You are on page 1of 10

Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

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

Liquid distributor design for random packed


columns

Florian Hanusch ∗ , Michael Künzler, Michael Renner, Sebastian Rehfeldt ∗ ,


Harald Klein
Technical University of Munich, Department of Mechanical Engineering, Institute of Plant and Process Technology,
85748 Garching, Germany

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

Article history: Liquid distributors are the second most important internal in packed columns, as they
Received 29 November 2018 ensure sufficient initial distribution, which is crucial to achieve optimum separation per-
Received in revised form 16 April formance. In this work, a liquid distributor design method is presented, which applies the
2019 TUM–WelChem Cell Model to consider the interaction between the liquid distributor and
Accepted 15 May 2019 the packing. Liquid distribution in the packing is estimated in dependence of the liquid dis-
Available online 23 May 2019 tributor design. An overall maldistribution quality is determined as a characteristic value for
evaluation. Two common methods, the coefficient of variation and the liquid distribution
Keywords: quality by Moore and Rukovena, are implemented for reference. A parameter study with
Liquid distributor an exemplary column design provides optimum drip point densities for different types of
Packed column packing. Moreover, the spacing between the outermost drip points and the column wall is
Random packing considered in detail, suggesting an optimum distribution quality for high wall spacing. Fur-
Cell model ther application of the method suggests an increase of the optimum drip point density for
Liquid maldistribution smaller column diameters. Application of the TUM–WelChem Cell Model to liquid distrib-
utor design provides a potent tool for an optimized distributor layout with sufficient initial
distribution at minimized material usage and high free space for the gas flow.
© 2019 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

1. Introduction flow through the drip points, which can be affected, e.g., by
plugging, fouling, or a tilted installation of the distributor.
Column fluid dynamic calculations are the main step in ran- Another design aspect is the free space for gas flow, which
dom packed column design, providing the results to choose is also linked to liquid carryover and foaming behavior. Fur-
the optimum column diameter and type of packing. Hence, thermore, uniform distribution is to be maintained also at
there are many state of the art column fluid dynamic calcula- high turndown rates (Eckert, 1961; Chen, 1984; Fadel, 1984;
tion methods available in literature. Aside from the packing, Bonilla, 1993; Olsson, 1999). However, calculation methods for
liquid distributors are the most important column internals in the estimation of liquid distribution quality cannot account for
distillation, absorption, and stripping (Chen, 1984). Sufficient constructional and installation aspects, thus it comes back to
initial distribution is crucial to achieve optimum separation drip point density, pattern uniformity, and liquid flow unifor-
performance (Fadel, 1984; Bonilla, 1993; Olsson, 1999). How- mity.
ever, liquid distributor design methods are few in number. The most common method to estimate liquid distribution
In general, liquid distributor design guidelines consider quality uses the coefficient of variation Cv , which describes
the drip point density and the uniformity of the drip point the standard deviations of local liquid loads in relation to the
pattern. Further attention is paid to the uniformity of liquid overall liquid load (Billingham et al., 1997). Another widely


Corresponding authors.
E-mail addresses: florian.hanusch@tum.de (F. Hanusch), sebastian.rehfeldt@tum.de (S. Rehfeldt).
https://doi.org/10.1016/j.cherd.2019.05.035
0263-8762/© 2019 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
690 Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698

Table 1 – Exemplary column design specification.


Symbols
Column diameter dC 1.0 m
Packing height hP 5.0 m
A, B, C characteristic values [–]
Liquid load uL 12.7 m3 m−2 h−1
AC column cross-sectional area [m−2 ]
Gas load factor F 1.78 Pa0.5
a grid dimensions [m] Type of packing Raflux Ring 50-5 Metal
CQ (coefficient of) variation quality [–]
Cv coefficient of variation [–]
DQ liquid distribution quality [–] a cell model for the prediction of liquid distribution in random
DPD drip point density [m−2 ] packing. He investigates the liquid distribution in the pack-
dC column diameter [m] ing for different initial distribution patterns and evaluates the
W wall spacing [m] drip point density against the number of layers to reach natu-
F gas load factor [Pa0.5 ] ral distribution. Another approach is presented by Klemas and
hP packing height [m] Bonilla (2000), who estimate the separation efficiency via an
lL liquid run length [m] empirical maldistribution model, which calculates the initial
Mf maldistribution factor [–] maldistribution from the drip point density.
MQ maldistribution quality [–] This general approach to evaluate the liquid distribution
NDP number of drip points [–] quality considering the interaction with the packing requires
r radius [m] a lot of preliminary work or, more precisely, a model for the
uL liquid load [m3 m−2 h−1 ] prediction of liquid distribution in random packing. Accord-
ingly, the current work is a follow-up investigation of the
Abbreviations completed project Cell Model for the Design of Packed Columns
HFP50 Hiflow Ring 50-6 Plastic and Liquid Distributors, funded by the Bavarian Research Foun-
HFP90 Hiflow Ring 90-7 Plastic dation (AZ-1033-12). The first part of the project covered an
RFM35 Raflux Ring 35-5 Metal experimental investigation of liquid maldistribution and its
RFM50 Raflux Ring 50-5 Metal influencing factors (Hanusch et al., 2018a), while in the second
RMSR50 RVT Metal Saddle Ring 50-4 part the TUM–WelChem Cell Model for the prediction of liquid
RMSR70 RVT Metal Saddle Ring 70-5 distribution in random packing was developed and validated
RSRM#2 Raschig Super-Ring No. 2 Metal (Hanusch et al., 2019). With the TUM–WelChem Cell Model
at hand, fields of application have been identified in liquid
distributor design as well as in process simulation (Hanusch
used method is introduced by Moore and Rukovena (1987), et al., 2018b). In this work, a method for optimum liquid dis-
which evaluates liquid distribution quality graphically. Each tributor design considering the interaction with the packing is
drip point is represented by a “distribution circle”, where the presented. Investigations focus on random packing only.
area of each distribution circle is proportional to the liquid
flow through the respective drip point and the sum of all dis- 2. Methods
tribution circle areas is equal to the column cross-sectional
area. This graphical setting is then evaluated by means of In packed column design the specification of liquid distribu-
characteristic values, accounting for drip point density as tors and, generally speaking, column internals is one of the
well as pattern and liquid flow uniformity. Perry et al. (1990) last steps. Column fluid dynamic calculations provide liquid
present a method to evaluate the pattern uniformity, espe- and gas loads, column diameter, and type of packing. With
cially close to the column wall. The column cross-section is these specifications available, liquid distributor design is con-
divided into three radial zones of equal area and the drip ducted. A method for the estimation of liquid distribution
point density in each zone is compared to the overall drip quality using the TUM–WelChem Cell Model is proposed and
point density. In most cases, this method reveals underir- compared against methods from literature. In this section, the
rigation in the outer ring, which can be compensated by workflow to estimate liquid distribution quality is explained
additional drip points close to the column wall. Killat and Rey with an exemplary column design, adopted from (Hanusch
(1996) propose a method to account for large-scale maldis- et al., 2018b; Schultes, 2000) and summarized in Table 1.
tribution by subdividing the column cross-section into three The liquid distributor is specified with a square grid drip
different types of zones—concentric rings, pie slices, and grid point pattern and a drip point density DPD of 100 m−2 . In gen-
squares—and then estimating liquid flow deviations between eral, the drip point density is defined as the number of drip
zones for each configuration. Billingham et al. (1997) further points NDP divided by the column cross-sectional area AC .
pursue the concept of the coefficient of variation Cv , which by However, for the parameter study an alternative definition for
itself can only identify randomly dispersed uneven distribu- the drip point density is required, which refers directly to the
tion. They introduce an additional coefficient, the coefficient grid dimensions DPD ≡ 1/a2 . Fig. 1a shows the drip point pat-
of maldistribution Cm , which relates the standard deviations tern of the liquid distributor and Fig. 1b a sketch of the square
of local liquid loads to average liquid loads in the vicinity. grid dimensions.
Consequently, the ratio of Cv /Cm is a measure for clustered
maldistribution. 2.1. TUM–WelChem Cell Model
All aforementioned methods have a drawback. While the
packing separation efficiency is linked to the liquid distribu- The TUM–WelChem Cell Model discretizes the packing with a
tion in the packing, they only consider the liquid distributor grid of honeycomb cells. Single random packing elements are
drip points, but not the interaction with the packing itself. positioned in each cell in random orientation. Initial distribu-
Albright (1984) takes this interaction into account by applying tion in the top layer is derived from the drip point coordinates
Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698 691

Fig. 1 – (a) Drip point pattern of the exemplary design liquid distributor with DPD = 100m−2 and (b) determination of the drip
point density DPD from square grid dimensions a.

of the liquid distributor. Liquid distribution in the packing is of MQ = 85.1%. Two further methods from literature, the coef-
calculated in a top down sequence, following a three-step dis- ficient of variation Cv and the liquid distribution quality DQ by
tribution mechanism, which considers the geometrical shape Moore and Rukovena (1987), are used for reference.
of the packing as well as the operating parameters liquid and
gas load. Detailed information about the TUM–WelChem Cell
2.2. Coefficient of variation
Model is presented in Hanusch et al. (2019).
Simulation results in the form of a liquid distribution pro-
Adjusted to the honeycomb cell grid of the TUM–WelChem
file, as shown in Fig. 2a, illustrate the radial liquid distribution
Cell Model, the coefficient of variation Cv is defined by Eq. (3),
in the longitudinal section. The most uniform liquid distribu-
assuming values between 0 and ∞ (Billingham et al., 1997).
tion is achieved just after the first few packing layers below the
distributor. Further down, an increased flow builds up at the 
 n  
column wall, resulting in a decreased liquid load in the bulk of 1  uL,i − uL 2
the packing. The cross-sectional liquid distribution is summa- Cv =  · (3)
n uL
rized in form of a maldistribution factor Mf , calculated from i=1
the local liquid loads uL,i via Eq. (1), with the number of cells
n in a cell layer. This maldistribution factor Mf is a modified
The coefficient of variation Cv is determined for the initial dis-
coefficient of variation, which by definition assumes values in
tribution of the liquid distributor only, not taking into account
a limited range between 0 and 2 (Hanusch et al., 2018b).
any interactions with the packing. Again, for better com-
 2
parison, a (coefficient of) variation quality CQ ≡ 1/(Cv + 1) is
1 
n
uL,i − uL defined to obtain values between 0% and 100%. For the exem-
Mf = · (1)
n uL plary liquid distributor the variation quality is CQ = 37.9%.
i=1

In Fig. 2b, the maldistribution factor Mf is plotted along the 2.3. Moore and Rukovena (1987)
liquid run length lL . The liquid distribution profile shows a
minimum maldistribution factor Mf after the first few layers The graphical method by Moore and Rukovena (1987) is used
below the liquid distributor, further down increasing towards as a second reference. In this method, the drip points are
a trend value along with the wall flow. For evaluation of the represented by distribution circles. While the sum of all dis-
liquid distributor, only the entry section up to a liquid run tribution circle areas is equal to the column cross-section, the
length lL = 2.0 m which equals to two times the column diam- area of each distribution circle is proportional to the liquid flow
eter DC = 1.0 m is considered. This entry section describes the through the corresponding drip point. In this work, the liquid
initial distribution as well as the distribution performance of flows through all drip points are assumed to be equal, resulting
the packing. The evaluation does not include the full packing in uniform sized distribution circles. Fig. 3 shows the graphical
height HP , which would shift the weight towards the natural setting for the evaluation of liquid distribution quality.
flow of the packing. Three characteristic values A, B, and C are determined
A mean maldistribution factor M̄f sums up the liquid distri- graphically, interpreting the coverage of the column cross-
bution profile in the entry section in one characteristic value, section by distribution circles as well as overlapping of
which again takes values between 0 and 2. Finally, for better distribution circles.
comparison with the other evaluation methods, a maldistri-
bution quality MQ is defined: covered column cross-section
A= (4)
column cross-section
MQ ≡ 1 − 0.5 · M̄f . (2)
B = min{B1 ; B2 } (5)
Here, 0% represents the worst possible and 100% a com-
pletely uniform liquid distribution. This approach considers
minimum distribution circlearea
liquid distributor design, interaction with the packing, and the
dependence on operating parameters. The distributor in the in1/12 column cross − section
B1 = (6)
exemplary column design achieves a maldistribution quality 1/12 column cross − section
692 Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698

Fig. 2 – (a) Local liquid loads uL,i plotted along the liquid run length lL against the radius r and (b) maldistribution factor Mf
plotted along the liquid run length lL for the exemplary column design.

Fig. 3 – Graphical evaluation of the exemplary liquid distributor with the method by Moore and Rukovena (1987);
segmentation by (a) 12 horizontal stripes and (b) 12 ring sectors of equal area for the determination of B according to
WelChem GmbH (2018).

1/12 column cross − section 3. Results and discussion


B2 = (7)
maximum distribution circlearea
in 1/12 column cross − section The methods to assess liquid distribution quality are now
applied to find optimum liquid distributor layouts for different
random packed column designs. At first, a parameter study is
distribution circle overlaparea
C= (8) conducted with varying drip point densities DPD for the exem-
column cross − section
plary column design. The approach is also applied to different
types of packing. Next, liquid distribution quality is evaluated
These three values are combined in Eq. (9) to obtain the
for varying wall spacing W . Furthermore, the influence of
liquid distribution quality DQ , which can take values between
column design parameters on the optimum liquid distributor
0% and 102.5%.
layout is examined. All designs are evaluated with the under-
lying assumption of uniformity of liquid flow through all drip
DQ = 0.40 · A + 0.60 · B − 0.33 · (C − 7.5%) (9) points.

While the determination of A and C is straightforward, 3.1. Drip point density


the right choice of 1/12 of the column-cross section for B
requires further consideration. A standardized method uses The primary parameter in liquid distributor design is the
both stripes and ring sectors of equal area, as shown in Fig. 3. drip point density. For the exemplary column design speci-
Stripes are evaluated in horizontal and vertical orientation, fied in Table 1, the drip point density is varied in the range of
while the circle sector pattern is rotated by 30◦ , 45◦ , and DPD = 1–200 m−2 for a strictly regular square grid, considering
60◦ (WelChem GmbH, 2018). With this method, the exem- no interruptions of this pattern for gas chimneys, distributor
plary liquid distributor achieves a liquid distribution quality troughs or parting boxes. The origin of the square grid is not
of DQ = 87.6%. defined by a single drip point in the column centre, but by
Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698 693

Fig. 4 – Liquid distribution qualities MQ , CQ , and DQ plotted against the drip point density DPD for Raflux Ring 50-5 Metal
(RFM50) at exemplary design specifications.

a centred rectangle of four drip points. Therefore, the mini- from a minimum value of 72% at low drip point densities up
mum drip point density for a column diameter of dC = 1.0 m to a trend value of about 86% at high drip point densities,
is DPD = 2 m−2 . Higher drip point densities than 200 m−2 are estimated by an asymptotic fit through the circles.
unusual, as they result in small drip hole diameters, which are An optimum distributor layout provides a high maldis-
susceptible to plugging. Results of liquid distribution qualities tribution quality at a sufficient number of drip points and,
MQ , CQ , and DQ according to the methods presented in Section therefore, a high free space for the gas flow and a low use
2 are at display in Fig. 4. of material. The optimum drip point density is defined as
The top graph in Fig. 4 shows the results for the mald- the minimum drip point density at which the maldistribu-
istribution quality MQ . Its progress, displayed by the short tion quality exceeds 99% of the trend value, referred to as
dashed line, shows sudden jumps, resulting from the step- MQ,99 . This method provides an optimum drip point density
wise appearance of additional drip points at the column wall of DPDopt = 72 m−2 for the exemplary column design, at which
while continuously increasing the drip point density DPD. a maldistribution quality of MQ = 86.5% is achieved. In the dia-
While the number of drip points NDP is constant for distinctive gram, the result is marked by a filled circle.
ranges of DPD, resulting from the definition DPD ≡ 1/a2 relating For the variation quality CQ , at display in the centre graph of
it to the square grid dimensions a, the maldistribution quality Fig. 4, such a method is not applicable. It shows an increasing
MQ varies strongly in these ranges. Thus, for each number of progress from 0% to 51% with increasing drip point densities.
drip points NDP a range of good and bad designs is obtained. For This can be explained by Eq. (3), in which the local liquid loads
further evaluation only the optimum design for each number uL,i decrease with increasing DPD. The variation quality can
of drip points NDP which scores the maximum maldistribution be used for, e.g., the evaluation of different distributor layouts
quality MQ is considered. These optimum designs are marked with similar drip point densities but irregular patterns, caused
by the hollow circles. The maldistribution quality MQ ranges by the presence of gas chimneys. However, for the presented
694 Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698

Table 2 – Optimum distributor design parameters for


RFM35, HFP50, HFP90, RMSR50, RMSR70, and RSRM#2
and exemplary design specifications.
Packing DPDopt [m−2 ] NDP,opt /AC [m−2 ] W,opt [mm] MQ [%]

RFM35 45 41 65 85.6
RFM50 72 66 51 86.4
HFP50 41 41 45 85.9
HFP90 120 112 30 86.5
RMSR50 73 66 54 87.7
RMSR70 72 66 51 82.3
RSRM#2 69 66 42 87.4

parameter study it shows no trend from which an optimum


could be derived. Thus it is not considered further in this work.
However, the liquid distribution quality DQ by Moore and
Rukovena (1987)—bottom graph in Fig. 4—shows a progress
similar to the maldistribution quality MQ , with a range from
DQ = 74% for low drip point densities up to a trend value of
about 89% for high drip point densities. Evaluation of the opti-
mum drip point density by the same procedure as for the
maldistribution quality results in a DPDopt of 66 m−2 for the
exemplary liquid distributor. With only a slight deviation from
the optimum drip point density DPDopt = 72 m−2 derived from
the maldistribution quality MQ and the identical number of
drip points per column cross-sectional area NDP /AC = 66 m−2
for both methods, the determined drip point density is in good
agreement with state of the art design rules (Kister, 1990; Perry
et al., 1990; Bonilla, 1993; Klemas and Bonilla, 1995).
This first result is valid for Raflux Ring 50-5 Metal (RFM50)
type of packing only. The interaction of liquid distributor and
packing is now further examined for different types of pack-
ing. The RFM50 from the exemplary column design is a 2nd
generation random packing. Optimum drip point densities are
now estimated for Raflux Ring 35-5 Metal (RFM35), 3rd gener-
ation Hiflow Ring 50-6 and 90-7 Plastic (HFP50, HFP90), RVT
Metal Saddle Ring 50-4 and 70-5 (RMSR50, RMSR70), and 4th
generation Raschig Super-Ring No. 2 Metal (RSRM#2) (Kister,
2013). Resulting maldistribution qualities MQ for drip point
densities DPD in the range of 2–200 m−2 are at display in Fig. 5,
with the optimum designs marked by the filled symbols. In
addition, Table 2 contains the results for the optimum dis-
tributor design parameters, including the optimum number Fig. 5 – Maldistribution quality MQ plotted against the drip
of drip points per column cross-sectional area NDP,opt /AC for point density DPD for RFM35, HFP50, HFP90, RMSR50,
reference to the standard definition. RMSR70, and RSRM#2 for exemplary design specifications.
Compared to the results for the RFM50, the smaller
RFM35 requires a lower optimum drip point density of
DPDopt = 45 m−2 , at which it achieves a slightly higher mald-
istribution quality MQ = 85.6%. The RFM35 features a higher
radial distribution of the liquid in the packing and is thus able and RSRM#2 cluster in the range of DPDopt = 69–73 m−2 , at
to compensate for the lower drip point density. What is more, which both RMSR50 and RSRM#2 score the highest maldis-
the wall flow is lower for smaller nominal size, resulting in tribution qualities MQ of 87.7% and 87.4%, while the RMSR70
a higher achievable maldistribution quality. An even smaller achieves the lowest MQ of 82.3%.
optimum drip point density of DPDopt = 41 m−2 is required for These observations are in good agreement with the theory
the HFP50. The grid-like structure of the HFP50 is least sus- that higher generation random packings are less susceptible
ceptible to liquid maldistribution in comparison with the other to liquid maldistribution, while an increase of the nominal
types of packing in the study. For the HFP90, on the other hand, size increases liquid maldistribution (Hanusch et al., 2018b).
a much higher optimum drip point density of DPDopt = 120 m−2 Moreover, these results show the advantage of the method
is determined. The radial distribution of liquid in the pack- presented in this work, which considers the interaction of
ing decreases with increasing nominal size, while the wall liquid distributor and packing, in contrast to the method by
flow increases. Thus, a higher drip point density is demanded, Moore and Rukovena (1987), which only considers the drip
since the packing itself is not able to compensate for an insuf- point pattern of the liquid distributor. An evaluation with the
ficient initial distribution. Optimum drip point densities for liquid distribution quality DQ would always show the identical
3rd and 4th generation saddle type packings RMSR50, RMSR70, result, as it does not consider the packing.
Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698 695

Fig. 6 – Maldistribution quality MQ and liquid distribution quality DQ plotted against wall spacing W for RFM35, HFP50,
HFP90, RMSR50, RMSR70, and RSRM#2 at optimum numbers of drip points per column cross-sectional area NDP,opt /AC for
exemplary design specifications.

3.2. Wall spacing HFP90, with the lowest W,opt of 30 mm, according to the max-
imum wall spacing available at the respective optimum drip
Another matter of interest in liquid distributor design is the point density NDP,opt /AC for each type of packing. These results
distance between the outermost drip points and the column align well with constructive restrictions, where a high distance
wall, specified by the wall spacing W . Due to constructive between the outermost drip points and the column wall is
restrictions, the positioning of the outermost drip points of the desirable. Design rules also state, that the liquid should not
square grid is restricted. The wall spacing W is investigated be distributed directly at the column wall, to counteract the
for the optimum drip point densities, or more precisely for the maldistribution phenomenon of wall flow (Kister, 1990).
range of wall spacings W at constant optimum numbers of A different trend occurs for the liquid distribution quality
drip points per column cross-sectional area NDP,opt /AC . In Fig. 6 DQ by Moore and Rukovena (1987). First, the liquid distribution
the maldistribution quality MQ for different types of packing quality DQ increases along with the wall spacing W up to an
and the liquid distribution quality DQ is plotted against the optimum value of 88.6%, only to decrease steeply for higher
wall spacing W . to maximum wall spacings. This trend can be explained by
The maldistribution quality MQ shows an increasing trend the change of the characteristic values A, B, and C in Eq. (9),
with increasing wall spacing W . Optimum maldistribution considering the position of the distribution circles of the out-
qualities are obtained at wall spacings W,opt in the range of ermost drip points. At a wall spacing W of 0 mm, additional
40–55 mm for all packings with an optimum drip point den- drip points appear at the column wall, with part of their distri-
sity NDP,opt /AC = 66 m−2 . The outermost drip points are as far bution circle areas outside of the column cross section. With
away from the column wall as possible, explaining the high an increasing wall spacing W , the distribution circle areas
maldistribution qualities, since the liquid is distributed uni- move inwards, simultaneously increasing A, B, and C, which
formly over the column cross-sectional area with the least improves the liquid distribution quality DQ up to its optimum.
possible initial wall flow. The same trend occurs for RFM35, Increasing the wall spacing beyond this optimum moves the
with the highest optimum wall spacing W,opt = 65 mm, and distribution circle areas so far inwards, that gaps in between
696 Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698

Fig. 7 – Maldistribution quality MQ and liquid distribution quality DQ plotted against the drip point density DPD for the
exemplary column design with varying column diameters dC .

distribution circles at the wall increase rapidly. This results umn diameter dC , the achievable maldistribution quality MQ
in a simultaneous decrease of A and B, an increase of C, and decreases. Contrary to the decreasing maldistribution quality
thus a steep decrease of the liquid distribution quality DQ . MQ a higher optimum drip point density of DPDopt = 97 m−2 is
The optimum wall spacing of W,opt = 31 mm conflicts with the required for the smaller column diameter of dC = 0.5 m. For the
constructive restrictions and does not account for the interac- bigger column diameter of dC = 2.0 m, a lower optimum drip
tion with the packing, considering the initial wall flow caused point density of DPDopt = 47 m−2 is sufficient to achieve the
by lower wall spacings. higher maldistribution quality of MQ = 89.4%. The same trend
is obtained for the liquid distribution quality DQ by Moore
and Rukovena (1987), only with an even higher optimum drip
3.3. Column design parameters
point density of DPDopt = 162 m−2 for the small column diam-
eter of dC = 0.5 m. This trend of decreasing optimum drip point
Beyond liquid distributor design for a fixed column design
densities for increasing column diameters is in line with the
specification, optimum drip point densities are determined for
decreasing ratio of outermost drip points in reference to the
varying design parameters. Based on the exemplary column
total number of drip points, resulting in a decrease of the wall
design, the column diameter dC , the liquid load uL , and the gas
flow. Altogether, an increasing drip point density for smaller
load factor F are changed individually. Fig. 7 shows the eval-
column diameters is in good agreement with state of the art
uation of optimum drip point densities for different column
design rules (Perry et al., 1990).
diameters. While the maldistribution quality MQ increases
A change of the operating parameters liquid load uL and
with the column diameter dC , the optimum drip point density
gas load factor F, on the other hand, does not lead to a signif-
DPDopt decreases.
icant change in the evaluation of DPDopt . For both increasing
The wall flow is the dominating effect in terms of the mald-
liquid load uL and gas load factor F, Fig. 8 shows the same
istribution quality. As the ratio of the wall flow in reference
progression for the maldistribution quality MQ , with an offset
to the column cross-section increases with decreasing col-
Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698 697

Fig. 8 – Maldistribution quality MQ plotted against the drip point density DPD for the exemplary column design with
varying liquid loads uL and gas load factors F.

to slightly lower values at medium to high drip point den- 4. Conclusion


sities. Both investigations show a runaway value—at liquid
load uL = 10 m3 m−2 h−1 and gas load F = 2 Pa0.5 —marked by the Established methods for the evaluation of liquid distributor
light filling. However, these individual points are not taken into designs like the coefficient of variation Cv and the liquid dis-
account, since liquid distributor design does not only consider tribution quality DQ by Moore and Rukovena (1987) take only
a single design case, but also turndown or maximum-load the liquid distributor into account. In this work, a liquid dis-
cases. With this exception, the optimum drip point densi- tributor design method is presented, which considers the
ties for all operating parameters fluctuate around the value interaction of the liquid distributor and the packing with the
DPDopt = 72 m−2 of the exemplary design case. There is no eval- TUM–WelChem Cell Model. As the TUM–WelChem Cell Model
uation for the liquid distribution quality DQ , as it does not includes liquid distribution characteristics for different types
consider the operating parameters. of packing and hydraulic calculations, it renders more detailed
An enhanced theoretical study examined different refine- investigations and optimizations in liquid distributor design
ments of the liquid distributor layout, including manual as possible.
well as automated positioning of additional drip points in The presented method provides an optimized liquid dis-
gaps close to the column wall and specific stretching or com- tributor design with a sufficient liquid distribution in the
pressing of rows in the square grid to optimize the wall packing at minimized material usage and a high free space
spacing. Beyond that, gradual stretching or compressing of for the gas flow. The following general trends can be derived
the whole square grid was investigated. Eventually, none of from the parameter studies. Packings of higher generations
these approaches leads to a significant improvement over the require lower drip point densities, as they are less susceptible
regular square grid pattern. Besides, more complex drip point to maldistribution. High drip point densities are suggested for
patterns are unlikely to be implemented in liquid distributor packings with a high nominal size, which lack the radial dis-
design, for which reason the results are not presented in this tribution of the liquid in the packing to compensate for poor
work. initial distribution. For a regular drip point pattern, maximum
698 Chemical Engineering Research and Design 1 4 7 ( 2 0 1 9 ) 689–698

wall spacings of the outermost drip points are the optimum, Hanusch, F., Rehfeldt, S., Klein, H., 2018a. Liquid maldistribution
not only from a constructive perspective. What is more, the in random packed columns: experimental investigation of
drip point density should increase for small column diam- influencing factors. Chem. Eng. Technol. 41, 2241–2249,
http://dx.doi.org/10.1002/ceat.201800467.
eters. Even without the TUM–WelChem Cell Model at hand,
Hanusch, F., Engel, V., Kender, R., Rehfeldt, S., Klein, H., 2018b.
these general trends should be considered in liquid distributor Development and application of the TUM-WelChem Cell
design for random packed columns. Model for prediction of liquid distribution in random packed
columns. Chem. Eng. Trans. 69, 739–744,
5. Acknowledgments http://dx.doi.org/10.3303/cet1869124.
Hanusch, F., Kender, R., Rehfeldt, S., Klein, H., Engel, V., 2019.
TUM-WelChem Cell Model for the prediction of liquid
The authors would like to thank the Bavarian Research Foun- distribution in random packed columns. AIChE J.,
dation for the financial support of the preceding research http://dx.doi.org/10.1002/aic.16598.
project “Cell Model for the Design of Packed Columns and Liq- Killat, G.R., Rey, T.D., 1996. Properly assess maldistribution in
uid Distributors” (AZ-1033-12). We would also like to thank our packed towers. Chem. Eng. Prog. 92, 69–73.
former project partners WelChem GmbH, RVT Process Equip- Kister, H.Z., 1990. Distillation Operation. McGraw Hill, New York.
ment GmbH and Linde AG for their support and the shared Kister, H.Z., 2013. Distillation Design. McGraw-Hill, New York.
Klemas, L., Bonilla, J.A., 1995. Accurately assess packed-column
knowledge of liquid distributor design.
efficiency. Chem. Eng. Prog. 91, 27–44.
Klemas, L., Bonilla, J.A., 2000. Distillation | packed columns:
References design and performance. In: Wilson, I.D., Adlard, E.R., Cooke,
M., Poole, C.F. (Eds.), Encyclopedia of Separation Science.
Albright, M.A., 1984. Packed tower distributors tested. Academic Press, San Diego, pp. 1081–1098,
Hydrocarbon Process. 62, 173–177. http://dx.doi.org/10.1016/B0-12-226770-2/00651-7.
Billingham, J.F., Bonaquist, D.P., Lockett, M.J., 1997. Moore, F., Rukovena, F., 1987. Liquid and gas distribution in
Characterization of the performance of packed distillation commercial packed towers. Chem. Plants Process., 11–15.
column liquid distributors. Inst. Chem. Eng. Symp. Ser. 142, Olsson, F.R., 1999. Detect distributor defects before they cripple
841–851. columns. Chem. Eng. Prog. 95, 57–61.
Bonilla, J.A., 1993. Don’t neglect liquid distributors. Chem. Eng. Perry, D., Nutter, D.E., Hale, A., 1990. Liquid distribution for
Prog. 89, 47–61. optimum packing performance. Chem. Eng. Prog. 86, 30–35.
Chen, G.K., 1984. Packed column internals. Chem. Eng. 91, 40–51. Schultes, M., 2000. Influence of liquid redistributors on the
Eckert, J.S., 1961. Design techniques for sizing packed towers. mass-transfer efficiency of packed columns. Ind. Eng. Chem.
Chem. Eng. Prog. 57, 54–58. Res. 39, 1381–1389, http://dx.doi.org/10.1021/ie990437j.
Fadel, T.M., 1984. Selecting packed-column auxiliaries. Chem. WelChem GmbH, 2018. Manual Design Software TrayHeart:
Eng. 91, 71–76. Version 3.08.01.

You might also like