Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Chemical Engineering Science 68 (2012) 624–639

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Science


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

Operation and dynamic behavior of wire mesh pads


A. Eddie Setekleiv, Hallvard F. Svendsen n
Department of Chemical Engineering, Norwegian University of Science and Technology, N-7491 Trondheim, Norway

a r t i c l e i n f o a b s t r a c t

Article history: Holdup distribution, pressure drop, separation efficiencies and droplet separation characteristics of six
Received 20 May 2011 different wire mesh pads were investigated experimentally. One of the mesh pads was tested with and
Received in revised form without a small gas bypass at the wall. The liquid holdup was recorded at minimum five locations
20 September 2011
inside the pads. The holdup distribution in the mesh pads were tested for time dependency. Separation
Accepted 12 October 2011
Available online 18 October 2011
efficiencies and droplet size distribution for the empty column was also investigated. The results show
variations in the holdup profile of the pads depending on running time, pad geometry, gas velocity and
Keywords: measuring position. A new equation was derived and a new parameter proposed to describe the
Multiphase flow characteristics of mesh pad operations.
Particle
& 2011 Elsevier Ltd. All rights reserved.
Droplet
Porous media
Mesh pad
Souders Brown equation

1. Introduction 2. Theory

Wire mesh pads are used in many chemical engineering and The theoretical tools used when characterizing wire mesh pads
oil production applications for removal of droplets. Increasing is the Stokes number and the Souders Brown equation. In the
capacity in a process plant or a gas/oil processing facility may following section a brief review of the Stokes number and the
result in higher concentrations of droplets in a gas stream and Souders Brown equation are given. For a thorough examination of
subsequently these droplets have to be removed. For downstream the origin of these two variables see Setekleiv et al. (2010).
equipment such as pumps and compressors, droplets in the gas A more rigorous derivation of the parameters affecting mesh
stream may result in poorer performance and possible breakdown pad operations is derived herein where the effect of surface
of the equipment. In other cases there is a desire to recover tension is emphasized. A presentation of the operating regimes
precious products suspended in the gas stream. Mesh pads are for wire mesh pads is also given.
ideal for removal of fine mists and sprays.
Gas flows containing small fractions of liquid and with droplet 2.1. Droplet capture
size distributions in the range of 52100 mm are considered
appropriate for wire mesh pad operation. Mesh pads are able to The main purpose of a wire mesh pad is to separate droplets
separate droplets down to a cut size of 2210 mm depending on from gas flows. The gas stream moves freely through the pad, but
the design of the mesh pad and operating conditions. A wire- if the inertia of the droplets is sufficient they impact on the wires.
mesh can also be used as a coalescer and pre-conditioner for The main mechanisms as stated by Holmes and Chen (1984) are:
down-stream equipment. Properties characterizing mesh pads are
low pressure drop, high porosity and long service life. An over-  inertial capture of droplets;
view of mesh pad selection criteria is given by Fabian et al.  direct interception.
(1993a, 1993b) and Anonymous (2005). They examine droplet
capture, structure of the wire mesh packing, design equations and Inertial capture, also called inertial impaction, is a process
industrial applications of wire mesh pads. whereby droplets leave the streamline around a target and impact
the target. Sufficient inertia is needed for inertial impaction to
occur. Inertial capture is the dominant mechanism for mesh pad
operation and is governed by the Stokes number:
2
ðrl rg ÞU g dp
n
Corresponding author. Tel.: þ47 73 59 41 00; fax: þ47 73 59 40 80. Stk ¼ ð1Þ
E-mail address: hallvard.svendsen@chemeng.ntnu.no (H.F. Svendsen).
18mg dw

0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.10.027
A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639 625

Fσ,sl

Fτ,lg
Fσ,gl

Fig. 1. Operational regimes in wire mesh pad operations. I. Gravitational settling


regime. II. Pre-flooding regime. III. Flooding regime. IV. Re-entrainment regime.

Here ðrl rg Þ has been included to account for the gas phase FG
influence on droplet separation (Hoffmann and Stein, 2002). The
Fig. 2. Force balance on an arbitrary liquid ligament suspended in a network
target diameter is dw , dp is the droplet diameter, mg the gas
of wires.
viscosity and U g is the gas phase velocity. Direct interception is a
mechanism that occurs when a droplet follows the streamline and
is collected when it sweeps close to the target.
The effect of direct interception is described by Lunde and assumed to be perpendicular to the flow direction. Gravity, F G ,
Lapple (1957); and the drag experienced by the ligament, F t,gl , are opposing
forces. Gravity tugs at the ligament downwards while the shear
dp
C¼ ð2Þ between the liquid and the gas results in an upward force. The
dw
surface tension forces acting on the ligament oppose the external
where dp is the particle diameter and dw denotes the wire forces on the drop and keep the droplet coherent. The surface
diameter. Direct interception is not dependent on droplet velocity. tension forces include the liquid gas inter-facial tension, F s,lg , and
the surface tension forces of the wire-liquid interface, F s,lg . de
2.2. K-value Gennes et al. (2004) described interactions between a solid
substrate and a liquid film of thickness e, with a free surface
The concept of K-value was developed by Souders and Brown exposed to a gas, in terms of energy per unit surface area:
(1934). In essence it is a force balance comprising the drag force
dE
and the gravity force acting on a droplet suspended in a gas flow ¼ ssl þ slg þ PðeÞ ð4Þ
dA
field. Souders and Brown (1934) assumed that the K-value would
be scalable to different flow fields by assuming that some of the where P(e) represents a function dependent on the thickness of
values in the derivation were constant: the liquid film. de Gennes et al. (2004) assumed that as the
sffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffi thickness of the film increases, Pð1Þ converges to 0. Above a film
4gdp rg thickness of 100 nm the term is negligible and it is assumed that
K¼ ¼ Ug ð3Þ
3C d rl rg Pð1Þ ¼ 0 at the surface of the liquid ligament. The surface forces
due to interfacial tension forces on the ligament then take the
Here U g can be seen as the terminal velocity, i.e. the velocity form
at which the drag force balance the gravity force. In this state a
given droplet would be suspended in the gas flow field. The F s,surface ¼ f ssl ksl @A þ slg klg @A ð5Þ
K-value is used to describe the different operational regimes of a
wire mesh pad. The characteristic regimes are: where f is a function dependent on the properties of the wires
embedded in the liquid ligament. At equilibrium we may write
 Gravitational settling. F s,lg þ F s,sl þ F t,gl ¼ F G ð6Þ
 Pre-flooding.
 Flooding. Here the inter-facial tension forces oppose the external force. If
 Re-entrainment regime. we consider a vertical up-flow separator operating in a regime
where gravity dominates and liquid drains from the wires the
Fig. 1 gives the characteristic pressure drop profile and the inter-facial tension forces will act co-currently with the gas flow.
specific regions for the given regimes, see Setekleiv et al. (2010) However if the gas–liquid shear dominates over gravity, the
for details. surface tension will work in the direction of gravity, and the
surface tension forces will be opposite to the flow direction.
2.3. Effect of surface tension in mesh pads Taking the forces at the liquid solid interface into special con-
sideration:
York (1954) established the use of Souders–Brown equation F s,sl ¼ ssl ksl @Asl ð7Þ
for wire mesh pad evaluation. He did not consider the effect of
surface tension on wire mesh demister design and this has not In order to simplify Eq. (6) it would be convenient to deter-
received any special attention in the literature on wire mesh pads. mine the area of the liquid solid interface, @Asl , as a function of the
In order to analyze the effect of surface tension, a force balance on volume of the liquid ligament, @V lg . To facilitate this transforma-
an arbitrary liquid ligament suspended in a wire network was tion in Eq. (7) a new parameter, pw , the wire packing density, was
considered, Fig. 2. The ligament is considered steady state. As in specified. The wire packing density, pw , [m/m3], is the length of
the Souders–Brown equation the effect of gravity and the influ- wire divided by mesh pad volume. By implementing this para-
ence of the gas phase have to be considered. The wires are meter we can describe the area of the liquid solid interface
626 A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639

in terms of ligament volume @V lg and circumference of the wires: the relation in terms of a characteristic length scale:
1 2 1
@Asl ¼ 2ppw r w @V lg ð8Þ ð2ppw ssl DrgÞk2
lg þ 2C d rg U g klg þ slg ¼ 0 ð13Þ
If we combine Eqs. (7) and (8) introducing the curvature of a Eq. (13) is similar to the capillary length as proposed by
cylinder, ksl ¼ 1=r w , we obtain an expression for the forces at the Batchelor (1967) which relates the magnitude of gravity forces
liquid solid interface: to capillary forces. Capillary length has been used by Lorenceau
F s,sl ¼ 2ppw ssl @V lg ð9Þ et al. (2006) to normalize the size of droplets draining from a wire
in a gravitational field. Similarly Eq. (13) may be used to normal-
If we combine this with Eq. (6) and introduce the more rigorous ize quantities such as flooding point and size of draining droplets.
implementation of the other forces we can write

slg klg @Alg þ 2ppw ssl @V lg þ 12C d rg U 2g @Alg ¼ Drg@V lg ð10Þ 3. Experimental apparatus

where we for simplicity consider that the forces in Eq. (6) are in The mesh pads were mounted in a column with an inner
equilibrium. By dividing by @V lg and introducing the geometric diameter of 102.5 mm in a vertical position, a typical scrubber
definition of curvature: configuration. The column was made of seven parts connected by
@A 1 flanges and packings.
k¼ ¼ ð11Þ The gas was fed through a horizontal pipe mounted 901 on the
@V @r
column and liquid was fed through a pneumatic nozzle above the
Eq. (10) simplifies to
inlet and centered in the middle of the pipe, see Fig. 4. The flow
slg k2lg þ 2ppw ssl þ 12C d rg U 2g klg ¼ Drg ð12Þ was measured by a volume flow transmitter, Flowtech DMV 6331.
Liquid entering the column was pumped by a ABVS 2-130
Eq. (12) shows a more faceted and detailed picture of mesh pad centrifugal pump and fed through a Lechler pneumatic nozzle.
operation than the Souders Brown equation. It introduces surface The gas pressure to the nozzle was kept constant at a gas pressure
tension of the liquid gas interface which most certainly should of 2.5 bar. The gas used to atomize the liquid was measured using
have an impact on mesh pad design. It also connects the effect of a direct mass flow meter, a Kobold WFM-22064M120, with an
gravity and shear forces between the gas liquid interface with the accuracy of 3% ð 7 6 N l=minÞ of full scale and the pressure was
effect of surface tension forces. The influence of the liquid solid measured with a Fuji pressure transmitter, FCX-C, with an
interface is shown and it introduces a new parameter, the wire accuracy 0.1% of full scale. The flow of liquid was measured with
packing density. By introducing the wire packing density only one a rotating flow meter, DPM-1507N22343, with an accuracy of 1%
curvature is needed, klg , which simplifies the equations. klg is of full scale.
related to the volume of the liquid ligament and should be related The mesh pads were all mounted inside a section of transpar-
to the amount of liquid trapped inside a wire mesh pad. In Fig. 3 ent PMMA, with a diameter of 100 mm, to allow visual inspection
an evaluation has been performed showing how the gas velocity of pad operation. Four rods around the section served to hold the
and wire packing density affect the curvature at the gas liquid weight exerted by the top part of the column. The rods also served
interface. It should be noted that the curvature scales as  D2 =D3 , as mounting for a gamma densitometer, see Fig. 8. The liquid that
if we consider the ligament as a sphere. This implies that when penetrated the mesh was collected by a separator, and by turning
the curvature decreases the volume of a liquid ligament increases. a valve downstream of the separator the liquid could be collected
In Fig. 3(a) it is seen that when the gas velocity increases, the in a basin for measuring the total amount of liquid penetrating the
curvature decreases and thus the gas flow should sustain larger mesh pad. The liquid moving downwards inside the column was
liquid ligaments. The same effect is seen in Fig. 3(b) as the wire collected in a tank and recycled to the nozzle. For a full overview
packing density increases, the curvature decreases and thus a of the system see Fig. 4. The pressure drop over the wire mesh was
higher wire packing density is able to sustain larger liquid measured using a differential pressure transmitter from Fuji, FCX-C
ligaments. These qualitative predictions are thus able to indicate with an accuracy of 0.1% of the total span ð 7 3:4 PaÞ.
how different parameters will affect wire mesh pad operations. A total of six different mesh pads were tested, Fig. 5. They were
The dependency of these parameters on mesh pad design is all located 112 cm above the air inlet. The geometric properties of
discussed in Section 5. If we divide Eq. (12) by k2lg we may write the pads are given in Table 1. The pads were all made of stainless

360 370

355
Curvature, κlg [m−1]

Curvature, κlg [m−1]

355
350

345 340

340
325
335

330 310
0 0.03 0.06 0.09 0.12 0.15 0 3 6 9 12
K−value [m/s] Wire packing density [m/m3] x 105

Fig. 3. The effect of varying velocity and the wire packing density upon curvature for an arbitrary liquid ligament suspended in a wire network. (a) shows the effect of
varying velocity when the wire packing density, pw , is 335 000 m/m3. (b) shows the effect of varying wire packing density when the K-value is 0.05 m/s.
A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639 627

Fig. 4. Layout of the experimental apparatus.

Fig. 5. Five of the wire mesh pads investigated. From rear left: LPB10, LPC10, LPA15, LPA5 and LPA7.

Table 1
Geometrical properties of six wire mesh pads.

Mesh (–) Height (mm) Dmesh (mm) dw (mm) rmesh (kg/m3) a


rmesh (kg/m3) Smesh (kg/m3) a
Smesh (kg/m3) E (%) Ea (%) pw (kg/m3)

LPA05 50 100 0.27 186.66 158 344.37 267 97.68 98.0 405.99
LPA07 70 100 0.27 162.96 158 300.64 267 97.97 98.0 354.44
LPA10 100 100 0.27 153.30 158 282.82 267 98.09 98.0 333.43
LPA15 150 100 0.27 154.05 158 284.22 267 98.08 98.0 335.07
LPB10 100 100 0.15 178.50 145 592.76 542 97.78 98.0 1257.88
LPC10 100 100 0.15 87.20 80 289.59 271 98.91 99.0 614.54

a
Values given by the manufacturer.

steel wire and were of a rolled configuration. They ranged in metal rods situated closest to the knitting, see Fig. 6. The purpose
porosities from 97.64% to 98.9%, and the range of heights was of using different mesh pads was to investigate how their
from 50 mm to 150 mm. Wire mesh pad height was fixed by different characteristics affected separation performance.
628 A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639

All mesh pads were made by Costacurta S.p.A Vico. The mesh assumed that the particle had to be uniform and that the waves of
pads were weighed and surface areas, porosities and densities the incident light were of constant frequency, amplitude and
were calculated and compared with the specifications given by parallel. The intensity of the scattered light may then be seen as a
the vendor, see Table 1. The parameters specified by the manu- function of the wavelength, l, particle size dp and the relative
facturer may differ somewhat from the real geometric parameters index of refraction n of a particle in a medium:
of the mesh pads, as discussed by Helsør and Svendsen (2007) and Iscattered ¼ Iincomming ðy, l,dp ,nÞ ð14Þ
shown in Table 1.
When a particle is illuminated by a monochromatic source, e.g.
a He–Ne laser, a diffraction pattern is created. The angle of the
3.1. Laser diffraction particle sizer
diffracted light wave depends on the particle size. The bigger the
particle, the lower the angle of the scattered light and vice versa
A laser diffraction particle sizer was used to characterize the
for smaller particles. By using an appropriate Fourier lens and a
ability of the mesh pads to separate droplets from the gas stream.
detector, the angle of the scattered light can be measured and a
Diffraction particle sizing is used in a wide range to give accurate
size for a specific particle can be given. When measuring on
measurements of particle sizes in many applications. The method
multiple particles the solution of the Mie equation will often lead
is usually applicable for a particle range from 0.1 to 3000 mm
to some smoothing of the results (Merkus, 2009). A rough sketch
depending on the setup of the system (Merkus, 2009). A particle
of the mounting of the laser diffraction instrument is seen in
loading from 0.001 to 1.0 vol% is appropriate. The laser diffraction
Fig. 7.
technique applied is based on Mie scattering developed by Mie
(1908) as a particular solution to the Maxwell (1861, 1873)
3.2. Gamma densitometry
equations. In order to solve for a spherical particle (Mie, 1908)
To measure the holdup, or liquid fraction, inside the wire mesh
pads, a gamma densitometer was utilized. The apparatus con-
sisted of a Cs-137 source encapsulated by a 110 kg steel container
and a cylindrical detector containing a scintillation crystal, photo
multiplier tube and necessary hardware. The scintillation crystal
had a diameter of 32 mm which is reflected in the ranges of
positions given in Table 2. The principle operation of the gamma
densitometer is to measure the attenuation of gamma rays
produced by a radiation source, in this case Cs-137. Gamma rays
passing through a medium, e.g. water, steel, etc. are attenuated by
the media. By using a photo multiplier and scintillation crystal the
photon flux can be measured, and by using appropriate reference
counts, the density can be calculated. In this case a 0% liquid
reference count was done on all mesh pads with air at ambient
conditions. This was done for all measuring positions. The same
procedure was done for the 100% reference count, but now the
mesh pads were immersed in water at ambient conditions. This
Fig. 6. The effective height of the wire mesh pads. gives a relationship between the attenuation of the two phases
(Linga, 1991):
Ig
¼ expðmi,g þ mi,l Þl ð15Þ
Il
In order to make a more accurate measurement the gamma
densitometer software corrected for temperature and background
radiation. The flux of gamma radiation from the source is a
random process so in order to reduce the random noise the signal
is smoothed using a rolling average with a specified time constant,
xi þ 1 Ai
Ai þ 1 ¼ Ai þ ð16Þ
t
where Ai þ 1 is the new rolling average, Ai is the previous rolling
average and xi þ 1 is the new measured value. For all the experi-
Fig. 7. Setup of the laser diffraction instrument. The test column is depicted in the ments a time constant of 240 s was used. This was done to ensure
middle with window access port for the laser. a low signal to noise error. The holdups of liquid in the mesh pads

Table 2
Measuring positions in different wire mesh pads, where 0 mm is the bottom of the pads.

Style mesh [–] LPA05 LPA07 LPA10 LPA15 LPB10 LPC10


Mesh height (mm) 50 70 100 150 100 100
Range position 1 (mm)  15–17  15–17  15–17  15–17  15–17  15–17
Range position 2 (mm)  10–22  5–27 5–37 5–37 5–37 5–37
Range position 3 (mm)  2.5–29.5 5–37 25-57 35–67 25–57 25–57
Range position 4 (mm) 5–37 20–52 45–77 65–97 45–77 45–77
Range position 5 (mm) 15–47 35–67 65–97 95–127 65–97 65–97
Range position 6 (mm) – – – 115–147 – –
A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639 629

were recorded for selected K-values and at least at five different to 0.20 m/s were investigated. Most mesh pads were investigated
positions, as given in Table 1. at K-values of 0.05, 0.08, 0.10, 0.15, 0.17 and at K-values of 0.13
and 0.20 m/s carryover and pressure drop were also recorded.
This was done in order to allow direct comparison of the result.
4. Uncertainty analysis of the gamma densitometry method Most pads were also investigated at the flooding point, Table 7,
and at a K-value 0.005 m/s below the flooding point. The measur-
To establish the validity of using a gamma densitometer to ing positions of the gamma densitometer, Table 2, were also
measure liquid holdup in wire mesh pads, a test of the repeat- chosen to allow direct comparisons of liquid fraction for some of
ability of the method was conducted. Mesh pad LPB10 was the measuring positions. Position 1 and the positions with range
mounted in the test column and the liquid fraction was recorded from 5 to 37 mm were the same for all pads investigated, see
at positions 1 and 2, the K-value was kept constant at 0.05 m/s and Table 2.
the volume fraction of liquid in gas was 0.02 vol%. The first run
was done with the mesh pad out of the box and mounted in the 5.1. Characterization of column with no internals
test column. In the second run the mesh pad was taken out of the
test column and left to dry in a drying cabinet overnight. The mesh In order to better evaluate the separation characteristics of the
was then again mounted in the test column in the same position mesh pads, an evaluation of the characteristic properties of the
and was kept running until the holdup stabilized. In the third run test column was performed with no internals present, only the
the mesh pad was in a pre-wetted condition and was allowed to inlet tube. Droplet size measurements were done at two posi-
run overnight. Table 3 shows the results of the repeatability tests. tions, 112 cm above the inlet and 156 cm above the inlet. The
It shows a very good repeatability of the method. lowest position corresponded to the droplet size distribution
An analysis of standard deviation was performed on all the entering the mesh pad, position 1. The characterization of posi-
holdup measurements to estimate the total dynamic error result- tion 2 was done to see if the column itself was able to separate
ing from averaging time dependent quantities, as seen in Table 4. droplets. Also a study of the efficiency of the column was carried
It must be said that Table 4 has a bias towards measurements out. To evaluate the performance of the nozzle, a study of the
made in the pre-flooding regime and especially towards measure- droplet size distribution at two nozzle pressures was tested. An
ments taken just below the flooding point. Just below the flooding analysis of the effect of reducing the measurement length of the
point the flow is highly unstable and thus large fluctuations in laser diffraction unit was performed by using two different
holdup are seen. The most extreme case being LPA15 which had a inserts.
standard deviation of 28.6% for position 6 at a K-value of 0.12 m/s.
A general observation from the results in Table 4 is that overall the 5.1.1. Droplet size distribution below mesh pad
highest uncertainties are found in the top position in the pads. In The droplet size distribution at position 1, just below the mesh
the re-entrainment regime the standard deviations are much pad, was recorded for several K-values as given in Table 5.
lower, in the range of 0.1–1%. Overall the dynamic uncertainty is In Table 5 dv,x signifies that x vol% of the droplets are below this
2.4%. Combining both the repeatability error and the dynamic diameter, d32 is the Sauter mean diameter and d43 is the Herdan
error gives a total uncertainty of 4.6% which is deemed acceptable. mean. For a description of these characteristic droplet sizes and
derivations see Liu (2000). The droplet size measurements had a
duration of 5 min for each velocity. Fig. 9 gives an overview of
5. Results and discussion
how the droplet size distribution changed with K-value. From a
K-value of 0.04 m/s up to 0.10 m/s the droplet size distribution
All experiments were conducted at 1 bar with air and water.
had only one peak. At a K-value of 0.10 m/s, bigger droplets were
The fraction of liquid in the gas flow was kept constant at
also created giving a bimodal distribution. When the K-value was
0.02 vol%. Gas velocities in the range of K-values from 0.04 m/s
increased the volume frequency of the smaller peak decreased
and the peak attributed to the larger droplets became more
Table 3
Test of repeatability in LPB10. Liquid fraction measured for several runs. prominent. The larger droplets were created by the nozzle as
discussed in Section 5.1.3. It should be noted that when the laser
Run Position 1 Position 2 diffraction instrument records droplets over dv,50 ¼ 1000 mm in
Liq. fraction (–) Liq. fraction (–) size the instrument is operating close to the upper detection limit
1 0.3057 0.5011
and may not be accurate if droplets of greater size passes the laser
2 0.3239 0.5111
3 0.3233 0.5018

Total repeatability 2.2%

Table 4
Averaging uncertainty in different wire mesh pads varying with position.

Position LPA10 LPB10 LPC10 LPA05 LPA07 LPA15

Uncertainty (%)
1 1.83 1.94 2.02 1.98 2.90 2.05
2 1.16 0.71 1.58 0.35 0.67 0.59
3 1.82 1.82 2.50 0.61 1.12 3.42
4 1.32 0.54 2.76 1.35 1.64 6.25
5 7.66 3.14 2.94 3.14 2.95 3.88
6 – – – – – 8.84

Dynamic uncertainty 2.4%


Fig. 8. Mounting of gamma densitometer in LP-rig.
630 A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639

Table 5 Separation efficiency of empty column


The characteristics of the droplet size distribution at recorded K-values. The dv,5 , 100
dv,50 and dv,95 show how many droplets by volume % are below the specific
number. d32 shows the typical diameter of a droplet and d43 is also shown.

K-value (m/s) dv,5 ðmmÞ dv,50 ðmmÞ dv,95 ðmmÞ d32 ðmmÞ d43 ðmmÞ 80

0.04 5.01 33.00 120.22 10.28 51.61

Carryover [%]
0.05 7.20 47.41 147.96 13.89 67.10
0.08 16.73 64.92 141.66 25.71 76.36 60
0.10 30.66 170.46 609.09 45.87 288.76
0.11 67.54 1077.72 1610.25 134.31 936.58
0.12 54.95 1108.15 1661.01 98.70 882.20
40
0.125 54.47 1041.77 1645.36 94.11 839.91
0.13 53.37 856.65 1613.64 92.99 776.40
0.14 67.99 801.50 1544.79 133.39 780.53
0.15 221.87 940.98 1550.48 336.18 926.15 20
0.17 241.74 946.02 1547.38 282.46 928.56

0
Droplet size distribution empty column 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
K−value [m/s]
20
K=0.04 m/s
Fig. 10. Efficiency of column with no mesh pad.
K=0.10 m/s
10 Table 6
Volume frequency [%]

The characteristics of the droplet size distribution at recorded K-values, at


position 2.
0
K-value (m/s) dv,5 ðmmÞ dv,50 ðmmÞ dv,95 ðmmÞ d32 ðmmÞ d43 ðmmÞ
100 101 102 103
0.05 6.63 46.73 166.97 13.15 71.58
20 0.10 27.41 102.02 873.58 40.71 292.36
K=0.11 m/s
0.12 50.67 824.52 1541.57 80.43 738.60
K= =0.17 m/s 0.15 283.20 1016.93 1570.69 349.31 983.88
10
Comparison of pos.1 and 2 at K=0.05 m/s

0 20 Pos. 1
100 101 102 103
Pos. 2
Droplet size [μm] 10
Volume frequency [%]

Fig. 9. Droplet size distribution in empty column, position 1, at K-value of 0.04,


0.10, 0.11 and 0.17 m/s. 0
100 101 102 103
beam. If the droplets are of greater size than the detection limit Comparison of pos.1 and 2 at K=0.10 m/s
they will be recorded as droplets of smaller size. However, these
20 Pos. 1
droplets will be separated by the mesh pad because of their
big size, in effect having a high Stokes number. Pos. 2
10

5.1.2. Position 2 0
In order to test the empty column ability to separate droplets, 100 101 102 103
measurements were also made at position 2, 156 mm above the Droplet size [μm]
inlet. The droplet size distribution was taken at K-values of 0.05,
0.10, 0.12 and 0.15 m/s, as given in Table 6. Fig. 11 shows a Fig. 11. Droplet size distribution compared, positions 1 and 2.
comparison of positions 1 and 2 at a K-value of 0.05 and 0.10 m/s.
At a K-value of 0.05 no separation of droplets due to height can be that the big droplets does not have the same velocity as the
seen, this trend is also apparent at a K-value of 0.10 m/s. By doing surrounding gas velocity. The stopping distance was calculated to
a comparison of Tables 6 and 5 we also see that this prevails at 20 m for a 1.5 mm droplet and thus the bigger droplets seen in
higher velocities. This might indicate that the speed of the big Fig. 11 is a result of the droplets not being harmonized with the
droplets out from the nozzle is higher than the surrounding gas gas flow. Another possibility is that the laser diffraction instru-
velocity and that a height of 156 mm is not sufficient for the ment smoothes the data so that the dv,95 cut off is set too high
droplet velocity of the big droplets to be harmonized with the gas compared with the actual values.
velocity. For large droplets as a consequence of size the stopping
distance will be long due to increased momentum. The maximum
droplet size at K¼ 0.10 m/s was calculated to be approximately 5.1.3. Test of nozzle pressure and inserts
700 mm when using the drag relation of Khan and Richardson To investigate the creation of the big droplets seen in
(1987) for solid particles and rearranging the Souders–Brown Figs. 9 and 11, Tables 5 and 6, a test of the effect of nozzle
equation to calculate the terminal velocity. This could indicate pressure on the droplet size distribution was made. The test was
A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639 631

performed at a nozzle pressure of 2.5 bar, which was the recom-


Stabilized
mended value from the vendor, and at 4 bar. Fig. 12 shows the
0.6 Not−Stabilized
results. We can clearly see that the reason for the big droplets
seen in the previous experiments is due to the lower pressure of
2.5 bar. There is not sufficient pressure and to much liquid in the 0.5
nozzle line to create a uniform distribution of droplets with a
single peak in the droplet size distribution. In effect the nozzle is 0.4
not able to atomize the liquid sufficiently at the recommended
specifications. This test was also performed at a K-value of 0.3
0.10 m/s with the same result.
Signal saturation of the laser diffraction instrument due to too 0.2
many droplets in the gas stream was also a possible problem. To
safeguard against this problem a set of inserts was made reducing 0.1
the measurement length from 90 mm 70 to mm. To test the effect
of using inserts, measurements were done at a K-values of 0.05, 0
0 10 20 30 40 50 60 70 80 90 100
0.10, 0.12 and 0.15 m/s. Running the experiments at K-value
Height [mm]
of 0.05 and 0.15 m/s gave comparable results to Table 5. For a
K-value of 0.10, Fig. 11, and a K-value of 0.12 the large diameter Fig. 13. The effect of not having long enough time to stabilize the holdup in the
peak shifted to lower values. For K-value 0.10 m/s the shift was wire mesh, mesh LPB10 K ¼0.05 m/s.
from 1500 mm to 750 mm and for a K-value of 0.12 the shift was
from 1500 mm to 850 mm. This shows that the nozzle propels the this situation more clearly. The result of not having long enough
bigger droplets so that they approach the wall and are carried time to stabilize the holdup leads to a difference at the three
away by the gas flow, this effect was also observed visually positions furthest from the gas inlet. The difference between a
through the PMMA transparent section of the column. stabilized mesh and a non-stabilized mesh is a holdup measure-
ment that has an error of several hundred percent. This very long
running time needed to reach steady state operation is probably
5.2. Time dependence due to the fact that the mesh pad separates most of the liquid in
the bottom part of the mesh and that this causes a much smaller
During preliminary studies of mesh pad operation discrepan- amount of liquid to enter the upper parts of the mesh pad. In turn
cies in the holdup measurements were found. An investigation of this leads to a very long stabilization time. This tendency is also
a possible time dependent mesh pad operation was executed. This clear in Fig. 16 where LPA10 and LPC10 at a K-value of 0.08 m/s
was done by first ensuring that mesh pad LPB10 was completely was pre-wetted and the liquid holdup was recorded for positions
dry, in effect putting the pad in a drying cabinet overnight. The 1 and 5 in both pads. The pre-wetting ensured that a stable
wire-mesh pad was then mounted in the test column and the gas operation of the pads was reached in a much shorter amount of
flow was set to a K-value of 0.04 m/s and the liquid/gas ratio of time. However, it is evident that position 1 reaches a steady state
0.02 vol%. The liquid holdup was recorded by the gamma densit- much faster than position 5. Steady state operations is almost
ometer at position 5. The mesh pad was allowed to run until it instantaneous at position 1 compared to position 5 for both pads.
reached a more or less stable state, Fig. 14. It can be clearly seen A typical evolution of liquid holdup in position 5 is also seen in
in Fig. 14 that the liquid holdup in LPB10 continually increased for Fig. 16. The liquid fraction steadily increases and the gradient
110 h before reaching steady state operation. The enlarged view flattens before the liquid fraction reaches a peak and settles at a
shows a typical profile for pads having reached a more or less mean value lower than the peak within a duration of a few hours.
stable regime of operation. This clearly shows that mesh pad Using this as a measure for how long mesh LPA10 and LPC10
operation may need a very long running time to reach steady needed to stabilize we see that mesh LPA10 needs about 3.5 h and
state conditions when starting from dry conditions. Fig. 13 shows mesh LPC10 needs 5.7 h. Considering that mesh LPC10 starts at a
higher liquid fraction than LPA10 we see that mesh LPC10 needs a
Test of Nozzle pressure at K=0.12 [m/s] longer stabilization time than LPA10. The longer stabilization
20 time may be due to the higher final liquid holdup reached in mesh
LPC10. These measurements show how important it is to reach a
2.5 bar
real steady state when measuring holdup in wire mesh pads.
10 4 bar To clarify the effect of liquid buildup inside the mesh pad on
the pressure drop, a measurement was conducted on LPB10 at
Volume frequency [%]

position 1, Fig. 15, where the pressure drop was calculated with a
0 rolling average so that it could be compared with the gamma
100 101 102 103 densitometer measurements. The experiment had a K-value of
Test of inserts at K=0.10 [m/s] 0.05 m/s and the mesh was completely dry at the start of the
measurement. Fig. 15 shows a clear relationship for the 1 h of
20 measurement. The liquid holdup inside the mesh increases and so
does the pressure drop. The variation produced from the 2 h until
No inserts
the end of the experiment is due to small velocity changes in the gas
10 Inserts flow and this of course effects the pressure drop over the mesh.

0 5.3. The effect of gas-bypass at the wall


100 101 102 103
Droplet size [μm] Mesh pad LPA15 was installed in two configurations. In one
configuration the wire mesh pad did not pack fully against the
Fig. 12. Test of nozzle pressure and inserts. wall of the column. For about 2/3 of the length of the pad, from
632 A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639

0.18
0.172
0.16 0.17
0.168
0.166
0.14 0.164
113 113.5 114 114.5 115 115.5 116 116.5
0.12
Time [h]
Liquid fraction [−]

0.1

0.08

0.06

0.04

0.02

0
0 20 40 60 80 100 120
Time [h]

Fig. 14. The time evolution of liquid holdup in LPB10 when the mesh was made completely dry, K¼ 0.04 m/s.

0.7 1000 LPA10 K−Value: 0.08 [m/s]


0.15
0.6
800
0.1
0.5
Liquid fraction [−]

Pressure [Pa]

600 0.05
0.4 Position 1 Position 5
Liq. fraction [−]

0.3 0
400 0 1 2 3 4 5
0.2 LPC10 K−Value: 0.08 [m/s]
0.15
200
0.1
0.1
0 0
0 2 4 6 8 10 0.05
Time [h] Position 1 Position 5
Fig. 15. Plot showing the dependence of liquid holdup upon pressure drop in wire 0
mesh LPB10, K¼0.04 m/s. 0 2 4 6 8 10
Time [h]

top to bottom, a gap of radius 1.5 cm, centered at the wall, Fig. 16. Liquid holdup versus time for LPA10 and LPC10 for positions 1 and 5,
K-value ¼ 0.08 m/s.
extended through the pad. The gamma densitometer did not
measure directly on the gap but was centered in the middle of the
mesh pad. In the other configuration care was taken to ensure positions. This is also seen at a K-value of 0.17 m/s. The most
that the pad was fitted against the wall. These two configurations significant change is that also at position 6 the holdup is higher
show differences both in holdup and efficiency. Fig. 17 shows a for the pad with bypass. The bypass configuration also show
comparison of liquid holdup in the mesh with bypass and without higher separation efficiencies, as seen in Fig. 18. The differences in
bypass. The bypass in the wire mesh pad was not directly holdup for LPA15 is in likelihood due to a different gas lift in the
measured by the gamma densitometer as the measurement path two configurations. If we compare the configuration of the two
did not go through that part of the wire mesh pad. Below the pads the gas flow in the configuration with no bypass will be
flooding point at a K-value of 0.08 m/s, the most significant more or less uniform entrapping liquid in the wires across the
difference is in the holdup profile at measuring position 6, from whole cross section. This will result in a higher holdup since
115 mm to 147 mm. LPA15 with bypass showed a significant entrapping liquid leads to a higher gas velocity, i.e. decreasing the
decrease in holdup compared to the configuration with no bypass. porosity. Comparing this situation with the configuration with
Measurements of holdup at a K-value of 0.05 and 0.10 m/s were bypass, which has a small gap which does not entrap liquid, this
performed as well and the same phenomenon prevailed. When leads to a lower overall gas velocity in the upper layers of the pad
the gas flow is increased to just below the flooding point, due to the absence of liquid ligaments sustaining a decrease in
K ¼0.12 m/s, and at the flooding point this pattern changes. actual porosity. The fact that the wires stabilize liquid holdup in
Throughout the pad with bypass we see a higher holdup for all wire mesh pads was shown in Eq. (12). The configuration with
A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639 633

bypass will let some of the gas go unhindered through the gap velocities in the mesh and thereby lower entrainment rates. This
between the wall and the pad. This will reduce the gas lift and as is reflected in the higher holdup of the configuration with bypass
such we see a lower holdup at low velocities. This will reduce the compared with the configuration with no bypass. This has a
gas lift in the pad and as such we see a lower holdup at low positive side effect when moving above the flooding point, the
velocities. Above flooding the bypass results in lower effective separation efficiencies increase for the configuration with bypass.

5.4. Flooding point

A normal method to evaluate the flooding point in mesh pad


operations is to say that once liquid is observed over the pad,
flooding has been achieved. However this method creates pro-
blems in terms of repeatability when observing flooding in
different pads. When conducting experiments some isolated
drops were observed above some pads as low as K ¼0.10 m/s.
However the separation efficiency was 100%. It should be noted
that the pads were mounted in a transparent PMMA section
which allowed for a detailed inspection of the operation of the
pad. Visual observation above the pad showed, together with
efficiency measurements, that these droplets were not carried
away with the gas flow. They traveled a couple of centimeters at
most and then settled down on the pad again. In order to have a
well defined reference determining the flooding point the flood-
ing point was defined as the moment liquid started pulsing out of
the pad, see Fig. 19. The pulsing was produced by very big
droplets which were thrown out of the pad and carried away
with the gas. These droplets could be of millimeter size. Also the
experiments were allowed to run overnight to establish that the
Fig. 17. Comparison of holdup in LPA15 for two configurations, with bypass at the
wall and no bypass at the wall.
flooding point was reached. LPA10 did showed early flooding
characteristics at a K-value of 0.12 m/s, but when the experiment
was allowed to run overnight no flooding behavior was seen. Thus
Efficiency of LPA15 another criterion for determining that the flooding point is
100 reached could be that the pulsing behavior should be sustained
LPA15 w/bypass for several hours. This new definition of flooding point corre-
LPA15 no bypass sponds to a change in the liquid holdup distribution in the pad.
80 A large increase in pressure drop was also seen, as shown in
Fig. 24. Fig. 22 shows that at the flooding point, the overall holdup
Carryover [%]

60 Before flooding At flooding

LPA10
Liquid fraction [−]

0.6 LPB10
40 LPC10
0.4

20 0.2

0 0 20 40 60 80 0 20 40 60 80 100
0.12 0.14 0.16 0.18 0.2 Height [mm]
K−value [m/s]
Fig. 20. Liquid fraction of 100 mm mesh pads before and at flooding. LPA10 at
Fig. 18. Comparison of efficiency of mesh pad and pipe below in LPA15 for two K-value of 0.12 m/s (BF) and 0.125 m/s (AF). LPB10 at K-value of 0.04 m/s (BF) and
configurations, with bypass at the wall and no bypass at the wall. 0.05 m/s (AF). LPC10 at K-value of 0.12 m/s (BF) and 0.125 m/s (AF).

Fig. 19. Mesh pad LPC10 at flooding. The pictures describe the pulsing action of liquid at the flooding point. On the left the buildup of the pulse is seen, the rightmost
picture show the behavior between pulses. The pictures have been taken by high speed camera, 1000 fps.
634 A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639

Before flooding At flooding Mesh pads with different heights were also investigated. From
0.25 Table 7 we see that increasing height from 100 mm to 150 mm
LPA15 has no effect on the flooding point. LPA10 and LPA15 have
Liquid fraction [−]

0.2 LPA10 according geometrical properties, only height differs. LPA7 have
LPA7 almost the same properties as LPA10 and LPA15 but the reduced
0.15
LPA5 height results in a flooding point occurring at a higher K-value.
0.1 This may be due to the shorter path the liquid has to drain. LPA05
shows a flooding point that occurs at an even lower K-value than
0.05 LPA7. This is probably due to the lower porosity which also
results in a higher wire packing density, see Table 1, compared
0 30 60 90 120 0 30 60 90 120 150 with the other LPA mesh pads. The porosity is around 0.4% lower
Height [mm] than LPA10 and LPA15. It is also 0.3% lower than LPA07. This
change in porosity is significant as shown by Helsør (2006).
Fig. 21. Liquid fraction of type A mesh pads with differing heights before and at
flooding. LPA15 at K-value of 0.12 m/s (BF) and 0.125 m/s (AF). LPA10 at K-value of
0.12 m/s (BF) and 0.125 m/s (AF). LPA7 at K-value of 0.13 m/s (BF) and 0.135 m/s 5.5. Liquid holdup
(AF). LPA5 at K-value of 0.11 m/s (BF) and 0.115 m/s (AF).

Holdup in all wire mesh pads was investigated at five different


positions and apart from the thickest mesh which was measured
Table 7 at six positions. An overview of the measuring range for each
K-value at flooding point for different mesh pads air/water 1 bar.
position is given in Table 2. Figs. 22, 23, 20 and 21 give an
Demister E (%) Smesh (m2/m3) pw (km/m3) K-value at overview of the results. Fig. 22 shows the results in terms of liquid
flooding point fraction versus K-value. Figs. 23, 20 and 21 show the liquid
fraction profile against measuring position. The measuring points
LPA05 97.64 344.37 405.99 0.115
shown in Figs. 23, 20 and 21 are the averages of the measuring
LPA07 97.94 300.64 354.44 0.135
LPA10 98.06 282.82 333.43 0.125 range at the selected positions, Table 2, and the designated
LPA15 98.05 284.22 335.07 0.125 velocities are the same as the velocities shown in Fig. 22. In order
LPB10 97.74 592.76 1257.88 0.05 to systematize the data obtained for liquid fraction in the various
LPC10 98.90 289.59 614.54 0.125 mesh pads it is necessary to study the holdup profiles below, at
and after the flooding point as the profiles are distinctly different.

in the wire mesh pads increase. This is also apparent in Figs. 20 5.5.1. Liquid holdup before flooding
and 21. The flooding point corresponds to a change of operating Evaluating the holdup profile of mesh LPC10, LPA15, LPA10
conditions in the pads. and LPA7 in Fig. 23 we observe a similar trend at a K-values of
The flooding point is understood as the moment the gas lift in 0.05, 0.08 and 0.10 m/s. For all the mesh pads and K-values
the wire mesh pads becomes sufficient for re-entraining liquid mentioned the highest liquid fraction is recorded in position 1,
from the pad. To understand what affects the flooding point which is the same position for all pads, the profiles dip until a
equation (12) gives a good representation of flooding mechanics. minimum is reached and then the holdup increases and reaches a
Considering that the experiments are conducted with one fluid maximum in the position furthest up in the pad, see Fig. 23. The
configuration and only steel wires, there are three variables that higher liquid fraction in position 1 is probably due to the fact that
could affect the flooding point, wire packing density, gas velocity most of the liquid that hits the pad is separated out in the bottom
and the curvature of the gas liquid interface. The curvature of the part of the pad. Also, any liquid that is separated further up in the
gas liquid interface should change to some degree with the pad also needs to drain and this will also create an effect of
holdup of liquid present in the mesh pads as a consequence of increased liquid fraction at positions lower down in the pads.
varying wire packing density and gas velocity, see Fig. 3. Increas- Position 1 is also situated in the bottom support grid, see Fig. 6.
ing the wire packing density should lower the flooding point as This grid will slightly increase the velocity inside the mesh pad
the liquid solid surface tension force, F s,sl , should increase with which again will affect the holdup as the gas lift will be higher.
higher wire packing density, as shown in Fig. 3 and Eq. (11). As The support grid also affects the holdup in position furthest up
the curvature decreases, the holdup in the pad should increase. the pad resulting in a higher holdup, this again is because of the
Table 7 shows that this is the case for LPB10, with its higher wire added gas lift caused by the constriction of the metal bars, see
packing density, and from Fig. 22 we see that the holdup in LPB10 Fig. 6. Another trend apparent in Fig. 22 is that for most
is much larger than in the other pads investigated. The large measuring positions an increase in gas velocity leads to a higher
holdup also results in a substantial increase in gas velocity inside liquid holdup below flooding. The LPA7 profile deviates from
the pad which compels re-entrainment of liquid in the pad and mesh LPC10, LPA15 and LPA10 at a K-value of 0.10 m/s in that at
thus the flooding point is reached at a much lower K-value than position 1 it shows a lower liquid fraction than in position 2. This
the other pads. This trend is also seen in LPC10. Helsør (2006) may be a result of the reduced height compared to the other three
showed that as the mesh pad porosities increased, the flooding mesh pads as this is the most prominent geometric feature that
point occurred at a higher K-value. Calculations show that this distinguishes LPA7 from LPC10, LPA15 and LPA10. If we focus our
also corresponds to a lower wire packing density. Mesh LPC10 has attention on LPA5 we see that before flooding, at K¼ 0.05–0.10 m/s,
a higher packing density compared to the highest porosity mesh the holdup profile is not at all similar to the other mesh pads
investigated by Helsør (2006). The higher wire packing density previously discussed. In the velocity range K ¼0.05 to 0.10 m/s the
leads to a higher holdup compared to LPA10, which has a lower holdup increases, when going up in the pad, to a maximum and
porosity, and thus a lower flooding point than shown by Helsør then decreases. The opposite effect is seen in LPC10, LPA15, LPA10
(2006). Needless to say the velocity in the pad also increases as a and LPA7. This is probably due to the lower mesh pad height of
result of higher holdup and re-entrainment follows, Fig. 22 shows LPA5 which would give liquid separated in the top part of the pad
the holdup behavior. a shorter distance to drain and this may affect holdup.
A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639 635

LPA5 LPA7
0.2 0.2

0.15 0.15

0.1 0.1

0.05 0.05

0 0

LPA10 LPA15
0.2 0.2
Liquid fraction [−]

0.15 0.15

0.1 0.1

0.05 0.05

0 0

LPC10
0.6 LPB10 0.2

0.15
0.4

0.1

0.2
0.05

0 0
0.04 0.08 0.12 0.16 0.2 0.04 0.08 0.12 0.16 0.2
K−value [m/s]

Position 1 Position 2 Position 3

Position 4 Position 5 Position 6

Fig. 22. Holdup versus K-value for all mesh pads, showing the selected positions inside the pads.

5.5.2. Liquid holdup at flooding and above point we see a general decrease in holdup as the gas velocity is
At the flooding point and above, the liquid holdup profiles increased, Fig. 22, for all positions. This corresponds with the re-
become more uniform for all pads, see Figs. 20, 21 and 6. They entrainment regime and lower separation efficiencies of the wire
start at a minimum in position 1 and increase to a maximum at a mesh pads are seen, as discussed in Section 5.7. The gas velocity
height between 14.75 and 21.00 mm (midpoint of range) depend- inside the mesh pads is now sufficient to re-entrain liquid from
ing on the pad. The profiles then decrease further up in the pad. the wires in the pads and this results in a lower holdup inside
For LPB10 position 1 remains at a maximum until reaching a the pads.
K-value of 0.10 m/s, see Fig. 22. This is likely due to the lower gas
velocity in the column, there is not enough gas lift to affect the 5.6. Pressure drop
holdup in the bottom part of the mesh. LPA15 also shows
deviating behavior at position 6 furthest up in the pad compared All the examined wire mesh pads follow the typical pressure
with the other mesh pads except LPA5. At the flooding point the drop pattern, as given in Fig. 1. LPB10 measurements started close
effect of the supporting grid may be seen at position 6. Looking at to flooding and so only a part of the typical s-shaped pressure
Figs. 20 and 21 we recognize a significant effect corresponding drop profile is seen, see Fig. 24. The profile of LPB10 exhibits
with the flooding point. For most positions investigated in the characteristics of regimes III and IV in Fig. 1. Looking at the results
various mesh pads we see an increase in liquid holdup as for the other mesh pads we clearly observe the pressure drop
mentioned in Section 5.4. The increased velocity together with evolution characteristic of regimes II–IV in Fig. 1. At low K-values
an increase in liquid fraction in the pads result in a stronger gas the pressure drop is slowly increasing. At and just below
lift which can sustain a higher liquid holdup. Above the flooding the flooding point, see Table 7, the pressure drop increases at a
636 A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639

0.2

K= 0.05 [m/s] K= 0.08 [m/s]


0.15

Liquid fraction [−] 0.1

0.05

0.2

K= 0.10 [m/s] K= 0.15 [m/s]


0.15

0.1

0.05

0 30 60 90 120 0 30 60 90 120 150


Height [mm]
0.25

K= 0.17 [m/s]
0.2 LPA15
Liquid fraction [−]

LPA10
0.15
LPA7

0.1 LPA5

LPC10
0.05

0 30 60 90 120 150
Height [mm]

Fig. 23. Holdup versus height for all mesh pads at given K-values.

Pressure drop vs. K−value, all mesh pads


higher rate. This is termed the flooding regime. Moving to higher
1600
LPA5 K-values, above flooding, the pressure drop increases more
slowly. From Fig. 24 we observe that the pressure drop is
1400 LPA7
dependent on pad geometry and liquid holdup. LPB10 has the
LPA10 largest pressure drop of all the pads. This is due to the higher
1200 LPA15 holdup of liquid in the pad, see Fig. 22, which is a result of the
LPB10 higher surface area, Smesh , and probably due to the high wire
1000
LPC10 packing density, pw , of the pad, see Table 1. Subsequently LPA15
Δ P [Pa]

follows with the second largest pressure drop which is the result
800
of the added height compared to the other pads. Surprisingly
mesh LPC10 follows with a higher pressure drop that LPA10. From
600
Fig. 23 we see that LPC10 has a larger holdup than LPA10. If we
compare the geometric properties of the pads, see Table 1, we see
400
that the surface areas are almost equal, the porosity differs by 1%
and the wire packing density of LPC10 is almost twice that of
200
LPA10. This difference in properties is a result of the lower wire
0 diameter of LPC10. Taking into account the results from LPB10
0.04 0.08 0.12 0.16 0.2 which has a slightly lower porosity than LPA10 but a much higher
K−Value [m/s] pressure drop, it is not likely that the porosity is the determining
factor. Results described by Setekleiv et al. (2010) showed that
Fig. 24. Pressure drop against K-value for all the tested wire mesh pads. increasing porosity should result in lower pressure drop. They did
A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639 637

however not test for wire packing density, which is the most Efficiency of mesh pads and pipe below
likely factor producing the higher pressure drop of LPC10 com- 100
pared to LPA10. A higher wire packing density results in a higher
holdup of liquid inside the pad and the actual porosity decreases,
thus increasing the gas velocity inside the pad which in turn leads 80
to a higher pressure drop. The pressure drop of LPA7 is lower than
LPA10 most likely due to the lower height of LPA7. Decreasing the
height has the effect of reducing the total amount of wires and

Carryover [%]
60
thus the total holdup of liquid. This leads to a reduction of the
total liquid/gas surface area and wire/gas surface area which
should give a lower pressure drop compared to LPA10. One would LPA5
expect that the lower height of LPA5 would also result in a lower 40
LPA7
pressure drop compared to LPA7. However LPA5 has a signifi-
LPA10
cantly lower porosity and also a higher surface area per cubic
meter and higher packing density than LPA7, see Table 1. The 20 LPA15
flooding point is also lower. The lower porosity leads to a overall LPB10
higher holdup and thus a larger actual porosity and this leads to a LPC10
pressure drop that is higher than LPA7 in the pre-flooding regime. 0
At higher K-values, above K ¼0.15 m/s, the holdup in LPA5 is 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
lower than LPA7 and because of the smaller height of LPA5 the K−Value [m/s]
pressure drop is lower.
Fig. 25. Efficiency of mesh pad and pipe below for all wire mesh pads.

5.7. Efficiency of mesh pads and pipe below


Mesh pad droplet size distribution
30
The efficiency of all the mesh pads in Table 1 was investigated,
K=0.05 [m/s]
Fig. 25. It was not possible to collect and measure any liquid 20
carryover up to the flooding point, Table 7, but as discussed in
Section 5.8 not all droplets are separated by the wire mesh. The 10
droplets are too small for the separator in the rig, Fig. 4, to separate 0
them from the main gas flow. When the mesh pads reach the 100 101
Volume frequency [%]

flooding point the efficiency of separation decreases, most rapidly 30


for LPA7. For the other mesh pads the decrease in efficiency is more K=0.08 [m/s]
gradual. LPB10 show the slowest reduction in efficiency and this is 20
a result of the very low flooding point of K¼0.05 m/s. LPA15 and
10
LPA10 do not show a big difference in efficiency indicating that
increasing the mesh pad height does not affect the efficiency to a 0
large degree. LPC10 has a lower efficiency than LPA10 but also has 100 101
a large increase in efficiency from a K-value 0.13 to 0.15 m/s. This 30
trend is contrary to the evolution of the efficiency profiles of the K=0.10 [m/s]
other pads, which are decreasing after reaching the flooding point. 20
It is not clear why this behavior is seen, but it corresponds with a 10
jump in holdup at position 5, top position in pad, in Fig. 22 at a K-
value of 0.13 m/s which decreases at a K-value of 0.15 m/s. When 0
the K-value is increased from 0.17 m/s to 0.20 m/s all mesh pads 100 101
exhibit an almost uniform efficiency, the efficiencies being around Droplet size [μm]
8.5%. In Fig. 10 we observe that at K¼0.17 m/s the empty column
LPA5 LPA7 LPA10 LPA15 LPC10
had an efficiency of 11%. At this velocity the empty column has a
better efficiency than the mesh pads investigated. If we compare Fig. 26. Droplet size distribution upstream of the mesh pads at a K-value of 0.05,
Fig. 10 with the performance of LPB10 we clearly see that above a 0.08 and 0.10 m/s.
K-value of 0.08 m/s this mesh pad has a larger carryover than an
empty column. If we look at Fig. 25 selecting mesh pad LPA10 or all other parameters are constant, will increase with increasing
LPA15 would give the best separation efficiencies. velocity. Increasing the Stokes number results in a higher prob-
ability for a droplet to impact a target. This trend has also been
5.8. Droplet separation shown by Brunazzi and Paglianti (1998). If we compare LPA10 and
LPC10 we also see a difference in the recorded data. LPC10 gives
In order to evaluate the ability of the mesh pads to separate complete separation at K¼0.08 m/s while LPA10 does not. This is
droplets, droplet size measurements were made above the pads most likely the result of the reduced wire diameter of LPC10,
with a laser diffraction instrument, see Fig. 26. The results are Table 1, which increases the Stokes number, see Eq. (1). There is
summarized in Table 8. All mesh pads, except LPB10, were tested also a trend that accompanies the increase in height. As the height
at K-values of 0.05, 0.08 and 0.10 m/s. At some of the velocities is increased from LPA5 to LPA15 the droplet cut size decreases,
studied no droplets were detected. We clearly see from Table 8 Fig. 26, and complete separation is evident for LPA10, at
and Fig. 26 that when the gas velocity is increased for LPA5 and K¼0.10 m/s, and for LPA15, at K¼0.08 m/s and K¼0.10 m/s. The
LPA7 and LPA10 the Sauter mean diameter, d32 , and volume increased height will result in a larger amount of wires available
median diameter, dv,50 , decrease and for LPA10 this results in no for direct interception. This trend is also seen if we compare LPC10
droplets recorded above the mesh at a K-value of 0.10 m/s. The and LPB10. They have the same wire diameter, dw , but the packing
trend is a result of increasing Stokes number, Eq. (1), which when density of wires, pw , for LPB10 is almost twice that of LPC10.
638 A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639

Table 8 size distribution. The droplet-size distribution after the liquid had
The characteristics of the droplet size distribution at recorded K-values above the been separated by mesh pads were also investigated. The droplet-
mesh pads. For LPA15 and LPC10 the droplet size distribution was also recorded at
size cut off showed a dependency on mesh pad height, wire
K¼ 0.08 m/s and K¼0.10 m/s but no droplets were detected.
diameter and gas velocity. In effect this showed that the droplet
Wire mesh pad K-value (m/s) dv,50 ðmmÞ d32 ðmmÞ cut size is dependent on Stokes number and direct interception.

LPA5 0.058 5.02 4.73


LPA5 0.081 4.97 4.54
Nomenclature
LPA5 0.101 4.50 3.94
LPA7 0.050 4.52 4.29
LPA7 0.081 3.87 3.58 Ai rolling average (–)
LPA7 0.100 3.57 3.22 Asl area of solid liquid interface (m2)
LPA10 0.052 3.89 3.66 Cd drag coefficient (–)
LPA10 0.081 2.13 1.68
Dmesh diameter of mesh pad (m)
LPA10 0.101 – –
LPA15 0.052 3.41 3.11 dv,5 5 vol% of the droplets are below this diameter ðmmÞ
LPC10 0.052 3.84 3.73 dv,50 50 vol% of the droplets are below this diameter ðmmÞ
LPB10 0.041 – – dv,95 95 vol% of the droplets are below this diameter ðmmÞ
d32 Sauter mean diameter, characteristic droplet size
ðmmÞ
This will also result in a larger amount of wires available for direct d43 Herdan mean ðmmÞ
interception. LPB10 has complete separation of droplets at K¼0.04 dp particle diameter (m)
and LPC10 is not able to separate all droplets at a K¼0.05 m/s. It is dw wire diameter (m)
also possible that the amount of liquid in the pad affects the FG gravity force (kg m/s2)
probability of interception since liquid holdup increases with F s,lg liquid gas interfacial tension force (N)
velocity at least up to the flooding point. F s,sg solid liquid interfacial tension force (N)
F t,gl shear force between gas and liquid (N)
g acceleration of gravity (m/s2)
6. Conclusion l length between gamma source and detector (m)
Ig intensity of incoming light (1/ms)
It was shown that wire mesh pads may need a long running Iincomming photon flux gas phase (1/ms)
time to reach steady state operation. Pre-wetting the wire mesh Il photon flux liquid phase (1/ms)
pads led to a faster attainment of stable operating conditions. The Iscattered intensity of scattered light (1/ms)
position in the pad closest to the inlet in the column showed a n refraction index (–)
fast response in reaching a stable state while the position that pw wire packing density (m/m3)
was closest to the outlet needed a long running time for reaching rw wire radius (m)
stable operation. Smesh surface area of mesh per volume (m2/m3)
The configuration with gas-bypass and without bypass showed Stk 2
Stk ¼ DrU g dp =18mg dcyl (–)
differences in operation. The gas-bypass affected holdup in the
wire mesh pad, but not the flooding point. The configuration with t time constant (s)
bypass had a lower carryover than the pad with no bypass. Ug velocity of gas phase (m/s)
A new equation was developed to describe the mechanisms xi þ 1 liquid fraction (–)
present in mesh pads. This equation showed that mesh pad design @A differential area (m2)
should be characterized with a new parameter, the wire packing @V differential volume (m3)
density. The equation also showed that gas velocity, gravity, surface Dr Dr ¼ rl rg (kg/m3)
tension between the liquid and gas and interfacial tension between E porosity of mesh pad (–)
liquid and solid affected mesh pad performance. The equation y angle of diffracted light (–)
showed that the curvature of the liquid gas interface should klg curvature of the liquid ligament at gas liquid
decrease when gas velocity was increased. The equation also interface (1/m)
showed that when the wire packing density was increased the ksl curvature of the liquid ligament at solid liquid
curvature of the gas liquid decreased. Decreasing curvature implies a interface (1/m)
larger gas liquid interface. These trends were also seen in the mg viscosity gas phase (kg/ms)
experimental data gathered. For a given pad increasing the gas l wavelength (m)
velocity resulted in higher holdup before the flooding point. This mi,g attenuation coefficient gas phase (–)
corresponds with a larger gas liquid interface and thus a lower mi,l attenuation coefficient liquid phase (–)
curvature. The wire packing density also affected mesh pad perfor- rg density of gas phase (kg/m3)
mance. When the wire packing density was increased the holdup in
rmesh density of mesh pad (kg/m3)
the pads increased and thus a larger gas liquid interface was created.
rl density of liquid phase (kg/m3)
This results in a lower curvature for a liquid ligament suspended in
slg surface tension at gas liquid interface (N/m)
the wire mesh pad and thus a larger gas liquid interface.
ssl surface tension at solid liquid interface (N/m)
The separation efficiency of the mesh pads and pipe below
were shown to be comparable or lower than the separation
efficiency of a column with no mesh pad. After the flooding point
the separation efficiencies decreased and above a K-value of
References
0.17 m/s the mesh pads exhibited an almost uniform separation
efficiency of 8.5%.
Anonymous, 2005. The engineered mist eliminator. Chem. Eng. Progr. 101(6), B1.
Droplet size distributions were recorded at two locations in the Batchelor, G., 1967. An Introduction to Fluid Dynamics. Cambridge University
empty column 44 cm apart. The distributions had the same droplet Press ISBN 0-89573-879-1.
A.E. Setekleiv, H.F. Svendsen / Chemical Engineering Science 68 (2012) 624–639 639

Brunazzi, E., Paglianti, A., 1998. Design of wire mesh mist eliminators. AIChE J. Liu, H., 2000. Science and Engineering of Droplets: Fundamentals and Applications,
44 (3), 505–512. 1st ed. William Andrew ISBN 978-0815514367.
de Gennes, P.-G., Brochard-Wyart, F., Quere, D., 2004. Capillarity and Wetting Lorenceau, E., Senden, T., Quéré, D., 2006. Wetting of fibres. In: Weiss, R.G., Terech, P.
Phenomena: Drops, Bubbles, Pearls, Waves. Springer ISBN 3-540-43326-0. (Eds.), Molecular Gels, Materials with Self-Assembled Fibrillar Networks1st ed. ,
Fabian, P., Hennesey, P., Neuman, N., 1993a. Demystifying the selection of mist Springer ISBN 1-0-1-4020-3352-4.
eliminators, Part 1: the basics. Chem. Eng. 100 (11), 148–156. Lunde, K.E., Lapple, C.E., 1957. Dust and mist collection. Chem. Eng. Progr. 53 (8),
Fabian, P., Hennesey, P., Neuman, N., 1993b. Demystifying the selection of mist 385–391.
eliminators, Part 2: the applications. Chem. Eng. 100 (12), 106–111. Maxwell, J.C., 1861. On physical lines of force. Philos. Mag. 21 (4), 75–161.
Helsør, T., 2006. Experimental Characterization of Wire Mesh Demisters. Ph.D. Maxwell, J.C., 1873. A Treatise on Electricity and Magnetisms, 1st ed. Clarendon
Thesis. NTNU. Press, Oxford.
Helsør, T., Svendsen, H.F., 2007. Experimental characterization of pressure drop in dry Merkus, H.G., 2009. Particle Size Measurements, Fundamentals, Practice, Quality.
demisters at low and elevated pressures. Chem. Eng. Res. Des. 85 (3), 377–385. 1st ed. Springer Science þBusiness Media, ISBN 978-1-4020-9015-8.
Hoffmann, A.C., Stein, L.E., 2002. Gas Cyclones and Swirl Tubes. Springer, New York Mie, G., 1908. Beiträge zur optik trüber medien speziell kolloidaler goldlösungens.
ISBN 3-540-43326-0.
Ann. Phys. 25, 377–445.
Holmes, T.L., Chen, G.K., 1984. Design and selection of spray/mist elimination
Setekleiv, A.E., Helsør, T., Svendsen, H.F., 2010. Liquid holdup in wire-mesh
equipment. Chem. Eng. 91 (21), 82–89.
demisters. Chem. Res. Des. 88 (11), 1523–1531.
Khan, A.R., Richardson, J.F., 1987. The resistance to motion of a solid sphere in a
Souders, M., Brown, G., 1934. Design of fractionating columns. i. Entrainment and
fluid. Chem. Eng. Commun. 62 (1–6), 135–150.
capacity. Ind. Eng. Chem. 2, 98–103.
Linga, H., 1991. Measurements of Two-phase Flow Details: Non-intrusive Methods
York, O.H., 1954. Performance of wire-mesh demisters. Chem. Eng. Progr. 50 (8),
Applied to Slug and Dispersed Flows. Ph.D. Thesis. Norwegian University of
Science and Technology. ISBN 82-7119-301-5. 421–424.

You might also like