See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/323700001

Separation performance investigation of packed distillation columns using

simple NEQ approach based on packing multicomponent efficiencies and
effective mass transfer coefficients

Article in Korean Journal of Chemical Engineering · March 2018

DOI: 10.1007/s11814-017-0315-7

CITATIONS READS

6 2,963

3 authors, including:

Hadi Poortalari Javad Karimi-Sabet

Semnan University Atomic Energy Organization of Iran
2 PUBLICATIONS 13 CITATIONS 126 PUBLICATIONS 2,716 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Hadi Poortalari on 12 September 2018.

The user has requested enhancement of the downloaded file.

Korean J. Chem. Eng., 34(7), 1-16 (2017) pISSN: 0256-1115
DOI: 10.1007/s11814-017-0315-7 eISSN: 1975-7220
INVITED REVIEW PAPER INVITED REVIEW PAPER

Separation performance investigation of packed distillation columns

using simple NEQ approach based on packing multicomponent efficiencies
and effective mass transfer coefficients
Hadi Poortalari*, Javad Karimi Sabet**,†, and Farshad Varaminian*,†

*Chemical Engineering, Oil, Gas Faculty, Semnan University, Semnan, Iran

**Material and Nuclear Fuel Research School (MNFRS), Nuclear Science and Technology Research Institute, Tehran, Iran
(Received 28 May 2017 • accepted 20 November 2017)

Abstract−A simple non-equilibrium modeling approach is proposed to simulate multicomponent distillation process
in packed columns. The real behavior of the column is simply considered by the evaluation of interphase mass transfer
rate based on the overall mass transfer coefficient. Two distinct methods are used to calculate this overall coefficient
including the effective mass transfer coefficient method and the packing efficiency method. The modelling procedure
consists of an iterative segment-wise algorithm implemented in a MATLAB home-code. For verification, the obtained
composition profiles from a structured and a random packed column are compared with reported experimental data.
Comparisons show that the packing efficiency-based model could acceptably predict the experimental profiles with an
average relative deviation of 18% and 25% for structured and random packed columns, respectively. This confirms that
our simple non-equilibrium approach is a reliable and robust model for the performance evaluation of packed columns.
Keywords: Packed Distillation Column, Multicomponent Separation, Simple Non-equilibrium Modeling, Overall Mass
Transfer Coefficient, Packing Efficiency

INTRODUCTION ate an existing design is very expensive and time-consuming, design-

ers have often used the simulation tool to examine their idea for
Among various separation techniques, distillation is the most the process advancement. In fact, simulation is a robust tool for
widely used industrial method for separation goals [1]. Distilla- the process analysis and the equipment initial design. A trustworthy
tion columns are extensively applied in chemical, petrochemical, model is essential for the performance evaluation of the distilla-
petroleum, and natural gas industries [2]. According to internal tion process [7].
devices which are applied as phase contactors, distillation columns The modeling of distillation is not new. Depending on the as-
are categorized as random or structured packed columns and trayed sumptions, distillation modeling approaches vary from the simple
or plate columns. The vapor and the liquid are in continuous con- to the quite complex. Until now, many of the challenging prob-
tact in the packed columns, while stepwise contact between phases lems related to the distillation process modeling have been widely
occurs in plate towers [3]. To achieve the efficient mass transfer investigated. The equilibrium (EQ) stage model is the most com-
between phases, trayed towers are designed to provide enough liq- mon method for the simple simulation of the distillation process
uid holdup. In contemporary industrial uses, packed columns are [3]. For more than a century, distillation columns have been simu-
applied more than trayed towers, especially when low-pressure lated by the EQ stage concept [8]. In this approach, it is assumed
drops across the column are required. In packed towers, a large that both of the vapor and the liquid phases are perfectly mixed so
surface area per unit volume is provided to facilitate the liquid- that the outlet streams from a theoretical (ideal) stage reach the
vapor mass transfer. Recently, packed distillation columns with equilibrium state. Since a detailed equipment design is not required,
structured packings have gained more attention for development this model is a simple on-hand modeling approach. Indeed, a mini-
of the existing trayed columns [4]. Structured packings are a good mum amount of data (only data for the calculation of K-values
substitute for existing trays when more capacity or higher degree and enthalpies) are required for the EQ modeling of the distillation
of separation is required [5]. process. Mass, equilibrium, summation and heat (MESH) rela-
The initial investment related to installed capacity and opera- tions make the governing equations of the EQ stage model [9].
tional costs due to energy usage of distillation operation are signifi- The EQ modeling approach is considered as a conventional method
cant [6]. Therefore, distillation columns should be designed precisely to simulate packed distillation columns with introducing HETP
as possible to achieve the optimum operational conditions. Since (the height equivalent to a theoretical plate) concept as a some-
conducting experiments to determine design parameters or evalu- thing similar concept to the stage efficiency in trayed columns
[10]. According to this, the packed tower is modeled similar to a
†
To whom correspondence should be addressed. staged column, so that the packed height is divided into a num-
E-mail: [email protected], [email protected] ber of sections, each of them considered as a discrete stage. Al-
Copyright by The Korean Institute of Chemical Engineers. though the HETP concept is widely applied, no theoretical justifi-

1
2 H. Poortalari et al.

cation is available and this leads to the deviation from experimen- umns by Ilme et al. [23], Jakobsson et al. [24]. Although several
tal data [11]. In fact, the performance of the EQ modeling approach works are dedicated to multicomponent efficiency calculations,
in the simulation of packed columns is extremely dependent to most of them have focused on the modeling of plate columns, and
the precision of HETP values. the packed distillation columns are seldom modeled using this
Although the EQ model has been applied as a conventional approach. Keskinen et al. [25] used an EQ model with multicom-
approach for steady-state simulation of the distillation process, sev- ponent efficiency factors for the evaluation of the separation per-
eral shortcomings are observed due to applied assumptions. In formance of packed distillation columns; however, they believed
fact, actual stages significantly deviate from ideal stages, and outlet that their method still needed extra work to be verified with more
streams from a real tray or a packed section are rarely in equilib- experimental data.
rium. Therefore, researchers have considered the non-ideality of In our previous study [26], we presented a simple NEQ model
real distillation columns in the modeling approaches. As previous based on rigorous efficiency calculations for the simulation of plate
studies have clearly shown, the actual separation in a distillation distillation columns. In this model, the phase equilibrium relations
column depends on interphase mass transfer rates [12]. These rates in the EQ stage method are replaced with the interphase mass trans-
correspond to the measure of the deviation of vapor and liquid fer equations based on overall mass transfer coefficients. The over-
streams from the equilibrium state. In packed distillation towers, all mass transfer coefficient is simply calculated from the simul-
the performance of mass transfer is often expressed by HETP val- taneous solution of the vapor phase material balance and the Mur-
ues. However, the HETP concept is not directly applicable in mul- phree efficiency definition. The main characteristic of the presented
ticomponent distillations, because the mass transfer rates are different model is to combine advantages of the efficiency-based modified
for various species in multicomponent mixtures. A proper alterna- EQ model and the rate-based NEQ model to obtain an effective
tive for HETP is HTU (height of a transfer unit) concept [13]. This and appropriate approach for the modeling of the distillation pro-
concept represents the mass transfer efficiency of the packed bed. cess. In the proposed model, the behavior of real plates is initially
HTU can be correlated much more easily than HETP and success- estimated in the form of stage efficiencies, and then column sepa-
fully used in the design of packed columns for separating binary ration performance is obtained. In the present work, we use this
mixtures. However, it has been experimentally shown that HTU efficiency-based NEQ approach for the modeling of packed distil-
values are commonly different for each component in multicom- lation columns. According to this, the mass transfer in any packed
ponent mixtures [14]. It means that each component has diverse segment is considered by multicomponent efficiency calculations to
mass transfer efficiency along the packed bed. To overcome this correct the deviation from the equilibrium stage assumption. Besides,
problem in a multicomponent packed distillation process, two another simple NEQ model based on effective mass transfer coeffi-
main approaches are commonly suggested: the efficiency-based cients is also applied for the further evaluation of the proposed effi-
modified EQ method, and the non-equilibrium (NEQ) stage ciency-based NEQ model. In this approach, the effective mass transfer
method, known as the rate-based model. As a first suggestion, the coefficients are initially calculated for each phase, and next they
multicomponent efficiency calculations [15] are introduced into are combined to compute the overall mass transfer coefficient. So,
the EQ modeling approach. Accordingly, the separation efficiency it can be said that the common face of both implemented NEQ
of each component is predicted in any calculation segment. The methods is that their modeling procedure is based on the evalua-
efficiency-based modified EQ model is one of the conventional tion of interphase mass transfer rates according to the overall mass
approaches to consider the mass transfer phenomena in the distil- transfer coefficients. The key feature of these modeling approaches is
lation process [8]. The characteristic feature of this approach is that the non-equilibrium nature of the distillation process is con-
that the multicomponent efficiencies are calculated together with sidered while the structure of EQ stage method is maintained.
the whole modeling procedure. In the second suggestion, the rate Therefore, our aim is the rigorous simulation of packed distilla-
based approach [16-18] is used in which the actual mass transfer tion columns by using two simple NEQ approaches based on the
rates are directly determined for a given packing section or a real simple calculation of interphase mass transfer rates. In this paper, a
plate. However, the rate-based model is quite heavy to solve. Mean- detailed comparison between results of our NEQ models and re-
while, many physical properties, such as diffusion coefficients, sur- ported measured data is performed for two different packed distil-
face tension, thermal conductivities, and viscosities, are required in lation columns: a structured packed column as case study I and a
this approach. Indeed, the rate-based modeling approach gives random packed column as case study II. By using these methods,
more detailed results about column separation at the expense of the mentioned packed columns are modeled in the same way as a
heavier computational cost and more equipment design details staged column so that the packed height is divided into a number
than EQ stage approach. of sections, each of which may be modeled as a discrete NEQ
In recent years, different researchers have been motivated to stage. In this way, the separation performance of both columns is
apply efficiency calculations along with EQ modeling approaches obtained, and then the estimated separations are compared with
to consider the non-ideal nature of multicomponent distillation actual measurements along the packed beds. Accordingly, the col-
process. The applying of multicomponent plate efficiencies on the umn design can be analyzed, and could be modified if needed.
platform of EQ stage model was initially introduced by Aittamaa
[19]. This modeling approach was later developed further by Ilme MODEL DESCRIPTION
[20]. The efficiency-based methods have been widely studied by
Klemola [21], Jakobsson [22], and also applied for industrial col- Basically, two classes of approaches are suggested for the packed

July, 2017
 Separation performance investigation of packed distillation columns using simple NEQ approach 3

column modeling: the discrete approach and the continuous ap- simultaneously with energy balance equation. Several assumptions
proach [27]. In the discrete method, the packed bed is divided into have been used to simplify the model, including:
several sections so that each section is treated as a stage in the plate 1) Both phases are perfectly mixed in each segment,
column, and the model contains a set of nonlinear algebraic equa- 2) Vapor-liquid equilibrium is established only at the interface,
tions as governing equations. Therefore, the stage-wise modeling 3) The mass transfer resistance mainly exists in the vapor phase,
approaches are applicable in the simulation of continuous contact 4) The finite flux effects on the mass transfer coefficients are
devices as well as staged contactors. However, in the continuous ignored, then the low flux mass transfer coefficients are applied
method, the packed column is considered as a continuous contac- into equations of the model,
tor while the differential balance equations are written for any small 5) The interphase heat transfer rates are assumed to be negligi-
element of packing. So, the model consists of a set of differential ble,
equations that are solved using a numerical integration scheme. 6) The condenser and the reboiler are treated as equilibrium
In the current work, we apply the discrete approach. Accord- stages.
ingly, the packed column is modeled in the same way as a staged 1. Governing Equations of the Simple NEQ Model
column where the packed height is vertically divided into a num- To consider the non-equilibrium behavior of a real packed bed,
ber of control volumes, each of them can be considered as a dis- it is necessary to include the interface mass transfer effects in the
crete stage. All of control volumes contain the liquid and the vapor component material balances written for any packing segment.
phase with homogeneous temperature and composition, and are Mass transfer on a two-phase contact segment may be described
referred to as packing segments. In this way, the balance equa- by two distinct material balance equations, one written for the liq-
tions for each packing segment are similar to corresponding equa- uid phase and one for the vapor phase. The component balance
tions for a single stage in a plate column. In the simulation procedure, equations for each phase are stated as:
the mass transfer stages can be identified with these packing seg- L L L
li, j−1− li, j(1+ rj ) + fi, j − Ni, jaj = 0 (1)
ments. The segments are interrelated via mass and energy balance
equations. The configuration of each segment is represented in V V V
vi, j+1− vi, j(1+ rj ) + fi, j + Ni, jaj = 0 (2)
Fig. 1.
In this work, the non-equilibrium behavior of a packed distilla- where li, j, vi, j, and fi, j are component flow rates for liquid, vapor
tion column is described by using the interphase mass transfer rates and feed, respectively, and rj is the dimensionless sidestream flow
calculated based on the vapor phase overall mass transfer coeffi- rate. The term related to feed (fi, j) should be considered in both
cient. So, the packed column consists of a sequence of non-equilib- equations since feed may enter either phase. Writing a material bal-
rium stages, each of which represents a section of packing. Indeed, ance around the entire interface resulted in Eq. (3):
the overall mass transfer coefficient approach is a simplified rate- L V
Ni, j = Ni, j = Nij (3)
based approach in which the temperature of the liquid and the
vapor phase is assumed to be the same. In addition, the temperature where Nij is the interphase molar flux that is determined by fol-
and composition at the interface are not variable in this approach. lowing simple expression in place of equations for rigorous multi-
The key feature of this simple NEQ model is that the structure component mass transfer rate used in rate-based approach:
of the EQ stage model is preserved while the non-ideality of the V OV
Ni, j = c Kij (y*ij − yij) (4)
distillation process due to mass transfer effects is considered. Ac-
cording to this approach, the material balance equations for each
segment are split into two parts, one for each phase, and solved

Fig. 2. Schematic representation of the interphase mass transfer in

Fig. 1. Schematic representation of a segment in the packed bed. a simple NEQ multistage operation.

Korean J. Chem. Eng.

4 H. Poortalari et al.

In the above equation, y*i, j=Ki, jxi, j where Ki, j is the K-value of com- 1− xij
κij = ---------------------------
-
L
(8)
ponent i at stage j, and KijOV is the overall mass transfer coefficient c
Σp=1(xpj/kip)
L

for the vapor phase. Eq. (4) is obtained based on the relation be-
In these equations, kijs represents the binary mass transfer coeffi-
tween the mass transfer driving force and the equilibrium condi-
cients that are estimated by using correlations correspond to the
tion. According to this relation, the rate of mass transfer decreases
type of packings applied in the column.
while conditions proceed toward equilibrium. Fig. 2 schematically
3. Method of Packing Multicomponent Efficiencies
depicts the flow pattern inside the segments according to the inter-
In this approach, the overall mass transfer coefficient for the
phase mass transfer effects [26].
vapor phase is determined using a specific relation obtained from
The heat transfer rates in all segments are ignored, so the energy
the simultaneous solution of the vapor phase material balance and
balance equations for our NEQ model are written similar to EQ
the definition of Murphree efficiency as follows [29]:
model as follows:
V OV MV
c Kij Eij
V V
c
L L
c
V
c --------------- = ---------------- (9)
Hj (1+ rj ) ∑ vk, j + Hj (1+ rj ) ∑ lk, j − Hj+1 ∑ vk, j+1 Vj 1− E
MV
ij
k=1 k=1 k=1
c c
(5)
L F
− Hj−1 ∑ lk, j−1− Hj ∑ fk, j + Qj = 0 where EijMV represents the Murphree efficiency of component i at
k=1 k=1 segment j. As mentioned, efficiencies are not the same for all com-
where c indicates the number of components of the mixture, and j ponents on each segment in multicomponent mixtures. So, the
represents the corresponding stage number. In the simulation, stage estimation of KijOV by Eq. (9) needs multicomponent efficiency cal-
1 corresponds to the condenser and stage N represents the reboiler. culations. For a multicomponent dilute system in a packed distilla-
These end stages are considered as ideal stages even though equi- tion column, the vapor phase material balance equation is written
librium is not achieved along the packed bed. Therefore, the equa- as below in the form of matrix notation:
tions for these stages differ from other equations representing a d( y)
V---------- = − (J )aeAc
V
general non-equilibrium stage. The equations for the condenser and (10)
dh
the reboiler are presented in Table 1.
We used two different approaches to predict the overall mass trans- where ae is the effective interfacial area of the packing, and Ac is
fer coefficients in Eq. (4): the method of effective mass transfer co- the column surface area. By substituting the diffusion flux vector,
efficients and the method of packing multicomponent efficiencies. (JV), with the matrix of overall mass transfer coefficients, [KOV], the
2. Method of Effective Mass Transfer Coefficients following equation is obtained:
According to this method, the “effective” mass transfer coeffi- d( y) V OV
cients for the vapor phase and the liquid phase are combined to V---------- = c [K ](y* − y)aeAc (11)
dh
calculate the overall mass transfer coefficient for the vapor phase
(KijOV) as follows [28]: This equation can be expressed as Eq. (12) by introducing the con-
cept of overall heights of transfer units (HTUs):
1 - ----------
1 Kij
-------------- = V V- + ---------
- (6) d( y) OV −1
V---------- = [H ] (y* − y)
V OV L L
c Kij c κ ij c κij (12)
dh
where κ and κ are the effective mass transfer coefficients for the
V
ij
L
ij
vapor and the liquid phase, respectively. These coefficients are ob- where y* represents the equilibrium vapor composition. Because
tained by the following equations: y* is different for each packing segment, Eq. (12) should be numer-
ically integrated to obtain the vapor composition profile along the
1− yij packed height. To avoid numerical integration, the term (y*−y) is
κij = ---------------------------
-
V
c V
(7)
Σp=1(ypj/kip ) approximated with the arithmetic mean value for a limited seg-

Table 1. Equations for the condenser and the reboiler as ideal stages
Condenser (j=1) Reboiler (j=N)
Material balance li, 1+ vi, 1− vi, 2 − fi, 1 = 0 li, N + vi, N − li, N−1− fi, N = 0
c c c
Energy balance ∑ lk, 1− RR ∑ vk, 1= 0 ∑ lk, N − W = 0
k=1 k=1 k=1

li, 1 vi, 1
Partial Ki, 1 ----------------
c
− -----------------
c
- =0
Σk=1lk, 1 Σk=1vk, 1
c lp, 1
∑ ⎛⎝Kp, 1---------------- ⎞ −1= 0 ( bubble point equation) li, N vi, N
Equilibrium c ⎠ Ki, N----------------
- − ------------------
- =0
p=1 Σk=1lk, 1 c
Σk=1lk, N Σk=1vk, N
c
Total
li, 1 vi, 1
----------------
c
− -----------------
c
- = 0 ( i = 2: c)
Σk=1lk, 1 Σk=1vk, 1

July, 2017
 Separation performance investigation of packed distillation columns using simple NEQ approach 5

ment height [25]. This approximation leads to following equation Rij = − zi⎛ ----- + -----⎞
1 1
for the segment number j: ⎝ kij kic⎠

OV −1 [ K ] j ( x )j + [ K ]j−1( x )j−1 ( y )j + ( y )j+1⎞ In this equation, z is the mol fraction of the appropriate phase and
(y )j − (y)j+1= hj[H ]j ⎛⎝ ------------------------------------------------
- − ------------------------- (13)
⎠
2 2 kij is the binary mass transfer coefficient of the same phase.
After the evaluation of the overall HTU matrix, the HETP val-
where (y)j and (x)j denote the concentration vector of the vapor
ues could be determined by following relation [15]:
and the liquid leaving the segment j, and (y)j+1 and (x)j−1 represent
the concentration vector of the vapor and the liquid entering the OV ln Λ
HETP = H ⎛⎝ -----------⎞⎠ (21)
segment j. [K]j and [K]j−1 are the diagonal matrix of K-values cor- Λ −1
responding to liquid concentrations xj and xj−1, respectively. [HOV]
4. Binary Mass Transfer Coefficients and the Interfacial Area
corresponds to the matrix of overall HTUs in the segment j and hj
is the height of segment j. Now the packing efficiency may be defined It is clear that both methods for the calculation of overall mass
similarly to the Murphree vapor phase plate efficiency as follows transfer coefficients require the estimation of binary mass transfer
[30,31]: coefficients. Several correlations have been proposed for the calcu-
lation of mass transfer coefficients and the interfacial area in packed
MV yj − yj+1 distillation columns. According to packing types, we applied two
Ej = -----------------
- (14)
y*j − yj+1 different correlations for the prediction of mass transfer coeffi-
Finally, the following expression is obtained to predict packing mul- cients: correlation of Bravo et al. 32] for structured packings, and
ticomponent efficiencies based on Eqs. (13), (14) for the segment j: correlation of Onda et al. [33] for random packings.
Correlation of Bravo et al. [32] for Structured Packings:
MV h OV −1 According to this correlation, binary mass transfer coefficients for
(E )j = [A]j ----j[H ]j {[K]j(x)j + [K ]j−1( x)j−1− (y)j − (y)j+1 } (15)
2 each phase are estimated as follows:
where [A]j is a diagonal matrix as follow, and its elements are the V
k = ShVD /deq
V
(22)
reciprocals of term (y*j−yj+1) for c−1 components:
L L
k = 2 D uLe/(πS ) (23)
1
------------------------- 0 L 0
y1, j − y1, j+1
* where ShV is the Sherwood number, deq is the equivalent diameter
1
of a channel, S is the packing corrugation spacing (channel side),
0 ------------------------
- L 0 uLe is the effective liquid velocity, and D is the diffusion coefficient.
[ A ]j = *
y2, j − y2, j+1 (16)
The dimensionless groups used in above equations are defined as:
M M O M
0.8 0.333
1 ShV = 0.0338ReV ScV
0 0 L --------------------------------- V V
y*c−1, j − yc−1, j+1 ReV = deqρ [(uVe + uLe)/μ ] (24)
V V V
ScV = μ /(ρ D )
Eq. (15) is the basic equation for the efficiency-based NEQ model-
ing of a packed distillation column. The main term in this equa- The method of Bravo et al. [32] assumes that the surface is com-
tion is the overall HTU matrix, [HOV], that should be calculated pletely wetted. Therefore, the interfacial area per unit of packing
before the estimation of segment efficiencies. [HOV] can be calcu- volume is equal to the specific packing surface (ae=ap).
lated based on multicomponent mass transfer models which are Correlation of Onda et al. [33] for Random Packings: The
best described with Maxwell-Stefan diffusion equations in connec- Onda correlation predicts the mass transfer coefficients by the fol-
tion with the two-film theory as follow: lowing equations:
[HOV]=[HV]+Λ[HL], Λ=[K] (V/L) (17) V 0.7 0.333
k - ReV ScV
----------- = A------------------------
- (25)
where [HV] and [HL] are the HTU matrix for the vapor and the ap D
V
( ap dp )
2

liquid phases, respectively. They are defined as:

0.667
L ρ ⎞
L 0.333
(Re'L )
k ⎛⎝ -------
- = 0.0051 ---------------------
0.4
[R ]usV
V
( ap dp ) (26)
[H ] = -----------------
- μ g⎠
V L
(18) ScL
ae
L
[R ]usL where dp is the nominal packing size, and ap is the specific surface
[H ] = ----------------
-
L
(19)
ae area of the packing. The parameter A is a constant that takes the
numerical value 2.0 if the nominal packing size is less than 0.012 m
where usV and usL are the superficial velocity for the vapor and the and has the value 5.23 if the nominal packing size is greater than
liquid phases, respectively. [R] is the inverse matrix of mass trans- (or equal to) 0.012 m. The dimensionless groups used to above
fer coefficients that its elements are given by: relations are defined as:
z c z V L
Rii = ----i- + ∑ -----p- (20) ρ uV ρ uL
kic p=1 kip ReV = -----------
V
-, Re'L = ----------
L
- (27)
k≠i μ ap μ ae

Korean J. Chem. Eng.

6 H. Poortalari et al.

V L
μ - μ - of the segment j as follows:
ScV = ------------
V V
, ScL = -----------
L L
ρ D ρD
aj=aevj (30)
The parameter ae is the interfacial area density (m2/m3 packing) that
is determined as follows: where vj is the volume of a packed segment represented by a NEQ
0.1 0.2 stage.
⎛ ⎧ σ 0.75
ReL WeL ⎫⎞
ae = ap⎜1− exp ⎨−1.45⎛⎝ -----c⎞⎠ -----------------------
0.005 ⎬⎟
- (28)
⎝ ⎩ σ FrL ⎭⎠ SIMULATION PROCEDURE
where σ is the surface tension of the liquid and σc is the critical
surface tension of the packing. The dimensionless groups are cal- The simulation procedure of a packed distillation column starts
culated as follows: with dividing the packed height into a number of segments. The
L 2 L 2 segments are considered then as a sequence of mass transfer stages
ρ uL ap uL ρu
- , FrL = ---------
ReL = ---------- - , WeL = ----------L- (29) in the calculations. Simulations consist of several steps that are
μ ap
L g ap σ
repeated for all packing segments. In each step, a set of values for
The interfacial area (aj) is given as the product of ae in the volume the overall mass transfer coefficients is initially obtained from either

Fig. 3. The segment-wise calculation procedure for calculating the overall mass transfer coefficients.

July, 2017
 Separation performance investigation of packed distillation columns using simple NEQ approach 7

effective mass transfer coefficients or packing multicomponent tered in the efficiency-based NEQ model. In the outer loop, over-
efficiencies. Fig. 3 represents step by step procedures for the calcu- all mass transfer coefficients are determined using each set of
lation of the segment-wise overall mass transfer coefficients using estimated efficiencies. Then, by using these values, whole model
both proposed methods. equations are solved in the inner loop to achieve the converged
The obtained overall mass transfer coefficients are used to esti- values of compositions and temperatures. Next, the new efficien-
mate the rate of mass transfer for any calculation segment. The cies are calculated again in the outer loop, and this procedure is
calculated mass transfer rates are then introduced to the mass bal- repeated until the mathematical model of the whole packed col-
ance equations to correct the equilibrium assumption. Eventually, umn is converged. On the other hand, the NEQ model based on
whole equations of the model are iteratively solved to obtain the effective mass transfer coefficients only includes an individual cal-
profiles of compositions and temperature along the packed height. culation loop, so that calculations proceed with determined over-
The sequence of simulation steps related to each NEQ model is all mass transfer coefficients in each iteration to reach the final results
depicted in Fig. 4 as flowcharts. (see Fig. 4(b)). The modified Newton-Raphson method [34] has
As it can be seen in Fig. 4(a), two calculation loops are encoun- been used to simultaneously solve the set of equations obtained

Fig. 4. The sequence of simulation steps. (a) The multicomponent efficiency-based NEQ model. (b) The NEQ model based on effective mass
transfer coefficients.

Korean J. Chem. Eng.

8 H. Poortalari et al.

for each NEQ model. The simulation process based on both NEQ ing and stripping sections is equal. Each of these sections contains
approaches has been implemented in our developed home-code six structure packing elements. The geometry of applied packings
within MATLAB environment. By applying this computational is similar to other commercial corrugated sheet-type packings, but
code, it is possible to simulate a packed column with any number sheets are sandwiched by gauze. Thus, the wetting characteristics
of segments, various packing types and sizes, and column diameter of applied packings are close to that of gauze-type packings. The
etc. Furthermore, the model is known exactly by using this home- second column (introduced by Arwikar [36]) separates another
code, so that it can be modified to examine the validity of various non-ideal mixture consisting of acetone, methyl acetate, and metha-
assumptions. There is also more flexibility when convergence prob- nol in a random packed bed containing Raschig ring packings.
lems are encountered. The column operates under total reflux condition in the normal
atmospheric pressure. The total reflux operation is stabilized so
METHOD OF VALIDATION that the measured internal flow rates are equal to 0.019 mol/s. A
simplified schematic of this laboratory-scale distillation column is
We used two experimental packed distillation columns to eval- shown in Fig. 5(b).
uate our modeling approaches and validate simulation results. The The specifications of both packed columns in addition to geom-
first case (presented by Mori et al. [35]) is a structured packed col- etries of the packing elements are presented in Table 2(a), (b). The
umn that separates a non-ideal ternary mixture containing metha- operating conditions and product compositions required for the
nol, ethanol, and water. a displays a schematic representation of simulation of continues distillation process in the case study I are
mentioned packed distillation column. As shown, the feed is intro- given in Table 3(a).
duced at the middle of the column; therefore, the length of enrich- As mentioned, the packed column in the case study II operates

Fig. 5. Schematic representation of mentioned packed columns in addition to packing specifications. (a) Structured packed column-case
study I (Methanol-Ethanol-Water system). (b) Random packed column-case study II (Acetone-Methyl Acetate-Methanol system).

July, 2017
 Separation performance investigation of packed distillation columns using simple NEQ approach 9

Table 2. Specifications of mentioned packed columns in addition to information about packing elements
(a) Structured packed column (case study I) (b) Random packed column (case study II)
Column specification Column specification
Column height 3m Packed height 68 cm
Packed height 2.2 m Diameter 79 mm
Diameter 21 cm Packing elements
Packing elements Random packing
Type
Type MC-250 (Mitsubishi Company) Raschig ring
Element diameter 0.199 m Nominal packing size 0.25 in (6 mm)
Element height 0.183 m Specific surface area 710 m2/m3
Height of triangle 9.9×10−3 m Material Ceramic
Base of triangle 25.4×10−3 m Critical surface tension 0.055 N/m
Void fraction 0.62
Corrugation spacing 15.6×10−3 m
Specific surface area 250 m2/m3
Void fraction 0.98
Channel flow angle 45o

Table 3. Operating conditions of mentioned packed distillation columns

(a) Structured packed column: continuous operation (b) Random packed column: pseudo-continuous operation
Specification Value Specification Value
Feed flow rate 1.11 mol/s Feed flow rate 1.9×10−4 mol/s
Feed composition Methanol 0.185 Feed composition Acetone 0.165
(mole fraction) Ethanol 0.045 (mole fraction) Methyl acetate 0.234
Water 0.770 Methanol 0.600
Feed temperature 333.15 K Feed temperature Saturated temp.
Feed entrance location 1.1 m (from top) Feed entrance location 58 cm (from top)
Feed quality Liquid Feed quality Liquid
Column pressure 101.4 kPa Column pressure 101.4 kPa
Distillate flow rate 0.19 mol/s Distillate flow rate F/2=9.5×10−5 mol/s
Reflux ratio 6.42 Reflux ratio 200
Reflux temperature 312.55 K Reflux temperature Saturated temp.

in the total reflux condition. The simulation of total reflux opera- The physical properties of pure components and mixture are
tions is complicated by the fact that there is no feed to the col- estimated by using various methods (see Table 4). Correlations for
umn at steady state. To overcome this problem, pseudo-continues the estimation of binary mass transfer coefficients and the effec-
approach is applied. According to this approach, the total reflux tive interfacial area correspond to applied structured packings are
operation is simulated by common approaches for the simulation also presented in Table 4.
of continuous distillation processes so that an arbitrary value with The simulation model of the column is specified by defining the
negligible flow in comparison with internal flow rates is chosen as thermodynamics of the system, the feed rate, the condition and com-
feed flow rate. The composition and entrance location of feed then position of the feed, the column geometry and the packing specifi-
correspond with conditions of one sample point for measuring cations. Since both considered mixtures are very non-ideal, the
compositions along the column. Indeed, the simulated composi- gamma/phi approach [38] is used to describe the vapor-liquid phase
tion profile of the total reflux run satisfies this specified composi- equilibrium (VLE). According to this, the vapor phase fugacity co-
tion. Since the column is operated at total reflux, the reflux flow efficients are computed by the Soave modification of Redlich-Kwong
rate determines the internal flow rates of vapor and liquid streams (SRK) equation of state [39], while the liquid phase activity coeffi-
on each section. Therefore, the reflux ratio is determined so that cients are calculated using two different models, including the
the calculated internal flow rates are adjusted to the experimen- UNIFAC group contribution method [40] and NRTL method
tally measured value (0.019 mol/s). The input parameters for pseudo- [41]. In addition, the vapor pressure is determined by Antoine
continuous simulation of random packed column in case study II equation [42]. The NRTL parameters applied in our simulations
are given in Table 3(b). are listed in Table 5. These parameters are used along with Gij=

Korean J. Chem. Eng.

10 H. Poortalari et al.

Table 4. Methods for estimating physical and mass transfer properties [37]
Physical properties
Vapor molar density Equation of State (SRK)
Liquid molar density Modified Rackett method
Pure gas viscosity Chung method
Mixture gas viscosity Wilke method
Pure liquid viscosity Correlation based on experimental data
Mixture gas viscosity Grunberg and Nissan method
Pure surface tension Sastri-Rao method
Mixture surface tension Tamura method
Mass transfer properties
Binary gas diffusion coefficient Brokaw method
Binary liquid diffusion coefficient Reddy and Doraiswamy method
Bravo et al. [32] for structured packings
Binary mass transfer coefficient Onda et al. [33] for random packings

Table 5. The NRTL parameters for binary mixtures at 101.3 kPa

Case study I; Methanol-Ethanol-Water [43] (τij =Bij/T)
Component i Component j Bij [K] Bji [K] αij
Water Ethanol 624.92 −29.17 0.294
Water Methanol 594.63 −182.61 0.297
Ethanol Methanol 073.41 0−79.17 0.303
Case study II; Acetone-Methyl acetate-Methanol [44] (τij=aij +bij/T)
Component i Component j aij aji bij [K] bji [K] αij
Acetone Methanol 0 −0 −101.886 114.135 0.3
Acetone Methyl acetate 1.4101 −1.9348 −592.968 835.809 0.3
Methanol Methyl acetate 0 −0 −130.505 234.866 0.3

exp(−αijτij). sen based on the effective mass transfer approach. According to these
results, two NEQ models are compared, eventually. For more investi-
RESULTS AND DISCUSSION gation of our proposed NEQ models, the results of HETP-based
EQ modeling approach are also presented.
According to our simple NEQ model, two experimental packed 1. The Segment Size
distillation columns conducted by Mori et al. [35] and Arwikar The number of segments can be considered as a parameter that
[25,36] are used to implement the simulation procedures. The main considers the backmixing in packed distillation columns. When
purpose of simulations is to evaluate the separation performance the number of segments is very high, basically no backmixing occurs
of considered packed columns under various conditions by the in the column, and, therefore, the separation becomes maximum.
efficiency-based NEQ model and the effective mass transfer NEQ Since backmixing and other non-ideal phenomena do exist in real
model. In the first step, the effect of packed segment size and liq- columns, an appropriate number of segments should be used.
uid activity model are investigated on the prediction of the separa- Indeed, the accuracy of the method is related to the number of
tion performance of the structured packed column (case study I). calculation segments specified for a certain packing height. There-
Since the efficiency-based NEQ model involves two calculation fore, the selection of this factor could have a significant impact on
loops and several matrix calculation operations, the complete sim- simulation results. Consequently, it is one of the parameters that
ulation of the packed column with this model requires more com- should be cautiously chosen in the mass transfer models. How-
putation time compared to the simulation by the effective mass ever, this factor cannot be determined in advance. When experi-
transfer model. However, the simulation with the efficiency-based mental data are available, the simulation results should be compared
approach gives more detailed results about the column separation to these data to determine the optimal number of segments. Accord-
performance, such as packing efficiency and HTUs. It also gives ingly, the simulations are performed several times while a differ-
more precise information about the mass-transfer phenomena ent number of segments are used in each run. Therefore, the size
such as the degree of backmixing on each segment. Therefore, the of segments is decreased step by step to minimize the differences
proper segment size and the appropriate activity model are cho- between simulation and observed data. The resulting profiles of tem-

July, 2017
 Separation performance investigation of packed distillation columns using simple NEQ approach 11

Fig. 6. The resulting profiles from the effective mass transfer NEQ model for several segment sizes in addition to some experimental data. (a)
Temperature profiles. (b) Liquid composition profiles.

perature and component compositions in addition to experimental properties of the system. In non-ideal systems, interactions exist
data along the column are presented in Fig. 6(a), (b). Simulations are between components that differ from other components in respect
performed based on the effective mass transfer NEQ model. to size, character, polarity and hydrogen bonding. The vapor-liq-
As shown in Fig. 6, there is a significant difference in the tem- uid equilibrium (VLE) of these mixtures is much more compli-
perature and composition profiles between the model results and cated than systems including non-polar species without hydrogen
the experimental data, especially for large segment sizes. However, bonding such as hydrocarbons. Indeed, in a thermodynamically
this difference at the top and the bottom of the column is less than non-ideal system, the thermodynamic and physical properties of
that in the middle section. It can be obviously seen that the pro- the vapor and the liquid phases are quite different for each com-
files obtained by the model incline to experimental results, as the ponent.
size of calculation segments decreases. Since the size of segments The considered structured packed column in case study I sepa-
directly affects the total computation time required for the conver- rates a ternary mixture consisting of methanol, ethanol, and water.
gence, excessive increase in the number of segments is also not This system is highly non-ideal due to the presence of polar com-
desirable. According to the obtained profiles for different segment ponents with hydrogen bonding. The non-ideal nature of the sys-
sizes, when the total packed height is divided into 25 segments tem directly affects the VLE calculations. In this paper, the effect
(meaning 8.8 cm for each segment), experimental profiles are pre- of VLE data on the simulation results has been analyzed using two
dicted properly. Therefore, this number of segments is used for specially chosen activity correlations: the UNIFAC model and the
further simulations. NRTL equation. A comparison between measured and calculated
2. The Liquid Activity Model liquid compositions using these liquid activity models is provided
The thermodynamic equilibrium constant (K-value) is normally in Fig. 7 for each component. The results are obtained by using
obtained from a variety of methods depending on the system com- the effective mass transfer NEQ model. Clearly, the disagreement
plexity. For most gas processing applications, an equation of state is between experimental data and simulation results obtained by the
often used, and this directly yields the fugacity coefficients. When UNIFAC equation is considerable. However, the profiles resulting
the liquid phase becomes more non-ideal, some applications will from the NRTL model show a good agreement with experimen-
require a Gibbs excess free energy model, which gives the fugacity tal results.
coefficients through the activity coefficients. Therefore, it is neces- A further comparison between outputs of two liquid activity
sary to select appropriate methods for estimating thermodynamic models is performed based on temperature profiles. The tempera-

Fig. 7. Liquid composition profiles resulting from the effective mass transfer NEQ model with different activity model in addition to experi-
mental data along the column.

Korean J. Chem. Eng.

12 H. Poortalari et al.

Table 6. The quantitative comparison of temperatures between sim-

ulation results and experimental data for structured packed
column (case study I)
h [m] Exp. NEQ 01 (%e) NEQ 02 (%e) EQ (%e)
0.37 340.78 341.34 −0.16 340.55 −0.07 342.05 −0.37
0.73 343.63 343.87 −0.07 343.08 −0.16 343.62 −0.00
1.47 349.99 350.43 −0.13 350.37 −0.11 350.37 −0.11
1.83 350.49 352.34 −0.53 351.92 −0.41 351.95 −0.42
2.20 358.18 356.50 −0.47 355.18 −0.84 355.08 −0.86
% Mean relative error −0.27 −0.32 −0.35
NEQ 01: Non-equilibrium modeling based on effective mass trans-
fer coefficients
NEQ 02: Non-equilibrium modeling based on packing efficiency
EQ: Equilibrium modeling based on HETP concept
Fig. 8. The temperature profile resulted from the effective mass trans-
fer NEQ model for different activity models in addition to
measured temperatures.
the bed height is 25 cm for the mentioned structured packed col-
umn, and 12.85 cm for the considered random packed column.
ture profiles resulting from NEQ simulations for each activity model Fig. 9(a), (b) show the temperature profiles and the liquid com-
are depicted in Fig. 8 where the measurement data along the col- position profiles of each component along the column height re-
umn are also presented. As is clear in the figure, the temperature sulted from the simulation of structured packed column in case
profile obtained from the UNIFAC model could not predict the study I. In this figure the measurement points taken along the packed
measured data while simulation results from the NRTL model bed are also presented.
have good agreement with experimental data. Therefore, it can be The temperature profiles (Fig. 9(a)) show that all modeling ap-
said that the NRTL equation is the appropriate model for VLE cal- proaches can predict the measured temperatures precisely. The quan-
culations to simulate the mentioned column. titative comparisons given in Table 6 confirm this where the maxi-
3. The NEQ Model mum average discrepancy is less than 0.5%. As shown in Fig. 9(a),
According to the obtained results for the proper number of pack- the predicted profiles from all modeling approaches are almost
ing segments and the appropriate model of liquid activity coeffi- similar though a slight difference is seen at the feed position.
cients, a detailed comparison between simulation results of each The detailed investigation in Fig. 9(b) denotes that the efficiency-
NEQ model was performed. Our investigations confirm that the based NEQ model predicts the measured compositions more pre-
chosen number of segments and activity model are also adaptable cisely, especially for water and methanol, while the NEQ model
for the simulation of random packed column in case study II. There- based on effective mass transfer coefficients could only predict the
fore, the results for both packed columns are presented in the fol- composition of methanol in enriching section somewhat better.
lowing figures. These results were obtained from both NEQ models The predictions of both NEQ models show some deviations rela-
in addition to results of HETP-based EQ modeling approach. The tive to experimental data for ethanol in enriching section. It is
average HETP value obtained from evaluation of HTU matrix along noticeable that the profiles resulting from EQ modeling approach

Fig. 9. The obtained profiles from simulation of structured packed column in addition to experimental data. (a) The temperature profiles.
(b) The liquid composition profiles.

July, 2017
 Separation performance investigation of packed distillation columns using simple NEQ approach 13

Table 7. The quantitative comparison of liquid compositions between simulation and experimental results for structured packed column
(case study I)
MeOH EtOH Water
h [m] Exp. NEQ 01 (%e) NEQ 02 (%e) EQ (%e) h [m] Exp. NEQ 01 (%e) NEQ 02 (%e) EQ (%e) h [m] Exp. NEQ 01 (%e) NEQ 02 (%e) EQ (%e)
Cond. 0.85 0.849 −0.3 0.886 04.1 0.891 4.7 Cond. 0.11 0.104 −5.6 0.088 −20 0.084 −24 Cond. 0.04 0.047 15 0.025 −38 0.025 −39
0.36 0.75 0.754 0.0 0.791 04.8 0.771 2.2 0.36 0.19 0.147 −22 0.143 −24 0.156 −17 0.36 0.06 0.096 62 0.064 7.3 0.071 18
0.73 0.59 0.627 5.6 0.651 09.6 0.609 2.6 0.73 0.27 0.187 −31 0.201 −26 0.220 −19 0.73 0.13 0.185 38 0.148 10 0.171 27
1.47 0.27 0.321 19 0.310 14.7 0.299 11 1.47 0.18 0.152 −18 0.173 −6.0 0.186 1.2 1.47 0.54 0.526 −3.1 0.517 −4.9 0.515 −5.2
1.83 0.23 0.262 11 0.248 05.6 0.241 2.7 1.83 0.20 0.155 −25 0.182 −11 0.202 −1.6 1.83 0.56 0.583 4.3 0.568 1.8 0.556 −0.4
Reb. 0.04 0.048 26 0.040 05.9 0.039 3.1 Reb. 0.03 0.033 9.3 0.036 20 0.037 24 Reb. 0.93 0.919 −1.4 0.924 −0.9 0.924 −0.9
% Mean error 10.5 07.5 4.4 18.3 17.9 14.5 20.5 10.5 15.2

Fig. 10. The liquid composition profiles obtained from simulation of random packed column in addition to experimental data (case study II).

approximately similar to the NEQ models. This originates from based NEQ method and HETP-based EQ method. According to
this fact that the HETP value for the considered structured packed the effective mass transfer coefficient method, the prediction of
column is accurately estimated. NEQ approach is accompanied by considerable departure relative
The quantitative comparisons between predicted and meas- to measurement data, particularly for acetone. This reveals that the
ured compositions are separately given in Table 7. The average dis- overall mass transfer coefficient has not been acceptably estimated
crepancies are also given in this Table. It is obvious that the accu- by effective mass transfer coefficients for mentioned non-ideal
racy of both the efficiency-based NEQ model and also HETP- mixture.
based EQ model is almost the same. A detailed comparison between simulation and measurement
Fig. 10 displays the variation of liquid mole fraction separately data is given in Table 8. As clearly observed, the simulation with
for each component along the random packed column in case study efficiency-based NEQ model has the lowest difference with experi-
II. Simulations were performed based on pseudo-continuous mental data, though the discrepancy of results obtained from HETP-
method so that the internal flow rates in the calculations corre- based EQ model is almost similar to the efficiency-based model.
sponded to experimentally measured value at total reflux opera- Note that the predictions based on all simulation models are
tion of column (0.019 mol/s). As seen, the trend of variations is accompanied by some deviations relative to experimental results.
fairly predicted by simulation methods, especially by efficiency- A source of these deviations can be related to simplifying assump-

Table 8. The quantitative comparison of liquid compositions between simulation and experimental results for random packed column (case
study II)
Acetone Methy acetate Methanol
h [m] Exp. NEQ 01 (%e) NEQ 02 (%e) EQ (%e) Exp. NEQ 01 (%e) NEQ 02 (%e) EQ (%e) Exp. NEQ 01 (%e) NEQ 02 (%e) EQ (%e)
Cond. 0.196 0.274 40 0.258 32 0.262 34 0.373 0.450 21 0.431 16 0.435 17 0.431 0.275 −36 0.311 −28 0.303 −30
0.25 0.198 0.302 52 0.260 31 0.268 35 0.315 0.374 19 0.352 12 0.360 14 0.488 0.324 −34 0.388 −20 0.372 −24
0.42 0.181 0.295 63 0.239 32 0.248 37 0.270 0.277 2.6 0.263 −2.6 0.271 0.2 0.548 0.428 −22 0.498 −9.1 0.481 −12
0.45 0.185 0.287 55 0.231 25 0.241 30 0.280 0.254 −9.4 0.244 −13 0.251 −10 0.535 0.459 −14 0.524 −2.0 0.509 −4.9
0.53 0.169 0.249 48 0.206 22 0.209 24 0.252 0.181 −28 0.191 −24 0.192 −24 0.579 0.569 −1.7 0.604 4.3 0.599 3.5
0.58 0.165 0.210 27 0.184 12 0.182 10 0.234 0.133 −43 0.158 −33 0.153 −35 0.600 0.657 9.5 0.658 10 0.666 11
0.63 0.154 0.162 5.4 0.161 4.2 0.156 1.4 0.213 0.087 −59 0.124 −42 0.118 −44 0.633 0.751 19 0.715 13 0.725 15
0.68 0.121 0.114 −5.6 0.136 12 0.131 8.2 0.140 0.051 −63 0.095 −32 0.087 −38 0.740 0.835 13 0.769 3.9 0.782 5.6
Reb. 0.073 0.056 −23 0.073 −0.5 0.069 −5.9 0.070 0.019 −73 0.038 −46 0.034 −51 0.856 0.925 8.1 0.890 3.9 0.897 4.8
% Mean error 35.5 19.0 20.6 35.4 24.4 25.9 17.4 10.5 12.2

Korean J. Chem. Eng.

14 H. Poortalari et al.

Fig. 11. Estimated efficiencies for different components as a functions of the packing height. (a) The structured packed column (case study
I). (b) The random packed column (case study II).

tions applied to develop simulation methods. However, the non- cients were determined for each segment. Results of temperature
uniform distribution of local fluid flows in different cross-sections and composition profiles were thus obtained and then compared
of the column is also accounted as another source of the discrep- with the experimental measurements from each packed column.
ancy of simulation results with experimental data. In fact, in packed The size of packing segments as an effective parameter in the
columns, particularly with large diameters, some uneven distribu- simulation procedure was investigated in the first step. The com-
tion of temperature and component concentrations may have arisen parison of different simulation results with various segment sizes
into the column. This issue could affect on the separation efficiency and experimental data for the structured packed column showed
of the mixture and consequently on final results of the simulation that 25 segments are a proper choice for further simulations. In
[45-49]. the next step, the non-ideality of liquid mixture separated in the
In Fig. 11, the estimated component efficiencies for both packed structured packed column was investigated based on two liquid
columns are displayed as a function of the packed height for 25 activity models, including the UNIFAC model and the NRTL
segments. These efficiency values are introduced to the efficiency- model. A comparison between measured and calculated profiles
based NEQ model for correcting the mass transfer in each seg- using these liquid activity models showed that the NRTL model
ment. According to this, the temperature and composition profiles could predict experimental results more precisely than the UNI-
are obtained. As observed, some component efficiencies change FAC model, and so this model was considered as a base for fur-
along the packed beds. This means that some components can have ther simulations. Our considerations illustrated that these choices
different mass transfer characteristics along the columns. Further- for number of segments and liquid activity model are also adapt-
more, in both columns, oscillations are observed in component able for the simulation of random packed column in case study II.
efficiencies at different sections of each column due to fluctuation Finally, the results of each NEQ model were compared with the
of compositions. experimentally measured data. For further comparisons, the results
obtained from so-called HETP-based EQ model are also presented.
CONCLUSION The detailed comparisons revealed that the efficiency-based NEQ
model had a better agreement with experimental results. How-
The separation performance of two multicomponent packed ever, the accuracy of the EQ model is almost equivalent to effi-
distillation columns, a structured packed column and a random ciency-based NEQ model, due to precise estimation of HETP values
packed column, was simply evaluated using a developed NEQ mod- for each packed column. The quantitative comparisons showed an
eling approach. Non-idealities of the distillation process due to inter- acceptable agreement between predicted and measured concentra-
phase mass transfer effects were evaluated by the estimation of tions along the columns, so that the maximum average difference
mass transfer rates for each phase. The rates of mass transfer were obtained by efficiency-based NEQ model was about 18% and 25%
determined using the overall mass transfer coefficients as simple, for the structured packed column and the random packed col-
on-hand method. We proposed distinct estimation methods to umn, respectively.
obtain the overall mass transfer coefficients, including the effec- By applying the efficiency-based approach, it was possible to
tive mass transfer method and the packing multicomponent effi- evaluate the mass transfer characteristics of each component along
ciency method. In this way, we could simulate both packed distillation the packed bed as packing efficiency values. These efficiency val-
columns with NEQ effects in a similar structure to EQ stage model. ues were successfully incorporated into the mass transfer models
A developed MATLAB home-code was used to implement the to predict the temperature and composition profiles along packed
simulation algorithms. For modeling, the packed height was divided columns. Accordingly, it can be concluded that our simple NEQ
into a number of segments and the overall mass transfer coeffi- approach provides a reliable way to estimate the performance of

July, 2017
 Separation performance investigation of packed distillation columns using simple NEQ approach 15

real packed columns. ical Engineers Symposium Series, 152, 800 (2006).
2. G. M. Cordeiro, S. R. Dantas, L. G. S. Vasconcelos and R. P. Brito,
NOMENCLATURE Adv. Chem. Eng. Sci., 3, 1 (2013).
3. J. Seader, E. J. Henley and D. Keith, Separation process principles:
ae : effective interfacial area [m2/m3] Chemical and biochemical operations, Hoboken, NJ, Wiley (2011).
ap : apparent specific surface area [m2/m3] 4. Y. S. Lee, M. G. Kim, D. M. Ha, A. Oda, C. Ito, T. Aragaki and H.
aj : interfacial area of segment j [m2/m3] Mori, Korean J. Chem. Eng., 14, 321 (1997).
Ac : column surface area [m2] 5. Y. Amini, J. Karimi-Sabet and M. Nasr Esfahany, Chem. Eng. Tech.,
cp : molar concentration for each phase [gmol/cm3] 40, 581 (2017).
c : the number of components 6. I. J. Halvorsen and S. Skogestad, J. Nat. Gas Sci. Eng., 3, 571 (2011).
Dp : diffusion coefficient for each phase [m2/s] 7. I. D. G. Chaves, J. R. G. López, J. L. G. Zapata, A. L. Robayo and
deq : equivalent diameter of a channel [m] G. R. Niño, Process analysis and simulation in chemical engineer-
(EMV)j : vector of Murphree segment efficiencies ing, Springer (2016).
EMV : Murphree vapor phase segment efficiency 8. R. Taylor, R. Krishna and H. Kooijman, Chem. Eng. Prog., 99, 28
f pi, j : component feed flow rate for each phase [mol/s] (2003).
hj : hsegment height [m] 9. R. Baur, A. Higler, R. Taylor and R. Krishna, Chem. Eng. J., 76, 33
Hj : enthalpy [J/mol] (2000).
[HOV] : matrix of overall HTUs [m] 10. M. F. Mendes, Hetp evaluation of structured and randomic packing
[Hp] : matrix of HTUs for each phase [m/s] distillation column, INTECH Open Access Publisher (2011).
(JV) : vector of vapor diffusion fluxes [mol/(m2·s)] 11. A. Gorak and A. Vogelpohl, Sep. Sci. Technol., 20, 33 (1985).
j : the corresponding stage number 12. C. Skowlund, M. Hlavinka, M. Lopez and C. Fitz, in Proceedings
[K] : matrix of vapor-liquid equilibrium constants (K-values) of Gas Processesors Association (2012).
Ki, j : vapor-liquid equilibrium constant 13. T. N. Mosorinac, J. J. Djurovic and J. B. Savkovic-Stevanovic, Petro-
[KOV] : the matrix of overall mass transfer coefficients [m/s] leum Coal, 53, 194 (2011).
KOV i, j : overall mass transfer coefficient [m/s] 14. R. Krishnamurthy and R. Taylor, Ind. Eng. Chem. Process Des. Dev.,
kpi, j : binary mass transfer coefficient for each phase [m/s] 24, 513 (1985).
κ i,pj : effective mass transfer coefficient for each phase [m/s] 15. R. Taylor and R. Krishna, Multicomponent mass transfer, Wiley
Lj : liquid flow rate [mol/s] (1993).
li, j : component liquid flow rate [mol/s] 16. R. Krishnamurthy and R. Taylor, AIChE J., 31, 449 (1985).
Qj : heat load [J/mol] 17. R. Krishnamurthy and R. Taylor, AIChE J., 31, 456 (1985).
rjp : side-stream flow rate for each phase, dimensionless 18. R. Krishnamurthy and R. Taylor, AIChE J., 31, 1973 (1985).
RR : reflux ratio 19. J. Aittamaa, Kemi, 8, 295 (1981).
[Rp] : inverse matrix of mass transfer coefficients for each phase, 20. J. Ilme, Estimating plate efficiencies in simulation of industrial scale
dimensionless distillation columns, Lappeenranta University of Technology (1997).
Re : Reynolds number, dimensionless 21. K. T. Klemola, Efficiencies in distillation and reactive distillation, Finn-
S : corrugation spacing [m] ish Academy of Technology (1998).
Sh : Sherwood number, dimensionless 22. K. Jakobsson and J. Aittamaa, Comparison of plate efficiency esti-
Sc : Schmidt number, dimensionless mation models to experimental results of pilot scale: A case study,
uVe : effective vapor velocity [m/s] American Institute of Chemical Engineers (2001).
uLe : effective liquid velocity [m/s] 23. J. K. Ilme, K. I. Keskinen, V. L. Markkanen and J. R. Aittamaa, in
uSV : superficial vapor velocity [m/s] Institution of Chemical Engineers Symposium Series, Hemsphere
uLV : superficial liquid velocity [m/s] Publishing Corporation, 142, 497 (1997).
Vj : vapor flow rate [mol/s] 24. K. Jakobsson, J. Aittamaa, K. I. Keskinen and J. Ilme, in Proceed-
vi, j : component vapor flow rate [mol/s] ings of the International Conference on Distillation & Absorption
vj : volume of a packed segment [m3] (on CD) (2002).
W : bottom product flow rate 25. K. I. Keskinen, A. Kinnunen, L. Nyström and J. Aittamaa, in Pro-
xi, j : actual liquid mole fraction ceedings of the International Conference on Distillation & Absorp-
yi, j : actual vapor mole fraction tion (on CD) (2002).
y*i, j : vapor composition in equilibrium with xi, j 26. H. Poortalari, J. K. Sabet and F. Varaminian, Sep. Sci. Technol., 52.11,
zi : mole fraction of the appropriate phase 1885 (2017).
ρV : vapor density [kg/m3] 27. C. D. Holland, Fundamentals of multicomponent distillation, McGraw-
μV : vapor viscosity [Pa·s] Hill (1981).
28. M. Powers, D. Vickeryt, A. Arehole and R. Taylor, Computers Chem.
REFERENCES Eng., 12, 1229 (1988).
29. R. Waggoner and G. Loud, Computers Chem. Eng., 1, 49 (1977).
1. G. Niggemann, S. Gruetzmann and G. Fieg, in Institution of Chem- 30. A. Medina, N. Ashton and C. McDermott, Chem. Eng. Sci., 33,

Korean J. Chem. Eng.

16 H. Poortalari et al.

331 (1978). contribution method and its application in distillation column design
31. A. Medina, N. Ashton and C. McDermott, Chem. Eng. Sci., 34, and steady state simulation, Doctoral dissertation, National Insti-
1105 (1979). tute of Technology, Rourkela (2010).
32. J. L. Bravo, J. Rocha and J. Fair, Hydrocarbon Process., 64, 91 41. J.-C. De Hemptinne and J.-M. Ledanois, Select thermodynamic
(1985). models for process simulation: A practical guide using a three steps
33. K. Onda, H. Takeuchi and Y. Okumoto, J. Chem. Eng. Jpn., 1, 56 methodology, Editions Technip (2012).
(1968). 42. G. W. Thomson, Chem. Rev., 38, 1 (1946).
34. J.-T. Lee and W.-K. Lee, Korean Chem. Eng. Res., 20, 473 (1982). 43. J. O. Nava and R. Krishna, Chem. Eng. Process., 43, 305 (2004).
35. H. Mori, A. Oda, T. Aragaki and Y. Kunimoto, J. Chem. Eng. Jpn., 44. A. Plus, Aspen plus user guide, Aspen Technology Limited, Cam-
29, 307 (1996). bridge, United States (2003).
36. K. J. Arwikar, Mass transfer in packed column absorption and mul- 45. A. Pavlenko, N. Pecherkin, V. Y. Chekhovich, V. Zhukov and S.
ticomponent distillation, Doctoral Dissertation, University of Cali- Sunder, Theor. Found. Chem. Eng., 43, 1 (2009).
fornia, Santa Barbara (1981). 46. A. Pavlenko, V. Zhukov, N. Pecherkin, V. Y. Chekhovich, S. Sun-
37. B. E. Poling, J. M. Prausnitz and J. P. O’connell, The properties of der and P. Houghton, Theor. Found. Chem. Eng., 44, 869 (2010).
gases and liquids, McGraw-hill New York (2001). 47. A. Pavlenko, X. Li, V. Zhukov, N. Pecherkin, O. Volodin, A. Sur-
38. J. M. Prausnitz, R. N. Lichtenthaler and E. G. de Azevedo, Molecu- taev, X. Gao, L. Zhang, H. Sui and H. Li, J. Eng. Phys. Thermo-
lar thermodynamics of fluid-phase equilibria, Pearson Education phys., 24, 210 (2015).
(1998). 48. Ž. Olujić and H. Jansen, Chem. Eng. Res. Des., 99, 2 (2015).
39. A. Danesh, Pvt and phase behaviour of petroleum reservoir fluids, 49. A. N. Pavlenko, V. E. Zhukov, N. I. Pecherkin, X. Li and H. Sui,
Elsevier (1998). Journal of Physics: Conference Series, IOP Publishing, 754, 042012
40. M. Behera, Vapour liquid equillibrium modeling using unifac group (2016).

July, 2017

View publication stats