Beruflich Dokumente
Kultur Dokumente
Jia-Lin Kang a , Ya-Cih Ciou b , Dong-Yang Lin b , David Shan-Hill Wong b,∗ ,
Shi-Shang Jang b,∗
a Department of Chemical and Material Engineering, TamKang University, New Taipei City, Taiwan, ROC
b Department of Chemical Engineering, National Tsing Hua University, Hsinchu, 30013, Taiwan, ROC
a r t i c l e i n f o a b s t r a c t
Article history: Computational fluid dynamics (CFD) simulations of countercurrent gas–liquid flow in ran-
Received 6 October 2018 dom packings of Raschig rings were carried out in this study. This model used gravity
Received in revised form 22 April simulation to construct the random packing structure, and a volume expansion-recovery
2019 method to improve the meshing quality. A simple feedback control scheme was applied to
Accepted 29 April 2019 control the gas inlet flow rate so that pressure drop can be estimated. The generated char-
Available online 7 May 2019 acteristics of the packing structure such as the number of packing elements, dry surface
area and porosity were found to be close to experiments. CFD predictions of hydrodynamics
Keywords: properties are also validated by the experimental data using a small column section. The
CFD simulation results indicate that our CFD model was able to capture the essential hydrodynamics behav-
Random packing ior of the gas–liquid countercurrent flow. Hence, the approach presented in this study can be
Multiphase used as a basis for studying the effect of detailed packing-geometry design on hydrodynamic
Automatic stacking and mass transfer characteristics.
Hydrodynamic behaviors © 2019 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
1. Introduction uid flow. The model can predict fairly well pressure drop, liquid hold-up
and wetted area of countercurrent gas–liquid flow in various structural
Packings are widely used in distillation, absorption, and stripping packings. Van Baten and Krishna (2002) presented using separated flow
columns. Effective mass transfer in packing reduces the height of a channels of the gas and liquid phases and interface between two chan-
column and leads to a reduction in capital cost. Alternatively, using a nels to study the mass transfer in KATAPAK-S structural packing. The
better packing can increase the number of transfer units, and hence, result showed that the predictions of Sherwood numbers of gas and
reduce the reflux in a distillation column and the absorption solvent liquid phases. The gas-phase Sherwood number agrees with empiri-
usage leading to reduced operating costs and increased throughput cal correlation, but the liquid phase Sherwood number is significantly
in revamping. Development of packings used to be empirical, relying lower. Valluri et al. (2005) adopted volume of fluid (VOF) to simulate
on experiments to collect data on the height of mass transfer units, the dynamic evolution of waves on laminar falling wavy films at low
liquid holdup, interfacial area, and pressure drop. Empirical correla- to moderate Reynolds numbers over corrugated surfaces in an REU of
tions of these data were developed. It is desirable that computational Mellapak-500.Y. Chen et al. (2009) reported that the VOF method devel-
fluid dynamics can be used to enable a more rational design of these oped a 3D gas–liquid countercurrent flow model for simulating the
packings. hydrodynamics and mass-transfer behavior in a typical REU of struc-
There have indeed been many studies using computational fluid tural packing. The gas inlet and the liquid outlet were on the same
dynamics (CFD) used to investigate the hydrodynamic behavior of side, whereas the gas outlet and the liquid inlet were on the same side.
structured packing. Structural packing can be easily simplified into a The result showed that the model can predict the liquid holdup, wet-
representative elementary unit (REU) to reduce computation load. Iliuta ted area, and separation efficiency. Haroun et al. (2012) investigated
and Larachi (2001) presented a one-dimensional two-fluid model of two mass transfer from the gas phase to the liquid along with a corru-
parallel inclined and interconnected slits to simulate the gas and liq- gated metal sheet with liquid and gas flowing in the same direction.
∗
Corresponding authors.
E-mail addresses: dshwong@che.nthu.edu.tw (D.S.-H. Wong), ssjang@mx.nthu.edu.tw, ssjang@che.nthu.edu.tw (S.-S. Jang).
https://doi.org/10.1016/j.cherd.2019.04.037
0263-8762/© 2019 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
44 Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 43–54
This was similar to a bridging approach. Partopour and Dixon (2017) where l and g are the viscosities of the liquid and gas,
presented the shrink-wrap method with an automated package for respectively. The continuity equation of the mixture fluid is
packed bed generation for handling the meshing challenges. The result expressed as follows:
showed that the shrink-wrap method can speed up the process of
the geometry and mesh generation while it maintains the structural ∂
(mx ) + ∇ =0
mx U (6)
features of the bed. Boccardo et al. (2019) investigated particle trans- ∂t
port and deposition in different catalytic systems using the geometric
models representing different porous media. The geometric models The momentum equation was modified as follows:
were a number of random packings of spheres created by rigid body
simulations. The result showed that the calculated particle deposi- ∂
+∇
U
mx U
= −∇P
+ t ∇
U
mx U
tion efficiency using the CFD matched well with the classical filtration ∂t
theory. U
+∇ mx g + F s
T + ∇ (7)
The aforementioned studies were usually applied to trickle bed
reactors with single phase flow or using two-phase flow in a co-current
where P is the pressure, t is the turbulence viscosity, g is the
manner. Hence the boundary condition at the inlet can be easily set
force of gravity, and F s is any additional body force.
by defining the mass flow or velocity of a given distribution geome-
try. An open boundary condition with given pressure can be ascribed The k − ε model was employed as the turbulence equa-
to the outlet. In this way, the pressure drop can be calculated. In a tion (Launder and Spalding, 1974). Turbulent kinetic energy,
structural packing, gas and liquid flow are relatively stratified. To sim- k, and turbulence dissipation rate, ε, were used to estimate
ulate countercurrent flow in a structural packing, Haroun et al. (2014) the Reynolds stress in the Reynolds-Averaged Navier–Stokes
used uniform velocity to inject liquid at the inlet surface and assumed equations (Kajishima and Taira, 2017). The equations are
a uniform negative gas velocity at the same inlet-surface. An open expressed as follows:
pressure-outlet surface was used. However, in countercurrent flow
through a random packing, we should not assume in a priori how the ∂
t
(mx k) + ∇ =∇
· (mx Uk) · mx +
∇k + Gk − mx ε (8)
air will flow out of the liquid inlet side. ∂t k
The purpose of this study was to present a CFD approach that can
be used to simulate a representative random packing section that can ∂
t
(mx ε) + ∇ =∇
· (mx Uε) · mx +
∇ε
study the effect of the detailed geometry of the packing elements. In ∂t ε
this approach, we have adopted a packing generation procedure that
can approximate the porosity of the actual packing but retain good ε2
+ C1ε Gk − C2ε mx (9)
meshing quality. A new method of setting boundary conditions for k
countercurrent gas–liquid flow was presented. Experimental data were
where k is the turbulent kinetic energy, ε is the turbulence
collected using a laboratory packed column to validate the simulation
results.
dissipation rate, C1ε and C2ε are numerical constants, and k
and ε are the effective turbulent Prandtl numbers for k and
2. CFD modelling approach ε, respectively. The term Gk is the generation of turbulence
kinetic energy owing to the mean velocity gradients and is
defined as follows:
2.1. Governing equations
Gk = t ∇ −∇
·U T
·U (10)
The VOF method in the multiphase flow model of Ansys
®
Fluent , a commercial CFD simulator was used. The standard
where U means the velocity vector. The turbulence viscosity,
k − ε method was adopted in the turbulence model. The gov-
t , is a function of k and ε, as shown below:
erning equations are described as follows.
The continuity equation of the water phase is given by
k2
t = mx C (11)
∂ ε
(˛ ) + ∇ =0
˛l l U (1)
∂t l l The model constants C1ε , C2ε , C , k , ε in the above equa-
®
tions are the default values of Ansys Fluent , which were
where ˛l and l are the volume fraction and density of the
provided by Launder and Spalding (1974).
liquid phase, and U is the velocity of the multiphase fluid
The additional body force considered in this work includes
mixture. The volume fraction of the liquid phase obeys the
surface tension forces between the gas–liquid interface which
transport equation:
was calculated by continuum surface force (Ataki and Bart,
∂˛l 2006; Brackbill et al., 1992) model as follows:
+∇ l = ˛l ∇
U˛
˙· U (2)
∂t
mx kn
F gl = gl 1
(12)
The volume fraction of the air phase is defined as the void 2 (g + l )
not occupied by water:
where gl is the surface tension coefficient between the gas
is the unit normal of the gas–liquid interface given
liquid and n
˛g = 1 − ˛l (3)
by
Ap,s Np,exp
ap,exp = (18)
Vcol
Fig. 2 – Schematic diagram of interval establishment. The dot pattern indicates the added volume.
2.3. Meshing quality marking in the blue of 5 mm diameter at the top. The 4 circu-
lar areas were evenly distributed around the range of 12 mm
If the packing rings are in contact, there will be direct contact diameter from the origin of the top area. The rest of the areas,
between the packing elements, a dead zone will be created, marking in the green, at the top were set as the outlet.
and a mesh grid cannot be generated by the ANSYS default The velocity at this inlet boundary was set to provide the
mesh generator. To overcome this bottleneck, the outer diam- specified liquid flow rate Lsp . Other parts of the top surface
eter and height of the packing were enlarged from 6 mm to were set as a pressure outlet with a fixed pressure of 0 Pa. The
6.3 mm, and the diameter of the absorber was reduced from entire bottom surface marking in the red was set as a pressure
25 mm to 24.7 mm before the automatic stacking procedure. outlet. For a pressure outlet, it is possible that the velocity is
After stacking, the packing and absorber were returned to negative, i.e. materials are coming in instead of going out. In
their original sizes to generate interval between packings. The such cases, we must designate what is the phase condition
schematic diagram of interval establishment is shown in Fig. 2. of the material flowing in. We assume that only gas can flow
We called this method “expansion and recovery”. into the section at both pressure outlets, i.e. the gas phase is
Dixon et al. (2013) reported that global enlargement or designated as the back flow phase. The net gas flow, G, into the
shrinkage methods would lead to large porosity changes, bottom pressure outlet can be calculated by the dot product of
resulting in substantial errors in the calculation of drag coef- the average velocity vector and the bottom surface-area vector
ficients. Local corrections at the contact points are preferred. with an outward normal vector:
However, in a random packing, contact points are random, cre-
ating caps or bridges at the contact point are cumbersome. In
elements at the bottom outlet
G=− n j ˛g,j
bot,j Abot,j · U (19)
our expansion and recovery method, all packings are returned
to its original size. While there will be inevitable disturbance j=1
Fig. 3 – (a) Schematic diagram of the approach presented by this study, (b) boundaries used by this work, (c) simulated
pressure distribution, and (d) simulated inlet gas streamlines.
Fig. 4 – Dynamic changes of (a) inlet gas flow rate, (b) bottom pressure, (c) the gas mass balance and (d) the liquid mass
balance at L/G = 0.5.
changes in the gas and liquid mass balance errors are shown ditions of the convergence were set as 10−3 of the absolute
in Fig. 4c and d. residual for all monitoring parameters including continuity,
velocity, volume fraction of phases, and k and ε for viscous
2.5. Convergence and termination model. The simulations started from an initial transient state
in which the packing was filled with a stationary gas phase,
A workstation (18 cores/CPU × 4 CPUs, 2.2 GHz of Intel Xeon) and liquid flowed from the top inlet until a steady state is
was used in this study. The time-step size set as 0.001 s while reached. Steady state is characterized by the fact that gas and
solving a maximum of 20 iterations per time step. The con- liquid mass balance errors become stable and close to zero;
Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 43–54 49
P̄ (Pa) 0.28
KP (Pa/m3 s−1 ) 60,000
KI (Pa/m3 ) 500
KD (Pa/m3 s−2 ) 120
ae Agl
= (21)
aP Atot, packing
˛ ıV l,j
V j l,j
εl = l = (22)
Vtot Vtot
Where Vl and Vtot are the volume of the liquid and the total
volume of the flow field, respectively, and ˛l,j and ıV l,j are the
liquid volume fraction of the jth mesh and the volume of the
jth mesh, respectively.
dP
The pressure drop, dh , is estimated by
dP P − Pt
= b (23)
dh hcol
Where Pb and Pt are the pressures of the bottom and top outlet,
Fig. 5 – Iso-surface for air volume fraction = 0.5. respectively.
5. Conclusions
The average and standard deviations of the steady state time- Acknowledgment
average pressure drop were compared with the pressure drop
of the experiment, and the correlation by Stichlmair et al. The authors acknowledged the financial support provided by
(1989) in Fig. 9. The result showed that the pressure drop of the Ministry of Science and Technology through the grant MOST
CFD simulations was lower than the experimental pressure 105-2221-E-007-141.
drop but was agreed with the pressure drop of the correlation.
At low L/G, the difference in the pressure drops of the CFD sim- Appendix A. Effect of length in the flow field
ulation and experiment was minor, but as the L/G increased to
greater than 0.3, the difference in the pressure drop between As Fig. A1 shows, the authors have carried out simulations of
the CFD simulation and experiment became larger. The exper- five heights: 10, 16, 22, 28, and 35 mm at a liquid flow rate of
imental pressure drop raised rapidly, while the CFD simulation 0.3 l/min, and a gas flow rate of 1 l/min. As figures shown, the
and correlation remained in low-pressure drops. The dif- interfacial area and liquid holdup can be reached stable after
ference between the CFD simulation and experiment was the height was higher than 22 mm. Thus, 22 mm height of the
caused by packing supporter in the small scale experimen- flow field was selected as the investigating length to reduce
tal equipment. The supporter was a porous disc. When the the computational cost.
52 Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 43–54
Fig. A1 – Effect of (a) fractional area, (b) liquid holdup, and (c) pressure drop in various lengths of the flow field.
Appendix B. Effect of diameter in the flow field to prove that the approach presented in this study can be
used in larger scales. The results showed that the interfa-
As Fig. B1 shows, the authors ran simulations of three diame- cial area, liquid holdup, and pressure drop are not affected in
ters: 25.4, 50.8, and 101.6 mm at the liquid flow rate of 0.3 l/min, the flow fields with different diameters. Thus, 22 mm diame-
and the gas flow rate of 1 l/min to investigate the differ- ter of the flow field was selected to reduce the computational
ences of the interfacial area, liquid holdup, and pressure drop cost.
Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 43–54 53
Fig. B1 – Effect of (a) fractional area, (b) liquid holdup, and (c) pressure drop in various diameters of the flow field.
tube/particle diameter ratios in the laminar flow regime. Haroun, Y., Raynal, L., Legendre, D., 2012. Mass transfer and
Chem. Eng. J. 155, 404–410. liquid hold-up determination in structured packing by CFD.
Atmakidis, T., Kenig, E.Y., 2012. Numerical analysis of mass Chem. Eng. Sci. 75, 342–348.
transfer in packed-bed reactors with irregular particle Haroun, Y., Raynal, L., Alix, P., 2014. Prediction of effective area
arrangements. Chem. Eng. Sci. 81, 77–83. and liquid hold-up in structured packings by CFD. Chem. Eng.
Bai, H., Theuerkauf, J., Gillis, P.A., Witt, P.M., 2009. A coupled DEM Res. Des. 92, 2247–2254.
and CFD simulation of flow field and pressure drop in fixed Iliuta, I., Larachi, F., 2001. Mechanistic model for
bed reactor with randomly packed catalyst particles. Ind. Eng. structured-packing-containing columns: irrigated pressure
Chem. Res. 48, 4060–4074. drop, liquid holdup, and packing fractional wetted area. Ind.
Billet, R., Schultes, M., 1993. Modelling of packed tower Eng. Chem. Res. 40, 5140–5146.
performance for rectification, absorption and desorption in Kajishima, T., Taira, K., 2017. Reynolds-Averaged Navier–Stokes
the total capacity range. The 3rd Korea-Japan Symposium. Equations, Computational Fluid Dynamics. Springer, pp.
Billet, R., Schultes, M., 1999. Prediction of mass transfer columns 237–268.
with dumped and arranged packings: updated summary of Launder, B.E., Spalding, D., 1974. The numerical computation of
the calculation method of billet and schultes. Chem. Eng. Res. turbulent flows. Comput. Methods Appl. Mech. Eng. 3,
Des. 77, 498–504. 269–289.
Boccardo, G., Augier, F., Haroun, Y., Ferre, D., Marchisio, D.L., 2015. Liu, J., Gao, H.-C., Peng, C.-C., Wong, D.S.-H., Jang, S.-S., Shen, J.-F.,
Validation of a novel open-source work-flow for the 2015. Aspen plus rate-based modeling for reconciling
simulation of packed-bed reactors. Chem. Eng. J. 279, laboratory scale and pilot scale CO2 absorption using aqueous
809–820. ammonia. Int. J. Greenhouse Gas Control 34, 117–128.
Boccardo, G., Sethi, R., Marchisio, D.L., 2019. Fine and ultrafine Lopes, R.J., Quinta-Ferreira, R.M., 2009. CFD modelling of
particle deposition in packed-bed catalytic reactors. Chem. multiphase flow distribution in trickle beds. Chem. Eng. J. 147,
Eng. Sci. 198, 290–304. 342–355.
Brackbill, J., Kothe, D.B., Zemach, C., 1992. A continuum method Ma, Y., Cao, X., Feng, X., Ma, Y., Zou, H., 2007. Fabrication of
for modeling surface tension. J. Comput. Phys. 100, 335–354. super-hydrophobic film from PMMA with intrinsic water
Bravo, J.L., Fair, J.R., 1982. Generalized correlation for mass contact angle below 90. Polymer 48, 7455–7460.
transfer in packed distillation columns. Ind. Eng. Chem. Onda, K., Takeuchi, H., Okumoto, Y., 1968. Mass transfer
Process Des. Dev. 21, 162–170. coefficients between gas and liquid phases in packed
Caulkin, R., Ahmad, A., Fairweather, M., Jia, X., Williams, R., 2007. columns. J. Chem. Eng. Jpn. 1, 56–62.
An investigation of sphere packed shell-side columns using a Partopour, B., Dixon, A.G., 2017. An integrated workflow for
digital packing algorithm. Comput. Chem. Eng. 31, 1715–1724. resolved-particle packed bed models with complex particle
Caulkin, R., Ahmad, A., Fairweather, M., Jia, X., Williams, R., 2009. shapes. Powder Technol. 322, 258–272.
Digital predictions of complex cylinder packed columns. Stichlmair, J., Bravo, J., Fair, J., 1989. General model for prediction
Comput. Chem. Eng. 33, 10–21. of pressure drop and capacity of countercurrent gas/liquid
Caulkin, R., Fairweather, M., Jia, X., Gopinathan, N., Williams, R., packed columns. Gas Sep. Purif. 3, 19–28.
2006. An investigation of packed columns using a digital Valluri, P., Matar, O.K., Hewitt, G.F., Mendes, M., 2005. Thin film
packing algorithm. Comput. Chem. Eng. 30, 1178–1188. flow over structured packings at moderate Reynolds numbers.
Chen, J., Liu, C., Yuan, X., Yu, G., 2009. CFD simulation of flow and Chem. Eng. Sci. 60, 1965–1975.
mass transfer in structured packing distillation columns. Van Baten, J., Krishna, R., 2002. Gas and liquid phase mass
Chin. J. Chem. Eng. 17, 381–388. transfer within KATAPAK-S structures studied using CFD
Dixon, A.G., Nijemeisland, M., Stitt, E.H., 2013. Systematic mesh simulations. Chem. Eng. Sci. 57, 1531–1536.
development for 3D CFD simulation of fixed beds: contact Wang, C., 2015. Mass Transfer Coefficients and Effective Area of
points study. Comput. Chem. Eng. 48, 135–153. Packing, M.S. CHE. University of Texas.