Cyclic Distillation Technology: Rasmussen, Jess Bjørn

Downloaded from orbit.dtu.
dk on: Aug 22, 2022
Cyclic Distillation Technology
Rasmussen, Jess Bjørn
Publication date:
2021
Document Version
Publisher's PDF, also known as Version of record
Link back to DTU Orbit
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

循环精馏技术
Jess Bjørn Rasmussen PhD Thesis
DTU Chemical Engineering
Department of Chemical and Biochemical Engineering

循环精馏技术
Jess Bjørn Rasmussen
PhD Thesis
November, 2021
Department of Chemical and Biochemical Engineering, Technical

University of Denmark (DTU)
SinoDanish Center for education and research (SDC)
Institute of Process Engineering, Chinese Academy of Sciences

(CAS)
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
Copyright: Reproduction of this publication in whole or in part must include

the customary bibliographic citation, including author attribution,
report title, etc.
Cover art: StudioXinyi, 2021
Published by: DTU, Department of Chemical and Biochemical Engineering,
Søltofts Plads, Builiding 228A, 2800 Kgs. Lyngby Denmark
www.kt.dtu.dk
Preface
This thesis was prepared at Process and Systems Engineering Centre (PROSYS),
DTU Chemical and Biochemical Engineering, in fulfilment of the requirements for
acquiring a PhD in Chemical and Biochemical Engineering from December 2018
to November 2021.
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.
Jess Bjørn Rasmussen,

København Ø
Signature
November, 2021
Date
Cyclic Distillation Technology iii

iv Cyclic Distillation Technology
Acknowledgements
I owe a big thanks to my family; my brothers Jon & Theis, their wives and children
and my parents.
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.
Cyclic Distillation Technology v

»Jeg’ fokuseret, holder øjnene på målstregen«
Pede B, Fokuseret
vi Cyclic Distillation Technology

Abstract
Distillation processes are one the most widespread separation technologies that
are used in the chemical and biochemical industry. It is a simple, well known
and effective method for separating various mixtures. It is also a high energy
demanding technology with a high environmental impact as the separation mix
tures are boiled in the bottom of the column and condensed in the top. In an
attempt to intensify the process, different alternatives have been proposed. One
of which is the cyclic distillation, where the phase movements inside the column
are separated, giving a high separation efficiency and thus allowing for a reduction
in energy demand, number of stages or an increase in throughput and conversion
for reactive distillation processes.
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
Cyclic Distillation Technology vii

cess. The distance to equilibrium could indicate whether an assumption of chem
ical equilibrium is valid or not. Based on existing feasibility conditions for reactive
and cyclic distillation, some helpful observations are made that could facilitate part
of a design method. However, this design method still requires iterations to find
some of the important specifications as currently available methods.
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.
viii Cyclic Distillation Technology

摘要
精馏作为一种简单高效分离混合物的方法，是典型的化工及生物过程分离技术。
精馏通过塔底再沸及塔顶冷凝的方式分离混合物，因此过程能耗较高，易对环境
造成一定影响，可通过多种过程强化手段降低能耗。循环精馏技术通过分离操作
塔内的不同相态，提高分离效率，从而降低能耗与塔板数，提升反应精馏过程转
化率。
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。
Cyclic Distillation Technology ix

x Cyclic Distillation Technology
Resumé
Destilleringsprocesser er en af de mest udbredte separationsteknologier, der bliver
anvendt i kemisk og biokemisk produktion. Det er en simpel, velkendt og ef
fektiv metode til at separere forskellige blandinger. Det er også en meget en
ergikrævende teknologi, med høj påvirkning på miljøet da separationsblandingerne
bliver kogt i bunden af kolonnen og kondenseret i toppen. I et forsøg på at intensi
vere denne proces er forskellige alternativer blevet fremlagt. Et af disse er cyklisk
destillering, hvor fasebevægelserne inde i selve kolonnen er adskilt. Det giver en
høj separationseffektivitet og tillader derved en reduktion i energikravet, antallet af
bunde eller en forøgelse i gennemstrømningen og omdannelsen for reaktiv cyklisk
destillering.
I denne afhandling er cyklisk destilleringsteknologi blevet studeret og analyseret

for at bedre forstå processen. Med en højere procesforståelse er det nemmere at
foreslå en cyklisk destillering som et alternativ til konventionel destillering.
En masse og energibalance bundmodel er blevet foreslået her. Det er en høj nø

jagtighedsmodel, der tager hensyn til tidsafhængige temperaturer og dampflows
og endvidere tillader flere fødestrømsplaceringer. Modellens præstation er eval
ueret og sammenlignet med tidligere modeller, hvor det er blevet vist at hvis der
en høj udvikling i bundtemperaturen eller dampflowet, så vil den præsenterede
model være et godt valg til at beskrive disse fænomener.
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.
Opførelsen af tre forskellige reaktive cykliske destilleringsprocesser er yderligere

blevet evalueret for når en forstyrrelse i inputtene er blevet introduceret. Yder
mere, er muliggørelsen af en reaktiv cyklisk destillering blevet diskuteret. Tre
opførelsesparametre til at analysere processen er foreslået: graden af reaktion
over en dampflowperiode, den relative afstand til kemisk ligevægt efter en dampflow
periode og det gennemsnitlige Damköhler tal over en dampflowperiode. Af disse
Cyclic Distillation Technology xi

parametre, er især graden af reaktion og Damköhler tallet vigtige værktøjer når
opførelsen og mulighedgørelsen af en reaktiv cyklisk destilleringsprocess skal un
dersøges. Afstanden til kemisk ligevægt kan hjælpe med at vurdere om hvorvidt
en reaktion kan antages at opføre sig som en ligevægtsreaktion eller ej. Baseret
på eksisterende kriterier for hvorvidt en reaktiv destillering kan lade sig gøre er
hjælpsomme observeringer, der er præsenteret som en del af en designmetode.
Men, denne designmetode kræver stadigvæk mange iterationer for at finde nogle
af de vigtige specifikationer.
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.
xii Cyclic Distillation Technology

Nomenclature
Parameters/Variables: Greek symbols:
ai activity of component i αi,j binary interaction parameter for compo
nents i and j
ai,j , bi,j binary interaction parameters for compo γi activity coefficient of component i
nents i and j
B bottoms [mol/cyc] ϵn,i difference in composition from equilib
rium for component i on stage n
BR boilup ratio νi stoichiometric coefficient
V
Cp,i specific heat capacity of component i χi conversion of component i
[J/(K mol)]
D distillate [mol/cycle] Subscripts:
Da Damköhler number i or j component ID
g helper function for vapour flow expres n stage number
sion [J/mol]
hn liquid enthalpy of stage n [J/mol] Superscripts:
Hn vapour enthalpy of stage n [J/mol] 0 initial value
∆H reac reaction enthalpy [J/mol] F feed
∆Hivap vaporisation enthalpy of component i [J/ LF liquid feed
mol]
Kn,i phase equilibrium constant of compo (LF P ) end of liquid flow period
nent i on stage n
Keq chemical equilibrium constant (L) liquid phase
kf forward rate constant SD side draw
L reflux [mol/cycle] VF vapour feed
Mn total liquid holdup on stage n (V F P ) end of vapour flow period
Mn,i liquid holdup of component i on stage n (V ) vapour phase
mcat
n catalyst loading on stage n Abbreviations:
N number of cycles DME dimethyl ether
NC number of components L liquid phase
NF number of feed locations LFP liquid flow period
NR number of reactive stage MeOAc methyl acetate
NT number of stages MESH mass, equilibrium, summation and heat
equations
Pn stage pressure [Pa] MTBE methyl tertbutyl ether
∆P column pressure drop [Pa] V vapour phase
sat
Pn,i saturation pressure of component i on VFP vapour flow period
stage n
P InI performance indicator I VLE vapour liquid equilibrium
P InII performance indicator II
P InIII performance indicator III
Qn external energy to stage n [J/s]
QB energy to reboiler [J/s]
QC energy to condenser [J/s]
qacid ionexchange capacity of catalyst (equiv.
(H+ /kg)
RR reflux ratio
rn,i reaction rate [mol/s]
Rn,i reaction rate [mol/s]
Sn side draw from stage n [mol/cycle]
Tn stage temperature [K]
Tref reference temperature [K]
t time [s]
tcyc cycle time [s]
tLF P liquid flow period duration [s]
tV F P vapour flow period duration [s]
Vn vapour flow rate from stage n [mol/s]
xn,i liquid composition of component i on
stage n
yn,i vapour composition of component i on
stage n
Cyclic Distillation Technology xiii

xiv Cyclic Distillation Technology
Contents
Preface iii
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
2 Cyclic distillation technology from the 20th to the 21st century 11

2.1 Progress from the early 20th century to the 21st century . . . . . . 12
2.2 State of the art in cyclic distillation technology . . . . . . . . . . . 15
2.2.1 Cyclic separation tray and operation design . . . . . . . . . 15
2.2.2 Modelling and designing of a cyclic separation process . . 17
2.2.3 Modelling of reactive cyclic distillation . . . . . . . . . . . . 22
2.2.4 Control of a cyclic distillation process . . . . . . . . . . . . 25
2.2.5 Experimental work . . . . . . . . . . . . . . . . . . . . . . . 26
3 Mass and energy balance stage models for cyclic distillation 29

3.1 Cyclic distillation model development . . . . . . . . . . . . . . . . 30
3.1.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.2 Balance equations . . . . . . . . . . . . . . . . . . . . . . . 31
Cyclic Distillation Technology xv

CONTENTS
3.1.3 Constitutive equations . . . . . . . . . . . . . . . . . . . . . 34

3.2 Model discussion and analysis . . . . . . . . . . . . . . . . . . . . 38
3.2.1 Applications of the model to cyclic stripping processes . . . 38
3.2.2 Liquid and vapour feed . . . . . . . . . . . . . . . . . . . . 38
3.2.3 Stage temperature deviations . . . . . . . . . . . . . . . . . 39
3.2.4 Model implementation, framework and analysis . . . . . . 39
3.2.5 Model variations . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 Mass and energy balance stage model for reactive cyclic distillation 43
3.4 Chapter conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4 Cyclic distillation case studies 49

4.1 Case studies for comparing proposed stage model to previous stage
models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.1 Dynamic open loop disturbance response . . . . . . . . . . 54
4.2 Model verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5 Analysis of reactive cyclic distillation stage behaviour for the methyl

tertbutyl ether production 65
5.1 Production of methyl tertbutyl ether . . . . . . . . . . . . . . . . . 66
5.2 Comparison of MTBE reactive distillation with conventional and cyclic
operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.2.1 Conventional reactive distillation model . . . . . . . . . . . 69
5.2.2 Equivalent cyclic column design . . . . . . . . . . . . . . . 70
5.2.3 Analysis of the separation and reaction behaviour on se
lected stages . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3 Analysis of the effect of key design variables . . . . . . . . . . . . 80
5.3.1 Improved cyclic operated column design . . . . . . . . . . . 82
6 Quantification of a reactive cyclic distillation performance and feasi

bility 87
6.1 Quantitative metrics for analysing reactive cyclic distillation . . . . 88
6.1.1 Performance indicators . . . . . . . . . . . . . . . . . . . . 88
6.1.2 Full column feasibility analysis . . . . . . . . . . . . . . . . 90
6.2 Case descriptions and results . . . . . . . . . . . . . . . . . . . . 91
6.2.1 Dimethyl ether production . . . . . . . . . . . . . . . . . . . 92
6.2.2 Methyl acetate production . . . . . . . . . . . . . . . . . . . 98
6.2.3 Methyl tertbutyl ether production . . . . . . . . . . . . . . . 104
xvi Cyclic Distillation Technology

CONTENTS
6.3 General trends and observations for the stage and column analysis 110
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
8 Thesis conclusion 121
9 Future perspectives 125
Bibliography 127
A Vapour flow rate derivation 133

A.1 Vapour flow rate derivation for reactive cyclic distillation . . . . . . 138
Cyclic Distillation Technology xvii

CONTENTS
xviii Cyclic Distillation Technology

1 Introduction
In this chapter, the scope of the PhD project is presented. A brief background and
motivation for the work on cyclic distillation technology presented in this thesis is
given. This leads to an overview of the contribution of the thesis. At the end of
this chapter, the outline of the thesis is shown, and a list of the scientific contribu
tions in the form of conference presentations and journal publications, which were
disseminated during the PhD project, is given. This chapter is partially based on
the published articles: ”A mass and energy balance stage model for cyclic distilla
tion”, AIChE Journal 66 (2020) e16259 [1] and ”Analysing separation and reaction
stage performance in a reactive cyclic distillation process”, Chem. Eng. Process
167 (2021) 108515 [2].
Cyclic Distillation Technology 1

1.1. BACKGROUND
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].
In short, a distillation process is a separation technique that employs the difference

in boiling points and the relative volatility of two or more components to separate a
feed stream in a low boiling product, removed in the top of the distillation column,
and a high boiling product, removed in the bottom. Midboiling products can be
removed from side draws. Most distillation columns are adiabatic, with energy
only supplied to the reboiler at a high temperature and removed in the condenser
at low temperature. As the mixture’s temperature needs to be at the boiling point
throughout the column, high energy consumption is often required.
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.
A stripping column is similar to a simplified distillation process, and, as the name

suggests, it acts like the stripping part of a conventional distillation column. Here a
vapour or gas, e.g. air or saturated steam, is fed to the bottom of the column, and
a liquid containing light or volatile component that is to be removed is added to
the top of the column. No reboiler or condenser is used in a stripping column, and
thus the temperature in a stripping column is not necessarily at the boiling point of
the mixture. Stripping processes differ from distillation by the driving force of the
separation. The driving force in stripping is the solubility, wherein the distillation it
is the vapour liquid equilibrium (VLE).
2 Cyclic Distillation Technology

CHAPTER 1. INTRODUCTION
A wellknown distillation process intensification is reactive distillation. Here, a re

active mixture of one or more reactants is introduced to the distillation column, so
reaction and separation occur simultaneously. This can have the advantage of
reducing the number of unit operations and reducing overall energy consumption
as multiple unit operations are combined in one. The production of methyl acetate
in a reactive distillation column is an excellent example of the advantages in com
bining reaction and separation as a single reactive distillation column can replace
a reactor, and multiple distillation columns [5, 6].
Schematics showing examples of the conventional distillation process (a), strip

ping process (b) and reactive distillation process (c) are presented in Figure 1.1.
A single liquid feed is assumed for the conventional distillation, while for the re
active distillation, both a liquid and a vapour feed are assumed. The reactive
distillation is an example of a hypothetical reaction, A + B ↔ C + D, with solid
heterogeneous catalyst on each stage in the reactive zone. Here the product C is
assumed to have the lowest boiling point and D the highest, while the reactants,
A and B, have intermediate boiling points. No azeotropes are assumed in either
of the processes shown in Figure 1.1.
Feed liquid
Vapour with
1 A,B 1
A
1
High purity A High purity C
Rectifying Rectifying
...
...
section section
A
...
...
...
...
...
...
Liquid feed
Reactive zone
...
...
...
A,B Feed location
B
...
...
...
...
...
...
Vapour feed
Stripping Stripping
...
...
section NT section
High purity B High purity D

NT NT
Vapour Stripped liquid B,
traces of A
(a) (b) (c)
Conventional Stripping Reactive distillation
distillation of A from B A+B↔C+D
of A and B
Figure 1.1: (a) Conventional distillation process of components A and B. (b)

Stripping process of A from B. (c) Reactive distillation process for the reaction
A + B ↔ C + D. Total number of stages is denoted N T .
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

1.1. BACKGROUND
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.
The United Nations has released a list of 17 Sustainable Development Goals

(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 12: Responsible consumption and production. This is related to estab

lishing and ensuring sustainable production in all industries. For a distillation
process, this translates to making the separation more green, reducing en
ergy consumption and reducing waste such as solvents.
• 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

1.3. CONTRIBUTION
efficiency and thus lower the energy consumption, increase throughput or reduce
the number of stages in a column [10].
Cyclic, or periodic, distillation is a promising emerging process intensification. In

short, the working principle behind the technology is a separation of phase move
ments, i.e. liquid and vapour flow, By doing this cyclic separation of phase move
ments instead of the conventional operation and avoiding liquid backmixing, it
has been shown that higher capacity, a better degree of separation or reduced
energy requirements depending on the process, column, and stage design can
be achieved [10–12]. During the mid 20th century, there was some interest in
cyclic distillation. However, the interest dwindled during the 19801990s. In 2011
a new tray design for cyclic distillation was proposed by Maleta et al. [13]. This
new tray design rekindled the interest in cyclic distillation, and since then, new
studies on the topic have been published. The tray design developed by Maleta
et al. [13] works by simultaneously draining the trays. However, in 2016 an al
ternative draining method of trays was proposed by Toftegård et al. [14]. This
tray operates with sequential draining, i.e. the trays are drained one by one start
ing from the bottom. Currently, the vast majority of modelling, experimental and
design studies are focused on a simultaneous draining method.
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.

1.4. THESIS STRUCTURE
1.4 Thesis structure

The thesis is divided into the following chapters, for each of which a small sum
mary is given:
• 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.
• Chapter 5: Analysis of reactive cyclic distillation stage behaviour for

the methyl tertbutyl ether production: With the reactive cyclic distilla
tion model presented in Chapter 3, the MTBE reactive distillation is inves
tigated, and compared to conventional process performance and the stage
behaviour is analysed. This is an example of a fast reaction. Hence either
an equilibrium reaction or a rate kinetic reaction can be assumed.
• Chapter 6: Quantification of a reactive cyclic distillation performance

and feasibility: Three reactive distillation cases are investigated and anal
ysed with regards to the stage and column response for changes in duration
of the vapour flow period, vapour flow rate and liquid holdup. The behaviour

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.
• Chapter 7: Discussion, Chapter 8: Conclusion and Chapter 9: Future

Works: A discussion of the model limitations for the presented model in
Chapter 3 is given, as well as a discussion of reactive cyclic distillation pro
cess design. Followed by the thesis conclusion and considerations for future
perspectives for the advancement in cyclic distillation technology.
• Appendix: The mathematical derivation of the vapour flow rate, which is

used in the mass and energy balance stage model in Chapter 3, is shown.
1.5 Publications and conference contributions

The journal publications and conference contributions that have been a part of
this PhD project are listed below.
1.5.1 Journal publications

• J.B. Rasmussen, S.S. Mansouri, X. Zhang, J. Abildskov, J.K. Huusom, A
mass and energy balance stage model for cyclic distillation, AIChE Journal
66 (2020) e16259.
• J.B. Rasmussen, S.S. Mansouri, X. Zhang, J. Abildskov, J.K. Huusom, Analysing

separation and reaction stage performance in a reactive cyclic distillation
process, Chem. Eng. Process. 167 (2021) 108515.
• J.B. Rasmussen, M. Stevnsborg, S.S. Mansouri, X. Zhang, J. Abildskov,

J.K. Huusom, Quantitative metrics for evaluating reactive cyclic distillation
performance, submitted to Chem. Eng. Process. (30112021).
1.5.2 Conference contributions

• J.B. Rasmussen, S.S. Mansouri, J. Abildskov, J.K. Huusom, Modelling of a
Cyclic Distillation Process, 22nd Nordic Process Control Workshop Kgs.
Lyngby, Denmark, 2019.
• J.B. Rasmussen, S.S. Mansouri, J. Abildskov, J.K. Huusom, Comparison of

sequential and simultaneous draining in periodic cycled separation columns,
12th European Congress of Chemical Engineering Firenze, Italy, 2019.
• J.B. Rasmussen, S.S. Mansouri, X. Zhang, J. Abildskov, J.K. Huusom, Bench

1.5. PUBLICATIONS AND CONFERENCE CONTRIBUTIONS
marking reactive distillation for MTBE using classical and cyclic operation,
Computer Aided Process Engineering FORUM 2020 Kgs. Lyngby, Den
mark 2020.
• J.B. Rasmussen, R.F. Nielsen, S.S. Mansouri, X. Zhang, J. Abildskov, J.K.

Huusom, Reactive cyclic distillation, International Congress of Chemical and
Process Engineering CHISA Prague, Czech Republic, 2021. (virtual)
• J.B. Rasmussen, T.S. Hansen, O. Bedryk, J. Abildskov, J.K. Huusom, Model

verification of cyclic separation processes, 13th European Congress of Chem
ical Engineering Berlin, Germany, 2021. (virtual)
• J.B. Rasmussen, M. Stevnsborg, S.S. Mansouri, X. Zhang, J. Abildskov,

J.K. Huusom, Moving from conventional to cyclic operation of reactive distil
lation processes, 13th European Congress of Chemical Engineering Berlin,
Germany, 2021. (virtual)

2 Cyclic distillation technology
from the 20th to the 21st century
This chapter gives a historical overview of the developments in cyclic distillation
technology from the 1930s to present. The theoretical, experimental and mod
elling work and challenges that have been done in the field of cyclic separation are
briefly discussed. With the recent increased interest in cyclic separation and ad
vances in equipment and computational power, new models and operation meth
ods have been proposed in the last few decades. An overview of the current
state of the art is presented. The recent work in the operation and tray design
of cyclic distillations, the modelling of nonreactive and reactive cyclic distillation
and the currently available experimental work is discussed. This chapter is par
tially based on the published articles: ”A mass and energy balance stage model
for cyclic distillation”, AIChE Journal 66 (2020) e16259 [1] and ”Analysing sep
aration and reaction stage performance in a reactive cyclic distillation process”,
Chem. Eng. Process 167 (2021) 108515 [2] and the manuscript ”Quantitative
metrics for evaluating reactive cyclic distillation performance” submitted to Chem.
Eng. Process. (30112021).

2.1. PROGRESS FROM THE EARLY 20TH CENTURY TO THE 21ST
CENTURY
2.1 Progress from the early 20th century to the

21st century
Despite the low efficiency in a distillation process, it is still widely used in industry.
It is a process that can handle a high process throughput and applies to many dif
ferent cases. The low separation efficiency in distillation columns has long been
a known issue. In 1936 Lewis [15] published a paper on the theoretical tray effi
ciencies for three different tray concepts assuming plug flow in the liquid phase,
i.e. ideal liquid flow behaviour. Ideal liquid flow is in itself difficult to achieve in a
conventional distillation. With these case studies, Lewis investigated whether the
distillation stage design could be improved upon. He established three different
tray scenarios:
• Case I: Inlet vapour is completely mixed.
• 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.
Case I Case II Case III

Figure 2.1: In and outlets of the three Lewis cases. Blue arrows indicates
the liquid flow and red arrows the vapour flow.

CHAPTER 2. CYCLIC DISTILLATION TECHNOLOGY FROM THE 20TH TO
THE 21ST CENTURY
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

2.1. PROGRESS FROM THE EARLY 20TH CENTURY TO THE 21ST
CENTURY
Case II can be directly substituted with a timedependent liquid composition as

in cyclic distillation assuming mixed liquid holdup. A tray in a cyclic operated
distillation process, therefore, behave as a Lewis Case II and have the highest
theoretical tray efficiency of the three cases investigated by Lewis [15].
Both the backmixing and thermodynamic limitations for distillation processes,

which was mentioned in the previous chapter, are dealt with in cyclic distillation to
give a high separation efficiency. By having a separated phase flow movement no
backmixing of the liquid holdup can occur. As shown by Lewis and later Robin
son and Engel [15, 19], the degree of separation of a cyclic distillation tray is much
higher than for a conventional distillation tray. Thus, the energy requirement can
be lowered, which in turn increases the thermodynamic efficiency.
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.

THE 21ST CENTURY
2.2 State of the art in cyclic distillation technology

The following section outlines the current state of the art in cyclic distillation mod
elling, column and tray design, operation, and experimental work.
2.2.1 Cyclic separation tray and operation design

As mentioned, Maleta et al. proposed a new tray design for cyclic separation [13]
with a sluice chamber beneath each tray. During the VFP, the vapour flow closes
the stage and prevents liquid from draining to the sluice chamber with the help of
pistons. After the VFP, the vapour flow is interrupted at the reboiler outlet to the
column, and the pistons in the tray move down to allow the liquid holdup draining
to the sluice chamber and closing the sluice. This way, the liquid is moved to
the sluice chamber without draining to the stage below. At the beginning of the
next VFP, the vapour flow moves the pistons up and opens the sluice chambers
so that the liquid holdups are moved to the stage below. With this tray design
that allows fullscale distillation columns to be operated in cyclic mode, together
with increased computational power, the interest in cyclic distillation has grown.
This method, using the Maleta trays, allows cyclic operation while avoiding the
backmixing of the liquid. However, with the vapour supply from the reboiler being
turned off and on again, there is a possibility of a significant pressure buildup and
sudden pressure release in the reboiler.
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.
Draining method Tray type Vapour flow Opening/ closing

mechanism
Simultaneous Maleta [13] During VFP Pistons
Sequential COPS [14] During VFP and LFP Pneumatics

2.2. STATE OF THE ART IN CYCLIC DISTILLATION TECHNOLOGY
Maleta
(a) (b) (c)

COPS
(d) (e) (f)

Figure 2.3: Maleta trays (top) and COPS trays (bottom) during a cyclic sepa
ration process. For the Maleta trays; (a): normal VFP, (b): normal LFP where
the liquid is drained to sluice chamber, (c): new VFP is initiated with initial
draining of liquid from sluice chamber. For the COPS trays; (d): normal VFP,
(e): LFP is started by opening and draining the bottom stage, (f): the bottom
stage is closed, and the stage above is drained. Blue arrows indicate liquid
flow, and red arrows indicate vapour flow.
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.
Wankat published a description of a continuous cyclic distillation for binary solvent

exchange [41]. This process is a threestage distillation column with a reboiler and
condenser. It works similar to the COPS tray concept with sequential draining of

THE 21ST CENTURY
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.
2.2.2 Modelling and designing of a cyclic separation process

With increased computational power since the early studies in cyclic distillation
technology, the complexity of the models has also increased. The first mass bal
ance based models from the 20th century described stripping processes with as
sumptions of linear equilibrium, binary mixtures and saturated liquid feed [12, 20,
44]. The newer models are capable of simulating full distillation columns with
nonlinear equilibrium and multicomponent mixtures.

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:
VFP molar balance:
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
LFP molar balance:
(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 21ST CENTURY
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

column can be designed by continuously solving the VFP balances backwards

in time, solving the algebraic LFP equations and finding the composition in the
stage above after VFP. The liquid feed stage is found when the stage concentra
tion matches the feed concentration. The top stage is found when the target top
product purity after VFP is reached. Figure 2.4 shows how the backwards inte
gration method can solve for the two bottom stages. More stages above these
two can be added similarly.
MNT-1(VFP) Stage NT-1 MNT-1(LFP)
VNT, yNT
MNT(VFP) Reboiler Reboiler

MNT(LFP) MNT(LFP)
Stage NT Stage NT
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.

THE 21ST CENTURY
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).
VFP molar balances with a single feed location:
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
VFP energy balances with a single feed location:
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
LFP energy balances with a single feed location.

( )
( LF P ) D+L
h1 = h1 1 − (V F P )
(2.19)
M1
(LF P ) L (V F P )
h2 = h
(V F P ) 1
(2.20)
M1
(V F P )
h(LF
n
P)
= hn−1 for n = [3, . . . , N F − 1; N F + 1, . . . , N T − 1] (2.21)
(LF P ) (V F P )
hN F +1 = hN F + hF (2.22)
( )
(LF P ) (V F P ) (V F P ) B
hN T =hN T −1 + hN T 1− (V F P )
(2.23)
MN T
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.
2.2.3 Modelling of reactive cyclic distillation

Reactive distillation is a wellestablished process of intensification. It combines
both reaction and separation in a single column. This makes it possible to re
duce the investment and energy cost [48]. Different industrial applications of
conventional reactive distillation already exist, such as the production of methyl
acetate, methyl tertbutyl ether (MTBE), ethyl tertbutyl ether (ETBE), tertamyl
methyl ether (TAME) and fatty acid metyhl esters (FAME) [6]. One wellknown
example of the benefits of reactive distillation is the Eastman process for the pro
duction of methyl acetate, in which a conventional setup with multiple reactors and
separation columns can be replaced by a single reactive distillation column and a
couple of subsequent separation columns for removal of impurities [5, 6].
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.
Current applications of reactive separations utilise that the separation enhances

the desired reaction path. Since the periodic operation of a column lead to more
efficient separations [11, 45] the integration of cyclic and reactive distillation is
a promising highly intensifying process alternative to conventional reactive dis
tillation. The additional design variables in a cyclic distillation process, i.e. the
duration of the two flow periods (VFP and LFP), allows for the time that liquid
holdup spends on a stage to be optimised concerning the reaction. This provides
flexibility in terms of catalyst loading and the residence time on the stages. Kiss
[49] outlined some benefits of various reactive distillation processes, including a
reactive cyclic distillation. Pǎtruţ et al. [45] presented a dynamic mass balance
model for reactive cyclic distillation that accounts for the combination of reactive
and cyclic distillation for the production of dimethyl ether (DME). Their extensive
study on DME investigates the reactive cyclic distillation performance and the pos

THE 21ST CENTURY
sibility of multiple periodic states. Table 2.2 shows the current literature available
on reactive (catalytic) cyclic distillation.
Table 2.2: Current literature available on reactive cyclic distillation.
Authors/Year Contribution Conclusions

Kiss [49] / Overview of various reac The DME synthesis is a suitable
2019 tions and novel alterna choice for catalytic cyclic distilla
tives to conventional reac tion, with benefits such as high col
tive distillation processes. umn throughput, low energy usage
and high separation performance.
Kiss, Job Overview of reactive distil Catalytic cyclic distillation can fa
son and lation equipment including cilitate slower reactions as the liq
Gao [50] / a catalytic cyclic distillation uid holdup and catalyst load per
2019 column. stage can be higher than conven
tional operation. With the cyclic
operation, more degrees of free
dom are introduced, complicating
the operation and control of the pro
cess. There is a need for the de
velopment and validation of pro
cess models, design approaches
and control strategies.
Pǎtruţ et al. First rigorous mass bal Benefits in catalytic cyclic dis
[45] / 2014 ance stage model and de tillation over conventional reac
sign method for cataylitic tive distillation include lower en
cyclic distillation. The syn ergy requirement and higher opera
thesis of DME was cho tional flexibility. Drawbacks include
sen as a case study to in the possibility of unstable periodic
vestigate the cyclic state states, meaning the column design
behaviour and a proposed should be made carefully.
design method.
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:

VFP molar balance:

dM1,i
= V2 y2,i for i = 1, . . . , N C (2.24)
dt
dMn,i
= Vn+1 yn+1,i − Vn yn, i + νi rn mcat n
dt (2.25)
for n = 2, . . . , N T − 1 and i = 1, . . . , N C
dMN T,i
= −VN T yN T,i for i = 1, . . . , N C (2.26)
dt
Vapour flow rate equation for stage n:
∑N C vap
Vn+1 i=1 ∆Hn+1,i yn+1,i + rn mcat reac
n ∆Hn
Vn = ∑N C vap
(2.27)
i=1 ∆Hn yn,i
The mole balances for LFP are the same as equations (2.4)(2.8).
The design of a conventional reactive distillation is in itself a difficult task, more

so than for the design of a nonreactive distillation, as the simultaneous reaction
and separation must be accounted for. Some general methods are available such
as graphical methods, heuristic methods, and optimisation methods [51–54]. The
number of rectifying and stripping stages can, for the most part, be found using
design methods for nonreactive columns. The number of reactive stages is much
more challenging to estimate.
As mentioned, a design method for nonreactive cyclic distillation columns was

proposed by Toftegård and Jørgensen [31] and later by Pǎtruţ [11]. This method
was also used by Pǎtruţ et al. [45] to find the number of stripping stages for a
reactive cyclic column for dimethyl ether production assuming a rate controlled
reaction. Here they assumed the change in liquid composition over the reactive
zone is small and based on the reaction rate. They find the total amount of cat
alyst needed, which is then divided over a suitable number of stages. The num
ber of rectifying stages was found by adding stages above the reactive zone one
by one until the top product purity target was reached. One drawback with this
method is the assumption of almost constant liquid composition over the reactive
zone, which might not hold for different reactions. They concluded their study by
showing that by operating the reactive distillation in a periodic mode instead of a
classical operation, the energy requirements could be lowered, and the flexibility
of operation was increased. They also noted that the design of a reactive cyclic
distillation column must be done carefully, as the periodic state must be stable.
For conventional reactive distillation processes, different feasibility, applicability

and process evaluating methods are available. Reactive distillation is feasible

THE 21ST CENTURY
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.
2.2.4 Control of a cyclic distillation process

The first studies for the control of a cyclic distillation process were made in the
20th century [23, 29, 34]. These early studies investigated the use of a variable
period control, for which tV F P was chosen as a manipulated variable to control the
bottom product composition. More recently, Pǎtruţ et al. [11] suggested control of
the temperatures near the top and bottom at the end of VFP. The use of discrete
PIcontrollers then manipulates the reflux and vapour flow rate. Andersen et al.

[46] also presented a discrete PIcontroller based on the work by Matsubara et

al. [29], with the reboiler duty as the manipulated variable and bottom product
concentration as the controlled variable. All these previously proposed controller
configurations for cyclic distillation present simple control strategies based on the
models available at the time.
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.
2.2.5 Experimental work

After the Maleta tray design was proposed in 2011 [13] a few studies concerning
the experimental work has been made and published using this tray design. Pilot
scale studies of cyclic operated separation experiments have been published by
Maleta et al. [59, 60]. The first one [59] was a study on a cyclic distillation column
that functions as a beer column, for which an ethanolwater mixture is distilled
in order to concentrate the ethanol. The study of a cyclic distillation beer col
umn was compared to a conventional beer column located at the State Enterprise
Kosarsky Distillery Plant in Ukraine. A total of 10 Maleta trays were used in the
cyclic column, and unlike other distillation processes, no reboiler was installed.
Instead, steam is fed to the bottom of the column. With the use of a cyclic beer
column instead of a conventional one, Maleta et al. [59] showed that an increase in
throughput and a decrease in energy requirements were obtained. Furthermore,
they showed that in the cyclic column, high tray efficiencies were possible.
In 2019 Maleta et al. published another experimental study related to alcohol

production in Ukraine [60]. Here an ethanol purification stripping process using
cyclic operation and Maleta trays was investigated. Many different impurities, al
cohols, ethers and esters, were removed from a waterethanol solution with air.
They showed that it is possible to employ cyclic stripping to purify ethanol, which
can recover high ethanol contents and remove the impurities from a waste stream

THE 21ST CENTURY
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.


3 Mass and energy balance stage
models for cyclic distillation
In this chapter, a rigorous dynamic mass and energy balance stage model for
cyclic distillation is derived and presented. With this it is possible to include mul
tiple feed stages, side draw and additional energy at each stage. The model
presented in this chapter is versatile. It can account for different scenarios, which
previous models could not. Based on the dynamic mass and energy balance stage
model, a further derivation of a stage model is made for reactive cyclic distillation
processes. The derived reactive cyclic distillation model applies to fast reactions
described by chemical equilibrium or slower reactions described by rate kinetics.
The work presented in this chapter is partially based on the journal article ”A mass
and energy balance stage model for cyclic distillation”, AIChE Journal 66 (2020)
e16259 [1] and the publication entitled ”Analysing separation and reaction stage
performance in a reactive cyclic distillation process”, Chem. Eng. Process. 167
(2021) 108515. [2].

3.1. CYCLIC DISTILLATION MODEL DEVELOPMENT
3.1 Cyclic distillation model development

The current state of the art models describing cyclic distillation processes pre
sented in Chapter 2.2 are still not generally applicable. So in order to describe
different processes, column design and specifications, a general mass and en
ergy balance stage model for cyclic distillation is needed. Such a model is pro
posed here, allowing different distillation relevant features to be described, e.g.
side draws, heat supplied to the stages, timedependent vapour flow rate, and
liquid holdups.
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.
• No backmixing of the liquid holdups during LFP.

CHAPTER 3. MASS AND ENERGY BALANCE STAGE MODELS FOR CYCLIC
DISTILLATION
• Liquid feed and reflux is added during LFP, vapour feed is added continu
ously during VFP.
• Negligible vapour holdup.
• Instantaneous displacement of liquid holdups during LFP.
• Side draws can be removed during LFP.
• Liquid products in top and bottom are removed during LFP.
• Mass and heat transfer between vapour and liquid phases occur only during
VFP.
• Negligible heat of mixing.
• Negligible heat loss to surroundings.
If a mixed vapourliquid feed is introduced to the column, it is assumed that the

liquid fraction will be fed during the LFP and the vapour fraction during the VFP.
In the case of a flash of the feed, e.g. for a highpressure stream fed to a low
pressure column, it is assumed the two phases can be separated before being fed
to the column. This way the liquid can be fed during LFP and the vapour during
VFP. The remaining noncyclic operationspecific assumptions shown above is
general for distillation processes [63].
3.1.2 Balance equations

From Figure 3.1 the variables for the cyclic distillation model that are used can
be seen. These variables are: molar holdup of component i on stage n, Mn,i ,
vapour flow rate from stage n, Vn , vapour and liquid compositions of component
i on stage n, yn,i and xn,i respectively. Liquid and vapour enthalpy on stage n is
denoted hn and Hn respectively, any side draw flow rate from stage n is Sn and
the energy supplied to or removed from stage n is Qn .
As previously mentioned, a cycle in a cyclic distillation process consists of a VFP,

where vapour is moving up through the column, and an LFP, where the liquid
holdups are drained to the stage below. The molar and energy balances for
the VFP and the LFP is given below, for components i = 1, . . . , N C and stage
n = 2, . . . , N T . The reboiler and condenser will be listed separately, although the
general balances are similar to the other stages. The stages are numbered from
top to bottom, with the condenser being stage 1 and the reboiler stage N T .

VFP mole balance:
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
VFP energy balance:
dMn (t)hn (t)

= Vn+1 (t)Hn+1 (t) − Vn (t)Hn (t) + Vn+1
F F
(t)Hn+1 (t) + Qn (t) (3.2)
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.
LFP mole balance:
(LF P ) (V F P ) (V F P )
Mn,i F
= Mn−1,i + Mn−1,i − Sn xn,i (3.3)
LFP energy balance:
(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.
VFP mole balances for n = 1 and n = N T :
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
VFP energy balances for n = 1 and n = N T :
dM1 (t)h1 (t)

= V2 (t)H2 (t) + QC (t) (3.7)
dt

DISTILLATION
dMN T (t)hN T (t)

= −VN T (t)HN T (t) + QB (t) (3.8)
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.
LFP mole balance for n = 1 and n = N T :
(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)
LFP energy balance for n = 1 and n = N T :

( )
(LF P ) (LF P ) (V F P ) (V F P )
M1 h1 = h1 M1− (D + L) (3.11)
( )
(LF P ) (LF P ) (V F P ) (V F P ) (V F P ) (V F P )
MN T hN T = MN T −1 hN T −1 + hN T MN T −B (3.12)
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:
Mn,i = Mn xn,i (3.13)

∑
NC
Mn = Mn,i (3.14)
i=1
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
The total mass balance model for the LFP is:
∑
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)
Where Kn,i is the phase equilibrium constant or separation factor dependent on

the chosen thermodynamics.
3.1.3 Constitutive equations

The constitutive equations relevant for applications of the mass and energy bal
ance stage model shown above are listed in the following section. These equa
tions are dependent on the pure component and binary mixture specific param
eters and relations, which can be found in the literature or a database. In this
thesis, these parameters are all found in Aspen Plus, unless otherwise is stated.
Vapour liquid equilibrium
In all of the case studies presented in this thesis, it is assumed that the vapour
phase is ideal, and the liquid phase is either described by an activity coefficient
model or is ideal. If the liquid phase is ideal, the activity coefficients are all unity
for each component in the mixture. With these assumptions the vapour liquid
equilibrium (VLE) relation can be written, for stage n and component i as:
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

DISTILLATION
component i on stage n can be written as:

∫ Tn (t)
Hn,i (t) = Hi◦ + V
CP,i dT = Hi◦ + CP,i
V
(Tn (t) − Tref ) (3.21)
T ref
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:
g(na , nb , t) =Hnb (t) − hna (t)

[
Tna (t) ∑ V
NC
+ Mna (t) − 2 C (Mna (t)ynb ,i (t) − Mna ,i (t))
Mna (t) i=1 P,i (3.25)
∑
NC V
CP,i Tref + ∆Hvap,i
+ 2 (t)
(Mna (t)ynb ,i (t) − Mna ,i (t)) ]
i=1
M n a
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:
∂f (t) dTn (t) ∑ ∂f (t) dMn,i (t)

NC
df (t)
= + =0 (3.28)
dt ∂Tn (t) dt n=1
∂Mn,i (t) dt
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.

DISTILLATION
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.
QC (t) = V2 (t) (H2 (t) − h1 (t)) (3.29)

MN0 T BR
QB (t) = (HN T (t) − hN T (t)) (3.30)
tV F P
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
The boilup flow in equation (3.31) is in units of moles per cycle.
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

3.2. MODEL DISCUSSION AND ANALYSIS
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.
3.2 Model discussion and analysis

3.2.1 Applications of the model to cyclic stripping processes
The presented model above can be rewritten to describe a cyclic stripping pro
cess. By removing the reboiler and condenser and having a single vapour feed in
the bottom of the column and a liquid feed in the top, the mass and energy balance
in equations (3.1)(3.4) can describe a stripping process. The model presented in
this chapter accounts for a timedependent vapour flow rate. Stripping processes
are often carried out at temperatures below boiling point, and the main driving
force is the solubility of the components rather than the vapour liquid equilibrium
as in distillation. If the process does not occur at boiling point, the effect of energy
transfer is likely not that high. Stripping processes are mainly used for the removal
of low concentration impurities. Hence the change in vapour flow rate and stage
holdups are likely negligible for many cases. So for the modelling of a stripping
process, it should be considered if the degree of energy transfer occurring be
tween the liquid and vapour phase is significant, and whether the vapour flow is
changing significantly across the column or not. If not, then it might be redundant
to use the mass and energy balance model described above. In a case like this
a mass balance model such as the one presented by Pǎtruţ et al. [11], shown in
Chapter 2.2, can be used instead.
3.2.2 Liquid and vapour feed

The proposed model can handle both vapour or liquid feed. It is assumed the
liquid will be fed during LFP to stage n = N F and, together with the molar holdup
on this stage, is drained to stage n = N F + 1 before the next VFP is initiated.
If there is a vapour feed to stage n = N F , it is assumed that the vapour is fed
above stage n = N F , so that it will mix with the vapour from this stage and enter
the stage above, n = N F − 1. Vapour feed is supplied during the entirety of the
VFP.

DISTILLATION
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.
3.2.3 Stage temperature deviations

If the stage temperature is above the boiling point, flash will occur. This is assumed
to happen during the LFP after the liquid feed has mixed with the holdup. The
vapour fraction that is generated will instantaneously go to the stage above with
the vapour flow during the VFP. On the other hand, if the temperature is too low,
then the liquid holdup will be heated during the VFP by the vapour flow rate until
the boiling point is reached, and the mass and energy transfer as described in the
model can continue. The stage and column material is assumed to not affect the
liquid holdup temperature. If the temperature is too low, the VFP should be long
enough to reach the boiling point.
3.2.4 Model implementation, framework and analysis

The mass and energy balance stage model presented above has been imple
mented in MATLAB® . The model framework is shown in Figure 3.2.

Is the column Is the period

suitable for this Yes times No
separation? satisfactory?
Yes
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+

DISTILLATION
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.
Table 3.1: Incidence matrix for VFP.
VFP Mn (t) Mn,i (t) yn,i (t) Tn (t) Vn (t)

Eq (3.18) × ×
Eq (3.17) ×
Eq (3.19) × × × ×
Eq (3.1) × × × ×
Eq (3.2) × × × × ×
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.
(LF P ) (LF P ) (LF P ) (LF P ) (LF P )

LFP Mn Mn,i yn,i Tn Ln
Eq (3.18) × ×
Eq (3.17) ×
Eq (3.19) × × × ×
Eq (3.3) × × × ×
Eq (3.4) × × × × ×
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

be specified and 2N T (3 + 2N C) variables that can be calculated for each cycle in

order to be able to simulate the process for any number of cycles.
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.
3.2.5 Model variations

The mass and energy balance model is a general description of a cyclic distilla
tion process that can be used for many applications. Table 3.3 show the features
of the full model compared to some of the currently available alternative models
as presented in Chapter 2.2. The developed mass and energy balance model in
Table 3.3 can readily be reduced to the mass balance model, proposed by Pǎtruţ
[11], by removing the energy balances, vapour feed terms and setting the vapour
flow rate constant. By removing the possibility of different feed/side draws and
energy to the stage and assuming all stages are always at bubble point, the pro
posed model reduces to the currently available mass and energy balance model
by Andersen et al. [46]. All the models are capable of handling multicomponent
mixtures with nonlinear phase equilibrium.

DISTILLATION
Table 3.3: Features of the general model and simplified versions.
Feature Full stage Current mass Mass

model and energy balance
balance model model [11]
[46]
Mass transfer Yes Yes Yes
Heat transfer Yes Yes No
Timedependent vapour flow Yes (Yes) No
rate based on the energy
transfer
Vapour feed Yes Yes No
Multiple feed Yes No No
Side draws Yes No No
Additional energy term Yes No No
Tray temperatures different than Yes No No
boiling point
Subcooled feed Yes No No
In the presented mass and energy balance model, it is possible to include an

additional energy term Qn , see equation (3.2). This energy term could be used to
describe heat integration, the heat of mixing, reaction energy or any combination
of these, which the existing models are not developed for and therefore cannot
incorporate.
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.
3.3 Mass and energy balance stage model for

reactive cyclic distillation
The model for cyclic distillation presented in Section 3.1 is intentionally made as
a widely applicable model. With some minor adjustments, i.e. introducing reac
tion terms in the mass and energy balances, a stage model for reactive cyclic
distillation can be derived from this cyclic distillation model.
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.

3.3. MASS AND ENERGY BALANCE STAGE MODEL FOR REACTIVE
CYCLIC DISTILLATION
• Reaction can either be described by rate kinetics or chemical equilibrium.
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 .
VFP mole balance
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

DISTILLATION
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.
VFP energy balance
dMn (t)hn (t)

=Vn+1 (t)Hn+1 (t) − Vn (t)Hn (t) + Vn+1
F F
(t)Hn+1 (t)
dt
∑
NC (3.37)
reac rt
+ ∆Hn Rn,i (t) + Qn (t)
i=1
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
Where g reac (n, t) is defined as:
∑
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

3.4. CHAPTER CONCLUSIONS
many reactive separation applications.
LFP mole balance
(LF P ) (V F P ) (V F P )
Mn,i F
= Mn−1,i + Mn−1,i − Sn xn,i (3.40)
LFP energy balance
(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.
The remaining constitutive equations, reboiler and condenser balance equations

and summation and equilibrium equations are the same as described in Section
3.1, see equations (3.5)(3.35).
This dynamic model constitutes a differentialalgebraic equation system (DAE)

as the nonreactive cyclic distillation model as described in Section 3.2.4. A de
tailed model and degree of freedom analysis can be seen in Section 3.2.4, where
the solving of the mass and energy balance stage model for nonreactive cyclic
distillation is described.
3.4 Chapter conclusions

In this chapter, it has been shown how a widely applicable model for cyclic distilla
tion accounting for both mass and energy transfer between phases on each stage
can be derived. The cyclic distillation model is based on the wellknown MESH
equation system for conventional distillation processes [62]. Due to the separa
tion of phase movements in cyclic operation, mass and energy balances are set
up for both the liquid and the vapour flow periods, which constitutes a cycle. It is a
highly versatile model for cyclic distillation processes, with simultaneous draining
of the trays, that can be used for many different cases with the possibility of mul
tiple liquid or vapour feed, and side draws. This model for cyclic distillation was
analysed with regards to the model features and solving strategies.
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

DISTILLATION
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.


4 Cyclic distillation case studies
The mass and energy balance stage model presented in Chapter 3, section 3.1 is
used to investigate different case studies and is compared to the performance of
previous models without energy balances. A preliminary model verification using
experimental data from literature is also shown. Currently, no other model valida
tion or verification for cyclic distillation have been published. The work presented
in this chapter is based on the journal article ”A mass and energy balance stage
model for cyclic distillation”, AIChE Journal 66 (2020) e16259 [1] and the confer
ence contribution ”Model verification of cyclic separation processes” presented at
the 13th European Congress of Chemical Engineering in 2021.

4.1. CASE STUDIES FOR COMPARING PROPOSED STAGE MODEL TO
PREVIOUS STAGE MODELS
4.1 Case studies for comparing proposed stage

model to previous stage models
For evaluating the cyclic distillation model presented in the previous chapter, Sec
tion 3.1, 4 cases are chosen and investigated. These cases present binary and
ternary mixtures with ideal vapour and both ideal and nonideal liquid phases. In
this chapter, the more unique features of the proposed cyclic distillation model,
such as side draws and heat integration/energy added to the stages, will not be
investigated.
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.
Table 4.1: Simulation parameters for the case studies.
Parameter Benzene/ Ethanol/ Benzene/ Ethanol/

Toluene n Toluene/ Methanol/
Pentanol oXylene Water
Pressure (P ) 101.325 101.325 101.325 101.325
kPa kPa kPa kPa
Number of trays (N T ) 14 21 14 15
Feed location (N F ) 7 11 7 12
NF
Feed flow (MF ) 375 mol/ 833.33 375 mol/ 200 mol/
cycle mol/cycle cycle cycle
F
x1 50 mol% 50 mol% 45 mol% 15 mol%
F
x2 50 mol% 50 mol% 45 mol% 5 mol%
xF3 10 mol% 80 mol%
Constant vapour flow (V ) 36.83 42.78 29.17 5 mol/s
mol/s mol/s mol/s
VFP duration (tV F P ) 12 s 25 s 12 s 25 s
LFP duration (tLF P ) 3s 5s 3s 5s
Distillate (D) 187.5 416.665 187.5 25 mol/cy
mol/cycle mol/cycle mol/cycle cle
Boilup ratio (BR) 0.09 0.12 0.08 0.07
Condenser holdup (M10 ) 5 kmol 10 kmol 5 kmol 2 kmol
0
Reboiler holdup (MN T ) 5 kmol 10 kmol 5 kmol 2 kmol
The cyclic distillation process has not been simulated with existing tools in com

CHAPTER 4. CYCLIC DISTILLATION CASE STUDIES
mercial simulation software. Currently, there are no such commercial simulation

tools for cyclic distillation available. The proposed model from Chapter 3 is in
vestigated with the cases shown in Table 4.1. Three of the four cases that are
analysed in this paper have previously been studied in another cyclic distillation
modelling work by Pǎtruţ et al. [11].
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
Figure 4.1: Stages in quasistationary steadystate for mass balance model

(blue) and mass and energy balance model (red) for an ideal mixture of ben
zene/toluene.

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.
4.1.1 Dynamic open loop disturbance response

For the ethanol/methanol/water case in Table 4.1, with saturated liquid feed, the
introduction of disturbances in the feed composition are investigated. A ±5%
change in the feed content of ethanol is introduced, while keeping the methanol
feed composition constant. The condenser and reboiler content after each VFP
for these disturbances and for the quasistationary steadystate are shown in Fig

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).
Table 4.2: Relative changes in composition of EtOH in condenser from

quasistationary steadystate for changes in feed composition for the
ethanol/methanol/water case.
Change in feed EtOH Simple model Extended model

+5% 0.64% 0.62%
5% 0.69% 0.67%
The relative changes and dynamic response in quasistationary steadystates are

similar for both models. The two models also showed a similar quasistationary

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.
After a 5% decrease in the feed temperature to 346.98 K, the composition of ben

zene changes drastically. It appears that when the temperature is decreased, the
separation of benzene and toluene is highly affected as well. When the feed tem
perature is below the boiling point, the feed stage temperature is also decreased
to below the boiling point after mixing. This means there is no mass transfer oc
curring until the liquid is heated to the boiling point.
4.2 Model verification

So far, there are no published journal articles or peerreviewed scientific contribu
tions regarding the verification or validation of the recently proposed models. In
this section, a preliminary attempt at a model verification using published experi
mental data and the proposed model presented in Chapter 3 is presented.
Maleta et al. [59] presented in 2015 an experimental study of a cyclic distillation

process for the purification of ethanol from water in an alcohol plant. It is a special
distillation case, as there is no reflux or reboiler. Instead, pure steam is supplied
at the bottom of the column, a liquid feed is fed to the top, and the vapour flow
exiting the column is condensed and sent further down the plant. The process
is, therefore, a sort of a stripping process at boiling condition. Maleta et al. [59]
presented nine different cases with differences in the liquid and vapour feed flow
rates. Table 4.3 shows the feed flows for the nine experiments. In all cases the
liquid feed has a concentration of 3.325 mol% ethanol and the vapour feed is pure

4.2. MODEL VERIFICATION
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].
Experiment no. LF (mol/cyc) V F (mol/cyc)

1 619.82 113.45
2 618.01 119.00
3 683.86 101.74
4 704.40 102.97
5 715.88 110.99
6 610.76 88.18
7 679.03 110.37
8 697.15 102.97
9 715.88 108.52
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.
ai,j EtOH H2 O bi,j EtOH H2 O αi,j EtOH H2 O

EtOH 0 0.8009 EtOH 0 246.18 EtOH 0 0.3
H2 O 3.4578 0 H2 O 586.081 0 H2 O 0.3 0
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.

4.2. MODEL VERIFICATION
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.
Experiment no. LF (mol/cyc) V F (mol/cyc)

1 594.36 87.99
2 596.84 92.78
3 660.32 78.18
4 689.92 73.33
5 632.00 78.31
6 598.68 85.90
7 649.24 74.27
8 656.04 73.33
9 738.00 70.52
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 4.7: Comparison of experimental and simulated product flow rates

and compositions with adjusted feed flow rates shown in Table 4.5. 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 10
logarithm scale.

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.

Comparisons of the proposed mass and energy transfer model and the current
mass transfer model for different cases have been carried out. These cases
showed that significant changes in the vapour flow rate over the column would
lead to different column profile predictions and different column designs. An ex
ample of this is the case of benzene/toluene, where the change in vapour flow
rate was found to be significant. The proposed mass and energy balance model
predicted that high purity products could be reached with fewer stages than the
mass transfer model predicted. On the other hand, for the ethanol/methanol/wa
ter case, where there was a small change in vapour flow rate, the column profile
predicted by the mass transfer and the mass and energy transfer models were
similar.
Furthermore, the timedependent stage temperatures can be defined as controlled

variables and the feed temperature as a disturbance, shown for the benzene/
toluene case. As for the quasistationary steadystate simulations, the dynamic
responses to disturbances showed that the vapour flow rate is important. If there
is no significant change in the vapour flow rate over the column, then the dynamic
responses for a change in the feed composition for the mass balance model and
the proposed model were similar, as shown for the ethanol/methanol/water case.

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.


5 Analysis of reactive cyclic
distillation stage behaviour for
the methyl tertbutyl ether
production
The model for reactive cyclic distillation presented in Chapter 3 is applied to the
wellknown production of methyl tertbutyl ether (MTBE) in a reactive cyclic dis
tillation process. The MTBE reaction is fast, and thus it can be described either
as a chemical equilibrium or a rate kinetics reaction. Both these reaction types
are explored in a reactive cyclic distillation process. General observations of the
reactive stage behaviour are made and analysed, and the key design variables
for a reactive cyclic distillation process are discussed. The work presented in
this chapter aims to develop a process understanding of the intensified reactive
cyclic distillation technology. The work presented in this chapter is based on the
scientific journal publication entitled: ”Analysing separation and reaction stage
performance in a reactive cyclic distillation process”, Chem. Eng. Process. 167
(2021) 108515 [2].

5.1. PRODUCTION OF METHYL TERTBUTYL ETHER
5.1 Production of methyl tertbutyl ether

The production of methyl tertbutyl ether (MTBE) is a wellknown implementation
of reactive distillation, which has been studied extensively in the literature using
both chemical equilibrium and rate kinetics based reactions due to its fast kinetics.
The case study presented by Jacobs and Krishna [70] has been used as a base
case in multiple publications, mainly focusing on the multiplicity in conversions of
isobutene [71–77].
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]:
(CH3 )2 CCH2 + CH3 OH ⇀

↽ (CH3 )3 COCH3 (5.1)
iBut + MeOH ⇀
↽ MTBE (5.2)
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.
i\ j Butenes MTBE MeOH

Butenes 434.01 5872.24
MTBE 204.73 3897.10
MeOH 294.14 731.95
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.

CHAPTER 5. ANALYSIS OF REACTIVE CYCLIC DISTILLATION STAGE
BEHAVIOUR FOR THE METHYL TERTBUTYL ETHER PRODUCTION
The temperature dependency of the reaction enthalpy is assumed to be negligi

ble over the column operation temperature. That is the heat of reaction is con
stant: ∆Hnreac = −43.6 kJ/mol [78]. As described in Section 3.3 the production or
rt
consumption of component i on stage n is defined as Rn,i , which is positive for
production and negative for consumption. Only one reaction is considered here,
so νn,i,r = νn,i and rn,r = rn .
If a chemical equilibrium reaction is assumed, the coefficient ϵn , i.e. the change

in the molar holdup or composition needed to reach equilibrium, must be found.
With the new variable ϵn an equation, which describes the chemical equilibrium,
for n = 1, ..., N T , is necessary:
an,M T BE γn,M T BE xn,M T BE

0 = Keq (Tn ) − = Keq (Tn ) − (5.3)
an,iBut an,M eOH γn,iBut γn,M eOH xn,iBut xn,M eOH
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

5.2. COMPARISON OF MTBE REACTIVE DISTILLATION WITH
CONVENTIONAL AND CYCLIC OPERATION
a chemical equilibrium reaction, with a equilibrium constant defined as [77]:
6820
ln Keq = −16.33 + (5.4)
T
With the temperature T in Kelvin. Since the reaction enthalpy is assumed to be

constant in this study, any correlation between the equilibrium constant and the
heat of reaction is not accounted for.
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).
5.2 Comparison of MTBE reactive distillation with

conventional and cyclic operation
In this section, the conventional column based on Wang et al. [77] has been re
produced, and a reactive cyclic column equivalent to the conventional is made
with the models described above. Both a reactive cyclic distillation process as
suming chemical equilibrium, as Wang et al. did [77], and a process where the
reaction is described by rate kinetics proposed by Rehfinger and Hoffmann [78]
are investigated. The chemical equilibrium case is abbreviated (chem. eq.) and
the rate kinetic case (kin.). These two reactive cyclic distillation cases are investi
gated in order to analyse how the proposed mass and energy balance model from
Chapter 3 can be used to describe a reaction described by chemical equilibrium

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.
LF (cyc) = LF (conv) · tcyc [mol/cyc] (5.8)

V F (cyc) = V F (conv) · tcyc [mol/cyc] (5.9)
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.
5.2.1 Conventional reactive distillation model

A conventional steadystate reactive distillation model has been set up and solved
to ensure comparable results with data available from the literature. The steady
state model, based on the MESH equations [62], is described by Taylor et al.
[68]. This conventional reactive distillation model have been solved in MATLAB®
with the lsqnonlin command. The lsqnonlin command is an algebraic least square
nonlinear equation solver that can handle upper and lower boundaries for all the

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].
5.2.2 Equivalent cyclic column design

When going from a conventional to a cyclic distillation process, all the column
inputs are kept. That is, the same feed per cycle time, same feed concentrations
and temperatures are kept. The column design in terms of rectifying, reactive and
stripping stages etc., are also maintained. The column specifications are shown
in Table 5.2 for both the conventional and the equivalent cyclic columns with either
chemical equilibrium (chem. eq.) or rate kinetics (kin.) reaction expressions. A
constant pressure drop is included across the column.
Table 5.2: Conventional and equivalent cyclic column design specifications.
Parameter Conventional Cyclic col Cyclic col

column umn (chem. umn (kin.)
eq.)
No. of trays (NT) 17 17 17
Rectifying stages 2 2 2
Reactive stages (NR) 8 (4:11) 8 (4:11) 8 (4:11)
Stripping stages 5 5 5
Overhead pressure (P ) 1110 kPa 1110 kPa 1110 kPa
Column pressure drop (∆P ) 50 kPa 50 kPa 50 kPa
Boilup ratio (BR) 0.223 0.2211
Reflux ratio (RR) 7 6.022 6.022
V/L (mol/cyc)/(mol/cyc) 0.529 0.617 0.612
Duration of VFP (tV F P ) 10 s 10 s
Duration of LFP (tLF P ) 5s 5s
Catalyst loading (mcat
n ) 1000 kg
First feed (liquid)
Flow rate (LF ) 198 mol/s 2970 mol/cyc 2970 mol/cyc
Composition (xFi ) Pure MeOH Pure MeOH Pure MeOH
Temperature (T LF ) 320 K 320 K 320 K
Feed to above stage (N F ) 10 10 10
Second feed (vapour)
Flow rate (V F ) 547 mol/s 8205 mol/cyc 8205 mol/cyc
Composition (yiF ) 36 mol% iBut 36 mol% iBut 36 mol% iBut
64 mol% nBut 64 mol% nBut 64 mol% nBut
Temperature (T V F ) 350 K 350 K 350 K
Feed to above stage (N F ) 12 12 12

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.

Conv. Cyclic chem. eq. Cyclic kin.

100 100 100
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
420 420 420
400 400 400

T (K)
380 380 380
360 360 360
340 340 340

5 10 15 5 10 15 5 10 15
Stage Stage Stage
Figure 5.1: Column profiles, temperature (bottom) and liquid composition

(top), for the MTBE process in a conventional column (left), a cyclic column
(chem. eq.) (middle) and (kin.) (right) after a VFP.
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.
Conventional Cyclic column Cyclic column

column (chem. eq.) (kin.)
B (mol/cycle) 2670 2946 2951
D (mol/cycle) 5835 5305 5298
xD
nBut (mol%) 89.5 99.0 99.1
xB
MTBE (mol%) 99.2 99.2 99.2
χiBut (%) 90.8 98.9 99.1
χMeOH (%) 89.0 98.4 98.5
QB (MJ/cycle) 648 619 614
QC (MJ/cycle) 1004 771 769
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.
5.2.3 Analysis of the separation and reaction behaviour on

selected stages
In the following section, reactive stages 4 and 11 are investigated in detail to anal
yse the difference in operating the reactive distillation in a conventional and a
cyclic operated column. These stages correspond to the top and bottom of the
reactive zone.
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.
10 -4 Stage 4, chem. eq. Stage 11, chem. eq.

total holdup (mol/(s mol))
Production of MTBE per
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.

10 -3 Stage 4, kin. Stage 11, kin.
total holdup (mol/(s mol))

Production of MTBE per
6 0.01
Cyc. VFP
Conv.
4 End of VFP 0
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).

10 -4 Stage 4, chem. eq. Stage 11, chem. eq.

4
dx iBut /dt (mol/s)
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)
Figure 5.4: Reactive, separation and combined contributions to dxn,i/dt for

stage 4 and 11. Top plots are for isobutene, middle for MTBE and bottom for
methanol for the (chem. eq.) case.

10 -3 Stage 4, kin. Stage 11, kin.

0.02
0
dx iBut /dt (mol/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)
Figure 5.5: Reactive, separation and combined contributions to dxn,i/dt for

stage 4 and 11. Top plots are for isobutene, middle for MTBE and bottom for
methanol for the (kin.) case.

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

5.3. ANALYSIS OF THE EFFECT OF KEY DESIGN VARIABLES
outputs, but it could have an effect on the operation and control.
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.
As previously mentioned, nonreactive cyclic distillation can benefit a distillation

process with higher tray efficiencies. This is due to the backmixing of liquid
holdups being eliminated in a cyclic operated process. This benefit in increased
separation also benefits the reactive cyclic distillation. The combined separation
and reaction in a periodic operated distillation process helps push the reaction
to form more and more MTBE throughout a cycle, as the mass transfer between
vapour and liquid phases coincides with the formation of MTBE. With the high
conversion in the cyclic case, there is an issue with MTBE consumption in stage
11 after approximately 46 seconds of VFP. Despite this, the overall conversion
with the periodic operation was still considerably higher than with the conventional
operation.
5.3 Analysis of the effect of key design variables

For a reactive cyclic distillation column, many variables must be considered in
order to propose a design. First of all, the number of stages, reactive and non
reactive, must be determined. Currently, there are no specific design methods for
reactive cyclic distillation columns. However, the backwards integration method
by Toftegård et al. [31] can be used for the stripping and rectifying stages, as also
done by Pǎtruţ et al. [45]. The reactive stages can then be estimated. With the
number of stages determined, there are still many design decisions to be made:
the feed locations, feed specifications, reflux, boilup and duration of the VFP and
LFP. If the cyclic column design is based on an existing reactive conventional col
umn, such as the MTBE case shown here, the feed locations and reactive stages
can be chosen as the same. Otherwise, the feed locations can be found from
other methods, e.g. graphical, heuristic or optimisation methods [52]. The feed
flow rate, composition and temperature are usually determined from the process,
i.e. desired throughput, conversion of reactants and outlets from processes before
the reactive distillation column.

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.
5.3.1 Improved cyclic operated column design

A new column design of the cyclic operated reactive distillation process for MTBE
is proposed based on the above observations for both the (chem. eq.) and the
(kin.) cases. These new realisations are not optimally designed. However, it
shows improvements with regards to reduction in column size without loss of prod
uct purity. For this new realisation, it was decided to keep the bottom product purity
of 99.2 mol% as a hard constraint. Without a systematic method for designing a re
active cyclic distillation column, the other outputs were allowed softer constraints
without significantly decreasing the performance in conversion, top product purity,
and product flow rates. A total of four stages were removed by trial and error, of
which one was rectifying, one was reactive, and two were stripping stages. The
liquid and vapour feed locations are kept above and below the bottom reactive
stage, respectively. The duration of both VFP and LFP were kept, despite the
partial consumption of MTBE in stage 11. It was deemed that the overall conver
sion of the reactants in the reactive zone was still high enough to compensate for

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.
Parameter Improved Improved

cyclic column cyclic column
(chem. eq.) (kin.)
No. of trays (NT) 13 13
Rectifying stages 1 1
Reactive stages (NR) 7 (3:9) 7 (3:9)
Stripping stages 3 3
Overhead pressure (P ) 1110 kPa 1110 kPa
Column pressure drop (∆P ) 31.25 kPa 31.25 kPa
Boilup ratio (BR) 0.225 0.22302
Reflux ratio (RR) 6.047 6.047
V/L 0.619 0.620
Duration of VFP (tV F P ) 10 s 10 s
Duration of LFP (tLF P ) 5s 5s
Catalyst loading per stage (mcat ) 1000 kg
First feed (liquid)
Flow rate (LF ) 2970 mol/cyc 2970 mol/cyc
Composition (xFi ) Pure MeOH Pure MeOH
Temperature (T LF ) 320 K 320 K
Feed to above stage (N F ) 8 8
Second feed (vapour)
Flow rate (V F ) 8205 mol/cyc 8205 mol/cyc
Composition (yiF ) 36 mol% iBut 36 mol% iBut
64 mol% nBut 64 mol% nBut
Temperature (T V F ) 350 K 350 K
Feed to above stage (N F ) 10 10
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.

Impr. cyclic chem. eq. Impr. cyclic kin.

100 100
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
Figure 5.6: Column profiles, temperature (bottom) and liquid composition

(top) for the proposed realisation of an improved cyclic column (left: (chem.
eq.) and right: (kin.) cases) after the VFP.
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.

This chapter showed how the cyclic operation of the MTBE reactive distillation
process affects both the reaction and separation when assuming either chemical
equilibrium or a rate kinetic reaction term. Cyclic distillation has previously been
shown to have an increase in separation efficiency compared to conventional dis

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.
Based on an existing column design for the conventional reactive distillation of

MTBE, the equivalent cyclic column was simulated. Two reaction descriptions
were investigated for the cyclic distillation column: one with a chemical equilib
rium assumption and the other with an assumption of a rate kinetics reaction.
The results showed that with the cyclic column, it was possible to lower the en
ergy requirements, increase the conversions of the reactants and increase the
throughput of the bottom product MTBE. Looking at the column profiles, it seems
the equivalent cyclic column is overdimensioned, which means there could be a
potential for improving the cyclic process further.
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.

6 Quantification of a reactive
cyclic distillation performance
and feasibility
The work presented in this chapter is based on the manuscript entitled: ”Quantita
tive metrics for evaluating reactive cyclic distillation performance” submitted to the
journal: Chemical Engineering and Processing: Process Intensification (3011
2021) and the conference contribution ”Moving from conventional to the cyclic op
eration of reactive distillation processes” presented at the 13th European Congress
of Chemical Engineering in 2021. In this chapter the reactive cyclic distillation
model described in Chapter 3 is used to analyse the performance of three differ
ent reactive distillation process in periodic operation. The processes are analysed
with a focus on column and stage behaviour when a disturbance is introduced in
the vapour flow period duration, the vapour flow rate or the liquid holdup. Three
performance indicators are defined to help quantify the process’ behaviour, ac
counting for the extent of reaction over a cycle, the reaction equilibrium and the
mean Damköhler number over a cycle. Furthermore, the feasibility of reactive
distillation in cyclic operated processes is considered. This chapter continues the
analysis of reactive cyclic distillation processes presented in Chapter 5.

6.1. QUANTITATIVE METRICS FOR ANALYSING REACTIVE CYCLIC
DISTILLATION
6.1 Quantitative metrics for analysing reactive

cyclic distillation
Different analysis methods are defined and presented here for the performance
evaluation of a reactive cyclic distillation process. Three quantitative metrics,
called performance indicators, are defined to evaluate important process phenom
ena. The effect of input disturbances on the process performance is investigated
and analysed. The performance indicators are: the total amount of product being
formed over a cycle, the relative distance to the equilibrium at the end of a VFP,
and the mean Damköhler number denoted P InI , P InII and P InIII , respectively. The
performance indicators are used in this chapter to evaluate the stage performance
for different perturbations in the duration of the VFP, the vapour flow rate and the
liquid holdup. Furthermore, different feasibility conditions are defined for reac
tive cyclic distillation. These feasibility conditions are based on existing reactive
distillation analysis methods.
6.1.1 Performance indicators

The first performance indicator (P InI ), i.e. the total amount of product being formed
over a cycle, is found as the integral of the reaction rate from t = 0 to t = tV F P for
the component i = product ID in stage n.
∫ t=tV F P ∫ t=tV F P
P InI = Rn,i (t) dt = νi rn,i (t)mcat
n dt [mol/cyc] (6.1)
t=0 t=0
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

CHAPTER 6. QUANTIFICATION OF A REACTIVE CYCLIC DISTILLATION
PERFORMANCE AND FEASIBILITY
nents:
∏
NC
′
Keq,n = aνn,i
i
(6.2)
i=1
The second performance indicator can then be determined as:



 reaction is moving towards reacting the product
′ 

>1:
Keq,n
P InII = = =1: at equilibrium (6.3)
Keq,n 


< 1 : reaction is moving towards forming the product
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

6.1. QUANTITATIVE METRICS FOR ANALYSING REACTIVE CYCLIC
DISTILLATION
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)
6.1.2 Full column feasibility analysis

In the framework for feasibility and technical evaluation for reactive distillation
presented by Shah et al. [55], they show that a reactive distillation is technically
feasible if either of the following are satisfied:
• Da ≤ 0.1 & Keq > 1
• Da > 0.1 & Keq > 1 or Keq ≤ 1
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.
6.2 Case descriptions and results

Three different reactive distillation cases have been chosen for the column and
reactive stage analysis. Two of these have previously been described in cyclic
operation, the production of dimethyl ether (DME) [45] and methyl tertbutyl ether
(MTBE) presented in Chapter 5. The third case presented in this chapter is the
production of methyl acetate (MeOAc), a wellknown classic industrial application
of reactive distillation [5, 80–82].
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

6.2. CASE DESCRIPTIONS AND RESULTS
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.
6.2.1 Dimethyl ether production

The reactive cyclic distillation for the production of dimethyl ether (DME) has pre
viously been described by Pǎtruţ et al. [45]. Methanol forms DME and water in a
reversible reaction over a heterogeneous catalyst, Amberlyst 35.
2 MeOH ⇀
↽ DME + H2 O (6.12)

With a constant reaction enthalpy of ∆Hnreac = −1911.15 J/mol [45]. Hosseininejad

et al. [79] described the reaction kinetics, which was also used by Pǎtruţ et al. [45]:
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.
Table 6.1: Binary interaction energies, aDM

i,j
E
and bDM
i,j
E
, for the DME case
[45].
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.
DME singlestage analysis

Stages 5, 7 and 9 are selected to be analysed in more detail for the DME case.
This corresponds to stages in the top and middle of the reactive zone. The re
sults for each of the chosen stages, regarding the performance indicators, when
a disturbance is introduced, can be seen in Figure 6.2.
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.

Stage 5 Stage 7 Stage 9

15 15 15
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
10-3 10-3 10-3

4 4 4
3 3 3
2 2 2
1 1 1
0 0 0
0.04 0.04 0.04
0.03 0.03 0.03
0.02 0.02 0.02
0.01 0.01 0.01
0 0 0
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.

15 15 15
10 10 10
5 5 5
0 0 0
tVFP RR BR tVFP RR BR tVFP RR BR
10-3 10-3 10-3

4 4 4
3 3 3
2 2 2
1 1 1
0 0 0
0.04 0.04 0.04
0.03 0.03 0.03
0.02 0.02 0.02
0.01 0.01 0.01
0 0 0
Figure 6.3: The effect of ±1 % perturbations in tV F P , RR and BR for full

column simulations on the performance indicators for the DME case in stages
The performance of stage 5 seems to be mostly unaffected by the changes in

tV F P , RR and BR. In stage 7 the performance indicators are a bit more affected,
while in stage 9, it is clear to see that the performance indicators are affected. The
change in the amount of DME formed over a cycle, P InI , is increasing slightly with
an increase in tV F P or RR or decrease in BR, but no significant changes can be
seen. The relative distance to equilibrium seems to increase for when tV F P and
RR is decreased, or BR is increased, opposite of what was shown in the single
stage analysis in Figure 6.2. Finally, a deviation from the base case steady state

can be seen in the third performance indicator for stage 9. Decreases in tV F P , RR

and increase in BR causes the mean Damköhler to increase, following the trend
for P InII . The slow forward reaction can also be seen from the P InII plot in Figure
6.3, where the reaction is shown to be very far from chemical equilibrium.
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:
L = 2098.4 < V 2 · tV F P = 2157.7 [mol/cyc] (6.16)

∑
NT
V N T · tV F P = 1130.0 < L + LFn = 2248.4 [mol/cyc] (6.17)
n=1
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.
6.2.2 Methyl acetate production

The reactive distillation of methyl acetate (MeOAc) is wellknown, with it being
one of the earliest industrial implementations of conventional reactive distillation
[5]. So far, the reactive distillation process has not been studied with a cyclic
operating mode. To form MeOAc, methanol and acetic acid react in a reversible
reaction with water being a byproduct:
MeOH + HOAc ⇀
↽ MeOAc + H2 O (6.18)

With a heat of reaction of ∆Hnreac = 3016.5 J/mol [81].
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.
Table 6.4: Binary interaction energies for the MeOAc case, aM

i,j
eOAc M eOAc
and αi,j
[80].
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
for MeOAc case.
MeOAc singlestage analysis

A 10 % change in the tV F P , Vn+1 and Mn−1 were introduced to the three stages.
The comparison of the three performance indicators can be seen in Figure 6.5.
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.


50 50 50
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
1 1 1
0.5 0.5 0.5
0 0 0
80 80 80
60 60 60
40 40 40
20 20 20
0 0 0
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.


50 50 50
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
1 1 1
0.5 0.5 0.5
0 0 0
80 80 80
60 60 60
40 40 40
20 20 20
0 0 0
Figure 6.6: The effect of ±1 % perturbations in tV F P , RR and BR for full col

umn simulations on the performance indicators for the MeOAc case in stages
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].
state of the MeOAc case gives:
L = 487.1 < V 2 · tV F P = 710.9 [mol/cyc] (6.22)

∑
NT
V N T · tV F P = 577.8 < L + LFn = 987.1 [mol/cyc] (6.23)
n=1
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.
6.2.3 Methyl tertbutyl ether production

The final case study presented here is the production of methyl tertbutyl ether
(MTBE). The case is already shown in Chapter 5. For the analysis in this chapter,
it is assumed the reaction is described by rate kinetics as presented by Rehfinger
and Hoffmann [78].
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
for the and MTBE case.
MTBE singlestage analysis

For the last case, MTBE, the selected reactive stages are 3, 6 and 9. This corre
sponds to the top, middle and bottom of the reactive zone.
The effect of the changes in the inputs on the performance indicators can be seen
in Figure 6.8.

1000 1000 1000
500 500 500
0 0 0
1 1 1
0.5 0.5 0.5
0 0 0
0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
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
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.
1000 1000 1000
500 500 500
0 0 0
1 1 1
0.5 0.5 0.5
0 0 0
0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
Figure 6.9: The effect of ±1 % perturbations in tV F P , RR and BR for full

column simulations on the performance indicators for the MTBE case in stages
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.
state of the MTBE case gives:
L = 32080 < V 2 · tV F P = 45832 [mol/cyc] (6.24)

6.3. GENERAL TRENDS AND OBSERVATIONS FOR THE STAGE AND
COLUMN ANALYSIS
∑
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.
6.3 General trends and observations for the stage

and column analysis
The tables 6.2, 6.5 and 6.7 shows the difference in the three chosen cases. There
is a high feed in the MTBE case, resulting in high stage holdups and thus high
throughput. This also means a high catalyst load is necessary and that the inter
nal vapour flow rate and condenser and reboiler duties are high as well. Despite
the high capacity and throughput, it is still possible to run a reactive cyclic distil
lation process for MTBE with a cycle time of just 15 seconds due to the fast rate
kinetics and high vapour flow rate. The two other cases have smaller capacities
and throughput, and due to the slower reaction kinetics, the cycle times are also
significantly longer, 65 seconds for DME and 125 seconds for MeOAc. With a
constant LFP duration of 5 seconds, the time for reaction and phase separation
(VFP) in the MTBE corresponds to 66.67 % of the cycle time. For the DME case it
is 92.31 % of the cycle time, and for the MeOAc case it is 96 % of the cycle time ded
icated to the VFP. The three cases also show different reaction stoichiometries.
For the DME case, two of the same reactants becomes two different products. In
the MeOAc case, two different reactants become two different products, and in
the MTBE case, two different reactants become one product in the presence of
an inert.
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.

Quantitative metrics in the form of performance indicators: the extent of reaction
over a VFP, distance to equilibrium after a VFP, and mean Damköhler number
over a VFP, are presented here as a way of evaluating a reactive cyclic distilla
tion process. These metrics were then applied to three reactive cyclic distillation
cases: dimethyl ether (DME), methyl acetate (MeOAc) and methyl tertbutyl ether
(MTBE) for when a disturbance is introduced in tV F P , RR or BR. Of these, both
the DME and MTBE have previously been studied in a reactive cyclic distillation
process. The description of a MeOAc production in a cyclic operating process is
new. The three different case studies present wellknown applications of reactive
distillation. It is assumed that all reactions are described by rate kinetics. The dif
ferent case studies represent different column, separation and reaction designs
with regards to the capacity, throughput, reaction rate, main product volatility and
reaction stoichiometries.

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.


7 Discussion
In this chapter, an overall discussion is presented of the main findings in this the
sis. These findings are the model limitations, and the challenges in designing
a reactive cyclic distillation process. The model limitations include assumptions
regarding the pressure drop, stage equilibrium, temperature independent parame
ters and the ideal vapour phase. The design of a reactive cyclic process is difficult
as many variables must be specified, as seen in the two previous chapters. Based
on the work presented in Chapter 5 and 6 a design method is derived for reac
tive cyclic distillation processes. This chapter is partially based on the published
articles: ”A mass and energy balance stage model for cyclic distillation”, AIChE
Journal 66 (2020) e16259 [1], ”Analysing separation and reaction stage perfor
mance in a reactive cyclic distillation process”, Chem. Eng. Process 167 (2021)
108515 [2] and the manuscript ”Quantitative metrics for evaluating reactive cyclic
distillation performance” submitted to Chem. Eng. Process. (30112021).

7.1. LIMITATIONS OF THE PROPOSED MASS AND ENERGY BALANCE
STAGE MODEL
7.1 Limitations of the proposed mass and energy

balance stage model
The presented model for cyclic distillation in Chapter 3 is subject to a number of
assumptions. Including pressure drop over each tray would require information
about the column design specifications, i.e. the internals, the tray diameter and
height of the liquid holdup. In a cyclic operation with simultaneous draining, the
pressure would equalise in the column during the LFP while the pressure profile
will build up dynamically over the VFP. Hence the magnitude for this pressure
dynamics over the length of the VFP will, in the end, determine if pressure drop
effects can be ignored. Alternatively, a more straightforward method would be to
set a specific static pressure drop per stage during the VFPs. This was done for
the MTBE reactive case presented in Chapter 5 and 6. In cyclic distillation trays,
there are no downcomers and thus no physical limit on the height of the liquid.
However, the pressure drop is still affected by the liquid level, and weeping can
still occur. The assumption of no pressure drop is often used in the modelling of
columns, when the pressure drops are low.
Another important assumption is the assumption of equilibrium trays. This means

the vapour exiting a tray is in equilibrium with the liquid holdup. This is rarely the
case, and a point efficiency less than unity is often introduced to account for this.
An assumption of point efficiency equal to one is suitable for modelling purposes
to simplify the implementation and calculations. When designing a column, the
assumption of equilibrium stages will most likely underdimension the column, and
the desired separation is not achieved with the calculated number of equilibrium
trays [83].
As mentioned in Chapter 3 the assumption of separated phase flow rates leads

to an assumption of a flash of feed that would be difficult to implement in real life.
However, this assumption is necessary in order to be able to allow a simulation of
the separated phase movements. As long as the amount of vapour released after
the flash is small or the duration of the VFP is short, this assumption is reasonable.
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.

CHAPTER 7. DISCUSSION
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.
7.2 Reactive cyclic distillation analysis and

process design
From the reactive cyclic distillation case studies presented in Chapter 5 and 6
detailed analysis of the stage dynamics over a VFP and the stage and column
performance were conducted.
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.
• Specifying the in and outlet specifications.

7.2. REACTIVE CYCLIC DISTILLATION ANALYSIS AND PROCESS DESIGN
• Specify the duration of the cycle, vapour and liquid flow periods.
• Estimate the boilup and reflux, i.e. internal vapour and liquid flow rate.
• Finding the number of nonreactive stages.
• Finding the number of reactive stages.
• Evaluating the design.
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.
In order to estimate the parameters tV F P , RR and BR, a feasibility check can be

conducted first by evaluating the mean Danumber over a VFP. To determine the
Da and K eq,n values the vapour flow rate, V , and tV F P needs to be estimated,
see equation (6.5) and (6.7). If the feasibility study does not give a viable reac
tive distillation, that is if both Da and K eq,n are low, then the Vn and tV F P can be
altered so that a reactive cyclic distillation can be carried out. This will give some
indications for the choice of V and tV F P .
For further determination of tV F P , RR and BR the feasibility condition by Pǎtruţ

et al. [45] can be evaluated. Equations (6.8) and (6.9) can be used for this. If an
estimate for the tV F P and V is known from the evaluation of the Davalue, then
this can be used to find an estimate for BR according to equation (3.31). The
estimated values of tV F P , BR and RR can be verified with equation (6.8)(6.9)
and an indicator of the reflux can also be made.
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

CHAPTER 7. DISCUSSION
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.
Firstly, the problem or process is defined, followed by a specification of the de

sired in and output streams. Then an estimate of cycle time, the VFP, and LFP
duration, as well as the boilup ratio is made. Using the Damköhler number and
equilibrium constant for the reaction, the feasibility of the reactive cyclic distillation
can be evaluated. If the process is not feasible, new estimates of tV F P and BR
can be made. Otherwise, the reflux ratio is estimated and if a distillation column
design exists this can be used to run simulation of the process. Alternatively, a
column design can be made by finding number of stages and feed locations, e.g.
with the backwards integration method. To simulate the process initial conditions
for the stage holdups and temperatures must be specified. If the target product
stream specifications can achieved with the derived column design then the de
sign procedure is done. If not, the choices of the number of stages, feed locations,
reflux ratio and boilup ratio and tV F P should be reconsidered. The feasibility con
ditions in equation (6.8) and (6.9) can be evaluated to ensure a proper reactive
cyclic distillation design has been made.
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.

7.2. REACTIVE CYCLIC DISTILLATION ANALYSIS AND PROCESS DESIGN
Problem Specify feed and

definition product streams
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
Design for reactive

cyclic distillation
process achieved
Figure 7.1: Design method for a reactive cyclic distillation process.

8 Thesis conclusion
Cyclic distillation technology has moved far in recent years with advancements
in tray design, operation design, modelling, and reactive and cyclic distillation
integration.
In this PhD project, the fundamental knowledge of cyclic distillation technology

has been advanced. A new model has been proposed, which accounts for each
stage’s mass and energy balances in a distillation column. The addition of the
energy balances compared to previous models allows for a timedependent stage
temperature expression and a timedependent stage vapour flow rate. Further
more, the general model can describe processes with multiple feeds, liquid or
vapour, side draws, and additional energy added to or removed from a stage.
With the incorporation of energy balances, the reboiler and condenser can also
be modelled regarding energy duties. Based on this proposed cyclic distillation
model, another model was developed for reactive cyclic distillation processes.
This model shares the same features as the nonreactive distillation model, such
as energy balances, multiple feed locations, timedependent stage vapour flow
rate and so on. A reaction term is added to the mass and energy stage balances.
This term can either be a rate kinetics expression or a chemical equilibrium re
action expression, depending on how fast the reaction rate is. To simplify, it is
assumed that the reaction and the mass transfer between phases only occur dur
ing the vapour flow period.
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

design were made. When designing a cyclic distillation process, two additional
variables must be determined compared to conventional distillation. That is the
duration of the liquid and vapour flow periods. The duration of the liquid flow period
is mainly the time it takes to drain a stage and fill it again completely. The duration
of the vapour flow period is much more difficult to specify, as one must account
for the dynamic separation and possible reaction. Due to the inherently dynamic
behaviour over a vapour flow period, it is impossible to reach a true steady state.
A pseudo or quasistationary periodic steady state is possible to find, where each
cycle is repeated, i.e. the column profiles for the compositions, temperatures,
vapour flow rates and so on are the same for every cycle.
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

CHAPTER 8. THESIS CONCLUSION
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.

9 Future perspectives
The cyclic distillation technology is still being developed and is emerging as an
alternative to conventional distillation that allows for an increase in separation
efficiency, column size, and energy requirements.
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.
Designing a cyclic distillation process is difficult, as no true steadystate known

from conventional distillation exists, making it difficult to propose a shortcut de
sign method. Current design methods are based on specifying the duration of
the vapour flow rate, estimating the boilup and reflux and iteratively building the
column from the bottom up by adding stage after stage and solving the mass
balances backwards through the VFP. It would be of great interest and help if a
design method for nonreactive and reactive cyclic distillation can help estimate or
find an optimal value of the duration of vapour flow rate, boilup and reflux, without
the need for many iterations. This is difficult, especially in reactive cyclic distilla
tion, as the reaction and separation coincide during the VFP. One approach for
designing a reactive cyclic distillation process was presented in this thesis. How
ever, it is still a lengthy and challenging task. For future studies, it would be of
great value if a shortcut method could be developed for reactive cyclic distilla
tion and nonreactive cyclic distillation as the current method also requires many
iterations.
Another aspect, which could benefit from future studies is the control of a cyclic

distillation process. Some proposals exist, but these are for specific processes.
It would be interesting to propose a control strategy for such a process that can
easily be applied to different processes. The duration of the vapour flow period
introduces a variable that potentially could be a suitable choice for a manipulated
variable. Furthermore, as the phase movements are separated, the introduction
of liquid feed and reflux could easily be manipulated as well in order to maintain
a consistent product flow.
Most of the current literature on cyclic distillation assumes simultaneous draining

of the trays. However, a sequential draining method is also available. It would
be interesting to advance this method to the same level of knowledge as the si
multaneous drained method presented in this thesis. If two models with either
simultaneous or sequential draining were available, a comparison of the two in
terms of performance, energy demand, throughput, and so on would be of high
interest to see if one of the two methods is more beneficial in different cases.

Bibliography
[1] J. B. Rasmussen et al. “A mass and energy balance stage model for cyclic
distillation”. In: AIChE Journal 66.8 (2020), e16259.
[2] J. B. Rassmussen et al. “Analysing separation and reaction stage perfor
mance in a reactive cyclic distillation process”. In: Chemical Engineering
and Processing Process Intensification 167 (2021), p. 108515.
[3] N. Kockmann. “History of Distillation”. In: Distillation: Fundamentals and
Principles. 1st ed. Elsevier Inc., 2014. Chap. 1, pp. 1–43.
[4] A. A. Kiss. “Distillation technology still young and full of breakthrough op
portunities”. In: Journal of Chemical Technology and Biotechnology 89.4
(2013), pp. 479–498.
[5] V.H. Agreda, L.R. Partin, and W.H. Heise. “HighPurity Methyl Acetate via
Reactive Distillation”. In: Chemical Engineering Progress 86.2 (1990), pp. 40–
46.
[6] A. A. Kiss. Advanced Distillation Technologies: Design, Control and Appli
cations. 1st. Chichester, UK: John Wiley & Sons, Ltd, 2013.
[7] D. S. Sholl and R. P. Lively. “Seven chemical separations to change the
world”. In: Nature News 532.7600 (2016), pp. 435–437.
[8] A.A. Kiss. “Rethinking Energy Use for a Sustainable Chemical Industry”. In:
Chemical Engineering Transactions 76 (2019), pp. 13–18.
[9] UN General Assembly. “Transforming our world : the 2030 Agenda for Sus
tainable Development”. In: A/RES/70/1 (2015), pp. 1–41.
[10] M. R. Cannon. “Controlled Cycling Improves Various Processes”. In: Indus
trial & Engineering Chemistry 53.8 (1961), pp. 629–629.
[11] C. Pǎtruţ et al. “Cyclic distillation Design, control and applications”. In:
Separation and Purification Technology 125 (2014), pp. 326–336.
[12] C. S. Bîldea et al. “Cyclic distillation technology A minireview”. In: Journal
of Chemical Technology and Biotechnology 91.5 (2016), pp. 1215–1223.
[13] V. N. Maleta et al. “Understanding process intensification in cyclic distillation
systems”. In: Chemical Engineering and Processing: Process Intensification
50.7 (2011), pp. 655–664.
[14] B. Toftegård et al. “New Realization of Periodic Cycled Separation”. In: In
dustrial & Engineering Chemistry Research 55.6 (2016), pp. 1720–1730.
[15] W. K. Lewis. “Rectification of Binary mixtures”. In: Industrial & Engineering
Chemistry 28.4 (1936), pp. 399–402.

BIBLIOGRAPHY
[16] J. R. Mcwhirter and M. R. Cannon. “Controlled Cycling Distillation:… in

a PackedPlate Column”. In: Industrial and Engineering Chemistry 53.8
(1961), pp. 632–634.
[17] R. A. Gaska and M. R. Cannon. “Controlled Cycling Distillation in Sieve and
Screen Plate Towers”. In: Industrial & Engineering Chemistry 53.8 (1961),
pp. 630–631.
[18] J. R. McWhirter and W. A. Lloyd. “New look at distillation 5. Controlled cy
cling in distillation and extraction”. In: Chemical Engineering Progress 59.6
(1963), pp. 58–63.
[19] R. G. Robinson and A. J. Engel. “An analysis of controlled cycling mass
transfer operations”. In: Industrial and Engineering Chemistry 59.3 (1967),
pp. 22–29.
[20] I. A. Furzer. “Periodic cycling of plate columns”. In: Chemical Engineering
Science 28.1 (1973), pp. 296–299.
[21] E.B. Dale and I.A. Furzer. “An application of Zakian’s method to solve the
dynamics of a periodically cycled plate column”. In: Chemical Engineering
Science 29.12 (1974), pp. 2378–2380.
[22] O. R. Rivas. “An Analytical Solution of Cyclic Mass Transfer Operations”.
In: Industrial and Engineering Chemistry Process Design and Development
16.3 (1977), pp. 400–405.
[23] E. B. Dale and I. A. Furzer. “Periodic cycling of plate columns. Control sim
ulations and experiments”. In: Chemical Engineering Science 33.7 (1978),
pp. 905–911.
[24] G. J. Duffy and I. A. Furzer. “Mass transfer on a single sieve plate column
operated with periodic cycling”. In: AIChE Journal 24.4 (1978), pp. 588–
598.
[25] I. A. Furzer. “The discrete residence time distribution of a distillation column
operated with microprocessor controlled periodic cycling”. In: The Canadian
Journal of Chemical Engineering 56.6 (1978), pp. 747–750.
[26] I. A. Furzer. “Periodic cycling of plate columns: Mass transfer with nonideal
liquid draining”. In: AIChE Journal 25.4 (1979), pp. 600–609.
[27] J. Larsen and M. Kümmel. “Hydrodynamic model for controlled cycling in
tray columns”. In: Chemical Engineering Science 34.4 (1979), pp. 455–462.
[28] D. W. Goss and I. A. Furzer. “Mass transfer in periodically cycled plate
columns containing multiple sieve plates”. In: AIChE Journal 26.4 (1980),
pp. 663–669.
[29] M. Matsubara et al. “Relay feedback periodic control of plate columns”. In:
Chemical Engineering Science 37.5 (1982), pp. 753–758.

BIBLIOGRAPHY
[30] B. Toftegård and S.B. Jørgensen. “Identification of Periodic Cycled Distil

lation for Control Purposes”. In: IFAC Proceedings Volumes 19.15 (1986),
pp. 221–224.
[31] B. Toftegård and S. B. Jørgensen. “Design algorithm for periodic cycled bi
nary distillation columns”. In: Industrial & Engineering Chemistry Research
26.5 (1987), pp. 1041–1043.
[32] B. Toftegård and S. B. Jørgensen. “Stationary profiles for periodic cycled
separation columns: linear case”. In: Industrial & Engineering Chemistry
Research 27.3 (1988), pp. 481–485.
[33] B. Toftegård and S. B. Jørgensen. “An integration method for dynamic sim
ulation of cycled processes”. In: Computers and Chemical Engineering 13.8
(1989), pp. 927–930.
[34] M. Matsubara et al. “Periodic control of continuous distillation processes”.
In: Chemical Engineering Science 30.9 (1975), pp. 1075–1083.
[35] M. Matsubara, N. Watanabe, and H. Kurimoto. “Binary periodic distillation
scheme with enhanced energy conservation—I”. In: Chemical Engineering
Science 40.5 (1985), pp. 715–721.
[36] V. N. Schrodt et al. “Plantscale study of controlled cyclic distillation”. In:
Chemical Engineering Science 22.5 (1967), pp. 759–767.
[37] G. Baron, S. Wajc, and R. Lavie. “Stepwise periodic distillation—I”. In: Chem
ical Engineering Science 35.4 (1980), pp. 859–865.
[38] I. A. Furzer. “Steady state flow distributions in a plate column fitted with a
manifold”. In: Chemical Engineering Science 35.6 (1980), pp. 1291–1298.
[39] L. Szonyi and I. A. Furzer. “Periodic cycling of distillation columns using a
new tray design”. In: AIChE Journal 31.10 (1985), pp. 1707–1713.
[40] V. P. Krivosheev and A. V. Anufriev. “Mathematical Modeling of the Cyclic
Distillation of Binary Mixtures with a Continuous Supply of Streams to the
Column”. In: Theoretical Foundations of Chemical Engineering 52.3 (2018),
pp. 307–315.
[41] P. C. Wankat. “Continuous Cyclic Distillation for Binary Solvent Exchange:
The Batch Stack”. In: Industrial and Engineering Chemistry Research 57.47
(2018), pp. 16077–16083.
[42] F. Fazlollahi and P. Wankat. “Cyclic operation of flash and column flash dis
tillation”. In: Industrial and Engineering Chemistry Research 59.50 (2020),
pp. 21914–21929.
[43] P. Bai et al. “A dynamic modeling for cyclic total reflux batch distillation”. In:
Chinese Journal of Chemical Engineering 18.4 (2010), pp. 554–561.

BIBLIOGRAPHY
[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.

BIBLIOGRAPHY
[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.

BIBLIOGRAPHY
[72] S. S. Mansouri et al. “Systematic Integrated Process Design and Control

of Binary Element Reactive Distillation Processes”. In: AIChE Journal 62.9
(2016), pp. 3137–3154.
[73] S. S. Mansouri et al. “Systematic integrated process design and control of
reactive distillation processes involving multielements”. In: Chemical Engi
neering Research and Design 115 (2016), pp. 348–364.
[74] S. S. Mansouri et al. “Integrated process design and control of reactive dis
tillation processes”. In: IFACPapersOnLine 48.8 (2015), pp. 1120–1125.
[75] S. A. Nijhuis, F. P.J.M. Kerkhof, and A. N.S. Mak. “Multiple Steady States
during Reactive Distillation of Methyl tertButyl Ether”. In: Industrial and En
gineering Chemistry Research 32.11 (1993), pp. 2767–2774.
[76] S. Schrans, S. de Wolf, and R. Baur. “Dynamic simulation of reactive dis
tillation: An MTBE case study”. In: Computers & Chemical Engineering 20
(2003), S1619–S1624.
[77] S. J. Wang, D. S.H. Wong, and E. K. Lee. “Effect of interaction multiplicity on
control system design for a MTBE reactive distillation column”. In: Journal
of Process Control 13.6 (2003), pp. 503–515.
[78] A. Rehfinger and U. Hoffmann. “Kinetics of methyl tertiary butyl ether liquid
phase synthesis catalyzed by ion exchange resinI. Intrinsic rate expression
in liquid phase activities”. In: Chemical Engineering Science 45.6 (1990),
pp. 1605–1617.
[79] S. Hosseininejad, A. Afacan, and R.E. Hayes. “Catalytic and kinetic study
of methanol dehydration to dimethyl ether”. In: Chemical Engineering Re
search and Design 90.6 (2012), pp. 825–833.
[80] C. Zuo et al. “Catalysts, kinetics, and reactive distillation for methyl ac
etate synthesis”. In: Industrial and Engineering Chemistry Research 53.26
(2014), pp. 10540–10548.
[81] R.S. Huss et al. “Reactive distillation for methyl acetate production”. In:
Computers and Chemical Engineering 27 (2003), pp. 1855–1866.
[82] W. Song et al. “Measurement of Residue Curve Maps and Heterogeneous
Kinetics in Methyl Acetate Synthesis”. In: Industrial and Engineering Chem
istry Research 37.5 (1998), pp. 1917–1928.
[83] V. Vishwakarma, M. Schubert, and U. Hampel. “Assessment of separation
efficiency modeling and visualization approaches pertaining to flow and mix
ing patterns on distillation trays”. In: Chemical Engineering Science 185
(2018), pp. 182–208.

A Vapour flow rate derivation
A derivation of the vapour flow rate, Vn , presented in Chapter 3 is given here.
The vapour flow rate is found from the VFP energy balance in equation (3.2). For
a simpler overview, the explicitly written time dependencies are neglected in the
following.
dMn hn
= Vn+1 Hn+1 − Vn Hn + Vn+1F F
Hn+1 + Qn
dt
Other useful and helpful relations and definitions for the vapour flow rate derivation
are shown here:
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 left hand side differential of equation (3.2) is split up:
dMn hn dhn dMn

= Mn + hn
dt dt dt
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 )
d ∑ ∑
NC
V V dTn
= Mn Tn xn,i CP,i + Mn xn,i CP,i
dt i=1 i=1
dt
[ NC ]
d ∑ ( V )
− Mn xn,i CP,i Tref + ∆Hvap,i
dt i=1
( NC )
∑ d ∑
NC
V V dTn
= Mn Tn xn,i CP,i + Mn xn,i CP,i
i=1
dt i=1
dt
[ NC ]
∑ d ( V )
− Mn xn,i CP,i Tref + ∆Hvap,i
i=1
dt
The expression for dxn,i/dt from above, can be inserted.

( NC ( )
dhn ∑ Mn dMn,i/dt − Mn,i dMn/dt )
V
Mn = Mn Tn 2
CP,i
dt i=1
M n
∑
NC
dTn
V
+ Mn xn,i CP,i
i=1
dt
N C ((
∑ ) )
Mn dMn,i/dt − Mn,i dMn/dt ( V )
− Mn 2
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
are treated individually.

( NC ( )
∑ Mn dMn,i/dt − Mn,i dMn/dt )
V
Mn Tn CP,i =
i=1
Mn2
( )
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
N C ((
∑ ) )
Mn dMn,i/dt − Mn,i dMn/dt ( V
)
Mn CP,i Tref + ∆Hvap,i =
i=1
Mn2
( )
∑
NC V
i=1
Mn2

APPENDIX A. VAPOUR FLOW RATE DERIVATION
( )
∑
NC V
− 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
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
− Mn Vn+1 (Mn yn+1,i − Mn,i )
i=1
Mn2
( )
∑NC V
−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
= Vn+1 Hn+1 − Vn Hn + Vn+1

F F
Hn+1 + Qn
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
+ 2
(Mn yn+1,i − Mn,i )
i=1
M n

[ ( ( )
Tn ∑
NC
− Vn Hn − hn + Mn − (Mn yn,i − Mn,i )
Mn2 i=1
)]
∑
NC V
+ (Mn y,i − Mn,i )
i=1
Mn2
[ ( ( )
Tn
∑
NC
F
+ Vn+1 F
Hn+1 − hn + Mn − F
Mn2 i=1
)]
∑
NC V
+ 2
F
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.
g(na , nb ) = Hnb − hna

[
Tna ∑ V
NC
+ Mna − 2 C (Mna ynb ,i − Mna ,i )
Mna i=1 P,i
∑
NC V
+ (Mna ynb ,i − Mna ,i ) ]
i=1
Mn2a
The energy balance can be written as:
∑
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
equation. Suppose the vapour liquid equilibrium can be defined as a function, f ,

such that:
∑
NC ∑NC
f =1− yn,i = 1 − xn,i Kn,i = 0
i=1 i=1
Where Kn,i is the equilibrium coefficient. For the work presented in this thesis

APPENDIX A. VAPOUR FLOW RATE DERIVATION
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
From this an expression for dTn/dt can be found.

∑N C
dTn ∂f /∂Mn,i dMn,i/dt
i=1
=
dt ∂f /∂Tn
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
The partial derivatives of xn,j with regards to Mn,i depends on whether i = j or

not.
∂xn,j 1
= (1 − xn,j ) if i = j
∂Mn,i Mn
∂xn,j 1
=− xn,j if i ̸= j
∂Mn,i Mn
The derivative of f with regards to Tn is:
∂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

A.1. VAPOUR FLOW RATE DERIVATION FOR REACTIVE CYCLIC
DISTILLATION
When an activity coefficient model is chosen, the partial derivatives of ∂ ln γn,i/∂Mn,i

and ∂ ln γn,i/∂Tn can be found and with a model for the saturated pressure the deriva
sat
tive dPn,i /dTn can be found. If the liquid is ideal then γn,i = 1 and ln γn,i = 0.
A.1 Vapour flow rate derivation for reactive cyclic

distillation
A similar derivation can be made for the reactive cyclic distillation model in Sec
tion 3.3). Most of the derivation shown above is the same for the reactive cyclic
distillation. The only difference is that new terms are added to the VFP mass and
energy balances, see equation (3.36)(3.37), these new reactive terms are shown
below:
dMn,i rt
= [. . .] + Rn,i
dt
dMn (t)hn ∑
NC
reac rt
= [. . .] + ∆Hn Rn,i
dt i=1
Where, the superscript rt denotes whether the reaction is described by chemical

equilibrium (rt = eq) or rate kinetics (rt = kin). Again the vapour flow rate for
stage n can be derived by inserting the terms for the mass balance in the energy
balance and so on, as shown above. By doing so a similar expression as above
can be found, however, the reaction is also included, such that:
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
Where the reaction function g reac (n) is defined as:
∑
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

Technical
University of
Denmark
Søltofts Plads, Builiding 228A

2800 Kgs. Lyngby
Tlf. 4525 2822
www.kt.dtu.dk

Cyclic Distillation Technology: Rasmussen, Jess Bjørn

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cyclic Distillation Technology: Rasmussen, Jess Bjørn

Uploaded by

Copyright:

Available Formats

Downloaded from orbit.dtu.

dk on: Aug 22, 2022

Cyclic Distillation Technology

Rasmussen, Jess Bjørn

Link back to DTU Orbit

Cyclic Distillation Technology

Cyclic Distillation Technology

Department of Chemical and Biochemical Engineering, Technical

Sino­Danish Center for education and research (SDC)

Institute of Process Engineering, Chinese Academy of Sciences

Copyright: Reproduction of this publication in whole or in part must include

Jess Bjørn Rasmussen,

Cyclic Distillation Technology iii

Cyclic Distillation Technology v

vi Cyclic Distillation Technology

Cyclic Distillation Technology vii

viii Cyclic Distillation Technology

Cyclic Distillation Technology ix

I denne afhandling er cyklisk destilleringsteknologi blevet studeret og analyseret

En masse­ og energibalance bundmodel er blevet foreslået her. Det er en høj nø­

Opførelsen af tre forskellige reaktive cykliske destilleringsprocesser er yderligere

Cyclic Distillation Technology xi

xii Cyclic Distillation Technology

Cyclic Distillation Technology xiii

2 Cyclic distillation technology from the 20th to the 21st century 11

3 Mass and energy balance stage models for cyclic distillation 29

Cyclic Distillation Technology xv

3.1.3 Constitutive equations . . . . . . . . . . . . . . . . . . . . . 34

4 Cyclic distillation case studies 49

5 Analysis of reactive cyclic distillation stage behaviour for the methyl

6 Quantification of a reactive cyclic distillation performance and feasi­

xvi Cyclic Distillation Technology

8 Thesis conclusion 121

9 Future perspectives 125

A Vapour flow rate derivation 133

Cyclic Distillation Technology xvii

xviii Cyclic Distillation Technology

Cyclic Distillation Technology 1

In short, a distillation process is a separation technique that employs the difference

A stripping column is similar to a simplified distillation process, and, as the name

2 Cyclic Distillation Technology

A well­known distillation process intensification is reactive distillation. Here, a re­

Schematics showing examples of the conventional distillation process (a), strip­

A,B Feed location

High purity B High purity D

Figure 1.1: (a) Conventional distillation process of components A and B. (b)

Cyclic Distillation Technology 3

The United Nations has released a list of 17 Sustainable Development Goals

4 Cyclic Distillation Technology

• Goal 12: Responsible consumption and production. This is related to estab­

Cyclic Distillation Technology 5

Cyclic, or periodic, distillation is a promising emerging process intensification. In

6 Cyclic Distillation Technology

Cyclic Distillation Technology 7

1.4 Thesis structure

• Chapter 5: Analysis of reactive cyclic distillation stage behaviour for

• Chapter 6: Quantification of a reactive cyclic distillation performance

8 Cyclic Distillation Technology

• Chapter 7: Discussion, Chapter 8: Conclusion and Chapter 9: Future

• Appendix: The mathematical derivation of the vapour flow rate, which is

1.5 Publications and conference contributions

1.5.1 Journal publications

• J.B. Rasmussen, S.S. Mansouri, X. Zhang, J. Abildskov, J.K. Huusom, Analysing

SinoDanish Center for education and research (SDC)

En masse og energibalance bundmodel er blevet foreslået her. Det er en høj nø

6 Quantification of a reactive cyclic distillation performance and feasi

A wellknown distillation process intensification is reactive distillation. Here, a re

Schematics showing examples of the conventional distillation process (a), strip

• Goal 12: Responsible consumption and production. This is related to estab

• J.B. Rasmussen, S.S. Mansouri, X. Zhang, J. Abildskov, J.K. Huusom, Bench

Case II can be directly substituted with a timedependent liquid composition as

Both the backmixing and thermodynamic limitations for distillation processes,

As mentioned, a design method for nonreactive cyclic distillation columns was

[46] also presented a discrete PIcontroller based on the work by Matsubara et

• No backmixing of the liquid holdups during LFP.

If a mixed vapourliquid feed is introduced to the column, it is assumed that the