Aerosol Science and Technology

ISSN: 0278-6826 (Print) 1521-7388 (Online) Journal homepage: http://www.tandfonline.com/loi/uast20

Optimization of Pleated Filter Designs Using a

Finite-Element Numerical Model

Da-Ren Chen , David Y. H. Pui & Benjamin Y. H. Liu

To cite this article: Da-Ren Chen , David Y. H. Pui & Benjamin Y. H. Liu (1995) Optimization of
Pleated Filter Designs Using a Finite-Element Numerical Model, Aerosol Science and Technology,
23:4, 579-590, DOI: 10.1080/02786829508965339

To link to this article: https://doi.org/10.1080/02786829508965339

Published online: 13 Jun 2007.

Submit your article to this journal

Article views: 1447

Citing articles: 41 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=uast20
Optimization of Pleated Filter Designs Using
a Finite-Element Numerical Model
Da-Ren Chen, David Y. H. Pui* and Benjamin Y. H. Liu
Particle Technology Laboratory, Mechanical Engineering Department,
University of Minnesota, Minneapolis, MN 55455

A numerical model has been developed to optimize the in the pleat spacings. Therefore, an optimal pleat count
design of pleated filter panels. In this model, the fluid for minimum pressure drop exists at a certain pleat
flow is modeled by a steady laminar flow and the filter height for each filter media type. The optimization of
media resistance is governed by the Darcy-Lapwood- rectangular pleated filters, e.g., mini-pleated filter pan-
Brinkman equation. A finite element method with a els, has been performed for six commercial filter me-
nine-node Lagrangian element is used to solve the dia. The optimal pleat count is shown to increase with
governing equations. For the rectangularly pleated fil- decreasing media permeability of the filter media. A
ter panel, the numerical results agree well with the generalized correlation curve has been found for the
analytical model of Yu and Goulding (1992) and with six filter media by using a nondimensional parameter
his experimental data. The pressure drop increases a t analysis. The results can be used to design pleated
small pleat count due to increased media face velocity, filter panels with minimum pressure drop.
and a t large pleat count due to increased viscous drag

INTRODUCTION The optimization parameters include

Pleated filter panels have been used in a pleat count, pleat height, pleat configura-
variety of industrial applications, includ- tion, and filter media characteristics,
ing air filtration in pharmaceutics and mi- among others. At a low pleat count, the
croelectronics cleanrooms, in computer- pressure drop across the filter media will
disk-drive micro-environments, and in in- increase due to the reduced filter media
let systems to the gas turbines. A pleated area and increased media face velocity. At
filter is more compact and has more fil- a high pleat count, the pressure drop will
tration area than a flat sheet media. With also increase due to the increased fluid
lower media face velocity due to increased viscous drag in the pleat spacings. There-
filtration area, the pleated filter has im- fore, an optimal pleat count can be found
proved particle collection efficiency and for a given pleat height, which gives the
reduced filter pressure drop, which can lowest pressure drop across the filter
result in an improved cleanliness level in panel.
a clean space while reducing operating There is limited information on the
costs. This paper provides a detailed para- pleated filter design in the literature.
metric study of the pleated filter design Raber (1990) developed a model using
using a numerical finite element method. similar approach as the heat exchanger
Optimization results are given for six design and neglecting the viscous effect
commercial filter media. for a moderate inlet velocity condition. In
his study, the pleated filter panel was as-
sumed to be a series of triangularly pleated
*Corresponding author. structure. Due to geometrical symmetry, a
Aerosol Science and Technology 23:579-590 (1995)
0 1995 American Association for Aerosol Rcsearch
Published by Elsevier Science Inc.
580 D.-R. Chen et al.

half-pleat could be selected for the analy- by a steady laminar flow, and the flow
sis. Finite element control volumes for passing through the filter media is mod-
upstream and downstream pleat spaces eled by the Darcy-Lapwood-Brinkman
were established by dividing the filter me- model. The governing equations are solved
dia into several equal surface areas, and by a numerical finite element method with
one-dimensional momentum and mass the nine-node Lagrangian element.
balances were used for each upstream and Therefore, the model includes all the ef-
downstream control volumes. Uniform ve- fects considered in Yu and Goulding's
locity profiles were assumed for the up- model, and other effects that were ne-
stream control volumes, and sinusoidal ve- glected. Since deformed finite elements
locity profiles for the downstream ones. are used to fit the curved boundaries or
Additionally, the flow passing through the interfaces, different configurations of pleat
filter media was assumed to be perpendic- designs can be analyzed using this method.
ular to the filter media surface and to
satisfy the classical Darcy's law. Due to
the simplified velocity profiles in the pleat NUMERICAL MODELING
spacings, the neglected viscous effect, and
the specific pleat configuration used, this Physical Model
model is not appropriate for use in our A typical mini-pleated filter panel is shown
work. Recently, Yu and Goulding (1992) in Fig. 1. Upstream inlet flow travels par-
has developed a semi-analytical method to allel to the pleated direction and then
solve this problem. In his method, the passes through the filter media to the
pleated filter panel was assumed to be a downstream region. The total pressure
series of channels and the flow field in the drop across the pleated filter is a result of
pleat space was simplified as channel flow flow contraction in the upstream region,
with suction or injection on the wall. By viscous drag between the fluid and porous
dividing the pleat height into finite ele- surface on both the upstream and down-
ments, it was assumed that the flow pass- stream channels, pressure drop across the
ing through the wall was uniform. The filter media, and flow expansion in the
pressure drop across the wall was first downstream region. Therefore, to set up
calculated and then the flow velocity at this problem, it is necessary to model the
the wall was calculated using the filter flow fields in the upstream and down-
media characteristics. For starting the cal- stream regions and in the filter media.
culation, the wall velocity at the first ele-
ment must be initially guessed. The con-
vergent solution would be obtained when
the total mass conservation was satisfied
by adjusting the wall velocity at the first
element. Although the method was suc-
cessfully used to analyze the mini-pleat
panel, it cannot be used to analyze the
general pleat panel. Further, the effects
of the developing flow, flow contraction,
flow expansion, and reduced permeability hot meit separators
at the corner of the pleat were neglected -.......
--.-.--.....
in the model.
In this study, the fluid flow is modeled FIGURE 1. Mini-pleated filter panel.
Optimization of Pleated Filter Designs

Governing Equations convective term into the classical Darcy

To simulate the fluid flow across the filter equation to remove the third limitation.
media, it is necessary to model the flow This modified equation is referred as the
field in several different domains. For the Darcy-Lapwood equation. In addition,
flow fields upstream and downstream of Katto and Masuoka (1967) modified the
the filter media, a steady laminar flow DL-equation by incorporating the viscous
model is used. The following assumptions force term to account for the fluid/por-
are made in this model: ous media interaction effect. The result is
known as the Darcy-Lapwood-Brinkman
1. Two-dimensional geometrical configu- equation. The derivation of this equation
ration. was given by Tam (1969) using the ensem-
2. Steady-state flow. ble average method or Rudraiah et al.
3. Incompressible flow. (1979) using the nonequilibrium thermo-
4. Isotropic, homogeneous, and saturated dynamic method. Therefore, in this study,
fluid. the flow field in the porous media is mod-
5. Uniform velocity profile at the far up- elled by DLK-equations and the continu-
stream. ity equation as follows:
The flow fields in the upstream and down- x-component:
stream domains are governed by the in-
compressible Navier-Stokes equations
and the continuity equation.
For modeling the fluid flow through the
filter media, the media is assumed to be a
porous media. The classical Darcy's equa-
tion can be used to calculate the flow field
in the porous media. However, the validity y-component:
of this equation is subject to the following
limitations:
1. The fluid is a homogeneous, or single-
phase fluid.
2. There is no interaction of the fluid and
the porous media.
3. The flow rate is kept relatively small.
4. In the case of gases at low pressures, Continuity equation:
there are deviations from the Darcy's
equation due to slip flow.
In this study, the first condition is satis-
fied and the fourth condition is not con- where
sidered. To remove the second and third
limitations, modifications to the Darcy's u = velocity in the x direction,
law are needed. According to the studies u = velocity in the y direction,
of Rudraiah and Srimani (1980), Rudraiah P = pressure,
and Rao (1982), Rudraiah and Friedrich p = gas viscosity
(1981) and Prasad et al. (19851, the Darcy's K = permeability of the porous media
law was modified by incorporating the E = porosity of the filter media
582 D.-R. Chen et al.

Finite Element Formulation Each coefficient in the matrix A has the

Various finite element methods for solv- form
ing the governing equations described
above have been proposed during the past
decade. These finite element formulations
can be classified into three classes based
on their approaches to treating the pres-
sure term in the governing equations:
Galerkin weighted residuals methods
[Taylor and Hughes 1980; Kim, 1987;
Zienkiewics et al. 19831, penalty methods where
(Zienkiewics et al. 1983; Engelman et al.
1982: Kikuchi et al. 1984), and velocity-
pressure segregated methods (Comini and
cl1= 7
E (
&.NkukU- + NiNkuk-
dx dy
dq. dN,

I
Giudice 1982; Benim and Zinser 1986;
Rice and Schnipke 1986).
The Galerkin weighted residuals meth-
ods do not require any approximation at
P
+ -NiNj7
dN;.
dx dx
dN. diy dN.
- - - L + - J .
dy dy I
K
different equation levels. However, sim-
plified pressure and/or pressure correc- dMj dN,
C,, = Ni- C,, =0 C,, = M,-
tion equations are used in the velocity- dx dx
pressure segregated methods, and the pe- dA$
nalized conservation of mass equation is C,, =0 C23 =M.- C,, =0
used in the penalty methods. Conceptu- dy
ally, the Galerkin weighted residuals
methods could satisfy the conservation of C,, = A$-
dMj
C,, =0
D =
" an:
-4.-
JY E dn
mass most rigorously. Since the pressure
distribution in the flow field is determined where u and u are the initial guesses or
by the conservation of mass equation for the values of u and u in the previous
the incompressible flow fields and, for this iteration. It is assumed to be constant
study, the key point is to find the pressure when formulating the matrices. The sur-
drop across the filter media, the Galerkin face integral term in the right hand side
weighted residuals formulation is used. In of A,, equation corresponds to that part
the Galerkin formulation, the governing of the element on an interface where the
equations are weighted with appropriate continuities of velocity and pressure ap-
discrete weighting functions and the re- ply. The matrix F is written as,
sulting equations integrated over the re-
gion of interest and equated to zero. By
using this method, the final results can be
written in the assembled matrix
where

where the chosen form for A is

)1:
(A;) = Pi
and S, is that part of the element on a
Optimization of Pleated Filter Designs

boundary where the natural boundary

conditions, i.e., du/dn and dv/dn, are
known.
In this formulation, N and M are the
interpolation functions for the velocity and
pressure fields, respectively. In addition,
this assembled matrix equation can be
easily switched to the incompressible fluid FIGURE 2. Nine-node Lagrangian element.
flow formulation by (1) setting 8 = 1, and
(2) dropping the third term in C,,.
The inlet velocity profile is assumed to ing functions for velocity and pressure
be uniform. Symmetrical boundary condi- variables. For error consistency reasons,
tions are imposed on the top and bottom the interpolating functions for the pres-
of the computational domain. For the out- sure should be a polynomial of lower de-
let boundary condition, only the pressure gree than that for velocity. This fact was
at one point on the outlet boundary is observed by several workers (Gartling and
specified as zero for reference and no Becker 1976; Nickel1 et a1. 1974) who se-
velocity profile is assumed. In the calcula- lected quadratic interpolating shape func-
tion, the outlet velocity profile is checked tions for velocity and linear shape func-
and the computational domain is ex- tions for pressure. Up to now, different
tended if the outlet velocity profile is not mixed-interpolation finite elements are
uniform. Along all the interfaces, the ve- commonly used. According to the study by
locity and pressure fields inside and out- Huyakorn et al. (1978), four types of mixed
side the filter medium are assumed to be interpolation finite elements in the
continuous. Galerkin formulation of incompressible
The set of nonlinear simultaneous viscous flows have been compared. The
equations described above is solved by a results indicate that, for similar grid ar-
simple iterative process. The procedures rangements, nine-node Lagrangian ele-
can be summarized as follows: ments give the most accurate pressure
and velocity distributions, and that eight-
Assume initial values of the primitive node serendipity elements give the least
variables denoted by u and u. accurate pressure distribution. Therefore,
Solve for the updated values u, u, and in this study, the nine-node Lagrangian
P. element shown in Fig. 2 is used.
Evaluate the relative error based on
the updated values at all points. If these
are within a specified tolerance, the Computational Domain and Mesh
calculation is completed. Distribution
If the convergence criterion is not sat- Although the finite element method can
isfied, then update u, u, and P ,to the analyze any pleat shapes, it is better to
new values and return to step 2. start with a simplified configuration in the
first stage of this study. As shown in Fig.
1, the pleat configuration of the mini-
Element Selection pleated panel can be simplified as a series
In dealing with incompressible flow prob- of rectangular pleats. Due to geometrical
lems with the Galerkin formulation, an symmetry, the computational domain used
important fact that has to be recognized is in this work can be simplified as shown in
the need to select appropriate interpolat- Fig. 3. The upstream calculation domain
 D.-R. Chen et al.

Filter media

Upstream channel h t

Downstream
channel

Inlet

W : Pleat Spacing
L : Pleat Height
t : Filter Media Thickness
h : Channel Half-width
FIGURE 3. A typical computational domain in this study.

is about 10-18 times the pleat spacing media modelled by Yu and Goulding
and the downstream domain is about (1992) and for which experimental data
10-15 times the pleat spacing. The num- are available for comparison with the the-
ber of elements used in the calculation is oretical results. Once the model is vali-
approximately 300-450. A typical mesh dated by comparing with the experimental
distribution is shown in Fig. 4. Finer ele- data, the calculations are then extended
ments are seen to distribute in the regions to six Lydall commercial filter media.
near the fluid/porous media interfaces.

RESULTS AND DISCUSSION Test Case with 95% DOP Filter Media
Initial calculations are performed for the Velocity Field at 100 fprn Panel Flow
95% DOP filter media. This is the filter Condition. A typical velocity field is shown

Filter Media

FIGURE 4. A typical mesh distribution in this study. (Only the

main part of computational domain is plotted).
Optimization of Pleated Filter Designs 585

in Fig. 5. The inlet panel flow is set at 100 will decrease with decreasing pleat spac-
fprn with the media face velocity being ing or increasing pleat height. The flow
approximately 16% of the panel flow ve- profile in the downstream pleat channel
locity. The flow velocity near the upper becomes fully developed after a short dis-
computation domain is seen to increase as tance and continues to increase due to
the fluid flow approaches the filter media. the mass injected from the interface. Upon
Flow contraction in the upstream region leaving the filter panel, the flow velocity
is observed when it reaches the pleated profile recovers to the uniform condition
filter region. Most of the flow will enter in a distance approximately 8-10 times
the upstream pleat channel while a small the pleat spacing. The flow expansion is
fraction will pass through the front end of also observed.
the pleated filter directly. For the flow in
the upstream pleat channel, it becomes Pressure Drop Optimization. In Fig. 6,
fully developed in a very short distance total pressure drop across the pleated fil-
(about 1-2 channel spacing). The velocity ter panel, A P , is shown as a function of
at the interface is almost zero in the hori- pleat count and pleat height for the 95%
zontal direction, i.e., the flow direction at DOP filter media (Lydall Grade no. 221).
the interface is almost normal to the filter Including in the Fig. 6 are the analytical
surface. This observation confirms the as- results of Yu and Goulding (1992). The
sumption used by Yu and Goulding al- numerical results are shown to agree well
though horizontal interface velocities will with those of Yu and Goulding's. At a
become more significant at a smaller pleat smaller pleat count, the pressure drop
spacing. Upon exiting the filter media, the predicted by the numerical model is
flow will then enter the downstream pleat slightly higher than that predicted by the
channel. There is a small fraction of the Yu and Goulding's model. This difference
flow that will pass through the back end is expected as Yu and Goulding's model
of the pleated filter, the amount of which neglected the pressure drops due to flow

FIGURE 5. U-component velocity profile for the pleated filter.

 D.-R. Chen et al.

' ' ' ' L ' ' ' ~ " " ' " " ' " "
LJd.U Fllter Medium
the pleat height causes the optimal pleat
(Grade No. 221) count to decrease with a corresponding
decrease in pressure drop. It is due to the
decrease in media face velocity with a
resulting decrease in the media pressure
drop.

Linear Relationship Between Pressure

Drop and Inlet Panel Velocity. Three uni-
form inlet panel velocities are investigated
in this study: 25 fpm, 100 fpm and 250
fpm. The pleat counts are 8 and 10
- -* - -Model of Yu and Godding (1992) pleats/inch and pleat height 0.875 in with
0.0 ' " " " ~ " " - " " " ' " "
0 5 10 15 20 25 the Lydall filter media (Grade no. 221).
Pleat Count/ in.
Figure 7 shows that the numerical results
FIGURE 6. Comparison of the present model agree well with the experimental data
with the model of Yu and Goulding (1992) for given by Yu and Goulding (1992). A lin-
mini-pleated filters with two different pleat ear relationship is observed between the
heights (100 fprn). total pressure drop and the inlet panel
velocity. Although a nonlinear term is in-
volved in the numerical model, its effect is
contraction, developing flow and flow ex- negligible.
pansion. At a higher pleat count, the pre-
dicted pressure drop is slightly lower than
that from Yu's model. This results from Optimization of Six Commercial
the fact that in our model the tangential Filter Media
velocity along filter surfaces has some fi- Six different commercial filter media have
nite values instead of zero assumed in been modeled. They include Lydall filter
Yu's model. For a certain pleat height,
there is an optimal pleat count corre-
sponding to the minimum pressure drop.
At a lower pleat count (or smaller filter
area), the filter media face velocity will be
higher, resulting in a higher pressure 'drop.
At a higher pleat count (or larger filter
-
95%DOP Filter Media
(Grade no. 2211
POI0.875 in. Pleat Height
/
area), the pressure drop due to the vis-
cous drag in the pleat spacing becomes
more significant, resulting in a higher
pressure drop. An optimal pleat count
therefore occurs when the combination of
viscous drag and media resistance is at a
minimum. The numerical results are
shown to give a slightly higher optimal
pleat count than the analytical results. It 10
- 100
is due to the effect of nonzero tangential Inlet Velocity, fprn

velocities at the interfaces which makes FIGURE 7. Pressure drop comparisons between
the viscosity effect lower than that given the present model results and the experimental
by Yu and Goulding's model. Increasing data of Yu and Goulding (1992).
Optimization of Pleated Filter Designs

media Grade no. 252 (ULPA 99.99% DOP AP, across the filter media can be ob-
media), Grade no. 213 (HEPA 99.985% tained as
DOP media), Grade no. 233 (98.5% DOP
media), Grade no. 220 (95% DOP media),
Grade no. 224 (90-95% A S H W media),
and Grade no. 229 (80-90% A S H W and
media). Table 1 lists the characteristics of Uc L-2t (W-t)
the Lydall filter media evaluated. The uni- - - -- h=
form panel flow velocity is set at 100 fpm. urn h 7 2
The results are shown in Fig. 8. The re- Therefore
sults show that, for a given pleat height, APc 8 ( L - 2t)(L - t)
the optimal pleat count increases if the ---- N -
permeability of the filter media decreases. APm K (w-t13
This is due to the decrease of the ratio of Finally, if L >> t, then
viscous to media pressure drop when the
filter media permeability decreases. The
optimal pleat count will then occur at the
condition when the viscous force becomes where
important relative to the media pressure
drop. L = Pleat height
W = Pleat space
h = Channel half-width
t = Filter media thickness
K = Permeability of the filter media per
NONDIMENSIONAL ANALYSIS unit media thickness per unit vis-
cosity
Nondimensional Pleating Parameter
K = Constant of the filter media, de-
As shown in Figure 3, the flow in the fined as the pressure drop per unit
upstream and downstream pleat spacing media face velocity, t / ~ .
can be considered as a channel flow. As-
suming that the characteristic velocity in
the pleat spacing is the mean velocity U, Replotting with the Derived Parameter
in the channel and the characteristic ve- Figure 8 is subsequently replotted using
locity across the filter media is the mean the nondimensional parameter as the ab-
medium face velocity, Urn,the ratio of the scissa and normalizing the total pressure
pressure drop AP, due to the viscous ef- drop across the pleated filter panel, AP,
fect in the channels to the pressure drop with the pressure drop across the filter

TABLE 1. Characteristics of the Lydall Filter Media Evaluated

Grade no. DOP Efficiency Permea, ~ ( c m ' ) Thickness, t b m ) Basic ~ e i ~ h t ( ~ / r n ' )
252 99.99% ULPA 7.25e-9 0.38 73
213 99.985% HEPA 1.03e-8 0.38 73
233 98.5% 2.26e-8 0.38 73
220 95% 3.20e-8 0.38 73
224 90-95% ASHRAE 7.68e-8 0.38 73
229 80-90% ASHRAE 1.10e-7 0.38 73
*Note that K (usually used in filtration industry) = t / ~
 D.-R. Chen et al.

q?..
' +.., /
'.*%.

--.a
....--
ra - Pleat Hdgbt

..-.---
-....
0.815 In
1.750 La.
3500La.
Pleat Helght
-om
--.o--
La.
1.750 tn

-.r..-&

0.001 .............................. ..
0 . ~ 1 ...........................f
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Pleatcouot / in. Pknt wunt / in.

I \ DIEPA 99.-
all Nter Media
mde no. 213
FUter mdh)
t
b;
.
,?.. .-. .-.*,
'.,'\...h . 9 .Y '
J

-m L
Pleat Hctght
..-*-- a . La.
1.750
---35001n.
-
.
-.I.- 5.250 Ln

0.001 ....,.... .... .... .... ....I

o 5 ib 15 2b 25 3b
Pkat count / in. P l a t count / in.

-Pleat Height
0675 in
-..*--1.750 in
-..- 3500 in.
- . " . - 5.250 In.
o . w d . . . . , . . . .................... 4
0 5 10 15 20 25 30
P l a t a n t / in.

FIGURE 8. Initial pressure drop as a function of pleat count

and height for six filter media at approaching velocity 100 fpm.
Optimization of Pleated Filter Designs 589

count in previous studies, such as flow

LydaU Filter Media contraction, flow developing and flow ex-
o Grade no. 213
0 Grade no. 220 pansion. The model can also allow us to
Grade no. 224
x Grade no. 252 investigate the local average flow field in
o Grade no. 233
a Grade no. 229 the filter media. Since the model uses a
.. rigorous finite element method with the
nine-node Lagrangian element, it can
handle many pleat shapes and other vari-
ables, including variable permeabilities in
the filter media.
For an initial test case with a 95%
DOP filter media, the numerical results
20.0 ' , '"'. " . , , ..,.'' , ' , ,,"" , , , ,,,,,' ' , ' ,,""
are shown to agree well with Yu and
@I Logarithmic Plot Goulding's analytical model and experi-
Lvdall Filter Media
Grade no. 213 mental data. An optimal pleat count cor-
Grade no. 220
Grade no. 224 responding to the minimum pressure drop
Grade no. 252
Grade no. 233 can be predicted for a given pleat height
Grade no. 229
and media characteristics. Six commercial
filter media have been evaluated by the
present model and their pressure vs. pleat
count curves obtained. Using a nondimen-
sional analysis, a general correlation curve
0.0 1 .......... m... ....................... has been found for all the six commercial
lo4 16.~ 16-1 ldO 1;fi
(8/KllLZ/~-t131 media evaluated. This correlation curve
FIGURE 9. Combined plot for the normalized can be used to design pleated filter panels
pressure drop vs. nondimensional pleating pa- with minimum pressure drop.
rameter: (a) linear plot to show the straight line, Future work will include (1) the study
(b) logarithmic plot to show the details of low of the triangularly pleated filter design,
value data. (2) the analysis and design of the cylindri-
cal cartridge filter, and (3) the experimen-
tal validation of the numerical model.
media, A Pm(= KpUm). All the six cases
are plotted together in Fig. 9. It is seen This research is supported by a grant from the Center
that a single general correlation curve is for Filtration Research at the University of Minnesota.
found for all the six filter media investi- Center members include 3M Company, Donaldson
gated. The correlation curve is tested to Company, Inc., Fleetguard, Inc., Honeywell and TSI
Inc.
be a straight line with a correlation coef-
ficient of 0.9995. This general correlation
curve can be used for the design of panel
filters with the rectangular pleats. REFERENCES
Benim, A. C, and Zinser, W. (1986). Cornput. Methods
Appl. Mech. Eng., 57:223.
SUMMARY AND FUTURE WORK Comini, G. and Giudice, S. D. (1982). Nurner. Heat
A numerical finite-element model has Transf. 5:463.
been successfully developed for calculat- Engelman, M. S., Sani, R. L., Greso, P. M. and
Bercovier, P. (1982). Znt. J . Nurner. Methods Fluids,
ing the total pressure drop across pleated 2:25.
filter panels. The model includes all the Gartling, D. K., and Becker, E. B. (1976). Cornput.
effects that have not been taken into ac- Methods Appl. Mech. Eng. 8:51.
590 D.-R. Chen et al.

Huyakorn, P. S., Taylor, C., Lee, R. L., and Gresho, Porous Media, PGSAM-2, Bangalore University,
P. M. (1978). Computers Fluid, 6:25. India.
Katto, Y., and Masuoka, T. (1967). Int. J. Heat Mass Rudraiah, N. and Friedrich, R. (1981). Proc. IV, GAMM
Transf. 10:297. Conference on Numerical Methods in Fluid Me-
Kim, S. W. (1987). NASA CR-179135. chanics, Paris, October 7-9, 1981.
Kikuchi, N., Oden, J. T., and Song, Y. J. (1984). Finite Rudraiah, N. and Srimani, P. K. (1980). Proc. R . Soc.,
Elements in Fluids, (R. H . Gallagher et al., Eds.), Lond. A 373:199.
Wiley, New York, Vol. 5, p. 107. Tam, C. K. W. (1969). J. Fluid Mech. 38537.
Nickell, R. E., Tanner, R. I. and Caswell, B. (1974). J. Taylor, C. and Hughes, T. G. (1980). Finite Element
Fluid Mech., 65:189. Programming of the Nauier-Stokes Equation, Pine-
Prasad, V., Kulacki, F. A. and Keyhani, M. (1985). 1. ridge Press, Swansea.
Fluid Mech. 150:89. Yu, H. H. S. and Goulding, C. H. (1992). Optimized
Raber, R. R. (1990). World Filtration Congress 111, ultra high efficiency filter for high-efficiency indus-
p. 52. trial combustion turbines, ASME Gas Turbine
Congress and Exposition.
Rice, J. C. and Schnipke, R. J. (1986). Comput. Methods
Appl. Mech. Eng. 58:135. Zienkiewics, 0 . C., Taylor, R. L. and Baynham,
J. M. W. (1983). in Hybrid and Mixed Finite Element
Rudraiah, N . and Balachandra Rao, S. (1982). Appl. Methods, (S. N . Atluri et al., Eds.), Wiley, New
Sci. Res. 39:21. York.
Rudraiah, N., Chandrasekhara, B. C., Veerabhadraiah,
R. and Nagraj, S. T. (1976). Some Flow Problems in Received January 3, 1995; revised April 11, 1995