56.075
16121
7.96
56.42
16022
8.16
56.075
0.890
0.891
0.886
0.861
0.867
0.861
0.861
0.859
0.855
Figure.22 illustrates the relationship between ppi and porosities. It can be found that CFFs
with a higher nominal pore count have lower porosity, e.g. for 80 ppi vs. 30 ppi. A constant
ceramic thickness deposited on the original polymer substrate is the likely cause of the
relationship.
0.9
0.89
Porosity
0.88
0.87
0.86
0.85
20
30
40
50
60
70
ppi
Figure.22 Fractional CFFs porosity versus ppi.
29
80
90
4.2 Cell, Window & Strut Diameter

200 measurements of the cell and window, and 40 measurements of the strut diameters were
taken for each filter type. Averages as well as sample standard deviations have been
calculated for each filter type. 95% confidence intervals have been estimated for each mean
using the students-T distribution, based on the number of actual degrees of freedom, to adjust
for the different sizes of the sample populations. These values are listed in Table V-VII, and
the measured experimental data are plotted as histogram in Figure.22-24.
Table V Information of cell size, dc
Measurement Times
Mean Size (m)
Minimum Size (m)
Median Size (m)
MaxmiumSize (m)
Standard Deviation
95%Confidence
30 ppi
200
1668
720
1730
2722
417
57.8
40 ppi
200
1306
525
1345
1975
251
34.8
50 ppi
200
1131
549
1181
1667
229
31.8
80 ppi
200
683
392
689
953
122
16.9
Table VI Information of window size, dw

Measurement Times
Mean Size (m)
Minimum Size (m)
Median Size (m)
MaxmiumSize (m)
Standard Deviation
95%Confidence
30 ppi
200
962
443
962
1878
190
26.4
40 ppi
200
699
273
714
1050
151
20.9
50 ppi
200
623
317
597
1308
130
18.0
80 ppi
200
384
212
381
689
87
12.0
Table VII Information of strut size, ds

Measurement Times
Mean Size (m)
Minimum Size (m)
Median Size (m)
MaxmiumSize (m)
Standard Deviation
95%Confidence
30 ppi
40
185
85
191
254
41
5.7
40 ppi
40
211
127
200
295
46
6.4
50 ppi
40
190
127
190
274
36
5.0
80 ppi
40
119
74
117
170
20
2.8
Histograms have been made of the 200 counts of window diameter (dw) as can be seen in
Figure.23 a) through d) for the 30-80 ppi filter types. Median values are indicated by dotted
lines.
30
80
80
a)
60
60
50
50
40
30
20
20
10
10
0
500
700
900
1100
Window size, d w (m)
1300
1500
250
80
350
450
550
650
750
Window size, dw (m)
850
950
1050
80
c)
70
60
60
50
50
Counts
Counts
40
30
70
b)
70
Counts
Counts
70
40
40
30
30
20
20
10
10
d)
350
450
550
650
750
Window size, dw (m)
850
950
1050
225
275
325
375
425
475
525
575
625
Window size, dw (m)
Figure.23 Histograms of a) 30, b) 40, c) 50 and d) 80 ppi window sizes (dw) [m]. Median values are
marked with dotted lines.
1900
1700
Cell size (m)
1500
1300
1100
900
700
500
300
300
500
700
900
1100
Window size (m)

Figure.24 Commercial ceramic foam filters cell size (dc) versus window size (dw) [m].
Average cell diameter (dc) has been plotted versus the average window diameter (dw) in
Figure.24 and the results correlated according to the following equation:
31
675
dc 1.787d w , R2 0.987
(30)
The linear relationship between cell and window diameters implies a simple geometric
relationship, likely originating with the original substrate used during the filter fabrication
process. Agreement with literature values for both cell and window sizes for similar alumina
CFFs is excellent.[80]
4.3 Pressure Drop Results

28 pressure drop experiments were done to achieve the following:
1) Develop and validate an adequate experimental procedure and apparatus.
2) Compare the pressure drop of CFF samples from the same filter and same position, but
with different diameters, i.e. the 101 and 49 mm filter elements to verify the adequacy
of the experimental and modelling procedure.
3) Correlate pressure drop with liquid velocity.
4) Determine the permeability of different types of filters, i.e. obtain k1 and k2 for
Equation (2).
5) Determine the variability of permeability between different filters of the same ppi.
6) Measure the permeability variation for the same CFF with changes in sample position.
(In order to describe the samples position clearly, different numbers represent
different positions. Figure.25 shows the detailed information.)
Figure.25 CFF sample position code.
32
The details describing each experiment are shown in the Table VIII. The sections 2-21 were
cut from the whole filters (23, 20 and 23 square). Sections 22-24 were cut from the
centre of sections 18, 20 and 21.
Table VIII Strut size and relevant information
Test Section
Filter
Section
Section
Inlet
Section
ppi
Method
No.
No.
No. Diameter (mm) Thickness (mm)
Length (mm) Position
1
2
30
I
48.57
50.18
1
106
6
2
15
80 IV
48.97
49.77
1
106
6
3
14
80
II
48.82
50.44
1
106
6
4
2
30
I
48.57
50.18
2
106
6
5
15
80 IV
48.97
49.77
2
106
6
6
14
80
II
48.82
50.44
2
106
6
7
4
50
I
48.61
50.6
2
106
2
8
4
50
I
48.61
50.6
2
306
2
9
20
50 N/A
~101
49.58
2
106
N/A
10
20
50 N/A
~101
49.58
2
306
N/A
11
1
30
I
48.71
50.48
2
106
1
12
3
50
I
48.95
50.37
2
106
1
13
5
50
I
48.65
50.43
2
106
3
14
6
50
I
48.86
50.56
2
106
4
15
7
50
I
48.80
50.26
2
106
6
16
8
50
I
49.37
50.49
2
106
5
17
11
80
I
48.00
50.59
2
106
6
18
12
80 III
48.94
50.07
2
106
6
19
13
50
II
49.39
49.97
2
106
1
20
16
80 N/A
48.84
50.06
2
106
6
21
17
50 N/A
49.12
50.41
2
106
1
22
17
50 N/A
49.12
50.41
2
106
1
23
18
30 N/A
~101
50.67
2
106
N/A
24
19
40 N/A
~101
47.65
2
106
N/A
25
21
80 N/A
~101
50.31
2
106
N/A
26
22
30 N/A
48.67
50.67
2
106
N/A
27
23
50 N/A
49.15
49.58
2
106
N/A
28
24
80 N/A
49.08
50.31
2
106
N/A
The section diameter and section thickness in Table VIII were measured by micrometer and
the average of 4 and 6 readings, were used respectively. The test number 22 which is marked
in yellow used the same filter as test 21, but in the opposite flow direction. The details of the
experimental data are presented in Appendix Tables 1-28. The velocity given in the table is
the calculated superficial velocity, and the temperature, density and viscosity are the fluid (i.e.
water) properties. In order to determine the effective superficial velocity for the 101 mm
filters, it was necessary to use computational fluid dynamics (CFD) to solve for average
diameter of the flow field. An effective flow field diameter was initially guessed for use
with Equation (2), the resulting k1 and k2 terms were then used with the CFD model and the
33
pressure gradient determined. If the results were in error, a new effective diameter was
guessed and the procedure repeated until convergence was achieved.[66] The details of the
CFD model that was used, are introduced in Chapter 5.
4.3.1 Improvement of Procedures

As mentioned previously, this research continues work begun during project Liquid
Permeability of Ceramic Foam Filters for course TMT 5500. The procedures used to
determine the experimental pressure gradient have been modified for these experiments. The
first method (e.g. Method 1) used this time was to daub thick high viscosity silicon grease
onto the side surface of CFF, in order to avoid bypassing of the fluid flow and obtain more
exact, i.e. higher pressure drop data.
The CFF number 2-30 ppi was used for the first test. The data are plot in Figure.26. It can be
seen that the highest point on the plot deviates significantly from the curve fitting the rest of
the data. This may be because the high velocity gave a very high pressure, which acted to
push the grease out of the filter housing, creating fluid bypassing, and resulting in a reduced
pressure drop.
1400
1200
Pressure Drop (KPa/m)
1000
800
600
400
200
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Velocity (m/s)
Figure.26 Experiment data of 2-30 ppi by using Method 1.
Two additional tests were done to prove the hypothesis from the first test. Filter numbers
14-80 ppi and 15-80 ppi were used in these tests. Figure.27 shows the data for 14-80 ppi. The
last 5 points, which are marked in the red circle, obviously deviate from the curve. It can be
34
concluded that operation at high velocity and high pressure caused a loss of the silicone
sealant and a reduced reading of the pressure gradient.
1600
1400
1200
1000
800
600
400
200
0.1
0.2
14-80ppi
0.3
0.4
0.5
0.6
0.7
Velocity (m/s)
Verification of loss of seal
Figure.27 Experiment data of 14-80 ppi by using Method 1.
Visual observation through the Plexiglas housing showed that the sealant was forced out of
the channel at the wall, as shown in Figure.28 b). The verification of loss of seal in Figure.27
was obtained by repeating the measurements, and after the loss of seal, the obtained pressure
drop data become smaller than was initially measured. These data indicate that the sealing of
the sides of the filter is critically important to obtaining accurate data, and the use of a high
viscosity silicone alone is not enough to avoid bypassing.
35
a)
b)
Figure.28
a) Filter housing before high velocity experiment.

b) Filter housing after high velocity experiment.
The final procedure (e.g. Method 2) was suggested by Mark W. Kennedy. Water swollen
cellulose and silicone were used to provide a very low permeability seal, which was not
subject to physical removal at pressures of up to 0.8 Bar. This was verified both visually and
by repeated measurements over the whole pressure range with test filters. Comparison of
results for the filter using Method 1 and Method 2 is presented in Figure.29 and Figure. 30.
From these figures it can be seen that, high flow and tight filters created large driving force
for bypassing. Experimental errors will tend to report lower pressure gradients and greater
permeability than reality.
36
1200
1000
800
600
400
200
0.1
0.2
0.3
Method 2
0.4
0.5
Velocity (m/s)
0.6
0.7
0.8
0.9
Method 1
Figure.29 Comparison of results for the 2-30 ppi filter using Method 1 and Method 2.
1400
1200
1000
800
600
400
200
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Velocity (m/s)
Method 2
Method 1
Figure.30 Comparison of results for the 14-80 ppi filter using Method 1 and Method 2.
37
4.3.2 Required Inlet Length to Fully Develop the Flow Regime

As shown in Figure.9. the filter holders were used in a piping circuit containing ten 90o bends,
with 5 located before and 5 after the filter holder apparatus. There was a total of 8.3 m of
piping with a 1.06 m straight section before the filter housing, and the pipe inner diameter was
0.0464 m.
Using Equation (19), EL=23.
Assumping a water temperature of 7oC, a water density of =999.9 kg/m3, and a viscosity of
=1.377510-3 Pas, the required entrance length number can be estimated by using Equation
(17), (23) and (24),. Details are shown in Table IX.
Table IX Required inlet length to fully develop flow
as a function of pipe Reynolds number
Velocity
(m/s)
0,03
0,05
0,1
0,2
0,4
0,8
1
Re
1010
1680
3370
6740
13470
26940
33680
L/D inlet L/D Inlet Estimated

Laimanal Turbulent L/D Inlet
30
N/A
30
50
9
29,5
N/A
10
10
N/A
12
12
N/A
15
15
N/A
17
17
N/A
18
18
From Table IX, it can be seen that except at very low velocity, the inlet length should be
adequate to achieve fully developed flow (i.e. 99% approach). The impact of back-to-back
and outof phase 90o bends on the flow, will be to promote turbulence even at low Reynolds
numbers.
38
Figure.31 Entrance length number versus Reynolds number.
The flow regimes during the experiments include laminar flow, transitional flow and turbulent
flow. As shown in Figure.31, it is hard to determine the entrance length number at the
transition region, and as a result longer pipes were used in the experiment to verify the
computation. The total pipe length was then 12.3 m, with 3.06 m of straight section before the
filter housing. The entrance length number became, EL=66. The long configuration should be
adequate to achieve fully developed flow at all Reynolds numbers. Filters 4-50 ppi and 18-50
ppi were used to verify this hypothesis. Figure.32 presents the comparison of the obtained
pressure drop data of filter 18-50 ppi using the two different circuit configurations.
39
1200
1000
800
600
400
200
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Velocity (m/s)
1.06m inlet length
3.06m inlet length
Figure.32 Comparison of results for the 20-50 ppi filter using both the short and long inlet lengths.
The regressions of the experimental data are shown below:

1.06 m inlet length: y 5917160.66x2 83380x, R2 0.99995 (31)
3.06 m inlet length: y 5837058.41x2 90658x, R2 0.99956 (32)

From the Figure.32, it can be seen that two experiments obtained virtually identical results.
The second order terms of the Equations (31) and (32) are ~1.35% different. This may be
caused by the accuracy of the data logger, different tightness of the filter housing or residual
gas bubbles in the pipes or may in fact be the correct impact of inlet length. Since this
research has been conducted primarily in order to test the permeability at high velocity,
second order effects are much more than the first order. So the difference in the first order of
the two regressions is not very significant. During the long configuration test, it was very
difficult to empty the gas bubbles from the system, and it was therefore decided to use the
short configuration for the remainder of the experiments.
4.3.3 Permeability of Different Types of Filters.

The obtained results from pressure drop experiment for test 23, 24, 9 and 25 are shown in
Figure.33. They are all for 101 mm diameter filters with 30 ppi, 40 ppi, 50 ppi and 80 ppi. The
40
effective flow field diameters for computing the superficial velocities are 65.5 mm, 66 mm,
66.1 mm and 66.5 mm respectively. At higher velocity, the pressure drop increases faster for
CFFs with a higher nominal pore count, e.g. for 80 ppi vs. 30 ppi, this apparently causes the
effective flow field diameter to also increase, in order to minimize the total pressure drop.
1400
1200
1000
800
600
400
200
0.1
0.2
19-30ppi
0.3
0.4
0.5
Velocity (m/s)
20-40ppi
18-50ppi
0.6
0.7
21-80ppi
Figure.33 Experimental data for tests 9, 23, 24 and 25.

19-30 ppi: y 1903861.95x2 24661x, R2 0.99979
(33)
20-40 ppi: y 2959253.57 x2 45880 x, R2 0.99966
(34)
18-50 ppi: y 5917160.66x2 83380x, R2 0.99995
(35)
21-80 ppi: y 10036001.67 x2 248300 x, R2 0.99959
(36)
Using the current equipment in this study, the pressure resolution during flow was 62.5 Pa, or
a gradient uncertainty of 625 Pa/m depending on the samples thickness. 625 Pa/m
represents more than 5% uncertainty at flow rates of <0.03-0.08 m/s, depending on filter type
(i.e. ppi). It was necessary to determine both k1 and k2 from flow conditions, where both terms
were simultaneously significant [66], i.e. it was not possible to operate in a purely Darcy
regime with this pressure measuring apparatus and obtain precise results.
41
Three different procedures were explored[66] to derive the Forchheimer coefficients, e.g. k1
and k2 in Equation (2):
1) An automated second order regression, with a zero intercept, using Excel 2003.
2) Ergun et al.s procedure of dividing Equation (2) by the velocity and performing a linear
regression. [81]
3) An iterative procedure to first guess k1 and then correlate the remainder for k2 using an
exponential regression.
The experimental parameters obtained for tests 9, 10, 23, 24, 25, 26, 27 and 28 were
calculated by all of the above procedures, and the values summarized in Table X.
Table X Comparison of three mathematical procedures to determine k1 and k2
Test
Filter
Number Number
26
23
24
27
9
10
28
25
22-30 ppi
18-30 ppi
19-40 ppi
23-50 ppi
20-50 ppi
20-50 ppi
24-80 ppi
21-80 ppi
Excel 2003
Forchheimer
k1
(m)
2.141E-08
2.463E-07
-2.035E-08
1.836E-08
1.631E-08
4.670E-08
1.897E-08
9.694E-09
Excel 2003 Ergun Method Ergun Method Method 3

Method 3
Forchheimer Forchheimer Forchheimer Forchheimer Forchheimer
k2
k1
k2
k1
k2
(m2)
5.827E-04
5.097E-04
3.047E-04
1.645E-04
1.693E-04
1.638E-04
1.077E-04
9.449E-05
(m)
4.582E-08
7.187E-08
1.090E-07
1.628E-08
1.244E-08
2.166E-08
8.676E-09
6.583E-09
(m2)
5.481E-04
5.177E-04
3.240E-04
1.653E-04
1.718E-04
1.667E-04
1.111E-04
9.680E-05
(m)
5.084E-08
5.572E-08
3.099E-08
1.568E-08
1.710E-08
1.525E-08
6.519E-09
5.441E-09
(m2)
5.459E-04
5.252E-04
3.379E-04
1.656E-04
1.690E-04
1.713E-04
1.148E-04
9.963E-05
It should be noted that it is clear that the negative coefficient k1 for the 19-40 ppi filter
obtained from Excel 2003 is physically meaningless. No physically meaningful correlations
could be performed on the values of k1 and k2 found using Excel, e.g. comparison with total
porosity or window size. Better results were obtained following the recommended method of
Ergun. The most physically meaningful results were obtained following the 3rd procedure,
where k1 was initially guessed, the first order component of Equation (2) subtracted from the
total and an exponential regression performed on the remainder. When the exponent on the
velocity became 2.00000, the procedure was deemed converged.
The balance of the experiment parameters were all computed using the 3rd procedure, and the
values of k1 and k2 are presented in Table XI.
42
Table XI Forchheimer empirical coefficients for Equation (2)

Test
Filter
Number Number
4
5
6
7
8
9
10
11
12
13
14
15
16
2-30 ppi
15-8 0ppi
14-80 ppi
4-50 ppi
4-50 ppi
20-50 ppi
20-50 ppi
1-30 ppi
3-50 ppi
5-50 ppi
6-50 ppi
7-50 ppi
8-50 ppi
Method 3
Method 3
Forchheimer Forchheimer Test
Filter
k1
k2
Number Number
(m)
3.045E-08
6.166E-09
6.762E-09
1.551E-08
2.326E-08
1.710E-08
1.525E-08
3.656E-08
1.296E-08
1.477E-08
1.381E-08
1.814E-08
2.229E-08
(m2)
4.404E-04
8.773E-05
1.170E-04
1.967E-04
2.007E-04
1.690E-04
1.713E-04
5.230E-04
1.751E-04
2.327E-04
1.790E-04
2.058E-04
2.142E-04
17
18
19
20
21
22
23
24
25
26
27
28
11-80 ppi
12-80 ppi
13-50 ppi
16-80 ppi
17-50 ppi
17-50 ppi
18-30 ppi
19-40 ppi
21-80 ppi
22-30 ppi
23-50 ppi
24-80 ppi
Method 3
Method 3
Forchheimer Forchheimer
k1
k2
(m)
5.250E-09
7.786E-09
1.679E-08
6.543E-09
2.537E-08
1.629E-08
5.572E-08
3.099E-08
5.441E-09
5.084E-08
1.568E-08
6.519E-09
(m2)
8.819E-05
1.143E-04
2.101E-04
1.442E-04
2.120E-04
2.155E-04
5.252E-04
3.379E-04
9.963E-05
5.459E-04
1.656E-04
1.148E-04
The average value of k1 and k2 for different types of filters are shown in Table XII.
Table XII Average value of k1 and k2 for different types of filters
Filter Type Forchheimer k1(m) Forchheimer k2 (m2)
30 ppi
4.339E-08
5.086E-04
40 ppi
3.099E-08
3.379E-04
50 ppi
1.748E-08
1.960E-04
80 ppi
6.352E-09
1.094E-04
The k1 and k2 values from Table XII have been plotted in Figure.34 and Figure.35 as functions
of the window area, dw2/4.
43
5.0E-08
Darcy Term, k 1
4.0E-08
3.0E-08
2.0E-08
1.0E-08
0.0E+00
200000
400000
600000
800000
Window area, m2
Figure.34 Correlation of Darcy term, k1 [m2] with window area.

6.0E-04
Darcy Term, k 2
5.0E-04
4.0E-04
3.0E-04
2.0E-04
1.0E-04
0.0E+00
0
200000
400000
600000
800000
Window area, m2
Figure.35 Correlation of Non-Darcy term, k2 [m] with window area.
Empirical correlations have been developed for k1 and k2 as functions of the window
diameter:
6.34 1014 d w2
(37)
k1
, R 2 0.928
4
7.311010 d w2
(38)
k2
, R 2 0.943
4
From Figure.34, and Figure.35, Equation (37) and (38), it is concluded that the Sivex CFFs
tested in these experiments behave like a series of orifices. An examination of Figure.36
would seem to support the concept of orifices, given the high percentage of closed windows,
particularly at higher ppis.
44
Figure.36 Representative SEM micrographs of 30 (a), 40 (b), 50 (c) and 80 (d) ppi Sivex CFFs. Cell or
pore sizes (dc) are indicated by solid circles and window sizes (dw) are indicated by dotted circles.
4.3.4 Correlation of Pressure Drop

Extensive work has been done to correlate the obtained results. All the equations listed in
Table II were compared, and only Dietrichs equation (Equation 12) was found to adequately
describe the data. Agreement is considered adequate being typically within 50%, except at
very low velocity.
Other equations result in much lower estimates of pressure drop than the experimental data
obtained using the final and well-sealed experimental procedure. Most previously published
literature did not describe the details of the experiment procedures used, e.g. did they use a
proper wall seal or correct for an expanding flow field? Since sealing arrangements and
analysis techniques were generally not explained in sufficient detail, firm conclusions can not
be drawn from comparisons with most of the previous studies.
The best empirical correlation obtained was a slightly modified version of Erguns equation.
P
(1 )2 Vs
(1 ) Vs2
8.385[150
1.75
], R 2 0.95
3
2
3
L
dw
dw
(39)
Equation (39) is plotted in Figure.37, and comparison is made with 181 measurements. The
45
average absolute error is 10.7 %.
Measured Pressure Drop Gradient, Pa/m
10000000
1000000
100000
10000
1000
1000
10000
100000
1000000
10000000
P
(1 )2 Vs
(1 ) Vs2
8.385[150
1.75 3
]
3
2
L
dw
dw
Figure.37 Overall correlation Equation (39), comparison to the 181 measured pressure gradients for test
4-8, 11-22 and 26-28, the absolute average error is 10.7%. Dotted lines in the figure indicate a range of
30%.
The Equation (39) is equivalent to using 23.4 and 2 (based on an average CFFs porosity of
88%), as the empirical constants, instead of the values 110 and 1.45 in Dietrichs Equation (12)
or the Ergun equivalent values of 66.7 and 1.17 from Equation (11):
V
V 2
P
23.4 2s 2 3 s
L
dh
dh
(40)
Applying Equation (40) to the obtained data indicates a significant reduction in error
compared with the original Equation (11), especially at low velocity. The overall error is
reduced from ~40% to ~30%.
Attempts were made to use cell diameter, strut diameter and tortuosity in various correlations;
however, no improvement could be made over the accuracy of Equation (39) or (40).
46
4.3.5 Permeability of Different Filters with the Same ppi.

Experiments 5, 6, 17, 18 and 20 all used 80 ppi filters-15-80 ppi, 14-80 ppi, 11-80 ppi, 12-80
ppi and 16-80 ppi. The filter sections were cut from the middle of 5 different filters, at
position No.6, as shown in Figure.25. The 3-50 ppi, 13-50 ppi and 17-50 ppi filter sections
were all cut from the corners of 3 different filters, at position No.1 as shown in Figure.25, and
these filters were used in experiments 12, 19 and 21. The pressure drop values obtained from
these experiments are plotted in Figure.38, and Figure.39.
1600
1400
1200
1000
800
600
400
200
0.05
0.1
11-80ppi
0.15
12-80ppi
0.2
0.25
0.3
Velocity (m/s)
14-80ppi
15-80ppi
0.35
0.4
0.45
16-80ppi
Figure.38 Pressure drop data from experiments 5, 6, 17, 18 and 20, filters were all cut from the middle,
at position number: 6.
47
1400
1200
1000
800
600
400
200
0.1
0.2
3-50ppi
0.3
0.4
Velocity (m/s)
13-50ppi
0.5
0.6
17-50ppi
Figure.39 Pressure drop data from experiments 12, 19, and 21, filters were all cut from the middle, at
position number 6.

11-80 ppi: y 11331160.53x2 219330 x, R2 0.99984
(41)
12-80 ppi: y 8750834.44 x2 176662 x, R2 0.99991
(42)
14-80 ppi: y 8549569.55x2 207140x, R2 0.99989
(43)
15-80 ppi: y 11397513.56 x2 218830 x, R2 0.99991
(44)
16-80 ppi: y 6931935.47 x2 203475 x, R2 0.99931
(45)
3-50 ppi: y 5709968.14 x2 107650 x, R2 0.99970
(46)
13-50 ppi: y 4758591.52x2 82210x, R2 0.99991
(47)
17-50 ppi: y 4716067.55x2 48920 x, R2 0.99994
(48)
It can be found that the regressions of the same ppi filters are not similar. 11-80 ppi, 15-80 ppi
and 3-50 ppi filters gave higher pressure drop than other filters with same ppi during the test,
and the pressure drop value of 16-80 ppi filter is much smaller than the rest 80PPI filters.
48
The k1 and k2 values in Equation (2) of these filters are shown in Table XIII.
Table XIII k1 and k2 values
Test Number Filter Number
17
18
6
5
20
12
19
21
11-80 ppi
12-80 ppi
14-80 ppi
15-80 ppi
16-80 ppi
3-50 ppi
13-50 ppi
17-50 ppi
k1 (m)
5.250E-09
7.786E-09
6.762E-09
6.166E-09
6.543E-09
1.296E-08
1.679E-08
2.537E-08
k2 (m2)
8.819E-05
1.143E-04
1.170E-04
8.773E-05
1.442E-04
1.751E-04
2.101E-04
2.120E-04
It is clear that the 16-80 ppi filter has a significantly higher k2 value than other 80 ppi filters.
The 16-80 ppi filter had the lowest measured pressure drop for this type of filter, since the
pressure drop is inversely proportional to the value of k1 and k2, and k2 is dominant at high
velocity, i.e. the velocities of primary interest in this study. Filters 11-80 ppi, 15-80 ppi and
3-50 ppi have lower k2 values and gave higher pressure drops. It should be noted that the
difference between 17-80 ppi and 18-80 ppi, 6-80 ppi and 16-80 ppi or 3-50 ppi and 13-50 ppi
are much larger than any error that can be attributed to the experiment procedure and are
therefore deemed to be caused by variation between the filters themselves.
Results between filters of the same ppi show some variability; however, results from each
type of filter are statistically different that other types of filters. This is shown in Table XIV
by the comparison of the mean and upper and lower 95% confidence intervals for the 50 and
80 ppi filter types.
Table XIV Average k1 and k2 values and 95% confidence area
Filter Type
Average k1 (m)
95% Confidence
Upper 95% Confidence
Lower 95% Confidence
Average k2 (m2)
95% Confidence
Upper 95% Confidence
Lower 95% Confidence
50 ppi
1.838E-08
5.570E-09
2.395E-08
1.281E-08
1.991E-04
1.820E-05
2.173E-04
1.809E-04
80 ppi
6.501E-09
8.078E-10
7.309E-09
5.694E-09
1.103E-04
2.060E-05
1.309E-04
8.967E-05
In conclusion, the permeability of filters with the same ppi are not identical. It is assumed that
natural variation during the manufacturing process is responsible, and has resulted in
variations in microstructure and tortuosity (e.g. greater or lesser numbers of blocked
windows). Visual observation also finds clear differences between the filters, as shown in
Figure.40.
49
a)
b)
Figure.40 Photo of 11-80 ppi and 12-80 ppi filters
In Figure.40, it is clear that two filters look different, 11-80 ppi filter has two layers which are
marked in the red circles. Figure.41 shows the microstructure of the layers.
Figure.41 Representative SEM micrographs of 11-80 ppi CFF indicating nearly complete blockage of all
windows.
From Figure.41, it can be found that in the lines from Figure.36 that most of the pores are
closed, and this makes it difficult for liquid to flow through this region, resulting in a very low
permeability and a very high pressure drop.
50
From above information, it is concluded that filters of the same ppi vary and that the
permeability for each type of the filter must be defined in a statistical manner, i.e. defined by
mean, standard deviation and confidence intervals.
4.3.6 Permeability for Different Samples taken from the Same Filter.
8 experiments were conducted to determine the relationship between permeability and sample
position. These 8 experiments numbers are 4, 7 and 11 through 16. The filters used are 1-30
ppi, 2-30 ppi, 3-50 ppi, 4-50 ppi, 5-50 ppi, 6-50 ppi, 7-50 ppi and 8-50 ppi. The position
information of these filters is presented in Table VII. The experiment data are plotted in
Figure.42 and Figure.43.
1200
1000
800
600
400
200
0
0
0,1
0,2
0,3
1-30ppi
0,4
Velocity (m/s)
0,5
0,6
0,7
0,8
2-30ppi
Figure.42 Pressure drop data for experiments 4 and 11, 1-30 ppi: position number 1, 2-30 ppi: position
number 2.
51
1400
1200
1000
800
600
400
200
0
0
0,1
0,2
3-50ppi
4-50ppi
0,3
Velocity (m/s)
5-50ppi
6-50ppi
0,4
7-50ppi
0,5
0,6
8-50ppi
Figure.43 Pressure drop data for experiments 7 and 12-16, 3-50 ppi: position number 1, 4-50 ppi:
position number 2, 5-50 ppi: position number 3, 6-50 ppi: position number 4, 7-50 ppi: position number6,
8-50 ppi: position number 5.

1-30 ppi: y 1911934.44x2 37360x, R2 0.99984
(49)
2-30 ppi: y 2270285.55x2 45114 x, R2 0.99985
(50)
3-50 ppi: y 5709968.14 x2 107650 x, R2 0.99970
(51)
4-50 ppi: y 5084123.17 x2 89000x, R2 0.99973
(52)
5-50 ppi: y 4297350.89 x2 92060 x, R2 0.99988
(53)
6-50 ppi: y 5586066.98x2 100785 x, R2 0.99993
(54)
7-50 ppi: y 4858509.54 x2 78150 x, R2 0.99990
(55)
8-50 ppi: y 4667185.24 x2 61325 x, R2 0.99992
(56)
It can be found that the regressions of the same ppi filters are not similar. 2-30 ppi which cut
from the centre of the filter gave a higher pressure drop than 1-30 ppi cut from the corner. For
50 ppi tests, 3-50 ppi cut from the corner of the filter has the highest pressure drop.
52
The k1 and k2 values in Equation (2) of these filters are shown in Table XV.
Table XV k1 and k2 values
11
1-30 ppi
4
2-30 ppi
12
3-50 ppi
7
4-50 ppi
13
5-50 ppi
14
6-50 ppi
15
7-50 ppi
16
8-50 ppi
k1 (m)
3.656E-08
3.045E-08
1.296E-08
1.551E-08
1.477E-08
1.381E-08
1.814E-08
2.229E-08
k2 (m2)
5.230E-04
4.404E-04
1.751E-04
1.967E-04
2.327E-04
1.790E-04
2.058E-04
2.142E-04
From above information it is concluded that the permeability of the filters vary both within a
given filter and even more between filters for a given ppi. There appears to be no correlation
between position number and permeability, i.e. the variations appear to be random in nature.
4.3.7 Permeability Changing with Fluid Flow Direction

Acosta[9] tested his filters in both the forward and reverse directions finding identical results.
In this study, two experiments were conducted to verify that flow direction would not alter the
pressure drop for well sealed filters. These two tests are experiments 21 and 22, and the filter
section used was 17-50 ppi. The experimental data is shown in Figure.44.
1400,00
1200,00
1000,00
800,00
600,00
400,00
200,00
0,00
0,00000
0,10000
0,20000
0,30000
0,40000
0,50000
Velocity (m/s)
17-50ppi forward direction
17-50ppi reverse direction
Figure.44 Pressure drop data for experiments 21 and 22, position number: 1.
53
0,60000
The regressions of the experimental data are:

17-50 ppi forward direction: y 4716067.55x2 48920 x, R2 0.99994 (57)
17-50 ppi reverse direction: y 4639062.59 x2 76410 x, R2 0.99982 (58)
From Figure.44, it can be seen that the experimental data of two tests are nearly identical. The
Equation (57) is also very similar to Equation (58), especially the second order term. The
difference of the second order term is about 1.6%, this different maybe caused by the
accuracy of the experiment. Since the error of the pressure drop obtained by the apparatus in
this study is 31.25 Pa, thus it will give more errors at very low velocity. This can explain
why there is a higher difference in the first order term. Table XVI shows the k1 and k2 values
that calculated by these two tests.
Table XVI k1 and k2 values
k1 (m)
k2 (m2)
21
17-50 ppi
2.537E-08 2.120E-04
22
17-50 ppi
1.629E-08 2.155E-04
Again, very similar k2 values were obtained from these two tests. Larger differences were
seen in k1 values, which maybe caused by the precision error of the experimental pressure
measuring equipment.
It can be concluded that the permeability of CFF will not change by the changing of fluid flow
direction.
4.4 Filter Tortuosity

Filter tortuosity has been determined for metal impregnated 30, 40, 50 and 80 ppi filters.
Measurements have been obtained with the metal in both liquid and solid states. The metal
used was commercially pure electrical grade aluminium, which was determined to have
61.70.1% IACS conductivity (average of 10 readings) prior to melting. After melting, the
solidified metal samples were found to have conductivities of 59.7% (30 ppi), 54.0% (40 ppi),
60.0% (50 ppi) and 61.0% (80 ppi). The hot metal experiments were conducted using the
apparatus already shown in Figure. 21.
Hot liquid metal and cold solid filter section measurements (average, vertical and horizontal
cuts) and calculated tortuosity results (using the hot metal data) are summarized in Table
XVII[66], for all four filter types. Experimental data have also been plotted in Figure.45.
54
Table XVII Summary of key tortuosity experimental results

Filter Filter
Filter
FEM Estimate of Cold Average
Cold Vertical
Cold Horizontal
Type Porosity Tortuosity Conductivity Ratio Conductivity Ratio Conductivity Ratio Conductivity Ratio
(opi)
(m/f)
(m/f)
(m/f)
(m/f)
30
0.892
1.30
1.70
1.64
1.77
1.400.06
40
0.900
2.29
2.15
2.27
2.02
2.540.18
50
0.863
2.19
2.70
2.34
3.05
2.540.12
80
0.865
3.20
2.79
2.91
2.66
3.700.06
4.00
80 PPI
3.50
m / f
3.00
40 PPI
2.50
80 PPI cold are consistent, but
low compared to hot values
2.00
50 PPI
30 PPI
1.50
1.00
300
400
500
600
700
800
900
1000
Window size, d w (m)

Liquid
Solid Avg.
Solid V
Solid H
Figure.45 Conductivity ratio of metal (m) and metal impregnated filter (f) versus window size (dw)
[m]. Comparison is shown between liquid metal values fitted to experimental data using FEM and
average, vertical, and horizontally cut solid filter sections (cold readings).[66]
Results show reasonable agreement between hot and average cold conductivity ratios, given
the different equipment, temperatures and frequencies involved in these two sets of
measurements. The conductivity ratio between the metal and metal impregnated filters for the
liquid metal experiments was correlated according to the following equation:
m
5.10 3.8 103 d w , R 2 0.981
f
(59)
No correlation could be found between the horizontal or vertical cuts. The observed variations
may the result of the random location of the sections through the pore structure of the filter
elements and the low electromagnetic penetration depth of the high frequency cold method,
which is less than one cell diameter.
55
Few measured values have been published previously for tortuosity. Moreia et al. measured
tortuosity values using an ionic conduction method equivalent to the induction method
described here. Their values for CFFs of 8, 20 and 45 ppi were 1.68, 1.71 and 1.84
respectively.[ 82 ] Diedericks et al. have theoretically studied tortuosity in some detail,
proposing a value of ~1.45 at =0.88, for foam like materials.[83] In P. Habisreuther et al.,
tortuosity has been numerically estimated for high porosity (=0.720) solid sponges material
to be 1.317.[63] Methods using water and ionic solutes, will likely underestimate the true filter
tortuosity, due to penetration of the water and ions into the micro and nano-porosity of the
filter structure itself. Liquid metal poorly wets the surface of the ceramic and in the absence
of intense pressure (e.g. 4000 Bar) will not penetrate the micro porosity.
56
Chapter 5
CFD Modelling
5.1 Introduction
Computational fluid dynamics (CFD) was used in this study to verify the experimental data
for the 49 mm filter design and to compute effective flow field diameter for 101 mm design.
To compute the effective flow field diameter for 101 mm design, an effective flow field
diameter was initially guessed for use with Equation (2), the resulting k1 and k2 terms were
then used with the CFD model and the pressure gradient determined. If the results were in
error, a new effective diameter was guessed and the procedure repeated until convergence
was achieved. Figure.46 shows this procedure.
Measure: P, Diameter, L, Mass Flow,

Temperature (i.e. and )
Guess an
effective
diameter
Calculate
k1 and k2
Guess a new
effective diameter
CFD
flow field,
P/L vs. V
No
Same as
measured?
Yes
57
Effective diameter
is correct
Figure.46 FEM CFD procedure applied to the 101 mm experimental results to determine the
Forchheimer, Equation (2), parameters k1 and k2.
5.2 Theory
5.2.1 Materials
The COMSOL material data water, liquid was used for the fluid properties in these
simulations. A new material called filter was built in manually for the filter domain,
containing the actual Darcy permeability and filter porosity.
The equation used to calculate the water dynamic viscosity in the COMSOL is[84]:
1.3799566804 0.021224019151T 1.3604562827 104 T 2 4.6454090319 107 T 3
(60)
8.9042735735 1010 T 4 9.0790692686 1013 T 5 3.8457331488 1016 T 6
The equation for water density is:
838.466135 1.40050603T 1 0.0030112376T 2 3.71822313 107 T 3
(61)
Temperature for use with Equations (60) and (61) were the actual average temperature
measured during each experiment.
The density and dynamic viscosity for the filter was the actual average water density and
dynamic viscosity from each experiment, entered manually. The porosities of the filters are
present in Table VIII.
5.2.2 Physics
This section has been extracted in large part from the documentation supplied by COMSOL
and the associated references.
There are two physics used in the modelling, Free and Porous Media Flow (fp) for the filter
and Turbulent Flow, k- (spf) for the pipe before and after the filter, covering the length of
the filter holder apparatus.
Free and Porous Media Flow
58
The free and porous media flow interface uses the Navier-Stokes equations for describing
the flow in open regions, and the Brinkman equations are used for the flow in porous
regions.[84]
The Navier-Stokes equation for incompressible flow is[85]:
V
V V ) P 2V f
t
(62)
Where V is the superficial flow velocity [m/s], P is the pressure [Pa], 2 is the vector
Laplacian and f represents other body forces (per unit volume) acting on the fluid, such
as gravity or centrifugal force.
The Brinkman equations are:
( ) ( V ) Qbr
t
(63)
V
V
2
1
[
(V ) ] P [ (V (V )T ( V ) I ]
t
(64)
k1
Qbr )V F
Vi
(65)
Where Qbr is a mass source or mass sink [kg/m3s], which is zero in case of these
experiments and F is the force term [kg/m2s2], this term is negligible (but non-zero) in
this study, due to the experimental equipment being positioned horizontal. Vi is the
interstitial velocity within the pores of the filter [m/s].
The Forchheimer drag option was added manually to the model and this option adds a
viscous force proportional to the square of fluid velocity, see Equation (2). So Equation
(64) can be written as:
V
V
2
1
[
(V ) ] P [ (V (V )T ( V ) I ]
t
(66)

k1
k2
V )V
59
Since the fluid is incompressible, the density is constant, so:
( ) V 0
t
(67)
Combine Equation (66) with Equation (67) get:
V
2
1
[(V ) ] P [ (V (V )T ( V ) I ]
k1
k2
(68)
V )V
Turbulent Flow, k- (spf)

As mentioned before, the fluid flow regimes experienced during the experiments included
laminar flow, transitional flow and complete turbulent flow. The maximum velocity
experienced in the filter during the experiment was ~0.8 m/s, and the Reynolds number,
Re is ~26000. Since the Re is not large, the Low Reynolds Number k- Turbulence
Model was chose for this modelling.
The flow fluid is incompressible and Newtonian, so Equation (62) can be written as:
V
V V [ P (V (V )T ] F
t
V 0
(69)
(70)
If the flow is turbulent, then all quantities fluctuate in time and space.[84] It is seldom
worth the extreme computational resources required to obtain details about the turbulent
fluctuations in time. A time averaged representation often provides sufficient information
about the flow.
The Reynolds-averaged representation of turbulent flows divides the flow quantities into
an averaged value and a fluctuation part,
(71)

Where represent any scalar quantity of the flow. The mean value can vary in space
and time generally. An example is shown for the fluid velocity in Figure.47[84], shows time
averaging of one component of the velocity vector for non-stationary turbulence. The
unfiltered flow has a time scale t1. After a time filter with width t2>> t1 has been
applied, there is a fluctuation part, ui, and an average part, Ui. Because the flow field also
60
varies on a time scale longer than t2, Ui is still time dependent but is much smoother
than the unfiltered velocity ui.
Figure.47 The unfiltered velocity component ui, with a time scale t1, and the averaged velocity
component, Ui, with time scale t2.
With Reynolds-averaging, Equation (69) can be re-written as:
V
V V ( u ' u ') [ P (V (V )T ] F (72)
t
Where V is the averaged velocity field and is the outer vector product. Compared with
Equation (69), only the last term on the left side is different. This term represents the
interaction between the fluctuating velocities, called the Reynolds stress tensor.
Assuming the turbulence to be of a purely diffusive nature, the deviating part of the
Reynolds stress is expressed by:
( u ' u ')
tracc(u ' u ') T (V (V )T )
(73)
Where T is the eddy viscosity, which is also called turbulent viscosity. The spherical part
can then be written as:
tracc(u ' u ')
2
k
3
(74)
where k is the turbulent kinetic energy. When simulating incompressible flows, this term
is included in the pressure, but in compressible flows, this term must be explicitly
included.
61
The Low Reynolds Number k- Turbulence Model introduces two dependent variables:
the turbulent kinetic energy, k, and the dissipation rate of turbulent energy, .[86]
The transport equation for k is[86]:
k
V k [( T )k ] Pk
t
k
(75)
The transport equation for is[87]:
2
V [( T ) ] C 1 Pk f C 2
t
k
k
k
(76)
Where:
2
2
Pk T V :[V (V )T ] ( V ) 2 k V
3
3
2
k
T f C
f 1 el
f 1 e l
/14
1 R5
2
e ( Rt /200)
3/4
t
/3.1
1 0.3e
2
( Rt /6.5) 2
l * ( V l ) /
Rt k 2 / ( )
V ( / )1/4
C1=1.5, C2=1.9, C=0.09, k=1.4, =1.4 and lw is the distance to the closest wall.
The damping terms in the equations for k and allows a no slip condition to be applied to
the velocity, so V=0. Since all velocity must disappear on the wall, so k=0 on the wall.
The boundary condition for is:
5.2.3
k
lw2
(77)
Modelling Approach
62
Some pertinent points to achieving adequate agreement between CFD, analytical models and
measured values were:
Iteration between high quality measurements and FEM to ensure validity of
assumptions and accuracy of final models.
Correct and validated boundary conditions, e.g.: no-slip walls, contiguous
velocity fields between liquid and porous media domains, and the inlet velocity
profile.
Use of the low Reynolds number k-, Reynolds Averaged Navier Stokes (RANS)
model for turbulence (k0 = 0.005 m2/s2 and 0 = 0.005 m2/s3), to adequately cover
the difficult range of velocities in the inlet region.
Use of dense meshes in regions of high velocity gradients (e.g. boundary mesh at
the no-slip walls).
Precise measurement and exact geometric reproduction of the actual apparatus.
5.3 FEM Model Details

5.3.1 49 mm Filter Finite Element Modelling
Figure.48 shows the geometry of the model for the 49 mm diameter filter holder apparatus.
Figure.48 Modelling geometry of 49 mm filter holder apparatus.
The mesh used in the FEM model is shown in Figure.49.
63
Figure.49 Modelling mesh used for the 49 mm filter apparatus.

Figure.50 and Figure.51 show the flow fields and the pressure gradients for the 49 mm diameter
filter apparatus at 7 oC and 0.054 m/s inlet water velocity. The filter properties used in this
simulation are those of 23-50 ppis as shown in Table VIII.
Please note in Figure.50 and Figure.51 that the flow field is contiguous over the boundary
between the turbulent fluid and porous media domains. This required some careful
manipulation of the FEM model in COMSOL, to force iteration between the inlet and outlet
conditions over these boundaries.
64
Figure.50 The flow field for the 49 mm diameter filter apparatus. The inlet velocity is 0.054 m/s, filter
number is 23-50 ppi.
Figure.51 The pressure gradient for the 49 mm diameter filter apparatus. The inlet velocity is 0.054 m/s,
filter number is 23-50 ppi.
65
5.3.2 101mm Filter Finite Element Modelling

Figure.52 shows the geometry of the model for the 101 mm diameter filter holder apparatus.
Figure.52 Modelling geometry of 101 mm filter holder apparatus.
The mesh used in the FEM model is shown in Figure.53.
66
Figure.53 Modelling mesh used for the 101 mm filter apparatus.

Figure.54 and Figure.55 shows the flow fields and the pressure gradient for the 101 mm
diameter filter apparatus at 7 oC and 0.7 m/s inlet water velocity. The filter properties used
here are those of the 20-50 ppis.
67
Figure.54 The flow field for the 101 mm diameter filter apparatus. The inlet velocity is 0.7 m/s, filter
number is 20-50 ppi.
Figure.55 The pressure gradient for the 101 mm diameter filter apparatus. The inlet velocity is 0.7 m/s,
filter number is 20-50 ppi.
68
5.4 Comparison of CFD Results with Experimental Data

The experimental data obtained from tests 26-28 are compared here with the results from
CFD modelling. The details of the comparison are shown in Tables 29 through 31 in the
Appendix. Figure.56 shows CFD results for tests 26-28, compared against the experimental
pressure gradients. The average error is ~1.74%
CFD Calculated Pressure Gradient, Pa/m
1400000
1200000
1000000
800000
600000
400000
200000
200000
400000
600000
800000
1000000
1200000
1400000
Measured Pressure Gradient, Pa/m
Figure.56 CFD calculated pressure gradients compared against the measured pressure gradients for test
26-28.
From Figure.56, it can be found that the CFD modelling results are almost identical with the
experimental values. It is important to note, that if significant bypassing had occurred during
these experiments, it would not have been possible to achieve agreement between the CFD
model and the experimental data for the 49 mm filter design, i.e. it would have been necessary
to add a bypass channel of finite thickness. The close agreement between experimental and
CFD results is thus taken as confirmation that the wall sealing arrangements were in fact of
negligible permeability.
For the 101 mm filter experiments, the final effective flow field diameters for use with
Equation (2) obtained by the CFD modelling were: 65.5 mm, 66 mm, 66.1 mm and 66.5 mm
for the 30 ppi, 40 ppi, 50 ppi and 80 ppi respectively.
Figure.57 shows the CFD results for test 9 and 23 through 25, compared against the
experimental pressure gradients. The average error is about 2.24%. The detailed information
69
for the comparison of the CFD results and experimental data are shown in Tables 32 through
35 in the Appendix.
CFD Calculated Pressure Gradient, Pa/m
1200000
1000000
800000
600000
400000
200000
200000
400000
600000
800000
1000000
1200000
Measured Pressure Gradient, Pa/m
Figure.57 CFD calculated pressure gradient compared against the measured pressure gradient for test 9,
and 23 through 25.
The 22-30 ppi, 23-50 ppi and 24-80 ppi filters were cut from the centre of the 18-30 ppi,
20-50 ppi and 21-80 ppi filters. Figure.58 shows the comparison of test 9 and 27. The
superficial velocity values for test 9 were calculated using the effective flow field diameter
determined by the CFD modelling.
70
1400.00
1200.00
1000.00
800.00
600.00
400.00
200.00
0.00
0.0000
0.0500
0.1000
0.1500
0.2000
20-50ppi
0.2500
Velocity (m/s)
0.3000
0.3500
0.4000
0.4500
0.5000
23-50ppi
Figure.58 Comparison between the 101 mm and 49 mm filter results.
The regressions are:

20-50 ppi: y 5917160.66x2 83380x, R2 0.99995
(78)
23-50 ppi: y 6036686.76 x2 87890 x, R2 0.99998
(79)
From Figure.58, Equation (78) and Equation (79), it can be found that the results obtained by
these two experiments are nearly identical. In the expanding flow field 101 mm apparatus
design, the side wall is a stagnant region and bypassing is low (80 ppi) to negligible (30-50
ppi) even in the absence of sealing, as shown in Figure.54. With proper wall sealing (Method
2 using grease and paper), it can be assumed that bypassing in the 101 mm design is always
negligible. Since there is nearly no bypassing for 101 mm filters, the agreement shown in
Figure.58, indicates that nearly no bypassing occurred during the 49 mm filter experiments as
well.
71
Chapter 6
Conclusions
The bypassing during the experiments affects the final results to a critical extent. Only the
well-sealed procedure gave meaningful pressure gradients. Two ways were found to avoid
bypassing in this study. The first one is use cellulose and silicon grease packed filters, as
shown in Figure. 16. A second choice is to use filters with a larger size than the inlet and outlet
pipes; however, in this case CFD modelling is necessary to compute the effective flow field
diameter in order to obtain the real relationship between fluid velocity and the measured
pressure gradients.
Different methods of calculating k1 and k2 in Equation (2) can give quite different answers.
The Method 3 was found to give the most physically meaningful mathematical results in this
study, i.e. the iterative procedure to first guess k1 and then correlate the remainder for k2 using
an exponential regression.
It was established that the porosities of the filters decrease consistently with increasing
number of ppi, and that when the velocity increased the pressure drop increased even faster
with CFFs of higher nominal pore count (e.g. for 80 vs. 30 ppi).
Forchheimers equation (2) was found to adequately describe the pressure drop for the CFFs
used in this study. Both k1 and k2 values vary inversely to filter ppi. The filters with lower ppi
proved to have lower tortuosities, which in agreement with literature
By testing different filters with the same ppi, it was found that every filter has a slightly
different k1 and k2 value. Even the filter sections cut from the same filter do not give exactly
the same k1 and k2 values. Thus there is no specific permeability for each type of filters, only a
range and a standard deviation can be obtained.
After modifying the Forchheimer and Ergun equations, Equation (39) was obtained. This
equation works well to predicate the pressure gradients in SIVEX filters. The average
absolute error is only 10.7 %.
It was also established that the permeability of a CFF does not change by the changing of
fluid flow direction (at least for an incompressible fluid like water or aluminium).
72
The COMSOL 2 D axial symmetric CFD model was used in this study to verify the
experimental methods. Comparing the CFD results with the experimental data, shows a
maximum error of only about 6.5%. It can be concluded that COMSOL can be accurately
used to obtain filter permeabilitys using the expanding flow field design and can accurately
simulate results for either type of apparatus, once the correct k1 and k2 values have been
determined and programmed.
73
Chapter 7
Further Work
It is recommended that additional experiments should be executed to explore the change of
pressure drop with the filter thickness. This would provide further validation that wall
bypassing has been prevented by the current experimental procedure and also help to
determine if the filter material is truly isotropic, for example: Does the original outer surface
(top and bottom) possess a different permeability than the balance of the filter?
Additional experiments should be conducted with filters produced from different suppliers
and with the same range of ppi to validate the correlation obtained by this study, or find a
more universal correlation suitable for all filters.
Additional experiments could be conducted with a low range, 0-0.1 Bar pressure transducer,
at low velocity (0-0.05 m/s), in order to study the Darcy and transitional regions in the greater
detail.
74
References
1
L.S. Aubrey and R. Schmahl: Application of Reticulated Silicon Carbide Foam Filters for Iron Casting
Applications, Foundry Trade J., September, 1991, 769-773.
2 L.S. Aubrey and J.E. Dore: Ceramic Foam- A Deep Bed or Caking Filter in Aluminium Cast Shop
Operations, Light Metals, 1993, 1009-1020.
3
W.H. Sutton, J.C. Palmer, and J.C. Morris: Development of Ceramic Foam Materials for Filtering High
Temperature Alloys, AFS Trans, 1985, vol. 51, 339-346.
R.W. Gairing and M. Cummings: Operational Experience Using Inclusion Filtration Technology for
Continuously Cast Steels Iron, Stell Maker, vol. 20, 1993, 55-58.
M. D. M. Innocentini, P. Sepulveda and F. S. Ortega, In: Scheffler M Colombo P (eds) Cellular ceramics:
structure, manufacturing, properties and application, Wiley-VCH Verlag GmbH, Weinheim, Germany, 2005,
313.
V. P. Rodrigues, L. P. Argiona, R. G. Pilegggi, R. C. Romano, J. R. Coury and M. D. M. Innocentini:

Permeability Optimization of Hot Aerosol Filters Prepared from Foaming of Ceramic Suspensions. In: 10th
International Conference and Exhibition of the European Ceramic Society, Berlin-Alemanha, 2007.
N. L. Freitas, J. A. S. Goncalves, M. D. M. Innocentini and J. R. Coury: Development of a Double-Layered

Ceramic Filter for Aerosol Filtration at High-Temperatures: The Filter Collection Efficiency, Journal of
Hazard Mater, Volume 136, Issue 3, 2006, 746-756.

V. W. Wang, G. J. Mako and A. L. Matthews: Pressed Cellular Filter Application in an Aluminum Foundry,
AFS Transact., 1996, 1045-1052.
F.A. Acosta G, A.H. Castillejos E., J.M. Almanza R., and A. Flores V: Analysis of Liquid Flow through
Ceramic Porous Media Used for Molten Metal Filtration, vol. 26B, 1995, 159-171.
10 P. Sandford and S.R. Sibley: Optimization of Al Casting Productivity Using Foam Filter Technology and
Application, AFS Transact, 1996, 1063-1068.
11 www.lanik.cz, accessed 26.09.2011.

12 A. Montillet, J. Comiti, and J. Legrand: Determination of Structural Parameters of Metallic Foams from
Permeametry Measurements, J. Mater. Sci., 27, 1992, 4460-4464.
13 M.D.M. Innocentini, P. Sepulveda, V.R. Salvini, J.R. Coury, and V.C. Pandolfelli: Permeability and
Structure of Cellular Ceramics: A Comparison between Two Preparation Techniques, J. Am. Ceram. Soc.,
vol. 81, 1998, 3349-52.
14 Martyn V. Twigg and James T. Richardson: Fundamentals and Applications of Structured Ceramic Foam
Catalysts, Ind. Eng. Chem. Res. vol. 46, 2007, 4166-4177.
15 L.J. Gibson and M.F. Ashby: Cellular Solids, Structure and Properties; Pergamon Press: Oxford, U.K.,
1988.
16 L.J. Gibson and M.F. Ashby: Cellular Solids, Structure and Properties, Second Edition; University of
Cambridge Press: Cambridge, U.K., 1999.
17 R. Viskanta, in: J.R. Lloyd, Y. Kurosaki (Eds.): Enhancement of Heat Transfer in Industrial Combustion
Systems: Problems and Future Challenges, Proceedings of the 3rd ASME/JSME, New York, 1991, 163-173.
75
18 F. Anderson: Heat Transport Model for Fibre Burners, Prog. Energy Combust. sci. 1992 , vol. 18, 1-12.
19 S.B. Bhaduri: Science and Technology of Ceramic Foams. Adv. Perform. Mater. vol. 1, 1994, 205-220.
20 T.B. Sweeting, D.A. Norris, L.A. Storm and J.R. Morris: Reticulated Ceramics for Catalyst Support
Applications. Mater Res. Soc. Symp. Proc. 1995, 309-314.
21 V. W. Wang, G. J. Mako and A. L. Matthews: Pressed Cellular Filter Application in an Aluminum Foundry,
AFS Transact., 1996, 1045-1052.
22 R. Fritzsch: Filtration of Aluminium Melts using CF-filters with Electromagnetic Field, Trondheim: s.n.,
2010 .
23 S. Ray, B. Milligan and N. Keegan: Measurement of Filtration Performance, Filtration Theory and Practical
Applications of Ceramic Foam Filters, Aluminium Cast House Technology, 2005, 1-12.
24 J.T. Richardson, Y. Peng and D. Remue: Properties of Ceramic Foam Catalyst Supports: Pressure Drop,
Applied Catalysis A: General 204, 2000, 19-32.
25 S.F. Ray: Recent Improvements in the Measurement and Control of Ceramic Foam Filter Quality, Presented
at the International Melt Quality Workshop, Spain, 2001.
26 Mark W. Kennedy: Detailed PhD plan 2010, Department of Materials Science and Engineering, NTNU,
Trondheim, 2010.
27 P. Sandford and S.R. Sibley: Optimization of Al Casting Productivity Using Foam Filter Technology and
Application, AFS Transact, 1996, 1063-1068.
28 E.A. Moreira and J.R. Coury: The Influence of Structural Parameters on the Permeability of Ceramic Foam,
Brazilian Journal of Chemical Engineering, vol.21, n. 01, 2004, 23-33.
29 L. Zhang and L.N.W. Damoah: Removal of Inclusion Rom Aluminium through Filtration, ASM International,
2010.
30 Jintai, J. Alumina ceramic foam Filter (CFF-Al). Special Material Co. Ltd., 2011.
31 A. Montillet, J. Comiti, and J. Legrand: Determination of Structural Parameters of Metallic Foams from
Permeametry Measurements, J. Mater. Sci., vol. 27, 1992, 4460-4464.
32 M.D.M. Innocentini, P. Sepulveda, V.R. Salvini, J.R. Coury, and V.C. Pandolfelli: Permeability and
Structure of Cellular Ceramics: A Comparison between Two Preparation Techniques, J. Am. Ceram. Soc.,
1998, 3349-3352.
33 Henry Darcy: Les Fontaines Publiques de la Ville de Dijon, Victor Dalmont, Paris, 1856, 647.
34 M.D.M. Innocentini, V.R. Salvini, V.C. Pandolfelli, and J.R. Coury: Assessment of Forchheimers Equation
to Predict the Permeability of Ceramic Foams, J. Am. Ceram. Soc. vol. 82, 1999, 1945-1948.
35 S. Liu, A. Afacan and J.H.Masliyah: Steady Incompressible Laminar Flow in Porous Media, Chem. Eng.
Science, vol. 49, 1994, 3565-3586.
36 P. Forchheimer: Wasserbewegungdurch Boden, Z. Ver. Deutsch. Ing. vol. 45, 1901, 1782-1788.
37 D. Ruth and H. Ma: On the Derivation of the Forchheimer Equation by Means of the Averaging Theorem,
Transp. Porous Media, vol. 7, 1992, 255-64.
38 S. Ergun: Flow through Packed Columns, Chem. Eng. Progress., 1952, vol. 48, n. 2, 89-94.
39 A. P. Philipse, H. L. Schram: Non- Darcian Airflow through Ceramic Foams, J. Am. Cer. Soc., vol. 74, n. 4,
1991, 728-732.
40 M. V. Twigg, J. T. Richardson: Preparation and Properties of Ceramic Foam Catalyst Supports, Studies in
Surface Science and Catalysis, vol. 91, 1995, 345-359.
76
41 M. D. M. Innocentini, V. R. Salvini, V. C. Pandolfelli and A. Macedo: Prdiction of Ceramic Foams

Permeability Using Erguns Equation, Material Research., vol.2 n. 4, Sao Carlos, 1999, 283-289.
42 E. Maire, P. Colombo, J. Adrien, L. Babout and L. Biasetto: Characterization of the Morphology of Cell
Ceramics by 3D Image Processing of X-ray Tomography, Journal of the European Ceramic Society, vol. 27,
issue 4, 2007, 1973-1987.
43 B. Dietrich, W. Schabel, M. Kind, and H. Martin: Pressure Drop Measurements of Ceramic
Sponges-Determining the Hydraulic Diameter, Chemical Engineering Science, vol. 64, 2009, 3633-3640.
44 B. Dietrich: Pressure Drop Correlation for Ceramic and Metal Sponges, Chemical Engineering Science, vol.
74, 2012, 192-199.
45 S. A. Shakiba, R. Ebrahimi and M. Shams: Experimental Investigation of Pressure Drop through Ceramic
Foams: An Empirical Model for Hot and Cold Flow, Journal of Fluids Engineering, vol. 133, 2011,.
46 D. Edouard, M. Lacroix, C.P. Huu and F. Luck: Pressure Drop Modelling on Solid Foam: State-of-the Art
Correlation, Chemical Engineering Journal, vol. 144, 2008, 299-311.
47 V.L. Streeter and E.B. Wylie: Fluid Mechanics, McGraw-Hill Ryerson, 1981, ISBN 0-07-548015-8, 8-10.
48 S. Keith: Mechanics (Third ed.), Addison-Wesley, 1971, ISBN 0-201-07392-7.
49 http://en.wikipedia.org/wiki/Viscosity#Types_of_visitscoy, accessed 13.04.2012.
50 http://en.wikipedia.org/wiki/Reynolds_number#cite_note-NASA-4, accessed 13.04.2012.
51 S. George: On the Effect of the Internal Friction of Fluids on the Motion of Pendulums, Transactions of the
Cambridge Philosophical Society 9, 1851, 8106.
52 O. Reynolds: An Experimental Investigation of the Circumstances which Determine whether the Motion of
Water shall be Direct or Sinuous, and of the Law of Resistance in Parallel Channels, Philosophical
Transactions of the Royal Society 174 (0),1883, 935982
53 http://www.engineeringtoolbox.com/reynolds-number-d_237.html, accessed 15.04.2012
54 J.P. Holman: Heat Transfer, McGraw-Hill, 2002, 207.
55 http://www.engineeringtoolbox.com/laminar-transitional-turbulent-flow-d_577.html, accessed 15.04.2012
56 L.F. Moody: Friction Factors for Pipe Flow, Transactions of the A.S.M.E, 1944, 671-684.
57 http://www.howequipmentworks.com/physics/fluids/flow/flow.html, accessed 16.04.2012.
58 http://www.engineeringtoolbox.com/entrance-length-flow-d_615.html, accessed 16.04.2012.
59 F.M. White: Fluid Mechanics (Fourth Edition), McGraw-Hill, 2010, ISBN 0073529346, 330-331.
60 W. Zhi-qing: Study on Correction Coefficients of Liminar and Turbulent Entrance Region Effect in Round
Pipe, Applied Mathematics and Mechanics, vol. 3, 1982, 433-446.
61 Z.Q. Wang: Study on Correction Coefficients of Liminar and Turbulent Entrance Region Effect in Round
Pipe, Applied Mathematics and Mechanics (English Edition, vol. 4, n. 3, 1982, 434-446.
62 M. Matyka, A. Khalili and Z. Koza: Tortuosity-porosity Relation in the Porous Media Flow, Phys. Rev. E,
vol. 78, 2008.
63 P. Habisreuther, N. Djordjevic and N. Zaralis: Statistical Distribution of Residence Time and Tortuosity of
Flow through Open-cell Foams, Chemical Engineering Science, vol. 64, 2009, 4943-4954.
64 G.S. Armatas: Determination of the Effects of the Pore Size Distribution and Pore Connectivity Distribution
on the Pore Tortuosity and Diffusive Transport in Mode Porous Networks, Chemical Engineering Science,
vol. 61, 4662-4675.
65 A. Scheidegger: The Physics of Flow through Porous Media, University of Toronto, Toronto, 1974.
77
66 M.W. Kennedy, K. Zhang, R. Fritzitch, S. Akhtar, J.A. Bakken and R.E. Aune: Characterization of Ceramic
Foam Filters used for Liquid Metal Filtration, to be submitted to Metallurgical Transactions B, 2012, 1-46.
67 P.D. Desai, H.M. James, and C.Y. Ho: Electrical Resistivity of Aluminum and Magnanese, Journal of
Physical and Chemical Reference Data, vol. 13, n. 4, 1984, 1131-1172.

68 http://www.henrys.com/61271-EPSON-PERFECTION-V330-PHOTO-FLATBED.aspx, accessed
10.07.2012.
69 http://en.wikipedia.org/wiki/Energy-dispersive_X-ray_spectroscopy, accessed 10.10.2011.
70 J.P. Anson, R.A.L. Drew and J.E. Gruzleski: The Surface Tension of Molten Aluminum and Al-Si-Mg Alloy
under Vacuum and Hydrogen Atmospheres, Metallurgical and Meterials Transacitions B, vol. 30B, 1999,
1027-1032.
71 D. Wang and R.A. Overfelt: Oscillating Cup Viscosity Measurements of Aluminum Alloys: A201, A319 and
A356, International Jouranal of Thermophysics, vol. 23, n. 4, 2002, 1063-1076.
72 V. Sarou-Kanlan, F. Millot and J.C. Rifflet: Surface Density of Oxygen-Free Liquid Aluminum at High
Temperature, International Journal of Thermophysics, vol. 24, no. 1, 2003, 277-286.
73 J. Huang, T. Mori and J.G. Conley: Simulation of Microporosity Formation in Modified and Unmodified
A356 Alloy Castings, Metallugical and Materials Transactions B, vol. 29B, 1998, 1249-1260.
74 M. W. Kennedy, S. Akhtar, J. A. Bakken, and R. E. Aune: Effect of Electromagnetic Force on the
Distribution of Inclusions from Liquid Aluminum by a Ceramic Foam Filter in Continuous Filtration Tests,
submitted to Metallurgical and Materials Transactions B, 2012.
75 M. W. Kennedy, S. Akhtar, J. A. Bakken, and R. E. Aune: Electromagnetically Enhanced Filtration of
Aluminum Melts, TMS Light Metals 2011, San Diego, California, 2011.
76 R. Fritzsch, M. W. Kennedy, S. Akhtar, J. A. Bakken, and R. E. Aune: Electromagnetically Modified
Filtration of Liquid Aluminium with a Ceramic Foam Filter, to be presented at the Electromagnetic
Processing of Materials, Beijing, China, 2012, 763-768.
77 http://en.wikipedia.org/wiki/Porosity, accessed 19.04.2012.
78 J. Groe, B. Dietrich, H. Martin, M. Kind, J. Vicente, and E. H. Hardy: Chemical Engineering & Technology,
vol. 31, 2008, 307-314.
79 B. Dietrich, G. I. Garrido, P. Habisreuther, N. Zarzalis, H. Martin, M. Kind, and B. Kraushaar-Czarnetzki:
Morphological Characterization of Ceramic Sponges for Applications in Chemical Engineering, Industrial &
Engineering Chemistry Research, vol. 48, 2009, 10395-10401.
80 N. Keegan, W. Schneider, and H. Krug: Evaluation of the Efficiency of Fine Pore Ceramic Foam Filters,
TMS Light Metals, San Diego, 1999, 1-10.
81 S. Ergun and A.A. Orning: Fluid Flow through Randomly Packed Columns and Fluidized Beds, Industrial
&Engineering Chemistry, vol. 41, 1949, 1179-1184.
82 E. Moreira, M. Innocentini, and J. Coury: Permeability of Ceramic Foams to Compressible and
Incompressible Flow, Journal of the European Ceramic Society, vol. 24, 2004, 3209-3218.
83 G. Diedericks and J. Du Plessis: Electrical Conduction and Formation Factor in Isotropic Porous Media,
Advances in water resources, vol. 19, 1996, 225-239.
84 COMSOL Multiphysics 4.2a documentation.
85 http://en.wikipedia.org/wiki/Navier%E2%80%93Stokes_equations#cite_ref-Ach_14-0, accessed 30.06.2012
86 D.C. Wilcox: Turbulence Modelling for CFD, 2nd ed., DCW Industries, 1998.
78
87 K. Abe, T. Kondoh, and Y. Nagano: A New Turbulence Model for Predicting Fluid Flow and Heat Transfer
in Separating and Reattaching FlowsI. Flow Field Calculations, International Journal of Heat and Mass
Transfer, vol. 37, no. 1, 1994, 139-151.
79
Appendix
Table 1 Test 1 2-30 ppi, Diameter: 48.57 mm, Thickness: 50.18 mm
Velocity (m/s) Pressure Drop (Pa/m) Temperature (oC) Density (kg/m3) Viscosity (kg/m.s)
0.029
2982
6.4
1000.0
1.401E-03
0.009
823
6.7
999.9
1.388E-03
0.063
12241
7.0
999.9
1.377E-03
0.016
1598
7.3
999.9
1.367E-03
0.084
20299
7.5
999.9
1.358E-03
0.133
46710
7.7
999.9
1.352E-03
0.250
155893
7.8
999.9
1.348E-03
0.334
273426
7.9
999.9
1.344E-03
0.438
493635
7.9
999.9
1.342E-03
0.550
683574
8.6
999.8
1.316E-03
0.654
942796
8.9
999.8
1.307E-03
0.691
1071006
9.0
999.8
1.304E-03
0.425
403214
9.0
999.8
1.302E-03
0.285
191152
9.1
999.8
1.298E-03

0.005
1067
6.9
999.9
1.383E-03
0.022
6091
7.3
999.9
1.367E-03
0.042
15132
7.5
999.9
1.357E-03
0.068
35601
7.8
999.9
1.345E-03
0.094
63438
8.1
999.9
1.336E-03
0.116
95251
8.3
999.9
1.328E-03
0.154
165072
8.4
999.9
1.324E-03
0.218
323551
8.6
999.8
1.318E-03
0.271
488434
8.7
999.8
1.314E-03
0.303
614326
8.9
999.8
1.307E-03
0.356
795660
9.0
999.8
1.303E-03
0.392
893485
9.2
999.8
1.296E-03
0.433
988171
9.5
999.8
1.284E-03
0.482
1123292
9.8
999.7
1.275E-03
0.526
1251507
9.9
999.7
1.270E-03
0.012
3893
10.2
999.7
1.261E-03
0.047
16451
10.4
999.7
1.253E-03
0.083
39055
10.5
999.7
1.248E-03
0.133
85937
10.6
999.7
1.247E-03
0.207
202557
10.7
999.7
1.242E-03
80

0.004
2145
6.6
999.9
1.392E-03
0.030
10046
7.1
999.9
1.374E-03
0.080
38283
7.4
999.9
1.363E-03
0.121
80527
7.6
999.9
1.354E-03
0.150
124960
7.8
999.9
1.345E-03
0.199
220398
8.1
999.9
1.337E-03
0.250
333215
8.2
999.9
1.332E-03
0.304
468173
8.4
999.9
1.325E-03
0.355
597009
8.5
999.8
1.320E-03
0.400
692701
8.7
999.8
1.314E-03
0.458
800349
8.8
999.8
1.308E-03
0.511
937777
9.0
999.8
1.302E-03
0.564
1049960
9.2
999.8
1.297E-03
0.609
1132377
9.2
999.8
1.294E-03
0.634
1145321
9.4
999.8
1.286E-03
0.334
334509
9.6
999.8
1.282E-03
0.189
116546
9.7
999.8
1.278E-03
0.072
21478
9.8
999.7
1.273E-03
0.684
1092146
6.0
1000.0
1.419E-03
0.048
7335
6.3
1000.0
1.407E-03
0.099
26821
6.5
999.9
1.396E-03
0.204
109275
6.7
999.9
1.389E-03
0.290
203839
7.1
999.9
1.373E-03
0.405
385956
7.5
999.9
1.358E-03
0.498
587554
7.7
999.9
1.350E-03
0.592
805953
8.0
999.9
1.338E-03
0.645
974158
8.2
999.9
1.332E-03

0.311
1179700
6.9
999.9
1.381E-03
0.335
1360744
7.3
999.9
1.364E-03
0.297
1069497
7.4
999.9
1.361E-03
0.253
782618
7.6
999.9
1.354E-03
0.193
468132
7.7
999.9
1.350E-03
0.153
299258
7.9
999.9
1.342E-03
0.102
139445
8.0
999.9
1.338E-03
0.074
76229
8.2
999.9
1.331E-03
0.034
20471
8.4
999.9
1.324E-03
81

0.378
1305519
5.7
1000.0
1.431E-03
0.343
1078644
5.9
1000.0
1.421E-03
0.288
774394
6.1
1000.0
1.413E-03
0.246
570100
6.3
1000.0
1.404E-03
0.199
379939
6.5
1000.0
1.398E-03
0.135
181663
6.7
999.9
1.390E-03
0.084
75497
7.0
999.9
1.379E-03
0.033
16655
7.2
999.9
1.370E-03

0.405
877748
6.3
1000.0
1.407E-03
0.043
13810
6.7
999.9
1.391E-03
0.118
78056
6.8
999.9
1.384E-03
0.211
242931
7.0
999.9
1.379E-03
0.285
443618
7.1
999.9
1.374E-03
0.361
692108
7.3
999.9
1.367E-03
0.483
1248950
7.5
999.9
1.360E-03
Table8 Test 8 4-50 ppi, Diameter: 48.61 mm, Thickness: 50.60 mm

0.449
1028147
4.5
1000.0
1.483E-03
0.055
18831
5.1
1000.0
1.455E-03
0.152
123763
5.5
1000.0
1.439E-03
0.224
257183
5.7
1000.0
1.430E-03
0.299
474541
5.9
1000.0
1.424E-03
0.370
698184
6.3
1000.0
1.406E-03
0.491
1240875
6.5
1000.0
1.399E-03
Table 9 Test 9 18-50 ppi, Diameter: ~101 mm, Thickness: 49.58 mm

0.393
937588
5.1
1000.0
1.454E-03
0.102
69810
5.5
1000.0
1.437E-03
0.156
156059
5.6
1000.0
1.432E-03
0.214
286174
5.8
1000.0
1.425E-03
0.284
505716
5.9
1000.0
1.421E-03
0.333
685889
6.2
1000.0
1.411E-03
0.420
1085478
6.4
1000.0
1.403E-03
82
Table10 Test 10 18-50 ppi, Diameter: ~101 mm, Thickness: 49.58 mm

0.366
816673
5.9
1000.0
1.422E-03
0.016
2992
6.2
1000.0
1.409E-03
0.047
15790
6.4
1000.0
1.401E-03
0.125
98103
6.6
999.9
1.391E-03
0.177
197459
6.9
999.9
1.383E-03
0.417
1074903
7.3
999.9
1.367E-03
0.372
864438
7.4
999.9
1.362E-03
0.302
572042
7.5
999.9
1.356E-03
0.242
359909
7.8
999.9
1.347E-03
Table11 Test 11 1-30 ppi, Diameter: 48.71 mm, Thickness: 50.48 mm

0.710
978343
6.5
1000.0
1.398E-03
0.046
5771
6.8
999.9
1.387E-03
0.117
30911
7.0
999.9
1.379E-03
0.218
99618
7.1
999.9
1.373E-03
0.325
215863
7.2
999.9
1.368E-03
0.427
375867
7.7
999.9
1.349E-03
0.523
549073
8.0
999.9
1.341E-03
0.603
690255
8.1
999.9
1.334E-03

0.367
776197
5.7
1000.0
1.431E-03
0.022
5233
6.3
1000.0
1.406E-03
0.067
31815
6.4
1000.0
1.400E-03
0.163
171054
6.6
999.9
1.394E-03
0.229
335865
6.7
999.9
1.389E-03
0.283
503703
6.8
999.9
1.387E-03
0.417
1067479
6.9
999.9
1.381E-03
0.469
1247505
7.0
999.9
1.377E-03
83

0.506
1152868
6.4
1000.0
1.401E-03
0.025
5251
6.9
999.9
1.383E-03
0.057
18846
7.1
999.9
1.373E-03
0.127
78605
7.3
999.9
1.366E-03
0.199
185871
7.5
999.9
1.359E-03
0.285
370994
7.6
999.9
1.355E-03
0.379
652171
7.7
999.9
1.349E-03
0.426
834240
7.9
999.9
1.343E-03
0.477
1034413
8.0
999.9
1.339E-03
0.523
1236086
8.2
999.9
1.333E-03

0.450
1185370
5.7
1000.0
1.428E-03
0.044
15610
6.1
1000.0
1.414E-03
0.102
67181
6.3
1000.0
1.406E-03
0.185
208637
6.4
1000.0
1.400E-03
0.243
352890
6.6
999.9
1.393E-03
0.298
540831
6.8
999.9
1.385E-03
0.362
766017
7.1
999.9
1.374E-03
0.411
971166
7.2
999.9
1.368E-03
0.467
1263214
7.4
999.9
1.361E-03

0.453
1045977
4.8
1000.0
1.468E-03
0.039
10604
5.5
1000.0
1.437E-03
0.112
69538
5.8
1000.0
1.428E-03
0.171
155384
6.0
1000.0
1.419E-03
0.246
317343
6.2
1000.0
1.409E-03
0.314
507861
6.4
1000.0
1.401E-03
0.382
746837
6.6
999.9
1.392E-03
0.507
1239343
6.8
999.9
1.386E-03
84

0.468
1039446
6.4
1000.0
1.403E-03
0.035
7550
6.7
999.9
1.389E-03
0.075
31264
6.9
999.9
1.381E-03
0.143
105264
7.2
999.9
1.370E-03
0.194
188586
7.4
999.9
1.363E-03
0.250
309131
7.5
999.9
1.358E-03
0.315
488208
7.7
999.9
1.352E-03
0.390
735226
7.8
999.9
1.348E-03
0.514
1241277
7.9
999.9
1.341E-03
Table 17 Test 17 11-80 ppi, Diameter: ~48.00 mm, Thickness: 50.59 mm

0.338
1355909
13.1
999.4
1.164E-03
0.317
1197566
13.2
999.4
1.160E-03
0.277
923720
13.3
999.4
1.158E-03
0.242
713898
13.4
999.4
1.155E-03
0.184
436260
13.6
999.3
1.151E-03
0.120
191626
13.7
999.3
1.148E-03
0.080
90628
13.9
999.3
1.141E-03
0.055
45271
14.0
999.3
1.136E-03

0.338
1355909
13.1
999.4
1.164E-03
0.317
1197566
13.2
999.4
1.160E-03
0.277
923720
13.3
999.4
1.158E-03
0.242
713898
13.4
999.4
1.155E-03
0.184
436260
13.6
999.3
1.151E-03
0.120
191626
13.7
999.3
1.148E-03
0.080
90628
13.9
999.3
1.141E-03
0.055
45271
14.0
999.3
1.136E-03
85

0.504
1264101
6.3
1000.0
1.406E-03
0.454
1022439
6.5
1000.0
1.399E-03
0.405
804974
6.6
999.9
1.395E-03
0.310
490138
6.7
999.9
1.388E-03
0.262
350434
7.0
999.9
1.379E-03
0.179
166969
7.1
999.9
1.375E-03
0.139
100367
7.3
999.9
1.366E-03
0.034
8313
7.4
999.9
1.361E-03
0.083
39380
7.6
999.9
1.355E-03

0.356
926569
7.1
999.9
1.376E-03
0.427
1318887
7.8
999.9
1.345E-03
0.368
1012900
8.1
999.9
1.336E-03
0.330
820307
8.2
999.9
1.332E-03
0.257
504143
8.4
999.9
1.326E-03
0.208
369229
8.4
999.8
1.323E-03
0.156
196134
8.5
999.8
1.320E-03
0.095
84645
8.7
999.8
1.312E-03
0.036
16075
8.7
999.8
1.312E-03

0.488
1157648
9.9
999.7
1.271E-03
0.507
1257931
10.4
999.7
1.253E-03
0.434
902559
10.5
999.7
1.248E-03
0.383
700781
10.6
999.7
1.246E-03
0.332
526432
10.7
999.7
1.242E-03
0.276
374688
10.8
999.7
1.239E-03
0.224
247365
10.9
999.6
1.235E-03
0.197
192677
11.0
999.6
1.231E-03
0.140
100579
11.2
999.6
1.227E-03
0.052
15443
11.3
999.6
1.222E-03
86
Table 22 Test 22 17-50 ppi, Diameter: 49.12 mm, Thickness: 50.41 mm (opposite direction)
0.313
491832
9.9
999.7
0.00127
0.246
304892
10.1
999.7
0.00126
0.185
169788
10.3
999.7
0.00126
0.136
95532
10.5
999.7
0.00125
0.083
39229
10.6
999.7
0.00125
0.052
16488
10.8
999.7
0.00124
0.426
866361
11.5
999.6
0.00122
0.366
637771
11.6
999.6
0.00121

0.031
2723
6.6
999.9
1.393E-03
0.074
11712
6.7
999.9
1.389E-03
0.156
50272
6.9
999.9
1.380E-03
0.242
116462
7.1
999.9
1.374E-03
0.343
233955
7.3
999.9
1.368E-03
0.445
386486
7.4
999.9
1.362E-03
0.610
735607
7.6
999.9
1.354E-03

0.408
501847
5.1
1000.0
1.456E-03
0.020
2179
5.5
1000.0
1.436E-03
0.093
28841
5.6
1000.0
1.435E-03
0.167
86720
5.8
1000.0
1.428E-03
0.249
187764
5.9
1000.0
1.421E-03
0.326
324686
6.0
1000.0
1.417E-03
0.396
485160
6.2
1000.0
1.411E-03
0.466
684740
6.2
1000.0
1.410E-03
0.511
822209
6.3
1000.0
1.405E-03
0.539
920699
6.5
1000.0
1.399E-03
87

0.295
948953
7.0
999.9
1.378E-03
0.021
9916
7.3
999.9
1.365E-03
0.058
46158
7.5
999.9
1.358E-03
0.091
101936
7.7
999.9
1.352E-03
0.137
217300
7.7
999.9
1.349E-03
0.195
428441
7.8
999.9
1.347E-03
0.236
621747
7.9
999.9
1.342E-03
0.314
1088535
8.0
999.9
1.339E-03
0.326
1181713
8.2
999.9
1.331E-03

0.434
1194904
6.3
1000.0
1.404E-03
0.460
1310433
6.6
999.9
1.395E-03
0.402
1008834
6.7
999.9
1.389E-03
0.341
726048
6.8
999.9
1.384E-03
0.265
445758
7.0
999.9
1.378E-03
0.204
268844
7.1
999.9
1.375E-03
0.172
193182
7.2
999.9
1.368E-03
0.097
65509
7.5
999.9
1.360E-03
0.055
23185
7.6
999.9
1.352E-03

0.637
807100
7.4
999.9
1.361E-03
0.769
1083791
7.9
999.9
1.343E-03
0.719
940447
8.0
999.9
1.338E-03
0.529
519563
8.2
999.9
1.333E-03
0.416
325407
8.3
999.9
1.329E-03
0.273
142851
8.4
999.9
1.324E-03
0.173
60003
8.6
999.8
1.317E-03
0.110
25014
8.7
999.8
1.314E-03
0.057
7345
8.9
999.8
1.307E-03
88

0.380
1379323
5.8
1000.0
1.427E-03
0.384
1382511
5.9
1000.0
1.423E-03
0.351
1172774
6.1
1000.0
1.415E-03
0.311
942852
6.1
1000.0
1.414E-03
0.262
629565
5.9
1000.0
1.423E-03
0.198
372652
6.0
1000.0
1.419E-03
0.131
174497
6.1
1000.0
1.414E-03
0.091
88998
6.2
1000.0
1.409E-03
0.036
19485
6.5
1000.0
1.398E-03
0.066
50886
6.8
999.9
1.385E-03
Table 29 Comparison of Experimental Data and CFD Results for Test 26, 22-30 ppi
Velocity
(m/s)
0.637
0.769
0.719
0.529
0.416
0.273
0.173
0.110
0.057
Experiment
Pressure Drop
(Pa/m)
807100
1083791
940447
519563
325407
142851
60003
25014
7345
89
COMSOL
Pressure Drop
(Pa/m)
754821
1095682
959122
524132
326559
143441
59490
25127
7413
Error
%
6.48
1.10
1.99
0.88
0.35
0.41
0.85
0.45
0.92
Superficial Experiment
Velocity Pressure Drop
(m/s)
(Pa/m)
0.434
1194904
0.460
1310433
0.402
1008834
0.341
726048
0.265
445758
0.204
268844
0.172
193182
0.097
65509
0.055
23185
COMSOL
Pressure Drop
(Pa/m)
1164648
1302957
1001339
726148
444351
267871
192433
65689
23182
Error
%
2.53
0.57
0.74
0.01
0.32
0.36
0.39
0.27
0.01
Table 31 Comparison of Experiment Data and CFD Results for Test 28, 24-80 ppi
(m/s)
(Pa/m)
0.380
1379323
0.384
1382511
0.351
1172774
0.311
942852
0.262
629565
0.198
372652
0.131
174497
0.091
88998
0.036
19485
0.066
50886
90
COMSOL
Pressure Drop
(Pa/m)
1327725
1349328
1136270
898688
649598
382152
178352
91165
18744
51944
Error
%
3.74
2.40
3.11
4.68
3.18
2.55
2.21
2.43
3.80
2.08
Table 32 Comparison of Experiment Data and CFD Results for Test 23, 18-30 ppi
(m/s)
(Pa/m)
0.031
2723
0.074
11712
0.156
50272
0.242
116462
0.343
233955
0.445
386486
0.610
735607
COMSOL
Pressure Drop
(Pa/m)
2622
12253
49979
116984
231887
387781
722911
Error
%
3.74
4.62
0.58
0.45
0.88
0.34
1.73
(m/s)
(Pa/m)
0.408
501847
0.020
2179
0.093
28841
0.167
86720
0.249
187764
0.326
324686
0.396
485160
0.466
684740
0.511
822209
0.539
920699
COMSOL
Pressure Drop
(Pa/m)
519931
2110
30466
91433
197494
335006
488576
674114
809295
896728
91
Error
%
3.60
3.17
5.63
5.44
5.18
3.18
0.70
1.55
1.57
2.60
(m/s)
(Pa/m)
0.393
937588
0.102
69810
0.156
156059
0.214
286174
0.284
505716
0.333
685889
0.420
1085478
COMSOL
Pressure Drop
(Pa/m)
953351
70079
157832
290568
504841
687470
1086506
Error
%
1.68
0.39
1.14
1.54
0.17
0.23
0.09
(m/s)
(Pa/m)
0.295
948953
0.021
9916
0.058
46158
0.091
101936
0.137
217300
0.195
428441
0.236
621747
0.314
1088535
0.326
1181713
COMSOL
Pressure Drop
(Pa/m)
966568
9539
48690
106651
225474
437538
629214
1088436
1173106
92
Error
%
1.86
3.80
5.49
4.63
3.76
2.12
1.20
0.01
0.73

Liquid Permeability of Ceramic Foam Filters

Uploaded by

Copyright:

Available Formats

Liquid Permeability of Ceramic Foam Filters

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Liquid Permeability of Ceramic Foam Filters

Uploaded by

Copyright:

Available Formats

Liquid Permeability of Ceramic Foam

Light Metals Production

Norwegian University of Science and Technology

Liquid Permeability of Ceramic

Norwegian University of Science and Technology

Trondheim, July 2012

Trondheim, July 2012

4.3.7 Permeability Changing with Fluid Flow Direction ....................................................................... 53

Figure.1 CFF and filter bowl.[22]

Figure.3 Sivex CFFs a) 30 ppi, b) 40 ppi, c) 50 ppi, d) 80 ppi.

Figure.4 Electron Probe Micro Analysis(EPMA) of CFFs.

* Others are assumed to be mainly Phosphate (P2O5) binders

Figure.5 Definition of CFF: cell, window and strut diameters.[42]

Substituting Equation (10) into Equation (8) yields:

Pressure drop on foams correlation proposed

J.T. Richardsom et al.

9.73 102 d w0.743 (1 )0.0982

12.979[1 0.971(1 )0.5 ]

3.68 102 d w0.7523 (1 )0.07158

Dp=equivalent spherical diameter

M.D.M Innocentini et al.

100 865,0.65 2.6

2.3 Fluid Mechanics

2.3.2 Reynolds Number

Figure.6 Moody diagram.[56]

Figure.7Diagrammatic sketch of laminar flow and turbulent flow.[57]

2.3.3 Entrance Length

These details are shown in Figure.8.

Substitution of Equation (17) in Equation (21) finally gives,

For turbulent flow one commonly accepted correlation is[61]:

Figure.9 Determination of tortuosity: fluid path and geometrical length.[63]

Figure.10 Porous structures with different flow toruosities.[63]

Scheidegger[ 65 ] suggested that the tortuosity in sponge structures can be obtained by

Figure.12 Modern digital microscope-LEICA MEF4M.

3.1.2 Optical Microscopy - Experimental Procedure

3.2. Optical Scanning

Figure. 13 Epson Perfection V330 scanner.[68]

3.3. Energy-Dispersive X-ray Spectroscopy

3.3.2 Energy-Dispersive X-ray Spectroscopy - Sample Preparation

3.4 Pressure Drop Experiments

Kinematic Viscosity (m /s)

Figure.14 Photograph of experimental set-up constructed using DN 50 mm x 1.8 mm plastic pipe.

Figure.15 Close up of filter housing.

a) 49 mm(diameter) CFF sample. b) 49 mm CFF, Method 1. c) 49 mm CFF, Method 2.

Differential Pressure Transducer

2 mm thick rubber gasket

Figure.17 Filter housing for ~101 mm diameter CFFs.

Differential Pressure Transducer

Figure.18 Filter housing for ~49mm diameter CFFs.

3.5 Tortuosity Experiments

Figure.20 Mounted crucible.

Two Bimex 400 fiber

Ceramic Foam Filter

Type K Inconel sheathed

information are listed in Table IV.

Table IV Porosity and other relevant information

Volume (cm3) A.Weight (kg) T.Weight (kg)

4.2 Cell, Window & Strut Diameter

Table VI Information of window size, dw

Table VII Information of strut size, ds