7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Modeling Mlist Eliminator Devices as Internals of Gravitational Separator Tanks
J.A.A. Oliveira Jr.*, C.B. Fonte*, C.E. Fontes* and C.A.C. Moraest
Engineering Simulation and Scientific Software Ltda., Rio de Janeiro, RJ, Brazil
Jf Research and Development Center (CENPES), Petr61eo Brasileiro S.A., Rio de Janeiro, RJ, Brazil
aguirre~,esss.com.br, clarissa~,esss.com.br, carlos.fontes~,esss.com.br and capela~,petrobras.com.br
Keywords: Mist Eliminator, Multiphase Flows, Computational Fluid Dynamics.
Abstract
Mist eliminators are very common internal devices in separator tanks and distillation columns in oil and gas industry. The main
objective of these equipment is to minimize the liquid carry over (LCO) of the gaseous streams by recovering the liquid that
would be carried by this effluent gas. This recovery is important for both economical (avoiding the waste of this valuable
liquid material) and safety reasons (for the gas branch equipment downstream of the tank). Mist eliminators, also known as
demisters, are available in two main versions: the vane pack, composed by a pack of corrugated plates defining a "tortuous"
path for the gas and mist stream, where the bigger liquid droplets are collected; and the wiremesh demister, composed of a
pack of crossed wires with high wire surface density and high void fraction used to collect smaller droplets.
This work deals with CFD modeling of a mist eliminator device used as an internal of gravitational separator tanks applied in
the oil and gas industry. The methodology developed in this work is focused in wiremesh demisters but part of it can be also
used when dealing with vanepack demisters. The analysis was focused at internal devices for tanks in primary oil processing.
The available design methodologies are based mostly on empirical knowledge, using experimental correlations heavily
simplified supplied by the main manufacturers. These correlations are based on averaged flow quantities and ideal conditions.
Such conditions (uniformity on the gas flow, narrow range of droplet diameters, etc.) are barely the reality in the operation of
offshore separation equipment, mainly due to fluctuations in oil and gas production conditions.
A mixed Euler/EulerEuler/Lagrange multiphase simulation was initially applied. Three phases were defined in the problem: a
continuous gas phase, the carrier of the liquid droplets; a Lagrangian (dispersed) phase, the liquid droplets (mist); and a
dispersed Eulerian phase, the liquid collected at the mist eliminator surface. This last phase was responsible for some gas
stream redistribution, for some changes of the local efficiency distribution and it was generated due to the saturation of the
mist eliminator device. All the three phases strongly interact.
The demister equipment is represented by a porous media and the collection efficiency is calculated based on modified
versions of the manufacturers' formulations. Local gas velocity, droplet diameter, equipment characteristics, etc., were
considered in this model. The collected droplets mass were defined to change into a mass source term for the Eulerian liquid
phase as far as the equipment' saturation had been reached. The presence of this liquid phase redistributes the gas flow and
changes the local efficiency distribution.
With this modeling implemented different configurations of
manufacturers' design point.
Introduction
Mist eliminator device, also called demister, is a commonly
used internal device in the oil and gas industry and in other
industrial segments to eliminate mist (very small disperse
liquid droplets) from the gaseous streams. In the oil and gas
industry one of the applications of this kind of equipment is
as an internal device to gravitational separators in primary
oil processing units, in order to minimize liquid carry over
by effluent gas stream. These separators usually work with
oil, water and gas mixtures (sometimes with sand also) from
the risers. Allowing the mist to follow the gas stream,
besides the economical losses, can produce liquid
accumulation inside the gas piping, gas compressor
problems and that can be a risk to the entire unit.
the demister equipment are tested and compared with the
There are two main kinds of mist eliminators: the vanepack
and the wiremesh demister. The main difference between
these two types is the range of droplet diameter that each
one can collect efficiently. The vanepack type is designed
to collect droplets bigger than the wiremesh type. It is not
uncommon to use both types in series, using the vanepack
to capture the bigger droplets and the wiremesh to collect
the smaller ones or using the wiremesh to capture and
coarsen small droplets and the vanepack for the final
collection. Figure 1 shows both types of demister.
In the design of mist eliminators as internal devices of
separators, just average values of the operating conditions
are applied. Values as average gas stream velocity, average
droplet diameter and fluids physical properties are used.
Additionally, some internal design .1l' !ices" are given in
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
In this kind of problem, it can be evaluated the influence of
the disperse flow on the continuous flow properties
(momentum, mass and temperature) by the calculation of a
series of dimensionless parameters (Crowe, 2006). The
application of these parameters in the operational conditions
defined to this case shows that the dispersed phase has no
significant influence on the continuous phase flow. Here,
and in many other industrial applications, the dispersed
phase flow can be solved in alter step after the solution of
the continuous phase flow.
The mass flow of gas and liquid droplets were known, only
the upper portion of the horizontal and vertical cylindrical
separators that is available to the gas flow was simulated
(the regions filled with liquid were discarded). Additional
data will be available in section Study Definition.. Mist
eliminator device, also called demister, is a commonly used
internal device in the oil and gas industry and in other
Physical Modeling
As commented before, in many practical industrial
applications either volume fraction (or loading) of the
dispersed phase and droplets size are very small. This very
small amount and small droplet size of dispersed phase let
us consider that their effects on the continuous phase motion
can be neglected without losing physical consistency or
even accuracy. In these cases, the solution of the discrete
phase motion can be performed after the solution of the
continuous phase motion, in an serial fashion. This
approach is known as "oneway coupling". This makes the
numerical simulation of multiphase flows very cheap, since
the main (continuous phase) flow can be computed as an
steadystate singlephase flow and the discrete phase flow
computed in a latter step using the Lagrangian approach.
The basic physical description and mathematical modeling
of the continuous and discrete phases is given in the
subsections below.
Continuous Phase Modeling
Since the "oneway coupling" approach can be applied, the
continuous phase flow can be solved as a singlephase flow.
The steadystate, isothermal and incompressible flow of a
Newtonian fluid without any source terms or body forces is
described by the conservation equations of mass and
momentum. These equations are written, as shown in Fox
and McDonald (2001), as Eq. (1) and Eq. (2).
order to improve the incoming flow over the demister
surface. Detailed analyses are very uncommon, although
sometimes necessary for very critical applications.
The application of Computational Fluid Dynamics (CFD) in
problems of the oil and gas industry has been a driving force
to the development of multiphase flow formulations and
methods and of entire CFD packages. The present group is
focused on the application of CFD techniques to problems
of primary oil processing in the oil and gas industry. This
paper shows a particular application of Computational Fluid
Dynamics in modeling of a demister equipment of the
wiremesh type using a multiphase simulation. The main
objective of this work is to develop a methodology to
simulate some special cases where the precise design of
demister equipment will be necessary. This is achieved
using the ANSYS Fluent CFD package in a multiphase
EulerLagrange simulation. The equipment itself is
represented by an internal surface with a porousjump type
of condition and a User Defined Function (UDF) to
determine the efficiency and collect the liquid droplets when
they reach the demister.
dui
=0
Where x, is the position and u, is the continuous fluid
(Eulerian) phase velocity in the direction i.
Figure 1: Mist eliminator equipment, vanepack type (top)
and wiremesh type (bottom).
dui 1 dp 8 dui dui
u= +v +
'dxi p dxi Oxi dxi dxi
Where p is the static pressure, p is the continuous fluid
phase density and v is the kinematic viscosity of the
continuous fluid.
These equations are solved in ANSYS Fluent using the
Finite Volume Method. Complete discussions of the method
The gravitational separators with mist eliminators will be
treated as a main gas flow (gas with constant properties,
since there is no significant variation of pressure and
temperature inside of the separator), the Eulerian phase, and
a secondary flow of liquid droplets, the Lagrangian phase.
Problem Description
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
problem, the only additional force considered in the
simulations was the turbulent dispersion force. In a practical
point of view, the turbulent dispersion force adds a
fluctuating velocity component to the continuous phase
velocity used in the parcel movement calculations. This
fluctuating component is based on the local turbulent
properties (turbulent kinetic energy, turbulent kinetic energy
dissipation rate, etc.). With this force active, two particle
tracks started at the same point will not present identical
trajectories. In order to have samples statistically
meaningful, the analyst has to inject a larger number of
parcels than without turbulent dispersion force. In cases
where the "oneway coupling" can be applied this additional
cost is not significant. Additional information on the
definition of the turbulent dispersion force can be found in
Equipment Modeling
The modeling of demister equipment is usually done during
the global design analysis using data provided by
manufacturers. The expressions applied to calculate the
equipment efficiency use only average values (average
droplet diameter, average gas velocity, etc.) and return a
global (average) efficiency. Although in the majority of the
cases this approach is satisfactory, in some cases a more
detailed data is necessary. For such cases very few works
are found in the literature.
Many manufacturers use as a base point to equipment
dimensioning the ideal velocity, urdaht of the gas stream,
given by Eq. (7).
can be found in the books of Patankar (1980), Versteeg and
Malalasekera (2=II =") and Maliska (II 4). No additional
discussion will be presented here. Turbulence effects were
accounted using the SST model as defined in Menter (1994).
Discrete Phase Modeling
In a Lagrangian frame, the motion of the discrete droplets is
solved by tracking many parcels (particles representing
small groups of droplets with the same characteristics)
throughout the domain (Fluent, Inc., 2006). Equation (3) is
used to calculate the droplet velocity at each solved
position,
dog 1
m,=pCDAlui v(u Vi) + F (3)
Here m, is the mass of the droplet, t is the time, v, is the
discrete fluid (Lagrangian) phase velocity, F, the sum of
forces in the direction i (other than the drag force), CD iS the
drag coefficient and 4 is the area of the droplet proj ected in
the direction of the main flow. For a spherical droplet,
Eq.(3) can be written as in Eq.(4).
do, 18p CDRe Fi
ui vi(ui agi) + (4)
at ppd2 24 my
Where p, is the discrete fluid phase density, d, is the droplet
diameter, pu is the continuous fluid phase viscosity and Re is
the relative Reynolds number. The relative Reynolds
number between continuous and discrete phases is given by
Eq.(5).
pd,vi u I
Re = (5)
The solution of the discrete fluid phase motion is obtained
calculating the local velocity of a parcel at the current
position using Eq. (4) and updating its position (using the
calculated velocity and a reference timestep). These steps
are repeated until the parcel leaves the domain or reaches
the limit number of calculations. For the drag coefficient,
there is a large number of correlations available, for
example, in the books of Crowe (1998) and Crowe,
Sommerfeld and Tsuji (2006). By analyzing the current case,
due to the very small droplets size, it is very unlikely that
these droplets will suffer any kind of deformation. Aware of
this, a simple standard drag model for a spherical particle
can be applied. The chosen model is the model from the
work of Morsi and Alexander (1972), available in the
ANSYS Fluent interface. This model defines the drag
coefficient using the expression below.
CD 1 a + + (6)
Re Re2
With the constants a;, at and a3 defined according to the
cited work.
The additional forces term can include many different
physical phenomena, such as lift, virtual mass and
thermophoretic forces. Crowe (1998) gives a physical
description of these forces and some alternatives for
mathematical modeling. Due to the configuration of the
(pp p)
uidaem = k
p
The constant k is a characteristic of the equipment chosen
and p, is the dispersed phase density. This velocity is called
ideal because is the average gas stream velocity which gives
the maximum average efficiency of the demister. Based on
the information of one manufacturer product sheet (ACS
Industries, Inc., 2004) and using some additional field
information from a private communication from the
Petrobras engineers (Moraes, 2009) the following
expressions were developed to calculate the local demister
efficiency.
U, Uiaeal
uzz > uideal
Here k, is an inertial parameter, which accounts for the
effect of the droplet inertia in the collisions with the
wiremesh, u,, is the magnitude of the local gas stream
velocity component normal to the demister surface and D is
the wire diameter of the wiremesh. This inertial parameter
is used in Eq. (9) to obtain the collision efficiency parameter
E = 1[ 1.09841~
1+ (0.74536k,)o.990so
1PP p)uzzd2
S(pp p)d2(uzz 3kuidaem)
18pD
Figure 2: Geometries and boundaries definition for each
tank/demister configuration.
Table 1: Operation conditions and fluids properties.
Operational conditions
Temperature (To) 343.15 [K]1
Pressure (Po) 0.9 [MlPa]
Gas properties
Gas density (p) 6.726 [kgm]
Gas viscosity (p)l 1.09105 [Pa s]
. ~Liquid roerte
Liquid density (p,) 800.0 [kgm]
Liquid viscosity (Per)1 6.0103 IPa S
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
horizontal tank
horizontal demister
A'=0.67A,e
(10)
mixture Inlet
gas outlet
duem ser
walls
S' is the effective wire surface area available for collisions
with the droplets, .4, is the equipment wire surface area
density and e is the equipment thickness. Using the results
from Eq. (8), Eq. (9) and Eq. (10), the local equipment
efficiency, rI, can be estimated by Eq. (11).
horizontal tanke
vertical demister
The demister equipment head loss in the gas stream was
defined using typical head loss values according to data
from manufacturers (ACS Industries, Inc., 2004) by a
porous medium approximation (Fluent, Inc., 2006) via the
definition of a medium permeability, a, the pressure jump
coefficient, Ct and applying Eq. (12) to calculate the local
head loss, Ap.
enixture Inlet
gas outlet
demister
liquid level
Ap=uz+ C2 p 7
cr=(2 2 >
(ia>
Only the dry head loss was included in this definition, since
almost no data is available on the dependency of the
equipment head loss on the collected liquid saturation.
Study Definition
Two separator tank geometries were initially defined for the
study. One horizontal separator and one vertical separator.
For each separator model two configurations of the demister
device were studied (also horizontal and vertical). All the
configurations respect the limitations cited by the
manufacturers (ACS Industries, Inc., 2004). Figure 2 shows
all the separator/demister configurations and the boundaries
definition,
The temperature and pressure operational conditions and the
fluids properties are shown in Tab. 1. For the liquid droplets
diameters an experimental distribution from field
measurements was used (Moraes, 2009) with droplet
diameters ranging from 0.31[pm] to 301.68[fan].
The principal dimensions and flow rates of liquid and gas in
each separator configuration are listed in Tab. 2. Once the
demister is chosen, the entire device data is specified by the
manufacturer, the data for the model used in this case is
given in Tab. 3.
The demister is modeled in the simulation using a
porousjump condition. The flow properties of the
porousmedium were defined using Eq. (12) and the values
in Tab 3. The collection of the droplets was defined using
User Defined Functions inside ANSYS Fluent. These UDFs
are executed each time a parcel collides with a surface
element of the demister. When executed the UDF calculates
the local efficiency, using Eq. (7), Eq. (8), Eq. (9), Eq. (10)
and Eq. (11), and decides whether the parcel is collected or
not. If the parcel is collected, information about the droplets
diameter, droplets mass and droplets number from that
parcel are stored to evaluate the average diameter, the total
mass and the total number of particles collected locally at
each demister surface element.
hor ze Taldtearster
mlixure In et
gassoutlet
demister
Liquid level
mil
mixture In et
gas outlet
demister
liqui leyel
vertical tank
vertical demister
7} = 100 
[ 100
exp(EA')
~I
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
low efficiency can be corrected by changing the equipment
model to other with more appropriate characteristics (which,
in practice, will change the values of k, 4o and e).
As a complement to the global efficiency analysis, a
detailed study of the efficiency variation with the droplets
diameter was performed. The results of this study are shown
in Fig. 5. One can see in this chart a typical efficiency curve,
showing the reduction of the collection efficiency at the
smaller diameters.
Figure 6 shows the distribution of collected droplets mass
over the demister surfaces for all configurations. In this case
the experimental size distribution was used. In all cases the
mass collection distribution is very irregular. As commented
before this can contribute to the poor efficiency of the
devices, since some parts of the demister, those that receive
more droplets, will saturate quickly, increasing the head loss
and reducing the efficiency drastically.
horizontal tank
horizontal demister
gas flow streamlines
Velocity distr bution dem ster surf ace
Figure 3: Gas flow streamlines and velocity distribution
over the demister surface for the horizontal
separator/horizontal demister and vertical separator/vertical
demister configurations.
veloc~tymagnitude [mis]
200
a so
1 00
ose
0.0,O
velocity d stribution demister surface
Table 2: Separator configurations data.
Separator configuration horizontal vertical
Separator diameter (D,) 3.()[n] 1.45[z]
Separator length (L,) 11.()[nz 4.5[nz]
Piping diameter (D,,) ()0.3()5[z ().3)5[n1
Gas flow rate (V)I 1.15[nzs~' ] 1.725[nz s
Liquid mass flow rate (nt,) 6.811(T [kgs~' ] 1.)21(~)3kgs ]
Device diameter (Di,) 1.()16[nz] 1.45[z]
The boundary conditions imposed were: A velocity inlet at
the mixture inlet surface; a pressure outlet at the gas outlet
surface; a free slip wall at the liquid level surface; a noslip
wall at the walls surface group and; a porousjump condition
at the demister surface using the UDF to calculate the
parcels fate. Due to the geometric symmetry over the center
plane xz, the actual geometry simulated was only half of the
tanks, as shown in Fig. 2. For each separator/demister
configuration, the main gas flow was solved and, using this
solution, a set of droplets simulations was performed to
evaluate the final demister efficiency for each droplet
diameter. A final simulation was performed using the
experimental droplet distribution to verify the collection
distribution on the demister surface. The results are
presented and discussed in the next section.
Table 3: Demister equipment data.
Velocity constant (k) ().1()7[ns ']
Area density() 278.871[
Medium permeability (a) 2.611(Y[
Pressure coefficient (Cz ) (.)m
Wire diameter (D) 2.791(T [nz]
Equipment thickness (e) ().15[nz]
Study Results
The simulations were executed as described before, solving
the main flow of gas and, in a later step, running all the mist
flow analysis using the same gas flow solution. The gas
streamlines and the velocity distribution over the demister
surface are shown in Fig. 3. In the case with the separator
and the demister both in the horizontal position the flow
over the demister surface is nearly uniform, the opposite can
be seen in the vertical separator/vertical demister
configuration, where the flow over the demister surface is
very irregular. In both configurations one can see
recirculation zones, suggesting that an additional internal
device to homogenize the flow can be used to improve the
flow distribution. A drastic case is the horizontal separator
with vertical demister, where the major part of the tank is
occupied by a big recirculation zone and there is a short
circuit flow to the demister. The formation of this
shortcircuit flow has a negative influence over the demister
efficiency because big droplets, that otherwise would fall to
the liquid level surface, reach the demister inducing a faster
saturation of the equipment.
One of the main results from the simulations is the overall
efficiency of the mist eliminators on each configuration.
This data is shoru1 in Fig. 4.
Figure 4 shows that the global efficiency of the mist
eliminator equipment used in this study is not appropriate
for this application. It was expected a low value of
efficiency, since calculations using global averages (average
gas velocity, average droplets diameter, etc.) and the
equipment data have returned a value of il = 62.4[%]. This
collected droplets mass [kg]
0.000e+000 7.500e006 1.500e005 2.250e005 3.000e005
hotrizonta r n
collected droplets mass [kg]
0.000e+000 1.500e 005 3.0100e005 4.500e 005 6.000e 005
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
vertical tank
horizontal demister
4 0
~20
Tank/demister configurations []
Figure 4: Global demister efficiency for each
separator/demister configuration.
collected droplets mass [kg]
0.000e+000 1.500e005 3.000e005 4.500e005 6.000e005
vertical tank(
vertical demister
d .r
100.3
90,3
80,3
70,3
60,3
0.00+001.0 e005 3.000e00 4.50e00 6.00e00
Figure 6: Distribution of collected droplets mass over the
demister area for each separator/demister configuration.
Conclusions
The lack of efficiency in the demister model applied in this
study is due to the main device characteristics, that are not
ideal to collect such small droplets as we had in this case.
Another demister model (even from the same manufacturer)
WOuld be more appropriate. Other strong factor in this case
was the nonuniform flow over the demister surfaces. This
shown in the details of Fig. 3 and Fig. 6. This
nonuniformity will also have influence over the collected
liquid removal from the device (as the liquid is collected in
the wiremesh it is conducted to some additional device and
removed from the demister) making it harder to design the
conductors.
The main objective of this work, to develop a methodology
to simulate the flow of gas and mist and modelate the
droplets collection over a demister equipment, was achieved.
The simulation parameters and additional user routines
(UDFs) required were defined. They can now be applied in
real critical cases where detailed information about
distribution of collection and efficiency data is necessary.
Next Steps
This work is still under development. The next step is to use
the collected mass information to create a source term for a
second Eulerian phase, the collected liquid. This will be
helpful to verify the influence of the dripping liquid over the
liquid surface (this dripping can perturb the separation of
the liquids) and to design the devices to remove the
collected liquid from the demister.
;,0+ 0 I
son "
& akvdrnser
*:nk enitr
O,0 5.0 10.C 15,0 20,C 25.0 30,0 31,0 40,0 41,0 53,0
d oplet diameter pm]
Figure 5: Demister efficiency variation with droplets
diameter for each tank/demister configuration.
horizontal tank
horizontal demister
__I_
I
't
t
L
'C
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
References
ACS Industries, Inc., 2004, "The Engineered Mist
Eliminator", www. acsseparations.com.
Crowe, C.T., 2006, "Multiphase Flow Handbook", CRC
Press LLC, Boca Raton, USA, 1156 p.
Crowe, C.T., Sommerfeld, M. and Tsuji, Y., 1998,
"Multiphase Flows with Droplets and Particles", CRC Press
LLC, Boca Raton, USA, 471 p.
Dehbi, A., 2008, "Turbulent Particle Dispersion in Arbitrary
WallBounded Geometries: A Coupled
CFDLangevinEquation Based Approach", International
Journal of Mutiphase Flow, Vol.34, pp. 819828.
Fluent, Inc., 2006, "Fluent 6.3 Documentation".
Fox, R.W and McDonald, A.T., 2001, "Introduplio g
Mecbnica dos Fluidos", 5 Edition, LTC Editora, Rio de
Janeiro, 504p.
Maliska, C.R., 2004, "Transferencia de Calor e Mechnica
dos Fluidos Computacional", 2 Edition, LTC, Rio de
Janeiro,472 p.
Menter, F.R., 1994, TwoEquation EddyViscosity
Turbulence Models for Engineering Applications", AIAA
Journal, Vol.32, No. 8, pp. 15981605.
Moraes, C.A.C., 2009, Private Communication.
Patankar, S.V, 1980, "Numerical Heat Transfer and Fluid
Flow", Hemisphere, Washington, D.C., 214 p.
Versteeg, H.K. and Malalasekera, W, 2007, "An
Introduction to Computational Fluid Dynamics: The Finite
Volume Method", 2 Edition, Prentice Hall, New Jersey,
520p.
