Professional Documents
Culture Documents
Publication date:
2021
Document Version
Publisher's PDF, also known as Version of record
Citation (APA):
Rasmussen, J. B. (2021). Cyclic Distillation Technology. Technical University of Denmark.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright
owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
You may not further distribute the material or use it for any profit-making activity or commercial gain
You may freely distribute the URL identifying the publication in the public portal
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
DTU Chemical Engineering
Department of Chemical and Biochemical Engineering
PhD Thesis
November, 2021
PhD Thesis
November, 2021
By
Jess Bjørn Rasmussen
Supervision by
Assoc. Prof. Jakob Kjøbsted Huusom, DTU Chemical Engineering
Assoc. Prof. Jens Abildskov, DTU Chemical Engineering
Prof. Xiangping Zhang, CAS Institute of Process Engineering
The SinoDanish Center for Education and Research (SDC) provided partial fund
ing for this PhD project.
The thesis consists of nine chapters and one appendix describing the outcome
from this PhD project, including references to the academic articles and contribu
tions prepared during this project, which have been peerreviewed and published
or submitted to scientific journals.
Signature
November, 2021
Date
Thanks a lot to my supervisors, especially Jakob and Jens, who have greatly sup
ported me in the three years of hard work. Also, thanks to Xiangping for giving me
the opportunity to visit Beijing, even though my total stay was shortened. Thanks
to Vladimir for taking an interest in my work and for sharing ideas on cyclic distil
lation.
Without the wonderful colleagues I have met at PROSYS, it would have been
three difficult years. Thanks to the good friends I made during my PhD for helping
me to get the best out of life as possible, to name a few I would like to thank (in
no particular order): Franz, Gisela, Simoneta, Niko, Vicente, Sigyn, Enrico, Pau,
Julie, and Andrea. And special thanks to Yuanmeng for translating services.
Thanks to my friends from high school; Quyen, Lauge, Malene, Jeppe and Casper,
and thanks my old friends from Havdrup; Mike, Rasmus, Jannik, Frederik and
Nicklas.
Thanks to everybody else who have been a part of my life and work at DTU and
for all of your help and support.
In this thesis, the cyclic distillation technology has been studied and analysed in or
der to be able to understand the process. It is easier to propose a cyclic distillation
as an alternative to conventional distillation with a higher process understanding.
A mass and energy balance stage model is proposed, which is a high fidelity model
accounting for timedependent temperature and vapour flow rate and allowing for
multiple feed locations, and side draws. The model performance was evaluated
and compared to previous models. It was shown that if there is a high development
in the stage temperature or vapour flow rate over a vapour flow rate, the proposed
mass and energy balance model would be a suitable choice.
The presented model is further expanded to account for reactive cyclic distillation
processes, including the reaction heat. Different reactive cyclic distillation cases
are presented, and a detailed analysis of the stage behaviour over the vapour
flow period is shown for the methyl tertbutyl ether case. This stage behaviour
analysis showed that as the vapour flow period progresses, significant changes
in the reaction and the separation affects the process.
The performance of three different reactive cyclic distillation cases was also eval
uated for a disturbance in the inputs. Furthermore, the feasibility of a reactive
cyclic distillation was discussed. Three performance indicators are proposed: the
extent of reaction over a vapour flow period, the relative distance to equilibrium at
the end of a vapour flow period and the mean Damköhler number over a vapour
flow period. Of these, the extent of reaction and the Damköhler number is useful
for investigating the performance and feasibility of a reactive cyclic distillation pro
All in all, the work presented in this thesis shows development of cyclic distilla
tion technology and how new models can help to make more high fidelity studies
useful for process analysis in terms of designing, control strategies and process
feasibility.
为进一步详细了解循环精馏技术,本论文对其进行了深入研究及分析。通过对该
新型精馏过程的深入了解,提出循环精馏替代传统精馏技术的可靠方案。
首先建立了该过程质量与能量平衡的高保真模型,该模型充分考虑过程中随时间
变化的温度及蒸汽流速参数,允许多位置进料及侧线取料。随后对该模型进行了
性能评估,并与之前的模型进行比较,结果表明,该模型对于塔板温度或蒸汽流
速变化较大的过程具有更好的适用性。
接着将该模型进一步扩展至包括反应热在内的反应循环精馏过程。通过研究不同
的反应循环精馏案例,尤其是甲基叔丁基醚循环精馏过程,对蒸汽流动期间的塔
板行为进行了详细分析,结果表明,随着蒸汽流动周期的变化,反应与分离效果
变化显著。
此外,本文还研究了输入波动时三种不同反应循环精馏过程的性能,并讨论了反
应循环精馏过程的可行性。为评估该动态过程,提出了衡量过程性能的三个指标:
蒸汽流动周期的反应程度、蒸汽流动周期结束时至达到平衡的相对距离以及蒸汽
流动周期内的平均达姆科勒数。其中,蒸汽流动周期的反应程度及平均达姆科勒
数对于研究反应循环精馏过程的性能及可行性至关重要。蒸汽流动周期结束时至
达到平衡的相对距离可验证该化学平衡的假设是否有效。
为进一步强化循环精馏设计方法,基于现有的反应循环精馏可行性条件增加了部
分研究,但该设计方法仍需深入迭代以探究可替代传统精馏过程的严格规范。
总之,本工作通过建立循环精馏过程的物质及能量高保真守恒模型,并从过程设
计、动态控制策略及工艺可行性方面对循环精馏技术进行了深入探究。
Den foreslåede model er yderligere blevet udvidet til at redegøre for reaktive cyk
liske destilleringsprocesser, således at modellen indeholder også reaktionsvar
men. Forskellige reaktive cyklisk destilleringsprocesser er præsenteret og en
detaljeret analyse af bundopførelsen over en dampflowperiode er vist for metyl
tertbutyl æter procssen. Denne bundopførelsesanalyse viste at når dampflow
perioden kører så vil væsentlige ændringer i både reaktionen og separation have
en stor påvirkning på processen.
Alt i alt viser det stykke arbejde præsenteret i denne afhandling et skridt i videreud
viklingen af cyklisk destilleringsteknologi og hvordan nye modeller kan hjælpe med
at simulere høj nøjagtighedsstudier, der kan bruges til procesanalyse af design,
regluering og evaluering.
Acknowledgements v
Abstract vii
摘要 (Chinese abstract) ix
Resumé xi
Nomenclature xiii
Contents xv
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Thesis structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Publications and conference contributions . . . . . . . . . . . . . . 9
1.5.1 Journal publications . . . . . . . . . . . . . . . . . . . . . . 9
1.5.2 Conference contributions . . . . . . . . . . . . . . . . . . . 9
6.3 General trends and observations for the stage and column analysis 110
6.4 Chapter conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7 Discussion 115
7.1 Limitations of the proposed mass and energy balance stage model 116
7.2 Reactive cyclic distillation analysis and process design . . . . . . . 117
Bibliography 127
1.1 Background
Distillation is a common fluid separation process in the chemical and biochemical
industries. Distillation processes have been in use for thousands of years, with
early civilisations distilling herbs and essential oils in small batches [3]. These
smallscale batch distillation processes continued up through thousands of years
until the industrial revolution in the 19th century when the use of large scale indus
trial continuous distillation columns became more and more widespread. Since
then, distillation has become an important and crucial separation technique in
many chemical and biochemical processes, from the oil and gas industry to the
pharmaceuticals and food industry. It is estimated that over 40,000 distillation
columns were in operation worldwide in 2013 in these different kinds of industries
[4].
A conventional distillation column can usually be divided into two parts: the lower
part of the column, known as the stripping section, and the upper part, known as
the rectifying section. In the stripping section of a distillation column, the more
volatile components are stripped from the liquid by the vapour. In the rectifying
section, the vapour is enriched or rectified by concentrating the light components.
The two sections are usually divided by the feed section, where a feed inlet into
the column is located.
1
High purity A High purity C
Rectifying Rectifying
...
...
section section
A
...
...
...
...
...
...
Liquid feed
Reactive zone
...
...
...
B
...
...
...
...
...
...
Vapour feed
Stripping Stripping
...
...
section NT section
While distillation is a versatile and widely used separation process, it is also often
expensive both regarding operational cost and capital cost due to low tray effi
ciencies and resulting high energy requirements. In 2016 it was estimated that
around 79 % of the energy consumption in the US and 1015 % in the world were
used for distillation processes alone [7].
The high energy consumption in distillation processes comes from two major lim
itations: the low thermodynamic efficiency and the backmixing of liquid. This
results in high energy demand and low separation efficiency. The low thermody
namic efficiency comes from a high energy supply in the bottom of the column
and the energy removal at a low temperature in the top, leading to an exergy loss
[8]. It is often costly to provide energy in the reboiler, typically steam is used, and
the energy can rarely be recovered in the top of the column in the condenser [4].
Generally, the thermodynamic efficiency is low for distillation processes, around
520 % [4, 8].
Backmixing occurs in the liquid holdup, which comes from the continuous flows
in a conventional distillation process. Any stage in a distillation column will have a
liquid phase holdup, and a liquid phase in and outlet. If the liquid flow on the stage
moves from the inlet to the outlet with a plug flow, then no backmixing occurs.
However, this is rarely, if ever, the case. Usually, there is a degree of backmixing
in the stage holdup, where the liquid is not moving in plug flow across the stage.
Some of the liquid stays on the stage for a long time, thus the separation efficiency
is reduced.
The high costs are especially true for difficult separations, such as highpressure
distillation or if an azeotrope is present. Azeotropes can be found in many non
ideal mixtures, where at a specific composition for a given pressure, the mixture
has a higher or lower boiling point than the pure component boiling points. It is
thermodynamically impossible to overcome such an azeotrope in a single con
ventional distillation column, so alternative distillation processes are often used.
These processes that can handle azeotropes are pressure swing distillation, where
two columns in series operate at different pressures, thus moving the azeotrope,
or extractive distillation, where a solvent is added to move one of the components
to the solvent phase where no azeotrope is present. The solvent can then be
separated in another distillation column. In a reactive distillation, the challenge of
azeotropes can, in some cases, be overcome, as the simultaneous reaction and
separation change the composition on the trays. However, there is a chance of
reactive azeotropes, for which neither the reaction nor separation can overcome
the azeotrope.
(SDG), including 169 targets, for the social, economical and environmentally sus
tainable development in a global perspective [9]. With these SDGs, the UN wants
to set goals for the current biggest challenges in the world today. Of the 17 SDGs,
the following are in particular relevant for a distillation process:
• Goal 13: Climate Action. This is to take urgent action to combat climate
change. If a distillation process, which generally has a high energy con
sumption, could be more efficient and thus reduce energy requirements, the
process would be more environmentally friendly. Therefore, a more environ
mentally friendly distillation process can help reduce the release of harmful
emissions.
The remaining goals deal with reducing world poverty (goal 15), improving in
frastructure (goal 9 and 11), improving social conditions (goal 8, 10, 16 and 17),
improving the utility consumption of water and electricity with regards to the clean
water and green electricity (goal 6 and 7) and improve the conditions for plant and
animal life in water and on land (goal 14 and 15). These SDGs are not directly
linked to the industrial use of distillation processes but could be indirectly linked.
In order to reduce the high energy requirements and improve the separation ef
ficiency of distillation processes, many different process intensification technolo
gies have been proposed. Examples of distillation process intensification tech
nologies that reduce the energy cost, column size, increase throughput or allow
multiple product streams include: divided wall columns, reactive distillation, heat
integrated columns, cyclic distillation columns, and mechanical vapour recom
pression columns [6]. Reactive distillation is an example of process integration,
combining reaction and separation in a single column, as described above.
1.2 Motivation
Distillation processes are often expensive with regards to capital and operational
expenditures due to low efficiency. In attempts to improve the process economics
and lower the environmental impact, different intensifying technologies have been
proposed that decrease the energy requirement, reduce capital cost, or increase
throughput. One of such technologies is cyclic distillation. It has been shown
that operating a distillation process in cyclic mode can increase the separation
efficiency and thus lower the energy consumption, increase throughput or reduce
the number of stages in a column [10].
The motivation for this PhD project is to further develop the knowledge and fun
damental tools necessary for cyclic distillation to become a viable alternative to
industrial conventional distillation processes and thus reduce energy consump
tion. With the possibility of employing cyclic operated separation on an industrial
scale, the process understanding must also be developed to model, simulate and
design these processes, columns and the control strategies. Furthermore, the
combination of reactive and cyclic distillation could be a promising process inte
gration as the residence time for the liquid holdup on each stage can directly be
specified along with an increase in the separation efficiency, which can help push
the reaction in the desired direction. However, in order to get a more precise
overview of how this can be done and which reactions might be potential can
didates for such a process integration, a rigorous model and model analysis for
reactive cyclic distillation is needed.
1.3 Contribution
This PhD project aims to develop the fundamental knowledge and understanding
of cyclic distillation technology. It is necessary to understand the cyclic distillation
process in order for it to become a viable alternative to conventional distillation for
industrial use.
The current models are mainly intended for a conceptual shortcut design mod
elling and preliminary studies of a cyclic distillation column, most of which do not
describe the energy transfer between liquid and vapour. Hence, there is a lack of
an energy and mass balance based stage model that can be used for high fidelity
rigorous simulation. Developing a dynamic rigorous mass and energy balance
stage model for cyclic distillation makes it possible to investigate the effect of heat
transfer between the liquid and vapour phases in the distillation processes. With
the energy transfer included an expression for the timedependent vapour flow
rate can be set up. A more general stage model will also be able to account for
multiple feed locations, side draws, and energy added to or removed from each
stage, e.g. from mixing, reaction or heat integration.
The research questions that compose and are answered in this PhD thesis are as
follows:
• Can new high fidelity models for cyclic distillation, accounting for mass and
energy transfer between phases, help to advance the technical knowledge
of the process further?
• How does a new model compare to the current state of the art models and
conventional operated distillation?
• What effect does the inherently dynamic vapour flow period have on the
process integration of reactive and cyclic distillation?
• Can a design method for reactive cyclic distillation processes be derived for
future studies?
In order to answer these questions, the following work is presented in this thesis:
High fidelity rigorous models for cyclic distillation and reactive cyclic distillation
are developed. These models are analysed and compared to existing models for
cyclic distillation and conventional distillation operation. Verification of the pro
posed model using published experimental data is conducted. Reactive cyclic
distillation processes are analysed w.r.t. the dynamic stage and column behaviour
over a single vapour flow period. This column performance is compared to con
ventional reactive distillation and other reactive cyclic distillation processes. A
design method or strategy for reactive cyclic distillation based on analysis of pro
cess performance is proposed.
• Chapter 2: Cyclic distillation technology from the 20th to the 21st cen
tury: A brief literature review of the historical development of cyclic sepa
ration technology is given, from the early 20th century to the 2010s. The
current state of the art in cyclic separation regarding the tray and operation
design, the modelling of a cyclic distillation process, the modelling of a reac
tive cyclic distillation process, and the experimental work in cyclic separation
are summarised.
• Chapter 3: Mass and energy balance stage models for cyclic distil
lation: A versatile rigorous dynamic mass and energy balance model is
proposed for modelling and simulation of cyclic distillation processes. This
mass and energy balance model is further expanded to a reactive cyclic
distillation model. This model can either account for an equilibrium or a rate
controlled reactions and present a versatile model for reactive cyclic distil
lation.
• Chapter 4: Cyclic distillation case studies: With the model for conven
tional cyclic distillation presented in Chapter 3, different case studies are
simulated and compared to the previously available models for cyclic distil
lation. These case studies include binary and ternary, ideal and nonideal
mixtures. Comparing the proposed model from Chapter 3 and the existing
model highlights the effect of including the stage energy balances. Further
more, a preliminary model verification using published experimental data is
presented.
is quantified with three performance indicators for selected stages: the ex
tent of reaction over a cycle, the relative distance to equilibrium at the end
of the vapour flow period and the mean Damköhler number over a cycle.
Finally, the feasibility of the three cyclic case studies are evaluated, and
general trends for reactive cyclic distillation processes are identified.
marking reactive distillation for MTBE using classical and cyclic operation,
Computer Aided Process Engineering FORUM 2020 Kgs. Lyngby, Den
mark 2020.
• Case II: Inlet vapour is not mixed, i.e. there is a concentration gradient over
the tray length, and liquid moves across the tray in the same direction on all
trays.
• Case III: Inlet vapour is not mixed, and the liquid moves countercurrently
on the trays.
The first case can describe the bottom stage in a distillation or stripping process,
where for example, pure steam or air is introduced, or the wellmixed vapour from
a reboiler enters a stage. The third case is what is typically seen in conventional
distillation columns. Figure 2.1 shows the vapour and liquid flow in each of the
three Lewis cases.
Lewis showed that the case with the highest tray efficiency is the second case,
followed by the first and then the third case. Thus, according to Lewis, a tray in a
conventional distillation column has the lowest tray efficiency and a tray behaving
as a Lewis Case II has the highest tray efficiency of the three cases he investi
gated. It is, however, challenging and inconvenient to have a distillation column
with internals for Case II, as shown in Figure 2.1.
In the early 1960s Cannon and McWhirter published the first papers describing the
cyclic, also referred to as periodic, distillation technology [10, 16–18]. With this
cyclic distillation technique, the vapour and liquid movements in the distillation
column are separated in a dedicated vapour flow period (VFP) and a liquid flow
period (LFP). These two flow periods constitutes a cycle, where one period is
immediately followed by the other. After a cycle, a VFP followed by an LFP, ends
a new cycle is initiated, and thus a continuous distillation process is achieved.
This cyclic behaviour between the vapour flow and liquid flow periods are shown
in Figure 2.2 with liquid feed added and products removed during LFP.
VFP
LFP
Products Feed
Figure 2.2: Vapour flow period (VFP) and liquid flow period (LFP) in a cyclic
continuous operation. Feed supplied and products removed during LFP.
During the VFP, only vapour is moved through the column, and the liquid holdups
are kept on each stage. This is also where mass and energy transfer between
the phases will occur as the vapour flows up through the trays. After the VFP, the
LFP is initiated. During this, only the liquid holdups are allowed to move inside the
column, with liquid feed and reflux being sent into the column and liquid products
in top and bottom are removed from the column. Cannon and McWhirter et al. [10,
16–18] showed that cyclic distillation could lead to higher tray efficiency compared
to the traditional distillation column due to the separation of phase movements.
Robinson and Engel showed the analogue behaviour in a cyclic distillation tray,
and Lewis Case II [19] and that the distancedependent liquid composition in Lewis
In the following decades from the 1960s to the 1980s many publications were
made focusing on the modelling, control, design and experimental work of cyclic
separation processes [20–35], mainly with simple linear VLE models and for bi
nary stripping processes or assuming reboilers with infinite holdups.
Despite all the work on cyclic distillation in the 20th century, the upscaling to
pilot plant or industrial scale was challenging to implement due to nonuniform
draining of the liquid holdups when using more than 12 ordinary sieve trays in
a pilot scale column as first reported by Schrodt et al. [36]. This nonuniform
draining meant that some stages would drain before others due to the uneven
pressure profile during the LFP, leading to a significant degree of backmixing
and thus reducing the tray efficiency. Different proposals for alternative operation
and tray designs for cyclic distillation were made, such as a stepwise periodic
draining of the stages [37], a pressure manifold to control the draining [38] and an
inclined tray with a closed downcomer [39]. However, none of these alternative
column and tray designs were able to give the necessary incentive to apply cyclic
separation in industry. The number of new publications on the topic decreased
from the 1990s to the 2000s.
In 2011 Maleta et al. [13] presented a new tray design with a sluice chamber
beneath each stage to collect the liquid holdup during the LFP, and then at the
start of the next VFP, this holdup is draining to the stage below. This new Maleta
tray made industrial applications of cyclic separation possible that ensured liquid
draining without backmixing, and thus a new era of cyclic distillation technology
was initiated.
An alternative tray was proposed by Toftegård et al. [14], which allows for continu
ous vapour flow and thus removes the issue with reboiler pressure buildup. This
tray design allows each tray to open and close onebyone so that the draining of
the liquid holdups doing an LFP can be done sequentially. By starting with drain
ing of the bottom stage, the stages are drained and filled with the liquid holdup
from the above stage one at a time, while the vapour flow through the column is
kept.
A comparison of the two operational modes are given in Table 2.1. Figure 2.3
shows a simplified depiction of the two different tray designs and how they behave
during a VFP and an LFP. The Maleta trays have multiple pistons to control the
VFP and LFP, but for simplification only one is depicted in Figure 2.3.
Table 2.1: Differences in the two draining methods for cyclic distillation.
So far most work, both recent and from the previous century, on cyclic distillation
have been done assuming a simultaneous draining of the trays with no vapour
flow present during LFP, as can now be achieved with the Maleta trays.
An alternative column setup that allows continuous vapour flow was proposed
by Krivosheev et al. [40]. This design has vertical partitions inside the column,
dividing all the stages into 3 or 5 sections each. The trays used are the Maleta
trays described above. Continuous vapour flow is provided from the bottom of
the column with a cutoff plate that rotates, and thus the sections of the trays that
undergoes VFP and LFP is continuously alternated.
the stages but without continuous vapour flow. Using this column design, Fa
zlollahi and Wankat also looked into the performance of different flash distillation
processes [42], i.e. batch distillation, conventional flash distillation, continuous
column flash distillation and cyclic flash distillation. They showed that batch or
continuous column flash distillation yielded significant improvements compared
to conventional flash distillation and modest improvement when cyclic flash distil
lation replaces the continuous column flash distillation.
Another cyclic separation process was proposed by Bai et al. [43]. They studied a
cyclic total reflux batch distillation with two separate reflux drums. At the beginning
of the operation, one reflux drum is full with the feed, and the other is empty.
The first period is initiated with the total reflux operation and continues until the
desired composition is reached. During this period, both vapour and liquid flow
rates are present in the column. When the holdup in the reflux drum has reached
the target composition, the reflux is interrupted, and the reflux drum is drained as
the product. The empty reflux drum is connected to the condenser, and a period
of filling the reflux drum is initiated. A new period with total reflux is initiated when
this is full, and by switching between filling the recycle drum, operating under total
reflux and emptying the reflux drum, a cyclic separation process can be achieved.
The most mature of the tray designs and operations is the one proposed by Maleta
et al. [13]. The Maleta trays work with simultaneous draining, which has also been
the idea of operation in most of the early work in the field, before the introduction of
the Maleta trays, where usually one would assume ordinary sieve or screen plate
trays. With the Maleta trays, it is assumed that simultaneous draining during the
LFP is employed. However, it would be interesting to compare the simultaneous
draining achieved with Maleta trays to a sequential draining, e.g. by using the
COPS trays. With a comparison of the two draining methods, the effect of the
continuous vapour flow could be analysed with regard to the separation efficiency.
For the purposes of the modelling and simulation in this thesis, only simultaneous
draining have been considered.
Most of the models describing cyclic separation are mass balance models. Pǎtruţ
et al. [11] presented a rigorous mass balance stage models for a single satu
rated liquid feed cyclic distillation process. This model is subject to assumptions
of vapour liquid equilibrium on all stages, an equal heat of vaporisation, perfect
mixing, negligible vapour holdup and saturated liquid feed. With the assumption
of equal heat of vaporisation, it is assumed there are constant molar holdups on
the stages, and the vapour flow rate is constant. Pǎtruţ et al. also included a term
for a stagedependent vapour flow rate based on the heat of vaporisation on each
stage, but not on the energy transfer between phases. The mass balance model
by Pǎtruţ, with stage 1 as the condenser, is written as:
dM1,i
= V y2,i for i = 1, . . . , N C (2.1)
dt
dMn,i
= V (yn+1,i − yn,i ) for n = 2, . . . , N T − 1 and i = 1, . . . , N C (2.2)
dt
dMN T,i
= −V yN T,i for i = 1, . . . , N C (2.3)
dt
(LF P ) (V F P ) (V F P )
M1,i = M1,i − (D + L)x1,i for i = 1, . . . , N C (2.4)
(LF P ) (V F P )
M2,i = Lx1,i for i = 1, . . . , N C (2.5)
(LF P ) (V F P )
Mn,i = Mn−1,i
(2.6)
for n = [3, . . . , N F ; N F + 2, . . . , N T − 1] and i = 1, . . . , N C
(LF P ) (V F P )
MN F +1,i = MN F,i + F xFi for i = 1, . . . , N C (2.7)
(LF P ) (V F P ) (V F P ) (V F P )
MN T,i = MN T,i − BxN T,i + MN T −1,i for i = 1, . . . , N C (2.8)
With this model it is possible to describe the cyclic distillation process for multi
component nonideal mixtures, however, there are no energy balances included,
so the vapour flow rate must be constant through the column, which means the
total molar holdup on each stage is also constant, or be expressed based on the
heat of vaporisation on each stage. Such an expression for the vapour flow rate
was also proposed by Pǎtruţ et al. [11].
∑ C
Vn+1 N vap
i=1 ∆Hn+1,i yn+1,i
Vn = ∑N C vap
(2.9)
i=1 ∆Hn yn,i
The model proposed by Pǎtruţ et al. [11] only allows a single liquid feed location.
For high fidelity studies the energy balance for each stage should be considered,
as this will affect the stage temperature and vapour flow rate. If different cyclic
distillation processes were to be studied that requires multiple feed locations, e.g.
some reactive distillation cases, the models should be able to also account for
multiple feed locations consisting of liquid and vapour or mixed feeds.
For a cyclic distillation process, the vapour flow rate over a cycle must be greater
than the reflux and lower than the reflux and liquid feed in order for the process
to be feasible and possible to set up. Pǎtruţ et al. [11, 45] have set up a feasi
bility condition for the choice of vapour flow, reflux and duration of VFP in cyclic
distillation process.
L < V · tV F P < L + F (2.10)
Equation (2.10) is only valid for systems with a single liquid feed and constant
vapour flow. Using this condition the vapour flow rate, reflux and duration of the
vapour flow period can be estimated.
Using their proposed model Pǎtruţ, et al. [11] studied a separation of an ethanol/n
propanol mixture in a cyclic distillation process, with a VFP time of tV F P = 25 s
and an LFP time of tLF P = 5 s, and compared it to a conventional distillation.
They found that the number of trays could almost be halved when going from
conventional to cyclic operation and that a larger range of product purities can
be obtained. Using a simple PIcontroller for the internal tray temperatures by
manipulating the distillate flow rate and vapour flow rate, they concluded that the
ethanol/npropanol mixture could easily be controlled regarding the product puri
ties.
Pǎtruţ et al. [11] also proposed a design algorithm, using the mass balance stage
model they proposed. This design algorithm is based on the previously devel
oped method by Toftegård et al. [31]. With this design method for cyclic distillation
columns, the bottom product purity is first specified. The VFP molar balance for
the reboiler is then integrated backwards in time, with the bottom product purity
as the initial condition, through the VFP and the reboiler composition at the begin
ning of VFP can be found. Using this and the knowledge of the bottom product
flow rate, the composition of the stage above the reboiler can be found from the
algebraic LFP mass balance. The reboiler and the stage above, N T − 1, can then
be integrated backwards in time through the VFP, and the composition in stage
N T − 1 at the beginning of VFP/end of LFP is found and used to solve the LFP
mass balance for the composition in stage N T − 2 after VFP. An entire distillation
VNT, yNT
Figure 2.4: First two steps of designing a cyclic distillation column using the
backwards integration method proposed by Pǎtruţ et al. [11] based on the
method by Toftegård et al. [31].
The first attempt to set up a mass and energy balance model was presented by
Andersen et al. [46]. They showed a mass and energy balance stage model
for cyclic distillation with a nonconstant vapour flow rate and the possibility for
vapour feed. However, they only included a single feed location for a liquid/vapour
feed. The purpose of their model was to describe the integrated column design
based on the driving force based design algorithm method by Nielsen et al. [47].
While considerations of the energy transfer on each stage are included in the
model, it does not give a general description of the process, and while the time
dependent vapour flow rate is mentioned, it is not explicitly presented. Using
this model, they showed a driving forcebased cyclic distillation column design
for a binary mixture of water and ethanol. The maximum driving force design
showed an improvement in response to feed flow rate disturbances, with a simple
PI controller for the bottom product purity with the reboiler duty as the manipulated
variable, over a design not based on the maximum driving force. Their mass
balances for the VFP and the LFP are similar to the model by Pǎtruţ et al. shown
in equations (2.1)(2.8). However, the VFP molar balances have a variable vapour
flow rate, and there is a vapour feed term for the stage above the feed stage.
Assumptions for their model included: vapour liquid equilibrium, perfect mixing,
negligible heat exchange with surroundings and a total condenser as Pǎtruţ et al.
[11] also assumed, and negligible pressure drop. The LFP mass equations are
the same as in equations (2.4)(2.8).
dM1,i
= V2 y2,i for i = 1, . . . , N C (2.11)
dt
dMn,i
= Vn+1 yn+1,i − Vn yn,i
dt (2.12)
for n = [2, . . . , N F − 2; N F, . . . , N T − 1] and i = 1, . . . , N C
dMN F −1,i
= VN F yN F,i − VN F −1 yN F −1,i + F V yN
F
F,i for i = 1, . . . , N C (2.13)
dt
dMN T,i
= −VN T yN T,i for i = 1, . . . , N C (2.14)
dt
dh1
= H2 − QC (2.15)
dt
dhn
= Hn+1 − Hn for n = [2, . . . , N F − 2; N F, . . . , N T − 1] (2.16)
dt
dhN F −1
= HN F − HN F −1 + H F (2.17)
dt
dhN T
= QB − HN T (2.18)
dt
While this model includes both mass and energy balances, a more general stage
model can be made for cyclic distillation that can account for multiple vapour or
liquid feeds, energy added to stage and side draws, such that any case can be
studied in a cyclic distillation process.
Kiss et al. [49, 50] suggests reactive cyclic, or catalytic cyclic in the case of a het
erogeneous catalyst, as a process intensification technology to lower the energy
usage and increase the separation efficiency compared to conventional reactive
distillation. With the adjustable duration of VFP, which is a design parameter in
a cyclic distillation process, the residence time for the liquid holdup on a stage
can be specified. This means that for the reactive stages, it is possible to di
rectly specify the residence time, which could be a benefit for slow reactions. In a
conventional reactive distillation process, the residence time on each stage is de
termined by the column and tray dimensions and the internal liquid flow rate. With
a reactive cyclic distillation process, it is possible to directly change the residence
time independent of column dimensions and liquid holdup.
sibility of multiple periodic states. Table 2.2 shows the current literature available
on reactive (catalytic) cyclic distillation.
The example of DME by Pǎtruţ et al. with a single feed location so far is the only
case investigated for reactive cyclic distillation. They presented a mass balance
based model for reactive cyclic distillation processes, similar to the one they pre
sented for nonreactive distillation [11]. The model for reactive cyclic distillation
they presented was under the assumption of vapourliquid equilibrium, negligible
vapour holdup, perfectly mixed liquid holdup, saturated liquid feed and instanta
neous liquid holdup displacement. The model equations are shown below, with a
simple expression for the vapour flow rate based on the heat of vapourisation of
the components and the heat of reaction:
The mole balances for LFP are the same as equations (2.4)(2.8).
when the reaction rate of the reaction and relative volatility of the separation mix
tures both allow a combined reactive separation [49]. For example, reactive dis
tillation is not feasible if the reaction rate is very slow and there is no significant
difference in relative volatility of the products and reactants. If the opposite is true,
i.e. fast reaction rate and significant difference in volatility, then reactive distilla
tion is feasible. If the reaction rate is very fast, then there might not be a clear
advantage in choosing reactive distillation over a reactor and a separation pro
cess, but this depends on the case, and a feasibility analysis should be made for
a given case before choosing the equipment. These feasibility analysis methods
can, for example, be based on the dimensionless Damköhler number and the re
action equilibrium constant as done by Shah et al. [55], who also presented a
framework for the technical evaluation of reactive distillation. In addition to eval
uating the Damköhler number and the equilibrium constant, they also propose to
evaluate the dimensionless Hatta number and the forward reaction rate constant
versus the mass transfer coefficient for the catalyst. This makes it possible to
both determine the feasibility of the reactive distillation and evaluate the control
ling mechanism. Muthia et al. [56] proposed first to evaluate the relative volatility
of the components, find the equilibrium constant and the Damköhler number and
then find the number of theoretical stages. The applicability range of the chosen
reactive distillation case can be found by performing a sensitivity analysis with
regard to the stage configuration and reflux ratio. In another study, Muthia et al.
[57] proposed a systematic method for assessing the applicability of quaternary
reactive distillation. This method is also based on evaluating the relative volatility
for the components, finding the equilibrium constant and Damköhler number, and
iteratively solving for a reactive distillation column design that allows the chosen
process to be applicable. Both the Damköhler number and equilibrium constant
are important parameters when assessing whether a reactive distillation process
is feasible or not. The values of the two parameters individually and together can
then help characterise the reactive distillation process, and the feasibility and the
controlling mechanism.
In order to develop a control strategy for the cyclic distillation process, the manip
ulated and controlled variables must be identified [58]. When designing a control
structure for a distillation column, there are many essential variables and many
options for choosing manipulated and controlled variables pairings. With cyclic
distillation, two variables, which conventional distillation does not have, is the time
for the LFP and the VFP period. Especially tV F P is important since a long time for
VFP increases the overall time for mass transfer but can decrease the through
put. The time for LFP should be set to ensure complete draining before the next
VFP is initiated to avoid backmixing. For cyclic distillation, it is vital to consider
the duration of VFP as a manipulated variable. Other possible variables that can
be used for control can also be found in conventional distillation, such as reboiler
duty, feed, product or reflux flow rates and stage temperatures.
in industrial alcohol production. With this cyclic stripper included in the produc
tion plant, the total waste from the plant can be reduced from 36 % of the plant
capacity to 11.5 %.
With the COPS tray design, Toftegård et al. also published experimental results
for an ammonia stripping process [14]. In this study, the COPS tray design was
proposed, and the experiments reported were focused on testing performance
in terms of weeping limits and tray and column efficiencies. Using the COPS
trays and cyclic operation instead of ordinary sieve trays in conventional opera
tion showed that high tray efficiencies could be achieved as well as high capaci
ties. They further list some advantages the COPS tray has over the Maleta tray
for cyclic operated separation processes, i.e. long and wellcontrolled LFP and
steady reboiler operation with no pressure buildup due to interruption of vapour
flow.
Nielsen et al. [61] further studied ammonia stripping using COPS trays. They
compared the tray efficiencies for ordinary sieve trays and COPS trays and found
the point efficiency for COPS trays under different values for the liquid holdup and
vapour flow rate. The tray efficiency for the COPS trays was almost double that
of the ordinary sieve trays.
So far, no extensive model verification studies using experimental data and the
current state of the art models, mentioned in Section 2.2.2, have been published.
The proposed mass and energy balance stage model for cyclic distillation pro
cesses is based on the wellknown MESH equation system for conventional dis
tillation [62]. It is assumed that simultaneously drained trays are used to develop
this model, as it is the most studied draining method and thus the most mature
technology. Figure 3.1 shows the in and outputs for a stage n in a cyclic distilla
tion column. With this overview of the variables, it is possible to set up both mass
and energy balances for each stage n. The terms mass and molar balances are
used interchangeably throughout this thesis. However, unless otherwise stated,
every equation is on a molar basis.
VFP LFP L
Vn,yn,i,TnV Mn-1,xn-1,i,Tn-1
VnF,yFn,i ,TnVF
MFn,xFn,i,TnLF
Qn Sn,xn,i,TnL
Stage n Stage n
V
Vn+1,yn+1,i,Tn+1 Mn,xn,i,TnL
Figure 3.1: In and outputs to stage n in a cyclic distillation process, during
VFP (left) and during LFP (right).
3.1.1 Assumptions
For the development of the rigorous dynamic model, the following assumptions
are made:
• Vapour liquid equilibrium on any tray, between the wellmixed liquid and the
vapour exiting a tray.
• Liquid feed and reflux is added during LFP, vapour feed is added continu
ously during VFP.
• Mass and heat transfer between vapour and liquid phases occur only during
VFP.
dMn,i (t)
= Vn+1 (t)yn+1,i (t) − Vn (t)yn,i (t) + Vn+1
F F
(t)yn+1,i (t) (3.1)
dt
For the VFP balances the initial conditions of the holdups and enthalpies are
0 (LF P ) (LF P )
Mn,i (t = 0) = Mn,i = Mn,i , Mn (t = 0) = Mn0 = Mn and hn (t = 0) =
0 (LF P )
hn = hn respectively.
(LF P ) (V F P ) (V F P )
Mn,i F
= Mn−1,i + Mn−1,i − Sn xn,i (3.3)
(V F P ) (V F P )
Mn(LF P ) hn(LF P ) = Mn−1 hn−1 F
+ Mn−1 hFn−1 − Sn hSD
n (3.4)
Here the superscripts (LF P ) and (V F P ) denotes the end of the LFP and the
VFP respectively. For stage n = 2 the holdup from the stage above, which is the
(V F P ) (V F P ) (V F P )
condenser, is equal to the reflux, M1 = L and M1,i = Lx1,i .
The mole and energy balance model for the reboiler and condenser, stage n = N T
and n = 1, can be set up similar to the above equations (3.1)(3.4), with the
following specifications: V1 = VN T +1 = 0, Q1 = QC , and QN T = QB . It is assumed
that there is no feed to the condenser or the reboiler. The mole and energy balance
for the reboiler and condenser during the VFP and the LFP are written below. It
is assumed that the condenser is a total condenser, and the reboiler is a partial
reboiler.
dM1,i (t)
= V2 (t)y2,i (t) (3.5)
dt
dMN T,i (t)
= −VN T (t)yN T,i (t) (3.6)
dt
During the LFP, the top and bottom liquid products are removed from the con
denser and reboiler, respectively, and the reflux is removed from the condenser
and fed to the stage below; n = 2.
(LF P ) (V F P ) (V F P )
M1,i = M1,i − (D + L)x1,i (3.9)
(LF P ) (V F P ) (V F P ) (V F P )
MN T,i = MN T,i + MN T −1,i − BxN T,i (3.10)
The molar holdup of a component i is the total molar holdup multiplied by the
liquid molar composition. The total molar holdup is the summation of the individual
components’ molar holdup:
Using equation (3.14) the total molar balance for the VFP and the LFP can be
found. The total mass balance for the VFP is:
dMn ∑ dMn,i
NC
= = Vn+1 (t) − Vn (t) + Vn+1
F
(t) (3.15)
dt i=1
dt
∑
NC
(LF P ) (V F P )
Mn(LF P ) = Mn,i = Mn−1 F
+ Mn−1 − Sn (3.16)
i=1
In order to ensure the model variables in the VFP and LFP balances can be found
and the degrees of freedom are satisfied, summation and equilibrium equations
must be included in the model. The summation equations for the liquid and vapour
compositions ensure that the total mole percentage does not exceed 100 mol%.
The vapourliquid phase equilibrium equation ensures the two phases are in equi
librium. The summation and equilibrium equations are also part of the MESH
equation system [62]. The summation equations for n = 1, . . . , N T are:
∑
NC
yn,i − 1 = 0 (3.17)
i=1
∑
N C
xn,i − 1 = 0 (3.18)
i=1
The vapour liquid equilibrium (VLE) equation for stage n = 1, . . . , N T and compo
nent i = 1, . . . , N C is:
yn,i − Kn,i xn,i = 0 (3.19)
sat
yn,i Pn = xn,i Pn,i γn,i (3.20)
sat
Where γn,i is the activity coefficient for component i on stage n, and Pn,i is the
saturation or vapour pressure for component i. The vapour pressure can be deter
mined from a vapour pressure equation, e.g. the Antoine or the extended Antoine
sat
equation. Equation (3.20) can be written as equation (3.19) with Kn,i = Pn,i γn,i/Pn .
Enthalpies
The liquid and vapour enthalpies, hn and Hn , respectively, must also be defined
to satisfy the energy balances. For simplification, it is assumed that the specific
V
heat capacities, CP,i , are independent of temperature. The vapour enthalpy for
The mass and energy balance presented here can be implemented with a more
V
general and temperature dependent expression for CP,i , however, for the case
V
studies that have been investigated in this thesis it is assumed Cp,i is temperature
independent. A reference temperature of Tref = 298 K and reference enthalpy
of Hi◦ = 0 are arbitrarily chosen. The reference enthalpy value can be chosen
arbitrarily as it is cancelled out in the energy balances. The summation of the
component vapour enthalpy gives the total vapour enthalpy of stage n.
∑
NC
( V )
Hn (t) = yn,i (t) CP,i (Tn (t) − Tref ) (3.22)
i=1
With the assumption of negligible heat of mixing the liquid enthalpy for stage n
can be found similar as the vapour enthalpy, whilst accounting for the heat of
vaporisation.
∑
NC
( V )
hn (t) = xn,i (t) CP,i (Tn (t) − Tref ) − ∆Hvap,i (3.23)
i=1
Tray hydraulics
With the energy balance in equation (3.2) a vapour flow rate expression for stage
n can be set up. Isolating Vn in equation (3.2) and inserting the enthalpy terms,
equations (3.22) and (3.23), and the molar balance, equation (3.1) gives the vapour
flow rate for stage n:
1 [ F
Vn (t) = Vn+1 (t)g(n, n + 1, t) + Vn+1 g F (n, n + 1, t)
g(n, n, t)
] (3.24)
∑NC
dTn (t)
−Mn (t) V
xn,i (t)CP,i + Qn
i=1
dt
The detailed derivation of this can be found in Appendix A. The vapour flow ex
pression consists of terms that account for the vapour flow from the stage below
and the vapour feed that enters the stage as well as the liquid enthalpy and the ad
ditional energy term Qn . The vapour flow rate is timedependent, as it will change
over a cycle when light components are removed from the liquid phase and heavy
components transfer from the vapour phase. With a change in vapour and liq
uid composition it follows that the total number of moles in the liquid and vapour
phases also change. The expression g(na , nb , t), in J/mol, describes the difference
in the vapour enthalpy Hn (t) and the liquid enthalpy hn (t). This is defined as:
Here, na and nb can be equal, and thus describe the same stage or nb = na + 1
for which g(na , nb , t) describes the enthalpy difference between the liquid holdup
on stage na and the vapour enthalpy of the vapour from stage nb .
The vapour flow rate from the reboiler can be written in a similar way. For the
reboiler, there is neither a vapour feed nor a vapour flow coming from a stage
below. ∑ C V dTN T (t)
QB (t) − MN T (t) Ni=1 xN T,i (t)CP,i dt
VN T (t) = (3.26)
g(N T, N T, t)
The total condenser has a vapour flow rate of zero, V1 (t) = 0.
The temperature derivative regarding the time can be found from the vapour liquid
equilibrium, equation (3.20). For any mixture the summation of the vapour com
∑ C
positions must give unity N i=1 yn,i = 1, and with the VLE from equation (3.20), a
function f can be defined, which must be true for any mixture.
∑
NC sat
Pn,i
f =1− xn,i γn,i =0 (3.27)
i=1
Pn
As the vapour pressure, molar composition, and activity coefficient are depen
dent on the temperature and molar holdup, so is the function f , thus taking the
derivative of the function f with regards to time gives:
From this equation the time dependent stage temperature derivative can be iso
lated. The molar holdup derivative, dMn,i (t)/dt, can be found from the balance equa
tion (3.1) and the partial derivatives, ∂f (t)/∂t and ∂f (t)/∂Mn,i (t), can be found from
the vapour liquid equilibrium in equation (3.20) with the knowledge of the activity
coefficient model.
The liquid flow rate from stage n − 1 to stage n is the total molar holdup that is
displaced during the LFP for each stage. This is also shown in the molar balance
for stage n in equation (3.3).
Duties
The reboiler and condenser duties are defined below. The condenser is assumed
to be a total condenser, so all vapour that enters is condensed. The reboiler
is assumed to be a partial reboiler, so only a fraction of the reboiler holdup is
vaporised. The fraction of the reboiler holdup that is vaporised is determined by
the boilup ratio, BR, which is a measure of the average vapour flow from the
reboiler during a VFP.
The initial total molar holdup in the reboiler, MN0 T , is used to relate the amount of
vapour sent from the reboiler to the boilup ratio and duration of the VFP. If BR = 1,
all of the reboiler holdup at time t = 0 will be vaporised over the duration of the
VFP.
For an estimate of the vapour flow rate from the reboiler, the definition of the
reboiler duty in equation (3.30) can be simplified:
MN0 T BR
VN T = (3.31)
tV F P
Periodicity condition
In order to ensure that a quasistationary steady state, i.e. repeating cycles, can
be reached, two periodicity conditions are defined. These conditions ensure that
the holdup in the reboiler and condenser are identical after each LFP. This way,
the reboiler and condenser will not run dry.
(V F P )
L = M1 − M10 − D (3.32)
(V F P (V F P )
B = MN T ) + MN T −1 − MN0 T (3.33)
By implementing these two periodicity conditions it is assumed the reflux and bot
tom flow rates perfectly control the condenser and reboiler holdups respectively.
If a reflux ratio, RR, instead of a reflux and distillate flow rate, L and D, is known
the periodicity condition for the condenser, equation (3.32), can be written as:
( )
(V F P )
M1 − M10 RR
L= (3.34)
1 + RR
(V F P )
M − M10
D= 1 (3.35)
1 + RR
Where the reflux ratio is a real positive number. The greater the value of the reflux
ratio, the higher is the amount of reflux that is sent back to the column compared
to the distillate.
A flash calculation with known column or stage pressure and feed enthalpy can
be made if the liquid feed is above the mixtures’ bubble point. This is done to find
the fraction of liquid that is vaporised and the respective temperatures and com
position of the liquid and vapour phases. It is assumed that liquid and vapour feed
is introduced separately during the LFP and VFP, respectively. Should a flashing
of the feed occur, it is thus assumed that the liquid and vapour fractions can be
separated before it is fed to the column. This assumption allows for modelling a
flashed feed in a cyclic distillation column. However, it is only reasonable for small
amounts of flashed vapour. Generally, the flashing of a feed in a cyclic distillation
process is undesirable due to the separated phase movement operation.
A cold liquid feed can also be fed to the column. In this case, the feed will enter
and mix with the holdup on a stage, and the stage temperature will likely drop
below the bubble point. It is assumed that no mass transfer between the liquid
and vapour phase will occur until the stage temperature is at the bubble point
again. The heating of the liquid will be facilitated by the vapour flow from the
stage(s) below. Alternatively, external heating can be applied to the stage, which
can be accounted for in the model with the variable for energy, Qn . This case is
not desirable, and if a cold feed is to be introduced to a cyclic distillation column,
a preheater should be considered.
No No
4. Cyclic 5. Rigorous
1. Problem 2. Choose 3. Column Results 6. Output
distillation dynamic Yes
definition VLE model design satisfactory? analysis
specifications modelling
Pure
component &
mixture
parameters
Figure 3.2: Model framework for mass and energy balance stage model for
cyclic distillation.
In step 12, the problem is defined, and the relevant properties for the given com
ponents and mixture are found in the literature or a database. The model com
plexity is also defined as part of the problem definition (see Table 3.3). In step
3, the column is designed, and the number of trays, initial conditions, feed loca
tion(s) and side draw(s) are specified. The cyclic distillation specific parameters,
time periods for the VFP and LFP, as well as the number of cycles simulated, are
chosen in step 4, and the rigorous dynamic model is solved. If the results are
not satisfactory, the model parameters must be reevaluated. First, it is evaluated
whether the number of cycles or VFP and LFP times is too high or low. If this is not
the case, then the column design must be verified, and finally, the choice of VLE
model and pure component/binary interaction parameters can be reconsidered.
In the implementation of the models, the VFP and LFP are conveniently solved
separately. One way to do this is to solve the VFP differential equations first, with
suitable initial values. After this, the algebraic equations for the LFP can be solved
since the states after the LFP depend on the states after the VFP. With the values
after the LFP as initial values, a new VFP can be solved and so forth, any number
of cycles can be simulated.
A model analysis, as described by Cameron and Gani [64], is carried out for a
singlestage both in VFP and LFP. This model analysis is necessary to perform
in order to ensure that the developed model is possible to solve. The separation
factor, Kn,i , is in this analysis treated as a parameter since the explicit temperature
dependency is only known when the appropriate VLE model has been selected.
V FP
The developed model for the vapour flow period contains Neqs = 3 + 2N C equa
V FP
tions for a single stage, (3.1)(3.2), (3.17)(3.19). There are a total of Nvar = 10+
4N C variables during VFP, which gives the following number of degrees of free
V FP
dom: NDoF = 7 +2N C. In addition to the variables shown in Figure 3.1, the stage
pressure, total holdup, individual component molar holdups and duration of VFP
are also variables. The feed and inlet variables and the duration of VFP, and the
stage pressure are specified. This leaves the following variables to be calculated
at each time step for stage n and component i: Mn (t), Mn,i (t), yn,i (t), Vn (t), Tn (t).
The liquid composition of the individual components can be found from the com
ponent holdups. The incidence matrix for the VFP is shown in Table 3.1.
LF P
A degree of freedom analysis can also be done for the LFP. There are Neqs =
LF P
3 + 2N C equations, (3.3)(3.4), (3.17)(3.19), for the LFP and Nvars = 10 + 4N C
LF P
variables, which gives the following number of degrees of freedom: NDoF =
7 + 2N C. The feed and inlet variables, as well as the duration of LFP and the
stage pressure are specified, leaving the following variables to be calculated:
(LF P ) (LF P ) (LF P ) (LF P ) (LF P )
Mn , Mn,i , yn,i , Tn , Ln . Since the entire stage is completely drained
(LF P ) (LF P )
during the LFP it follows that Mn = Ln tLF P . The incidence matrix for the
LFP is shown in Table 3.2.
Table 3.2: Incidence matrix for LFP.
The total molar holdup balances shown in equation (3.15) for the VFP and equa
tion (3.16) for the LFP can substitute either equation (3.17) or (3.18).
For a cyclic distillation column with N T stages and N C components, there are
N T (3 + 2N C) equations to be solved for both the VFP and the LFP, with N T (10 +
4N C) variables each. This leaves a total of 2N T (7 + 2N C) variables that need to
The two incidence matrices, Table 3.1 and 3.2, do not have lower tridiagonal
forms, which means the model equations can not be solved sequentially [64]. The
model must therefore be solved simultaneously. The differential equations and the
algebraic constraints for the VFP constitutes a DAE system of equations. This
DAE system is not specific to cyclic distillation but is generally seen in dynamic
distillation processes as well as other complex chemical processes, where the
ODEs describe the mass and energy balances and algebraic constraints describe
the VLE. It is possible to simplify this DAE system to an ODE system by solving
the algebraic constraints at each time step from the current states [63, 65–67].
The solving of the ODE system is as follows: Initially for time t = 0, the values of
Tn (0), Mn,i (0), Mn (0) and xn,i (0) are known from the initial conditions with Mn (0)
and xn,i (0) found from equations (3.13) and (3.14). The vapour flow rate and
vapour compositions can then be calculated for t = 0 using equations (3.19) and
(3.24). Then the ODEs, equations (3.1)(3.2), are calculated and for the next
time step the algebraic equations can be solved again. With this simplification, by
separating the ODEs from the algebraic constraints, a certain level of uncertainty
is allowed in the solving for each time step. However, as long as the time steps
are reasonably small, this should not give any significant deviations compared to
solving the DAE system. The ODEs and algebraic equations and balances for
the VFP and LFP are solved using MATLAB® , with, for example, the ode15s or
ode45 solvers for the numerical solving of the VFP, dependent on the stiffness of
the case.
By including the energy balances new possibilities for process control strategies
emerge. For example, the boilup and reflux ratios can be used as manipulated
variables and the stage temperatures as controlled variables.
The derived model for reactive cyclic distillation is based on the mass and energy
balance stage model shown in Section 3.1, with the same assumptions, expanded
to include reactions [62].
• Reaction only takes place in the liquid phase and the reactive zone of the
column.
It is further assumed that the reaction only occurs during the VFP. This is a reason
able assumption if the reaction is catalysed by a heterogeneous catalyst placed
on the stages. The dynamic stage model includes an energy balance to describe
the timedependent vapour flow on each stage and further allows multiple feed
locations and incorporates the heat of reaction. These features are essential in
many reactive distillation processes, where reactants commonly are introduced at
two or more locations in the column. Figure 3.3 shows a stage n in the reactive
cyclic distillation process during the VFP and LFP, similar to the stage shown in
Figure 3.1. As shown the reaction only occurs during the VFP in the liquid phase.
VFP LFP L
Vn,yn,i,TnV Mn-1,xn-1,i,Tn-1
VnF,yFn,i ,TnVF
MFn,xFn,i,TnLF
Qn Sn,xn,i,TnL
Stage n Stage n
Reaction
V
Vn+1,yn+1,i,Tn+1 Mn,xn,i,TnL
Figure 3.3: Stage n in reactive cyclic distillation during the VFP (left) and LFP
(right).
The mass and energy balance for tray n = 1, ..., N T and component i = 1, ..., N C
is shown below for the VFP and the LFP. The presented model is a general re
active cyclic distillation model, with rate kinetics or equilibrium reaction and pos
sibility for side draws, Sn . The reactive cyclic distillation model is a vapour liquid
equilibrium distillation model, similar to the nonreactive MESH equation model by
Wang and Henke [62] and the reactive MESH equation model presented by Taylor
and Krishna [68], but for cyclic operation. The stages are numbered from top to
bottom, with the condenser being stage n = 1 and the reboiler stage n = N T .
dMn,i (t)
= Vn+1 (t)yn+1,i (t) − Vn (t)yn,i (t) + Vn+1
F F
(t)yn+1,i rt
(t) + Rn,i (t) (3.36)
dt
0 0
With the initial conditions Mn,i (0) = Mn,i and yn,i (0) = yn,i for n = 1, ..., N T and
rt
i = 1, ..., N C. The reaction rate for stage n and component i is denoted Rn,i .
The superscript ”rt” denotes the reaction type, which can either be a reaction rate
kin
expression determined by reaction kinetics, where Rn,i = νn,i,r rn,r mcat
n with rt =
eq
kin, or by a chemical equilibrium term, where Rn,i = νn,i,r ϵn with rt = eq.
With the initial conditions Tn (0) = Tn0 , Mn (0) = Mn0 and yn,i (0) = yn,i0
for n =
1, ..., N T and i = 1, ..., N C. This energy balance is used to describe the time
dependent vapour flow rate as also described in Section 3.1. With the new mole
and energy balance equations for the reactive cyclic distillation, a new expression
for the reaction heat must be written for the vapour flow rate. By defining a new
helper function g reac (n, t) that accounts for the reaction heat the vapour flow rate
for reactive cyclic distillation can also be written:
1 ( F
Vn (t) = Vn+1 (t)g(n, n + 1, t) + Vn+1 g F (n, n + 1, t)
g(n, n, t)
) (3.38)
∑
NC
dTn (t)
+g reac (n, t) − V
xn,i (t)CP,i + ∆Hnreac + Qn (t)
i=1
dt
∑
NC
g reac (n, t) = −hn (t) rt
Rn,i (t)
i=1
[ ( )
Tn (t) ∑
N C ∑
NC
+ Mn (t) − 2 V
CP,i rt
Mn (t)Rn,i (t) − Mn,i (t) rt
Rn,i (t) (3.39)
Mn (t) i=1 i=1
( )]
∑
NC V
CP,i Tref
+ ∆Hvap,i ∑
N C
+ rt
Mn (t)Rn,i (t) − Mn,i (t) rt
Rn,i (t)
i=1
Mn2 (t) i=1
The derivation of the vapour flow rate for a nonreactive cyclic distillation is shown
in Appendix A. An equivalent derivation that can be made for a reactive cyclic
distillation to find a vapour flow rate expression is also described in the appendix.
The reaction takes place in the liquid phase, so this does not directly affect the
vapour flow rate. The heat of reaction, ∆Hnreac , is included in the energy balance
since it will affect the liquid temperature as the reaction takes place. It is included
here to ensure a high fidelity model, although this contribution is usually small for
(LF P ) (V F P ) (V F P )
Mn,i F
= Mn−1,i + Mn−1,i − Sn xn,i (3.40)
(V F P ) (V F P )
Mn(LF P ) hn(LF P ) = Mn−1 hn−1 F
+ Mn−1 hFn−1 − Sn hSD
n (3.41)
The reaction terms are zero for nonreactive stages, i.e. the reboiler, condenser,
rectifying and stripping stages. It is further assumed energy is only added to the
reboiler and removed from the condenser, Qn = 0 for n = 2...N T − 1.
Using the proposed model for cyclic distillation as a basis, a model for reactive
cyclic distillation was also derived. This reactive cyclic distillation model is, as the
cyclic distillation model, similar to the already established MESH equation system
for reactive distillation [68]. The reactive cyclic distillation model accounts for the
reaction enthalpy on all reactive stages and the development in vapour flow rate
over the vapour flow period. As a generalisation, the reaction rate can be either for
a fast reaction, using a chemical equilibrium assumption, or based on a reaction
rate kinetic expression.
Examples from Pǎtruţ et al. [11] have been used for comparing the two mod
els. The chosen case studies from Pǎtruţ et al. are: Ideal mixtures of ben
zene/toluene and benzene/toluene/oxylene, and a nonideal mixture of ethanol/n
pentanol. Furthermore, a nonideal mixture of ethanol/methanol/water has also
been analysed. The nonideal mixtures are described by the Wilson activity coef
ficient model, with the binary interaction parameters found in Aspen Plus. All the
simulation parameters for the case studies are shown in Table 4.1.
The cyclic distillation process has not been simulated with existing tools in com
The chosen case studies, summarised in Table 4.1, are analysed in terms of dif
ferences between the mass balance model and the mass and energy balance
model. First the differences in the cyclic behaviour are investigated for the case
of benzene/toluene. Figure 4.1 shows 10 cycles for the benzene/toluene mixture,
when using the mass balance model (blue) and the proposed mass and energy
balance model (red).
100 100
95
99.5
90
99 85
80
98.5
75
98 70
0 5 10 0 5 10
60 2
1.5
55
50
0.5
45 0
0 5 10 0 5 10
In Figure 4.1 the quasistationary steadystate can be seen as the identical re
peating cycles. There is a clear difference in the mass and the mass and energy
balance models. For example, the mass and energy balance model predict higher
purity top and bottom products. There is, however, no trend in the magnitude of
change in the composition over the VFP over time, where the simple mass balance
model predicts a greater change in the composition in stage 4, a lower change is
predicted in stage 7.
The main reason for this difference between the two models is the vapour flow
rate. For the mass balance model, it is assumed that the stage vapour flow rates
are constant over the VFP, whereas the flow rates are timedependent for the
mass and energy balance model. The vapour flow rate depends on the stage
temperature and the liquid and vapour phase compositions. When the vapour
flows up through a column and interacts with the liquid holdups, the two phases
reaches an equilibrium. The timedependent vapour flow rate for the model, see
equation (3.24), changes with the time derivative of the stage temperature, which
in turn depends on the change in composition and phase equilibrium. It follows
that a timedependent vapour flow rate will be more important for long vapour flow
periods since for a short VFP, the change in Vn (t) will be small. Furthermore, a
mixture with high differences in volatility will experience a higher degree of mass
transfer between the vapour and liquid phases. Significant temperature changes
over a cycle would also result in large changes in the vapour flow rate. Significant
changes in the vapour flow rate will, in turn, also mean a high degree of change in
composition and temperature over time. Therefore a timedependent vapour flow
rate is necessary for high fidelity studies.
The differences in the mass balance and the mass and energy balance models
are further investigated in the following comparisons between the quasistationary
steadystates for the simple mass balance model and the mass and energy bal
ance model for the four cases in Table 4.1. These comparisons of the molar com
position in each stage after a VFP with the cycles in quasistationary steadystate.
are shown in Figure 4.2.
0.8
0.6
0.4
0.2
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0.8
0.6
0.4
0.2
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
0.8
0.6
0.4
0.2
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14
0.8
0.6
0.4
0.2
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Figure 4.2: Stage compositions after VFP for the cases in Table 4.1 for the
mass balance model (blue) and the mass and energy balance model (red) in
quasistationary steadystate.
There is a clear difference between using a simple mass balance model and an
extended model with both mass and energy balances, as is evident from the quasi
stationary steadystate plots in Figure 4.2. For the two binary cases, the mass and
energy balance models show a higher separation degree than the mass balance
model. This is due to the timedependent vapour flow rate as mentioned above.
In both cases, it seems the columns are overdimensioned when using the mass
and energy balance model. It also means that by using the mass and energy
balance model with a timedependent vapour flow rate, fewer trays are needed
for the same process compared to the prediction from the mass balance model
for this case.
For the case with benzene/toluene/oxylene there is also a clear difference be
tween the two models. While the top and bottom compositions are similar, the
column profile is not. The mass balance model predicted the midboiling com
ponent, toluene, to have the highest composition at stage 13. In contrast, the
mass and energy balance model predict the highest composition at stage 12. This
means if the midboiling component is desired to be removed as a sideproduct,
the two models will give two different optimal stages.
The last case, ethanol/methanol/water, does not seem to differ between the two
models significantly. The reason for this is the low vapour flow rate. The mass
balance was set constant to 5 mol/s, and for the mass and energy balance, it
was between 5.53 mol/s and 5.73 mol/s, on average for all stages around 5.57
mol/s. Since the vapour flow rates are very similar for the two models, the dif
ference between the results is minor. Hence, for this case with minor changes
in the vapour flow rate, as mentioned previously, the mass balance model can
be used as a reasonable approximation. For comparison, the vapour flow rate
for the benzene/toluene case ranged from 37.49 mol/s to 45.17 mol/s over the
stages, with an average of 41.20 mol/s. For the mass balance only model, it was
a constant value of V = 36.83 mol/s. Here the timedependent vapour flow rate
has a more significant effect, and the mass and energy balance should therefore
be used. This can also be seen as the difference between the compositions for
the two models in Figure 4.2.
ure 4.3 for the simple mass balance model and the mass and energy balance
model. The relative changes from the initial quasistationary steadystate to the
new ethanol composition are summarised in Table 4.2.
Figure 4.3: EtOH compositions in condenser (top) and reboiler (bottom) after
VFP in quasistationary steadystate (black) and for +5% (red) and 5% (blue)
changes in the feed EtOH composition, using the simple mass balance model
(left) and the extended mass and energy balance model (right).
steadystate profile as shown in the bottom plot in Figure 4.2. This was due to
the almost constant vapour flow rate for the mass and energy balance model. For
the ±5 % disturbances in the ethanol feed composition, with the mass and energy
balance model, the vapour flow rate was almost constant with a flow rate around
5.32−5.73 mol/s. Since the vapour flow rate was almost constant for the mass and
energy balance model, it is as expected that the two models would show similar
dynamic responses as shown in Figure 4.3.
For the benzene/toluene case, changes in the feed temperature have been in
vestigated with the presented model. The composition of benzene in the top and
bottom stage after VFP are shown in Figure 4.4 for the quasistationary steady
state with saturated liquid feed (black) and a +5% (red) and a 5% (blue) change
in the feed temperature.
Figure 4.4: Benzene composition in condenser (top) and reboiler (bottom) for
saturated liquid feed (black), +5% in feed temperature (red) and 5% in feed
temperature (blue).
When the feed temperature is increased to a value above the boiling point, a flash
will occur. As described in the previous chapter, a fraction of the liquid will flash
to vapour and be introduced to the column during VFP. After the flash, a new
boiling point will be found. In this case, for an equimolar benzene and toluene
mixture at atmospheric pressure, the saturation temperature is 365.24 K, and the
temperature after a 5% increase followed by flash is 365.62 K. This is a small
temperature change. However, the vapour feed goes from 0 mol/s to 1.89 mol/s
over the 12 second duration of VFP. Increasing the feed temperature thus lowers
the top concentration and increases the bottom concentration of benzene. Here
it is assumed that the flashed feed can be separated, allowing the liquid to be
introduced during the LFP and the flashed vapour during VFP. When the feed is
flashed, only a fraction of the total feed is introduced as liquid (93.94%), which
means less liquid holdup will be on the trays below the feed stage. On the other
hand, the flashed vapour will mix with the vapour from the feed stage and rise.
However, it was found that the flashed vapour had a lower composition of benzene
than the vapour from below the feed location. This means a higher fraction of
toluene will be in the vapour phase from the feed stage to the condenser, thus
lowering the benzene concentration.
steam. The duration of the flow periods are tV F P = 38 s and tLF P = 2 s, with a
combined cycle time of 40 s.
Table 4.3: Liquid and feed flow rates for the experiments presented by Maleta
et al. [59].
The experimental data presented by Maleta et al. [59] for the nine experiments
shown in Table 4.3 are used for verification of the proposed model in Chapter 3.
The NRTL thermodynamic model is used for the simulated process, with binary
interaction energy parameters from Aspen, shown in Table 4.4.
Table 4.4: Binary NRTL interaction parameters for the ethanol water mixture
studied for the model verification.
It is assumed that the point efficiency is equal to one. The top of the column has
a pressure of 101 kPa, and the bottom 109 kPa, with a column pressure drop of
8 kPa. Ten stages are used, with the liquid feed to the top stage and the vapour
feed to the bottom stage.
Figure 4.5 shows a comparison of the experimental and the simulated data using
the proposed model for the product flow rates and ethanol compositions.
150 40
35
30
100 25
20
15
50 10
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
750 10 0
700
10 -3
650
10 -6
600
550 10 -9
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
Figure 4.5: Comparison of experimental and simulated product flow rates and
compositions. Blue dots are experimental data and red dots are simulated
data using the proposed model. NB: the yaxis for the bottom composition
(bottom right) is on a 10logarithm scale.
There is some difference between the experimental and the simulated model data.
However, it seems the general trends are present for most of the experiments.
Especially experiment no. 6 seems to have a good agreement between the ex
perimental and the simulated data. The bottoms flow rates, and distillate compo
sitions seem to follow the general trends. The composition in the bottom product
is very low in the simulated results compared to the experimental results. For
the experiments, Maleta et al. [59] states that there was a measurement error of
0.002 vol% in the waste (or bottoms) stream and 0.1 vol% in the liquid feed and
distillate streams. This measurement error is not enough to fully explain the dif
ference in the experimental and the simulated data. However, from looking at the
in and outputs data provided by Maleta et al. [59] it was clear that their presented
mass balance was not satisfied. Figure 4.6 shows the feed flow inputs and the
experimental and simulated outputs.
850
800
750
700
650
1 2 3 4 5 6 7 8 9
Figure 4.6: Experimental data inputs (yellow) and the experimental (blue) and
simulated (red) data outputs for the nine experiments described by Maleta et
al. [59].
From Figure 4.6 it is clear that the experimental data has a significant difference
between the process inputs and outputs, which may have affected the results.
The simulated model mass balance will inherently be satisfied. Experiment no. 6,
in Figure 4.6, has the lowest error in the mass balance for the data. It was also
experiment no. 6, which had the best match in the model predictions shown in
Figure 4.5.
In order to compensate for the mass balance error from the experiments, new sim
ulations are run, which have the feed streams scaled to the experimental distillate
and bottoms flow rate instead. So the difference in the experimental reported in
and outlets have been subtracted from the model input. The new feed flow rates
are shown in Table 4.5. The output flow rates and product compositions are found
again for the nine experiments with the new adjusted feed flows. The results of
these are summarised in Figure 4.7.
Table 4.5: Adjusted liquid and vapour feed flow rates for the experiments
presented by Maleta et al. [59] based on the experimental outputs.
150 40
35
30
100 25
20
15
50 10
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
750 10 0
700
10 -3
650
10 -6
600
550 10 -9
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
With the adjusted feed flows, the distillate and bottom product flow rates are much
better fitted for the model and experimental outputs. Furthermore, it seems the
distillate composition also is fitted better. However, the bottom product compo
sition is still relatively far from the experimental values, although this could be
attributed to the low values and the related uncertainty.
The model has not been completely verified with experimental data. However,
one of the main issues is that the mass balances from the experiments are not
closed. Other reasons for this deviation between the experimental and the model
data could be the model thermodynamics or assumptions, e.g. point efficiency
equal to one. The issue with the experiments’ mass balance not being closed
could come from the pump or valves used to control the feed to the column, but it
is not easy to say. In any case, it seems that both the experimental data and the
simulated data has some degrees of uncertainty that could explain the difference
between the model and experiment outputs. One approach to get better fitting of
the data is to use data reconciliation, as done by Bisgaard et al. [69]. However,
to do it for this case by Maleta et al. [59] key information is missing, such as the
standard deviation of the flow inputs.
Model verification was attempted using experimental data from the literature. How
ever, it was shown that due to the uncertainty in the experimental data regarding
the mass balance, a proper fitting of the model could not be carried out.
The case study in this chapter is based on the rate kinetic study published by
Rehfinger and Hoffmann [78], with the use of a heterogeneous catalyst. Isobutene
(iBut) and methanol (MeOH) can react over an acidic catalyst, e.g. Amberlyst 15,
to form MTBE in a fast reversible reaction [78]:
Inert nbutene (nBut) is mixed with isobutene in the vapour feed to help with the
reactive separation process [70, 71].
The UNIQUAC model is used for liquid activities, and the vapour is assumed to be
ideal. Due to the similar interactions between nbutene/isobutene and methanol
or MTBE, it is assumed that nbutene and isobutene have the same UNIQUAC
interaction parameters [78]. It should be noted that both the butenes and MTBE
form azeotropes with methanol. The binary interaction parameters from Rehfinger
and Hoffmann [78], are listed in Table 5.1. Two additional parameters are needed
for the UNIQUAC equation: the Van der Waal volume, ri , and surface area, qi , of
the molecules of component i. These have been found in Aspen Plus.
Table 5.1: UNIQUAC interaction parameters aij from Rehfinger and Hoffmann
[78] in J/mol.
The UNIQUAC thermodynamic model has been used to describe the liquid activ
ities in many previous studies on a reactive distillation of MTBE [70, 71, 75, 77,
78]. The reaction rate expression, which was proposed by Rehfinger and Hoff
mann [78], includes the activities of the reactants and product. Thus, the choice
of thermodynamic model affects both the separation and reaction.
If this is not satisfied the direction of the reaction can be determined. If Keq (Tn ) <
an,M T BE
an,iBut an,M eOH
then the activity of MTBE is too high and iBut and MeOH are formed.
an,M T BE
Otherwise, if Keq (Tn ) > an,iBut an,M eOH
then the activity of MTBE is too low and iBut
an,M T BE
and MeOH are consumed. At equilibrium, Keq (Tn ) = an,iBut an,M eOH
, no changes in
the molar holdups occur.
an,M T BE
If Keq (Tn ) <
an,iBut an,M eOH
[ ] [ ]
then νn,nBut , νn,iBut , νn,M T BE , νn,M eOH = 0, 1, −1, 1
an,M T BE
If Keq (Tn ) >
an,iBut an,M eOH
[ ] [ ]
then νn,nBut , νn,iBut , νn,M T BE , νn,M eOH = 0, −1, 1, −1
an,M T BE
If Keq (Tn ) =
an,iBut an,M eOH
[ ] [ ]
then νn,nBut , νn,iBut , νn,M T BE , νn,M eOH = 0, 0, 0, 0
The equation (5.3) and the variable ϵn must be included in the model for an equi
librium reaction, thus adding another algebraic equation and variable that must
be solved for at each time step during the VFP. No reaction is assumed to occur
during the LFP as the catalyst is assumed to be a solid placed on the stages.
Wang et al. [77] investigated the multiplicity in MTBE reactive distillation assuming
6820
ln Keq = −16.33 + (5.4)
T
For a rate kinetics controlled reaction the reaction rate in mol/(s kg catalyst) is
given by Rehfinger and Hoffmann [78].
( )
aiBut 1 aM T BE
r = qacid kf (T ) − (5.5)
aM eOH Keq,kin (T ) a2M eOH
( )
−11110 [ ]
kf (T ) = 3.67 · 10 exp
12
mol/(s equiv. H+ ) (5.6)
T
[ ( ) ( )
1 1 T
Keq,kin (T ) = 284 exp −1492.77 − − 77.4002 ln
T 298.15 298.15
(5.7)
+0.507563(T − 298.15) − 9.12739 · 10−4 (T 2 − 298.152 )
]
+1.10549 · 10−6 (T 3 − 298.153 ) − 6.27996 · 10−10 (T 4 − 298.154 )
Where qacid is the amount of acid groups on the resin per kilogram catalyst, qacid =
4.9 equiv. H+ /kg catalyst [75]. The reaction rate presented by Rehfinger and Hoff
mann [78] is a Langmuir Hinshelwood rate expression, and is used in previous
studies on reactive distillation of MTBE [70, 75, 76]. Hauan et al. [71] and Wang
et al. [77] assumed chemical equilibrium reaction with temperature dependent
equilibrium constants based on the work by Rehfinger and Hoffmann [78], see
equation (5.4).
or rate kinetics. The column performance and the stage behaviour over a cycle
are compared for the two reactive cyclic distillation cases and the conventional re
active distillation process. The column design proposed by Wang et al. consists
of 17 stages, of which eight are reactive, and has both a liquid methanol and a
vapour butene (nBut/iBut) feed. The overall conversion from column inlet to outlet
of iBut was 90 %, and the bottom product composition of MTBE was 99.2 mol%.
The number of stages, feed specifications, pressure, and pressure drop for the
equivalent cyclic column design are kept constant. The desired bottom product
purity of 99.2 mol% MTBE is kept as a constraint, while the other outputs, such
as product flow rates, reactant conversion and top product composition, are not
constrained. This is done since the product of interest is the MTBE in the bot
tom, while the top product is mainly inert nbutene and unreacted methanol and
isobutene. This means that for the periodic operated column, the product flow
rates, reactant conversions and top products can increase or decrease compared
to the conventional column as long as the bottom product purity is kept. In order
to achieve the bottom product specification in the equivalent cyclic column, the
reflux and boilup ratios are adjusted.
When going from classical to the cyclic operation of a distillation column, two new
variables are introduced: the duration of the VFP and the LFP. These need to be
specified. Here the cycle time is also used to describe the in and outputs so that
the liquid and vapour feeds are scaled with regards to the duration of the cycles.
Where the duration of the complete cycle is tcyc = tV F P + tLF P in seconds per
cycle. With these scaled feeds, the same processing capacity as the conventional
column in terms of the volumetric feed rate can be accommodated in the cyclic
operation over a cycle time. The liquid feed will be introduced to the column during
the LFP and the vapour feed continuously during the VFP.
variables. In order to solve this nonlinear algebraic equation system, initial esti
mates for the output variables must be determined. For the conventional reactive
distillation, an assumption of chemical equilibrium reaction has been used, as is
done in the work by Wang et al. [77].
The MTBE reactive distillation has multiple steady states with both a low and a
high state of conversion for isobutene as shown by Jacobs and Krishna [70]. The
reflux and boilup should be chosen or adjusted to ensure the desired high state
of conversion of isobutene. The boilup ratios for the cyclic cases were chosen as
a low value that gives the high conversion state. The boilup ratios for the (chem.
eq.) and the (kin.) cases slightly differs, but in both cases, the target bottom
product purity of 99.2 mol% are reached. This similar operation implies that the
assumption of chemical equilibrium is reasonable for the MTBE reactive distillation
case. A lower boilup ratio would result in the low conversion of the reactants, and
a higher boilup ratio would lead to the product, MTBE, being sent up through the
column as vapour. The choice of reflux must also be chosen carefully, as the
reflux flow rate is equal to the liquid holdup on stage 2 after an LFP. The vapour
flow over liquid flow ratio, V /L, calculated as averaged vapour flow from reboiler
divided by the reflux plus liquid feed over a cycle, shows the internal vapour and
liquid dynamics. There is a higher amount of vapour going through the column in
the cyclic columns, despite a shorter time for vapour flow movement. Overall the
two cyclic column designs for the different reaction types are quite similar, with
only minor differences in the boilup ratio and V /L ratio. The duration of the VFP
and LFP are here chosen as tV F P = 10 s and tLF P = 5 s. The duration of the
LFP is assumed to be sufficient enough to allow the liquid to drain completely.
In a pilot study by Maleta et al. [60] a draining period of 2 seconds was found
to be satisfactory for complete draining. This was, however, a pilot study with a
liquid feed of 50100 L/hr, so in order to account for a higher liquid holdup in the
reactive MTBE column, the tLF P is set to 5 seconds. As mentioned, the VFP is
the period where vapour is sent up through the column while the liquid holdups
are kept on each stage. During the VFP, there is both mass and heat transfer
between the phases and reaction in the liquid phase. The duration of the vapour
flow period should therefore be selected with care. Here it is set to tV F P = 10 s,
as the MTBE reaction is fast. During the LFP, there is no mass/energy transfer
between the phases nor a reaction taking place. Thus the effect of the duration
of the LFP is less noticeable than the duration of the VFP. However, tLF P should
be selected so that there is time for each stage to be drained and the liquid feed
can be introduced to the column, and products can be removed. Furthermore,
the duration of the LFP also influences the liquid and vapour feed, see equations
(5.8)(5.9).
The temperature and composition profiles for the equivalent cyclic column are
shown in Figure 5.1 compared to the conventional column.
80 80 80
x (mol%)
60 60 nBut 60
iBut
40 40 MTBE 40
MeOH
20 20 20
0 0 0
5 10 15 5 10 15 5 10 15
It appears that both cyclic column cases, (chem. eq.) and (kin.), yield similar
results in the profiles shown in Figure 5.1. The equivalent cyclic columns are
not optimally designed as shown on the profiles in Figure 5.1, where stages 12
14 seem to have a low difference in temperature and compositions compared to
the conventional column. For the cyclic cases, it can also be seen that there is
a higher nbutene purity in the top product compared to the conventional case.
The explanation for this is that the cyclic column has a higher conversion of both
methanol and isobutene. Thus fewer reactants are lost in the product streams.
The heat of reaction was set to be constant at ∆Hnreac = −43.6 kJ/mol, this gives a
relatively small contribution to the energy balance in equation (3.37). For example,
for stage 11 in the cyclic (chem. eq.) column at time t = 0 the total value of the
energy balance was found to be dM11 h11/dt(t = 0) = 1.66 MJ/s. Of this the reaction
reac
∑ eq
heat contribution ∆H11 R11,i (t = 0) is only 1.41 kJ/s or 0.09 % of the total
value of dM11 h11/dt(t = 0). The effect of the heat of reaction is negligible for the
MTBE case; however, the contribution to the internal column flows is rigorously
accounted for.
Key results from the conventional and cyclic columns are shown in Table 5.3.
Table 5.3: Summary of reactive distillation cases. The cycle time is 15 sec
onds, of which 10 seconds is the vapour flow period.
As shown in Table 5.3, the cyclic columns have a significant lower energy re
quirement per cycle. Where the conventional column needs a continuous energy
supply over 15 seconds per cycle, the cyclic column only has energy supplied and
removed during the vapour flow period, which is 10 seconds in this case. There is
a higher bottom output flow rate and a lower distillate flow rate for the cyclic cases
compared to the conventional. This is related to the conversions of the reactants,
which are also higher for the cyclic cases, meaning fewer reactants are removed
from the column than for the conventional case. While the same bottom product
purity of MTBE is reached in all cases, the cyclic column has a higher purity of
nbutene in the top product. Both the (chem. eq.) and the (kin.) cyclic cases have
similar outputs.
Figure 5.2 shows the number of moles of MTBE per second per total molar holdup
formed on stage 4 and 11 in the conventional case and the equivalent cyclic col
umn (chem. eq.) case over a cycle and Figure 5.3 shows the conventional case
compared to the (kin.) case. The figures also show the liquid and vapour compo
sitions of the reactants and products isobutene, MTBE and methanol. The solids
lines shows the cyclic operated process and the dotted lines the conventional op
eration. The conventional operation is in steadystate, i.e. there is no change in
how much MTBE is formed over time, nor is there change in the compositions,
and the production and compositions will be steady for the entire cycle time of 15
seconds. For the cyclic operation only the development in production and com
positions during the vapour flow period is plotted. Note that the reaction rate in
the conventional column stage 4 is not zero, but merely very low.
0.04
Cyc. VFP
4
Conv.
End of VFP 0.02
3
2
0
1
0 -0.02
0 5 10 15 0 5 10 15
Time (s) Time (s)
5 50
iBut
4 40
MTBE
x i (mol%)
3 MeOH 30
2 20
1 10
0 0
0 5 10 15 0 5 10 15
4 25
3 20
y i (mol%)
15
2
10
1 5
0 0
0 5 10 15 0 5 10 15
Time (s) Time (s)
Figure 5.2: MTBE produced in stage 4 and 11 for the chem. eq. case (top).
Dotted line indicates the end of VFP. Middle plots shows the liquid compo
sition, and bottom plots are the vapour composition. For middle and bottom
plots, dotted lines show the conventional operation and solid lines the cyclic
operation.
2 -0.01
0 -0.02
0 5 10 15 0 5 10 15
Time (s) Time (s)
5 50
iBut
4 40
MTBE
x i (mol%)
3 MeOH 30
2 20
1 10
0 0
0 5 10 15 0 5 10 15
5 30
4
y i (mol%)
20
3
2
10
1
0 0
0 5 10 15 0 5 10 15
Time (s) Time (s)
Figure 5.3: MTBE produced in stage 4 and 11 for the kin. case (top). Dotted
line indicates the end of VFP. Middle plots shows the liquid composition, and
bottom plots are the vapour composition. For middle and bottom plots, dotted
lines show the conventional operation and solid lines the cyclic operation.
With the cyclic column cases, an increase can be seen in the overall formation of
MTBE in stage 4 compared to the conventional case. For stage 4, 1.49·10−5 moles
MTBE per cycle per total holdup for the conventional case and 1.17 · 10−3 moles
MTBE per cycle per total holdup for the cyclic column (chem. eq.) are produced.
If a rate kinetics expression, the (kin.) case, is used instead, the total production
over a cycle is 5.52 · 10−3 moles MTBE per cycle per total holdup. As shown in
Figure 5.2 and 5.3 for stage 11 with the cyclic operation, for both the (chem. eq.)
and the (kin.) cases, the equilibrium is surpassed and MTBE goes from being
produced to being consumed. This means the conventional case has a production
of 6.12 · 10−2 moles MTBE per cycle per total holdup, the cyclic (chem. eq.) case
has a production of 9.73 · 10−3 moles MTBE per cycle per total holdup in stage 11.
The total production of MTBE over a cycle for the rate kinetics cyclic (kin.) case in
stage 11 is 2.38 · 10−2 moles MTBE per cycle per total holdup. For the (chem. eq.)
case, the equilibrium is crossed after approximately 4 seconds, and for the (kin.)
case, it is crossed after approximately 6 seconds, which is also shown with the
increase in the overall production of MTBE over a cycle. So while the top stage in
the reactive zone shows an improvement in MTBE production for the cyclic case,
the reaction in stage 11 goes from producing MTBE to consuming MTBE, thus
reducing the overall production per cycle. In the reactive cyclic column, liquid
feed enters stage 10 during the LFP, mixes with the holdup, and is then drained to
stage 11 when a VFP is initiated. The vapour feed is introduced above stage 12,
i.e. the first stage this vapour will come into contact with is the stage above, stage
11. For the cyclic operated column on stage 11 at the start of a VFP, there is a liquid
with high content of methanol and vapour with a high concentration of isobutene.
Studying the reactive stages between stages 4 and 11, which is not shown here,
revealed that in all the other reactive stages, an increase in production of MTBE
over a cycle was achieved for the cyclic distillation cases. Furthermore, it is only
in stage 11 that the equilibrium is crossed.
For stage 11 in Figure 5.2 and 5.3 the liquid and vapour compositions follow the
reaction rate for the cyclic column. At the beginning of the cycle, methanol and
isobutene react to form MTBE and eventually, the reaction equilibrium is crossed,
and MTBE is now reacting to form methanol and isobutene. The isobutene in
the liquid is almost constant due to the constant vapour feed from below. The
compositions in the vapour all increase over the cycle due to the simultaneous
reaction and separation as well as the continuous supply of vapour from the stages
below. In stage 4 there is only a low degree of reaction as is shown in Figure 5.2,
which is also seen in the low change in compositions.
For further analysis of the contributions from both the reactive part and the sep
aration part of the mass balance, see equation (3.36), dxn,i/dt, are plotted for the
stages 4 and 11 in Figure 5.4 and 5.5 for isobutene, MTBE and methanol for the
(chem. eq.) and the (kin.) cases respectively. The inert component nbutene is
not shown. Both the reactive part, the separation part and the combined contribu
tions of dxn,i/dt are plotted.
/ The change in liquid composition over time are found
∑
as dxn,i/dt = dMn,i/dt dMn,i/dt, see equation (3.36).
2 0.01
0 0
-0.01
-2 react
sep -0.02
-4 reac+sep
-0.03
0 5 10 0 5 10
10 -4
6
dx MTBE/dt (mol/s)
0.06
4 0.04
0.02
2 0
-0.02
0 -0.04
0 5 10 0 5 10
10 -4
2
dx MeOH/dt (mol/s)
0.04
0
0.02
-2 0
-0.02
-4
-0.04
0 5 10 0 5 10
Time (s) Time (s)
0.01
-2
react 0
sep
-4 reac+sep -0.01
-6 -0.02
0 5 10 0 5 10
10 -3
3 0.06
dx MTBE/dt (mol/s)
0.04
2
0.02
1
0
0 -0.02
0 5 10 0 5 10
10 -3
dx MeOH/dt (mol/s)
0 0.04
-2
0.02
-4
0
-6
0 5 10 0 5 10
Time (s) Time (s)
It should be noted that the reactive part for MTBE in Figure 5.4 and 5.5 is the
same as the reaction rates in the top rows of Figure 5.2 and 5.3 for the cyclic
process. A negative dxn,i/dt means, for the reactive part that the component is
being consumed and for the separation part that the component will go to the
vapour phase.
Generally, the reactive part of the mass balances is lower than the separation
part in stage 4. For stage 4, there is a clear order in the reactive and separation
parts, with the reactive parts with the lowest contribution to the mass balance. For
MTBE, where the reaction rate is positive in stage 4, the separation part is also
positive, giving a more significant positive combined effect, meaning an increase
in the liquid composition of MTBE. For the isobutene and methanol, the reaction
rates are negative, i.e. the reactants are consumed, the separate parts of these
mass balances are positive, giving a combined effect somewhere between the
reactive and separation parts.
In stage 11, where the reaction equilibrium is being surpassed, significant changes
in dxn,i/dt can be seen in Figure 5.4 and 5.5 as well. Looking at MTBE in stage 11,
the separation part of the mass balance is slowly increasing, i.e. more and more of
the MTBE in the vapour phase is going to the liquid phase. When the equilibrium
is surpassed, and high concentrations of MTBE are present in the liquid holdup,
the separation part of MTBE will slowly decrease, indicating less MTBE are moved
from the vapour to the liquid phase due to the amount of MTBE in the liquid. The
reaction part goes from positive to negative, corresponding to the reaction from
producing MTBE to consuming it. The methanol composition in stage 11 also
follows the observations made in Figure 5.2 and 5.3. The separation part of the
isobutene plots shows that after the reaction equilibrium is reached, the isobutene
in the liquid goes to the vapour phase since isobutene is a light component.
Looking at Figure 5.4 and 5.5, if the combined effects of the reaction and the
separation in the mass balance is equal to zero, no change in liquid composition
occurs. If it is positive, the liquid composition will increase, and if negative, it will
decrease. This follows the liquid compositions shown in Figure 5.2 and 5.3.
Regarding the two different reaction cases, (chem. eq.) and (kin.), it seemed from
the profiles shown in Figure 5.1 and outputs in Table 5.3 that there was almost no
difference between the two reaction kinetics assumptions. Thus the assumption
of a chemical equilibrium reaction seem to be valid in this case. However, in the
stage analysis over a VFP shown here, see Figures 5.25.5, a clear difference
in the reaction and separation was shown. This does not necessarily affect the
The analysis of the stage behaviour for reactive cyclic distillation of MTBE shows
that both the reaction and separation plays important parts. This is evident even
with the assumption of equilibrium reaction. For the bottom reactive stage 11, it
was clear that while reaction and separation coincide, they also contribute sig
nificantly to the stage behaviour. For example, as the production rate of MTBE
decreases, the mass transfer of MTBE to the liquid phase increases, but since
the consumption rate of MTBE is higher than the mass transfer rate, the overall
combined liquid composition decreases at the end of the VFP.
In order to better understand the reactive cyclic distillation process and identify the
variables that are important to keep in mind when designing a reactive cyclic dis
tillation column, the key design variables must be found. When these key design
variables are identified, it is easier to study the effect of each and propose suit
able settings for them. When the effect of each of these key variables is known,
it is possible to suggest a column design, although not necessarily an optimised
column design.
From the above results for the reactive cyclic distillation process, it can be seen
that the cyclic column is overdimensioned. Figure 5.1 shows that there are three
stripping stages, 1214, where little change in compositions and temperatures oc
cur. These stages could potentially be removed without any significant loss in
overall separation efficiency. The rectifying stages, 23, also seem to undergo a
very low change in composition and temperature. Regarding the reactive stages,
it is clear from Figure 5.2 and 5.3 that the bottom stage in the reactive zone, stage
11, is not optimally designed since there is a negative production of MTBE, mean
ing a shift in the equilibrium towards the reactants. Further evidence is that the
MTBE produced per cycle per total holdup in stage 11 in the cyclic column is not as
high as in the conventional case due to MTBE consumption after approximately 4
seconds. However, this is the stage in the column where both the butene vapour
feed and methanol liquid feed mixes, so it is an essential reactive stage, and it
might not necessarily be a benefit to remove the bottom reactive stage. In the top
of the reactive zone, stage 4, there is a very low production of MTBE. As shown
in the previous section, the liquid compositions here do not change significantly
either, so this might be a possibility for a reduction of the reactive zone.
The reflux and boilup ratios must also be determined. The reflux can almost be
freely chosen to ensure a high purity product as long as the vapour flow rate from
the reboiler is chosen carefully as well [45]. In a cyclic operating process, the
reflux flow is the liquid holdup on stage 2 after an LFP, i.e. the amount added
to the column from the condenser over an LFP. The reflux will affect the boilup
and vice versa. Thus, when choosing the reflux and boilup ratios, it might be a
good idea to first determine a suitable reflux value and then find the corresponding
boilup that gives a high purity product or vice versa. In the case of MTBE, there
are multiple states of reactant conversions [70], so it is necessary to keep this in
mind when choosing the boilup and reflux and ensure the chosen values give the
high state of reactant conversion. One approach is to initially keep the boilup and
reflux ratios from the conventional column and gradually decrease or increase
them until the desired product purity is reached.
The duration of the LFP and VFP should also be chosen. The liquid flow period
duration, tLF P , is determined from the draining time of the stages. It should be at
least set to the time it takes to drain the slowest draining tray to ensure all liquid
is drained in the LFP, thus avoiding backmixing of the holdups. When setting the
tLF P , it must also be remembered that this is also where liquid feed and reflux are
added, and products are removed. The duration of the vapour flow period is more
difficult to determine. One must take account of the reaction time and separation
while considering the liquid holdup and vapour flow rate. In the case of MTBE
studied in this chapter, it was assumed the liquid and vapour feed were scaled
with regards to the cycle time, i.e. a higher cycle time means higher feed, which
in turn means higher liquid holdups on the stages. The duration of the VFP will
influence both the reaction and separation. A long VFP will mean more time for
the reaction to occur. In the case of MTBE, which is a fast reaction, the MTBE
would be consumed after approximately 46 seconds in stage 11. For slower
reactions, it might be a benefit to set a long VFP. A long VFP will also influence
the separation, so more components will experience mass and energy transfer
between the phases. This could lead to problems with more heavy components
transferring to the vapour phase or if the mass transfer rate of light reactants from
the liquid to the vapour phase is faster than the reaction. On the other hand, a
short tV F P will reduce the reaction and separation time, so that there might not be
a suitable conversion in the reactive stages or not enough mass transfer between
the liquid and vapour phases.
this. By keeping the duration of the VFP and LFP, the improved column can be
compared to the conventional and equivalent column. The boilup and reflux ratios
were increased slightly compared to the equivalent cyclic column. The specifica
tions for this new realisation is shown in Table 5.4. The column pressure drop in
Table 5.4 corresponds to the same pressure drop per stage in Table 5.2.
Table 5.4: Design specifications for the improved realisation of the cyclic dis
tillation column.
With these specifications from Table 5.4 it was shown that the target bottom prod
uct purity of 99.2 mol% for MTBE could be reached without significant reduction in
the other output parameters and conversions relative to the equivalent cyclic op
erated column. Figure 5.6 and Table 5.5 shows the column profiles and outputs
of the improved realisation of the cyclic column.
80 80
x (mol%)
60 60
nBut
iBut
40 MTBE 40
MeOH
20 20
0 0
2 4 6 8 10 12 2 4 6 8 10 12
420 420
400 400
T (K)
380 380
360 360
340 340
2 4 6 8 10 12 2 4 6 8 10 12
Stage Stage
Table 5.5: Summary of reactive distillation with the improved cyclic column.
The cycle time is 15 seconds, of which 10 seconds is the vapour flow period.
Improved Improved
cyclic column cyclic column
(chem. eq.) (kin.)
B (mol/cycle) 2943 2948
D (mol/cycle) 5313 5305
xD
nBut (mol%) 98.8 98.9
xB
MTBE (mol%) 99.2 99.2
χiBut (%) 98.8 99.0
χMeOH (%) 98.3 98.4
QB (MJ/cycle) 625 619
QC (MJ/cycle) 775 773
Again the two cases with either chemical equilibrium or rate kinetic reaction terms
yield similar profiles and results as shown in Table 5.5. As shown in Figure 5.6 the
improved realisation of the cyclic column does not have the stripping stages with a
low degree of separation, which was present in the equivalent cyclic column pro
files for stage 1214 in Figure 5.1. This implies that the stages where a low degree
of separation and reaction occurs can be removed without a significant reduction
in performance. Comparing the results from the smaller cyclic column in Table 5.5
and the cyclic column in Table 5.3 with 17 stages, it shows that while the bottom
product purity of MTBE is maintained at 99.2 mol%, the reactant conversions, top
product purity and bottom product flow rate are slightly reduced. The top product
flow rate and energy consumption are slightly increased from the reduction of the
total number of trays.
For the design of this short cyclic column, there is a tradeoff regarding column
size and energy consumption. When going from the equivalent reactive cyclic dis
tillation column, shown in Table 5.2 and results shown in Table 5.3, to the improved
column design, a small increase in energy requirement can be seen. Compared
to the conventional reactive column, the small cyclic column still has a lower en
ergy requirement. This new proposed realisation of a cyclic column for the MTBE
production is not optimal, as is it derived from observations of the cyclic column
equivalent to a conventional column. The backwardsintegration method used by
Pǎtruţ et al. [45], described in Chapter 2.2, could also have been used. However,
this approach would only estimate the number of the nonreactive stages followed
by an estimation or trialanderror approach to find the number of reactive stages.
In this case, a good initial estimate would be to keep the number of reactive stages
as in the conventional case and try to remove one stage at a time.
It is expected that reactive separation processes other than the MTBE production
can benefit from moving to periodically operated distillation. This was also the
case for the DME production shown by Pǎtruţ et al. [45]. However, based on the
currently available literature and the observations made in this study, it is difficult
to make general conclusions for other cases, as each reactive distillation case
behaves differently.
tillation. This higher separation efficiency was also shown in the MTBE case, and
it helped push the extent of reaction so that a high conversion of reactants was
reached.
By analysing the stage performance in the reactive section of the column, it was
found that the overall production of MTBE was indeed higher in the cyclic case.
This is due to the cyclic operation, which does not have backmixing of the liq
uid holdups and thus has a higher separation efficiency, favouring the desired
reaction. It was further shown that although the chemical equilibrium and the rate
kinetics cases had similar outputs, the dynamic behaviour over a cycle was not
the same. However, the assumption of a chemical equilibrium reaction is valid for
the MTBE case as the outlet results are very similar to the rate kinetics case.
Based on the observations made for the performance of the cyclic column cases,
improved and shorter cyclic column designs were proposed, which could lead to
a reduction in a number of stages without a reduction in the target bottom product
purity of 99.2 mol%. However, a small increase in the energy demand is necessary
when going from the large cyclic column to the small cyclic column design.
The value of this performance indicator show the extent of reaction that takes
place on a given reactive stage n during the VFP. If it is negative it means a overall
consumption of the product takes place.
The second performance indicator (P InII ), i.e. the relative distance to the equi
librium at the end of a VFP, is found by comparing the temperaturedependent
equilibrium constant, Keq,n , to the actual value of the equilibrium constant at the
end of a VFP. The temperaturedependent equilibrium constant at the end of the
VFP is found by using the definition of the Keq,n found in literature, based on the
Gibbs free energy of the components in the standard state, the activity coefficients
refer to. Here the temperaturedependent equilibrium constants are found from
the reference articles that each presented case is based on, see equations (6.15),
(6.21) and (5.7) [45, 78–80]. The actual value of the equilibrium constant at the
′
end of a VFP denoted Keq,n , is found from the activities of the reactive compo
nents:
∏
NC
′
Keq,n = aνn,i
i
(6.2)
i=1
This performance indicator gives the relative distance from chemical equilibrium at
the end of a VFP for stage n. The closer P InII is to one, the closer the liquid holdup
composition is to equilibrium. If this performance indicator is close to one after a
VFP it could indicate that the stage can be assumed to be controlled by chemical
equilibrium. Thus if multiple reactive stages have P InII ≈ 1, the reactive cyclic
distillation process can be simplified to the assumption of chemical equilibrium
reaction.
The third and last performance indicator for the stage analysis is the mean Damköh
ler number. Shah et al. [55] describe the use of a Damköhler number for use in the
evaluation of conventional reactive distillation processes. The Damköhler number
is defined as:
M0 kf
Da = (6.4)
V
The liquid holdup, M0 , is multiplied by the forward reaction rate constant, kf and
divided by the vapour flow rate, V . The forward rate constant in equation (6.4) is
in units of s−1 [55]. It is a dimensionless number that describes the reaction rate
compared to the mass transport phenomena rate. Shah et al. [55] classifies the
reaction based on the value of the Danumber as:
• Da < 0.1: the reaction rate is slow compared to the residence time, and the
phase equilibrium is the main effect.
• 0.1 < Da < 10: the process is controlled by the reaction rate kinetics.
• 10 < Da: the reaction is fast, and the chemical equilibrium is the controlling
effect.
In reactive cyclic distillation, where the residence time can be specified as the
tV F P , the expression for Da in equation (6.4) is slightly adjusted. The forward rate
constant is in units of mol/(s kg cat) instead of s−1 as shown in equation (6.4).
This is merely to account for the forward reaction rates used in this chapter, which
are all based on catalyst loading. Since no true steadystate exists in a cyclic
distillation, the Danumber is timedependent.
kf (t)mcat
n
Dan (t) = (6.5)
Vn (t)
A mean Davalue over the VFP is found by integrating the time dependent Da over
one cycle, and divide by the VFP duration in order to quantify the effect. Since
reaction only occurs during the VFP and not the LFP, it is only the duration of VFP
and not the entire cycle that is of interest.
∫ t=tV F P
1 kf (t)mcat
n
P InIII = Dan = dt (6.6)
tV F P t=0 V (t)
If, however, Da ≤ 0.1 & Keq ≤ 1 then reactive distillation is not applicable. In
the following analysis, it would be of great interest to see if the stagewise values
of the mean temperaturedependent equilibrium constant, K eq,n , and Damköhler
number, Dan , would result in a feasible design or not. The mean equilibrium
constant is found as:
∫ t=tV F P
1
K eq,n = Keq,n (t) dt (6.7)
tV F P t=0
The vapour flow rate is not assumed constant for the model presented in Chapter
3. However, the feasibility condition proposed by Pǎtruţ et al. [11] based on a
constant vapour flow rate, see equation (2.10), can still be of interest. In this
chapter, the following rewritten condition have been used.
L < V 2 · tV F P (6.8)
∑
NT
V N T · tV F P < L + LFn (6.9)
n=1
Where L is the reflux and LFn the liquid feed to stage n. The boilup vapour flow,
V N T , and vapour flow from stage 2, V 2 , are averaged over the VFP duration. It
is assumed any vapour feed is already implicitly accounted for in the value of V 2 .
∑ T F
Note that V N T + N n=1 Vn is not necessarily equal to V 2 , since there can be a
significant increase in vapour flow rate over a column. In the design study by
Pǎtruţ et al. [45] for the DME case, they accounted for the change in vapour flow
rate by multiplying the feed with the conversion of the reactant methanol and thus
assuming all of the product DME will go to the vapour phase.
The mass and energy balance stage model for reactive cyclic distillation pro
cesses, see Chapter 3, has been used for modelling and simulating the case
studies. For the model implementation, a rate kinetics reaction expression is as
sumed.
When a change in the vapour flow period is introduced, the liquid and vapour
feed can either be kept constant, or the feeds can be scaled to the cycle time so
that a constant throughput relative to the time is ensured. In this work, constant
throughput is desired, so the feed flow rates are defined as:
tcyc,new
LFnew = LFini · [mol/cycle] (6.10)
tcyc,ini
tcyc,new
F
Vnew F
= Vini · [mol/cycle] (6.11)
tcyc,ini
Where the subscript ini denotes the initial values and the subscript new denotes
the new values.
The duration for the LFP is chosen to be 5 seconds in all cases. This is chosen
to have comparable results, which only depends on the duration of the VFP. Fur
thermore, the initial reboiler and condenser holdups are chosen to be 50 kmol for
all cases, and the vapour phase is assumed to be ideal.
In the following section of this chapter, each of the three chosen case studies is
presented, the cyclic steady state is shown, and the stage and column perfor
mance, as well as the feasibility of the reactive case, are analysed. Perturba
tions in the duration of the VFP, tV F P , and the liquid and vapour flow rates are
introduced to the three reactive cases. The stage behaviour for selected reactive
stages for each case has been analysed regarding the performance indicators
mentioned in Section 6.1. Furthermore, full column simulations with disturbances
in the same inputs have been carried out. The individual stage performance of
this was also investigated and the feasibility of the reactive cyclic distillation pro
cesses.
Three stages for each case has been selected and isolated so that a change in
the tV F P , liquid holdup or vapour flow rate can be analysed. For the single stage
analysis perturbations of ±10 % in the tV F P , Mn−1 , Vn+1 have been introduced to
the selected stages. The parameter Mn−1 is the liquid holdup on the stage above
the selected stage, which will be fed to the selected stage at the next LFP. The
parameter Vn+1 is the vapour flow coming from the stage below. The investiga
tions are made as disturbances in a single parameter at a time, and then the new
periodic steady state is found. By isolating a single stage like this, it behaves as
a sort of a batch reactor with a vapour input in the bottom and a liquid feed in the
top after tV F P seconds. It is assumed that the vapour and liquid feeds are con
stant regarding flow rate, compositions, and temperatures, unless a perturbation
is introduced. This assumption means that a slight deviation from the full column
simulation is expected, as the vapour and liquid flows are timedependent.
For the full column analysis, a ±1% disturbance has been introduced to the tV F P ,
RR and BR parameters. A change in RR corresponds to a change in the liquid
holdup, and a change in BR corresponds to a change in the vapour flow rate. Only
a small disturbance of 1% instead of 10% is introduced, since a full column simu
lation is much more sensitive to these changes than the single stages. Significant
effects can still be identified and analysed. With the introduction of a disturbance,
the full column model was simulated until a new cyclic steady state was achieved,
which could take up to 1000 cycles, depending on the case. The same stages
chosen for the singlestage analysis is once again investigated, with the same
assumptions as before.
2 MeOH ⇀
↽ DME + H2 O (6.12)
kf
rDM E = ( )2 (6.13)
KW CW
1+ KM eOH CM eOH
) (
−98000
kf = 6.12 · 10 exp 7
[kmol/(s [kg cat)]] (6.14)
8.31T
( )
KW 2964
= Keq = exp −6.46 + (6.15)
KM eOH T
The concentrations of the water and methanol are denoted CW and CM eOH in
mol/L, respectively.
Pǎtruţ et al. [45] proposed a reactive cyclic distillation column design, with 17
stages of which seven were reactive, that gives a high purity water product in the
bottom. They also tried to see if a column design for a high purity DME product
(> 99 mol%) in the top would be feasible but concluded that such a design would
not be stable. The Wilson activity coefficient model is used to describe the liquid
phase. Pǎtruţ et al. mentions that DME and water are immiscible, which the
Wilson model can not account for. However, they also mentioned that only one
liquid phase exists at 313 K if the methanol mole fraction is above 0.07. In the
entire column, the methanol content is above 0.07 and the temperature above 313
K, so it is assumed that the Wilson activity coefficient model is sufficient for this
case study. The activity coefficient parameters are given by Pǎtruţ et al. [45], the
interaction parameters, aij and bij , are shown in Table 6.1.
aDM
i,j
E
DME MeOH H2 O
DME 0 1.617 2.2595
MeOH 7.90976 0 5.8335
H2 O 3.58498 3.365 0
bDM
i,j
E
(K) DME MeOH H2 O
DME 0 640.759 281.331
MEOH 3065.95 0 2057.172
H2 O 1629.76 783.498 0
The in and outputs for the DME case is shown in Table 6.2 and Figure 6.1 show
the column profiles in cyclic steady state at the end of a VFP.
Table 6.2: Column and process specifications for the three case studies in a
reactive cyclic distillation process.
DME
Catalyst load 10 kg/stage L1 (mol/cyc) 2098.4
Liquid phase Wilson D (mol/cyc) 82.54
NC 3 xD (mol%) 81.7 mol% DME
DME 18.27 mol% MeOH
MeOH 0.02 mol% H2 O
H2 O B (mol/cyc) 67.46
NT 17 xB (mol%) 0.00 mol% DME
NR 7 (4:10) 0.04 mol% MeOH
P 10 atm 99.96 mol% H2 O
NF 7 BR 0.0226
F
x 100 mol% MeOH RR 25.423
F (mol/cyc) 150 (L) Conversion, χi 89.93 % MeOH
tV F P (s) 60 QB (MJ/cyc) 45.94
tLF P (s) 5 QC (MJ/cyc) 49.55
tcyc (s) 65
V 2 (mol/cyc) 2157.72
V N T (mol/cyc) 1130.04
The column profiles at the end of VFP shows that the bottom product is high purity
water and the DME purity in the top is above 80 mol%. Furthermore, the vapour
flow rate at the end of VFP shows that there is a considerable change from the
vapour leaving the reboiler to the vapour entering the condenser. The average
values over a VFP of V 2 and V N T are shown in Table 6.2. In the DME process, a
midboiling component, MeOH, forms a light product, DME, and a heavy product,
water. Thus the separation of light components from the liquid phase to the vapour
phase is further aided by the reaction as a mid boiling component forms a light
product, which then can be transferred to the vapour phase.
In the DME case study that was previously published by Pǎtruţ et al. [45] another
model that does not include energy balances for each stage was used. Small
deviations from the results presented by Pǎtruţ et al. [45] are therefore expected.
100
50
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
500
450
400
350
300
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Figure 6.1: Base case column profiles, top: liquid composition, middle: tem
perature, bottom: vapour flow rate, at the end of a VFP in cyclic steady state
for the DME case.
Starting with the first performance indicator, the duration of the VFP has the high
est effect compared to the vapour flow and holdup. In all three stages, an increase
in tV F P results in higher production of DME over a cycle and a lower production if
tV F P is reduced by 10 %. As the VFP duration increases, so does the residence
time, meaning the reaction has a longer time producing DME. The effect of the
vapour flow rate and the liquid holdup is small. The DME produced is highest in
the top of the reactive zone, stage 5, and decrease in stage 7 and 9.
10 10 10
5 5 5
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
3 3 3
2 2 2
1 1 1
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
Figure 6.2: The effect of ±10 % perturbations in tV F P , Vn+1 and Mn−1 for
single stage studies on the performance indicators for the DME case in stage
5 (left), stage 7 (middle) and stage 9 (right).
For the second performance indicator, it seems all the stages are very far from
equilibrium. Stage 9 is closest and is also affected the most by changes in the
inputs tV F P , Vn+1 and Mn−1 . With an increase in the distance to equilibrium for
when tV F P and Vn+1 are increased or when Mn−1 is decreased can be seen.
Finally, the mean Damköhler number over a cycle is investigated. Again the most
significant effects are seen for stage 9. The vapour flow rate, Vn+1 , seems to
have a higher effect in stage 5 and 7 than in stage 9 when comparing tV F P and
Mn−1 . Lower holdup or higher tV F P results in higher Damköhler values. A lower
Vn+1 also gives a higher Davalue as expected from equation (6.5). In the three
stages, the value of P InIII is all below 0.1, meaning, according to Shah et al. [55]
that the reaction rate is slow relative to the residence time. The main driving force
is, therefore, the phase equilibrium, which also corresponds to a longer VFP giving
higher values of the performance indicators and that the reaction is far from the
equilibrium as shown for P InII .
DME full column analysis
The three performance indicators are plotted for different changes in the inputs
(±1 %), the results are shown in Figure 6.3.
10 10 10
5 5 5
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
3 3 3
2 2 2
1 1 1
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
For evaluation of whether the DME process is feasible, according to the framework
by Shah et al. [55], the mean Damköhler and the mean equilibrium constant over
a cycle have been calculated for each of the selected stages. The results are
shown in Table 6.3.
Table 6.3: Values of K eq,n and Dan for DME case in cyclic steady state.
K eq,n Dan
Stage 5 2.1 0.01
Stage 7 2.1 0.01
Stage 9 1.7 0.03
In all cases, the value of Da is below 0.1, and the value of K eq is above 1. This indi
cates a slow forward reaction with an even slower reverse reaction [55], meaning
this reactive cyclic distillation process is feasible.
Evaluating the feasibility conditions in equations (6.8) and (6.9) for the cyclic steady
state of the DME case gives:
This shows that the vapour flow rate and reflux are satisfactorily chosen. The
significant change in vapour flow rate from the reboiler to the condenser can be
seen from the values of V 2 · tV F P = 2157.7 mol/cycle and V N T · tV F P = 1130.0
mol/cycle.
MeOH + HOAc ⇀
↽ MeOAc + H2 O (6.18)
Zuo et al. [80] described the reaction kinetics using a heterogeneous catalyst,
NKC9.
[ ]
a aH2 O
kf aM eOH aHOAc − M eOAc
Keq
rM eOAc = (6.19)
(1 + 4.95aM eOH + 3.18aHOAc + 4.16aM eOAc + 5.24aH2 O )2
( )
−55275.94
kf = 5.1093 · 10 exp 10
[mol/(min [kg cat])] (6.20)
RT
( )
2565.1
Keq = exp − 4.7335 (6.21)
T
Zuo et al. [80] proposed a reactive conventional distillation column design with
45 stages, of which 16 stages were reactive. As previously mentioned, one of the
benefits of cyclic distillation operation is the reduced number of stages. It is there
fore suspected that a shorter reactive cyclic distillation column can be designed
that gives the same or similar outputs. By using the design method described by
Pǎtruţ et al. [11, 45] combined with trial and error, a new column configuration for
the reactive cyclic distillation of MeOAc has been proposed. This column configu
ration has 26 stages, with 17 reactive, ensuring a high purity MeOAc product can
be removed from the top. In the cyclic operated distillation column, one additional
reactive stage has been added compared to the conventional column proposed
by Zuo et al. [80]. However, the total number of stages has been reduced by 19
stages. The liquid phase is described by the NRTL equation, with the interaction
parameters given by Zuo et al. [80] and shown in Table 6.4.
aM
i,j
eOAc
(K) MeOAc MeOH H2 O HOAc
MeOAc 0 234.8660 269.5857 415.2702
MeOH 130.5047 0 57.8859 310.2822
H2 O 866.2183 292.9637 0 385.2682
HOAc 239.2462 342.0151 48.5157 0
M eOAc
αi,j MeOAc MeOH H2 O HOAc
MeOAc 0 0.30 0.35 0.30
MeOH 0.30 0 0.30 0.30
H2 O 0.35 0.30 0 0.30
HOAc 0.30 0.30 0.30 0
A side reaction that can occur is the DME reaction. It has, however, been ne
glected in this work as it only has a low reaction rate [81, 82]. Furthermore, the
production of MeOAc can, in some cases, be regarded as a chemical equilibrium
reaction [81, 82]. However, since this work aims to analyse the reaction kinetics
effect, this assumption has not been used.
The column in and outputs for the MeOAc case are shown in Table 6.5 and Figure
6.4 show the column profiles in cyclic steady state at the end of a VFP.
Table 6.5: Column and process specifications for the MeOAc case in a reac
tive cyclic distillation process.
MeOAc
Catalyst load 25 kg/stage L1 (mol/cyc) 487.08
Liquid phase NRTL D (mol/cyc) 223.8
NC 4 xD (mol%) 99.21 mol% MeOAc
MeOAc 0.34 mol% MeOH
MeOH 0.44 mol% H2 O
H2 O 0.00 mol% HOAc
HOAc B (mol/cyc) 276.2
NT 26 xB (mol%) 0.00 mol% MeOAc
NR 17 (6:22) 9.84 mol% MeOH
P 1 atm 80.03 mol% H2 O
NF 5 & 22 10.12 mol% HOAc
F
x 100 mol% HOAc BR 0.0115
100 mol% MeOH RR 2.1764
F (mol/cyc) 250 (L) & 250 (L) Conversion, χi 88.81 % MeOH
tV F P (s) 120 88.81 % HOAc
tLF P (s) 5 QB (MJ/cyc) 23.30
tcyc (s) 125 QC (MJ/cyc) 21.02
V 2 (mol/cyc) 710.88
V N T (mol/cyc) 577.78
For the MeOAc case in Figure 6.4 a high MeOAc purity top product at the end of the
VFP can be seen. However, the vapour flow rate also changes from the reboiler
to the condenser with a smaller magnitude compared to the DME case. In the
MeOAc, a heavy component, HOAc, and a lighter component, MeOH, react and
form another heavy component, water, and the light product MeOAc. The change
in vapour flow rate over the stages mainly comes from the separation, moving the
light components, such as MeOAc, to the vapour phase and the heavy component
to the liquid phase.
For the MeOAc case, the reactive stages selected for further analysis are stages
10, 16 and 19, covering the reactive zone’s middle stages.
100
50
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
400
350
300
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
Figure 6.4: Base case column profiles, top: liquid composition, middle: tem
perature, bottom: vapour flow rate, at the end of a VFP in cyclic steady state
for MeOAc case.
Starting with the plots for P InI it can be seen that the highest amount of MeOAc is
produced at the bottom of the reactive zone. Furthermore, as for the DME case,
the tV F P has the largest effect on the P InI , with a higher tV F P giving a higher
production of MeOAc over a cycle.
For the second performance indicator, the higher placed stage 10 is much closer
to equilibrium than stage 19 at the end of VFP. This also corresponds to the low
amount of MeOAc being produced over a cycle for stage 10, while high for stage
19. By changing the tV F P , Vn+1 or Mn−1 by 10 % it seems the effect on P InII is low
in all stages. A very small increase in P InII can be seen when tV F P is increased
or when Vn+1 or Mn−1 are decreased.
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
1 1 1
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
80 80 80
60 60 60
40 40 40
20 20 20
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
Figure 6.5: The effect of ±10 % perturbations in tV F P , Vn+1 and Mn−1 for single
stage studies on the performance indicators for the MeOAc case in stage 10
(left), stage 16 (middle) and stage 19 (right).
Looking at the third performance indicator, the mean Damköhler number, all the
stages have values above 10. This means the main driving force is the chemical
equilibrium, which is also evident from the investigations for P InII . For the P InIII
analysis, the tV F P and Mn−1 parameters have almost no effect. However, the
vapour flow rate, Vn+1 , does have an significant effect, since it also is directly
integrated in the expression for Da, see equation (6.5). A lower vapour flow rate
results in a higher Davalue as expected.
MeOAc full column analysis
Stages 10, 16 and 19 are again chosen to analyse the column performance for the
MeOAc case. The three performance indicators are plotted for different changes
in the inputs for the full column analysis, the results are shown in Figure 6.6.
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
1 1 1
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
80 80 80
60 60 60
40 40 40
20 20 20
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
As for the DME case, the largest effect on the performance indicators, when a
change in tV F P , RR or BR is introduced, occurs in the lowest placed stage, stage
19 for the MeOAc case, although the effect can also be seen in stage 16. It
is shown that decreases in tV F P or RR or an increase in BR leads to a higher
production of MeOAc over a cycle, P InI . For the second performance indicator,
it can be seen that an increase in tV F P or decreases in RR and BR leads to
getting closer to the equilibrium at the end of a cycle. Finally, the plots for the
last performance indicator shows similar results as for P InII , where an increase
in tV F P or a decrease in RR or BR increases the mean Damköhler number for a
cycle.
To evaluate the feasibility of the MeOAc process, the mean Damköhler and mean
equilibrium constant over a cycle have been calculated for each of the selected
stages. The results are shown in Table 6.6.
Table 6.6: Values of K eq,n and Dan for MeOAc case in cyclic steady state.
K eq,n Dan
Stage 10 15.5 42.5
Stage 16 15.2 46.3
Stage 19 13.3 67.9
In all cases, there is a Davalue above 10 and a K eq value above 1. This corre
sponds to a fast forward reaction, and a slow backwards reaction [55].
Evaluating the feasibility conditions in equations (6.8) and (6.9) for the cyclic steady
state of the MeOAc case gives:
Based on the Da and K eq values and the flow rate feasibility check above, it is
evident that the MeOAc production is feasible in a reactive cyclic distillation pro
cess.
In the MTBE case the forward rate constant in equation (5.6) is dependent on
the catalyst ionexchange capacity, meaning kf (t) = kf (t)qacid for the MTBE case
in equation (6.1) and (6.6). The catalyst ionexchange capacity is qacid = 4.9
equiv./kg catalyst [75]. The conventional reactive distillation of MTBE assuming
rate kinetics for the reaction has been studied by Jacobs and Krishna [70], and
others [71, 75–77]. They assume a rather large catalyst loading of 1000 kg per
stage, which is necessary due to the high throughput and capacity.
For this case, the UNIQUAC thermodynamic model has been used to describe
the liquid phase with the interaction parameters given by Rehfinger and Hoffmann
[78], which is the same as used in the previous study on MTBE in reactive cyclic
distillation shown in Chapter 5. The UNIQUAC interaction parameters, aij , are
shown in Table 5.1.
The reboiler and condenser holdup after each cycle are chosen to be 50 kmol for
all cases. This, along with the choice of a rate determined reaction instead of an
equilibrium reaction, also mean the boilup ratio for the MTBE case deviates from
the case shown in Chapter 5, which had a reboiler holdup of 100 kmol. Both the
boilup ratio and reboiler holdup affect the reboiler duty, as can also be seen in
the estimation of the vapour flow rate equation (3.31). This means that a small
deviation in the output results is expected compared to the results presented in
Chapter 5.
The column in and outputs for the MTBE case are shown in Table 6.7 and Figure
6.7 show the column profiles in cyclic steady state at the end of a VFP.
Table 6.7: Column and process specifications for the MTBE case study in a
reactive cyclic distillation process.
MTBE
Catalyst load 1000 kg/stage L1 (kmol/cyc) 32.08
Liquid phase UNIQUAC D (kmol/cyc) 5.31
NC 4 xD (mol%) 98.98 mol% nBut
nBut (inert) 0.58 mol% iBut
iBut 0.01 mol% MTBE
MeOH 0.43 mol% MeOH
MTBE B (mol/cyc) 2946.7
NT 13 xB (mol%) 0.00 mol% nBut
NR 7 (3:9) 0.00 mol% iBut
P 10.95 atm 99.19 mol% MTBE
Column ∆P 0.40 atm 0.81 mol% MeOH
NF 8 & 10 BR 0.4683
F
x 100 mol% MeOH RR 6.047
(36 mol% iBut Conversion, χi 98.96 % iBut
+ 64 mol% nBut) 98.42 % MeOH
F (mol/cyc) 2970 (L) & 8205 (V) QB (MJ/cyc) 648.46
tV F P (s) 10 QC (MJ/cyc) 774.74
tLF P (s) 5
tcyc (s) 15
V 2 (kmol/cyc) 37.63
V N T (kmol/cyc) 22.40
For the MTBE case in Figure 6.7, the conversion of the reactants iBut and MeOH
is very high, as the top product is high purity inert nBut and the bottom product
is high purity product MTBE. The vapour flow rate changes significantly over the
column height, as shown in the bottom plot for the MTBE case in Figure 6.7. This
is mainly due to the introduction of a vapour feed and the molar phase transfer on
the stages. The vapour feed contains about onethird of the reactant iBut, and the
remaining content is the inert volatile component nBut. The volatile component
iBut reacts with the MeOH to form the heavy product MTBE in the MTBE reaction.
The temperature and composition profiles for the MTBE case with the rate kinet
ics in Figure 6.7 is very similar to the profiles in Figure 5.6, despite the different
reboiler holdup and boilup ratio.
100
50
0
1 2 3 4 5 6 7 8 9 10 11 12 13
450
400
350
300
1 2 3 4 5 6 7 8 9 10 11 12 13
4000
3000
2000
1000
0
1 2 3 4 5 6 7 8 9 10 11 12 13
Figure 6.7: Base case column profiles, top: liquid composition, middle: tem
perature, bottom: vapour flow rate, at the end of a VFP in cyclic steady state
for the and MTBE case.
The effect of the changes in the inputs on the performance indicators can be seen
in Figure 6.8.
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
1 1 1
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
0 0 0
tVFP Vn+1 M n-1 tVFP Vn+1 M n-1 tVFP Vn+1 M n-1
Figure 6.8: The effect of ±10 % perturbations in tV F P , Vn+1 and Mn−1 for
single stage studies on the performance indicators for the MTBE case in stage
3 (left), stage 6 (middle) and stage 9 (right).
For the first performance indicator, two things stand out: The high production rate
is due to the high holdups, capacity and catalyst load, and the effect of changes in
the inputs have on the production in stage 9. Unlike the other cases, an increase
in tV F P or Vn+1 does not increase the amount of produced MTBE, but in fact,
reduces this. The effect of Vn+1 and Mn−1 is also much more evident in this case
compared to the two previous ones, likely due to the high holdups and flow rates.
For the second performance indicator, there is not much change in stage 3 as
it is already very close to equilibrium P InII ≃ 1. For stage 9, a high change in
P InII is observed for changes in the inputs. For an increase in tV F P and Vn+1
or a decrease in Mn−1 in stage 9, it seems the equilibrium is crossed, meaning
consumption of the product will occur, for example, if tV F P is increased by 10 %.
This also explains the behaviour of P InI in the top row of plots, where the MTBE
production rate is decreased for increases in tV F P and Vn+1 or a decrease in Mn−1 .
Finally, looking at the last performance indicator, not a lot of change is seen in
stages 3 and 6, for when a change in any of the inputs is imposed. In stage 9, it is
shown that an increase in tV F P and Vn+1 or decrease in Mn−1 gives an increase
in the P InIII as well. Furthermore, the value of the Danumber is either above
or very close to 0.1 for the three stages and below 10, meaning the reaction is
controlled by the reaction kinetics [55]. For the P InIII it is further evidence that in
all cases, the lowest placed reactive stage has the highest Davalue and is the
most sensitive to changes in the inputs.
MTBE full column analysis
Stages 3, 6 and 9 are again chosen to analyse the column performance for the
MTBE case. The three performance indicators are plotted for the different changes
in the inputs, the results are shown in Figure 6.9.
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
1 1 1
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
In this MTBE case, the effects of changing the inputs are much more evident than
in the previous two cases. For the performance indicator, P InI different things can
be observed. In stage 3, a reduction in tV F P or RR or an increase in BR results in
an increase in the production rate of MTBE, while for stage 6, only a decrease in
RR or increase in BR also benefits the production rate. For stage 9, it seems that
in no case has the MTBE production rate over a cycle increased. This is mainly
due to the reaction being very close to or even over the equilibrium. Thus, a small
change in any input can easily cause the composition to get closer to and cross
the equilibrium, resulting in product consumption.
For the relative distance to equilibrium, stage 3 is already close, and there are no
significant changes here. For stage 6, the distance to the chemical equilibrium is
getting smaller for any changes in the inputs. While for stage 9, only a decrease
in RR or an increase in BR results in a reduction of distance to equilibrium at the
end of VFP. For an increase in tV F P or RR or a decrease in BR, the equilibrium
distance is greatly increased. This could be an indication that the new periodic
steady state is a low conversion steady state.
For the Damköhler number, stage 3 and 6 undergo little change, but the P InIII
significantly changes in stage 9. As in the DME case, a decrease in tV F P , RR or
an increase in BR increases the mean Damköhler number. The opposite causes
the P InIII to reduce significantly, similar to the investigations of the P InII .
For evaluating the feasibility of the MTBE process, as for the two previous cases,
the mean Damköhler and mean equilibrium constant over a cycle have been cal
culated for each selected stage. The results are shown in Table 6.8.
Table 6.8: Values of K eq,n and Dan for MTBE case in cyclic steady state.
K eq,n Dan
Stage 3 28.9 0.1
Stage 6 28.2 0.1
Stage 9 17.0 0.5
In all cases, there is a Davalue slightly above or equal to 0.1 and a K eq value
above 1, meaning the process is controlled a fast forwards reaction and a slow
backwards reaction.
Evaluating the feasibility conditions in equations (6.8) and (6.9) for the cyclic steady
state of the MTBE case gives:
∑
NT
V N T · tV F P = 22395 < L + LFn = 35050 [mol/cyc] (6.25)
n=1
These feasibility checks also shows that the MTBE case is feasible in a reactive
cyclic distillation process.
The vapour flow rates from the reboiler and stage 2, V N T and V 2 , in tables 6.2, 6.5
and 6.7 are taken as the average vapour flow over one cycle. Due to the dynamic
cyclic operation, this will change slightly over a cycle. It can also be seen that
there is a relatively high difference between the amount of vapour that leaves the
reboiler and enters the condenser. For the MTBE, there is also a vapour feed,
but for the two other cases, the difference was only due to the vaporisation on the
stages. The percentage difference of the vapour flow rates from top to bottom is
62.52% for the DME case and 20.66% for the MeOAc case, while it is 68.70% for the
MTBE case. The average vapour flow rate from the reboiler, shown in tables 6.2,
6.5 and 6.7, can be predicted from equation (3.31), as mentioned previously. With
this equation the vapour flow rate from the reboiler is calculated to be V N T = [1130,
575, 23415] mol/cycle for the DME, MeOAc and MTBE case, respectively, which is
quite close to the values shown in tables 6.2, 6.5 and 6.7.
Looking at the singlestage analysis, it seems overall in the three cases that the
vapour flow duration, tV F P , was perhaps the most important input parameter. The
higher the tV F P , the higher amount of products are produced for the DME and
MeOAc cases, the closer to equilibrium at the end of VFP the reaction is and the
higher the value of the Da is. This is as expected as an increased vapour flow
duration also means more time for reaction and separation. Especially in the DME
and MeOAc cases where the products have low boiling points compared to the
other components, it seems to be beneficial to have a long tV F P . In the MTBE
case, it was shown that a longer VFP meant the equilibrium was crossed in stage
9, and thus MTBE is consumed instead of produced. As the MTBE production is
a fast reaction, the effect of a longer VFP also have a higher effect on both reac
tion and separation. As shown in Chapter 5 the MTBE production in a reactive
cyclic distillation column did result in negative production of MTBE in the lowest
reactive stage during a VFP. However, as the overall production of MTBE in the
lowest reactive stage during a VFP is still positive and combined with the increase
in production rate for the above reactive stages compared to conventional distil
lation, it is deemed an improvement. The two phase movements, liquid holdup
and vapour flow rate, also affected the performances of the stages. In general it
seems an increase in Vn+1 or a decrease in Mn−1 leads to an increase in P InI , P InII
and P InIII , with some exceptions. As both the liquid and vapour compositions are
constants, an increase in one only means more species are present. The effect
of Vn+1 and Mn−1 on the performance indicators will therefore depend on whether
an increase or decrease in the holdup or vapour flow means a higher reaction rate
due to an increase in reactants and a decrease in the products. Furthermore, a
large Vn+1 will mean a smaller Davalue, see equations (6.5) and (6.6).
Looking at the full column analysis, it can be seen that the value of the temperature
dependent equilibrium constant does not seem to have a significant effect on the
feasibility of the reactive distillation on its own, as shown with the plots of P InII in
Figures 6.3, 6.6 and 6.9. For the DME case the value of P InII was very low, and
for the MTBE and MeOAc cases almost equal to 1. For the MeOAc and MTBE
case, there is further evidence that the reaction can be described as a chemical
equilibrium reaction [2, 77, 81, 82]. This can also be seen in Figures 6.6 and 6.9,
where the P InII is close to one in some of the stages. If the P InII is close to one,
as, in the MTBE and MeOAc cases, it could potentially be an indication that the
reaction can be described by chemical equilibrium rather than rate kinetics. The
changes introduced in tV F P , RR and BR showed that the MTBE case was most
affected with regards to P InII , especially in the bottom reactive stage 9. This is
also where both liquid and vapour feed mixes. Thus there is a high concentra
tion of reactants, which means the reaction can be highly affected by changes
in the input. The extent of reaction over a cycle, P InI , showed that the reaction
rate could differ significantly over the stages. For the DME case, the production
of DME increased higher up in the reactive zone, while for the MeOAc and MTBE
case, the product production decreased. Again the MTBE case was the most
sensitive with regards to changes in the inputs. It shows that the MTBE case is
the most unstable case when a disturbance is introduced. This is due to the short
cycle time and high column capacity and throughput, For the mean Damköhler
number, P InIII , it was seen that the value of this was very sensitive to changes
in the inputs. However, in none of the cases, the threshold values determined
by Shah et al. [55] for the controlling phenomena were crossed. This means the
phase equilibrium case still controlled the DME case, the MeOAc by the chemical
equilibrium and the MTBE case by the reaction kinetics for any introduced input
disturbances. In all of the shown cases, the Davalue and the effect of changes
in the inputs were all highest in the lowest of the reactive stage. These different
performance indicators all showed that a reactive cyclic distillation process must
be carefully designed and that the three inputs, tV F P , RR and BR are not easily
chosen. From the feasibility analysis, partly from Shah et al. [55] and Pǎtruţ et
al. [45] all the three cases, with the chosen VFP duration, boilup and reflux, were
feasible. The choice of these values used in this chapter was based on estimates,
iterations and trial and error to find appropriate values.
The work presented in this chapter shows how different reactive cyclic distillation
processes behave when a disturbance in the internal phase flows or the duration
of VFP is introduced. The tV F P parameter seems to be the most important, as
it specifies the time for both reaction and separation. It was shown that, when
determining whether or not a reactive cyclic distillation is feasible, the relative
distance to equilibrium was not that important, as long as the production rate is
still high. The mean Davalue for the stages is an important parameter that can
easily indicate whether or not the process is feasible. Due to cyclic operation, if
the Davalue shows the process is not feasible, the tV F P can be altered to ensure
a sensible reactive cyclic distillation process.
The vapour phase was assumed to be ideal for all the cases presented in this
thesis. In addition to the effect on the vapour liquid equilibrium, this assumption
also affects the vapour flow rate in equation (3.24). The temperature derivative
in equation (3.24) requires the knowledge of the vapour liquid equilibrium and
if a nonideal vapour phase is assumed, then this must be accounted for. This
requires a new derivation of the vapour flow rate as the one shown for ideal vapour
phase mixtures in the appendix.
Another significant assumption used in this thesis is the constant pure compo
nent and mixture parameters. The specific heat capacities, heat of vaporisation
and heat of reaction were all assumed to be constant. Instead of this assump
tion, temperaturedependent parameters could have been used. This would also
complicate the model additionally, and a new derivation of the vapour flow rate is
needed to account for this.
The assumption of reaction only occurring during the VFP is suitable for when a
heterogeneous solid catalyst placed on top of the stages. During the LFP, the
liquid holdup will drain to the sluice chambers below each stage as described by
Maleta et al. [13]. The reactive distillation cases presented in this thesis were
all catalysed by solid catalysts. If instead a liquid homogeneous catalyst is used,
then the reaction will most likely continue during the LFP, as the catalyst will follow
the liquid holdup. Thus, if a reactive cyclic distillation case with a liquid catalyst is
to be described, reaction during the LFP must be considered.
It was shown that while a reduction in energy demand and column size could be
achieved when going from conventional to the cyclic operation of a reactive dis
tillation process, it is not trivial to design or optimise such a column. The stage
behaviour over the vapour flow period can be highly dynamic for the reaction and
the separation parts. The performance of the stage and column are also affected
by the internal liquid and vapour flows and especially the duration of the VFP.
Pǎtruţ et al. [11] have proposed a design method for nonreactive cyclic distil
lation processes based on the method proposed by Toftegård et al. [31]. This
method can be used for the stripping and rectifying stages in a reactive cyclic dis
tillation process. However, the number of reactive stages is not that easy to find.
Alternative methods for estimating the number of reactive stages could be based
on existing conventional columns or estimating the overall production rate and the
total amount of catalyst needed and dividing it across a suitable number of stages.
There are some key challenges in designing a reactive cyclic distillation process.
For a quick overview, some of the main ones are listed below.
• Specify the duration of the cycle, vapour and liquid flow periods.
• Estimate the boilup and reflux, i.e. internal vapour and liquid flow rate.
The key design variables for reactive cyclic distillation were also discussed in
Chapter 5. As mentioned, the number of nonreactive stages can be found using
the proposed method by Pǎtruţ et al. [11] and the number of the reactive stages
can be estimated based on existing conventional column designs or the overall
production rate. The duration of the liquid flow period must be long enough for
draining each stage entirely and allow for the liquid feed and reflux to enter the
column and the products to be removed. The remaining parameters are difficult
to specify as they all have a dependency on the others. A detailed analysis of
these three parameters, the duration of the VFP, tV F P , and the boilup, BR, and
reflux ratio, RR, was shown in Chapter 6.
When a guess for the tV F P and BR have been made, the reflux ratio needs to
be determined as well. This is not easy to estimate, as it depends on the vapour
flowing into the condenser during a VFP. A method for determining this is to use the
already estimated values of tV F P and BR, simulate the process with a guess of the
reflux and iteratively adjust the reflux ratio until satisfactory product purities, and
flow rates are achieved. It is a demanding task that can require many iterations,
but by doing this, an appropriate reflux ratio for the case can be found. For any
specified boilup ratio, or vapour flow rate inside the column, a reflux ratio value
can be found that satisfy the process constraints [2, 11, 45], within the feasibility
constraints in equations (6.8)(6.9). This is because the reflux ratio in a cyclic
distillation process directly sets the stage holdup. Thus for any holdup, there is
a vapour flow rate that can provide the degree of separation necessary for the
process, provided that the cycle time is also known and constant. Figure 7.1 gives
an overview of the design method discussed in this section. It is assumed that the
pure component and mixture properties as well as thermodynamic specifications
are known. The catalyst loading has not been considered here, but it is of course
also an important design decision.
The method described here for estimating the internal flows from reboiler and
condenser, and VFP duration is tedious and can require many iterations. Due to
the development of vapour flow rate over the column and the inherently dynamic
behaviour over the VFP, the feasibility conditions in equation (6.8)(6.9) can be
difficult to evaluate. One method to better predict the changes in the vapour flow
rate over a column is to take the different components’ relative volatility and reac
tion stoichiometries into account. It may require several cycles to reach a quasi
stationary steadystate, making this design method even more tedious and slow.
Estimate
tVFP and
tLFP
Estimate New
boilup estimates of
ratio (BR) tVFP and BR
Reactive cyclic No
distillation feasible?
Yes
Estimate
reflux ratio
(RR)
Estimate
Find no. of Existing
no. of
stripping No column
reactive
stages design?
stages
Yes
Find no. of
Find feed Process
rectifying
locations simulation
stages
New estimates of
Are the results tVFP, BR and RR
No
satisfactory? Evaluate column
design
Yes
The performance of the proposed mass and energy stage balance model for four
different case studies was compared to that of a mass balance only model. By
including the energy balance, and therefore the timedependent vapour flow rate,
the proposed model showed clear benefits over a mass balanceonly model, espe
cially when there was a high development of the vapour flow rate over the column.
A model verification was presented. However, due to uncertainties in the available
data, it was not possible to conclude on a verification of the model. Despite the
inconclusiveness of the model verification, similar trends for the experimental and
model output data was shown. This indicates that a proper and complete model
verification can be made, if the necessary data is available.
The proposed reactive cyclic distillation model was also analysed for different case
studies. Here the effect of the simultaneous separation and reaction on the re
active stages was analysed, and observations regarding the column and process
The reactive cyclic distillation analysis found that the same process in a conven
tional operation can be moved to cyclic operation, which in turn allows for a re
duction in the number of stages (reactive and nonreactive), a reduction in energy
demand and an increase in reactant conversion. However, the cyclic operation
and dynamic behaviour over a vapour flow period might also cause problems. For
example, in the MTBE case shown in Chapter 5 the reaction goes from producing
MTBE to consuming it over the VFP in the bottom reactive stage. The overall
production rate of MTBE was high, also higher than the conventional case. How
ever, it is counterintuitive to have a consumption of the production in a reactive
distillation process. The conversion over a cycle for the two reactants was also
high (>98%), thus the cyclic process was deemed satisfactory. With the MTBE
case it was shown that the proposed model for reactive cyclic distillation could
handle a reaction determined either by chemical equilibrium or rate kinetics. Both
reaction types yielded similar quasistationary steadystate column profiles after a
VFP with regards to the composition and temperature. It was also shown that the
cases differed in stage behaviour over a cycle, meaning the operation and control
of a reactive cyclic distillation process depends on the reaction type.
A different analysis approach was also made for reactive cyclic distillation. For
this, the performance of a reactive cyclic distillation process was evaluated and
quantified in terms of the extent of reaction, the relative distance to equilibrium
and the Damköhler number for disturbances in the duration of the VFP, boilup,
and reflux. Three reactive cases were investigated in cyclic distillation, the pre
viously studied reactive cyclic distillation cases of dimethyl ether and methyl tert
butyl ether and the methyl acetate case. It was found that the relative distance
to chemical equilibrium did not significantly affect whether a reactive cyclic dis
tillation is feasible or not but could indicate if an assumption of equilibrium reac
tion is valid. The Damköhler number over a cycle was shown to be a significant
performance indicator and feasibility check. It can also be used to evaluate the
controlling mechanism in the reactive distillation process. The extent of reaction
was also an important performance indicator and highly dependent on the input
disturbances in the tV F P , BR and RR. The duration of the vapour flow period was
shown to have a high effect on the performance, especially in the case of MTBE,
which was already close to equilibrium. A small change in the inputs here could
cause the reaction to go from a high conversion state to a low one.
With the presented reactive cyclic distillation process analysis regarding stage
behaviour over a single VFP and the performance of the column, a design method
was proposed. It is not a shortcut design method, and it can be tedious to estimate
all the design specifications, i.e. number of stages, boilup and reflux ratio and
duration of flow periods. This method does require many iterations of the entire
column simulation before a final design can be estimated.
New models have been proposed in recent years, including the ones presented
in this thesis. A model verification with the proposed model was presented in this
thesis. However, with some degree of uncertainty in the experimental data, it was
difficult to evaluate or conclude anything. The cyclic distillation technology is still
not widely spread, making it difficult to get these experimental data. As more and
more cyclic distillation columns are being put in operation, access to experimental
data should be easier to get. However, only a few published papers with such data
are available for now and only for nonreactive distillation and stripping processes.
The presented dynamic reactive cyclic distillation model is suitable for other case
studies and even new reactive distillation processes. The model is intentionally
made general and can handle fast reactions, described by reaction equilibrium
such as the MTBE case, and slower reactions that are described by reaction rate
kinetics as shown in Chapter 6.
Another aspect, which could benefit from future studies is the control of a cyclic
[44] H. H. Chien et al. “Study of Controlled Cyclic Distillation: II. Analytical Tran
sient Solution and Asymptotic Plate Efficiencies”. In: Separation Science
1.23 (1966), pp. 281–317.
[45] C. Pǎtruţ, C. S. Bîldea, and A. A. Kiss. “Catalytic cyclic distillation A novel
process intensification approach in reactive separations”. In: Chemical En
gineering and Processing: Process Intensification 81 (2014), pp. 1–12.
[46] B. B Andersen et al. “Integrated Process Design and Control of Cyclic Dis
tillation Columns”. In: IFAC Advanced Control of Chemical Processes 51.18
(2017), pp. 542–547.
[47] R. F. Nielsen, J. K. Huusom, and J. Abildskov. “Driving Force Based Design
of Cyclic Distillation”. In: Industrial and Engineering Chemistry Research
56.38 (2017), pp. 10833–10844.
[48] G. J. Harmsen. “Reactive distillation: The frontrunner of industrial process
intensification. A full review of commercial applications, research, scaleup,
design and operation”. In: Chemical Engineering and Processing: Process
Intensification 46.9 SPEC. ISS. (2007), pp. 774–780.
[49] A. A. Kiss. “Novel Catalytic Reactive Distillation Processes for a Sustainable
Chemical Industry”. In: Topics in Catalysis 62 (2019), pp. 1132–1148.
[50] A. A. Kiss, M. Jobson, and X. Gao. “Reactive Distillation: Stepping Up to
the Next Level of Process Intensification”. In: Industrial and Engineering
Chemistry Research 58.15 (2019), pp. 5909–5918.
[51] A. Avami et al. “Shortcut design of reactive distillation columns”. In: Chem
ical Engineering Science 71 (2012), pp. 166–177.
[52] T. Keller. Reactive Distillation. Elsevier Inc., 2014, pp. 261–294.
[53] N. Nazemzadeh et al. “Graphical tools for designing intensified distillation
processes: Methods and applications”. In: Process Intensification: Design
Methodologies. Ed. by F. I. GómezCastro and J. G. SegoviaHernández.
Berlin, DE: De Gruyter, 2019. Chap. 6, pp. 145–179.
[54] H. Subawalla and J. R. Fair. “Design guidelines for solidcatalyzed reac
tive distillation systems”. In: Industrial and Engineering Chemistry Research
38.10 (1999), pp. 3696–3709.
[55] M. Shah et al. “A systematic framework for the feasibility and technical eval
uation of reactive distillation processes”. In: Chemical Engineering and Pro
cessing: Process Intensification 60 (2012), pp. 55–64.
[56] R. Muthia et al. “Novel method for mapping the applicability of reactive dis
tillation”. In: Chemical Engineering and Processing: Process Intensification
128 (2018), pp. 263–275.
[57] R. Muthia, M. Jobson, and A.A. Kiss. “A systematic framework for assess
ing the applicability of reactive distillation for quaternary mixtures using a
mapping method”. In: Computers and Chemical Engineering 136 (2020),
p. 106804.
[58] S. Skogestad. “Control structure design for complete chemical plants”. In:
Computers and Chemical Engineering 28.12 (2004), pp. 219–234.
[59] B. V. Maleta et al. “Pilotscale studies of process intensification by cyclic
distillation”. In: AIChE Journal 61.8 (2015), pp. 2581–2591.
[60] V. N. Maleta et al. “Pilotscale experimental studies on ethanol purification
by cyclic stripping”. In: AIChE Journal 65.9 (2019), pp. 1–7.
[61] A.em A. R. Nielsen et al. “Analysis and Evaluation of Periodic Separations
Using COPS Trays”. In: Chemical Engineering Transactions 69 (2018), p. 6.
[62] J.C. Wang and G.E. Henke. “Tridiagonal Matrix for Distillation”. In: Hidro
carbon Processing 45.8 (1966), pp. 155–163.
[63] S. Skogestad. “Dynamics and control of distillation columns A critical sur
vey”. In: Modeling, Identification and Control 18.3 (1997), pp. 177–217.
[64] I. Cameron and R. Gani. Product and Process Modelling. 2011.
[65] T. Bisgaard, J. K. Huusom, and J. Abildskov. “Modeling and analysis of con
ventional and heatintegrated distillation columns”. In: AIChE Journal 61.12
(2015), pp. 4251–4263.
[66] R. Gani, C. A. Ruiz, and I. T. Cameron. “A generalized model for distillation
columnsI. Model description and applications”. In: Computers and Chemi
cal Engineering 10.3 (1986), pp. 181–198.
[67] L. T. Biegler, J. J. Damiano, and G. E. Blau. “Nonlinear parameter estima
tion: A case study comparison”. In: AIChE Journal 32.1 (1986), pp. 29–45.
[68] R. Taylor and R. Krishna. “Modelling reactive distillation”. In: Chemical En
gineering Science 55.22 (2000), pp. 5183–5229.
[69] T. Bisgaard et al. “Adding Value to Bioethanol through a Purification Process
Revamp”. In: Industrial and Engineering Chemistry Research 56 (2017),
pp. 5692–5704.
[70] R. Jacobs and R. Krishna. “Multiple Solutions in Reactive Distillation for
Methyl tertButyl Ether Synthesis”. In: Industrial and Engineering Chemistry
Research 32.8 (1993), pp. 1706–1709.
[71] S. Hauan, T. Hertzberg, and K. M. Lien. “Multiplicity in reactive distillation of
MTBE”. In: Computers and Chemical Engineering 21.10 (1997), pp. 1117–
1124.
dMn,i
= Vn+1 yn+1,i − Vn yn,i + Vn+1
F F
yn+1,i
dt
dMn
= Vn+1 − Vn + Vn+1 F
dt
∑
NC
( V )
Hn = yn,i CP,i (Tn − Tref )
i=1
∑
NC
( V )
hn = xn,i CP,i (Tn − Tref ) − ∆Hvap,i
i=1
Mn,i
xn,i =
Mn
dxn,i dMn,i/dt Mn n,i/dt − Mn,i n/dt
dM dM
= dM =
dt n/dt Mn2
The term dMn/dt is known from the overall molar balance for stage n:
dMn ( )
hn = hn Vn+1 − Vn + Vn+1
F
dt
The other term, dhn/dt, can be found from the liquid enthalpy definition.
[ NC ]
dhn d ∑ ( V )
Mn = Mn xn,i CP,i (Tn − Tref ) − ∆Hvap,i
dt dt i=1
[ NC ] [ NC ]
dhn ∑ d ∑ ( )
= Mn V
xn,i CP,i Tn − Mn V
xn,i CP,i Tref + ∆Hvap,i
dt i=1 dt i=1
∑
NC
dTn
V
+ Mn xn,i CP,i
i=1
dt
N C ((
∑ ) )
Mn dMn,i/dt − Mn,i dMn/dt ( V )
− Mn 2
CP,i Tref + ∆Hvap,i
i=1
M n
The total and component molar balances can now be inserted for dMn/dt and
dMn,i/dt. As this will become quite extensive, the terms on the right hand side
N C ((
∑ ) )
Mn dMn,i/dt − Mn,i dMn/dt ( V
)
Mn CP,i Tref + ∆Hvap,i =
i=1
Mn2
( )
∑
NC V
CP,i Tref + ∆Hvap,i
Mn Vn+1 (Mn yn+1,i − Mn,i )
i=1
Mn2
( )
∑
NC V
CP,i Tref + ∆Hvap,i
− Mn Vn 2
(Mn yn,i − Mn,i )
i=1
M n
( )
∑
NC
C V
T
P,i ref + ∆Hvap,i ( )
F
+ Mn Vn+1 2
F
Mn yn+1,i − Mn,i
i=1
Mn
The derived terms for hn dMn/dt and Mn dhn/dt can be inserted in the stage energy
balance.
dhn dMn
Mn + hn = Vn+1 Hn+1 − Vn Hn + Vn+1
F F
Hn+1 + Qn
dt dt
[ ( )
Tn ∑
NC
Mn Vn+1 (Mn yn+1,i − Mn,i )
Mn2 i=1
( )
Tn ∑
NC
−Mn Vn (Mn yn,i − Mn,i )
Mn2 i=1
( )]
Tn ∑ ( )
NC
F
+Mn Vn+1 Mn yn+1,i − Mn,i
F
Mn2 i=1
∑
NC
dTn
V
+ Mn xn,i CP,i
i=1
dt
[ ( )
∑
NC V
CP,i Tref + ∆Hvap,i
− Mn Vn+1 (Mn yn+1,i − Mn,i )
i=1
Mn2
( )
∑NC V
CP,i Tref + ∆Hvap,i
−Mn Vn 2
(Mn yn,i − Mn,i )
i=1
M n
( )]
∑
NC
C V
P,i Tref + ∆Hvap,i ( )
F
+Mn Vn+1 2
F
Mn yn+1,i − Mn,i
i=1
M n
( )
+ hn Vn+1 − Vn + Vn+1 F
F
From this all the terms containing Vn+1 , Vn and Vn+1 can be found and isolated on
the right hand side.
[ ( ( )
Tn ∑
NC
0 = Vn+1 Hn+1 − hn + Mn − (Mn yn+1,i − Mn,i )
Mn2 i=1
)]
∑
NC V
CP,i Tref + ∆Hvap,i
+ 2
(Mn yn+1,i − Mn,i )
i=1
M n
∑
NC
dTn
− Mn V
xn,i CP,i + Qn
i=1
dt
From this the helping function g, see equation (3.25), can be inserted.
∑
NC
dTn
0 = Vn+1 g(n, n + 1)) − Vn g(n, n) + F
Vn+1 g F (n, n + 1) − Mn V
xn,i CP,i + Qn
i=1
dt
From this an expression for the vapour flow rate can easily be found.
[ ]
∑
NC
dT n 1
Vn = Vn+1 g(n, n + 1)) − +Vn+1
F
g F (n, n + 1) − Mn V
xn,i CP,i + Qn
i=1
dt g(n, n)
A similar derivation for the reboiler vapour flow rate can also be made, where
Vn+1 = 0.
In this term there is still the differential stage temperature with regards to the time,
dTn/dt. An expression for this can be derived from the vapour liquid equilibrium
Where Kn,i is the equilibrium coefficient. For the work presented in this thesis
it is assumed that the VLE can be described by ideal vapour and an ideal liquid
or a nonideal liquid described by an activity coefficient model. This function f
depends on the molar holdup and temperature on stage n. The derivative of f
with regards to the time is found:
∂f dTn ∑ ∂f dMn,i
NC
df
= + =0
dt ∂Tn dt i=1
∂M n,i dt
The component molar balance gives the dMn,i/dt. The partial derivatives must be
found from knowledge of the vapour liquid equilibrium. First the partial derivative
of f with regards to Mn,i is found.
∂f ∑
NC
∂xn,j ∂Kn,j
=− Kn,j + xn,j
∂Mn,i j=1
∂Mn,i ∂Mn,i
∂f ∑
NC
∂xn,j ∂ ln Kn,j
=− Kn,j + xn,j Kn,j
∂Mn,i j=1
∂Mn,i ∂Mn,i
The pressure and saturated pressure are independent of the molar holdup.
∑
NC ( )
∂f ∂xn,j ∂ ln γn,j
=− Kn,j + xn,j
∂Mn,i j=1
∂Mn,i ∂Mn,i
∂xn,j 1
= (1 − xn,j ) if i = j
∂Mn,i Mn
∂xn,j 1
=− xn,j if i ̸= j
∂Mn,i Mn
∂f ∑ NC sat
dPn,i sat ∂γn,i
= −Pn xn,i γn,i + xn,i Pn,i
∂Tn i=1
dTn ∂Tn
∑
NC ( sat )
dPn,i sat ∂ ln γn,i
= −Pn xn,i γn,i + Pn,i
i=1
dTn ∂Tn
dMn,i rt
= [. . .] + Rn,i
dt
dMn (t)hn ∑
NC
reac rt
= [. . .] + ∆Hn Rn,i
dt i=1
1 [ F
Vn = Vn+1 g(n, n + 1) + Vn+1 g F g(n, n + 1) + g reac (n)
g(n, n)
]
∑
NC
dT n
− V
xn,i CP,i + ∆Hnreac + Qn
i=1
dt
∑
NC
g reac
(n) = −hn rt
Rn,i
i=1
[ ( )
Tn ∑
N C ∑
NC
+ Mn − 2 V
CP,i rt
Mn Rn,i − Mn,i rt
Rn,i
Mn i=1 i=1
( )]
∑
NC V
CP,i Tref+ ∆Hvap,i ∑
NC
+ rt
Mn Rn,i − Mn,i rt
Rn,i
i=1
Mn2 i=1
www.kt.dtu.dk