See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/272390393

A high-precision method for calculating the pressure drop across wire mesh
filters

Article  in  Chemical Engineering Science · May 2015

DOI: 10.1016/j.ces.2015.01.023

CITATIONS READS

15 7,337

3 authors, including:

Yigang Luan
Harbin Engineering University
38 PUBLICATIONS   141 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Yigang Luan on 10 May 2018.

The user has requested enhancement of the downloaded file.

 Chemical Engineering Science 127 (2015) 143–150

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: www.elsevier.com/locate/ces

A high-precision method for calculating the pressure drop

across wire mesh filters
Haiou Sun, Shi Bu, Yigang Luan n
College of Power and Energy Engineering, Harbin Engineering University, 145 Nantong Street, Harbin 150001, Heilongjiang Province, China

H I G H L I G H T S

Experiments are accomplished with wire mesh parameters precisely controlled.

A creative numerical method for wire mesh is proposed.

A new mathematical correlation is derived for wire mesh pressure drop.

art ic l e i nf o a b s t r a c t

Article history: Three structural parameters were defined based on the characteristics of wire mesh filters: wire
Received 7 October 2014 diameter, layer spacing and mesh size. The pressure drops across two different mesh pads were
Received in revised form measured under different superficial air velocities in a wind tunnel. A new geometrical simplification
12 January 2015
method for numerical simulations was proposed. Using the large amount of calculated pressure drop
Accepted 13 January 2015
data, a new expression was derived for predicting pressure drops across mesh pads. Good agreement
Available online 22 January 2015
was achieved among the experimental data, simulated data and expression data which demonstrated
Keywords: that the derived expression can accurately predict pressure drops.
Wire mesh & 2015 Elsevier Ltd. All rights reserved.
Pressure drop
Experiment
Numerical simulation
Expression

1. Introduction be composed of glass and plastic fibers, such as polypropylene,

Wire mesh filters have many applications in industrial and
dynamic systems, such as gas processing facilities (Al-Dughaither
et al., 2010), multi-stage flash desalination plants (Janajreh et al.,
2013), nuclear electric plants and offshore intake-air filtration
systems (Brekke et al., 2009, 2010). These filters are generally
installed in the form of multi-layers for steam purification and for
the removal of liquid droplets or aerosols from airflows. Compared
with other separation devices, such as wave-plate separators or
cyclones, wire mesh filters provide better separation efficiencies
down to sizes of 2 μm to 10 μm; therefore, they can also be used as
coalescers or pre-filters positioned upstream of other separation
devices (Setekleiv and Hallvard, 2012).
Wire meshes can be shaped as either cylinders or as multi-layer
pads with certain thicknesses to suit the actual spatial pattern of
the installation site. In addition to stainless steel, wire meshes can
n
Corresponding author. Mobile.: þ 86 18646306466.
E-mail address: shandong-313@163.com (Y. Luan).
Dacron and Teflon. The wire mesh material primarily depends on
the operating environment, including the gas temperature, caus-
ticity and liquid loading (Brunazzi and Paglianti, 1998, 2000).
Despite these factors, the knit forms of wire meshes are approxi-
mately the same, which makes it possible to adopt some analo-
gous approaches during investigations.
The three main separation mechanisms are inertial capture,
direct interception and droplet diffusion; among these, inertial
capture is the prominent separation mechanism (Holmes and
Chen, 1984). When a gas stream that contains droplets moves
through wires, the inertia of the droplets causes their directions to
change from the streamline and impact the target wire (Brunazzi
and Paglianti, 2001). Therefore, it can easily be deduced that the
separation efficiency has a strong relationship with the structural
features of the mesh pads, the droplet size distribution, the gas
velocity and the liquid loading (Setekleiv and Hallvard, 2012;
Setekleiv et al., 2010). In practice, when designing a wire mesh
filter, a good balance must be achieved between the pressure
drop and the removal efficiency. Knowing the effects of various
factors on the mesh pad performance will be helpful in the design

http://dx.doi.org/10.1016/j.ces.2015.01.023
0009-2509/& 2015 Elsevier Ltd. All rights reserved.
144 H. Sun et al. / Chemical Engineering Science 127 (2015) 143–150

process. This study attempts to explore the structural and aero- mesh test section instead of circular tubes because the investigated
dynamic parameters that affect the pressure drop across wire-mesh wire mesh filters were installed in square ducts. For actual applica-
filters and to develop a method for predicting pressure drops. tions, this type of wire mesh filter usually works at a velocity of
Because numerical simulations have shortcomings and diffi- nearly 10 m/s; thus, a slightly wider velocity range was chosen in
culties, experimental and theoretical analyses are generally used this study.
to investigate the removal efficiencies and pressure drops of wire
mesh filters. Some methods for predicting the separation efficien-
2.1. Wire mesh section
cies of mesh pads were derived based on the droplet collection
efficiency of a single wire, which was related to the Stokes number,
Three structural parameters of wire mesh filters were defined:
Stk. However, in most cases, certain assumptions must be made
wire diameter (d), layer spacing (σ) and the mesh size, including
(Brunazzi and Paglianti, 2000, 2001; Feord et al., 1993; Langmuir
the length (h) and width (w), as shown in Fig. 2.
and Blodgett, 1944–1945; Carpenter and Othmer, 1955; Bradie and
To perform the experiment as accurately as possible, the above
Dickson, 1969). With the advancements in measurement technol-
structural parameters must be precisely adjusted. The wire mesh
ogies, researchers have been able to experimentally investigate the
used for the experiment had a wire diameter of 0.08 mm. Because
relationship between separation efficiency and some macro-
wire mesh pads are easily deformed, the other two parameters were
structural parameters (pad thicknesses, densities, specific surface
controlled using a statistical method. The cross-sectional area of the
areas, porosities, etc.). Some empirical and semi-empirical formu-
tested wire mesh section was 200 mm  200 mm. The required
las were obtained based on large amounts of measured data
mesh length h and width w can be obtained by controlling the
(Al-Dughaither et al., 2010; EI-Dessouky, 2000). One typical
mesh number in both the vertical and horizontal directions. Alth-
theoretical tool for studying pressure drops across mesh pads is
ough there must be shape and size nuances among individual
the famous Darcy equation, which has long been used as a basic
meshes, the through capacities of each layer are approximately
method in the field of porous media. For example, when conduct-
identical because they have the same tensity. Plastic plates with
ing numerical simulations of flow fields inside devices that contain
fixed thicknesses were placed between successive layers; therefore,
wire meshes, regions of the mesh pads are typically treated as
the average layer spacing could be adjusted by changing the number
porous media with permeabilities and drag coefficients that are
of plates and the total thickness of the mesh pad. Fig. 3 shows the
determined using experimental data (Janajreh et al., 2013; Rahimi
tested wire mesh pad, which was pressed with stainless steel plates
and Abbaspour, 2008). The Darcy equation contains linear and
and bolted. Twelve sticks were welded perpendicular to one of the
quadratic terms for velocity, which are also known as the viscous
plates, with each stick passing through a specific mesh to obtain the
and inertia terms. However, this binomial derived from regression
requested mesh tension. The geometric parameters of the tested
analysis has limitations because it is only applicable to a specific
wire mesh are listed in Table 1.
porous media structure (Helsør and Svendsen, 2007). Based on
a literature review, most experimental studies provide macrosco-
pic parameters of wire mesh such as thickness, density, specific 2.2. Measurement principle
surface area and porosity. The only microscope parameter of wire
mesh included in these studies is wire diameter. In this study, the The experiments were performed at room temperature, which
relationships between the pressure drop across the wire mesh and was measured using an anemometer. The anemometer was also
the following three microscope parameters are established: wire
diameter, layer spacing and mesh size.
According to the majority of previous investigations, when the
liquid loading is low, the effect of droplets on the continuous gas
phase can be neglected because neither flooding nor re-entra-
inment will occur. Therefore, the pressure drops of a single-phase
flow and diluted gas–liquid flow should be equal. For the wire
mesh studied here, the maximum volume fraction of the liquid
phase is less than 3.67e 6% (30 ppm); thus, experiments and sim-
ulations were performed based on single-phase airflow.

Fig. 1. Layout of the experimental apparatus.

2. Experimental apparatus

The experimental setup was designed and constructed at the

Department of Power and Energy Engineering of Harbin Engineer-
ing University. The horizontal wind tunnel contains two round
tubes and a square tube connected using a transition section, which
changes their shape gradually from the round end to the square end
to avoid flow distortion and to minimize measurement error. A
lemniscate flow tube was installed at the inlet part of the tunnel to
indirectly measure the gas velocity. The tested wire meshes were
installed inside the square tube of the tunnel, and adhesive tapes
were used to seal the tube to prevent gas leakage. Each part of the
wind tunnel was connected and fixed using bolts. The tunnel wall
was composed of organic glass. The gas flow was driven by a
centrifugal fan connected to an electric motor positioned at the end
of the tunnel. The superficial gas velocity was maintained within a
range of 2–8 m/s and was adjusted by changing the output power of
the centrifugal fan (Fig. 1). The square tubes were used for the wire Fig. 2. Definition of the structural parameters.
 H. Sun et al. / Chemical Engineering Science 127 (2015) 143–150 145

The converted pressure drop across the wire mesh ΔP (Pa)

related to Pr and Tr can be calculated using:
ρa p T
ΔP ¼ ΔP ¼ a r ΔP ð3Þ
ρr e pr T a e
where ΔPe is the measured pressure drop under local pressure Pa
and temperature Ta.

3. Numerical simulation

3.1. CFD simulation setup

The commercial software FLUENT, which is based on the finite

volume method, was used for the flow field simulation of the
mesh pads. Because the Mach number of the gas velocity in the
mesh pads is not greater than 0.05, the flow could be regarded as
incompressible. The two-equation standard k–ε model with the
standard wall function was adopted as the turbulence model to
close the governing equation set of hydromechanics. The yþ value
near the surface of the wires was approximately 40. When a
pressure-based separated solver was applied, various discretiza-
tion schemes were provided for the governing equations. To
Fig. 3. Tested wire mesh pad. minimize truncation error, a second-order scheme was selected
for the pressure interpolation and a second-order upwind scheme
Table 1 selected for the spatial discretization of momentum, volume
Geometrical properties of the tested wire mesh. fraction, turbulent kinetics and turbulent dissipation equations.
The SIMPLE algorithm was used to solve the pressure–velocity
Mesh pads d (mm) σ (mm) h (mm) w (mm) n
coupling. The velocity and pressure were chosen as the inlet and
A 0.08 0.60 4.74 3.28 80 outlet boundary conditions of the computational domain, respec-
B 0.08 0.42 4.74 3.28 80 tively. During the simulations, the operating conditions were set at
1 atm and 300 K.
used to measure the gauge static pressures. Because there were slight
fluctuations in the real-time atmospheric pressure, a mercury barom- 3.2. Geometrical simplification
eter was used to measure the local atmospheric pressure before every
set of pressure drop data was collected. This procedure is necessary The numerical investigation of mesh mist eliminators has long
because the later numerical simulations were all based on the ass- been a difficult task because of the following reasons: (a) the
umption of standard pressure and temperature (Pr ¼101,325 Pa, Tr ¼ geometrical model is difficult to establish. Although all of the
300 K) and the experimental data must be converted into the corr- meshes have a similar shape during installation and operation, the
esponding ones at standard conditions before being compared with actual size of the mesh may be completely different due to the
the simulated data. Because the tubes both upstream and down- flexibility of the mesh pads. It is impossible to draw all of the
stream of the wire mesh have equal cross-sectional areas, the inco- meshes based on their actual shapes and size. (b) Because the wire
mpressible air velocity is constant, which means that the dynamic diameter is considerably smaller than the mesh size and layer
pressures up- and -downstream of the wire mesh are equal. In this space, the establishment of the grid should be a problem because
way, the total pressure drop can be obtained by measuring the static the minimum interval size must be substantially smaller than the
pressure on the tubes located upstream and downstream of the wire wire diameter to ideally reveal the flowing features, which con-
mesh. The discharge coefficient of the lemniscate is 1. The diameter of sequently results in a large total grid number. (c) Generating a hex
the lemniscate was 120 mm, and it possessed advantages of a small grid cell for the complex geometrical structure of the mesh is time
resistance, rectifier capacity, constant pressure and equal variations in consuming. To overcome the above problems, the geometrical
the dynamic and static pressures. model must be simplified to a certain degree, and the rationality of
Based on the local pressure (Pa), the temperature (Ta), flow tube the simplification should be verified.
differential pressure (ΔPV), and the cross-sectional areas of the
flow tube (S) and the tested wire mesh (A), the superficial gas 3.2.1. Three dimensional original model
velocity (V) could be calculated using The original three-dimensional model was established using
sffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the Pro/Engineer software based on the actual shape of the knit
S 2ΔP V S 2ΔP V  R  T a mesh. The statistical average values of the wire diameter and layer
V¼ ¼ ð1Þ
A ρa A pa space were 0.08 mm and 0.42 mm, respectively. The length and
breadth of the mesh were 3.28 mm and 4.74 mm, respectively. All
where R is the molar gas constant (R¼8.3145 J/mol/K) and ρa is of the structural parameters were in agreement with those of the
the local gas density. experimental wire mesh. The wires from the different layers were
The converted gas velocity Vr related to reference pressure Pr placed randomly. Considering the similarity in mesh shape, the
and temperature Tr can be calculated using overall computational domain was composed of mesh cell units. A
ρa p Tr periodic boundary condition was introduced to save computa-
Vr ¼ V¼ a V ð2Þ tional resources, and only one mesh array was considered, as
ρr pr T a
shown in Fig. 4. Both 80-layer and 20-layer domains with the same
where ρr is the reference gas density. structural parameters were established to investigate the effects of
146 H. Sun et al. / Chemical Engineering Science 127 (2015) 143–150

the layer number on the calculation results. Extensions were of layers, and using 20 layers is appropriate for investigating the
added to the inlet and outlet of the computational domain to resistance characteristics. The effects of layer number on the pressure
stabilize the flow field. drop of wire mesh depend on the interactions between the layers.
The meshes were generated in the Gambit software. A grid sens- The numerical results indicate that the energy dissipation of each
itivity analysis was performed to assess the grid solution conver- layer is equal to the range of the layer spacing in this investigation,
gence and its solution dependency. Three types of tetrahedral unstr- which indicates that the pressure drop should increase linearly with
uctured mesh models were constructed for the 20-layer domain the layer number. To demonstrate the rationality of the method for
with grids number of 2310,000, 3690,000 and 5930,000. The result establishing the geometric model, the simulated data must be
indicated that the pressure drop increases with increasing grid compared with the experimentally obtained data. As indicated by
number under the same superficial gas velocity; however, because the resistance curves in Fig. 6, a good agreement is reached between
the difference between 3690,000 grids and 5930,000 grids is very the experimental and numerical results. Therefore, the assumptions
small, the former one was adopted to reduce computation time and of a basic mesh unit with the same size and shape and the intro-
the consumption of resources. Similarly, an equal grid scale was app- duction of periodic boundary conditions are appropriate.
lied for the 80-layer domain. Fig. 5 shows the local grid of the com-
putational domain, more intensive meshes were used near the sur-
3.2.2. Three-dimensional simplified model
face of the wire.
The three-dimensional original model revealed the actual stru-
The calculated total pressure drop was divided into single layer
cture of the mesh, but the computational domain and established
average values, which are given in Table 2. The relative errors of
grids are very complex. For numerical convenience, the original
the single layer average pressure drops between 20 layers and 80
three-dimensional model was simplified to a structure with inte-
layers are also listed.
rweaved orthogonal wires. The shape of each mesh was trans-
The maximum relative error is 1.59%, which may lead to the
formed to a rectangle whose length and width were equal to the
conclusion that the pressure drop increases linearly with the number
corresponding sizes in the original mesh, as shown in Fig. 7.
A grid sensitivity study was performed, and a grid number of
Table 2 3060,000 grids was selected. The simulated single layer average
Simulated pressure drop (three-dimensional original model).
pressure drop data based on the three-dimensional original model
V (m/s) 2 3 4 5 6 7 8 and simplified model and their relative errors are listed in Table 3.
As shown in Table 3, the maximum relative error between the
ΔP (Pa) n¼ 80 0.339 0.643 1.026 1.481 2.004 2.592 3.242 three-dimensional original model and the simplified model is 2.73%.
ΔP (Pa) n¼ 20 0.344 0.653 1.040 1.501 2.030 2.623 3.279
A comparison between the three-dimensional simplified model and
δ (%) 1.59 1.42 1.38 1.32 1.26 1.16 1.13
the experiment was performed, as shown in Fig. 8. The exact agre-
ement between the model and the experimental results demonstra-
tes the high precision of the three-dimensional simplified model.

Fig. 6. Comparison between experimental measurements and 3D original model.

Fig. 4. Mesh array in flow direction.

Fig. 5. Local grids in the computational domain. Fig. 7. Three-dimensional simplified model consisting of orthogonal wires.
 H. Sun et al. / Chemical Engineering Science 127 (2015) 143–150 147

3.2.3. Two-dimensional simplified model
The geometrical structure of the three-dimensional simplified
model might be relatively regular, but the overall grid number is
still large. To obtain a far more convenient computation, a two-
dimensional model was proposed. The model was altered to form
a structure composed of parallel wires, as shown in Fig. 9.
The equivalence principle could be stated as the “constant ratio
of the windward area of the wire to that of the entire mesh”; in
this way, the same flow capacity could be guaranteed. Fig. 10
shows the single basic mesh unit of the 3D and 2D simplified
models. The simplification method was illustrated more intuitively
by using Fig. 10 and the following equations.
The equivalent two-dimensional mesh size was calculated
using the following formula:
wh l
¼ ð4Þ
ðw þ dÞðh þ dÞ l þ d
Eq. (4) can be re-written as:
hw
l¼
h þ w þd
The Two-dimensional simplified model provides the most con-
venient method for numerically simulating the pressure drop across
the wire mesh. The 2D equivalent mesh size could be determined
(l¼1.92 mm) from the original mesh size of the experimental wire
Table 3
Simulated pressure drop (thee-dimensional original and simplified model).
V (m/s) 2 3 4 5 6 7
mesh (3D original model) (d¼0.08 mm, h¼4.74 mm, w¼3.28 mm).
The wires are represented by a series of circles. A pair of parallel
lines represents the periodic boundaries, and the distance between
the lines is the 2D mesh size l. The wire positions were randomly
arranged between the two lines, as shown in Fig. 11. Only one mesh
array was considered; therefore, the computational domain is a
rectangular area encircled by an inlet, outlet and the periodic
boundaries. By adopting the 2D model, the number of grids could
be significantly reduced. The final number of grids for the 20-layer
domain was only 40,000 after the grid sensitivity analysis. The
numerical results of the 2D model, the 3D original model and the
relative error between them are listed in Table 4.
In this study, the geometrical simplification of wire mesh was
necessary because the correlations derived in this article are based
on a large amount of pressure drop data (32 mesh with different
structural parameters and five superficial velocities for each mesh).
For a 24-core computer, approximately 2 h are needed to accom-
plish a 3D simulation for one wire mesh under one superficial
velocity. However, only approximately 7 min are necessary for the
2D model. Consequently, 300 h could be saved by using the 2D
ð5Þ
method to complete all of the computational work. Table 4 and
Fig. 12 show that the 2D model is very accurate.
3.3. Influence of the wire position randomness
During the construction of the geometrical model, the wires in
each layer were arranged randomly according to the actual condi-
tions. To improve the accuracy of the numerical results, the influence
8

ΔP (Pa) (original) 0.344 0.653 1.040 1.501 2.030 2.623 3.279

ΔP (Pa) (simplified) 0.350 0.661 1.055 1.528 2.085 2.679 3.361
δ(%) 1.60 1.23 1.39 1.83 2.71 2.15 2.50

Fig. 10. Single mesh unit of the 3D simplified model and the 2D model.

Fig. 8. Comparison between experimental measurements and 3D simplified model.

Fig. 11. Computational domain of the two-dimensional simplified model.

Table 4
Calculated pressure drop (three-dimensional original model and two-dimensional
model).

V (m/s) 2 3 4 5 6 7 8

ΔP (Pa) 3D 0.344 0.653 1.040 1.501 2.030 2.623 3.279

ΔP (Pa) 2D 0.345 0.653 1.043 1.506 2.050 2.653 3.332
δ (%) 0.29 0.08 0.24 0.33 0.99 1.16 1.62
Fig. 9. Two-dimensional simplified model composed of parallel wires.
148 H. Sun et al. / Chemical Engineering Science 127 (2015) 143–150
of randomness was investigated. Domains with different numbers of
layers were established based on several groups of random num-
bers. The results suggest that randomness has a greater influence on
the 3D simplified model than on the original model, which is quite
understandable because the original model better reveals the actual
shape of the wire mesh. Dimensionality reduction would increase
the influence of randomness; thus, the result from the 2D model is
the most unstable. The pressure drop simulated using the 2D model
tends to decrease rather than linearly increase with the number of
layers because, as the number of layers increases, the wires are more
likely in the flow-affected zone of the upstream wires; this phenom-
enon can be explained by the theory of flow around a circular
cylinder. The low velocity zone behind a wire leads to a relatively
low turbulence intensity caused by the downstream wire in this
zone, which results in a reduction in the energy dissipation. During
geometrical modeling, an improved method called “equal probabil-
ity” was adopted for the 2D simplified model, which means that
there is an equal distance between adjacent wires in the direction
perpendicular to the periodic boundary. In other words, the position
of each wire is determined by a group of arithmetic random
numbers. In this way, the wires in each layer stay away from the
flow-affected zone of the upstream layer such that the numerical
results of the 2D model could be more stable, and a strict linear
relationship between the pressure drop and the number of layers
could also be obtained.
4. Correlation with the numerical data
Based on the analysis of the above three geometric models,
the discrepancy in the numerical results for the pressure drop
between the three-dimensional original model and the two-dime-
nsional simplified model could be neglected. Because the two-
dimensional model is sufficiently accurate and has a significant
advantage in computational efficiency, this model should be the
best choice when implementing a parametrical study of the pres-
sure drop across wire meshes. Each of the remaining three para-
meters has five values: d (0.04 mm, 0.06 mm, 0.08 mm, 0.10 mm,
and 0.12 mm), σ (0.2 mm, 0.4 mm, 0.6 mm, 0.8 mm, and 1.0 mm)
and l (1.6 mm, 1.8 mm, 2.0 mm, 2.2 mm, and 2.4 mm). A reason-
able combination of parameters could improve the accuracy of the
fitted correlation.
As shown in Fig. 13, the pressure drop across mesh pads
significantly increases as the wire diameter increase. First, a slight
increase in the diameter will cause a substantial increase in the
specific area of the mesh pad, and the viscous effect of the flow field
becomes more apparent as the gas–solid interfacial area increases.
In addition, because a series of turbulent eddies exist behind each
wire, a thicker wire can produce more eddies and higher turbulent
pulse amplitudes, which increase the energy dissipation. A thin wire
Fig. 12. Comparison between the experimental measurements
simplified model.
can help improve the turbulence suppression ability of the mesh
pads. Assuming that the average gas velocity inside the mesh pads
is Vave ¼ V/ε and that the wire mesh hydraulic radius is Rh ¼ ε  d/4
(1 ε) (Helsør and Svendsen, 2007), the Reynolds number could be
calculated. The result shows that for the same gas velocity, the
Reynolds number increases as the wire diameter increase, from
which it could be inferred that the pressure drop is primarily caused
by inertia resistance.
Fig. 14 shows that the pressure drop decreases slightly as the
layer spacing increases. When the spacing extends to a certain
degree, the pressure drop remains almost constant, indicating that
blindly increasing the layer spacing cannot reduce the flow
resistance because the layer number is certain. Analysis of the
flow field shows that eddies formed in the recirculation zone
behind the wires interact with the downstream disturbance and
cause additional energy dissipation. This interaction becomes
weaker as the layer spacing increases, until it gradually disappears.
The pressure drop decreases as the mesh size becomes larger,
as shown in Fig. 15. The pressure drop will not change when the
mesh size exceeds a certain value. However, compared with the
layer spacing, the mesh size has a more pronounced effect on the
pressure drop. Based on flow field analysis, the local velocity in the
mesh increases as the pore size is reduced; thus, the turbulent
intensity will be enhanced behind the wires, which is the primary
reason for the increase in the pressure drop. Because the wires are
considerably smaller than the mesh, the interaction between the
flow disturbances caused by adjacent wires is extremely weak.
From the previous discussion, the pressure drop across a wire
mesh is influenced by the wire diameter, the layer spacing, the mesh
size and the superficial gas velocity. Taking the void fraction range
Fig. 13. Pressure drop against wire diameter (layer spacing σ¼ 0.6 mm, mesh size
l ¼2.0 mm, superficial gas velocity V ¼ 6 m/s).
and 2D Fig. 14. Pressure drop against layer spacing (wire diameter d¼ 0.08 mm, mesh size
l ¼2.0 mm, superficial gas velocity V ¼ 6 m/s).
 H. Sun et al. / Chemical Engineering Science 127 (2015) 143–150 149

into consideration, a non-dimensional multivariate correlation was

developed to predict the pressure drop across wire meshes via a
reasonable selection of structural parameters. The correlation is
obtained based on a collection of numerical data of two-dimen-
sional simplified geometrical models of a 20-layer computational
domain. The regression analysis provides the following correlation:

ρV 2
ΔP ¼ 10nRed 0:77 Reσ  0:09 Rel  1:03 ð6Þ
2
where n is the number of layers, ρ is the gas density, and V is the
superficial gas velocity. Red, Reσ and Rel are the Reynolds numbers
calculated using the wire diameter d, layer spacing σ and mesh size l,
respectively, as the characteristic length scale:
ρVd ρV σ ρVl
Red ¼ ; Reσ ¼ ; Rel ¼ ð7Þ
μ μ μ
where μ is the kinetic viscosity of the gas flow. The ranges of the
parameters are d (0.04–0.12 mm), σ (0.2–1.0 mm), l (1.6–2.4 mm) and
V (2–10 m/s)
To verify the accuracy of the correlation, the calculations were
compared with the experimental data. Fig. 16 shows the compar-
isons between the experimental data and the correlation calcula-
tions, and good agreement is obtained. The maximum relative
errors for the tested wire meshes A and B are 4.26% and 2.65%,
respectively. The correlation could be used as an effective tool for
evaluating pressure drops across wire mesh mist filters.

5. Conclusions

Three structural parameters (wire diameter, layer spacing, mesh
size and number of layers) were defined. The pressure drop across two
tested wire meshes with the same wire diameter, mesh size and num-
ber of layers but with different layer spacing (0.60 mm and 0.42 mm)
were measured in a wind tunnel. Experimental data were collected to
verify the numerical method and the resulting correlation.
The numerical results for the three-dimensional original model,
the simplified model and the two-dimensional model all reached
high levels of agreement with the experimental data. The results
for 20-layer and 80-layer domains imply that the pressure drop
linearly increases with the number of layers. For convenience and
to increase the efficiency of the numerical simulation, the original
geometrical model was simplified. The relative errors among the
computational results of the 3D original model, the 3D simplified
model and the 2D model were within the range of acceptable
error. Therefore, the 2D model is the best choice for numerically
studying the pressure drops across wire meshes.
The parametrical study based on the 2D model indicated that
the pressure drop increases as the wire diameter increases and
Fig. 16. Comparison between the experimental data and the calculations: (a) wire
mesh A; (b) wire mesh B.
decreases as the layer spacing and pore size increase. The influence
of the above three parameters on the pressure drop from the str-
ongest to the weakest are as follows: wire diameter, mesh size and
layer spacing.
Using the large amount of pressure drop data from the 2D model,
a correlation was proposed using regression analysis. To make this
mathematical expression non-dimensional, the Reynolds numbers,
which were calculated using the wire diameter d, layer spacing σ and
pore size l were defined as Red, Reσ and Rel. The calculations were
constant with the experimental data; therefore, the correlation could
be applied to predict the pressure drop when designing wire mesh
filters or when evaluating the performance of wire mesh filters.
Nomenclature

d wire diameter (mm)

l 2D mesh size (mm)
h mesh length (mm)
w mesh width (mm)
n number of layers
P reference pressure (Pa)
T reference temperature (K)
ΔPV flow tube differential pressure (Pa)
S cross-sectional area of the flow tube (m2)
A cross-sectional area of the tested wire mesh (m2)
V superficial gas velocity (m/s)
Vr converted gas velocity (m/s)
ΔP pressure drop (Pa)
Fig. 15. Pressure drop against mesh size (wire diameter d ¼ 0.08 mm, layer spacing k turbulent kinetic
σ ¼0.6 mm, superficial gas velocity V ¼ 6 m/s). Re Reynolds number
150 H. Sun et al. / Chemical Engineering Science 127 (2015) 143–150

Greek symbols
σ layer spacing (mm)
ρ gas density (kg/m3)
μ kinetic viscosity of the gas (Pa s)
ε turbulent dissipation rate
Subscripts
r reference value
a local value
e measured value
References
Al-Dughaither, Abdullah S., Ibrahim, Ahmed A., Al-Masry, Waheed A., 2010.
Investigating droplet separation efficiency in wire-mesh mist eliminators in
bubble column. J. Saudi Chem. Soc. 14, 331–339.
Bradie, I.K., Dickson, A.N., 1969. Removal of entrainment liquid droplets by wire
mesh demisters. Proc. Inc. Mech. Eng. 184, 195–203.
Brekke, Olaf, Bakken. Lars. E. Performance deterioration of intake air filters for gas
turbines in offshore installations. In: Proceedings of ASME Turbo Expo2010:
Power for Land, Sea and Air, 2010, pp. 685–694.
Brekke, Olaf, Bakken, Lars E., Syverud, Elisabet. Filtration of gas turbine intake air in
offshore installations: the gap between test and actual operating conditions. In:
Proceedings of ASME Turbo Expo 2009: Power for Land, Sea and Air, 2009,
pp. 371–379.
Brunazzi, E., Paglianti, A., 1998. Design of wire mesh mist eliminators. AIChE J. 44
(3), 505–512.
Brunazzi, E., Paglianti, A., 2000. Design of complex wire mesh mist eliminators.
AIChE J. 46, 1131–1137.
Brunazzi, E., Paglianti, A., 2001. Conventional and complex knitted mesh mist
eliminators. Chem. Eng. Technol. 24, 1199–1204.
Carpenter, C.L., Othmer, D.F., 1955. Entrainment removal by a wire mesh separators.
AIChE J. 1 (December), 549–557.
EI-Dessouky, H.T., 2000. Performance of wire mesh mist eliminator. Chem. Eng.
Process. 39, 129–139.
Feord, D., Wilcock, E., Davies, G.A., 1993. A stochastic model to describe the
operation of knitted mesh mist eliminators, computation of separation effi-
ciency. Trans. IChemE. Res. Des. 71, 282–295.
Helsør, T., Svendsen, H., 2007. Experimental characterization of pressure drop in
dry demisters at low and elevated pressures. Chem. Eng. Res. Des. 85 (A3),
377–385.
Holmes, T.L., Chen, G.K., 1984. Design and Selection of spray/mist elimination
equipment. Chem. Eng. 91 (21) (82 789).
Janajreh, I., Hasania, A., Fath, H., 2013. Numerical simulation of vapor flow and
pressure drop across the demister of MSF desalination plant. Energy Conver.
Manage. 65, 793–800.
Langmuir, I., Blodgett, K.B., 1944–1945. Mathematical Investigation of Water
Droplet Trajectories. Technical Report RL-225, General Electric Research Lab,
Schenectady, NY.
Rahimi, R., Abbaspour, D., 2008. Determination of pressure drop in wire mesh mist
eliminator by CFD. Chem. Eng. Process. 47, 1504–1508.
Setekleiv, A.Eddie, Hallvard, F. Svendsen, 2012. Operation and dynamic behavior of
wire mesh pads. Chem. Eng. Sci. 68, 624–639.
Setekleiv, A.Eddie, Thomas, HelsØr, Hallvard, F. Svendsen, 2010. Luid holdup in
wire-mesh pads. Chem. Eng. Res. Des. 88, 1523–1531.

View publication stats