Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Available online at www.sciencedirect.

com

Chemical Engineering and Processing 47 (2008) 1504–1508

Determination of pressure drop in wire mesh mist eliminator by CFD


R. Rahimi ∗ , D. Abbaspour
Department of Chemical Engineering, Faculty of Engineering, University of Sistan and Baluchestan, Iran
Received 24 February 2007; received in revised form 26 June 2007; accepted 26 June 2007
Available online 22 August 2007

Abstract
With this article, pressure drop was predicted in a mist pad by CFD and compared with the available experimental data and empirical model.
A CFD study was made to determine the pressure drop of wire mesh mist eliminator as a function of broad ranges of inlet velocity 1–7 m/s. The
simulation was carried out in an industrial scale layered type demister pad made of stainless steel wires. These variables were measured as a
function of vapor velocity (1–7 m/s), packing density (200 kg/m3 ), pad thickness (200 mm), wire diameter (0.31 mm) and distance between mesh
wires of (5 mm). The numerical solutions were carried out using spreadsheet and commercial CFD code FLUENT 6.0. The CFD simulations predict
excellently the wire mesh mist eliminator pressure drop under inlet velocity with a maximum deviation of 14% from the Frashband’s refinery data
and 21% from empirical formula. The efficiency of the demister is at highest at the gas inlet velocity of 6 m/s.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Demister pad; CFD; Pressure drop; Inlet velocity; Efficiency

1. Introduction El-Dessouky [1], Feord et al. [4], Buerkholz [5], Belden [6],
Bhatia [7], Tennyson [8], Holmes and Chen [9], Zhao et al. [10]
Wire mesh mist eliminator, in the most general sense, is a are amongst a few who have studied demister conceptually and
simple porous blanket of metal or plastic wire retains liquid experimentally. The main objective of primordial investigators
droplets entrained by the gas phase. As the gas passed the mist of wire mesh mist eliminator system is comparing the separation
eliminator, droplets impinged on the extensive surface of the efficiency of the wire mesh with that of other mist eliminator
wire, retained until they coalesce into large drops. When liquid units, though El-Dessouky [1] has presented an empirical model
drops reach sufficient size, break away from the wire mesh and for presser drop for the wire mesh demister pad.
fall back against the rising gas stream [1]. In this study, pressure drop calculations are performed using
Demisters are widely used in gas–liquid operations such as CFD and compared with empirical model of El-Dessouky [1].
distillation, absorption, evaporation. They are expected to have This empirical model and CFD prediction are compared with
low cost, low pressure drop and high efficiency for the reten- the Frashband’s refinery data and also dropping of liquids from
tion of liquid droplets entrained by the gas phase. Resistance demister pad as a second target, presented in the literature. In
to flooding whilst having high capacity is also a requirement of this study, the CFD calculations are carried out using commercial
demisters [2]. finite volume code FLUENT 6.0 and the empirical models are
The performance of wire mesh eliminators depends on many performed in Microsoft Excel spreadsheet. The pressure drop
design variables such as supporting grids, vapor velocity, wire data for this case reached from refinery whom were used instru-
diameter, packing density, pad thickness and material of con- ment PD cell sensor for stabilizer tower at the top and the bottom
struction. Due to flexibility of wire mesh their supporting grids of demister pad of ethylene glycol tower1 .
should have a high free passage for lowering pre mature flood- It is good to insists that CFD simulation for demister pad
ing. The required free passages for wire mesh and for the grids causes difficulty because of tiny diameter of wire mesh and it is
are 98 and 90%, respectively [3]. so hard to simulate because of limitation of computing and in
some cases complex structure of wire meshes.

∗ Corresponding author.
E-mail address: Rahimi@hamoon.usb.ac.ir (R. Rahimi). 1 Farashband’s gas refinery, Fars Province, Iran.

0255-2701/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.cep.2007.06.016
R. Rahimi, D. Abbaspour / Chemical Engineering and Processing 47 (2008) 1504–1508 1505

Even though reports of CFD simulation of wire meshes done The five constant of the model that were set to the standard
by some of manufactures are reported, but are not in public reach. k–ε model values are: Cμ = 0.09, σ k = 1.00, σ ε = 1.30, Cε1 = 1.44,
The resent study of Zhao et al. [10] on the van type demister pad Cε2 = 1.92.
is worth to mention.
3. CFD simulation setup and solution strategy
2. Theory
Amongst a few methods of solving the set of governing equa-
In the two-fluid approach, the different phases are treated tions, the finite approaches was chosen as the commercial CFD
mathematically as interpenetration continua. The derivation of software package, FLUENT 6.0, which is based on the finite
the conservation equations for mass, momentum and energy for volume approach, was available.
each of the individual phases are done by ensemble averaging FLUENT provides the flexibility in choosing discretization
the local instantaneous balances for each of the phases [11]. schemes for each governing equation. The discretized equations,
The general form of continuity equation is: along with the initial and boundary conditions, were solved
∂(rk ρk ) ∂(ρk rk Uki ) ∂(Γk ρk (∂rk /∂Xj )) using the segregated solution method to obtain a numerical solu-
+ = + Sk (1) tion. Using the segregated solver, the conservation of mass and
∂t ∂Xj ∂Xj
momentum were solved iteratively and a pressure-correction
where ρk is the density, rk the volume fraction, Uki the mean equation was used to ensure the conservation of momentum and
velocity in the i direction of phase k, Γ k the turbulent diffusivity conservation of mass. The k–ε model was used to treat turbu-
and Sk is a mass source for phase k. Assuming no source term lence phenomena in both phases. In this simulation, pressure
exists and neglecting mass transfer, the continuity equation at and temperature of the inlet gas to tower was fixed at 100 atm
steady state condition simplifies to: and 55 ◦ C, respectively. The gas entrained with liquid droplets
∂(ρk rk Uki ) of about 0.1 mm. Physical properties of the liquid and saturated
=0 (2) gas are given in Table 1.
∂Xj
With liquid phase contains of water, propane, butane, pentane
The momentum conservation equation is: to heavy component C16 + . The structure used in this simulation
∂(rk ρk Uk ) is shown in Fig. 1. The ethylene glycol tower consist of five
+ ∇(rk ρk Uk Uk ) trays filled with ethylene glycol to remove the liquid vapors of
∂t
the gas. The area shown in the figure is from the gas chamber to
= rk ∇p − ∇(rk τk ) + rk ρk g + Fk + Fg (3)
the bottom of the first tray. This height is 6 m whilst the height to
In Eq. (3), Fk represents the inter phase momentum exchange the demister is 3.21 m. The area above the demister is also taken
terms between phase k and the all other phases present in system. into simulation to check its effect on the demister performance.
Fg include the terms of the momentum source and dispersion of Tower diameter is 2.5 m.
the multiphase flow relation. The heat of dispersion and disso- This simulation carried out in an industrial scale layered type
lution of liquid in gas have been neglected. demister pad made of stainless steel wires. The pressure drop
The standard k–ε turbulence model used in all simulations. across the demister pad were measured as a function of vapor
Two equations of this model, the turbulent kinetic energy, k, and velocity (1–7 m/s), packing density (200 kg/m3 ), pad thickness
the turbulent dissipation rate are: (20 cm), wire diameter (0.31 mm) and the distance between
∂ ∂ ∂ wires of (5 mm).
(rk ρk kk ) + (rk ρk uk,i kk ) − A first-order upwind discretization scheme was used for the
∂t ∂xi ∂xi
    momentum equations whilst a first-order upwind discretization
μk,tur ∂kk was sued for volume fraction, turbulent kinetic and turbulent
× rk μk,lam + = rk (Gk − ρk εk ) + Sk,k
σk ∂xi dissipation energy.
(4) In the present calculations, the threshold values were set
  to a thousandth of the initial residual value of each variable.
∂ ∂ ∂ The PRESTO and SIMPLE algorithms are used to solve the
(rk ρk εk ) + (rk ρk uk,i εk ) − rk μk,lam
∂t ∂xi ∂xi coupling between velocity and pressure and Schiller–Newman
  method used for drag coefficient. The numerical computation
μk,tur ∂εk εk
+ = rk (Cε1 Gk − Cε2 ρk εk ) + Sk,ε (5)
σk ∂xi kk
Table 1
The standard model is taken without any further modifications. Properties of phases at T = 55 ◦ C and P = 100 atm
The source terms Sk,ε and Sk,k on the right-hand side of the
Gas Liquid
equations are not considered in the mean of interphase turbu-
lence exchange. By using the standard k–ε model the turbulent Phase fraction 0.9964 0.0036
viscosity of the continuous phase is calculated by ρ (kg/m3 ) 81 931.8
Cp (kJ/(kg K)) 2.179 3.775
kk2 μ (cP) 0.001638 0.5374
μk,tur = Cμ ρk (6) W (W/(m K)) 0.004751 0.587
εk
1506 R. Rahimi, D. Abbaspour / Chemical Engineering and Processing 47 (2008) 1504–1508

Fig. 3. Evolution of pressure drop with inlet velocity. Comparison between data
presented by refinery and the predictions of CFD and data presented by empirical
model of El-Dessouky [1] (P = 100 atm, T = 55 ◦ C).

4. Result and discussion


Fig. 1. Wire mesh configuration. Boundary conditions used for the simulation:
(1) velocity inlet, (2) pressure outlet, (3) demister pad (wall) and (4) wall.
Measurement of the demister pressure drop was carried out
for inlet velocity ranging of 1–7 m/s in the refinery. The compar-
was considered converged when the residual summed over all ison between the refinery experimental data which are collected
the computational nodes at nth iteration have difference about for this special case and CFD prediction are shown in Fig. 3.
10−5 to previous for P and V. The liquid diameter set con- The CFD numerical calculations were compared with the
stant as 10−1 mm. In Fig. 2 structured meshes are showed. experimental data, and El-Dessouky [1] empirical correlation.
Meshes are quadric with interval size of 0.1 mm. The num- As Fig. 3 shows, CFD prediction are comparable with the refin-
bers of quad meshes are 113,357,621. The geometry created ery data, even though CFD predictions are about 14% lower than
using Gambit. This small interval size is necessary because the experimental data. But for the empirical formula this devi-
our demister had very small edges and those edges have to ation after velocity of 5 m/s gets harder and reaches to 21% in
be meshed. Interval sizing of less than 0.1 mm is not used average. This deviation is because of carrying out of liquids from
because of just taking more CPU time, whilst interval sizing the demister pad at higher inlet velocity by gas and these liquids
of more than 0.1 mm do not cover the thickness of the wires, close holes of demister and cause more pressure drop which is
too. not accounted for by the El-Dessouky’s equation El-Dessouky’s
The boundary condition pressure outlet is place of the first [1] models show a good prediction on demister pressure drop
tray counted from the demister. The volume between the demis- under low operational inlet velocity. However, their model is
ter and the tray is used for the prediction of the flow regime unable to predict accurately the pressure drop at high velocities.
above of the demister. As another result in this paper accumulation of collected droplets

Fig. 2. Meshed structure—notice that the rectangular are metal wires of 0.31 mm in diameter. Meshed size is 0.1 mm and the distance between wire rows are 5 mm.
The entire calculation domain is shown.
R. Rahimi, D. Abbaspour / Chemical Engineering and Processing 47 (2008) 1504–1508 1507

Fig. 4. Counter of density, accumulated liquid at the bottom of tower as predicted


by CFD, gas velocity 6 m/s.
Fig. 5. Counters of fluid density, prediction of liquid separation by CFD, gas
are predictable. This can lead to the calculation of demister sep- velocity 6 m/s.
aration efficiency. Entered liquids to the demister as droplets,
after coalesced and reached certain size would detach from the tends to exit from a certain areas of the demister outlet, but at
demister pad and were accumulated at the bottom of the gas higher velocities more uniformity are reached.
chambers. Fig. 4 which is a contour of liquid density clearly The demister pad efficiency, η, is the separation of liquid from
shows the accumulated liquid that is not passed the demister gas that occurs due to the effect of demister pad. This is defined
pad. This effect is further pronounced as Fig. 5 shows the fluid by Eq. (7).
density in the vicinity of the demister. Fig. 5 is a counter of Min − Mout
detaching fluid density for gas velocity of 6 m/s. Variation of η= (7)
Min
this phenomena for the other gas velocity are not shown for
the sake of brevity. As is clear from Fig. 5 the gas chamber is Min and Mout are the mass flow rates of entering and leaving
an aid to the demister as some of the liquid due the change of the specific area, respectively. The calculated efficiencies as a
momentum that occurs in the gas chamber are separated from function of gas velocity are shown in Fig. 7.
the entering gas. The efficiency increases steadily by increase of gas velocity
Fig. 6 is a vector plot of mixture velocity for the gas inlet up to velocity of 6 m/s and then decreases. The increase of effi-
velocities of 1 and 6 m/s. At low gas inlet velocities eddies in ciency by increase of velocity is due to the increase of uniformity
the gas chamber are detectable. These eddies are increasing as of gas outlet from the demister, but at higher velocities a carry
gas inlet velocities are increased. At the low gas inlet the gas over of gas lowers the efficiency.

Fig. 6. Vectors of velocity in simulated tower by CFD.


1508 R. Rahimi, D. Abbaspour / Chemical Engineering and Processing 47 (2008) 1504–1508

P pressure, Pa
rk the volume fraction of phase k
Sk mass source term for phase k, kg/m3
t time, s
Uk flow velocity, m/s
Uki flow velocity in i direction, m/s
W thermal conductivity, W/m K
Xj spatial coordinate, m

Greek symbols
Γk turbulent diffusivity of phase k, m2 /s
ε turbulent energy dissipation rate
η separation efficiency
μ viscosity, cP
Fig. 7. Effect of gas velocity on the efficiency of demister pad.
μk,jam laminar viscosity of phase k, cP
μk,tur turbulent viscosity of phase k, Pa s
5. Conclusions σk constant in k–ε model (turbulent Prandtl number for k),
l
The predicted pressure drop in demister by CFD code FLU- τ shear stress
ENT have a good relationship by experimental data and can be
used in demister pad design for any operating conditions. In References
the CFD numerical calculations a small pressure drop devia-
tion were observed, with about 14–21% of deviation between [1] H.T. El-Dessouky, Performance of wire mesh mist eliminator, Chem. Eng.
experimental and empirical method. Process. 39 (2000) 129–139.
The demister pressure drop can be rewritten as a function of [2] E.E. Ludwig, Applied Process Design for Chemical and Petrochemical
inlet velocity. The model used for the prediction of pressure drop Plants, vol. 2, Gulf Publishing Co., Huston, 1977.
[3] P. Fabian, R. Cusack, P. Hennessey, M. Neuman, Demystifying the selection
depends on the demister properties and operating condition. The of mist eliminators. Part I. The basics, Chem. Eng. 11 (1993) 148–166.
optimum velocity for this kind of demister pad was find to be [4] D. Feord, E. Wilcock, G.A. Davies, A stochastic model to describe the
6 m/s. operation of knitted mesh mist eliminators, computation of separation
efficiency, Trans. IChemE Res. Des. 71 (1993) 282–295.
Appendix A. Nomenclature [5] A. Buerkholz, Droplet Separation, VCH, New York, 1989.
[6] R.D. Belden, Mist eliminators in evaporates and pans, in: Proceedings of
SIA Meeting, Savannah, GA, USA, May 1988, pp. 219–229.
[7] M.V. Bhatia, Entrainment separators, in: Process Equipment Series, VCH,
C1ε ; C2ε ; Cμ constant in k–ε turbulence model New York, 1989.
Fg sum of momentum source and dispersion of multi- [8] R.P. Tennyson, Mist eliminator design and application, in: Proceedings of
phase, kg/(m3 s) the 70th Annual Meeting of the Air Pollution Control Association, Toronto,
Canada, June 1977, pp. 20–24.
Fk momentum exchange parameter, kg/(m3 s) [9] T.L. Holmes, G.K. Chen, Design and selection of spray: mist elimination
g gravitational acceleration, m/s2 equipment, Chem. Eng. (1984) 82–86.
k turbulent kinetic energy, m2 /s2 [10] J. Zhao, B. Jin, Z. Zhong, Study of the separation efficiency of a demister
Min mass of entrained liquid droplet by the upstream of mist vane with response surface methodology, J. Hazard. Mater. 147 (2007)
eliminator, kg/s 363–369.
[11] T.B. Anderson, R. Jackson, The nature of aggregative and particulate flu-
Mout mass of entrained liquid droplet by the downstream of idization, Chem. Eng. Sci. 19 (7) (1964) 509–511.
mist eliminator, kg/s

You might also like