Você está na página 1de 6

Chinese Journal of Chemical Engineering, 19(5) 815ü820 (2011)

CFD Simulation and Experimental Validation of Fluid Flow


in Pre-distributor*

ZHANG Lühong (჆ে‫)ۿ‬1, GAO Guohua (‫)ܟڳغ‬1, SUI Hong (ഺ‫)܃‬1,2, LI Hong (ह‫)܀‬1,2
and LI Xingang (ह᳨‫)ر‬1,2,**
1
School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China
2
National Engineering Research Centre of Distillation Technology, Tianjin 300072, China

Abstract Liquid distributor is a very import internal for distillation columns. Pre-distributor is usually set on the
top of distributor for initial distribution. Fluid flow in pre-distributor is a complex system of variable mass flow
with many orifices and sub-branches. Consequently, the two phase modeling of pre-distributors was carried out and
the homogeneous model with free surface model was applied. The numerical method was validated by comparing
with experimental data. Using the simulated results for different pre-distributors, the impacts of inflow rate, location
and orientation upon the outflow distribution were investigated. Furthermore, influences of the outflow distribution
for pre-distributor on liquid uniformity in trough were also analyzed. The conclusions can be adopted for the struc-
tural design of liquid distributor and pre-distributor of large scale.
Keywords pre-distributor, computational fluid dynamics, orifices, trough, outflow distribution

1 INTRODUCTION effects of flow status upon outflow uniformity.


Therefore, the outflow through the orifices and
Liquid distributor is one of the most important the flow characteristics of pre-distributors were inves-
internals placed on the top of packings. It allows the tigated numerically using CFD method. Three-
feed liquid and return liquid to evenly distribute on dimensional (3D) two-phase flow simulation was car-
packing surface in order to form initial distribution. ried out with Euler-Euler formulation provided by
There are many different types and structures for liq- ANSYS CFX for multiphase calculations. A VOF-like
uid distributors, but trough liquid distributor is the method [6] named the free surface model was introduced
most commonly used in industry [1]. Typically, to this algorithm for the calculation of free surface
pre-distributor is expected to set above trough liquid flows. As the first part of the study, the simulated out-
distributor to ensure better distribution of liquid. flow distribution through orifices was compared with
Fluid flow in pre-distributor is a system of vari- experimental data to verify the numerical method. Then,
able mass flow including many orifices and sub- the effects of flow rate, feed direction upon the flow
branches with complex flow conditions. Identification field, and the outflow distribution within pre-distributor
of the flow pattern in the pre-distributor is critical for were discussed. Additionally, pre-distributors connected
better understanding the mass transfer in the distilla- with a distribution trough were modeled to study the
tion column and is beneficial for the column structure influence of two-phase flow field in pre-distributor on
optimization. The outflow distribution could be ob- the flow pattern and the outflow distribution of the
tained by experiments, however, the flow characteris- trough. All the conclusions gained from the CFD
tics inside the liquid distributor are difficult to capture modeling can be used to guide structural optimization
by experimental methods due to the difficulty of of industrial distributors.
measurement. In addition, it is difficult and expensive
to guide structural design of fluid distributors in large 2 OBJECTS OF SIMULATION
scale by experiments. The development of theory and
software for computational fluid dynamics (CFD)
renders it possible to acquire inside flow status so that Three kinds of pre-distributor (Fig. 1) were mod-
these problems can easily be solved by numerical eled. Pre-distributor I [Fig. 1 (a)] was a straight pipe
simulation method. However, there are only a few with orifices for water outflow. As for this
research reports [15] and still many deficiencies about pre-distributor, two pipes with different sizes and dif-
the CFD study of pre-distributor. For example, liquid ferent numbers of outlets were concerned in this study.
single-phase flow was mostly used for the CFD mod- One was 4 m long with diameter of 100 mm and 20
eling with simulation results biased to experimental orifices of M22 mm, which was used for the validation.
data; only single orifice or small number of orifices The other one of industrial size was 10.3 m long with
were studied. Moreover, studies are needed on the diameter of 203 mm and 12 orifices of M55 mm. In
flow field within the pre-distribution pipe and the fact, short pipes are usually applied as liquid outlets in

Received 2011-06-12, accepted 2011-08-31.


* Supported by the National Key Basic Research and Development Program of China (2009CB219905, 2009CB219907), and
the Program for Changjiang Scholars and Innovative Research Teams in Universities (IRT0936).
** To whom correspondence should be addressed. E-mail: lxg@tju.edu.cn
816 Chin. J. Chem. Eng., Vol. 19, No. 5, October 2011

(a) Pre-distributor with outlet orifices

(b) Pre-distributor with outlet pipes

(c) Pre-distributor and troughs


Figure 1 Different computational subjects for two-phase flow modeling of pre-distributor

order to ensure accurate falling points. Pre-distributor II 3 CFD SIMULATION MODEL


[Fig. 1 (b)] was a combination of a 11.5 m long
pre-distributor with 10 outlet pipes of 240 mm diame- There are two different sub-models, the homo-
ter and a trough with a cross section of 0.72 m width geneous model and inhomogeneous model, available
by 0.75 m depth. In addition, 8 simplified shunts were for Euler-Euler multiphase flow. The former is a
set up on both two sides of the trough as water outlets VOF-like multiphase flow model developed for the
with 0.3 m width and 0.35 m depth. calculation of free surface flows. Free surface model
Pre-distributor III [Fig. 1 (c)] was used to study [6] was introduced to the homogeneous model for the
the affect of flow characteristics in pre-distributor on two-phase flow modeling of pre-distributor.
liquid distribution of the trough, which was also 11.5 The conservation equations based on the
m long with only 4 outlet pipes of 229 mm diameter Euler-Euler homogeneous model are listed below:
and fed water to 4 distribution troughs at their center Continuity equation:
points. These distribution troughs located symmetri-
cally on both sides of the main feed pipe with different wU
 ’ ˜ ( UU ) 0 (1)
numbers of outlet pipe and same size for cross section wt
of 0.6 m width by 0.45 m depth, the two troughs away
from the main feed pipe was 7.78 m long and had 17 Momentum equation:
short pipe outlets while the two troughs near the The homogeneous model for momentum trans-
main feed pipe was 10.16 m long and had 21 short port assumes
pipe outlets. UD U, 1İ D İ Np (2)
Liquid distributor in industrial scale has so large
pipe size but so small outlet orifices that it is difficult and the momentum equation:
to capture the structure of the small orifices with a w
limited number of meshes. ANSYS ICEM CFD was
adopted for meshing with the octree mesh method. wt ¬
( UU )  ’ ˜ ª UU … U  P ’U  (’U )T º
¼ S M  ’p

This method is based on an algorithm that ensures (3)


refinement of the mesh where necessary, but maintains with
larger cells where possible, allowing for faster com-
Np Np
putation. The mesh numbers for these three kinds of
computational subjects are 500, 1000, 1800 thousands, U ¦ rD UD , P ¦ rD PD (4)
respectively. D 1 D 1
Chin. J. Chem. Eng., Vol. 19, No. 5, October 2011 817

Volume conservation equations :


Np

¦ rD 1 (5)
D 1
Pressure constraint:
pD p (6)
For free surface flow, the following equation is
used for interfacial area density:
2 ’rD ’rE
ADE (7)
’rD  ’rE
Figure 3 Compare of simulated mass flow through the
If there are just two phases in the simulation, it is outlet orifices and experimental data
ƵCFD results;ƽexperimental data
simplified to
ADE ’rD (8)
still exist especially at the end of pre-distributor. It is
The standard k-İ turbulence model was used for assumed that the disagreement is particularly caused
the turbulence simulation. As for boundary conditions, by the remaining air bubbles in the vessel that is un-
velocity for water inlet, pressure for outlet and avoidable during the test setup. On the other hand, the
non-slip setting for walls were applied in accordance numerical model employed in this study focuses on
with operation conditions. As for the initial condition, the liquid phase, but the gas phase is neglected by de-
computational domains were filled with immobile air. activating the cells containing gas only.

4 MODEL VALIDATION 5 RESULTS AND DISCUSSION

Experiments were carried out to validate the per- 5.1 Liquid feed rate
formances of modelling including physical model,
multiphase model and boundary conditions setting. Kister [7] pointed out that variable mass flow
Fig. 2 shows the setup for hydrodynamics test of with multi-orifices can be divided into five general
Pre-distributor I. The volume flow rate at the liquid flow patterns due to different inlet flow rates for the
inlet was 14 m3·h1 (UL 0.5 m·s1) and the corre- same flow channel. Therein, the outflow increases
sponding Reynolds number was about 5.6×104. Drain- from the first orifice to the last one for large inflow
age pipe was used to prevent liquid spills from di- velocity, while it tends to be decrease due to small
rectly outflow near orifices. Liquid mass flux of each inflow velocity and the resulting lower pressure drop
outflow was measured to compare with modeling re- at orifices.
sults for validation. The simulation results for the Pre-distributor II
with different inflow velocities are presented in Fig. 4.
As for the smaller velocity (UL 1.5 m·s1), the out-
flow distribution is consistent with Kister’s clarifica-
tion, whereas the outflow increases for first several
orifices and then decreases for the remaining ones for

Figure 2 The setup for testing the hydrodynamic per-


formance of the pre-distributor for validation
1üpool; 2üelectronic scale; 3übucket; 4üdrainage pipe; 5ü
pre-distributor; 6ürotor flowmeter; 7üvalve; 8üwater pump

The fundamental behaviour of the simulated re-


sults (square symbols) could be validated by experi-
mental data (circles) in Fig. 3. Obviously, the simula- Figure 4 Outflow distribution of the orifices for different
tion results have the same trend as the experimental inflow velocities
results. Although they agree well in general, errors Liquid velocity at inlet/m·s1:Ƶ1.5;ƽ2.0;Ʒ2.5;ͩ3.0
818 Chin. J. Chem. Eng., Vol. 19, No. 5, October 2011

Figure 5 Simulated results of distribution planes of water volume fraction for different inflow velocities

the same pre-distributor with larger inflow velocities Figure 6 shows the simulation results for these
(UL 2.0 m·s1, 2.5 m·s1, 3.0 m·s1). Furthermore, the four pre-distributors with the same inflow velocity (1
larger the velocity is, the more orifices in the increas- m·s1). The outflow distributions are apparently dif-
ing trend there are. ferent for these 4 types of pre-distributors. A trend of
In simulation, the distributions of liquid volume decreasing is clearly observed for the straight pipe
fraction are different from inflow velocities. The first with lateral feeding. But for the pre-distributors with
part of pre-distributor close to inlet is full of water horizontal feeding and 45° feeding from the central
(Fig. 5), but the water layer is gradually thinner for the pipe, the outflow distributes with increasing trend in
remaining part. Additionally, the length of this full both left and right ends of the pipe. Moreover, mass
filled part increases with the increase of inflow veloc- flows of outlets for the left segment are mostly larger
ity. Combination of Figs. 4 and 5 helps to conclude than the right ones due to the bias flow caused by the
that the outflow increases for the orifices in the fully orientation of entrance. For the pre-distributor with ver-
filled section of pipe and the outflow for orifices un- tical feeding in the centre, the outflow is low for the
der half filled pipe is vice versa. This result is in ac- centre orifices 6 and 7 due to the central inflow of the
cordance with Kister’s viewpoint. pre-distributor. However, it is the most uniform dis-
When the inflow velocity is above 2.0 m·s1, liq- tribution for the outlet orifices in vertical feeding pipe.
uid mass flow is large enough to fill the whole pipe at
the beginning and a layer distribution occurs for the
rest pipe segment. As for the segment full of liquid,
the outflow increases due to the decreasing of liquid
velocity caused by decreasing liquid mass flow and
the resulting higher static pressure at first orifices. As
for the segment covered by a layer of liquid, the out-
flow presents a downward trend because of decreasing
height of the layer and the resulting lower pressure
along the pipe.

5.2 Different directions of feeding pipe


Figure 6 Outflow distribution of the orifices for different
It has been found in different test series that the feed pipe directions
outflow is sensitive to the exact location and orienta- inlet pipe direction: Ƶ direct feed; ƽ horizontal feed; Ʒ 45°
tion of the feeding pipe [2]. Therefore, four different feed;ͩvertical feed
pre-distributors were simulated. One of these
pre-distributors is a straight pipe with lateral feeding Different outflow distribution is resulted from
and industrial size shown in Fig. 1 (a); the other three different flow characteristics within the pre-distributors.
pipes are of the same size but using central feeding In Fig. 7, it is showed that simulated planes of water
with different inlet designs. volume fraction differed from each other for those 4
Chin. J. Chem. Eng., Vol. 19, No. 5, October 2011 819

Figure 7 Simulated results of distribution of water volume fraction for feed pipes with different feeding directions

pipes although they all belong to layer distribution


under such inflow velocity. Using a central feeding,
the liquid volume fraction is smaller at the central part
and larger at the end part because of the high lateral
velocities resulted from centre inflow and the liquid
accumulation resulted from end wall reflection. Ob-
viously, there is serious bias taking place in the centre
of pipes with horizontal feeding and 45°feeding.

5.3 Liquid distribution in the trough

Pre-distributor with multi-orifices is the common Figure 9 Outflow distribution for the shunts of upper
type of inflow for the top row of trough of a large trough with different inflow velocities
scale liquid distributor. Therefore, the uniformity of 1.5 m·s1; 2.0 m·s1; 2.5 m·s1; 3.0 m·s1
liquid level within the upper troughs is directly related
to the outflow distribution of pre-distributor. Fig. 8
shows a distribution of the height of free surface in the feeding; this results in the nonuniformity of outflow
upper trough for the pre-distributor II which indicated through the shunts. Furthermore, an uneven distribu-
that the height of liquid level increased with the in- tion of the upper trough may further affects the distri-
crease of inflow velocity. However, liquid levels are butional performance of liquid in the trough below.
all decreasing for different velocities; this leads to a The computational domain of Fig. 1 (c) including
same decreasing trend for outflow of the shunt troughs, both pre-distributor and troughs was modelled for the
which can be found from Fig. 9. In a word, liquid study whether the pre-distributor with central feeding
level presents serious inconsistencies for the first is better for uniform liquid level of the trough. The
trough under the pre-distributor with long lateral simulated distribution for water volume fraction is
presented in Fig. 10. As shown in this figure, these
two troughs located on both sides are almost full of
water with small bubbles scattered, while these two
close to the entrance are only half full with a serious
fluctuation in the centre. The reason is that impact on
the wall for the inflow makes large lateral velocity;
this leads to smaller outflow through the outlet pipes
near entrance.
Figure 11 is for the outflow distribution within
those four troughs. It is shown that there is a very
large outflow in the centre of trough 1 and 4 resulted
from high liquid level. On the contrary, the outflow
through the central outlets is very small for trough 2
and 3 due to the serious fluctuation and the resulting
thinner liquid layer. In addition, the outflow of the
remaining outlets has an increasing trend.
Figure 8 Simulated liquid level within the trough for Since the central feeding still can not obtain more
different inflow velocities uniform liquid distribution, it is suggested for structural
UL/m·s1: 1ü1.5; 2ü2.0; 3ü2.5; 4ü3.0 optimization with tapered design for the troughs or
820 Chin. J. Chem. Eng., Vol. 19, No. 5, October 2011

Figure 10 Simulated planes for water volume fraction within different troughs

the trough. Structural optimization was recommended


for more uniform distribution of liquid in the distributors.

NOMENCLATURE

A interfacial area density, m2·m3


g gravity acceleration, m·s2
Np total number of phase
p pressure, Pa
r volume fraction
SM momentum sources, kg·m1·s2
Figure 11 Outflow distribution for different troughs t time, s
trough
Ƶ 1;ƽtrough 4;Ʒtrough 2;ͩtrough 3 U velocity vector, m·s1
ȡ density, kg·m3
ȝ viscosity, kg·m1·s1
variable sizes for the outlet pipes. Moreover, it is bet- Subscripts
ter to avoid outflow close to the central entrance. Į phase Į
ȕ phase ȕ

6 CONCLUSIONS
REFERENCES
With free surface model for the calculation of
gas/liquid interface, homogeneous model has been 1 Gao, G.H., Li, Y., Zhao, J.W., Jiang, B., Li, X.G., “CFD simulation
used for the two phase modeling of pre-distributors. and optimal design of pre-distributor of straight pipe”, Chem. Ind.
The simulated results agreed well with the experi- and Eng. Prog., 28 (Suppl.), 355359 (2009).(in Chinese)
2 Zhang, W.Q., “Computational fluid dynamics on vapor-liquid dis-
mental data. The affects of liquid inflow rates and
tributor in large-scale packed column and optimum design”, M.S.
feeding pipe directions upon outflow distribution have Thesis, Tianjin (2004). (in Chinese)
been analyzed by modeling different pre-distributors. 3 Heggemann, M., Hirschberg, S., Spiege, L., Bachmann, C., “CFD
It was concluded that liquid distributed differently simulation and experimental validation of fluid flow in liquid dis-
under various flow rates which could induce different tributors”, Chem. Eng. Res. Des., 85 (A1), 5964 (2007).
outflow characteristics. And the outflow was sensitive 4 Fan, Z.W., Zhou, X.G., Luo, L.A., Yuan, W.K., “Numerical investi-
to the exact location and orientation of the feeding gation of constructal distributors with different configurations”, Chin.
pipes. The simulation results showed that vertical J. Chem. Eng., 17 (1), 175178 (2009).
5 Luo, L., Tondeur, D., “Optimal distribution of viscous dissipation in
feeding in the centre was the best feeding way for the
a multi-scale branched fluid distributor”, Int. J. Therm. Sci., 44 (12),
pre-distributor. In addtion, the influences of outflow 11311141 (2005).
performance of the pre-distributor on the liquid distribu- 6 Hirt, C.W., Nichols, B.D., “Volume of fluid (VOF) method for the
tion in the trough have also been analyzed. It was found dynamics of free boundaries”, J. Comput. Phys., 39, 201225 (1981).
that uneven outflow of the pre-distributor could fur- 7 Kister, H.Z., Distillation Operation, McGraw-Hill, New York,
ther affect the distributional performance of liquid in 434435 (1990).

Você também pode gostar