Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
Technische Universität Berlin
Politechnika Krakowska im. Tadeusza Kościuski
20th International Conference of Process Engineering and
Chemical Plant Design
2014 Berlin
Conference Proceedings
Berlin
October 15 – 17, 2014
Published by
Technische Universität Berlin
Process Science
Chair of Chemical & Process Engineering
Strasse des 17. Juni 135
10623 Berlin - Germany
http://www.verfahrenstechnik.tu-berlin.de
Tel.: +49 (0)30 / 314 23701
Fax.: +49 (0)30 / 314 21134
Editors:
Prof. Dr.-Ing. Matthias Kraume (Chairman)
Dipl.-Ing. Gregor D. Wehinger
Publishing Company & Distribution
Copy Print, Berlin
Berlin 2014
ISBN 978-3-00-047364-7
Technische Universität Berlin
Politechnika Krakowska im. Tadeusza Kościuski
20th International Conference of Process Engineering and
Chemical Plant Design
2014 Berlin
Conference Proceedings
Berlin
October 15 – 17, 2014
i
PREFACE
40 years ago the first workshop of this series of conferences started to bring together
scientists from Polytechnika Krakowska and Technische Universität Berlin. What has been a
challenging approach during cold war time developed intensively over the following decades.
Nowadays, the formerly bilateral workshop became an international conference with
contributions from more than two countries. Additionally, the collaboration was intensified by
student exchange programs and various knowledge transfers between the participating
scientists.
The conference venue alternates between Polytechnika Krakowska and Technische
Universität Berlin. The 20th
International Conference of Process Engineering and Chemical
Plant Design is held in Berlin from 15-17 October 2014.
This year, the scope of topics treated in the 21 contributions comprises special aspects
of reacting systems, multiphase systems, heat transfer and separation processes. Experimental
as well as theoretical papers using different corresponding methods are presented.
Fundamental and application-driven contributions will be discussed. Hence, the scale of the
investigated systems varies between microfluidic devices and full technical scale. The papers
of all these presentations are given in this book. All of them are peer-reviewed by the
Scientific Advisory Committee.
The conference was organised by Technische Universität Berlin. The editors would
especially like to thank all authors for their excellent contributions and members of the
Scientific Advisory Committee for their expertise. Additionally, the financial support of the
office for foreign relations (ABZ) and the DAAD is gratefully acknowledged.
MATTHIAS KRAUME
Chairman
iii
CHAIRMAN
Prof. M. Kraume (Technische Universität Berlin)
SCIENTIFIC ADVISORY COMMITTEE
Prof. J. Bałdyga (Warsaw University of Technology)
Prof. A. Drews (HTW Berlin)
Prof. M. Dyląg (Institute of Advanced Manufacturing Technology, Cracow)
Prof. S. Enders (Technische Universität Berlin)
Prof. M. Kraume (Technische Universität Berlin)
Prof. K. Kupiec (Politechnika Krakowska)
Prof. Z. Matras (Politechnika Krakowska)
Prof. F. Rieger (Czech Technical University in Prague)
Prof. K. Rup (Politechnika Krakowska)
Prof. R. Schomäcker (Technische Universität Berlin)
Prof. G. Wozny (Technische Universität Berlin)
ORGANIZATION TEAM
Prof. M. Kraume
G. Wehinger
U. Herrndorf
v
CONTENTS
PREFACE .................................................................................................................................. I SCIENTIFIC ADVISORY COMMITTEE ...................................................................................... III
1. REACTIVE SYSTEMS ........................................................................................................... 1 ALTERNATIVE USE OF COKE OVEN GAS – EXPERIMENTS WITH A MODULAR TEST PLANT AND
REAL PROCESS GAS E. Speelmanns; M. Rieger; H. Thielert; G. Wozny ............................................................. 3
HYDROFORMYLATION OF 1-DODECENE IN MICROEMULSIONS: LONG-TERM MINI-PLANT
OPERATION RESULTS D. Müller; E. Esche; T. Pogrzeba; T. Hamerla; T. Barz; R. Schomäcker; G. Wozny ...... 15
ENABLING ONLINE-OPTIMIZATION FOR A MULTIPHASE SYSTEM IN A HYDROFORMYLATION
MINI-PLANT D. Müller; M. Illner; A. Fleck; E. Esche; T. Barz; R. Schomäcker; G. Wozny ................ 25
EFFECT OF THE GEOMETRICAL AND TECHNOLOGICAL PARAMETERS ON THE FLOCS SIZE IN THE
CONTINUOUS TUBULAR REACTOR WITH BAFFLES
W. Szatko; M. Dyląg; J. Kamieński; J. Rosiński; J. Talaga; R. Wójtowicz ..................... 37 RHODIUM-CATALYZED HYDROFORMYLATION OF 1-DODECENE IN MICROEMULSION SYSTEMS
WITH NON IONIC SURFACTANTS T. Pogrzeba; T. Hamerla; R. Schomäcker ........................................................................ 49
INVESTIGATIONS TO INCREASE THE SELECTIVITY OF SUZUKI CROSS COUPLING REACTIONS BY
SEMI-BATCH EXPERIMENTS M. Schmidt; M. Schwarze; R. Schomäcker ....................................................................... 59
2. MULTIPHASE SYSTEMS ..................................................................................................... 67 MASS TRANSFER IN TWO-PHASE LIQUID-LIQUID SYSTEMS – CHALLENGES AND SOLUTIONS
J. Bałdyga;M. Jasińska; W. Kowaliński ............................................................................ 69
ELECTROCOAGULATION METHOD FOR TREATMENT OF OIL-IN-WATER EMULSIONS – PROCESS
MODELLING FROM ‘ENLIGHTENED EMPIRICISM’ TO MOSAIC Ł. Janczewski; M. Dyląg; .................................................................................................. 85
DISPERSION AND COALESCENCE IN STIRRED MICELLAR MULTIPHASE SYSTEMS
L. Hohl; N. Paul; M. Kraume ........................................................................................... 97 RELATIVE PARTICLE-TO-FLUID VELOCITY IN A TURBULENT FLUID
P. Ditl; J. Skřivánek; V. Pešava ...................................................................................... 107 SYSTEMATIC ANALYSIS OF COALESCENCE IN LIQUID/LIQUID DISPERSIONS
J. Villwock; J. Kamp; M. Kraume ................................................................................... 119
INFLUENCE OF POLYMER-SURFACTANT ADDITIVES ON PRESSURE DROPS IN PIPE FLOW Z. Matras; B. Kopiczak ................................................................................................... 127
INFLUENCE OF A NON-IONIC SURFACTANT TRITON X-100 ON FLUID DYNAMICS AND MASS
TRANSFER OF SINGLE RISING DROPS
S.-J. Kim; N. Paul; M. Kraume ....................................................................................... 141
vi
3. HEAT TRANSFER ............................................................................................................ 149
HEAT TRANSFER IN HORIZONTAL GROUND HEAT EXCHANGER – ONE-DIMENSIONAL MODEL B. Larwa; K. Kupiec; M. Gwadera; T. Komorowicz ...................................................... 151
OPTIMIZATION OF THE LUCID DETECTOR HEAT SHIELD COOLING R. Kantor......................................................................................................................... 165
HEAT TRANSFER ENHANCEMENT IN NATURAL CONVECTION IN MICROPOLAR NANOFLUIDS K. Nering; K. Rup ........................................................................................................... 179
4. SEPARATION PROCESSES ............................................................................................... 191 FILTERABILITY OF W/O PICKERING EMULSIONS IN MULTIPHASE REACTIONS
A. Drews; L. Schumacher; T. Skale; D. Zedel ................................................................ 193 SEPARATION OF CO2 FROM THE OCM PRODUCT STREAM USING HYPERBRANCHED
POLYMERS SOLUTIONS C. Walowski; S. Enders .................................................................................................. 199
DETERMINATION OF RHEOLOGICAL DATA FROM MIXING EXPERIMENTS F. Rieger; J. Moravec ..................................................................................................... 211
IMPROVED DESORPTION CONTROL VIA RAMAN SPECTROSCOPY E. Esche; B. Kraemer; D. Müller; K. Meyer; N. Zientek; M. Maiwald; G. Wozny ....... 223
COMPARISON OF DIFFERENT MEMBRANES FOR THE REMOVAL OF SURFACTANTS FROM
ORGANIC SOLVENTS BY ORGANIC SOLVENT NANOFILTRATION
D. Zedel; A. Drews; M. Kraume ..................................................................................... 235
LIST OF AUTHORS .............................................................................................................. 243
1
1. REACTIVE SYSTEMS
3
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
ALTERNATIVE USE OF COKE OVEN GAS – EXPERIMENTS WITH A MODULAR
TEST PLANT AND REAL PROCESS GAS
E. SPEELMANNS (1); M. RIEGER (2); H. THIELERT (2); G. WOZNY (1)
(1) Technische Universität Berlin, Chair of Process Dynamics and
Operation (dbta), Straße des 17. Juni 135, D-10623, Berlin, Germany
(2) Thyssen Krupp Industrial Solutions, Coke Plant Technologies, Uhde-
Str. 15, D-44141, Dortmund, Germany
Abstract. A mobile and modular test plant is presented that was used for the
investigations of two different processes concerning the reduction of sulphur
emissions of coking plants. After completion of experiments for the first process,
the plant was modified to investigate the hydrolysis of organosulphur compounds
in industrial coke oven gas. We will show that due to the strictly modular set-up,
we were able to reconstruct the plant both time and cost efficient. Finally, the
successful hydrolysis of COS and CS2 in coke oven gas is demonstrated by
showing the results of the first measurement campaign. This is the first step to
enable the use of coke oven gas for other processes.
Keywords. Modular plant design, desulphurisation, organosulphur comounds
INTRODUCTION
In the production process of steel in blast furnaces, an initial and inevitable step is the
production of coke. This is realized by a carbonization process where coal is heated up at high
temperatures out of contact with oxygen while volatile matter and coal tar are driven off the
coal. In this process, with each ton of coke about 200 kg of coke oven gas (COG) is produced.
4
This mixture of multiple gas components is treated in modern coking plants in a by-product
plant with stripping processes to reduce emissions and gain valuable products.
Hence, compounds such as tar, hydrogen sulfide, benzene and ammonia are removed
and converted to products. Hydrogen sulphide is often converted to elemental sulphur in a
Claus process, but also to sulphuric acid that is used for production of ammonium sulphate.
The entire by-product plant of a coking plant is an established Verbund of different processes.
However, it carries great potential for modifications and optimizations to reduce emissions
and improve production of by-products and therefore is an inherent part of state of the art
research [1–3].
At our chair two processes concerning the reduction of sulphur emissions were subject
to research in the recent past and in the present.
The first one, that has already been completed successfully, is the emission-free
sulphuric acid plant that reduces the SO2-emissions of a sulphuric acid plant to zero [4]. This
is realised by a heterogeneous reaction in a fixed bed reactor. Experiments with a mobile test
plant using industrial off-gas at the coking plant ArcelorMittal in Bottrop were conducted and
the results were used for process development. The test plant that was constructed in this
project was planned in a strictly modular setup so the flexible usage in other applications is
possible with only minor and therefore time and money effective adaptations [5]. This is
demonstrated by introducing the second project concerning sulphur emissions of a coking
plant for which the test plant was modified and operated. The project aims to reduce
organosulfur compounds (COS and CS2) that are found in the coke oven gas in small
concentrations (100-200 ppm), but nevertheless have to be eliminated for different reasons.
Up to now, coke oven gas is used for heating of the coke ovens and blast furnaces of
integrated steel plants or is exported as a low valuable gas. It consists of mainly hydrogen
(60%) and methane (20%) and several other substances (carbon monoxide, olefins, oxygen,
etc.) in varying concentrations. COS and CS2 cannot be washed out of the gas by the
established procedures and are obstructive for two main reasons. Due to combustion they
increase the emissions of the whole coking plant in form of SO2. Secondly they are toxic for
nickel or copper catalysts that are used for the water gas shift (WGS) reaction and steam
methane reforming (SMR) [6]. Hence, it is not possible to use coke oven gas for the
production of valuable products such as e.g. methanol, synthetic natural gas or urea.
Therefore, it is of great interest to remove organosulfur compounds from the gas [7]. Lab
experiments showed that a heterogeneous catalysis that forms H2S is effective. However,
since coke oven gas contains numerous different compounds in varying concentrations, it is
crucial to perform tests with real process gas.
5
For this purpose, the mobile and modular test plant of the previously introduced
project “emission free sulphuric acid plant” was modified. In the following we introduce the
set-up of the plant and show that due to the strictly modular concept, we were able to
reconstruct the plant and put it into operation in short time. Furthermore, the first experiments
on the conversion of organosulfur compounds that were conducted in bypass with the real
process on the coking plant ArcelorMittal Bottrop GmbH are presented.
MATERIALS AND METHODS
A mobile and modular test plant was constructed to conduct experiments with industrial
process gas. The main unit of the plant is a fixed-bed reactor; hence heterogeneous reactions
with various catalysts can be investigated. Since the plant was used in two processes, the
general set-up will be explained first, followed by the modifications that were made due to
differences in operating condition.
Modular Set-up
The test plant was planned and constructed in a strictly modular setup. It consists of three
different modules that each perform a specified task (Figure 1 and Figure 2):
1. Gas-mixing. Since different process gas streams from the coking plant as well as
pure gases from gas bottles are needed for the experiments, gas streams are mixed at
defined flow rates in module 1. (Gasbottles: SLA5850, Nitrogen: SLA5853; each
Brooks Instrument GmbH. COG: control valve: RC200, Badgermeter, measurement
device: EX-FLOW Bronkhorst Mättig GmbH).
2. Reactor. In module 2 the mixture of different gases is heated to the specified process
temperature by a gas heater. It then flows to the main item of the plant, the fixed- bed
reactor.
3. Down-stream. After reaction, the gas is cooled to ambient temperature in the third
module by two gas coolers that also include condensate traps in case water is formed
in the reactor.
6
Figure 1: Schematic drawing of the mobile test plant. The three modules including their functions are
indicated.
Figure 2. Photo of the miniplant on the coking plant ArcelorMittal Bottrop GmbH. The first module is on
the left, connected by a tube with module 2 and 3.
Module 1
Module 2+3
7
Part of the modular set-up is the sole use of standard pipe sizes and standard connection types.
Furthermore, the connections between the modules are reduced to a minimum. In addition, all
parts where constructed to realise a wide operating range. This includes mass flow devices,
pipe diameter, temperature range, capacity of the gas heater and the cooling devices. The
process measurement and control technology was chosen to be robust against disturbance
(e.g. vibrations). The test plant is controlled by industrial process control system (PCS) ABB
Freelance 800F. Moreover, the plant is weatherproof, so it can be set into operation on the
field. Due to corrosive substances stainless steel (1.4571) was chosen for all parts.
The fixed-bed reactor is the main item of the plant. Its diameter was chosen to 0,1 m so the
ratio of dR/dP is > 20 for most catalyst particle sizes. Under this condition radial diffusion can
be neglected [8]. Hence, a model development of the process will be simplified. With a length
of 1 m, the reactor can be loaded with up to 5 catalyst beds of each 0,1 m height. The
temperature is measured at the entrance of the reactor as well as behind every bed inside the
reactor by PT 100 and at the outer wall by thermocouples (Figure 4 and Figure 3). This
enables the detection of heat of formation inside the reactor as well as the calculation of heat
loss over the wall. For chemical analysis gas samples can be taken before the reactor and
behind it. Hence, the chemical composition of the gas including reaction products can be
determined by gas chromatography or online measurement devices. In our case we used
online Fourier-transformation infrared spectroscopy (FT-IR) (ANSYCO analytische Systeme
und Componenten GmbH) for some experiments to measure several components at the same
time in the product and reactant stream. In addition, the plant contains a redundant mass flow
meter (SLAMF63, Brooks Instrument GmbH) to enable data reconciliation and detection of
errors. Oxygen sensors (VisiFerm DO Arc 120, Hamilton Messtechnik GmbH) were installed
before and after the reactor to supervise the oxygen concentration in the gas, so in case of
critical rise emergency procedures can be initiated.
8
Figure 3. Schematic drawing of the fixed-bed
reactor. The maximum number of five catalyst beds
is indicated as well as temperature measurement
(TR) positions and sample points (QI).
Figure 4. Photo of the reactor
Variation of Pilot Plant
The original pilot plant was used to examine the process of the emission free sulphuric acid
plant. Modifications were made afterwards to conduct experiments to hydrolyse
organosulphur compounds in COG. Both processes with their specifications of the test plant
are compared in Table 1.
TR
TR
TR
TR
TR
TR
QI
QI
TR
TR
TR
TR
TR
TR
Bed 1
Bed 2
Bed 3
Bed 4
Bed 5
9
Table 1. Comparison of the two projects emission free sulphuric acid plant and hydrolysis of
organosulphur compounds.
Emission free sulphuric acid
plant
Hydrolysis of org. sulphur
compounds
Reaction SO2 + 3H2 → H2S + 2H2O COS + H2O ↔ H2S + CO2
CS2 + 2H2O ↔ 2H2S + CO2
Gas bottle 1 SO2 COS, CS2
Gas bottle 2 H2S H2S
Vol. fraction COG 8-10 % 50 – 100 %
Add. Feed from industries Off-gas sulphuric acid plant none
After completion of the initial project, the test plant had to be modified due to different
process specifications:
Range of flow rate for COG
Pure COG instead of dilution with nitrogen
The first specification had the sole consequence that a new mass flow measurement and
regulation device for COG flow with higher flow rate had to be installed. The second change
lead to a new hazard and operability analysis (HAZOP). The result was an adaption of the
safety concept. COG contains both inflammable and toxic substances. Therefore, new gas
warning systems, emergency shutdown procedures and a flame arrestor had to be included in
the modification. The gas heater was replaced for safety reasons by Typ HK/SE-4 (Elmess
Thermo System Technik) for which a specified maximum surface temperature is guaranteed.
Experiments with COG
First experiments on the heterogeneous catalysis of organosulphur compounds with real COG
were conducted on the coking plant ArcelorMittal Bottrop GmbH for a period of over 100
operating hours. At a constant pressure of 0.2 bar the parameters inlet temperature of the
reactor and gas hourly space velocity (GHSV) were varied in a previously determined range.
Gas analysis was realised by gas chromatography in which main components of the gas as
well as organosulfur compounds (COS and CS2) and H2S were quantified.
10
RESULTS & DISCUSSION
Due to the modular set-up of the plant it was possible to modify certain parts of single
modules without enforcing any consequences for the other modules. Nearly all of the
equipment of the first project was reused. This was enabled by the initial foresighted planning
of the mobile plant. Due to previously gained experience in the construction of the plant, the
emerging costs and time of modification was estimated accurately. It took only four month to
reconstruct the plant and take it into operation. During this period the process was designed,
the HAZOP was undertaken, new components were configured and ordered and the plant was
reconstructed. Moreover, the PCS was adapted and permit application for the experiments
was placed. After reconstruction, the mobile plant was transported to the industrial coking
plant in Bottrop by truck.
This demonstrates that the characteristics of modularization: protection of know-how,
reusability, flexibility, reduction of complexity, higher accuracy of cost estimation that were
stated by [9] were met. Consequently, costs and time of development was reduced.
The first experiments with industrial COG lead to trend-setting results for the process. It was
demonstrated that conversion of organosulphur compounds is successful with real process
gas, i.e. outlet concentration is below 7 mol ppm. This is shown exemplary in Figure 5 where
COS and CS2 were measured by FT-IR in the product stream and in the reactant stream
consecutively. The graph indicates that both components are present in the reactant stream
(i.e. in COG) and almost completely react in the fixed-bed. Hence, the first step to alternative
usage of COG is successful.
11
Figure 5. Concentrations of COS and CS2 in product stream and reactant stream measured with FT-IR
over time.
Secondly, a large rise of the temperature (ΔT > 100 K) was observed in the reactor (Figure 6).
This leads to the conclusion that beside the desired main reactions exothermic side reactions
do occur. Gas analysis by FT-IR and the oxygen sensor showed that oxygen and olefins are
involved in the reactions. Moreover, it was found out that by increasing the GHSV, the rise of
temperature is even higher. This indicates a prompt reaction kinetic of the side reactions.
The conversion of organosulphur compounds is increased by higher temperatures [10].
However, it was shown that the reaction is relatively slow. Therefore, the GHSV can only be
raised up to a point where a sufficient residence time in the reactor is still guaranteed.
Hence, to find optimal operating conditions, a second measurement campaign has to be
conducted where the mutual dependency of GHSV, inlet temperature and exothermic side
reactions have to be systematically investigated.
Due to constantly varying concentrations of all components in COG it is also necessary to
conduct further investigations to identify the relevant side reactions. Lab experiments under
well defined conditions are advantageous in this case to avoid the influence of the multiple
components in COG.
9:17 9:46 10:14 10:43 11:12 11:41
con
cen
trat
ion
[p
pm
]
time
CS2
COS
produkt stream
reactant stream
produkt stream
12
Figure 6. Temperature profile in the reactor over time by adding COG feed. Measurement points are inlet
temperature and behind beds 1 and 2 (Figure 3)
CONCLUSION
By constructing the mobile test plant in a modular set-up, the adaption to the new process of
hydrolysis of organosulphur compounds in COG was time and cost saving. Most
modifications were due to higher safety risk of the gas composition; however most of the
equipment of the initial plant was reused. The test plant was taken into operation on the
coking plant ArcelorMittal Bottrop GmbH and a first measurement campaign was conducted.
The results were trend-setting for the process: it was demonstrated that conversion of the
organosulphur compounds COS and CS2 was sufficient with industrial gas. Moreover, it was
found that exothermic side reactions occur that have a relevant impact on the process
temperature. This knowledge is used to plan a second measurement campaign in which
optimal operating conditions will systematically be investigated.
ACKNOWLEDGEMENTS
The authors kindly thank ArcelorMittal Bottrop GmbH for their hospitality and the
opportunity to conduct experiments with industrial gas. A special thank to the cokers from
Bottrop for their help during our stay.
0 0.5 1 1.5 2 2.5 3
tem
per
atu
re [
°C]
inlet
behind bed 1
behind bed 2
0 0.5 1 1.5 2 2.5 30
0.5
1
time [h]
vol.
frac
tio
n C
OG
[-]
13
REFERENCES
[1] Müller, M., et al., Experimental investigations on Biodiesel as an alternative absorbent for
the recovery of aromatic hydrocarbons under industrial conditions, Distillation Absorption
(2010).
[2] G. Zhang, Y. Dong, M. Feng, Y. Zhang, W. Zhao, H. Cao, CO2 reforming of CH4 in coke
oven gas to syngas over coal char catalyst, Chemical Engineering Journal 156 (2010) 519–
523.
[3] J.M. Bermúdez, B. Fidalgo, A. Arenillas, J.A. Menéndez, CO2 reforming of coke oven
gas over a Ni/γAl2O3 catalyst to produce syngas for methanol synthesis, Fuel 94 (2012)
197–203.
[4] R. Günther, J. Schöneberger, H. Thielert, G. Wozny, Process development and catalyst
testing under industrial conditions, Clean Techn Environ Policy (2014) 1-11.
[5] R. Günther, Effektive Prozessentwicklung durch modulare Versuchsanlage im realen
Prozessverbund, Dortmund, 2012.
[6] M.V. Twigg, M.S. Spencer, Deactivation of supported copper metal catalysts for
hydrogenation reactions, Applied Catalysis A: General 212 (2001) 161–174.
[7] J.C. Schöneberger, H. Thielert, Koksofengas - Zu wertvoll zum Verheizen?, Chemie
Ingenieur Technik 8 (2012) 1235.
[8] J.C. Schöneberger, Entwicklung und Analyse katalytischer Abgasbehandlungsprozesse am
Beispiel einer emissionsfreien Schwefelsäureanlage, Dissertation, TU Berlin, Berlin,
2010.
[9] Ł. Hady, G. Wozny, Multikriterielle Aspekte der Modularisierung bei der Planung
verfahrenstechnischer Anlagen, Chemie Ingenieur Technik 84 (2012) 597–614.
[10] M. Kilian, G. Wozny, Optimize your search for sulfur-recovry technology, Hydrocarbon
processing special report (2003) 45–49.
15
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
HYDROFORMYLATION OF 1-DODECENE IN MICROEMULSIONS: LONG-TERM
MINI-PLANT OPERATION RESULTS
D. MÜLLER (1); E. ESCHE (1); T. POGRZEBA (2); T. HAMERLA (2); T. BARZ (1); R.
SCHOMÄCKER (2); G. WOZNY (1)
(1) Technische Universität Berlin, Chair for Process Dynamics and
Operations, Sekr. KWT9, Straße des 17. Juni 135, D-10623 Berlin,
Germany
(2) Technische Universität Berlin, Department for Chemistry, Sekr.
TC8, Straße des 17. Juni 124, D-10623 Berlin, Germany
Abstract. Within the Collaborative Research Center InPROMPT / TRR 63 in
Germany the process concept known as hydroformylation in microemulsions is
investigated. The aim is to perform the hydroformylation of long-chain alkenes
while utilizing water-soluble rhodium-based catalysts. The surfactant enables the
mixing of two immiscible liquids and after the reaction the separation of the two.
To perform the proof of concept, a mini-plant consisting of a reactor-decanter
system has been constructed at TU Berlin. In this contribution, results from long-
term miniplant operations are shown. The plant was operated continuously for 130
and 100 hours. The phase separation was successfully kept stable for a long time
period resulting in a rhodium loss of below 0.01% in the oily product phase. In the
second operation, the selectivity of the hydroformylation reaction was found to be
98:2 i.e. linear vs. branched aldehydes. No byproducts were observed during the
entire time span.
Keywords. Hydroformylation, mini-plant, microemulsions, operation
16
MOTIVATION AND INTRODUCTION
Hydroformylation producing aldehydes from alkenes is one of the most important industrial
reactions and usually realized as a homogeneous catalytic process. The hydroformylation
reaction itself is the coincidental addition of carbon-monoxide and hydrogen to an alkene.
Hereby, either a linear or a branched aldehyde is produced. The original hydroformylation
reaction was discovered and patented by Otto Roelen in 1938 [1]. In Fig. 1 the general
reaction equation is shown.
Figure 1. Reaction equation of the hydroformylation reaction [2]. R is an alkyl group.
A process concept currently under investigation in the Collaborative Research Center
InPROMPT / TRR 63 in Germany is concerned with the hydroformylation of long chain
alkenes in microemulsions. Herein, the issues of reaction selectivity and catalysts recyclability
are tackled by using ligand-modified water-soluble rhodium catalysts in combination with a
surfactant in the reaction system. The surfactant increases the miscibility between the
nonpolar long chain alkene and the aqueous catalyst solution, thus enabling the reaction.
Afterwards, as discussed in [3, 4], the phase separation characteristics of oil-water-surfactant
systems are exploited to separate and recycle the catalyst from the oily product. Fig. 2
displays the general concept.
Figure 2: Process concept for the hydroformylation of long chain olefins in microemulsions [5].
To investigate the technical and economical viability of this concept, a mini-plant has been
built. In this contribution, results from two long-term mini-plant operations are presented.
Special attention is given to the reaction selectivity and the stability of the separation. For this
purpose, the mini-plant will shortly be introduced.
17
MINI-PLANT AT TECHNISCHE UNIVERSITÄT BERLIN
As described in [6], the plant consists of three sections, a feed section, a reaction and
separation section, and a product storage section. Fig. 3 shows a simplified P&ID of the mini-
plant at Technische Universität Berlin. The plant is automated with the process control system
PCS7, sponsored by SIEMENS. In the process control room the operators receive information
from over 50 sensors. In the plant, 15 manipulable variables exist. Among these are pump set
points for liquid mass flows, control valves for product and gas streams, and temperature
control via thermostats.
The feed section mainly consists of feed tanks and feed pumps for all liquids as well as gas
bottles. In Fig. 3, the three feed pumps are labelled 4a, 4b, and 4c. Pump 4a, feeding the
alkene, can feed up to 1500 g/h. The feeds of the catalyst solution (4b) and non-ionic
surfactant (4c) can be regulated to a maximum value of 500 g/h each.
The reaction and separation section consists of a reactor and a decanter. The former has a
maximum capacity of 1000 ml, whilst a drain is fixed at 70% capacity. The reactor features a
gassing stirrer with a maximum speed of 2880 rpm, usually operated at 1100 rpm. In contrast,
the decanter has a volume of 300 ml. The maximum flow of the recycle lies at 500 g/h.
Apart from online measurements for temperatures, pressures, flows, and levels, two offline
GCs are employed for hourly composition analysis. In addition, an ICP-OES is used for
estimating the rhodium amount in the product phase. The sample points of the liquids are
positioned at 11a and 11b shown in Fig. 3.
18
Figure 3: Simplified P&ID of the mini-plant at Technische Universität Berlin.
19
APPLIED SUBSTANCES
In InPROMPT, 1-dodecene (C12 alkene) is exemplarily used as an unsaturated, long chain
hydrocarbon. The syngas for the reaction has a composition of 1:1 vol.-% of CO:H2 with a
purity of 5.0. The applied catalyst consists of a rhodium based precursor [Rh(acac)(CO)2]
(CAS: 14874-82-9), sponsored by Umicore, and a water soluble ligand called SulfoXantPhos
(sulfonated form of XantPhos, CAS: 161265-03-8). The ligand was purchased from MOLISA
GmbH. The miscibility of the alkene and the catalyst solution is enabled by the non-ionic
surfactant Marlipal 24/70 (CAS: 68439-50-9), sponsored by Sasol Germany GmbH. The
desired product is the linear aldehyde 1-tridecanal. Expected byproducts during the reaction
are various isomers of tridecanal, the hydrogenated form of 1-dodecene, and n-dodecane.
MINI-PLANT OPERATION RESULTS AND DISCUSSION
The discussion of the mini-plant operation is divided between reaction and separation
behaviour.
Reaction Results
In a first operation, the mini-plant was operated for 130h. The reaction was initiated at
1.5MPa induced syngas (H2&CO) pressure. The reactor temperature hereby was constantly
set to 368K and the reactor residence time to slightly above one hour. This was achieved by
setting a constant alkene feed of 100g/h and recycle of 500g/h. Hereby, an expected yield of
14% was successfully obtained. The selectivity of the desired linear aldehyde vs. the
undesired branched aldehyde was around 93.3:6.6, which was lower than in previously
performed lab-scale batch experiments (98:2, [3]). During the operation, large amounts (up to
20wt.-%) of the undesired hydrogenation product n-dodecane were produced.
The selectivity of the reaction as well as minimization of byproduct production could
successfully be improved in a second mini-plant operation (100 hours), for which several
aspects of the plant set-up and set-points. First of all, the reaction temperature was lowered to
358K. Thus, the hydrogenation reaction should be avoided. Secondly, a gas purge was
installed to ensure a constant syngas renewal and thus avoid a shift of the gas composition.
Thirdly, the reactor residence time increased to two hours to increase the yield. Additionally,
a sample point in the reactor was added. The n- to iso-aldehyde selectivity during the entire
operation is at 100:0, meaning that the amount of iso-aldehyde lies below the detectability
limit of the GCs. The selectivity regarding all other byproducts is roughly 97:3 to 98:2. Fig. 4
shows the selectivity for seventy hours of operation. Obviously, n-dodecane production is
20
considerably smaller compared to the first mini-plant operation. Fig. 4 proves that the
catalyst-complex is stable as well as selective for a long period in time.
Figure 4: Selectivity of the catalyst towards tridecanal vs. all byproducts (iso-aldehyde, n-dodecane).
Phase Separation Results
In the first operation, the phase separation was systematically analyzed. As described in [5],
the ideal separation region for the applied mixture with the surfactant Marlipal 24/70 is fairly
small (roughly 4K). In this temperature interval the separation is fast and pure. In the regions
outside of this interval the separation is inapplicable for a continuous process. Thus,
controlling the phase separation for a long time period was challenging. Nevertheless, during
the operation, both the phase separation as well as the catalyst recycling were successfully
implemented. Fig. 5 exemplarily shows the separation by comparing the mass fraction of
tridecanal in the product and the water phase of the decanter. For over 24 hours the separation
was kept stable. The oily components in the product phase amounted to values between
95wt.-% and 99 wt.-%. Hereby 99.99% of the catalyst remained in the catalyst phase. This
mimicks lab-scale test-tube results.
21
Figure 5: Successful phase separation for over 24 hours showing a constant difference in concentration of
1-tridecanal between the oil and water phase as well as a minimal rhodium loss in the oil phase.
As mentioned before, the control of the separation in the decanter proved to be challenging,
since the ideal separation region shifts with temperature and changing concentrations of
reactants and products. Since the analysis of the GC samples took one hour, a
predetermination of the current state of the plant was difficult. Thus, a shift of the optimal
separation region could not be predicted. Fig. 6 shows the instability of the phase separation
due to these concentration changes. Obviously, with a non-functioning phase separation, the
rhodium loss increased drastically.
22
Figure 6: Instability of the phase separation during the first mini-plant operation.
CONCLUSIONS AND OUTLOOK
In this contribution, the results from two mini-plant operations testing the process concept of
hydroformylation in microemulsions were shown. Firstly, the proof of concept has been
made. The reaction, applying the water-soluble rhodium-based catalyst modified with
SulfoXantPhos as ligand is stable and remains selective towards the desired product for a long
time period. Using the technical-grade surfactant Marlipal 24/70, the phase separation was
also succesfully implemented resulting in minimal rhodium losses. Secondly, it must be
mentioned, that the phase separation region of the surfactant is very small, thus making the
stable operation challenging. The selection of an adequate surfactant, which functions well in
the reaction and has a large ideal phase separation area, is critical for the success of this
process concept.
For future research, a new decanter will be developed in which coalescence accelerators will
be used. With these, a faster and more controlable phase separation is to be enabled.
Furthermore, since the mini-plant is fully automated, advanced process operating strategies
are to be implemented. These are to be achieved with the help of dynamic models to obtain a
robust long-term operation of the plant.
23
ACKNOWLEDGEMENTS
This work is part of the Collaborative Research Center "Integrated Chemical Processes in
Liquid Multiphase Systems" coordinated by the Technische Universität Berlin. Financial
support by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) is
gratefully acknowledged (TRR 63). The authors gratefully acknowledge the support of the
company Umicore for sponsoring the rhodium catalyst “Acetylacetonatodicarbonylrhodium(I)
(CAS: 14874-82-9)” and the company Sasol for the surfactant used in the described
experiments. Finally, the authors also acknowledge the support of SIEMENS for sponsoring
the entire process control system SIMATIC PCS7 for the mini-plant.
REFERENCES
[1] Roelen, O. US Patent, (1943), 2.327.066.
[2] Kupka, J. A.: Hydroformylierung von 1-Octen in Mikroemulsion. Ph.D. Thesis,
Technische Universitat Braunschweig (2012).
[3] Hamerla, T.; Rost, A.; Kasaka, Y.; Schomäcker, R.: Hydroformylation of 1-dodecene
with water-soluble rhodium catalysts with bidentate ligands in multiphase systems,
ChemCatChem, 5, (2012) 7, 1854 - 1862. DOI: 10.1002/cctc.201200744.
[4] Müller, M.; Kasaka, Y.; Müller, D.; Schomäcker, R.; Wozny, G.: Process Design for
the Separation of Three Liquid Phases for a Continuous Hydroformylation Process in a
Miniplant Scale. Ind. Eng. Chem. Res. (2013), 52, 7259 – 7264. DOI:
10.1021/ie302487m.
[5] Müller, D.; Esche, E.; Müller, M.; Wozny, G. Development of a Short-Cut Model for
Three-phase Liquid Separation Dynamics for a Hydroformylation Mini-Plant,
Presentation at the AIChE 2012, Pittsburgh, USA.
[6] Müller, D.; Minh, D.H.; Merchan, V.A.; Arellano-Garcia, H.; Kasaka, Y.; Müller, M.;
Schomäcker, R.; Wozny, G.: Towards a novel process concept for the
hydroformylation of higher alkenes: Mini-plant operation strategies via model
development and optimal experimental design, Chem. Eng. Sci. (2013). In Press. DOI:
10.1016/j.ces.2013.05.022.
25
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
ENABLING ONLINE-OPTIMIZATION FOR A MULTIPHASE SYSTEM IN A
HYDROFORMYLATION MINI-PLANT
D. MÜLLER (1); M. ILLNER (1); A. FLECK (1); E. ESCHE (1); T. BARZ (1); R. SCHOMÄCKER
(2); G. WOZNY (1)
(1) Technische Universität Berlin, Chair for Process Dynamics and
Operations, Sekr. KWT9, Straße des 17. Juni 135, D-10623 Berlin,
Germany
(2) Technische Universität Berlin, Department for Chemistry, Sekr.
TC8, Straße des 17. Juni 124, D-10623 Berlin, Germany
Abstract. In the Collaborative Research Center InPROMPT / TRR 63
hydroformylation in microemulsions is investigated. For this purpose, a mini-
plant has been constructed at Technische Universität Berlin. This process concept
has proven to be challenging from an operation point of view. Therefore, model-
assisted process operating strategies are applied, to control the process. In this
contribution, the required infrastructure for implementing dynamic online-
optimization is discussed and carried out in the mini-plant at Technische
Universität Berlin. The successful implementation during a mini-plant operation is
shown and open challenges are discussed.
Keywords. Online-Optimization, Hydroformylation, Mini-plant
INTRODUCTION AND GOAL
Homogeneous catalytic reactions offer important advantages over their heterogeneous
counterparts. However, the commercial feasibility depends on the ability to separate the
reaction product from the often valuable catalyst and reaction solvent [1]. One possible
approach is the use of a multiphase system, where the reactants and the catalyst are dissolved
in different phases. Whilst the reaction is carried out in a continuously stirred tank reactor to
ensure maximum contact between the substrate and catalyst, a settler is used to remove the
catalyst from the product by decanting. The catalyst can then be recycled to the reactor.
26
In the Collaborative Research Center InPROMPT / TRR 63 a mini-plant has been
constructed to investigate the continuous operation of a multiphase liquid-liquid system for
the hydroformylation of long chain olefines using a rhodium-based catalyst. The continuous
operation of multiphase systems with phase separation is a challenging task from a process
control point of view. Often, these systems consist of a multitude of components, thus making
thermodynamic modeling of phase equilibria challenging. This is especially the case for oil-
water-surfactant systems. As discussed in [2], the bottleneck for a continuous mini-plant
operation is the stabilization of the phase separation. Therefore, model-assisted process
operating strategies are taken into consideration.
In this contribution a mini-plant model for the hydroformylation of long-chain alkens
in micro-emulsions is presented, the hierarchical automation concept for online-optimization
is discussed, and details on the actual implementation in the mini-plant are shown.
BACKGROUND INFORMATION
In this section some background information on the system in hand is given. The component
system is discussed, the mini-plant is described, and the phase separation is characterized.
Analyzed Multiphase System: Hydroformylation in Micro-Emulsions
Hydroformylation is an important reaction in the chemical industry for producing aldehydes
from short chain alkenes. The rhodium-assisted production of long chain aldehydes is
currently not practiced, mainly due to the lack of applicable concepts covering both, efficient
catalytic reactions and catalyst recycling. A promising concept is hydroformylation in micro-
emulsions. Here, a surfactant acts as a mediator between the hydrophobic alkene and
hydrophilic catalyst solution. The reaction is initiated by feeding syngas (H2&CO) into the
mixture. After the reaction, a phase separation is initiated to remove the hydrophobic products
from the system and to recycle the hydrophilic catalyst for the reaction. In order to test the
technical feasibility of this process concept, a mini-plant has been built at TU Berlin.
Mini-plant at Technische Universitat Berlin
Fig. 1 shows the mini-plant at TU Berlin. It is divided into three parts: a feed section for gas
and liquid feeds, a mixer-settler section, and the product storage section.
27
Figure 1. Hydroformylation mini-plant at TU Berlin: simplified flow sheet used for modeling (left) and 3D
image and photo of the mixer-settler section of the mini-plant (right) [2].
The aim is to control the decanter via online optimization. In the product stream of the
decanter, a Raman laser is installed to observe the concentration changes and thus manipulate
the temperature of the thermostat connected to the decanter. Of high relevance for operating
the process and especially the decanter in a stable manner is the knowledge regarding the
phase separation states.
Phase Separation Characteristics of Surfactant-Oil-Water Systems
Fig. 2 shows a general image of Kahlweit’s fish describing the possible separation states [3].
The applicability of the different phase separation states for a mixer-settler process varies.
Figure 2. Qualitative image of Kahlweit’s fish from [3] indicating the different possible separation states
of oil-water-surfactant systems. The lower 2-phase and the 3-phase state are desired for the separation
step of the process.
At moderate surfactant concentrations and low temperatures, a two phase region is formed (a
pure oil and a water/surfactant rich phase). The surfactant is mainly dissolved in the lower
28
phase. This separation region is desired for the product separation step in the process as a pure
product phase can be removed. If the temperature is increased, the three phase region is
established. The lower phase mainly consists of water and the surfactant is suspended in the
middle phase. Just as in the lower two phase region, an oil-rich top phase is created, making
this region also applicable for the separation step.
The other two regions in Kahlweit’s fish are undesired. These are the single-phase and the
upper two-phase regions. The upper two-phase region lies above the three-phase region and
the surfactant is dissolved in the top phase. Thus, a separation would lead to greater surfactant
and catalyst loss with the product phase.
The challenging aspect of operating a process as described above is that the ideal separation
region shifts depending on the composition. During the operation of the process, several shifts
occur simply due to concentration changes, i.e. when the reaction starts and product is
produced, when the surfactant concentration changes due to surfactant leaching, or when the
water to oil ratio moves. Therefore, a stable operation requires the assistance of a model,
which is speedily updated with information.
MINI-PLANT MODEL DEVELOPMENT
Performing an online-optimization for a non-stationary process requires a dynamic model of
the mini-plant. In this work, simplified first principle models are used to describe the dynamic
behavior of the plant units and peripheral equipment: the reactor, involving kinetic equations
determined in [4], the decanter, using a derived empirical model to calculate optimal
separation conditions, and auxiliary equipment such as vessels, pumps, and valves.
The general mini-plant model consists of 16 units, 14 components and is modeled in the web-
based modeling environment MOSAIC [5]. For this purpose, first principle equations
regarding the units are set up. The model is discretized with orthogonal collocation on finite
elements using fourth order Lagrangian polynomials on shifted Radau roots. That way a
nonlinear system consisting of 16 times 2540 equations is created. The entire model is written
in AMPL using IPOPT as a solver. Regarding the objective two units are modeled in greater
detail: the reactor and the decanter.
29
Rigorous Reactor Model
The reaction network (Fig. 3) as well as the structure of the reaction kinetics implemented
into the CSTR model stem from [4]. The network, if systematically set up, should consist of
seven reactions. The desired reaction is r1. In experimental investigations though, reaction 2
could not be observed.
Figure 3. Reaction Network based on [4].
The reactions displayed in Fig. 3 can be described as follows: The rate for 1-dodecene is
shown in Eq. (1) and for its isomers in Eq. (2), for dodecane in Eq. (3), for n-tridecanal in Eq.
(4), and for iso-aldehyde in Eq. (5).
54311 rrrrr dodecene
(1)
7643 rrrrr dodeceneIso
(2)
65 rrr dodecanen
(3)
11 rr tridecanal
(4)
7rr tridecanaliso
(5)
The kinetic equations for the reactions presented above stem from [4] and are displayed in Eq.
6 to 11. The gas solubilities are obtained using a linearized form of Henry’s law regarding
temperature and pressure dependency.
COdodecenedodecene
HCOdodecene
ref
A
ref
CO
cat
ccKcK
cccTTR
Ek
cK
cr
11,11,
1
1,
1,
11
11exp
1
2
(6)
dodecene
dodecene
ref
A
ref
CO
cat
cK
cTTR
Ek
cK
cr
13,
1
3,
3,
31
11exp
1
(7)
30
dodecene
dodeceneiso
ref
A
ref
CO
cat
cK
cTTR
Ek
cK
cr
14,
4,
4,
41
11exp
1
(8)
COdodecenedodecene
HCOdodeceneiso
ref
A
ref
CO
cat
ccKcK
cccTTR
Ek
cK
cr
11,11,
5,
5,
51
11exp
1
2
(9)
dodecene
Hdodecene
ref
A
ref
CO
cat
cK
ccTTR
Ek
cK
cr
16,
1
6,
6,
61
11exp
1
2
(10)
dodecene
Hdodeceneiso
ref
A
ref
CO
cat
cK
ccTTR
Ek
cK
cr
17,
7,
7,
71
11exp
1
2
(11)
The reaction parameters of this network are fitted based on non-published lab experiments
performed by Tobias Hamerla at Technische Universität Berlin. This reactor model is then
combined, with a multiphase separation model of the decanter.
Decanter Modeling: Multiphase Separation
An empirical model is developed for the phase separation, due to lack of thermodynamic data.
To systematically prepare a model for optimization purposes, the workflow discussed by
(Esche et al., 2013) is applied. The aim is to identify the feasible operating area for the
process. Hence, several experiments are carried out. Hereby, the relevant variables of the oil
to water ratio (α), surfactant mass fraction (γ), and reaction conversion (X) are taken into
consideration. These ratios are shown in Eq. 13 to 15, whereby mi is the mass of the
component i.
surfactantwaterolefin
surfactant
mmm
m
(12)
waterolefin
olefin
mm
m
(13)
olefinaldehyde
aldehyde
mm
m
(14)
31
As discussed in [2], an area exists in which a sufficiently large oil phase is established (Fig.
4). Two surfaces are visible, which represent the phase boundaries between the three possible
phases. The ideal operating area is the region in the upper surface where the “valley” is
created. Two functions are formulated, which describe the shift of the upper and the lower
temperature bound in which an oil phase height of at least 20 vol.-% is reached after 20
minutes of separation (schematically shown in Fig. 4 on the right).
Figure 4. Phase separation: experimental results (left) from [2] as well as the desired lower and upper
temperature bounds (right).
ONLINE-OPTIMIZATION PLATFORM
An online-optimization platform is developed and tested, to achieve a stabilization of the
multi-phase system. The workflow of the online-optimization can be divided into several
steps. Firstly, data from the plant is transferred to the process control system (PCS) and is
stored. Secondly, the gathered data is sent to the optimizer at a specific, user-defined point in
time. After a consistency check this data is used for the setting of design values in the model
to determine the current state of the plant. In the fourth step, a simulation is carried out to
predict the behavior of the plant for the next hours. Afterwards, these results are used as initial
values as well as starting values for the dynamic optimization of the system. If this
optimization is successful, a matrix of set-points for the controllers of the process is sent back
and imported into the PCS of the plant. Here, the operator has the possibility to check the
results for consistency and to implement them in the process. The communication scheme is
shown in Fig. 5.
32
Figure 5. Online-optimization scheme implemented into the hydroformylation mini-plant at Technische
Universität Berlin.
These steps can be repeated, as soon as new information regarding the concentrations in the
process is available. Thus, the model-predicted state of the plant is repeatedly refitted to the
new operating point and the optimization results are adjusted accordingly.
The plant at Technische Universität Berlin is fully automated with the process control system
Siemens PCS7 and therefore suitable for the integration of online-optimization. To enable the
communication between optimization solver, process control system, and plant, a Matlab
script has been written to perform the optimization and communication routine. For defined
time steps plant data will be exported from the process control system. This data is then used
to update the model regarding the current plant state. Subsequently a simulation of the current
plant state is carried out, to generate initial values for the next optimization step. In total,
seven optimization variables exist: The feed, recycle, and outlet streams, as well as the reactor
and decanter temperature. Fig. 6 shows the matrix for three controls (feed streams 1 - 3) for a
total time of one hour. The matrix is divided into eight time slots, which can each be set
individually depending on the optimization result. As soon as the operator activates the
optimization framework, the set points are implemented into the controllers of the plant and
the countdown for duration of these set-points is started. Afterwards, the next set-points in the
matrix are implemented.
33
Figure 6. Set point matrix implementation into the process control system. Top: initial time sequence
lasting 300s, bottom: start of the second time sequence.
Fig. 7 depicts the time-procedure of the optimization routine, covering a (discretized) time
span of 4 hours of plant operation. Considering the initially availability of a set of optimized
control set points for feed flow rates and temperatures of reactor and decanter, the plant will
be run with this solution for 4 hours. However, to cover deviations and dead times for sample
analysis and calculations, resimulations are needed to achieve the next optimization solution.
Therefore, a whole set of measurements is taken after 2 hours. These values are used to
update the current state of the plant model (t=2h). Further, including the optimized controls
for t=2-4h, a resimulation is carried out to predict an updated plant model state at t=4h. Using
these as initial values, an optimization result of the next 4 hour time period can be calculated
and implemented towards the end time of the first optimization solution.
34
Figure 7. Time-procedure of the optimization routine.
To test the functionality of the implemented online-optimization platform, the mini-plant was
operated continuously for 100 hours. Hereby, a pressure of 15 bar and a temperature of 85°C
was initiated in the reactor. The results of the online-optimization are shown in Fig. 8. Here,
the mass fraction of the summed up oily components in the product (oil phase) and recycle
stream (water phase) are displayed. Initially, a near to perfect separation was achieved at
which 100 wt.-% oil was measured in the product stream for the first 10 hours of operation.
The recycle stream on the other hand showed minimal oil concentrations, which is highly
desirable in terms of economic viability of the process concept.
As the reaction was initiated though, the phase separation broke down. This is apparent after
roughly 10 hours of operation. The concentrations of oily components in the oil and water
phase are more or less equal. This shows that the model for the phase separation needs to be
further developed regarding the influence of the catalyst activation during the reaction.
However, it can be said, that the infrastructural implementation of the online-optimization
platform functions very well. Improvements of the reaction models must be performed
though, to guarantee optimal operation for a longer time span.
35
Figure 8. Mini-plant operation showing the cumulated concentrations of oily components in the recycle
stream (water phase) and the product stream (oil phase).
CONCLUSIONS AND OUTLOOK
In this contribution, the implementation of an online-optimization platform in a
hydroformylation mini-plant was shown. The steps of data transmission, optimization, and
result implementation were successfully tested during a mini-plant run. In the future,
modifications of the model will be introduced, to correctly describe the behavior of the system
during the reaction.
ACKNOWLEDGEMENTS
This work is part of the Collaborative Research Center "Integrated Chemical Processes in
Liquid Multiphase Systems" coordinated by the Technische Universität Berlin. Financial
support by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) is
gratefully acknowledged (TRR 63). The authors gratefully acknowledge the support of
Umicore N.V. for sponsoring the rhodium catalyst “Acetylacetonatodicarbonylrhodium(I)
(CAS: 14874-82-9)” and the Sasol GmbH for the surfactant used in the described
experiments. Finally, the authors also acknowledge the support of Siemens AG for sponsoring
the entire process control system SIMATIC PCS7 for the mini-plant.
36
REFERENCES
[1] Cole-Hamilton, D. J.: Homogeneous Catalysis - New Approaches to Catalyst
Separation, Recovery, and Recycling. Science (2003) 299.5613, 1702-1706.
[2] Müller, M.; Kasaka, Y.; Müller, D.; Schomäcker, R.; Wozny, G.: Process Design for
the Separation of Three Liquid Phases for a Continuous Hydroformylation Process in a
Miniplant Scale. Ind. Eng. Chem. Res. (2013), 52, 7259 – 7264. DOI:
10.1021/ie302487m.
[3] Rost, A.; Müller, M.; Hamerla, T.; Kasaka, Y.; Wozny, G.; Schomäcker R.:
Development of a continuous process for the hydroformylation of long chain olefins in
aqueous multiphase systems, Chem. Eng. Process. Process Intensif. (2012), 67, 130 –
135.
[4] Kiedorf, G.; Hoang, D.M.; Müller, A.; Jörke, A.; Markert, J.; Arellano-Garcia, H.;
Seidel-Morgenstern, A.; Hamel C.: Kinetics of 1-dodecene hydroformylation in a
thermomorphic solvent system using a rhodium-biphephos catalyst. Chem. Eng. Sci.
(2013). In Press. DOI: 10.1016/j.ces.2013.06.027.
[5] Kuntsche, S.; Arellano-Garcia, H.; Wozny G.: MOSAIC, an environment for web-
based modeling in the documentation level, Comp.-Aided Chem. Eng. (2011), 29,
1140 – 1144.
37
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
EFFECT OF THE GEOMETRICAL AND TECHNOLOGICAL PARAMETERS ON THE
FLOCS SIZE IN THE CONTINUOUS TUBULAR REACTOR WITH BAFFLES
W. SZATKO (1); M. DYLĄG (2); J. KAMIEŃSKI (1); J. ROSIŃSKI (1); J. TALAGA (1); R.
WÓJTOWICZ (1)
(1) Institute of Thermal and Process Engineering, Faculty of
Mechanical Engineering, Cracow University of Technology, 31-
155 Kraków, ul. Warszawska 24
(2) Institute of Advanced Manufacturing Technology, Cracow, 30-011
Kraków, ul. Wrocławska 37A
Abstract. This study presents the results of experimental research of the
flocculation process in a tubular reactor with baffles. The variable parameters of
the research were: flocculant concentration, suspension flow intensity and the
geometry of the reactor baffles. The study presents the measurements of the size
of produced agglomerates, statistical analysis of the size of the flocs and, on the
base of the above, the size increase rate of the flocs in three reactors with different
designs. Guidelines for the design and operation of reactors were formulated in
relation to the experimental and model research of the liquid flow rate through the
reactor. The study is a part of a wide research project on the description of the
formation and decomposition of the solid body aggregates in multi-phase
dispersed systems.
Keywords. flocculation, agglomeration, tubular reactor with baffles
INTRODUCTION
The processes of forming the agglomerates of solid bodies in liquids (flocculation) are one of
the most complex phenomena found in nature or technology. They are found in single
operations of various industries: chemical, food, pharmaceutical, cosmetics, biotechnology,
mineral processing, power and heat generation and environment protection, however they are
38
of special importance in surface water treatment – removal of colloids and slowly settling
suspensions. During the flocculation process, the pollution particles combine into
agglomerates (flocs) of variable shapes and properties [1]. Water treatment processes use
equipment with periodic functioning. Agglomeration is defined as a process whereby small
solid particles combine into aggregates – complex and heterogeneous three-dimensional
structures of joined original particles.
The purpose of the study is the analysis of the process of agglomerates generation during
continuous flocculation in a tubular reactor with baffles. The analysed flocculation case
pertains to a system in which the interaction of particles and floccules – leading to their
aggregation or decomposition – in fact, appearing simultaneously; in a state of a dynamic
quasi-equilibrium [2]. Based on the microscope measurements, the characteristic parameters
describing the size and structure of generated agglomerates were selected and the results –
presented as empirical distribution histograms – were approximated using continuous
theoretical distributions.
Flow rate distribution verification tests by using the LDA method for the flow simulations
with the CFD method were also conducted. This part of the research was concerned with
assessing the rate of turbulent kinetic energy dissipation and flow stability in relation to the
elements of the kinetics of the processes of aggregation and agglomerate decomposition in the
tubular reactor with baffles.
RESEARCH INFRASTRUCTURE
Experimental research was conducted in the Department of Industrial Apparatus on a
universal research station presented in Fig. 1. The station allows the measurements of the
generated flocs for various process parameters and testing the rate distribution using the LDA
method.
39
sample module C
C1
C2
water
solid
NaOH, CaCl2
pH
T
water
SK 16
focculant
A5
A6
F
A2A3A4
suspension
A1
A8
A7
module A
A2module BB2 B1 B3
module E
E1
p
D1
D2 D3X,YLDA
module D
Fig. 1. Measurement stand diagram
module A: A1 – suspension tank, A2 – suspension pump, A3 – pulsation damper, A4 – flow meter, A5 –
flocculant tank, A6 – flocculant pump, A7– temperature meter, A8 – pH meter, module B: B1 – tubular
reactor with baffles, B2 – baffle, B3 – suspension sampler, module C: C1 – digital camera, C2 – optical
microscope, module D: D1 – pressure drop meter, D2 – LDA probe and optical system, D3 – LDA probe
position meter, module E: E1 – computer for data analysis and archiving.
The primary functional systems of the measurement station are: suspension preparation
system (module A), mixing system (module B) as well as the analysis and logging systems –
of the size of the generated flocs (module C) and the flow rate of the liquid through the
tubular reactor (module D).
The suspension preparation system (module A) allows the preparation of the suspension of
quartz grains in distilled water – detailed specifications on the properties of original particles
are presented in [3]. For the purposes of the study, an anionic flocculant- Sokoflok (SK16) -
was used, typical for water treatment processes. This required the change of control of the pH
indicator by using water solutions of NaOH and CaCl2.
The mixing system (module B) was composed of the tubular reactor (B1) with baffles (B2) as
presented in Fig. 2.
40
a)
260 36 260
700
3
6
2 36
b)
h=1/3D h=1/2D h=2/3D Fig. 2. Tubular reactor a) reactor geometry, b) baffles geometry
The station was equipped with the analysis and logging systems – of the size of the generated
flocs (module C) and the flow rate of the liquid through the tubular reactor (module D) as
well as analog-digital converters and special software.
RESEARCH METHODOLOGY AND PARAMETERS
The study on the flocculation process in a tubular reactor was conducted for two flocculant
concentrations in the suspension; c = 2 and 4 ml/dm3, three baffle widths; 1/3D, 1/2D, 2/3D
and four flow rates of the suspension through the reactor – Re = 2000, 3000, 4000 and 5000.
Module C was used to analyse the structure, dimensions and size distribution of the generated
flocs in different sections of the continuous tubular reactor – the pictures of the samples (from
the outlet port or behind second, fourth, sixth baffle) were taken with a digital camera and an
optical microscope. The pictures were analysed with a special image analysis software
package – Image Pro Plus 5.1. [4]. Based on the obtained measurement data and using the
Statistica 9.0 [5] software, empirical histograms were created for the distribution of analysed
parameters with further approximation using the known, continuous theoretical distributions
[6,7]. The histogram for the mean floccule area was approximated by the Rayleigh
distribution; for the Feret’s diameter – by the logarithmic-normal distribution and the
histogram for the fractal dimension distribution – by the normal distribution. The results are
presented in Fig. 3 and Table 1. The measures of the connection “quality” between the
theoretical distribution and the experimental data were the compliance tests: chi-square and
Kolmogorov-Smirnov [5].
Module D of the measurement station, Fig. 1, was used to determine the velocity field of
suspension flow through the tubular reactor (B1) – measurements taken with the LDA
method, using a two-channel laser doppler anemometer (D2) [8,9]. Based on the results of
41
measurements of momentary liquid flow rates, the mean flow rates were determined as well
as the fluctuation velocity components and the kinetic energy of the turbulence.
The obtained empirical results of the flow velocity field distribution were compared with
the results of CFD digital modelling of the rate of liquid flow through a tubular reactor with
baffles. The geometrical model of the reactors and numerical grid were created in the Gambit
2.4 preprocessor. The selection of a turbulence model adequate to the characteristics of the
suspension flow through the reactor was a serious research problem. The nature of the
problem was the fact that the turbulent stress tensor introduces six additional unknowns into
the Reynolds equations, resulting in the equations forming an open system and the lack of
physical indications of the relation between turbulent stress and other characteristic features
of the liquid forms a problem for setting additional six, three or two equations. Decision was
made to use empirical models, where the model suitability is determined by the simplicity or
even possibility of measuring the introduced coefficients. In result, the flow of the liquid in
the apparatus was described mathematically by standard Reynolds-averaged Navier–Stokes
equations of mass and momentum transport, whereas the turbulence was modelled using the
classic standard turbulence model k- [10,11,12,13].
RESEARCH RESULTS
The distributions of Feret’s diameter and area are characterised by right-side asymmetry and
unimodality with mode in the lower distribution classes. The best matches were obtained by
Rayleigh distribution for the area and logarithmic-normal for Feret’s diameter. In the case of
the fractal dimension, the distributions are characterised by symmetry. The mode for the
population was determined closer to the mid-range of the histogram and the approximation of
the experimental data used the Gaussian theoretical normal distribution.
Selected data of floc sizes generated in the tubular reactor presented in Fig. 3 show that in
general – the increase in the baffle size results in the increase in the mean size of the flocs –
detailed data presented in Table 1.
The effect of the rate of suspension flow through the reactor is not quite as unequivocal and
depends on the value used to describe the size of the generated floc – turbulence energy
dissipation seems to be the key parameter. For the fractal dimension, the increase in the flow
rate results in the increase of size, for the mean area and Feret’s diameter, the largest values
were obtained for the lowest flow intensities.
42
aver
age
Fer
et’s
dia
met
er
μm
44
,9
D
opa
so
wan
ie:
roz
kład
log
no
rma
lny
Śre
dn
ica F
ere
ta =
16
40*2
2,8
30
3*lo
gn
orm
(x; 3
,63
69
; 0,5
36
6)
57
%
25
%
9%
5%
2%
1%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
14,2
00
8
59,8
61
4
105
,52
21
151
,18
27
196
,84
33
24
2,5
039
28
8,1
646
33
3,8
252
37
9,4
85
8
Śre
dnic
a F
ere
ta
m
0%
6%
12
%
18
%
24
%
30
%
37
%
43
%
49
%
55
%
61
%
Procent obserwacji D
opa
so
wan
ie:
roz
kład
log
no
rma
lny
Śre
dn
ica F
ere
ta =
16
40*2
2,8
30
3*lo
gn
orm
(x; 3
,63
69
; 0,5
36
6)
57
%
25
%
9%
5%
2%
1%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
14,2
00
8
59,8
61
4
105
,52
21
151
,18
27
196
,84
33
24
2,5
039
28
8,1
646
33
3,8
252
37
9,4
85
8
Śre
dnic
a F
ere
ta
m
0%
6%
12
%
18
%
24
%
30
%
37
%
43
%
49
%
55
%
61
%
Procent obserwacji
Fe
ret’s d
iam
ete
r µ
m
log
no
rma
l d
istr
ibu
tio
n
observations
Re
= 5
00
0, 1
/3D
54
,2
D
opa
so
wanie
: ro
zkł
ad logn
orm
aln
y
Śre
dnic
a F
ere
ta =
134
8*5
4,5
584
*log
no
rm(x
; 3,7
254; 0,6
40
6)
81%
12%
4%
2%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
15
,891
8
125
,0086
23
4,1
25
4
34
3,2
422
452,3
590
561
,4758
670
,5926
779
,709
4
88
8,8
262
Śre
dnic
a F
ere
ta
m
0%
15
%
30
%
45
%
59
%
74
%
89
%
104
%
119
%
Procent obserwacji D
opa
so
wanie
: ro
zkł
ad logn
orm
aln
y
Śre
dnic
a F
ere
ta =
134
8*5
4,5
584
*log
no
rm(x
; 3,7
254; 0,6
40
6)
81%
12%
4%
2%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
15
,891
8
125
,0086
23
4,1
25
4
34
3,2
422
452,3
590
561
,4758
670
,5926
779
,709
4
88
8,8
262
Śre
dnic
a F
ere
ta
m
0%
15
%
30
%
45
%
59
%
74
%
89
%
104
%
119
%
Procent obserwacji
Fe
ret’s d
iam
ete
r µ
m
log
no
rma
l d
istr
ibu
tio
n
observations
Re
= 5
00
0, 1
/2D
66
,9
D
opa
so
wan
ie: ro
zkła
d lo
gn
orm
aln
y
Śre
dn
ica F
ere
ta =
271
*25
,517
5*l
og
no
rm(x
; 3
,95
22
; 0,6
58
6)
49%
21
%
12%
5%
4%
2%
1%
2%
1%
1%
1%
0%
0%
0%
0%
0%
17,5
828
68
,61
77
11
9,6
52
6
17
0,6
875
221
,72
24
272
,757
3
32
3,7
92
3
37
4,8
272
42
5,8
621
Śre
dn
ica F
ere
ta
m
0%
7%
15%
22%
30%
37%
44%
52%
Procent obserwacji D
opa
so
wan
ie: ro
zkła
d lo
gn
orm
aln
y
Śre
dn
ica F
ere
ta =
271
*25
,517
5*l
og
no
rm(x
; 3
,95
22
; 0,6
58
6)
49%
21
%
12%
5%
4%
2%
1%
2%
1%
1%
1%
0%
0%
0%
0%
0%
17,5
828
68
,61
77
11
9,6
52
6
17
0,6
875
221
,72
24
272
,757
3
32
3,7
92
3
37
4,8
272
42
5,8
621
Śre
dn
ica F
ere
ta
m
0%
7%
15%
22%
30%
37%
44%
52%
Procent obserwacji
Fe
ret’s d
iam
ete
r µ
m
log
no
rma
l d
istr
ibu
tio
n
observations
Re
= 5
00
0, 2
/3D
Fig
. 3
. C
om
pa
riso
n o
f th
e m
ean
flo
c d
imen
sio
ns
for
thre
e d
iffe
ren
t b
aff
les
Re=
50
00
aver
age
frac
tal
size
1,1
35
D
opa
sow
an
ie: ro
zkła
d n
orm
aln
y
Wym
iar
Fra
ktaln
y =
271
*0,0
184
*no
rma
l(x
; 1,1
35
2; 0
,053
7)
7%
13%
13%
15
%
14
%
10
%
6%
6%
6%
3%
3%
1%
0%
1%
0%
1%
1,0
51
61
,08
83
1,1
25
01,1
61
71,1
984
1,2
35
11,2
71
81,3
085
1,3
45
2
Wy
mia
r F
rakt
aln
y
0%
2%
4%
6%
7%
9%
11
%
13
%
15
%
17
%
Procent obserwacji
fra
cta
l siz
e
no
rma
l d
istr
ibu
tio
n
observations
1,1
96
D
opasow
anie
: ro
zkł
ad n
orm
aln
y
Wym
iar
Fra
ktaln
y =
1348*0
,0245*n
orm
al(x; 1,1
926; 0,0
624)
2%
5%
9%
14%
16%
16%
12%
9%
8%
5%
2%
2%
1%
0%
0%
0%
1,0
556
1,1
046
1,1
537
1,2
027
1,2
518
1,3
008
1,3
499
1,3
989
1,4
480
Wym
iar
Fra
ktaln
y
0%
1%
3%
4%
6%
7%
9%
10%
12%
13%
15%
16%
18%
Procent obserwacji D
opasow
anie
: ro
zkł
ad n
orm
aln
y
Wym
iar
Fra
ktaln
y =
1348*0
,0245*n
orm
al(x; 1,1
926; 0,0
624)
2%
5%
9%
14%
16%
16%
12%
9%
8%
5%
2%
2%
1%
0%
0%
0%
1,0
556
1,1
046
1,1
537
1,2
027
1,2
518
1,3
008
1,3
499
1,3
989
1,4
480
Wym
iar
Fra
ktaln
y
0%
1%
3%
4%
6%
7%
9%
10%
12%
13%
15%
16%
18%
Procent obserwacji
fra
cta
l siz
e
no
rma
l d
istr
ibu
tio
n
observations
1,1
93
D
opas
ow
anie
: ro
zkł
ad n
orm
aln
y
Wym
iar
Fra
ktaln
y =
1640*0
,0276*n
orm
al(x;
1,1
96;
0,0
662)
2%
6%
12
%
15%
15
%1
5%
13%
9%
6%
3%
2%
1%
0%
0%
0%
0%
1,0
530
1,1
082
1,1
633
1,2
185
1,2
737
1,3
288
1,3
840
1,4
392
1,4
943
Wym
iar
Fra
ktaln
y
0%
1%
2%
4%
5%
6%
7%
9%
10
%
11
%
12
%
13
%
15
%
16
%
17
%
18
%
Procent obserwacji D
opas
ow
anie
: ro
zkł
ad n
orm
aln
y
Wym
iar
Fra
ktaln
y =
1640*0
,0276*n
orm
al(x;
1,1
96;
0,0
662)
2%
6%
12
%
15%
15
%1
5%
13%
9%
6%
3%
2%
1%
0%
0%
0%
0%
1,0
530
1,1
082
1,1
633
1,2
185
1,2
737
1,3
288
1,3
840
1,4
392
1,4
943
Wym
iar
Fra
ktaln
y
0%
1%
2%
4%
5%
6%
7%
9%
10
%
11
%
12
%
13
%
15
%
16
%
17
%
18
%
Procent obserwacji
fra
cta
l siz
e
no
rma
l d
istr
ibu
tio
n
observations
aver
age
area
μm
2
838,9
D
opa
so
wan
ie: ro
zkła
d R
ay
leig
ha
Pow
ierz
ch
nia
= 1
64
0*1
307
,967
1*r
ay
leig
h(x
; 1
34
4,5
68
8)
86%
8%
3%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
86,2
06
9
270
2,1
41
0531
8,0
75
2 79
34
,00
001
054
9,9
43
4 13
16
5,8
77
6157
81
,81
17 18
397
,745
921
01
3,6
80
0
Po
wie
rzc
hn
ia
m2
0%
12%
24%
37%
49%
61%
73%
85%
98%
Procent obserwacji
are
a µ
m2
Ra
yle
igh
dis
trib
utio
n
observations
1789,1
D
opaso
wanie
: ro
zkła
d R
ay
leig
ha
Pow
ierz
chn
ia =
1348
*8758,5
46
7*r
ayle
igh
(x; 5540
,8895)
97%
2%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
59,4
530
1757
6,5
46435093
,63
98 52610
,73
3170127
,82
65 87645
,00
001
,0516E
5
1,2
268E5
1,4
02
E5
Pow
ierz
ch
nia
m
2
0%
15
%
30
%
45
%
59
%
74
%
89
%
104
%
Procent obserwacji D
opaso
wanie
: ro
zkła
d R
ay
leig
ha
Pow
ierz
chn
ia =
1348
*8758,5
46
7*r
ayle
igh
(x; 5540
,8895)
97%
2%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
59,4
530
1757
6,5
46435093
,63
98 52610
,73
3170127
,82
65 87645
,00
001
,0516E
5
1,2
268E5
1,4
02
E5
Pow
ierz
ch
nia
m
2
0%
15
%
30
%
45
%
59
%
74
%
89
%
104
%
Procent obserwacji
are
a µ
m2
Ra
yle
igh
dis
trib
utio
n
observations
2329,9
D
op
aso
wa
nie
: ro
zkł
ad
Ray
leig
ha
Pow
ierz
ch
nia
= 2
71*3
53
2,2
53
7*r
ayle
igh(x
; 40
47
,3103
)
85
%
6%
3%
3%
1%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
112
,960
8
717
7,4
6821
42
42
,000
0 213
06,4
83028
37
1,0
000 35
435
,497
8425
00,0
000 49
56
4,5
12656
62
9,0
00
0
Pow
ierz
ch
nia
m
2
0%
7%
15
%
22
%
30
%
37
%
44
%
52
%
59
%
66
%
74
%
81
%
89
%
96
%
Procent obserwacji D
op
aso
wa
nie
: ro
zkł
ad
Ray
leig
ha
Pow
ierz
ch
nia
= 2
71*3
53
2,2
53
7*r
ayle
igh(x
; 40
47
,3103
)
85
%
6%
3%
3%
1%
1%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
112
,960
8
717
7,4
6821
42
42
,000
0 213
06,4
83028
37
1,0
000 35
435
,497
8425
00,0
000 49
56
4,5
12656
62
9,0
00
0
Pow
ierz
ch
nia
m
2
0%
7%
15
%
22
%
30
%
37
%
44
%
52
%
59
%
66
%
74
%
81
%
89
%
96
%
Procent obserwacji
are
a µ
m2
Ra
yle
igh
dis
trib
utio
n
observations
The comparison of the results of experimental research and numerical simulations – Figures 4
and 5 – shows a satisfactory compliance as to the tendencies of the changes and the resulting
values.
43
Tab. 1. The coefficient values for the empirical distributions
Rayleigh’s distribution
density function 2
2
22
)( b
A
eb
AAf
b - scale coefficient
Reynols
Number
H = 1/3D H = 1/2D H = 2/3 D
b b b
2000 2044 7693 19122
3000 5742 5153 6537
4000 4672 14668 9239
5000 1344 4047 5540
lognormal distribution
density function
2
2
2
))((
2
1)(
k
bFln
ekF
Ff
b - scale coefficient
k - shape coefficient
b k b k b k
2000 3,6673 0,541 3,9008 0,7443 4,2268 0,9382
3000 3,9201 0,7432 3,81 0,6541 4,2214 0,6529
4000 3,7809 0,6874 4,3667 0,9103 3,9559 0,7643
5000 3,6369 0,5366 3,9522 0,6586 3,7254 0,6406
normal distribution
density function 2
2
2
))((
2
1)(
D
eDf
µ - mean value coefficient
σ - standard deviation coefficient
µ σ µ σ µ σ
2000 1,1756 0,059 1,1376 0,0499 1,1495 0,0493
3000 1,1623 0,0578 1,1501 0,0508 1,1457 0,0473
4000 1,1697 0,0625 1,1317 0,0397 1,1495 0,0613
5000 1,926 0,0624 1,1352 0,0537 1,196 0,0662
44
x6
Fig
. 4
. C
on
tou
r m
ap
s o
f tu
rbu
len
ce i
nte
nsi
ty I
% i
n t
he
rea
cto
rs u
sed
in
th
e re
sea
rch
, fo
r th
e h
ori
zon
tal
(z1
) a
nd
ver
tica
l (x
1 –
x6
) se
cti
on
pla
nes
x5
x4
x3
x2
x1
a) h
= 1
/3D
b)
h =
1/2
D
c) h
= 2
/3D
d)
h =
1/3
D
e) h
= 1
/3D
f) h
= 2
/3D
45
a) b)
LDA
baffle 1
CFD
reactor axis
reactor
with baffles
h = 1/2D
rea
cto
r a
xis
kinetic turbulence energy k [m2/s2]
rea
cto
r a
xis
reactor
with baffles
h = 1/2D
reactor axis
LDA
baffle 3
CFD
kinetic turbulence energy k [m2/s2]
c) d)
reactor
with baffles
h = 1/2D
reactor axis
rea
cto
r a
xis
kinetic turbulence energy k [m2/s2]
LDA
baffle 5
CFD
LDA
baffle 5
CFD
reactor axis
kinetic turbulence energy k [m2/s2]
rea
cto
r a
xis
Fig. 5. Comparison of the results of the CFD simulation and the LDA measurements of the turbulence
kinetic energy., a, b, c) in the selected cross-sections in the reactor tube located in the baffles planes,
d) for three reactors with different widths of baffles
The comparison of the results of turbulence kinetic energy value calculations based on
the numerical simulations and measurements allows to verify their high compliance and
attests to the validity of the assumptions made for the numerical calculations. For reactors
with baffle widths: h = 1/3D i h = 1/2D a very high compliance between the results of the
simulations and measurements was obtained (mean relative error was, respectively, 15.8%,
13.2%). However, for the reactors with the widest baffles, i.e. h = 2/3D, the resulting
46
discrepancy between the results of the simulations and actual measurements was higher,
especially near the reactor tube wall. In this case, the maximum relative error was 47.2%.
Obtained results provide data and allow the qualitative and quantitative assessment of the
actual liquid flow turbulence parameters, such as: mean flow rates, fluctuation velocities,
kinetic energy of the turbulence and the turbulent kinetic energy dissipation in the reactor
with different widths of internal baffles. The CFD computer simulation, analysing the
turbulent motion of the liquid in the tubular reactor with baffles of different widths allowed to
identify the zones of intensive mixing, flow velocity components and turbulence parameters.
Based on the obtained results of simulation and experimental research, it was found that the
aggregation and decomposition processes in the tubular mixer approach the state of quasi-
equilibrium and the turbulence kinetic energy value k decides which of these processes will
dominate. Independently of the width of the baffles, the dominant flow in each of the reactors
was the longitudinal flow – conforming with the direction of liquid supply to the apparatus.
Lower flow rate values (less intensive flow) were also found in the crosswise direction,
conforming with the reactor radius. Vertical movement of the liquid – along the height of the
reactor was negligible and its velocity was near zero.
CONCLUSIONS
The results of the simulation (CFD) and experimental (LDA) research of the flow of the
suspension through a tubular reactor with baffles show very good convergence – relative error
does not exceed 25%. The CFD simulation calculations packages baed on the codes of the
numerical fluid mechanics are a useful tool which may be successfully applied to identify
liquid flow and optimise the designs of tubular reactors with internal baffles. However, their
application for a full modelling of a complex process, such as flocculation, is yet quite
limited. In this respect, their further and intensive development is required, especially in the
form of creating theoretical models describing the processes of particles interactions and flocs
generation.Based on the obtained results, the crucial impact of geometry and dimensions of
the baffles on the turbulence of the liquid flow in the tubular reactor was clearly confirmed.
The baffles intensify the flow of the suspension within the apparatus and initiate additional
flow turbulence in specific zones of the reactor.
The lowest values of the rotation and helicity of the flow were observed for the narrowest
baffles and the highest values for the widest baffles – this verifies the correlation between the
liquid flow rate field and the width of baffles.
The smallest flocs, independently of the change in the flow intensity, were obtained in the
reactor with baffles: h = 1/3D.
47
The increase in the concentration of the flocculant causes the increase of all the parameters
determining the size of generated agglomerates (mean area of the flocs, Feret’s diameter,
fractal dimension) for all flow intensity values.
The largest values of the mean area and Feret’s diameter were obtained for the lowest flow
intensities. This is a result of the lower rotations in the spaces between the baffles, facilitating
the increase in the size of the flocs.
The largest values of fractal dimension were obtained for the highest flow intensity – the
generated agglomerates flow between the baffles at high speeds and are pulled apart into
smaller agglomerated with uneven edges, resulting in the increase of their fractal dimension.
Based on the change in the sizes of the agglomerates generated behind the consecutive baffles
of the tubular reactor there is an observable increase in the mean area of the particles after
each consecutive baffle. For the large values of the flow rate, there is an increase in the
turbulence, advantageous to the good distribution of the flocculant in the full volume of the
suspension, resulting in the increased size of the flocs.
The tubular reactor with baffles allows a continuous process operation, resulting in a highly
increased efficiency. Furthermore, the reactor is free from the typical disadvantages of the
tank systems – uneven distribution of the kinetic energy of the turbulence in the mixed
contents of the tubular reactor.
The tubular reactor with baffles is an efficient apparatus for continuous flocculation if the
following principles are observed: the width of the baffles in the tubular reactor should be at
least 1/2 D, higher flocculant concentrations are to be used, low suspension flow intensities
are to be avoided.
ACKNOWLEDGMENT
These investigations were conducted with financial support from the national research
program (0264/B/H03/2011/40).
REFERENCES
[1] A.L. Kowal, M. Świderska-Bróż: Water purifying. PWN, Warszawa 2005.
(in Polish)
[2] Thomas D. N., Judd S. J., Fawcett N.: Flocculation modeling: A review, Water
Research, 1999, Nr. 33, S. 1579/1592
48
[3] The final report of the national research project No. 0192/B/H03/2008/34, 2011.
(in Polish)
[4] Media Cybernetics: Image-Pro Plus Start-up Guide. Media Cybernetics, Inc.,
Silver Spring, 1993.
[5] StatSoft: Statistica – opis systemu. StatSoft, Inc., Tulsa 2002.
[6] Dyląg M., Kamieński J., Rosiński J., Szatko W.: Identification and modeling of
particle size distributions for established terms of flocculation process, Inż. i Ap.
Chem., 2010, Nr. 4, S.22/23. (in Polish)
[7] Dyląg M., Kamieński J., Rosiński J.: Modelling of the particle size distribution in
the process of perikinetic coagulation, Inż. i Ap. Chem., 2010, Nr. 2, S. 35/36.
(in Polish)
[8] Albrecht H.-E., Borys M., Damaschke N., Tropea C.: Laser Doppler and Phase
Doppler Measurement Techniques. Springer, Berlin-Heidelberg 2003.
[9] DynamicStudio 3.2 User’s Guide. Dantec Dynamics, Skovlunde 2011.
[10] Jaworski Z.: Computational Fluid Dynamics in Chemical and Process
Engineering. Exit, Warsaw 2005. (in Polish)
[11] Ansys Fluent 14.0 User’s Guide. Ansys Inc., Lebanon 2011.
[12] Gambit 2.4 User’s Guide. Ansys Inc., Lebanon 2009.
[13] Ansys Fluent 14.0 Theory Guide. Ansys Inc., Lebanon 2011.
49
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
RHODIUM-CATALYZED HYDROFORMYLATION OF 1-DODECENE IN
MICROEMULSION SYSTEMS WITH NON IONIC SURFACTANTS
T. POGRZEBA; T. HAMERLA; R. SCHOMÄCKER
Technische Universität Berlin, Department for Chemistry, Sekr. TC8,
Straße des 17. Juni 124, D-10623 Berlin, Germany
Abstract. We investigate microemulsion systems with nonionic surfactants for
catalytic gas/liquid reactions, namely the hydroformylation of the long-chain
olefin 1-dodecene. The selection of an appropriate surfactant is crucial, since the
catalyst activity in microemulsion systems is strongly influenced by the phase
behaviour. Furthermore, the surfactant affects the quality and time of phase
separation as well, which are important parameters for the design of a continuous
process. Based on these systems, we present an integrated concept of reaction
management and catalyst recycling. Batch-recycling experiments showed that the
rhodium catalyst remains stable for at least 4 runs, with a rhodium leaching into
the organic phase less than 1 ppm and minor loss in activity. The reaction showed
TOFs of >1000 h-1
and high selectivities of 98:2 to the desired linear aldehyde.
Keywords. Hydroformylation, Microemulsion systems, Surfactants
INTRODUCTION
At the end of 1950’s, about 20 years after the discovery of the cobalt-catalyzed “oxo”
reaction by Otto Roelen [1], the investigations on rhodium-catalyzed hydroformylation
started. From the beginning, it was obvious that the rhodium-based catalysts are more active
compared to cobalt catalysts. The crucial discovery by Wilkinson’s group [2] that active and
selective rhodium catalysts for hydroformylation under mild reaction conditions can be
obtained in the presence of triphenylphosphine (TPP) as a ligand triggered a lot of research on
50
phosphine modified rhodium catalysts. The synthesis and use of different phosphorous- and
sulphur-containing ligands was extensively investigated until today and generated an
enormous amount of knowledge [3].
The hydroformylation of short-chain olefins using two-phase catalysis was established
over 30 years ago in industry (Ruhrchemie/Rhône-Poulenc-Process [4]). In this process, the
challenge is the quantitative recovery of the expensive rhodium catalyst and it was solved by
immobilization of the hydrophilic catalyst complex in the aqueous phase. Unfortunately, this
concept is limited to short-chain olefins (<C5) since the solubility of higher olefins in water is
very poor. Microemulsion systems are an approach to overcome this limitation and detailed
studies are currently under investigation in the Collaborative Research Centre InPROMPT /
TRR 63. The addition of a surfactant to the reaction mixture increases the solubility of the
olefin in water, which leads to the formulation of microemulsion systems that can act as
tuneable solvents. With these microemulsions not only the interfacial area is increased during
the reaction, but also the phase separation behaviour can be adjusted through temperature
changes, thus allowing for an easy separation of the expensive rhodium complex from the
organic phase after reaction. Therefore, a process concept for the hydroformylation of long-
chain olefins (in this case using 1-dodecene as the substrate) on the basis of these aqueous
multiphase systems was developed (see Figure ).
CO, H2
Surfactant
Olefin
Phase separation
Recyle stream
(aqueous catalyst solution)
Product
Figure 1: Process concept for the hydroformylation of long-chain olefins in microemulsions.
In the following, we will present a systematic approach to optimize microemulsion
systems for a certain application, e.g. hydroformylation of long-chain olefins, by lab-scale
experiments. In contrast to other reaction media the influence of different process parameters
on the reaction in microemulsion systems is more complex. The variation of parameters such
as temperature or concentrations of reactants can change the state of the microemulsion, thus
51
forming a new system with different modalities for reaction and mass transfer. Figure 2
illustrates this issue with two possible states of microemulsion systems. In the case of an oil-
in-water microemulsion the surfactant forms “normal” micelles in the aqueous phase,
dissolving a part of the oily reactant into the catalyst solution. If the system forms a water-in-
oil microemulsion, inverse micelles disperse the catalyst solution into the oily phase. Hence, a
microemulsion system gives the possibility to choose the phase where the reaction should
take place by offsetting the appropriate parameters. It has to be mentioned that this choice is
very crucial for the reaction performance. Since the ability to form micelles is different for
each surfactant, it is necessary to thoroughly investigate the phase behaviour of each
microemulsion system before the actual optimization of the parameters for the reaction.
Furthermore, we will discuss the importance of choosing an appropriate surfactant for
the reaction and its influence on the catalytic activity, since the surfactant is not only a
solubilizer but also interacts with the catalyst in solution.
Figure 2: Two different states of microemulsion systems with non-ionic surfactants. Oil-in-water
microemulsion with an oil excess phase (left) and water-in-oil microemulsion with an aqueous excess
phase (right).
EXPERIMENTAL SETUP
The hydroformylation reactions in lab-scale are performed in a 100 ml stainless steel high
pressure vessel from Premex Reactor AG, equipped with a gas-dispersion stirrer and mounted
in an oil thermostat from Huber (K12-NR). An overview of the complete reactor set-up is
given in Figure 3. Mass flow controller (4) and a pressure transmitter (5) in the syngas feed
line enable isobaric reaction management (semi-batch mode). For an alternative batch-mode
(decreasing pressure) a 300 ml gas-reservoir (7) is connected to the reactor. A connection for
the inertisation of reaction mixture is implemented as well (9).
52
Figure 3: Set-up of hydroformylation reactor for lab-scale experiments.
The hydroformylation of 1-dodecene is investigated in semi-batch experiments, at
temperatures of 75 to 110 °C and at pressures of 3 to 40 bar overpressure of syngas (1:1
mixture of CO and H2) and a stirring speed of 1200 rpm. The reaction mixture usually
consists of 1-dodecene (0.120 mol, 20.0 g, 26 mL), water (1.11 mol, 20 mL), non-ionic
surfactant, rhodium precursor [Rh(acac)(CO)2] (0.05 mmol, 12.9 mg) and ligand
SulfoXantPhos (0.20 mmol, 158 mg). The metal-to-ligand ratio of the water-soluble catalyst
complex is usually held constant at 1:4.
For the evaluation of experiments, samples are taken at several time intervals and analyzed by
gas chromatography (GC) on a Hewlett Packard model 5890, series II equipped with a RTX-
5MS capillary column, a FID analyzer, and nitrogen as the carrier gas. Inductively coupled
plasma optical emission spectrometry (ICP-OES) is used to determine the amount of rhodium
and phosphor in the product phase.
The parameters to characterize the composition of a microemulsion system are the weight
fractions α and γ [Eq. (1) & (2)]; moil is the mass of oil, mH2O is the mass of water and msurf is
the mass of surfactant.
53
OHoil
oil
mm
m
2
(15) (1)
surfOHoil
surf
mmm
m
2
(16) (2)
APPLIED CHEMICALS
The substrate 1-dodecene was purchased from the company VWR (purity 95 %). The applied
water has HPLC grade. The syngas (1:1 mixture of CO and H2, purity 1.6 for CO and 3.0 for
H2) was purchased from Air Liquide. The applied catalyst consists of a rhodium based
precursor [Rh(acac)(CO)2], sponsored by the company Umicore, and the water-soluble ligand
SulfoXantPhos, purchased from the company MOLISA GmbH. A list of various technical
grade nonionic surfactants used for the experiments is given in Table 1.
Table 1: Names and structures of different applied surfactants.
Abbreviation/
trade name
Surfactant type Structure
Marlipal O13/80-
200
Fatty alcohol
ethoxylate
Marlipal 24/70 Fatty alcohol
ethoxylate
Marlophen NP 5-9 Nonylphenol
ethoxylate
Triton X-100 Nonylphenol
ethoxylate
54
HYDROFORMYLATION IN MICROEMULSION SYSTEMS
Surfactant selection
The hydroformylation of long-chain olefins in microemulsion systems has been extensively
investigated for many years [5–8]. In general, the presence of a surfactant makes the
performance of a reaction difficult to predict. As mentioned before, each surfactant has
different abilities to form micelles, thus affecting the mass transfer in the reaction system
differently. Moreover, the surfactant can change the phase behaviour of the reaction system
and/or interact with the catalyst on a molecular level and vary its catalytic behaviour. The
compatibility of surfactant and catalyst is important for the catalyst activity and selectivity.
Hence, the choice of a surfactant for a reaction system is very crucial and has to be done
carefully.
A surfactant screening is the general procedure to find an appropriate candidate for the
reaction. We investigated different technical-grade nonionic surfactants for the rhodium
catalyzed hydroformylation of 1-dodecene in microemulsion systems under standard
conditions and the results are shown in Table 2.
Table 2: Result of surfactant screening for the hydroformylation of 1-dodecene in microemulsion systems.
Nr.
Surfactant
Conversion
1-dodecene
[%]
aldehyde
selectivity
[%]
linear/
branched
selectivity
1 Marlophen NP 5 4 99 99:1
2 Marlophen NP 6 8 75 99:1
3 Marlophen NP 7 10 90 98:2
4 Marlophen NP 9 15 87 98:2
5 Triton X-100 18 89 98:2
6 Marlipal O13/80 20 95 99:1
7 Marlipal O13/100 23 96 98:2
8 Marlipal O13/200 17 65 96:4
9 Marlipal 24/70 15 93 98:2
Standard reaction conditions: tR = 240 min, T = 110 °C, p = 40 bar, 0.05 mmol
Rh(acac)(CO)2, 0.25 mmol SulfoXantPhos, 180 mmol 1-dodecene; VR = 50 mL.
55
The reaction shows varying behaviour for apparently similar surfactants (e.g. surfactants 1-4).
This clearly illustrates the strong impact of the surfactant on the reaction. As a result of the
screening, we selected the surfactant Marlipal 24/70 for the hydroformylation reaction due to
its easy handling and catalyst compatibility. Apart from the choice of the appropriate
compatible surfactant for the reaction, it is important to investigate the phase behaviour of the
resulting product containing microemulsion system. In these systems generally a strong
variation of the catalytic activity is obtained with respect to the region in the phase diagram
where the reaction is performed. A detailed analysis of the influence of phase behavior on this
reaction is already given by Hamerla et al. [8]. The three-phase region of the microemulsion
system was found to be most suitable for the hydroformylation.
Variation of parameters
The preliminary investigations resulted in an adequate reaction system for the
hydroformylation of 1-dodecene with respect to activity and selectivity. In the following,
several experiments have been conducted to investigate the influence of different parameters
on the reaction and to gain more knowledge of the reaction kinetics.
The temperature of the reaction mixture has a strong influence on the reaction rate; especially
in microemulsion systems temperature affects the phase behaviour and mass transfer
processes. The experimental results shown in Figure 4 are illustrating this complex
dependency. As obvious, the reaction rate increases with rising temperatures, which is
expected in terms of kinetics. At 85 °C the reaction rate increases drastically and remains on
that level until 95 °C. The reason for this is the existence of the three-phase region for the
investigated microemulsion system in this temperature range. At 100 °C the reaction rate
decreases significantly, what is in agreement with the preliminary experiments that the
reaction rate decreases with leaving the three-phase region [8].
56
60 70 80 90 100
1,5x10-4
2,0x10-4
2,5x10-4
3,0x10-4
3,5x10-4
4,0x10-4
4,5x10-4
5,0x10-4
Re
action
ra
te (
mo
l/l*
s)
Temperature (°C)
Figure 4: Influence of temperature on the hydroformylation. Test conditions: 15 bar, 1200 rpm, 0.05
mmol Rh(acac)(CO)2, 0.2 mmol SulfoXantPhos, 120 mmol 1-dodecene; α = 0.5, γ = 0.08, VR = 50 mL.
Another important parameter for the hydroformylation is the syngas pressure. Experiments
were performed in a pressure range of 2 to 40 bar. In Figure 5 only the results for 3, 15 and 30
bar are shown to gain a better overview of the main trend; it has to be mentioned that the
outstanding low conversion after 180 min at 3 bar is a result of insufficient sampling. In
general, the reaction rate increases with higher pressures. However, we found that the
difference of the overall reaction performance for each experiment is not very significant,
which can be explained by the formation of inactive catalyst species with increasing carbon
monoxide concentrations. Thus, the positive effect on the reaction kinetics of more carbon
monoxide in solution is suppressed by an increasing amount of inactive catalyst. For a better
understanding of this issue we refer to [9], where the authors analyzed the complex reaction
network of the rhodium-catalyzed hydroformylation of 1-dodecene and suggested a catalytic
cycle including the two main side reactions (isomerization and hydrogenation) as well as the
deactivation of catalyst.
57
0 50 100 150 200 250
0
5
10
15
20
25
3 bar
15 bar
30 barC
on
ve
rsio
n (
%)
Time (min)
Figure 5: Influence of pressure on the hydroformylation. Test conditions: 85 °C, 1200 rpm, 0.05 mmol
Rh(acac)(CO)2, 0.2 mmol SulfoXantPhos, 120 mmol 1-dodecene; α = 0.5, γ = 0.08, VR = 50 mL.
SUMMARY AND CONCLUSION
In summary, we showed that microemulsion systems are feasible reaction media for the
rhodium-catalyzed hydroformylation of 1-dodecene. Our results establish the basis for a
continuous reaction process using surfactants systems. Nevertheless, the first step is the
selection of an appropriate surfactant that primarily supports the catalytic reaction and, in
addition, gives the best performance in the separation step. From the large pool of available
surfactants with different degree of purity and characteristics, an appropriate surfactant is
mostly obtained by a detailed screening procedure, because the prediction of the catalytic
performance in a surfactant system is not yet possible. The second step includes the study of
the phase behaviour of the applied microemulsion system. These two aspects, surfactant
selection and phase separation, should emphasize that the application of surfactant systems
requires a high degree of knowledge about the phase behaviour of the formulated system.
After establishing an appropriate reaction medium, the investigation and optimization of the
reaction parameters is consequentially the next step. We showed that changing the reaction
temperature in microemulsion systems is not only influencing the kinetics, but also the phase
behaviour of the reaction system.
58
ACKNOWLEDGEMENTS
This work is part of the collaborative research centre InPROMPT (SFB/TRR 63). Financial
support by the Deutsche Forschungsgemeinschaft (DFG) is gratefully acknowledged.
REFERENCES
[1] O. Roelen, Production of oxygenated carbon compounds, US Patent 2.327.066, 1943.
[2] D. Evans, J.A. Osborn, G. Wilkinson, Hydroformylation of Alkenes bye Use of
Rhodium Complex Catalysts, J. Chem. Soc. (1968) 3133.
[3] P.W.N.M. Van Leeuwen, C. Claver, Rhodium Catalyzed Hydroformylation, Kluwer
Academic Publishers, New York, Boston, Dordrecht, London, Moscow, 2002.
[4] B. Cornils, J. Hibbel, W. Konkol, B. Lieder, J. Much, V. Schmidt, et al., Verfahren zur
Herstellung von Aldehyden, DE3234701, 1982.
[5] M. Haumann, H. Koch, P. Hugo, R. Schomäcker, Hydroformylation of 1-dodecene
using Rh-TPPTS in a microemulsion, Appl. Catal. A Gen. 225 (2002) 239–249.
[6] C.C. Miyagawa, J. Kupka, A. Schumpe, Rhodium-catalyzed hydroformylation of 1-
octene in micro-emulsions and micellar media, J. Mol. Catal. A Chem. 234 (2005) 9–
17.
[7] M. Gottardo, A. Scarso, S. Paganelli, G. Strukul, Efficient Platinum(II) Catalyzed
Hydroformylation Reaction in Water: Unusual Product Distribution in Micellar Media,
Adv. Synth. Catal. 352 (2010) 2251–2262.
[8] T. Hamerla, A. Rost, Y. Kasaka, R. Schomäcker, Hydroformylation of 1-Dodecene
with Water-Soluble Rhodium Catalysts with Bidentate Ligands in Multiphase Systems,
ChemCatChem. 5 (2013) 1854–1862.
[9] J. Markert, Y. Brunsch, T. Munkelt, G. Kiedorf, a. Behr, C. Hamel, et al., Analysis of
the reaction network for the Rh-catalyzed hydroformylation of 1-dodecene in a
thermomorphic multicomponent solvent system, Appl. Catal. A Gen. 462-463 (2013)
287–295.
59
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
INVESTIGATIONS TO INCREASE THE SELECTIVITY OF SUZUKI CROSS
COUPLING REACTIONS BY SEMI-BATCH EXPERIMENTS
M. SCHMIDT; M. SCHWARZE; R. SCHOMÄCKER
Technische Universität Berlin, Department for Chemistry, Sekr. TC8,
Straße des 17. Juni 124, D-10623 Berlin, Germany
Abstract. We investigate the catalytic Suzuki cross coupling reaction of
2-chloronitrobenzene (2-CNB) and 4-chlorophenylboronic acid (4-CBA), an
intermediate step of the Boscalid synthesis, aiming for an improvement of the
selectivity. Therefore, we developed a feed strategy for a semi-batch reactor in
order to suppress the homocoupling of the boronic acid, the main side reaction.
The choice of appropriate semi-batch parameters, especially the time of dosage, is
crucial to obtain a high selectivity with good reaction rates. Batch experiments
showed that substituents of the coupling reactants influence the homocoupling and
that oxygen is a factor, which enhances the formation of the undesired product.
We performed semi-batch experiments, which lead to an increase of the
selectivity. Contrary to our expectations an extended time of dosage of boronic
acid increases the conversion because of suppression by the boronic acid.
Keywords. Suzuki reaction, Cross coupling, Palladium, Boscalid, Semi-batch
reactor, Homocoupling
INTRODUCTION
Since the discovery in the beginning of the seventies by Heck [1], coupling reactions
of aromatic compounds play an important role in organic synthesis. The formation of a
carbon-carbon bond is essential for the production of fine chemicals, especially in the
pharmaceutical industry. A wide range of fine chemicals is industrially synthesized via a cross
coupling reaction such as the production of the plant protection product prosulfuron by
Novartis [2] or the fabrication of the fungicide Boscalid by BASF, which has with 1000 tons
60
per year a huge impact [3]. The fungicide is synthesized via a complex reaction sequence [4]
in which the Suzuki cross coupling is the initial step (see Figure 1).
Figure 1: Reaction sequence of the Boscalid synthesis.
In general the Suzuki cross coupling reaction is one of the most powerful methods for
the production of functionalized biphenyls. A wide range of aryl halides with different
substituents can be used. Compared to other cross coupling reactions the use of boronic acids
offer the advantage that they are nontoxic, cheap, easily accessible and most notably not water
and air sensitive. Furthermore the Suzuki cross coupling reaction can be done in a
homogeneous or heterogeneous way, in which different metals can be used as catalysts. The
use of palladium as catalyst is one approach, which is most frequently investigated until
today. But also other metals such as nickel or rhodium [5] are used for the Suzuki cross
coupling reaction. For the stabilization of the palladium as a homogeneous catalyst different
ligands are applied. In the beginning, the use of triarylphosphan-based ligands was
investigated, but studies show that the structure of the ligand is crucial for the activity and
selectivity of the cross coupling reaction. So the synthesis and the use of complex ligands
such as dialkylbiarylphosphines were investigated recently [6].
The homocoupling of the boronic acid, which is an oxidative coupling, is the problem
of the Suzuki cross coupling reaction. It causes a low selectivity and a difficult purification of
the cross coupling product. On the one hand the synthesis of the undesired side product is
affected by the addition of oxidizing agents, wherein uncoordinated metal catalyst can act as
oxidizing agent. On the other hand oxygen plays an important role for the homocoupling. An
intermediate generated by the reaction of oxygen with the catalyst can initiate the
homocoupling [7]. Hence, the side reaction must be suppressed by the choice of a suitable
catalyst-ligand system as well as the conditions of the reaction. The use of special ligands is
very expensive and sometimes they are difficult to obtain. Moreover the homocoupling occurs
despite inert conditions.
61
Our approach to ensure a high selectivity and to suppress the undesired product is a
new one and we try to control the selectivity by a feed strategy for the reactants rather than a
ligand controlled process. Therefore, the Suzuki cross coupling should not be performed as a
batch but as a semi-batch reaction and the boronic acid is dosed to the reactor in order to
suppress the side reaction and to obtain high selectivities with simple ligands. Based on the
low concentration of the boronic acid caused by dosage, the reaction rate of the homocoupling
is to be expected low and high selectivities should be achieved.
EXPERIMENTAL
Chemicals
The reactants 4-chlorophenylboronic acid (4-CBA, purity 98 %) from ABCR,
2-chloronitrobenzene (2-CNB, purity 99 %) from Sigma-Aldrich, phenylboronic acid from
Fluka (purity 97 %) and p-tolylboronic acid from Sigma-Aldrich were used for the coupling
reactions. The base potassium carbonate (K2CO3, purity 99 %) was purchased from Roth. The
applied water and acetonitrile (ACN) had HPLC grade. The palladium precursor Pd(OAc)2
from ABCR (purity 99 %) and the water-soluble ligand Tris(3-sulfophenyl)phosphine
trisodium salt (TPPTS) were used for the preparation of the active catalyst. The TPPTS was
obtained as a 29.77 wt% aqueous solution. 4,4´-dimethylbiphenyl from ABCR,
4,4´-dichlorobiphenyl (DCBP), 4-chloro-2´-nitrobiphenyl (CNBP) and biphenyl from Merck
were used as calibration substances for the HPLC.
Catalyst preparation
For the preparation of the water-soluble catalyst complex, the precursor of palladium acetate
Pd(OAc)2 (4.8 mg, 0.021 mmol, 0.005 eq.) was evacuated three times in a Schlenk tube and
flushed with argon. 485.7 mg TPPTS solution, containing 144.6 mg (0.254 mmol, 0.06 eq.)
TPPTS, was added by a syringe through a septum and additionally degassed water (3.0 g,
0.166 mmol) was injected. Then the mixture was evacuated and flushed with argon again. The
catalyst solution was stirred at room temperature and under argon atmosphere over night.
Thereby, the solution turned from light yellow to yellow-orange.
Suzuki coupling
All reactions were carried out in a 200 mL glass reactor with double jacket to adjust the
reaction temperature to 75 °C. The lid of the reactor had connections for vacuum, nitrogen,
sampling and dosage. The coupling components 2-CNB (668 mg, 4.24 mmol, 1 eq.), 4-CBA
62
(729 mg, 4.66 mmol, 1.1 eq.), which was dosed during the semi-batch experiments, and
potassium carbonate (732 mg, 5.3 mmol, 1.25 eq.) were put in to the reactor and the reactor
was evacuated and flushed with nitrogen three times. Afterwards, the solvent was filled in,
which consists of a one to one mixture of water (51.7 g, 2.87 mol) and ACN (51.7 g,
1.26 mol). The reaction started, when the catalyst was added by a syringe. Samples were
taken with time via a syringe and then analyzed by high-performance liquid chromatography
(HPLC) using an Agilent series 1200 chromatograph with a Multospher 120 RP15-5µL
column. The reactants were detected with a spectrophotometer at a wavelength of 225 nm at
25 °C and the flow rate was 1 mL/min. The mobile phase was a mixture of 30 % water and
70 % ACN.
The conversion X [Eq. (1)] and selectivity S [Eq. (2)] were calculated from the results of the
HPLC; c2-CNB is the concentration of 2-CNB, c2-CNB,0 is the initial concentration of 2-CNB,
cCNBP is the concentration of the main product CNPB and cDCBP is the concentration of the
side product DCPB.
𝑋2−𝐶𝑁𝐵 = 1 −𝑐2−𝐶𝑁𝐵
𝑐2−𝐶𝑁𝐵,0 (1)
𝑆 =𝑐𝐶𝑁𝐵𝑃
𝑐𝐶𝑁𝐵𝑃 + 𝑐𝐷𝐶𝐵𝑃 (2)
RESULTS AND DISCUSSION
Effect of different substituents on the boronic acid
The influence of different boronic acids was investigated to study the effect of substituents
and to get more insight into the homocoupling reaction. The conversion of 2-CNB and the
concentrations of the side products are shown in Figure 2. On the one hand there is no
significant difference in the conversion rates of different boronic acids. It is known that the
coupling of ortho-substituted boronic acids is difficult due to steric hindrance [8]. Para- or
meta-substituted boronic acids have no effect on reaction rate because the oxidative addition
of the aryl chloride is the rate-determining step [9]. Thus, the applied para-substituted boronic
acids have nearly the same conversion. On the other hand, the substituent of the boronic acid
influences the homocoupling reaction. For the chloro-substituted boronic acid, the amount of
the side product formed during the reaction was doubled. Apparently, the chlorine substituent
63
activates the homocoupling, because in contrast to the other boronic acids it has an electron-
withdrawing substituent (-I-effect).
0 20 40 60 80 100 120
0
10
20
30
40
50
60
70
80
90
100
4-chlorphenylboronic acid
p-tolylboronic acid
phenylboronic acid
X2-C
NB [%
]
t [min]
0 20 40 60 80 100 120
0,0
3,0x10-4
6,0x10-4
9,0x10-4
4,4´-dichlorobiphenyl
4,4´-dimethylbiphenyl
biphenyl
c [m
ol/L]
t [min]
Figure 2: Conversion of 2-CNB (left) and concentrations of the side product (right) in use of different
boronic acids. Reaction conditions: T = 75 °C, c2-CNB = 0.36 mol/L, cboronic acid = 1.1 eq., cbase = 1.25
eq., cPd = 0.005 eq., cTPPTS/cPd = 12, mwater = mACN = 51.7 g.
Effect of oxygen
Since oxygen in the reaction medium can amplify the homocoupling, the influence of inert
conditions was examined. For this, two identical experiments were performed with and
without inert conditions, whose results are summarized in Table 1.
Table 2: Conversion and selectivity of the Suzuki coupling under air and nitrogen atmosphere
No. Atmosphere
Conversion of
2-CNB
[%]
Selectivity
[%]
1 air 19.7 53.8
2 nitrogen 25.9 90.6
Reaction conditions: tR = 120 min, T = 75 °C, c2-CNB = 0.36 mol/L, c4-CBA = 1.1 eq., cbase = 1.25 eq.,
cPd = 0.005 eq., cTPPTS/cPd = 12, mwater = mACN = 51.7 g
The conversion under a nitrogen atmosphere (Table 1, No. 2) is slightly higher than the
conversion under air (Table 1, No. 1). However, it should be noted that the reaction solution
64
becomes black after ten minutes when using an air atmosphere, because the palladium
precipitates. Thereby, the catalyst is inactivated. Under a nitrogen atmosphere, the catalyst
stays stable and a higher conversion can be achieved. But the selectivity shows fundamental
differences. Under the air atmosphere, the selectivity is much lower compared to the reaction
under inert conditions. This fact indicates the formation of a peroxo complex by the presence
of oxygen [7], which allows only the oxidative addition of the boronic acid and not of the aryl
halide. As a result of the transmetalation, the formation of the undesired side product is
facilitated.
Effect of the time of dosage tD
Based on the results of the batch experiments, we know that the homocoupling is amplified by
the use of 4-CBA, which is applied in the Boscalid synthesis. Hence semi-batch experiments
were performed to increase the selectivity of this reaction. The results are shown in Figure 3.
The dosed volume Vm and the initial volume V0 inside the reactor was held constant. The
corresponding batch experiment is also plotted.
0 20 40 60 80 100 120
0
10
20
30
40
50
60
70
80
90
100
tD= 0 min
tD= 15 min
tD= 30 min
tD= 60 min
X2-C
NB [%
]
t [min]
0 20 40 60 80 100 120
60
65
70
75
80
85
90
95
100
tD= 0 min
tD= 15 min
tD= 30 min
tD= 60 min
S [%
]
t [min]
Figure 3: Effect of the time of dosage on conversion of 2-CNB (left) and selectivity (right). Reaction
conditions: T = 75 °C, n2-CNB = 4.24 mmol, n4-CBA = 1.1 eq., nbase = 1.25 eq., nPd = 0.005 eq.,
nTPPTS/nPd = 12, V0 = 100 mL, Vm = 18 mL, V0,Batch = 118 mL, mwater = mACN = 51.7g
It is surprising that the conversions of the semi-batch experiments are nearly twice as high as
the conversion of the batch run. Usually, the longer the time of dosage, the lower is the
conversion and higher the selectivity. Therefore, these results were evaluated only
qualitatively. On the one hand the applied palladium complex can insert in the chlorine-
65
carbon bond of 4-CBA and not in 2-CNB. This inhibition could be reduced by a feed strategy
and hence we can expect higher conversions. So 4-chloro-4'-biphenylboronic acid should be
formed as main product, but it was not detected in the HPLC. On the other hand an inhibition
through the boronic acid could lead to low rates in the batch mode. This is prevented by the
semi-batch strategy and results in higher rates and thus higher conversions in comparison to
the batch run. It can be seen clearly in the experiments that the rates of the batch experiment
are not transferable to the semi-batch reaction. Furthermore the experimental results show a
higher selectivity compared to the batch experiment. However, no tendency toward an
increase in selectivity with extended time of dosage is recognizable. The selectivity increases
to values of around 95 %, which means an increase of 5 % compared to the batch run.
CONCLUSION
In summary, the feed strategy of the Suzuki cross coupling reaction leads as expected to a
higher selectivity in comparison to the batch reaction. Nevertheless, the batch experiments are
necessary to investigate the undesired homocoupling reaction and to determine the
parameters, which enhance it. The chlorine substituent on the boronic acid and the presence of
oxygen amplify the rate of the side reaction leading to a low selectivity. Based on these facts,
we can transfer the Suzuki cross coupling reaction to a semi-batch reactor to achieve a higher
selectivity. Due to the feed strategy, also unexpected higher conversions are achieved and it
seems that the boronic acid inhibits the Suzuki cross coupling reaction in the batch reaction
mode. To get a higher degree of knowledge of this inhibition process, further experiments are
required.
ACKNOWLEDGEMENTS
This work is part of the project “Auslegung von Suzuki-Kupplungen in
Mikroemulsionen unter den Aspekt der Katalysatorrückführung und Produktisolierung”.
Financial support by the „Arbeitsgemeinschaft industrieller Forschungsvereinigungen“ (AiF)
is gratefully acknowledged.
REFERENCES
[1] R. F. Heck, J. P. Nolley, Palladium-catalyzed vinylic hydrogen substitution reactions
with aryl, benzyl, and styryl halides, J. Org. Chem. 37 (1972) 14 2320-2322
66
[2] M. Röper, Homogene Katalyse in der Chemischen Industrie. Selektivität, Aktivität
und Standzeit. Chemie in unserer Zeit 40 (2006) 2 126-135
[3] C. Torborg, M. Beller, Recent applications of palladium-catalyzed coupling reactions
in the pharmaceutical, agrochemical and fine chemical industries. Adv. Synth. Catal.
351 (2009) 18 3027-3043
[4] T. N. Glasnov, C. O. Kappe, Toward a continuous-flow synthesis of Boscalid. Adv.
Synth. Catal. 352 (2010) 17 3089-3097
[5] K. Ueura, T. Satoh, M. Miura, Rhodium-catalyzed arylation using arylboron
compounds: efficient coupling with aryl halides and unexpected multiple aylation of
benzonitrile. Org. Lett. 7 (2005) 11 2229-2231
[6] R. Martin, S. L. Buchwald, Palladium-catalyzed Suzuki-Miyaura cross-coupling
reactions employing dialkylbiaryl phosphine ligands. Acc. Chem. Res. 41 (2008) 11
1461-1473
[7] C. Adamo, C. Amatore, I. Ciofini, A. Jutand, H. Lakmini, Mechanism of the
palladium-catalyzed homocoupling of arylboronic acids: key involvement of a
palladium peroxo complex. J. Am. Chem. Soc. 128 (2006) 21 6829-6836
[8] T. Watanabe, N. Miyaura, A. Suzuki, Synthesis of sterically hindered biaryls via the
palladium-catalyzed cross-coupling reaction of arylboronic acids or their esters with
haloarenes. Synlett 03 (1992) 207-210
[9] G. B. Smith, G. C. Dezeny, D. L. Hughes, A. O. King, T. R. Verhoeven, Mechanistic
studies of the Suzuki cross-coupling reaction. J. Org. Chem. 59 (1994) 26 8151-8156
67
2. MULTIPHASE SYSTEMS
69
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
MASS TRANSFER IN TWO-PHASE LIQUID-LIQUID SYSTEMS – CHALLENGES AND
SOLUTIONS
J. BAŁDYGA;M. JASIŃSKA; W. KOWALIŃSKI
Faculty of Chemical and Process Engineering, Warsaw University of
Technology, ul. Waryńskiego 1, 00 645 Warsaw, Poland
Abstract. The model describing external mass transfer coefficients in liquid-
liquid systems should include effects of surface mobility and fluid deformation on
acceleration of molecular diffusion and chemical reactions. Traditionally effects
of surface mobility and fluid deformation were treated separately in well known
phenomenological models proposed by Levich and Batchelor for droplets and
solid particles. General method of modelling including these effects is proposed,
discussed and applied to mass transfer between a droplet and continuous phase.
In the case of microstructured reactors (MSR) the model explains effects of
internal circulation in the drop or slug and film saturation on the external mass
transfer. A possibility of a priori prediction of mass transfer is analysed and two
adequate models are proposed. The model predictions are compared with
experimental data and interpreted using CFD VOF modelling.
Keywords. deformation, droplets, liquid-liquid system, mass transfer, surface
mobility.
INTRODUCTION
The problem considered in this work is related to prediction of the rate of diffusional
mass transfer to the mobile surface in the liquid–liquid two-phase heterogeneous system
including effects of fluid deformation on mass transfer. Figures 1a and 1b show schematically
70
the velocity field near the droplet with mobile surface and solid particle, respectively. Starting
from works of Levich [1] and Batchelor [2] there are two basic methods available in the
chemical engineering and fluid mechanics literature that are applied for modeling of mass
transfer at small values of the Reynolds number. The first method considers surface mobility
but neglects fluid deformation (so neglects velocity variation in the vicinity of droplet surface
that is shown in Figure 1a), whereas the second method neglects surface mobility but takes
into account deformation of fluid as shown in Figure 1b. The first method is applied to
describe external mass transfer of bubbles and not very viscous liquids, the second method is
applied to model mass transfer between ambient fluid and solid particles, very viscous drops
or drops with immobilized surface.
a) b)
Figure 1. Schematic of mass transfer (a) to a drop with mobile surface, and (b) to a solid, spherical
particle.
In the case of mass transfer of substance A from a spherical drop or particle moving
with velocity u in the continuous phase and assuming mobile surface Levich [1] obtained for
moving drop
1 2
1 2Sh=0 65 PeC
d C
.
(1)
and for moving solid sphere
1 3Sh 0 988Pe. (2)
where the Sherwood number is defined by L
A
k dSh
D , and the Péclet number by
A
udPe
D .
In the first case an effect of surface mobility on mass transfer was considered in
modeling, whereas in the second case effects of fluid deformation in particle proximity on
mass transfer were considered. Notice that as a result of assumptions leading to eq.(1) this
equation for infinitely large viscosity of the dispersed phase, d , does not converge to eq.(2);
u
u0
u
d
71
moreover in the first case one has L Ak D 1 2 for any viscosity ratio, in the second case L Ak D 2 3
For small values of viscosity of dispersed phase, 0d C and eq.(1) converges to the
form of the penetration Higbie model , 1 2
2 AL
Dk
with d u . The film models do not
take into account effects of surface mobility and fluid deformation directly, rather the film
thickness in L
A c
k d dSh
D is adjusted or fitted to experimental data. Hence three kinds of
relation between the Sherwood number, Sh, and the Péclet number are observed in the
literature: Pe1/2
, Pe1/3
and Pe0. More complex methods of modeling are based on applying an
approach proposed originally by Levich to solve more complex problems, considering for
example detailed description of fluid deformation [3, 4, 5]. For example Batchelor [3]
considered mass transfer to particles smaller than the Kolmogorov scale k 3 4 1 4 ,
Polyanin [4] described a three dimensional diffusive boundary-layer (also referred to as a
concentration boundary layer), and Polyanin and Shevtsova [5] described mass transfer
accompanied by chemical reactions.
In this work we apply modeling that includes effect of both surface mobility and fluid
deformation. We expect then continuous change of effect of molecular diffusivity, DA, on the
mass transfer coefficient, kL, between L Ak D 2 3and L Ak D 1 2
with increasing effects of
surface mobility. Schematic of model assumptions is shown in Figure 2; for simplicity a flat
surface is considered. In this simple flow deformation of fluid is expressed by and effects of
surface mobility are expressed by the surface velocity u0.
Figure 2. Schematic of model assumptions.
L
u0
x,u
y
u
72
The model is described by the system of equations:
2
2
A AA
c cu D
x y
(3)
0u u y (4)
with the boundary conditions:
0
0 0 0
0 0
0 0
A
A A
A
x , y c
x , y , c c
x , y , c
(5)
Levich [1] in his approach to external mass transfer to droplets assumed that 0 0u
and 0.
Similar problem was considered in ref. [6] to derive approximated correlation with
adjusted constant. Recently solutions of this problem were presented by authors of this work
using either transformation of variables with some simplifications [7] or direct numerical
integration [8]. In the case of microstructured reactors (MSR) the model will be extended to
explain effects of internal circulations in the drop or slug on the external mass transfer. Also
the problem of film saturation, and resulting effect on the rate of mass transfer will be
considered. We will show in what follows extension of recent work [9].
MASS TRANSFER TO FALLING DROPLET
Consider now spherical drop mowing with velocity u in the ambient fluid. The Hadamard-
Rybczyński model (valid for Stokes regime) can be applied to express the interfacial velocity
and local rate of deformation in continuous phase close to the drop surface. Equations (3) and
(4) can be now transformed by introducing a new variable, and applied to describe mass
transfer along the drop surface for 02
dx . This is possible for Sc>>1, when the layer of
variable concentration lies near the sphere and is of small thickness comparing to the layer of
velocity variation.
1 3
9 A
yD x
(6)
where
02
dx for (7)
73
2 2
d dy r for r (8)
02
C
C d
uu sin
(9)
3 d
C d
usin
d
(10)
Denoting now by m a viscosity ratio C dm and defining APe ud D , where d is the
drop diameter, one can calculate the local Sherwood number L ASh k d D .
3 20
1 3
1 3
1 3
0
2 1
3 1
sinPe
mSh( )
e d
(11)
where
1 3
2 3 4 3 1 3
0 1 32 3
1
m sinPe
m
(12)
and the average Sherwood number.
0
1
2Sh Sh sin d
(13)
The solution has been obtained under assumption that variation of 0 , eq.(12) with , so
variation of 1 3
sin with , is small comparing to variation of with in eq.(11), which
is true, indeed. Results of model predictions are shown on Figures 3 and 4.
2
4
8
100 1000
Sh
Pe
m = 0
m = 1m = 5
m = 100
74
Figure 3. Effect of the Péclet number, Pe, and the viscosity ratio, m, on the Sherwood number, Sh.
Figure 4. Effect of viscosity ratio, m, on the exponent on Pe.
Figure 3 illustrates effect of viscosity on this part of the Sherwood number that depends on
the Péclet number. Notice that such an approach neglects the fact that in the stagnant
continuous phase one has 2Sh . Also effect of variation of 1 3
sin should be
compensated. Hence, to compare model prediction with experimental data the following,
extended equation should be used
1 3
0
1 22
2 1
mSh exp f Sh sin d
m
(14)
where
1 2
1 1
m mf
m m
. Predictions of eq.(14) can be compared with results presented
in Figure 3.10, page 48 in Ref. [10], where the curves for intermediate m were prepared by
locating them between the exact solutions for m=0 and m= .
The results presented in Figure 4 show transition of exponent on the Péclet number between
exponents observed in equations (1) and (2) as viscosity of dispersed phase decreases in
relation to viscosity of the continuous phase.
Mass transfer in a slug flow concentric microreactor
By enhancing mass transfer one can change significantly the product distribution and decrease
or remove creation of by-products [9, 11, 12]. Significant enhancement of mass transfer can
be achieved by applying micro-devices. In what follows we consider effects of fluid
deformation and surface mobility using experimental results, CFD simulations and
0.3
0.35
0.4
0.45
0.5
0.55
0.001 0.1 10 1000
slo
pe
m
75
phenomenological models. Experiments were performed by the Group of Catalytic Reaction
Engineering (GGRC), Ecole Politechnique Federale de Lausanne (EPFL) of Professor A.
Renken and are reported by Kowaliński [9]. In experiments a non-reacting system of water-
acetone-toluene was used with two fluids introduced to a concentric flow contactor shown in
Figure 5a. The inner diameter of the glass capillary was 1.6 mm while the inner and outer
diameter of the steel syringe was 0.5 and 0.8 mm, respectively. The feed concentration of
acetone in water was 3.5 wt% (mole fraction xAW=0.011) and no acetone in toluene was fed
into the system (xAT=0). Hence, acetone was transferred from water to toluene. After having
established an operating window for each flow regime, experiments were repeated to
investigate the mass transfer performance. The samples containing biphasic mixture were
collected at the outlet and analyzed using gas chromatography after separating two phases.
The mass transfer coefficient, Lk a , was calculated from experimental data.
Effect of velocity of fluids on measured value of the mass transfer coefficient are shown in
Figures 5b and 6.
(a)
(b)
76
Figure 5. a) details of the concentric MSR, b) Volumetric mass transfer coefficient obtained in concentric
MSR
Figure 6. Mass transfer coefficient as a function of average velocity in MSR The experimental snapshots
show correponding flow patterns observed in MSR.
The performance of slug flow reactor shows significant increase of the mass-transfer
coefficient with the flow velocity. This is due to the increased internal circulations and
increased deformation rate. Also interfacial mass transfer area increases with increased flow
velocity. Analysis of results shows that the mass transfer is faster and mass transfer
coefficients higher on the toluene side of the interface. There is also much smaller solubility
of acetone in water. Combination of these two effects results in negligible resistance to mass
transfer on the toluene side of the interface and for this reasons the measured mass transfer
coefficients characterize external mass transfer.
It should be noted that there is well observed variation of the structure of the two-phase flow,
and such variation was predicted using the Volume of FLUID (VOF) method in our previous
paper [13].
In what follows we consider two models: Model I is constructed under condition that the
diffusion time (here the time necessary to saturate or de-saturate the film between the slug or
drop and the wall) is shorter than the contact time
2
s
A
L
D u
(15)
or
A
B
C
D
E
F
77
2
i i sc
A i
uD D LPe
D D
(16)
The resulting model equation reads
1 2
0.354 0.097
1 2 1 2
1
4 1 6
c
c dL i
s i
k a DC Ca We C
u PeL D
(17)
where C1 and C2 are constants that should be estimated based on experimental data. The
specific surface a is defined as a surface per unit volume of the dispersed organic phase.
Notice that effect of the viscosity ratio is that characteristic for creeping flow. However, as
recommended by Levich [1], pages 408-409 “the formula is valid for Re >1 provided that the
motion remains laminar….”but with an undetermined numerical coefficient”. One can
consider further extension of this model to higher values of the Reynolds number by
replacing the Stokes flow by the Oseen or Proudman-Perason stream functions [10].
Model II is for faster flow when the film is not saturated and condition (16) is not fulfilled.
Similar procedure as in derivation of model I gives in this case
1 2
0.177 0.0485
3 4 1 2
1
4 1 6
c
c dL i s
ii
k a D LC Ca We C
u D PeL D
(18)
There is some similarity between eq.(18) and the model proposed by van Baten and Krishna,
[14]. However, the effect of viscosity ratio is inluded here and not considered by van Baten
and Krishna, which was not necessary when the gas-liquid system was considered, as the term
1C C d
. Also the effect of the film fluid deformation, neglected by van Baten and
Krishna [14] is included in eq. (18). C3 and C4 are constants that should be estimated based on
experimental data.
Based on experimental data the following values for model constants were obtained:
C1=2.96 ∙10-3
, C2=0.226, C3=0 and C4 =0.34. Comparison of results of modeling with
experimental data is shown in Figure 7. Model I gives better agreement with experimental
data, however, model II much better predicts tendencies observed in experiments at higher
flow velocities [15].
To analyse model assumptions also the volume of fluid (VOF) model implemented in
commercial CFD package ANSYS was used. Computations were performed for the geometry
presented in Figure 5a. The 2D version of the VOF model was applied to show details of the
78
flow within the slug or droplet and in continuous phase nearby. The numerical grid was build
of 353250 quad cells and 355039 nodes.
Figure 8 shows the slug flow at 8 28u . mm s and circulation loops on the front and back of
the slug. One can also see velocity gradient in the film between the slug and tube wall, which
shows that there is some significant deformation rate. There is much less circulation in the
cylindrical part of the slug, which can explain why the constant C3 in eq.(18) that includes
circulations in this cylindrical part has been estimated as equal to zero.
Figure 7. Comparison of tendencies predicted by Models I and II with experimental data.
Figure 8. Flow pattern for 8 28u . mm s , water-toluene system; velocity vectors show velocities
calculated in relation to the velocity of the slug center of mass.
79
Figure 9. Flow pattern for 99 5u . mm s , water-toluene system; velocity vectors show velocities
calculated in relation to the velocity of the slug center of mass.
Figure 9 shows that at 99 5u . mm s the cylindrical part of the slug disappears and there are
rather drops than slugs present. Circulation loops are observed on the front, in the central part
of the drop and on the back. In both cases, at low and higher velocity, there is intensive
circulation in ambient fluid between the drops and slugs. Results of CFD modelling of the
flow pattern confirm assumptions of models I and II.
For further confirmation of model assumptions mass transfer of acetone from water to toluene
has been simulated using an Euler-Euler frame of reference for the flow structures presented
in Figures 8 and 9. In simulations at first the flow that was passive for the mass transfer was
developed and then at the time t = 1.3639 s for 8 28u . mm s and t = 5.430∙10-1
s for
99 5u . mm s , the transfer of acetone was initialized. Uniform concentration of acetone in
aqueous phase was assumed (xAW=0.011) and in the organic phase the concentration of
acetone was set to zero (xAT=0). The same concentration values were set to the inlet boundary
conditions. The variation of acetone concentration is simulated using the User Defined
Scalars.
Results of computations are presented in Figures 10 ab. Figure 10a shows that there is mass
transfer between the slug and the wall film and that at low velocity the film can be either
completely unloaded or completely saturated, which supports assumptions of model I. At high
velocity mass transfer around the drop has everywhere the same mechanism based on surface
mobility and fluid deformation in agreement with model II, Figure 10b, eq.(18).
One can conclude that predictions of CFD support assumptions of models I and II.
80
Figure 10. Concentration [mol fraction] of acetone in the aqueous phase for: (a) ū = 8.28 mm/s and (b) ū
= 99.5 mm/s. The first slug on the right hand side (the oldest one) is presented in both cases. Euler-Euler
approach. The feed concentrations were equal to xAW=0.011 and xAT=0.
CONCLUSIONS
A new model of mass transfer is proposed. In the case of no fluid deformation and complete
surface mobility the model becomes equivalent to the penetration model with surface
renovation rate depending on viscosity ratio. For not mobile surface the model is equivalent to
the Levich-Batchelor model accounting for fluid deformation around the rigid particle.
Phenomenological models proposed in this work were applied to interpret experimental data
for mass transfer in microreactors and further verified by using CFD. This has been done by
predicting details of the flow; results have shown fluid deformation and local circulation
loops. At small values of the flow velocity one can observe the slug flow with well developed
film, at high values of the flow velocity there is transformation of the slug flow to the drop
flow observed with the film length, Ls, shrinking to zero. Predicted details of mass transfer
outside the slugs and drops further support assumptions applied to derive phenomenological
models I and II.
(a)
(b)
81
LIST OF SYMBOLS
Ca capillary number, c dμ uCa
σ
cA concentration of A
DA diffusion coefficient
d drop diameter
kL mass transfer coefficient
Ls slug length
m viscosity ratio, C dm
Pe Péclet number, A
udPe
D
Sh Sherwood number, L
A
k dSh
D
u, ud velocity, dispersed phase velocity
We Weber number,
2
c d iu D
We
x, y coordinates
AWx mole fraction of acetone in aqueous phase
ATx mole fraction of acetone in organic phase
rate of deformation
K Kolmogorov microscale
C continuous phase viscosity
d dispersed phase viscosity
0, variables defined in equations (6) and (12)
ACKNOWLEDGEMENT
The authors acknowledge the financial support from Polish National Science Centre (Grant
agreement number: DEC-2013/11/B/ST8/00258).
82
REFERENCES
[1] Levich V.G., Physical hydrodynamics. Englewood Cliffs, N.J., Prentice-Hall 1962.
[2] Batchelor G. K., Mass transfer from a particle suspended in fluid with a steady linear
ambient velocity distribution. Journal of Fluid Mechanics, 95 (1979). Nr. 2, S. 369/400.
[3] Batchelor G. K., Mass transfer from small particles suspended in turbulent flow, J. Fluid.
Mech. 98 (1980) Nr. 3, S. 609/623.
[4] Polyanin A. D., Three-dimensional diffusive boundary-layer problem. Zhurnal Prikladnoi
Mekhaniki i Tekhnicheskoi Fiziki, 4 (1984), S.71/81.
[5] Polyanin A. D., Shevtsova V.M., Mass transfer between particles and a flow in the
presence of a volume chemical reaction. Izvestiya Akademii Nauk SSSR, Mekhanika
Zhidkosti i Gaza. 6 (1987) S.109/113.
[6] Ueyama K., Ogawa K., Hatanaka J.I., Mass transfer in the continuous phase around a
single drop. Journal of Chemical Engineering of Japan. 6 (1973), nr.2, S. 167/171.
[7] Bałdyga J., Jasińska M., Effect of model structure on on complex liquid-liquid
heterogeneous reactions, Proceedings of the third European Process intensification
Conference, EPIC 2011, 20-23 June 2011, Manchester, UK, CD, S. 175/181.
[8] Jasińska M., Lewandowski P., Bałdyga J., Nowy model wnikania masy z reakcją
chemiczną w układach heterofazowych ciecz-ciecz, Inżynieria i Aparatura Chemiczna. 52
(2013) Nr.4, S. 325/327.
[9] Kowaliński W., Effects of Multiphase flow on mass transfer and chemical reactions, PhD
Thesis, Warsaw University of Technology, Faculty of Chemical and Process Engineering,
2012.
[10] Clift R., Grace J. R. and Weber M. E., Bubbles, Drops, and Particles, New York,
Academic Press, 1978.
[11] Bourne, J.R., Mixing and the Selectivity of Chemical Reactions. Organic Process
Research & Development. 7 (2003) S. 471/508.
[12] Jasińska, M., Bałdyga, J., Cooke, M., Kowalski A., Investigations of mass transfer with
chemical reactions in two-phase liquid-liquid systems. Chem. Eng. Res. Des. 91 (2013) S.
2169/2178.
[13] M.N. Kashid, W. Kowaliński, A. Renken, J. Baldyga, L. Kiwi-Minsker, Analytical
method to predict two-phase flow pattern in horizontal micro-capillaries, Chem. Eng. Sci.
74 (2012) S. 219-232.
[14] van Baten, J.M., Krishna, R., CFD simulations of mass transfer from Taylor bubbles
rising in circular capillaries. Chem. Eng. Sci. 59 (2004) S. 2535/2545.
83
[15] M.N. Kashid, A. Renken, and L. Kiwi-Minsker, 2011, Gas-liquid and liquid-liquid mass
transfer in microstructured reactors. Chem. Eng. Sci. 66 (2011) S. 3876/389.
85
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
ELECTROCOAGULATION METHOD FOR TREATMENT OF OIL-IN-WATER
EMULSIONS – PROCESS MODELLING FROM ‘ENLIGHTENED EMPIRICISM’ TO
MOSAIC
Ł. JANCZEWSKI; M. DYLĄG;
The Institute of Advanced Manufacturing Technology, Wrocławska 37a,
PL 30-011 Cracow, [email protected]
Abstract. Physical model and challenge of the mathematical modelling with
MOSAIC software was presented. Investigations are related to known models
found in the literature. Based on the methodology being essence of the MOSAIC
software, mathematical basic equation system, which encompasses all relevant
parameters, was presented. Important model parameters like: physical and
chemical properties of the emulsion, geometry of the reactor, relevant electrical
parameters, flow pattern, electrodes configuration parameters, resident time,
efficiency and initial and boundary aluminium ions concentration have been also
considered in presented model.
Keywords. Electrocoagulation, Flotation, Coagulation, Modelling, MOSAIC.
INTRODUCTION
High demand on the metalworking fluids which are widely used in the mechanical
engineering technologies (sale of cutting fluids in Europe (with Russia) in 2010 was estimated
to 601.1 thousand tonnes [17]) effects in generation of high amount of harmful and difficult to
neutralize used oil-in-water emulsions. Problem of used metalworking fluids separation was
interchangeably defined and is embraced in the European Union law; used metalworking
fluids are qualified as hazardous and have to be neutralized before they are drained to the
wastewater treatment plant. State-of-the-art analysis in regard to industrial methods of
86
metalworking fluids neutralization showed four basic groups of methods: chemical, physical,
mechanical and thermal. Validity of method application and its process variants is related to
ecological, economical and technical aspects.
GENESIS AND PROBLEM FORMULATION - OBJECTIVES
Presented in [9] literature analysis and discussion on electrocoagulation (EC) modelling
problem showed lack of mathematical models describing all unit operations of EC process.
Most of model equations considers only selected unit operations, and methodology of their
creation signalize use of empirical approach of modelling; which intrinsically do not consider
occurrence of different interactions, constituent mechanisms. Presented ascertainment was
recognized as justification for attempt to synthesise complex electrocoagulation model, which
contains all unit operations and interactions between mechanisms.
Attempt will be made to evaluate oil-in-water emulsions EC process efficiency defined as
Chemical oxygen demand (COD) abatement as a measure of break-up result; determined as a
result of conversion of COD to normative process efficiency. Resulted emulsion break-up
efficiency is connected with energetic efficiency of the process. This ensures model, which
guarantees description of cause and effect interactions. Energetic efficiency of the process,
which is a measure of EC process applicability for technical problems is defined as electrical
energy E demand necessary for COD abatement to predefined level. Possible ways to
minimize of electrical energy consumption are interelectrode gap, conductivity and electrode
material change.
ELECTROCOAGULATION
One of the methods used in chemical engineering is EC – it belongs to wide group of
electrochemical processes. This method is described by the synergy interactions of many
mechanisms – connected with electrochemistry, coagulation, flotation. Electrocoagulation is
based, in its first period on electrolysis of polluted wastewater with use of metal anode
(usually made of iron or aluminium). Metal ions dissolved from the anode bind with OH- to
make metal hydroxides. Insoluble in water hydroxides exhibit high adsorption capability of
pollution particles. Hydrogen and oxygen bubbles produced during electrolysis, rise
aggregates on the fluid surface; where they can be disposed by e.g. skimming or blowing off
(Fig. 1a).
87
Figure 1 a) Pollutant separation processes in an ECF reactor [6]; b)Venn diagram of EC process -
„synthesis” of technology [8].
Transfer of the electrical charge between electrodes embedded in the emulsion induce
electrode reactions, which for electrodes made of metal M can be written in form of equations
according to Mollah [13]:
Anode: M(S) → M(aq)n+
+ ne-
Side process, anodic: 2H2O(l) → 4H+
(aq) + O2(g)↑+4e-
Cathode: M(aq)n+
+ne- → M(S)
2H2O(l) + 2e- → 2OH
- + H2(g)↑
Hydroxide ions OH- produced on cathode bind with metal ions M(aq)n+ (product of anodic
reactions) to make metal hydroxides and polyhydroxides – necessary for oil sorption process.
Aluminium ions (Al3+
) made in anodic reactions during electrolytic dissolution process
hydrolyse spontaneously to produce different monomeric species. The kind of the hydrolysis
process species and their amount is related inter alia to pH of the solution.
Because of energy efficiency and process efficiency increase electrocoagulation is
competitive and efficient also in treatment of wastewaters with heavy metal, dyes, organic
matter (COD, BOD5), solid suspensions, colloids and even arsenates [14].
Electrocoagulation is an evolving technology although it is known for about hundred years.
However in last several years increase of an importance of the EC method and relevant
technical EC applications can be observed. Nevertheless no EC process based models (which
encompass complex physical and chemical relationships) and systematic reactor design rules
(based on this relatively cheap method) have been elaborated. Synthesis technology of EC
process is presented in form of Venn diagram (made of three fundamental technologies -
electrochemistry, coagulation and flotation). Venn diagram presents complexity of EC
process (Fig 1b); EC process is made of three foundation areas which bring certain prospects
88
to electrocoagulation’s complexity. It also presents how EC can be simplified with use of
reductionist approach. Each lobe of the Venn diagram explains relevant phenomena,
characterization methods and tools. In the intersection between two lobes, knowledge that
links the foundation areas is presented. Present state in regard to modelling of EC process can
be called ‘enlightened empiricism’. To cause technology progress in this matter some level of
mechanistically based mathematical modelling is almost certainly required. It is necessary to
quantify the interactions, that occur between each of the three underlying technologies for a
range of the systems, where the pollutant itself can be readily quantified [8].
ELECTROCOAGULATION MODELLING IN LITERATURE
In many fields of process engineering short cut methods (based on correlation of experimental
data with consideration of dimensional analysis) have been successfully applied as a
design/operation tool; before finding overall physical description of the process and
mechanisms and interactions. However EC related literature is a lack of this methodology –
what is a confirmation of the process complexity. Authors think, that above mentioned model
building methodology will not cause EC development beyond actual state of “enlightened
empiricism”. Correct way is an application of mathematical modeling and concentration of
works on quantifying of mechanical interaction mechanisms of unit operations being
foundation of EC method; necessity of model validation impose of limitation of its
applicability for case of unequivocally quantitative defined pollution [8]. Analysis of known
and newly developed methods of model building and its simulation investigations, indicates
big chance for solving of this problem could be the use of web based environment MOSAIC.
This environment is a platform which enables creation, renewable use, cooperation between
different models and results of experimental investigations. Actually works are made to
implement in MOSAIC environment of systematic scheme of model building presented in [7];
this scheme can be used by the preparation, simplification and modification of complicated
nonlinear optimization models.
Literature research shows existence of models connected with electrochemistry,
coagulation and flotation problematic. During initial analysis and evaluation we distinguished
following modeling approaches which purpose was estimation of :
a) electrolysis voltage necessary for EC process [5],
𝑈𝑜 = 𝐴 + 𝜂𝑎,𝑝 +𝑑
𝜅𝑗 + 𝐾1ln𝑗 – Total required electrolysis voltage for passivated electrodes
b) pollution disposal efficiency with consideration of electrophoretic velocity of colloidal
particle move toward anode (Fe) [12],
89
c) oil suspension removal from water [2]. Final model is formed by three equations:
𝜕𝐶𝐴𝑙(𝐼𝐼𝐼)𝑐
𝜕𝑡= −𝑢
𝜕𝐶𝐴𝑙(𝐼𝐼𝐼)𝑐
𝜕𝑥+
Φ
ℎ
𝑗
𝑛𝐹𝑀𝐴𝑙(𝑂𝐻)3
𝑑𝐶𝐴𝑙(𝐼𝐼𝐼)𝑉
𝑑𝑡= 𝐹𝑟(𝐶𝐴𝑙(𝐼𝐼𝐼)
𝐶 |𝑥=𝐿 − 𝐶𝐴𝑙(𝐼𝐼𝐼)
𝐾𝐼𝐼𝐶𝐴𝑙(𝐼𝐼𝐼)𝐶∗𝑚 + 𝐶∗ − 𝐶0 = 0 – for Freundlich isotherm
where: 𝐶𝐴𝑙(𝐼𝐼𝐼)𝑐 – aluminum hydroxide concentration in the electrocoagulation cell
𝐶𝐴𝑙(𝐼𝐼𝐼)- aluminum hydroxide concentration in the reactor, j – current density, L – electrode
length, u – superficial velocity, V – volume of the stirred tank reactor, Fr – liquid recirculation
flow rate, KII – conditional Freundlich stability constant, m – mass action stoichiometric
coefficient, C* - liquid phase equilibrium concentrations.
d) pollution disposal efficiency with consideration mechanisms of reactions of wastes and
aluminum coagulants produced in the EC process [3, 4, 12],
Khemis in [11] proposed following equations for COD abatement calculation:
𝑋 =[𝑆0]−[𝑆1]−[𝑆2]
[𝑆0]
where:
[𝑆2] =− {1 + 𝐾 (
𝐶𝐴𝑙 − 𝐶𝑚𝑖𝑛𝑛 − [𝑆𝑜])} + √{1 + 𝐾 (
𝐶𝐴𝑙 − 𝐶𝑚𝑖𝑛𝑛 − [𝑆𝑜])}
2
+ 4𝐾[𝑆𝑜]
2𝐾
𝐶𝐴𝑙(𝑡) =𝑀𝐴𝑙𝐼 ∫ Φ𝐴𝑙𝑑𝑡
𝑡
0
𝑧𝐴𝑙𝐹 𝑉 , where subscript Al is related to aluminum
Canizares in [4] proposed model for estimation oil-in-water emulsions break–up
efficiency. Model is built of following equations:
𝐶0 = 𝐶1 + 𝐶2 + 𝐶3 – total COD of the suspension
𝐶𝑟 = 𝐶1 + 𝐶3 – remaining COD
𝐶3 + 𝑚𝐴𝑙1 ⇄ 𝐶2 – remaining COD i.e. amount of oil that can be eliminated - but it is not ,
under given operation conditions
𝑚 = −47,83 ∙ ln(|𝐶𝑛𝐴𝑙|) − 360,05
𝐶2 = 𝐶0 − 𝐶1 − 𝑚(𝐴𝑙1 − 𝐴𝑙1𝑚𝑖𝑛) – denotes the amount of COD that is removed by the
attachment or adsorption of the oil droplets to particles of the precipitate, under given
operation conditions where: C1 – corresponds to the refractory COD, i.e., the COD that cannot
be removed by electro-destabilization, m – overall coefficient of adsorption (or attachment) of
oil droplets on the particles of precipitate. In order to quantify the effect of aluminum ions
present in the system, and the overall effect of their net charge, a parameter called CnAl (net
charge of the aluminum ions, mol dm-3
) was introduced:
90
𝐶𝑛𝐴𝑙 = 3 ∙ [𝐴𝑙3+] + 2 ∙ [𝐴𝑙𝑂𝐻2+] + [𝐴𝑙(𝑂𝐻)2−] − [𝐴𝑙(𝑂𝐻)4
−]
e) pollution disposal efficiency based on the empirical model built with consideration of
operation parameters of the process [6], Defluoridation efficiency was described with
equation 𝐸 = 1 −𝐶𝑡
𝐶0 , where fluoride concentration at any time t is expressed by equation
𝐶𝑡 = 𝐶0𝑒−10−5[5,9(𝐼
𝑉)−37,1(𝐶0)−82,1(𝑑)+2746,4]𝑡
, where Co is initial fluoride concentration,
detention time can be expressed by equation 𝑑𝑡 =1
𝐾ln (
𝐶0
𝐶𝑡)
f) pollution disposal efficiency with consideration of hydrodynamic conditions related to
electrochemical properties of EC method; with particular consideration of additional
pneumatic mixing in the reactor volume [16],
g) measures, which describe EC sedimentation process (np. sedimentation velocity) [11],
h) chemical composition of aluminum anodes– with consideration of energy efficiency and
unit productivity of the process – empirical model [15].
MODEL SYNTHESIS WITH MOSAIC
Based on the literature survey following parameters variables were chosen:
• process variables like electrolysis current, voltage, COD abatement level – which can be
treated as quantifier of emulsion break – up,
• construction and material variables like volume of the reactor, electrode area,
interelectrode gap,
• properties of neutralized fluid like conductivity, variable which describes assumed COD
quantifier.
Model should enable estimation of optimal design parameters of the reactor and selection of
optimal process parameters under consideration of: electrolysis time, energetic efficiency of
the process and emulsion break-up efficiency. Presented Venn diagram analysis which
signalized multiplicity of unit operations being part of EC process – authorize to make an
attempt of choosing MOSAIC software for synthesis of models: model used for estimation of
necessary EC electrolysis voltage [5] and model which describes emulsion break-up
efficiency; which measure is COD abatement – used by Khemis [12]. Presented models
synthesis can be used for estimation of relationship between COD abatement, current
efficiency and electrolysis time. As presented in Figure 2a time necessary for abatement of
pollution increases with minimizing electrolysis current value. It has to be mentioned that
form of presented surface is related at first to Faradays law, which combines amount of
dissolved material from electrode in anodic process with current and electrolysis time value
91
and fact, and at second to the fact that initiation of the emulsion break-up process occurs after
exceeding minimal amount of aluminium hydroxides concentration CAl > Cmin in the
emulsions volume. Begin of the emulsion break-up process is related with energy
consumption necessary to produce necessary aluminium hydroxides concentration (Figure
2b).
Energetic process efficiency is important measure of EC method applicability for technical
problems. It can be defined as an amount of energy E necessary to achieve predefined COD
abatement level X. Energetic efficiency change is possible through a change of interelectrode
gap, emulsions conductivity and selection of electrodes material etc. Electrolysis resistance
decreases with minimizing of interelectrode gap d, with consideration that produced sludge
assembling in interelectrode area can relevantly block current flow between electrodes and
therefore decrease positive effect of interelectrode gap decreasing. Way for partial
elimination of this negative influence of break-up products is increase of volumetric flow in
the reactor.
Figure 2 a) COD abatement in function of current I and time t; b) Electrical energy demand for
realization of electrocoagulation process in function of electrolysis time for different COD abatement X
values.
Influence of emulsions conductivity (which relevantly influence electrolysis resistance and
therefore results in increase of energetic efficiency of the process) has been presented in
Figure 3; relationship between electrolysis voltage from electrolysis current and interelectrode
gap; conductivity of the emulsion was set on the diagram in form of the parameter
Results of different investigations [4, 10] signalize, that amount of electro dissolved
aluminium is related to electrochemical (related to current value) and intermetallic corrosion
(related to chemical composition of the anode) component. Knowing that emulsion break-up
process initiation occurs after exceeding minimal concentration of aluminium hydroxides in
emulsion volume – it can be concluded, proper selection of chemical composition of anode
92
material, advantageous from intermetallic corrosion point of view can cause increase of
Faradayic yield . It means that electrolysis time necessary to provide necessary Al ions
concentration CAl > Cmin can be decreased. In Figure 4 comparison of electrical energy
necessary for emulsion break-up process realization achieved for four kinds of aluminium
electrodes with modified chemical composition is shown. Histogram analysis justifies
formulation of the question about time necessary for creation of aluminium compounds
concentration which leads to predefined COD abatement; and therefore expected emulsion
break-up efficiency.
Figure 3 Electrode potential in function of current I and net distance between electrodes: a)
𝜿 = 0,04 mho/m, b) 𝜿 = 0,12 mho/m
Figure 4 Electrical energy demand for break-up 1dm3 of oil - in – water emulsion during
electrocoagulation process for different anode materials (cathode material: Aluminum), A1 – Al 100%
(Al), AlCu4MgSi alloy – Al 95%, Cu 4,3% (P6), ZnAl28Cu4 alloy - PN-H-87102:1980 – Al 28%, Cu 4,2%,
Zn 62% (ZN), alloy made of Al 90%, Mg 10% (AG).
For presentation of electrolysis current, interelectrode gap, emulsion conductivity on the
electrolysis time and energy consumption (necessary for emulsions break-up to the predefined
abatement level X) diagrams E in relationship to t were presented in logarithmic scale (see
Figure 5 to 6). Variables values were taken from set d = {4; 6; 8} mm, ={0,04; 0,08; 0,12}
mho/m, = {1; 1,5; 2}, I = {0,5; 1; 2}A.
127.5
103.12
49.69
24.37
0
50
100
150
Al. P6 ZN AG
Ele
ctri
cal e
ne
rgy
ne
cess
ery
fo
r p
roce
ss
real
izat
ion
[W
h/d
m3]
Anode material
93
It has to be mentioned that low current electrocoagulation process is advantageous from
electrical energy consumption point of view, however in consequence it results in longer
electrolysis time. Consideration of limitations in form of process productivity optimal
process parameters selection will be compromise between energy consumption and
electrolysis time.
Figure 5 Electrical energy demand for X=0,86 COD abatement of 1dm3 of oil-in-water emulsion during
electrocoagulation process in function of electrolysis time: a) for different conductivity values , b) for
different net distance between electrodes d.
Figure 6 Electrical energy demand for X=0,86 COD abatement of 1dm3 of oil-in-water emulsion during
electrocoagulation process in function of electrolysis time a) for different Faradaic Yield values, b) for
minimal and maximal energy conditions (d, ).
Emulsions conductivity influence on energy consumption and electrolysis time for set
electrolysis current, interelectrode gap and current efficiency was presented in Figure 5a.
Analogically in Figure 5b influence of interelectrode gap on energy consumption and
electrolysis time was presented. Decrease of energy consumption (for set electrolysis time)
can be achieved through conductivity increase of emulsion (e.g. use of NaCl) and decrease of
94
interelectrode gap. As Faradayic yield increases emulsions break-up time shortens (Fig.6a).
Considering above presented relationships optimal operation, construction parameters can be
found which are advantageous for criterion of electrical energy demand and emulsion break-
up time. For example in Figure 6b relationship electrical energy and electrolysis time was
presented for best and worst case scenario (under consideration energetic energy efficiency
and electrolysis time). Application of d=4 mm, =0,12mho/m, and =2 results in over 10
time decreasing of energy consumption in comparison to worst variant d=12mm,
K=0,04mho/m, = 1.
CONCLUSIONS AND FINAL REMARKS
Presented considerations acknowledged in general possibility of mathematical model building
of EC process, which connects energy consumption with assumed process effects; with
consideration of all process relevant variables and parameters.
In particular use of MOSAIC methodology acknowledged numerous interactions between
constituent process unit operations – and signalized agreement with phenomenological model.
For assumed EC process efficiency measure as COD abatement level in water, minimal
electrical energy necessary for emulsions break-up process initiation and termination, and
maximal possible COD abatement in relationship of voltage-current relation, electrolysis
time, emulsions conductivity, interelectrode area and Faradayic Yield were presented.
Additionally analysis of Figures 3 to 6 showed numerous interaction mechanisms, signalized
possibility of finding optimal EC process parameters based on min- max criterion. Necessity
of consideration in modelling – next to voltage-current parameters and time; construction
parameters (interelectrode gap, electrode material) and emulsion conductivity was
acknowledged.
REFERENCES
[1] Bard, A. J. Stratmann, M.: Encyclopiedia of Electrochemistry, Volume 5:
Electrochemical Engineering, Edited by Macdonald D. D., Schmuki P. 1 Auflage.
Weinheim Wiley-VCH Verlag GmbH & Co. KGaA 2007
[2] Carmona M., Khemis M., Leclerc J-P, Lapicque F., 2006. A simple model to predict the
removal of oil suspensions from water using the electrocoagulation technique. Chem.
Eng. Sci., 61, 1237-1246
95
[3] Canizares P., Martinez F., Rodrigo M. A., Jimenez C., Saez C., Lobato J., 2008.
Modelling of wastewater electrocoagulation processes Part I. General description and
application to kaolin – polluted wastewaters. Sep. and Pur. Technol., 60, 155-161
[4] Canizares P., Martinez F., Rodrigo M. A., Jimenez C., Saez C., Lobato J., 2008.
Modelling of wastewater electrocoagulation processes Part II. Application to dye-
polluted wastewaters and oil-in-water emulsions. Sep. and Pur. Technol., 60, 147-154
[5] Chen X., Chen G., Yue P.L., 2002. Investigation on the electrolysis voltage of
electrocoagulation. Chem. Eng. Sci., 57, 2449-2455
[6] Emamjomeh M.M., Sivakumar M., 2006. An empirical model for defluoridation by
batch monopolar electrocoagulation/flotation (ECF) process. J. of Haz. Mater., B 131,
118-125
[7] Esche E., Müller D., Kraus R., Wozny G. 2014. Systematic approaches for model
derivation for optimization purposes. Chem. Eng. Sci., DOI: 10.1016/j.ces.2013.11.041
[8] Holt P. K., Barton G. W., Mitchel C. A., 2005. The future for electrocoagulation as a
localized water treatment technology. Chemosphere, 59, 355-367
[9] Janczewski Ł., Dyląg M.. 2014. Wybrane zagadnienia modelowania procesu
elektrokoagulacji w zastosowaniu do rozdziału emulsji olejowo-wodnych. Inż. Ap.
Chem. 53, nr 2, 76-79
[10] Khemis M., Leclerc J.-P., Tanguy G., Valentin G., Lapicque F., 2006, Treatment of
industrial liquid wastes by electrocoagulation: Experimental investigation and an overall
interpretation model. Chemical Engineering Science, 61, 3602-3609.
[11] Lai C. L, Sheng H. L., 2004, Treatment of chemical mechanical polishing wastewater
by electrocoagulation: system performances and sludge settling characteristics,
Chemosphere 54, 235–242
[12] Matteson J. M., Dobson R. L., Glenn R. W., Kukunoor N. S., Waits III W. H., Clayfield
E. J.: Electrocoagulation and separation of aqueosus suspensions of ultrafine particles,
1995, Colloids and Surfaces, A: Physiochemical and Engineering Aspects, 104, 101 –
109
[13] Mollah M. Y.A., Morkovsky P., Gomes J. A. G., Kesmez M., Parga J., Cocke D. L.,
2004. Fundamentals, present and future perspectives of electrocoagulation. J. Hazard.
Mater., B 114, 199-210
[14] Mohammad M. E., Muttumcaru S., 2009. Review of pollutants removed by
electrocoagulation and electrocoagulation/flotation processes. J. of Env. Manag.,90,
1663-1679
96
[15] Polowski W., Janczewski Ł., Czechowski K., Wronska I., 2014. Neutralizacja zużytych
emulsji olejowych metodą elektrochemiczną – analiza i ocena wpływu materiału
elektrod na efektywność procesu. Inż. Ap. Chem. 2014, 53, nr 3, 170-171
[16] Szpyrkowicz L., 2005, Hydrodynamic effects on the performance of electro-
coagulation/electro-flotation for the removal of dyes from textile wastewater . Ind. Eng.
Chem. Res., 44, 7844–7853
[17] Raport of Kline Group: Report II Y650B, Metalworking Fluids Global Series 2010:
Europe Market Analysis and Opportunities, 2011.
NOTATION
CAl Concentration of dissolved Al, mg l-1
Cmin minimal value of Al concentration to allow start – up of electrocoagulation, mg l-1
F Faraday`s constant, 96487 A.s eq-1
I current, A
K equilibrium constant of the electrocoagulation process, l mg-1
M molecular weight kg mol-1
n overall coefficient of the coagulation process, mg Al (III)/mg O2
So intrinsic COD of the suspension at time t, mg O2l-1
S1 COD variation of the suspension induced by pH change, without Al dissolution, mgO2l-1
S2 COD of the suspension which can be removed by electrocoagulation, mgO2l-1
X COD abatement,
z number of electrons involved in the electrode reaction
V Volume of treated liquid, [m3]
d net distance between electrodes, m
j current density, A/m2
A, K1 constants
U total required electrolysis voltage of an electrocoagulation process, V
Uo electrolysis voltage between electrodes, V
Greek letters
Φ Faradaic yield
𝜅 conductivity of water/wastewater treated, mho/m
97
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
DISPERSION AND COALESCENCE IN STIRRED MICELLAR MULTIPHASE
SYSTEMS
L. HOHL; N. PAUL; M. KRAUME
TU Berlin, Chair of Chemical & Process Engineering, Ackerstraße 76,
13355 Berlin
Abstract. Micellar multiphase systems are one of several promising possibilities
to enable the hydroformylation of long-chained alkenes. Aim of this study was to
investigate relevant characteristics of micellar systems. Therefore, the interfacial
tension and phase behavior were determined. Dispersion conditions and
coalescence were analysed using an endoscope measurement technique in a stirred
tank. The mean sauter diameters of the dispersed phases were quantified as a
function of stirring frequency and amount of surfactant.
Keywords. Micellar multiphase systems, Coalescence, Endoscope, Stirred tank
List of symbols.
α mass fraction of organic phase [-] t time [min]
c molar concentration [mol/l] T temperature [°C]
d32 mean sauter diameter [μm] γ mass fraction of surfactant [-]
m mass [g] y interfacial tension [mN/m]
INTRODUCTION
The utilisation of renewable resources as educts is a major advantage concerning the
production of basic chemicals. The hydroformylation is a chemical reaction used for the
conversion of alkenes to aldehydes, which may furthermore be converted into different
secondary products. One of the most common industrial applications is the
98
Ruhrchemie / Rhône-Poulenc process, which relies on a water soluble rhodium complex as
catalyst. This catalyst offers significant benefits, because it enables a homogenous liquid-
liquid reaction with high and specific yield among mild reaction conditions [1]. Furthermore,
it is preferable to use water as reaction medium from an economic and environmental point of
view. One major disadvantage of the process is that only alkenes with relatively short chains
of carbon atoms can be converted. Due to their hydrophobic structure, long-chained alkenes
are hardly soluble in water. Thus, the catalysis is hindered and the process economical
unviable due to low reaction rates. Micellar multiphase systems are one possibility to
overcome these obstacles and to utilise the mentioned advantages of the Ruhrchemie / Rhône-
Poulenc process [2-4].
Disregarding the catalyst and synthesis gas, the micellar systems mainly consist of
water, organic phase and surfactant. The use of surfactant reduces the interfacial tension and
enhances the solubilisation of the organic phase in water. The phase behaviour of the systems
is mainly dependent on the temperature and composition, and can be specified using the mass
fractions:
𝛼 =𝑚𝑜𝑟𝑔𝑎𝑛𝑖𝑐
𝑚𝑜𝑟𝑔𝑎𝑛𝑖𝑐 + 𝑚𝑤𝑎𝑡𝑒𝑟
𝛾 = 𝑚𝑠𝑢𝑟𝑓𝑎𝑐𝑡𝑎𝑛𝑡
𝑚 𝑠𝑢𝑟𝑓𝑎𝑐𝑡𝑎𝑛𝑡 + 𝑚𝑜𝑟𝑔𝑎𝑛𝑖𝑐 + 𝑚𝑤𝑎𝑡𝑒𝑟
Figure 1 illustrates the phase behaviour of a micellar system at a constant ratio of
water to organic phase (e.g. α = 0,5). At low temperatures, the system develops an aqueous
phase which contains the surfactant and is covered by the organic phase. At high
temperatures, a phase inversion of this two-phase system occurs. If very high amounts of
surfactants are used, the system only consist of one phase (microemulsion) which exists
within a wide range of temperature.
Figure 1: Phase behavior as a function of temperature and amount of surfactant [4]
99
Within the central area of the T-γ-graph, three different phases can be observed.
Previous studies discovered that optimal reaction conditions for the hydroformylation occur
within this three-phase system [1]. It consists of one aqueous and one organic phase
respectively, which are separated by a bicontinuous middle phase containing high amounts of
surfactant (Fig. 2). The interfacial area is crucial for mass transport of educts and catalyst
during the hydroformylation, since a vast mass transfer enables high reaction rates and
economic feasibility. Therefore, detailed knowledge of the dispersion conditions is necessary.
Figure 2: Possible dispersion conditions of micellar multiphase systems [5]
Hamerla et al. investigated dispersion conditions and mass transport in micellar
systems in a stirred tank [5]. In this paper, a similar approach was chosen to evaluate the
influence of the surfactant Marlophen NP7. The aim of this research is to investigate the
relevant characteristics and quantify the impact of different parameters on phase behaviour,
dispersion and coalescence.
EXPERIMENTAL SECTION
In this chapter, the chosen micellar systems as well as the experimental methods and set-ups
are described. The characterisation of the systems included the determination of the interfacial
tension and phase behaviour. Additionally, the dispersion conditions and coalescence were
estimated via an endoscope measurement technique.
Materials
The micellar system consisted of distilled water, 1-dodecene (Merck GmbH, purity ≥ 96%)
and Marlophen NP7 (Sasol Germany, technical grade). All components were used without
further treatment.
100
Interfacial tension
The interfacial tension is one important characteristic of micellar systems, because it
influences the dispersion conditions and solubilisation. Therefore, the interfacial tension was
estimated using pendant drop and spinning drop method for different solutions of
water / Marlophen NP7 in air and in 1-dodecene. The impact of surfactant concentration on
the interfacial tension was quantified [6-7].
Phase behaviour and conductivity measurements
The phase behaviour and change in conductivity was documented for different amounts of
surfactant as a function of temperature. A small double-walled sample container made of
glass was used, which could be tempered via a thermostat (Lauda E200). The temperature was
varied from 40 to 65°C. The observed phase behaviour was compared to the results of
conductivity measurements in stirred systems. To enable this, small amounts of KCl were
added. Conductivity should only occur when water is the continuous phase.
Endoscope Measurement Technique
To evaluate the dispersion conditions and the drop size distribution, an endoscope
measurement technique was used. Figure 3 shows the experimental set-up and dimensions. A
double-walled glass tank was tempered via a thermostat. A Rushton Turbine was used for
stirring at different frequencies.
Figure 3: Experimental set-up and dimensions of the stirred tank [8]
101
Images of the dispersion were taken using an endoscope and a camera. Afterwards, the mean
sauter diameters of the dispersion were estimated via Image analysis. The endoscope
measurement technique and details of image analysis were developed and reviewed by Maaß
et al. in several journal articles[8-9].
RESULTS AND DISCUSSION
In this chapter, the interfacial tension was determined as a function of surfactant
concentration. Additionally, the impact of temperature and amount of surfactant on the phase
behaviour was investigated. The temperature interval of the three-phase system was identified
and used for the investigation of dispersion and coalescence in the stirred tank.
Interfacial tension
The interfacial tension of the drops with different amounts of surfactant was measured via
pendant drop method. The surfactants adsorb at the interfacial area, until it is fully occupied
and the interfacial tension reaches a stable value. The critical micelle concentration (cmc) of
the system water / Marlophen NP7 in air was determined to ccrit = 10-4
mol/l at room
temperature. For higher surfactant concentrations, a nearly constant value was observed. The
interfacial tension of water / 1-dodecene was y ≈ 33 mN/m. The interfacial tension of
water / Marlophen NP7 in 1-dodecene could not be entirely estimated by pendant drop
method, because high amounts of surfactant led to very low surface tensions and instable
drops. For this reason, the spinning drop method (SVT-20, Dataphysics) was used for further
investigations.
Figure 4: Snapshot of a spinning drop in three-phase system. Two different phase interfaces could be
observed at T=57-60°C
y ≈ 0,34 mN/m
y ≈ 0,06 mN/m
continuous phase
phase 1 phase 2
102
Thus, the interfacial tension of the systems could be estimated even with high amounts of
surfactant. A mixture of water and Marlophen NP7 (cMarlophen = 0,1 mol/l) was used as the
continuous phase. Figure 4 shows a cylindrical drop in the capillary tube. At a temperature of
approximately 57 - 60 °C, a second interface within the drop could be observed. Apparently, a
drop of 1-dodecene (phase 1) was surrounded by a drop of bicontinuous fluid (phase 2). The
interfacial tensions were determined to y ≈ 0,34 mN/m and y ≈ 0,06 mN/m.
Phase Behaviour
The phase behaviour was determined as a function of temperature for different amounts of
surfactant (Fig. 5). Starting from low temperatures, the conductivity was almost constant and
water was the continuous phase. Afterwards, the conductivity dropped rapidly within a
specific temperature interval, which was 50 to 63°C for γ = 0,1 and 52 to 54°C for γ = 0,075.
Thus, the length of the interval is a function of the surfactant concentration. A three phase
system was observed within these temperature intervals. The low conductivity at higher
temperatures indicated a phase-inversion, so 1-dodecene became the continuous phase.
Observed hysteresis effects were comparatively small.
The temperatures leading to three-phase systems were chosen for further experiments to
evaluate the drop size distributions and dispersion conditions.
0.0
0.2
0.4
0.6
0.8
1.0
45 50 55 60 65
no
rme
d s
pe
cif
ic c
on
du
cti
vit
y κ
[-]
Temperature T [°C]
y = 0,1
y = 0,075
Figure 5: Conductivity measurements using KCl-ions in a stirred system (α = 0,5)
103
Drop Size Distribution
Two different kinds of drops were observed, representing the two dispersed phases within the
continuous aqueous phase. The drops could be distinguished optical and were identified as 1-
dodecene (dark, clear drops) and bicontinuous phase (bright, hazy drops). Different types of
drop interaction were observed in the stirred tank (Fig. 6).
To evaluate the drop sizes, the phases were separately counted and measured. The ratio of 1-
dodecene to water was held at a constant value of α = 0,5. The used amounts of surfactant
were γ = 0,075 and γ = 0,1. Figure 7 shows the mean sauter diameters as a function of time
for both dispersed phases.
Figure 7: Mean sauter diameters of both dispersed phases
Figure 6: Observed drops and interactions in the stirred tank
104
The mean sauter diameter reached a nearly constant value after approximately 40 minutes of
stirring at a constant frequency of 400 min-1
. After 60 minutes, the frequency was changed
abruptly to 200 min-1
leading to a considerable increase of the mean sauter diameters of both
dispersed phases. Even with high amounts of surfactant, this coalescence effect was observed.
The drop interactions seemed to depend on the amount of surfactant, but not on the stirring
frequencies. Beneficial for the hydroformylation are interacting drops, because of the larger
contact area of the phases, which support mass transport and therefore reaction rates.
CONCLUSIONS
In this work, micellar multiphase systems consisting of water / Marlophen NP7 and 1-
dodecene were characterised. First, the interfacial tension and critical micelle concentration of
water / Marlophen NP7 in air was measured via pendant drop method. Thus, the interfacial
tension of water / Marlophen NP7 in 1-dodecene was analysed using the spinning drop
method. A three-phase system was observed at specific temperatures, leading to two different
cylindrical drops within the continuous phase.
The phase behaviour of the systems was determined as a function of temperature and amount
of surfactant. The continuous phase was identified and the phase inversion examined via
conductivity measurements. The temperature interval leading to three-phase systems was
quantified and used to investigate the dispersion conditions and coalescence behaviour of the
system.
Using an endoscope measurement technique, two dispersed phases were observed within the
ambient aqueous phase. The different drops could be identified optical due to their
appearances. The drop interactions depended mainly on the amount of surfactant, but not on
the stirring frequency. The mean sauter diameters of both dispersed phases were determined
as a function of time. Therefore, different amounts of surfactant were used and the stirring
frequency was varied. The mean sauter diameter of both dispersed phases reached a stable
value after a specific stirring time. Even with high amounts of surfactant, coalescence could
be observed while reducing the stirrer frequency.
ACKNOWLEDGMENTS
This work is part of the Collaborative Research Center “Integrated Chemical Processes in
Liquid Multiphase Systems” coordinated by the Technische Universitat Berlin. Financial
support by the Deutsche Forschungsgemeinschaft is gratefully acknowledged (TRR 63).
105
REFERENCES
[1] Kohlpainter, C.W. et al.: Aqueous biphasic catalysis: Ruhrchemie/Rhone-Poulenc oxo
process. Appl. Catal. A: General, vol. 221, no. 1–2, pp. 219–225, 2001.
[2] Zagajewski, M. et al.: Continuously operated miniplant for the rhodium catalyzed
hydroformylation of 1-dodecene in a thermomorphic multicomponent solvent system
(TMS). Chem. Eng. Sci., vol. 115, pp. 88–94, 2014.
[3] Hamerla, T. et al.: Katalyse in modifizierten Flüssig/flüssig-Mehrphasensystemen (in
german), Chemie Ing. Tech., vol. 84, no. 11, pp. 1861–1872, 2012.
[4] Rost, A.: Rhodium-katalysierte Hydroformylierung von 1-Dodecen mit zweizähnigen
Ligangen in Mikroemulsionssystemen (in german). Technische Universität Berlin,
2013.
[5] Hamerla, T. et al.: Aufklärung der Stofftransportwege in mizellaren
Mehrphasenreaktionen am Beispiel der Hydroformylierung (in german). Chemie Ing.
Tech., no. 10, pp. 1530 – 1539, 2013.
[6] Dörfler, H.-D.: Grenzflächen und kolloid-disperse Systeme: Physik und Chemie (in
german). Berlin, Heidelberg: Springer Verlag, 2002.
[7] Princen, H.M.: Measurement of Interfacial Tension from the Shape of a rotating drop.
Colloid Interface Sci., vol. 23, no. 1, pp. 99–107, 1967.
[8] Maaß, S. et al.: Influence of the dispersed phase fraction on experimental and predicted
drop size distributions in breakage dominated stirred systems. Chem. Eng. Sci., vol. 76,
pp. 140–153, 2012.
[9] Maaß, S. et al.: Automated drop detection using image analysis for online particle size
monitoring in multiphase systems. Comput. Chem. Eng., vol. 45, pp. 27–37, 2012.
107
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
RELATIVE PARTICLE-TO-FLUID VELOCITY IN A TURBULENT FLUID
P. DITL; J. SKŘIVÁNEK; V. PEŠAVA
Czech Technical University in Prague, Faculty of Mechanical
Engineering, Department of Process Engineering, Technická 4, 166 07
Prague 6, email: [email protected]
Abstract. The settling particle velocity in a turbulent field affects many transfer
phenomena, e.g. off-bottom particle suspension, particle break up and attrition,
mass and heat transfer at dissolution and crystallization, polymerization,
fluidization, biochemical reactions, heterogeneous chemical reactions, etc. Most
of these operations occur in turbulently agitated tanks or in vertical tubes, so it is
of crucial importance to determine the relative particle-to-fluid velocity. Most of
the existing approaches use the settling particle velocity in a still liquid for
designing processes. However, the settling particle velocity in a still liquid and the
settling particle velocity in a turbulent field may differ significantly from each
other. The goal of this work is to develop a methodology for calculating the
settling particle velocity in a turbulent field. The model of Ditl and Skrivanek [1]
proposed for sinusoidal turbulent velocity profile is used in this methodology.
Calculations reported in this paper proved that the ratio of a real sedimentation
velocity in a turbulent field to the terminal settling velocity drops to a value of
0.45 and then increases back to 1 when the particle size increases. This course is
not in contrast with our previous LDA/PDA experiments and practical
experiences gained in agitated and fluidized systems.
Keywords. Settling particle, liquid, turbulence, Matlab
INTRODUCTION
The design of process equipment requires an accurate determination of the relative
particle-to-fluid velocity. This is often achieved by an experimental investigation of the
influence of a turbulent field on the settling particle velocity. The measured relative particle-
108
to-fluid velocity is usually reported as a function of the Stokes number or the Kolmogoroff
micro-length scale [13]. The Stokes number is defined as the ratio between the particle
relaxation time and the turbulent integral scale. An alternative method is to calculate the
relative particle-to-fluid velocity using the motion equation of a particle in a fluid. This
method is utilised in this work. The input values are the measured time dependencies of the
local axial component of the turbulent fluctuations. The results of the calculations are
subsequently compared with LDA/PDA measurements of the fluid and particle velocities in
an agitated vessel. Unfortunately, there can be considerable uncertainty of the relative
velocity determined during these measurements. The relative velocity is calculated as the
difference between particle velocity and fluid velocity. The uncertainties of each of these two
measurements have a significant impact on the uncertainty of the relative velocity that is
determined.
A SHORT CRITICAL LITERATURE REVIEW
An extensive critical search for publications dealing with this problem was recently accepted
for publication [13] in Chem.Eng.Sci.
Methods for measuring hindered sedimentation can be divided into two groups. The first
group comprises direct methods for measuring velocities using PIV, LDA/PDA or a high-
speed video camera. The second group comprises indirect methods for measuring velocities
using particle distribution measurements, particle suspension and mass transfer
measurements. The measured values are usually depicted using the Stokes number or the rate
of particle size and the integral length scale of the turbulence.
DIRECT METHODS
Šedivý et al [2] measured the difference between liquid and particle velocities in a
mixed vessel using the LDA/PDA method. A velocity decrease was observed by them in
comparison to the settling velocity in stagnant water.They observed a decrease in the settling
velocity caused by turbulence for glass beads and iron particles 1.85 mm and 0.33 mm in size
respectively for glass beads 1.85 mm in diameter and iron particles 0.333 mm in diameter.
This decrease was proportional to the relative particle size d/T and the rate of densities Δρ/ρ.
Yang and Shy [3] reported an increase in settling velocity in a turbulent field
generated by two oscillating grids. Heavy tungsten and glass particles in an aqueous near-
isotropic turbulence were investigated. The maximum increase was observed as the Stokes
number was near approached unity.
109
Doroodchi et al. [4] investigated the impact of turbulence on the drag coefficient for
particles of different sizes and densities, using a high speed video camera. Turbulence was
generated by two oscillating grids. Nylon and teflon spherical particles from 2.38 to 7.94 mm
in size diameter were examined. They reported that the settling velocity decreases with
increasing particle size. From the point where the particle size was equal to the integral length
scale of turbulence settling, the velocity increased with increasing particle size. A reduction in
settling velocity in a turbulent field was reported by these authors.
Ghatage et al. [5] measured the hindered settling slip velocity of steel particles in a
solid-liquid fluidized bed. The drag coefficient was increased by turbulence for all
experimental conditions.
Indirect methods
Magelli et al. [6] investigated the solid distribution for solid-liquid suspensions in tanks
stirred by multiple Rushton turbines. The solid concentrations were measured using an optical
technique, and the Peclet number was calculated from experimental data. The settling velocity
was subsequently calculated from the Peclet number and the dispersion coefficient. For larger
particles, the retardation of the settling velocities sank to 40 percent of the settling velocity in
a still liquid. However, for tiny particles (less than 10λ) the settling velocity remained
unchanged.
Brucato et al. [7] determined the settling velocity and the drag coefficient in Couette
flow, using the residence time technique. The particles used in the experiment were glass
beads 63-500 µm in size diameter and silica particles 180-500 µm in size diameter. They
reported that the settling velocity decreases with increasing turbulence intensity. An influence
of particle size and turbulence intensity on particle drag was found. Particle drag was either
unaffected or was increased by free stream turbulence. A new correlation for estimating
particle drag coefficients was proposed.
3
41076.8
0
0
d
C
CC
D
DD
Solid concentrations of glass and plastic particles of various sizes in tanks stirred by
multiple impellers were investigated both by Nocentini et al. [8] and by Pinelli et al. [9].
Particle settling velocities were subsequently determined from the definition of the Peclet
number. The retardation of the settling velocity was measured in agreement with the previous
correlation:
110
6.01/16tanh4.0/ dUU tS
THEORETICAL
The Ditl - Skrivanek model
The model of Ditl and Skrivanek [1] valid for sinusoidal velocity fluctuation is enlarged and
optimized in this paper. The balance of the forces acting on a particle forms the basis of this
model. Fluctuations in the local axial velocities are used to determine the drag force acting on
a particle.
According to [1], the basic equation for particle motion in a turbulent liquid velocity field is
as follows:
95.6880Re
Re318825.0)6459.0Re186.01(Re18
2
Re*
Arddt
d
S
Lv
(1)
where the dimensionless number is defined in [1] as:
2dDi
L
S and
uv ReReRe (2, 3)
In [1], the local turbulent velocity is simplified by a sinusoid, so the values of fluctuating Reu
can be expressed as:
)2sin(Re tad
u
(4)
The value of velocity amplitude a and frequency ω for the corresponding turbulent flow must
be determined experimentally or from CFD calculations.
The ratio X between the particle sedimentation velocity hindered by turbulence and the
terminal velocity used can be expressed as:
sedsed
uv
u
uv
vX
Re
ReRe
(5)
The terminal sedimentation velocity used is calculated in MATLAB from Equation 6, using the
Newton method.
095.6880Re
Re4251.0Re186.01Re24
3
4 36459.0
Ar
(6)
Equation (1) was solved in MATLAB using the Runge-Kutta-Fehlberg method (ODE45). The
same method was also used for the solution in EXCEL.
111
Calculation procedures
The velocities have been newly expressed in this work, using the Fourier transform. The
measured records of real velocities obtained by LDA [10] were first re-sampled to a constant
time step by the sample-and-hold method [11]. This treatment is a necessary condition for the
Fourier transform to work properly. Then the measured time dependencies were decomposed
into the frequency domain, using the Fourier transform. In the next step, the calculated
coefficients of the Fourier series were transformed using equations (7 - 9) into parameters
suitable for utilization in the MATLAB computational model. The computational model was
created using [12].
22)Im()Re( iii FTFTa (7)
T
ii (8)
i
i
iFT
FT
)Re(
)Im( (9)
These parameters are iia , and i in equation (10). FT is the Fourier transform of the
measured data. Finally, the recomposed time dependency of the velocities is the sum of the
sines created using the parameters.
)2sin(Re iiiu ta
d
(10)
The process data inputs for the calculations are listed in Table 1. The amplitudes and
frequencies for modeling the simplified sine liquid velocity are taken from [1]. The measured
data obtained by LDA for calculations with a real fluid velocity were chosen according to the
corresponding revolutions of the impeller. LDA measurements were performed on a tank
T=0.3 m in diameter agitated by a pitched six-blade turbine at impeller speeds of 450 and 600
rpm. The iron and glassy particles are considered in an aqueous suspension with density 998.7
kg.m-3
and viscosity 1.05477x10-6
m2.s
-1.
112
Table 1. - Input process data
Material Solid phase density
[kg.m-3
].
Amplitude
[m.s-1
]. Frequency [s
-1].
Revolutions of
impeller [min-1].
Glass 2640 0.2 12.8 450
Glass 2640 0.27 14.5 600
Iron 7400 0.2 12.8 450
Iron 7400 0.27 14.5 600
RESULTS
The results are plotted as dependencies of X on Di, or of X on a particle diameter, for
different materials and revolutions. A comparison of results calculated in Matlab and in Excel
[1] for glassy particles at revolutions of 450 rpm and 600 rpm are depicted in Figure 1.
Fig. 1. X as a function of Di for glassy beads at 450 rpm and 600 rpm. Comparison of Excel and Matlab
results
Comparisons of the results of calculations with a simplified sine liquid velocity and the real
liquid velocity performed in MATLAB are depicted in Figures 2 and 3.
113
Fig. 2. X as a function of particle diameter for glassy beads at 450 rpm and 600 rpm. Comparison of
experimental results with simplified and real velocities.
Fig. 3. X as a function of particle diameter for iron spheres at 450 rpm and 600 rpm. Comparison of
experimental results with simplified and real velocities.
Considering the 3D flow, we have to multiply the fluctuating velocities by 3 to express the
diagonal value of the fluctuating velocities. These correlated results are depicted in Fig. 4.
114
Fig.4. X as a function of particle diameter for iron spheres at 450 rpm and 600 rpm. Comparison
between experimental results and predicted velocities after 3D correction.
CONCLUSIONS
The following conclusions can be drawn:
A model has been proposed for describing the particle movement in a turbulent fluid,
where the local velocities are represented by a sine or an actual velocity profile.
The Fourier transform has been applied for evaluating the amplitudes, frequencies and
phase shifts for single sine waves.
Good consistency of the results obtained by Matlab and by Excel has been proved.
Ratio X aproaches a value of 1 for very fine particles and also for large particles,
where fluid fluctuations have only a small effect on the particle motion.
The medium region of particle sizes is the most interesting both from the theoretical
and from the practical point of view. In this region, X drops to a value of 0.45 and
then increases back to 1.
This fact has a significant impact on technical calculations of particle sedimentation
hindered by turbulence and mass and heat transfer coefficients in fluid-particulate
systems.
115
A comparison of X vs. d obtained for sinusoidal and real liquid velocities shows that
they have a similar course. However, the curves are postponed and results with a
sinusoidal velocity take lower values. A correction taking into account 3D flow
significantly improved the consistency between the predicted and measured values.
Experimentally obtained values for X using PDA [2] are not in contrast with the
calculated values.
ACKNOWLEDGEMENT
This research has been supported by the Grant Agency of Czech Republic under grant No.
P101-12-2274, and by the Grant Agency of Czech Republic under grant No.
SGS14/061/OHK2/1T/12.
SYMBOLS
Roman letters
Ar Archimedes number (1)
a amplitude of the fluctuating fluid velocity component (m·s-1
)
CD drag coefficient (1)
D shaft diameter (m)
Di dimensionless number defined by Eq. (2) (1)
d particle diameter (m)
g acceleration due to gravity (m·s-2
)
L shaft length (m)
n revolutions of impeller (s-1
)
Revsed particle Reynolds number at sedimentation by its terminal velocity
in a still fluid (1)
Reu local fluid Reynolds number defined by Eq. (4) and (10) (1)
Rev particle Reynolds number defined by Eq. (1) (1)
∆Re difference between the particle and fluid Reynolds numbers - see Eq. (3) (1)
T tank diameter (m)
t time (s)
116
u local fluid velocity (m·s-1
)
v particle local velocity (m·s-1
)
X ratio between the sedimentation velocity in a turbulent field and the
terminal velocity in a still liquid, -see Eq. (5) (1)
Greek letters
φ phase shift (1)
λ Kolmogoroff microscale (m)
ν kinematic viscosity (m2·s
-1)
L fluid density (kg·m-3
)
L particle density (kg·m-3
)
frequency (s-1
)
REFERENCES
[1] Ditl, P. - Skrivanek, J.: Sedimentation hindered by a turbulent sinusoidal velocity field. In
XXI Ogólnopolska Konferencja Inżynierii Chemicznej i Procesowej - Materiały
konferencyjne. Szczecin: Zachodniopomorski Uniwersytet Technologiczny w
Szczecinie,2013, p. 1-5.ISBN 978-83-7518-596-6.
[2] Šedivy, V. - Ditl, P. - Rieger, F. - Severa, M.: Dimensionless Flow Characteristics in
Mixed Suspension Obtained by LDA/PDA. In: Fluid Mixing 6. Rugby: Institution of
Chemical Engineers. 1999. p. 373-382.
[3] Yang, T.S., Shy, S.S.: The settling velocity of heavy particles in an aqueous near-isotropic
turbulence. Physics of Fluids, 15, 868-880, 2003.
[4] Doroodchi, E., Evans, G. M., Schwarz, M. P., et al.: Influence of turbulence intensity on
particle drag coefficients. Chemical Engineering Journal, 135, 129-134, 2008.
[5] Ghatage, S.V., Sathe, M.J., Doroodchi, E., Joshi, J.B., Evans, G.M.: Effect of turbulence
on particle and bubble slip velocity. Chem. Eng. Sci., 100, 120-136, 2013.
[6] Magelli, F., Fajner, D., Nocentini, M., Pasquali, G.: Solid distribution in vessels stirred
with multiple impellers. Chem. Eng. Sci., 45(3), 615-625, 1990.
[7] Brucato, A., Grisafi, F., Montante, G.: Particle drag coefficients in turbulent fluids,
Chemical Engineering Science, 53, 3295-3314,1998.
117
[8] Nocentini, M., Pinelli, D., Magelli, F.: Dispersion coefficient and settling velocity of the
solids in agitated slurry reactors stirred with multiple Rushton turbines. Chem. Eng.
Sci., 57, 1877-1884, 2002.
[9] Pinelli, D., Montante, G., Magelli, F.: Dispersion coefficients and settling velocities of
solids in slurry vessels stirred with different types of multiple impellers. Chem. Eng.
Sci., 59(15), 3081-3089, 2004.
[10] Pešava, V., Ditl, P.: Určení parametrů fluktuačních rychlostí turbulence v míchané
nádobě. (Determination of parameters of fluctuation velocities in an agitated vesel.) In
Procesní technika 2013. Praha: České vysoké učení technické v Praze, Fakulta strojní,
2013, s. 1-11. ISBN 978-80-01-05285-3(in Czech).
[11] Benedict, L.H., Nobach, H., Tropea, C.: Estimation of turbulent velocity spectra from
laser Doppler data, MEASUREMENT SCIENCE & TECHNOLOGY, 2000, vol. 11,
Issue: 8, p. 1089-1104.
[12] Zaplatílek, K., Donar, B.: MATLAB: Začínáme se signály. (We begin with signals) 1st
Edition. Praha: BEN - technická literatura, 2006. 271 s. ISBN 80-7300-200-0 (in
Czech).
[13] Swapnil V.G., Mayur J.S., Elham D., Joshi J.B., Geoffrey M.E.: Effect of turbulence on
particle and bubble slip velocity, Chem.Eng.Sci.,100 (30) 2013, Pages 120–136.
119
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
SYSTEMATIC ANALYSIS OF COALESCENCE IN LIQUID/LIQUID DISPERSIONS
J. VILLWOCK; J. KAMP; M. KRAUME
Chair of Chemical & Process Engineering, Technische Universität
Berlin, FH 6-1, Fraunhoferstr. 33-36, 10587 Berlin, Germany
Abstract. Dispersions of two immiscible liquids are an integral part of several
unit operations. The drop size distribution determines a decisive part of the overall
process efficiency and product quality. Hence, a reliable prediction of the drop
size distribution is necessary for the plant design and later the process control.
Using population balance equations, which base upon the two microscopic
phenomena drop breakage and coalescence, different scales of liquid/liquid
dispersions can be simulated. Consequently, a successful simulation of large scale
processes can be performed with this modelling approach if a detailed
understanding of the small scale single drop behaviour is gained. This work
presents the general bottom-up approach which is enabled by the use of
population balance equations and focusses on the coalescence process and its
investigation and description.
Keywords. coalescence, population balance equation, batch settling test, binary
collision.
INTRODUCTION
Population balance equations (PBE) were introduced in the 1960s to describe the
number and size distribution of a particle species (the population) varying with position and
time in a process and thus overcome the widely used simple assumption that the particle
environment is homogenous [1-3]. Balancing a certain control volume, the particle size
distribution is determined by the entering and leaving convective flows and particle source
and sink terms. These terms are used to shift particles from one size class to another, which
120
could be caused for instance by particle breakage, nucleation, growth or agglomeration. This
generic approach has been used since then to describe particles in different multiphase
systems (solid/gas, solid/liquid, crystallisation, liquid/liquid, gas/liquid) and even biological
and geological systems [4]. For liquid/liquid systems a wide range of submodels is available
[5-7] to describe the source and sink terms of droplet breakage and coalescence. Each of these
submodels can be implemented independently into the population balance equations, which
allows a specific assembly of proper submodels for each application. The breakage rate
consists of the submodels for the breakage frequency and the daughter drop size distribution,
whereas the coalescence rate is divided into the collision frequency, which describes how
often droplets collide with each other in the system, and the coalescence efficiency, which
designates the probability of two droplets to confluence if brought into contact.
Existing submodels for the coalescence rate depend on the physical phase properties
and various process parameters. These influencing parameters were implemented in existing
models with different proportionalities and in some cases even contradictorily [7]. With single
drop experiments the coalescence process can be investigated independently from drop
breakage. For this reason, the results can be used to evaluate the models for the coalescence
rate directly for the first time.
FROM SINGLE DROPS TO LARGE SCALE SIMULATIONS
By modelling the whole population with submodels describing the microscopic interaction
between single entities, the population balance equation offers a bottom-up approach from
small scale detailed experiments to the simulation and design of technical applications: the
different submodels can be validated and fitted by lab scale experiments (single drop
investigations and/or shaken bottle test) using only small amounts of the real process liquids
and then be used for the simulation of the whole droplet swarm [8]. This approach is depicted
schematically in Figure 1 and allows skipping expensive pilot plant experiments with a
significant amount of (possibly toxic, hazardous and/or valuable) original process
components.
121
Figure 1: Scheme for systematic analysis of liquid/liquid dispersions for large scale simulations
In the past years several investigation methods were developed at our institute to perform
fundamental single drop experiments. As the observation of breakage and coalescence events
needs a high spatial (order of magnitude: millimetres and smaller) and temporal resolution
(order of magnitude: milliseconds and below), the realisation of experiments is not a trivial
task. Maaß et al. [9] developed a breakage channel to analyse the break-up of droplets at the
stirrer blade of a Rushton turbine. From these findings a detailed breakage time and daughter
drop size distribution were determined and used to develop breakage rate submodels
successfully [10; 11]. Concerning the coalescence investigations, Kamp and Kraume [12]
built a test cell in which a rising droplet collides with a pendant one, recorded by high speed
imaging (see Figure 2). This set-up is a result of the compromise between good observability
due to the locally determined droplet collision and the dynamic collision process of droplets
in free flow. Even though, the microfluidic conditions in the test cell are not identical with the
ones of a collision of two freely moving drops, the acting forces can be assumed to be
mechanically similar at equivalent relative velocities. To conduct serial examinations under
different system conditions (varying e.g. drop sizes, ion species and concentrations in
continuous phase), the drop generation, detachment and the triggered recording of the drop
collision event was fully automated.
Population Balance Equation
Coalescence rate
Single drop experiments
Breakage channel Coalescence cell
Batch settling tests
Collision frequency
Coalescence efficiency
Breakage rate
Breakage frequency
Daughter size distribution
Prediction of drop size distribution
Experimental validation
Lab scale: stirred tank
Plant designProcess control
122
Figure 2: Coalescence of two droplets recorded with high speed imaging (150.000 frames/s)
As the coalescence and contact time of drops, which affect the coalescence probability [13],
have a broad variation, a significant number of experiments have to be conducted to provide a
statistically solid data base. Analysing the recorded images of the coalescence process,
important quantities like coalescence probability, contact and coalescence time, relative
velocity, momentum and deformation can be evaluated. Villwock et al. [14] developed this
test cell further and carried out systematic experiments at two different research laboratories.
The main task was to ensure the comparability and reproducibility of the results. Besides the
continuous control of the used chemicals, the prevention of contaminations of the setup was
mandatory. Impurities (e.g. surfactants) have a strong impact on coalescence since they
change the mobility of the drop surface [15]. The determination of terminal drop rise
velocities was found a reliable standard parameter to evaluate the purity of the system. The
systematic single drop experiments included the investigation of the influence of drop size,
ions in varying concentrations and the pH value, respectively on the coalescence probability
in the system toluene/water with toluene as dispersed and water as continuous phase.
In addition to single drop experiments, Henschke et al. [16] stated that a coalescence
efficiency parameter can be determined from sedimentation curves in batch settling tests.
Therefore, Villwock et al. [14] also conducted systematic standardized batch settling
experiments to identify the coalescence behaviour of different system compositions and to
validate the feasibility as a screening system.
To apply and validate the gained population balance equations, lab scale experiments in
stirred tanks were performed, as large scale applications are not always available. Maaß and
Kraume [11] successfully determined breakage rates from single drop experiments and
showed the application in stirred tanks varying the geometry, dispersed phase and phase
123
fraction [17-19]. For the coalescence in turbulent systems Kamp et al. [20] developed an
eligible model considering droplet charge, which is discussed in the following.
RECENT COALESCENCE INVESTIGATIONS
With the developed single drop test cell a large number of experiments with different system
parameters could be carried out. One main aspect was the investigation of the influence of the
pH value on coalescence probability. Therefore, NaOH in concentrations from 10-5
to 10-1
mol/L (pH 9 - 13) in the continuous phase were prepared. As it is shown in Figure 3, the
coalescence probability strongly decreases with increasing NaOH concentration or pH value
respectively. Since the OH- ions have a strong affinity to adsorb at the drop surface, they
induce a surface potential [21] which causes an electrostatic repulsion. Drop sizes from
1.5 mm to 3.0 mm were investigated. The influence of the drop size on coalescence
probability appears to be dominated by the electrostatic effects. Only at the transition to a
coalescence inhibition due to electrostatic repulsion (here pH 9), where the repulsive and
attractive forces are in the same order of magnitude, the influence of the drop size is
noticeable. Hence, the high standard deviation at pH 9 (30%) is the result of the influence of
the drop size ratio.
Figure 3: Left: coalescence probability depending on pH value from single drop experiments. Right:
transient Sauter mean diameter at different pH comparing experiments and simulations in a stirred tank
DN 150 (from [22])
With the standardised batch settling tests it was possible to achieve comparable results in two
different laboratories. Within the analysis several influencing parameters (e.g. CO2
124
absorption, energy input) were investigated. The results and those of preliminary tests
regarding the influence of ions on the separation behaviour of the toluene/water system were
in good agreement with the literature (e.g. [21; 23-24]). With these findings it was possible to
predict the coalescence behaviour qualitatively for the single drop experiments even though a
universal coalescence parameter could not be derived so far. Furthermore, experiments with
different organic solvents showed that settling tests are also feasible as a screening test.
A coalescence inhibition at high pH values caused by electrostatic interactions was also found
in experimental investigations in a stirred tank (toluene/water system) [25-27]. Two models
from literature [13; 28] were used to simulate the transient drop size distribution. Kamp et al.
[20] showed, that the coalescence inhibition could not be predicted with constant numerical
parameters and thus not satisfactorily by these models. Based on this finding a new
mechanistic model was developed, which implements the DLVO theory [29; 30] into the PBE
framework. The model divides the coalescence efficiency into a hydrodynamic and an
electrostatic part, which are independent from each other. Thus, the hydrodynamic part can be
described by existent models in literature. The new electrostatic part relates the ratio of the
repulsing force (due to the overlap of the electrical double layers of two colliding drops) to
the attractive van der Waals force between them. Using this electrostatic extension of the PBE
the hindered coalescence at high pH values can be described with constant numerical
parameters and the transient and steady-state Sauter mean diameters of the drop size
distribution fit well with the experimental values, as can be seen in Figure 3 exemplarily for
pH 7 and 13 at a stirrer frequency n = 550 min-1
.
Conclusion
The presented results show that a systematic analysis leads to a better and profound
understanding of the microscopic and macroscopic characteristics of coalescence processes.
The bottom-up approach with the combination of single drop experiments, standardised batch
settling tests and investigations in stirred tanks is a challenging but promising way to
successfully describe liquid/liquid dispersions. It is possible to achieve comparable results in
different research laboratories by developing universal strategies and standards.
However, further optimisation of experimental procedures and setups, the evaluation of
existing and development of new submodels for the population balance framework are still
part of our current research and future work.
ACKNOWLEDGMENT
Partially funded by DFG project KR 1639/19-1 “Coalescence efficiency in binary systems”.
125
REFERENCES
[1] HULBURT, H. & KATZ, S.: Some problems in particle technology: A statistical mechanical
formulation. Chem. Eng. Sci. 19 (1964), 8, 555-574
[2] RANDOLPH, A. D.: Effect of Crystal Breakage on Crystal Size Distribution in Mixed
Suspension Crystallizer. Ind. Eng. Chem. Fund. 8 (1969), 1, 58-63
[3] RAMKRISHNA, D.: The Status of Population Balances. Rev. Chem. Eng. 3 (1985), 1, 49-95
[4] RAMKRISHNA, D.: Population Balances: Theory and Applications to Particulate Systems in
Engineering, San Diego: Academic Press, 2000
[5] LIAO, Y. & LUCAS, D.: A literature review of theoretical models for drop and bubble breakup
in turbulent dispersions. Chem. Eng. Sci. 64 (2009), 15, 3389-3406
[6] LIAO, Y. & LUCAS, D.: A literature review on mechanisms and models for the coalescence
process of fluid particles. Chem. Eng. Sci. 65 (2010), 10, 2851-2864
[7] KOPRIWA, N.; BUCHBENDER, F.; AYESTERÁN, J.; KALEM, M. & PFENNIG, A.: A Critical
Review of the Application of Drop-Population Balances for the Design of Solvent Extraction
Columns: I. Concept of Solving Drop-Population Balances and Modelling Breakage and
Coalescence. Solvent Extr. Ion Exch. 30 (2012), 7, 683-723
[8] BART, H.-J.; GARTHE, D.; GRÖMPING, T.; PFENNIG, A.; SCHMIDT, & STICHLMAIR, J.: Vom
Einzeltropfen zur Extraktionskolonne. Chem. Ing. Tech. 78 (2006), 5, 543-547
[9] MAAß, S.; GÄBLER, A.; ZACCONE, A.; PASCHEDAG, A. & KRAUME, M.: Experimental
Investigations and Modelling of Breakage Phenomena in Stirred Liquid/Liquid Systems.
Chem. Eng. Res. Des. 85 (2007), 5, 703-709
[10] MAAß, S.; WOLLNY,; SPERLING, R. & KRAUME, M.: Numerical and experimental analysis of
particle strain and breakage in turbulent dispersions. Chem. Eng. Res. Des. 87 (2009), 4, 565-
572
[11] MAAß, S. & KRAUME, M.: Determination of breakage rates with single drop experiments.
Chem. Eng. Sci. 70 (2012), 146-164
[12] KAMP, J. & KRAUME, M.: Influence of drop size and superimposed mass transfer on
coalescence in liquid/liquid dispersions - Test cell design for single drop investigations. Chem.
Eng. Res. Des. 92 (2014), 4, 635-643
[13] COULALOGLOU, C. A. & TAVLARIDES, L. L.: Description of interaction processes in agitated
liquid-liquid dispersions. Chem. Eng. Sci. 32 (1977), 11, 1289-97
[14] VILLWOCK, J.; GEBAUER, F.; KAMP, J.; BART, H.-J. & KRAUME, M.: Systematic analysis of
single droplet coalescence. Chem. Eng. Technol. 37 (2014), 7, 1103-1111
[15] WEGENER, M.; PAUL, N. & KRAUME, M.: Fluid dynamics and mass transfer at single droplets
in liquid/liquid systems. Int. J. Heat Mass Transfer 71 (2014), 475-495
[16] HENSCHKE, M.; SCHLIEPER, L. H. & PFENNIG, A.: Determination of a coalescence parameter
from batch-settling experiments. Chem. Eng. J. 85 (2002), 2-3, 369-378
[17] MAAß, S.; METZ, F.; REHM, T. & KRAUME, M.: Prediction of drop sizes for liquid-liquid
126
systems in stirred slim reactors - Part I: Single stage impellers. Chem. Eng. J. 162 (2010), 792-
801
[18] MAAß, S.; REHM, T. & KRAUME, M.: Prediction of drop sizes for liquid/liquid systems in
stirred slim reactors - Part II: Multi stage impellers, Chem. Eng. J. 168 (2011), 827-838
[19] MAAß, S.; PAUL, N. & KRAUME, M.: Influence of the dispersed phase fraction on experimental
and predicted drop size distributions in breakage dominated stirred systems. Chem. Eng. Sci.
76 (2012), 140-153
[20] KAMP, J.; NACHTIGALL, S.; MAAß, S. & KRAUME, M.: Modelling of coalescence in turbulent
liquid/liquid dispersions considering droplet charge. Czasopismo Techniczne Mechanika 109
(2012), 5, 113-124
[21] MARINOVA, K. G.; ALARGOVA, R. G.; DENKOV, N. D.; VELEV, O. D.; PETSEV, D. N.; IVANOV,
I. B. & BORWANKAR, R. P.: Charging of Oil-Water Interfaces Due to Spontaneous Adsorption
of Hydroxyl Ions. Langmuir 12 (1996), 8, 2045-2051
[22] KAMP, J.; NACHTIGALL, S.; MAAß, S. & KRAUME, M.: Modelling of coalescence in turbulent
liquid/liquid dispersions considering droplet charge. 19th International Conference Process
Engineering and Chemical Plant Design (2012), Cracow, 25-27th September
[23] PFENNIG, A. & SCHWERIN, A.: Influence of Electrolytes on Liquid-Liquid Extraction. Ind.
Eng. Chem. Res. 37 (1998), 8, 3180-3188
[24] KUMAR, M. K.; MITRA, T. & GHOSH, P.: Adsorption of Ionic Surfactants at Liquid-Liquid
Interfaces in the Presence of Salt: Application in Binary Coalescence of Drops. Ind. Eng.
Chem. Res. 45 (2006), 21, 7135-7143
[25] TOBIN, T. & RAMKRISHNA, D.: Coalescence of charged droplets in agitated liquid-liquid
dispersions. AIChE J. 38 (1992), 8, 1199-1205
[26] GÄBLER, A.; WEGENER, M.; PASCHEDAG, A. & KRAUME, M.: The effect of pH on
experimental and simulation results of transient drop size distributions in stirred liquid-liquid
dispersions. Chem. Eng. Sci. 61 (2006), 9, 3018-3024
[27] KRAUME, M.; GÄBLER, A. & SCHULZE, K.: Influence of physical properties on drop size
distributions of stirred liquid-liquid dispersions. Chem. Eng. Technol. 27 (2004), 3, 330-334
[28] TOBIN, T. & RAMKRISHNA, D.: Modeling the effect of drop charge on coalescence in turbulent
liquid-liquid dispersion. Can. J. Chem. Eng. 77 (1999), 6, 1090-1104
[29] DERJAGUIN, B. & LANDAU, E.: Theory of the stability of strongly charged lyophobic sols and
of the adhesion of strongly charged particles in solutions of electrolytes, Acta Physicochitnica
U.R.S.S. 14 (1941), 633-662
[30] VERWEY, E. J. W. & OVERBEEK, J. T. G.: Theory of the Stability of Lyophobic Colloids, New
York: Elsevier, 1948
127
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
INFLUENCE OF POLYMER-SURFACTANT ADDITIVES ON PRESSURE DROPS IN
PIPE FLOW
Z. MATRAS; B. KOPICZAK
Cracow University of Technology, Institute of Thermal and Process
Engineering, Al. Jana Pawla II 37, 31-864 Cracow, Poland;
Abstract. The aim of this paper was to present the possibility to enhance the drag
reduction effect in pipe flow by simultaneous addition of high molecular polymers
and surfactants to the transported liquid medium.
The mechanism of drag reduction by polymeric and micellar aggregates was
presented. Qualitative analysis of polymer and micellar additives influence on the
shape and character of flow resistance curves was performed. Multicomponent
polymer-micellar solution flow resistance curves were compared with the
appropriate single additive polymer or surfactant solution flow resistance curves.
The results indicate that the presence of polymer-micellar aggregates causes flow
laminarization in the initial phase of turbulent flow, leading to the extension of the
stable transition zone. Simultaneous addition of surfactant with salt and a high
molecular polymer to the solvent significantly reduces mechanical degradation of
the internal structure of polymer-micellar solution.
Keywords. Drag reduction, flow resistance, internal friction, pipe flow, polymer,
surfactant, aggregate.
INTRODUCTION
The discovery of the abnormal drag reduction phenomenon in pipe flow by adding
trace amounts of high molecular weight polymers or surfactants to the solvent was an
inspiring starting point in researches of hydraulic transport possibilities with simultaneous
significant cost reduction. Within the framework of the scientific discipline of technical
rheology, a large-scale study was performed, whose aim was either reducing the cost of the
liquid flow in the pipeline or increasing the flow efficiency/flow rate without the necessity of
increasing the power demand. It is commonly known that drag reduction effect is caused by
128
high molecular weight polymers, surfactants, fine solid particles and fibrous structure
additives [1-7].
In scientific research as well as in engineering applications, the most commonly used
drag reduction agents are high molecular weight polymers and surfactants. Adding even a
small amount of these substances to the transported solvent induces significant and extremely
advantageous, from the energetic point of view, reductions of pressure losses caused by fluid
internal friction in the turbulent range of flow. This results in a significant increase in the flow
rate without the necessity of increasing the power demand, or vice versa – to reduce the power
demand while maintaining a constant flow rate. Therefore, it provides large potential
possibilities for application of this effect in different industry branches, particularly in oil
industry [3, 8], or in heating [9], fire-fighting [10], transport of slurries, sludge and brines
[2, 11], while increasing the efficiency of sewerage and anti-flood systems in time of heavy
rainfall [12]. The drag reduction effect caused by polymers was utilized in designing an
industrial installation for crude oil transportation in Norway, USA and India [2, 3, 8].
The new internal structure, which occurs when special additives are introduced into
the solution, was perceived to be the cause of the drag reduction effect. The addition of
surfactants agents into the solvent results in micelles structure formation. In case of
application of a high molecular weight polymer as a drag reducing additive, formation of
macromolecules is observed.
At no motion condition, when the fluid is at rest, the above mentioned structures are
chaotic. Only during fluid flow shearing, both macromolecules and micelles start to arrange in
a characteristic orientation, in accordance with the principle of minimum resistance. Solution
concentrations are very small. In case of high molecular polymers it is order of magnitude of
several or several dozen of ppm, and for surfactants it is order of tenths of mM/dm3.
Furthermore, in order to obtain more favourable conditions for the formation of micelles,
small amounts of electrolytes are added into the solution. Salts (e.g. sodium salicylate or
sodium bromide) or alcohols (e.g. α – naphthol) are the typically used electrolytes.
The effect of drag reduction of flow by high molecular polymers or surfactants has
also some limitations resulting from the properties of additives. The main disadvantage of
high molecular polymers is their susceptibility to mechanical and thermal degradation. In case
of surfactant solutions it is a certain critical Reynolds number. When exceeded, micelles
orientation is lost, and the drag reduction effect collapses. The number depends not only on
the pipe's diameter, but also on the solution concentration.
An important distinguishing factor for the phenomenon of drag reduction caused by
surfactant additives comparing to drag reduction effect induced by high molecular polymers is
129
the reversibility of solution degradation [1, 13]. In case where the value of Re number falls
below critical Re value, the drag reduction effect occurs again. Moreover, the drag reduction
effect in surfactant solution is not weakened in time of long-term liquid pumping comparing
to the polymer solution. This is a great advantage for the transport of liquids over long
distances.
INFLUENCE OF ANALYSED ADDITIVES INTERNAL STRUCTURE ON
PIPE FLOW RESISTANCE
Chain structure of polymer macromolecules
Adding a small amount of a polymer to the solvent (usually water) causes a change of the
internal structure of the solution creating so called macromolecules. Due to the characteristic
chain structure of polymers in solutions, macromolecules form different conformations. They
can take shape of randomly coiled spherical structures or stretched elastic threads.
Macromolecules in a solution of low concentration (c<0,1%) form spherical polymer coils -
fig.1(a). Their influence on the rheological properties of the solution is negligible.
Rheological changes are noticed only when the polymer macromolecules take the form of
elastic fibres – Fig.1(b).
Figure 1. Polymer macromolecules: a) in shape of coiled spherical balls, b) in shape of elastic threads
This effect occurs upon increasing the polymer concentration in the solution. Such a solution
exhibits non-Newtonian and viscoelastic characteristics. Changes in its viscosity are also
observable. By the application of slight shear stress to the solution induced by simple
solution shearing, the polymer threads are stretched and start to arrange in a characteristic
orientation, in accordance with the principle of minimum resistance – Fig.2.
Figure 2. Macromolecules of polymer elongated under the shear stress
a) b)
130
Molecules with such conformation contribute to velocity fluctuation suppression in turbulent
range of flow causing its laminarization and reducing fluid flow pressure losses at the same
time. Having exceeded a certain critical value of shear stress at pipe wall the polymer chains
are ripped apart and an irreversible degradation of polymer macromolecule structure is
observed [5,7,14]. It should also be noted that the polymer macromolecules are also
susceptible to irreversible thermal degradation for solution temperatures exceeding 40°C [1].
Due to the mechanical and thermal degradation of the polymer solution, the drag reduction
effect gradually vanishes.
The effect of abnormally reduced friction losses in turbulent flow can appear in different
ranges of the Reynolds number, while the shape and location of friction losses curve depends
not only on the Reynolds number definition, but most of all on the pipe diameter value,
solution concentration, type of the drag reducer applied, its molecular structure and
conformation, intermolecular bonds of the applied chemical additive, its molecular weight,
rate of solution degradation, etc.
The source of the Toms effect is lying in different rheological properties of Toms liquid
comparing to the properties of purely viscous liquid. The presence of elastic polymer
macromolecules in the solvent leads to a locally distorted averaged linear velocity profile and
a deformation of power of the low empirical Ostwald - de Waele formula [7]. It allows to
assume the real flow of the polymer solution to the flow of the so-called pseudo purely
viscous liquid, defined as a purely viscous suspension of stiff molecules with dimensions of
nondeformed polymer macromolecules. A new hypothesis [9] was formulated as a
consequence of comparing the distorted Toms liquid velocity profiles and the corresponding
velocity profiles of the pseudo purely viscous liquid. According to this hypothesis, during the
flow of the polymer solution an additional negative shear stress is observed. It is induced by
the polymer's elastic properties and the reduction of power law consistency constant value
associated with this effect. The hypothesis of additional shear stress explains all effects
accompanying the Toms phenomenon.
Micellar structure of surfactants
The drag reduction effect caused by small amounts of surfactant addition is particularly
interesting from a cognitive point of view, as well as its utilitarian character. Unlike the Toms
effect, it is not susceptible to mechanical degradation [1, 2, 13]. Surfactants added to the
solvent (usually water) create a low concentrated solution. When the surfactant concentration
exceeds a certain critical level of concentration, individual surfactant molecules start to
organize into new structures called micelles. This concentration of structure transition is
131
called first critical micelle concentration CMC. The so-formed single micelle consists of
several dozen to several thousand individual surfactant molecules. The drag reduction effect
in the flow of such a micellar solution takes place only when cylindrical threadlike/rodlike
micelles are present in the solution – fig. 3
Figure 3. Formation of spherical and cylindrical/threadlike micelle structures in aqueous solution of
surfactant
The length of rodlike micelles is in the range of 25÷200[nm], and their diameter is in the
range of 2÷5[nm]. The time of micelles formation varies from milliseconds to hours.
Threadlike micelles are formed when the surfactant concentration exceeds second critical
micelle concentration CMC2. Both CMC and CMC2 values depend on the type of surfactant
and temperature – fig.4.
Figure 4. Critical concentration CMC and CMC2 of aqueous solution of surfactant in temperature
function [15]
Reformation of spherical micelles into threadlike micelles can be induced by adding a small
amount of electrolytes (e.g. salts or alcohols) [15, 16].
Under the shear stress caused by the solution flow, the rodlike micelles start to arrange in a
characteristic orientation which produces the drag reduction effect. According to [17] the drag
reduction effect is caused by the viscosity anisotropy in the boundary layer. On the other
CMC2CMC
0
1
2
3
4
5
6
7
8
9
10
0 10 20 30 40 50 60 70 80 90 100
c [m
M/d
m3]
T [˚C]
HTASal
ODASal
CMC2
CMC
132
hand, in [13-15] the friction losses reduction effect is explained by the characteristic
orientation of threadlike micelles in the flow direction.
Under the shear stress micelles are untangled and oriented toward the flow direction. It
impedes the development and relocation of vortices in the direction perpendicular to the flow
direction, thereby decreasing the amplitude and frequency of turbulent fluctuation. Hence
turbulence intensity and turbulent energy dissipation are reduced.
Beyond a certain critical value of the shear stress on the wall the drag reduction effect
disappears. Micelles lose orientation in the flow direction. Comparing to a polymer solution, a
micellar solution is insusceptible to mechanical and thermal degradation [1].
According to the functional relationship of the friction coefficient derived in [13,14], after
exceeding a certain critical Reynolds number value a sudden increase in the flow resistance is
observed. This critical value of Re number also defines the flow conditions at which the
maximal drag reduction is obtained. The presented concept assumes that the drag reduction
effect appears only in presence of threadlike micelles. In the laminar flow micelles arrange in
the orientation parallel to the main flow direction. The flow resistance in this range of flow
can be assigned using the same equations, which are valid for the laminar flow of non-
Newtonian purely viscous liquid. Exceeding the first critical Reynolds number causes
appearance of the laminar flow disturbances which are suppressed by threadlike micelles
oriented in the flow direction.
The flow resistance in this range of flow is only slightly larger than the flow resistance in the
laminar range and significantly smaller comparing to the flow resistance in the turbulent zone.
Exceeding the second critical Reynolds number leads to the loss of rodlike micelles specific
orientation. Micelles then coil into a globular shape structure. This causes a sudden increase
in the flow resistance and the disappearance of the drag reduction effect.
EXPERIMENTAL ANALYSES OF FLOW RESISTANCE CURVES
Measurements were performed using modern capillary-pipe rheometer, designed and
constructed in the Division of Fluid Mechanics laboratory at the Cracow University of
Technology [18]. After the preliminary study, the following drag reducers were used for
experimental analysis: the non-ionic polymer with high molecular weight Mv=8·106 –
poly(ethylene oxide) – [CH2 CH2 O]n (PEO) and the cationic surfactant – cetyltrimetyl
ammonium bromide – [CH3(CH2)i5N(CH3)3]+Br~ (CTAB).
In order to lower the CMC value, salt sodium salicylate C7H5NaO3 (NaSal) was used.
Distilled water was used as the solvent.
133
After the application of the appropriate additives to the solvent, solutions were mixed gently
so as not to cause mechanical degradation of polymer chains. Mixtures were left at rest for 24
hours before measurements. Adiabatic steady flow of homogenous solutions were examined
in 8 different straight pipes with diameters between 1.8[mm] and 21[mm], in temperature
equal to 25˚C.
In order to perform a better interpretation of the effect of the simultaneous addition of the
polymer, the surfactant and salt, comparing to the adequate addition of pure polymer or pure
surfactant, measurement data was presented in the modified system of “pseudorheostable”
numbers [ReM, cfM], and described by formulas (1) and (2) [7]:
2,5
2m
fM13n
1n2
2
ρu
4L
pD
c (1)
2,5
1n
n
n2
m
n
13n
1n2
84n
13nK
ρuD
MRe
(2)
where: D is pipe diameter, r is fluid density and n and K are respectively flow index and
consistency constant of power-law fluid model.
The flow resistance curves of rheostable (purely viscous) non-Newtonian fluids in such
defined dimensionless numbers system are boiled down to a single curve – in the whole range
of the modified Reynolds number (2) – identical to the classical Newtonian curve described in
the laminar range by Fanning equation and in the turbulent flow by Blasius formula. The
choice of such a coordinate system was dictated additionally by the fact that it facilitates the
identification and description of the characteristic drag reduction flow zones. In this modified
system of pseudorheostable numbers [ReM, cfM] each deviation of the experimental flow
resistance curve which indicates abnormal drag reduction from pseudorheostable Blasius
curve, allows the identification of specific additives' influence (polymers or/and surfactants
with salt) on a range of analyses of the drag reduction effect.
Representative flow resistance curves of the analysed polymer-micellar solution and the
corresponding purely polymer solution and purely micellar solution are presented in fig.5.
134
Figure 5. The flow resistance curves of polymer, surfactant and polymer-surfactant water solutions,
defined in the system of cardinal numbers (1) and (2)
The analysis of presented flow resistance curves indicates that in the laminar range of flow
simultaneous addition of polymer and surfactant to the solvent causes inconsiderable increase
of flow resistance and significant extension of the stable transitional zone of flow. The
existence of the third significantly extended reduction zone is also observed, in which the
viscoelastic properties of the solution are the dominant factor. It is particularly well illustrated
by fig.6.
Figure 6. Comparison of the flow resistance curves of polymer-surfactant water solutions for different
polymer concentration
135
In comparison to a purely micellar solution, the collapse of drag reduction is normally
observed in this region. It should also be noted that addition to the micellar solution of even
small amounts of high molecular weight polymer (about 10[ppm]) causes a reduction of the
non-Newtonian properties of the solution.
The results of drag reduction measurements analysis indicate the effect of pipe diameter
influence on drag reduction efficiency. Increasing the pipe diameter d results in clear
extension of the stable transitional zone towards higher values of the Reynolds number.
Moreover, decreasing the pipe diameter value d results in an increase of the drag reduction
effect in the third additional turbulent range of flow – fig.7.
Figure 7. Pipe diameter effect on the flow resistance curves of polymer-surfactant water solutions
HYPOTETIC MECHANIZM OF DRAG REDUCTION CAUSED BY
POLYMER-MICELLAR ADDITIVES
Simultaneous introduction of small amounts of polymer and surfactant additives to the solvent
provides initiation of micellarization process at much lower concentration, comparing CMC.
This concentration is called the critical aggregation concentration (CAC) [13, 19].
In case of an ionic surfactant mixed with a counter charged polyelectrolyte only a small part
of the polyelectrolyte is adsorbed by the surfaces of the micelles. Furthermore, CAC has a
lower order of magnitude than the original CMC due to the following facts:
electrostatic interaction occurs between the electrolyte and the surface of the micelles,
no concentration of surfactant counter-ions on the micelles surface appears; their place
is taken by the polyelectrolyte,
136
highly charged polyelectrolytes can trigger a certain amount of concentrated counter-
ions, as they combine with the micelles.
In the initial state, there is a long highly charged polyelectrolyte chain with a great number of
concentrated counter-ions. The final state of the mixture has single threadlike micelles with a
part of polymer macromolecule chain coiled around rigid micelles – fig. 8(a). According to
[13, 19, 20], these molecules form the so-called aggregates.
Such newly created polymer-micellar solution can be characterized by a lower susceptibility
to mechanical degradation during flow or its degradation can be almost invisible.
Schematically illustrated aggregates (fig. 8(a)) subjected to the shear stress take orientations
consistent with the aforementioned principle of minimum resistance – fig. 8(b). With
increasing value of the Reynolds number, internal friction forces stretch and extend the
aggregates leading to the laminarization of the initial phase of the turbulent flow.
Therefore, it may be hypothesized that the rigid rodlike micelles, which create the core of the
aggregates, are responsible for reducing the flow resistance in the extended transitional zone
between the laminar and turbulent flow.
Figure 8. Polymer-micellar aggregate: a) in shape of spherical balls, b) during shear stress action
The aggregates and micelles are responsible for transmission of internal friction in the liquid.
The value of the critical Reynolds number for which the transition to the turbulent zone is
observed is greater for polymer-micellar solutions. This means that the stable transition zone
is extended. The reason for such behaviour can be the partial disintegration of aggregates to
original forms, i.e. micelles (formed from the surfactant) and macromolecules (formed from
the polymer) due to a significant increase of the shear rate. From this moment, both micelles
and macromolecules interact separately on the transported solution causing a further drag
reduction effect. Passing further in the turbulent range of flow micelles lose their orientation
and no longer have a major impact on the drag reduction. A key role is played in this zone by
the polymer. Not having undergone an earlier degradation, the polymer macromolecules still
cause the low reduction.
In drag reduction caused by the use of polymer-surfactant solution, one cannot talk about the
so-called collapse of the drag reduction. It occurs permanently over a wide range of Reynolds
numbers. In the turbulent zone polymer macromolecules undergo a certain mechanical
b)a)
137
degradation. Decreasing the shear rate leads to the reconstruction of the solution's internal
structure. As a result of electrostatic interaction, the recreated micelles are combined with the
polymer chains by coiling around them. These chains are much shorter and such newly
created aggregates do not have the same rheological properties as the original ones. This
results in a slight increase of the flow resistance in comparison with a freshly prepared
solution.
CONCLUSIONS
The experimental analysis of the influence of simultaneous addition of polymer and surfactant
additives to a water solvent indicates that in the laminar range of flow the simultaneous
addition of a polymer and a surfactant to the solvent causes an inconsiderable increase of the
flow resistance comparing to the pure solvent's flow resistance. A significant extension of the
stable transitional zone of flow has been observed, however. The analysis of experimental
flow resistance curves (fig.5–fig.7) allows to observe the existence of the third significantly
extended reduction zone in the turbulent range of flow. In this zone the viscoelastic properties
of the solution are the dominant factor. It is found that in this zone the drag reduction is
influenced by the polymer concentration in the solvent. The higher concentration of polymer
in the solution is applied, the greater drag reduction in the flow is obtained.
The results of drag reduction measurements analysis indicate the effect of the pipe diameter
change on the drag reduction. Increasing the pipe diameter d results in clear extension of the
stable transitional zone towards higher values of the Reynolds number, while decreasing the
pipe diameter value induces an increase of the drag reduction effect in the third additional
turbulent range of flow.
Simultaneous addition of a polymer and a surfactant with salt significantly reduces the
mechanical degradation of the polymer-micellar solution's internal structure.
The comparison analysis indicates that polymer-micellar solutions combine and intensify the
positive features of their pure polymer and micellar equivalents. Moreover, they provide
efficient reduction of flow resistance in a wider range of the Reynolds number.
BIBLIOGRAPHY
[1] Wang Y., Yu B., Zakin J. L., Shi H.: Review on Drag reduction and Its Heat Transfer
by Additives, Advances in Mechanical Engineering, no. 10, (2011), 17pages
[2] Zakin J. L., Ge W.: Polymer and Surfactant Drag Reduction in Turbulent Flows,
Polymer Physics: From Suspensions to Nanocomposites and Beyond, John Wiley &
Sons, Inc., Hoboken, NJ, USA (2010) 89-127.
138
[3] White C.M.: Mungal M.G., Mechanics and Predictions of Turbulent Drag Reduction
with Polymer Additives, Annular Review of Fluid Mechanics, no. 40 (2008) 235-256.
[4] Borostow W.: Drag reduction in flow: Review of applications, mechanism and
prediction, Jurnal of Industrial and Engineering Chemistry, vol.14 (2008) 409-416.
[5] Shu-Peng C.: Drag reduction of a cationic surfactant solution and its shear stress
relaxation, Jurnal of Hydrodynamics, 24(2) (2012) 202-206
[6] Tamano S., Ito M., Kato K., Yokota K.: Turbulent drag reduction in nonionic
surfactant solutions, Physics of Fluids, 22(5), 055102 (2010)
[7] Matras Z., Przepływ cieczu Tomsa w przewodach kołowych, Politechnika Krakowska,
Monografia 29, 1984.
[8] Dujmovich T., Gallegos A.: Drag reducers improve throughput, cut costs, Offshore,
vol. 65, no. 12 (2005) 1–4.
[9] Motier J.F., Chou L. C., Kommareddi N.S.: Commercial drag reduction: past, present,
and future, Proceedings of the ASME Fluids Engineering Division Summer Meeting,
San Diego, Calif, USA (1996).
[10] Figueredo R. C. R., Sabadini E.: Firefighting foam stability: effect of the drag reducer
poly(ethylene) oxide, Colloids and Surfaces A, vol. 215, no. 1–3 (2003) 77–86.
[11] Golda J.: Hydraulic transport of coal in pipes with drag reducing additives, Chemical
Engineering Communications, vol. 43, no. 1–3 (1986) 53–67.
[12] Dembek G., Bewersdorff H.W.: "Short-time increase of sewer capacity by addition of
water-soluble polymers," GWF, Wasser/Abwasser, vol. 122, no. 9, 1981, pp. 392–395.
[13] Matras Z., Malcher T., Gzyl-Malcher B.: The influence of polymer-surfactant
aggregates on drag reduction, Thin Solids Films, vol. 516 (2008) 8848–8851.
[14] Matras Z., J. Głód, T. Malcher, The influence of surfactant additives on friction losses
in the turbulent flow, Applied Mechanics and Engineering, vol. 4, 441, 1999, 441-446.
[15] Bębenek B., Reologia roztworów kationowych środków powierzchniowo czynnych:
Część I. Stan wywołany ścinaniem, Chem. and Proc. Engineering, 2, No 179, (1994)
[16] Knop A. Colby R.: Role of Condensed Counterions in the Thermodynamics of
Surfactant Micelle Formation with and without Oppositely Charged Polyelectrolytes.
The ACS Journal of Surfaces and Colloids, Volume 15, (1) (1999) 58-65.
[17] Myska J., Anisotropy of Viscosity of Drag Reducing Solution. AIChE Journal, 44, No
6. 1467 (1998)
[18] Matras Z., Walczak S.: Reometr kapilarno-rurowy, Czasopismo Techniczne z.5-M
(2003) 359-370.
139
[19] Hou Z., Li Z., Wang H.: Interaction between poly(ethylene oxide) and sodium dodecyl
sulfonate as studied by surface tension, conductivity, viscosity, electron spin reso-
nance and nuclear magnetic resonance, Colloid Polym. Sci., 277 (1999) 1011-1018
[20] Jung T. K., Chul A. K., Ke Z., Chun H. J., Hyoung J. C., Effect of polymer–surfactant
interaction on its turbulent drag reduction, Colloids and Surfaces A: Physicochemical
and Engineering Aspects, ISSN 0927-7757, Vol. 391, Issues 1–3, 5, (2011), 125-129
141
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
INFLUENCE OF A NON-IONIC SURFACTANT TRITON X-100 ON FLUID DYNAMICS
AND MASS TRANSFER OF SINGLE RISING DROPS
S.-J. KIM; N. PAUL; M. KRAUME
Technische Universität Berlin, Chair of Chemical & Process
Engineering, Ackerstraße 76, 13355 Berlin
Abstract. Micellar systems can be applied for the enhancement of the reaction
rate of reactions in liquid/liquid systems. As the amphiphilic molecules adsorb at
the interface the transport processes are influenced. This work focuses on a
systematic investigation of a homologous row of long chain alcohols to improve
the understanding of the occurring phenomena. Next to the change of the fluid
dynamics and the additional mass transfer resistance caused by the adsorption
layer of surfactants, the formation of a microemulsion layer must be taken into
consideration; otherwise the mass transfer rate will be overestimated.
Keywords. Surfactants, micellar systems, fluid dynamics, mass transfer
INTRODUCTION
Fast and specific reactions are the main advantages of homogeneous catalyzed
reactions compared to heterogeneous catalyzed reactions. Nevertheless, the biggest challenge
of homogeneous processes is the separation of the catalyst and the product. For the realization
of the separation process liquid/liquid systems can be applied which represent a special type
of homogeneous reaction systems. In most cases these reactions systems are designed in the
way that the catalyst is solved in one of the liquid phases while the reactants and the products
are solved in the other liquid phase. Therefore, a simplified separation process is possible.
Nevertheless, due the additional phase transport processes across the liquid/liquid interface
must be taken into consideration to understand the reaction mechanisms, completely.
142
Furthermore, the reactants need to have certain solubility in the phase where the catalyst is
solved; otherwise the reaction will not proceed.
Using water as a solvent is one of the basic principles of the “Green Chemistry” [1].
Hence, a strong polar solvent must be regarded. Considering an aliphatic reactant (long chain
alkene) the solubility of the reactant in the aqueous phase where the catalyst is solved is low
which results in a slow reaction rate. To improve the reaction rate micellar systems can be
applied [2]. The amphiphilic additives adsorb at the liquid/liquid interface where these
molecules influence the transport processes which influences the yield and the selectivity of
the reaction again.
For a fundamental understanding of the influences caused by surfactants on the
transport processes many researchers observed the occurring phenomena at single droplets
which reduce the complexity; thus this is the smallest transfer. In most cases a reduction of
the mass transfer is reported [3-4]. Two effects are taken into consideration: the
physicochemistry effect and the fluid dynamic effect. The adsorption layer of the surfactant
molecules causes an additional mass transfer resistance which is called the physicochemistry
effect [5]. Furthermore, the formation of the adsorption layer results in a reduction of the
interfacial mobility. Therefore, the droplet behaves like a particle with a rigid interface. Shear
stress which is transported across the liquid/liquid interface is reduced as a result the inner
circulations of fluid particles disappear. This effect causes a reduction of the mass transfer [6].
In systems with high non-ionic surfactant concentrations (micellar systems) these effects are
insufficient to describe the occurring phenomena [7]. Micellar reaction systems often have
much higher surfactant concentrations than the critical micelle concentration (CMC). At these
surfactant concentrations the visco-elastic modulus [7] and the spring constant of the interface
[8] change which is explained by a change in the phase behaviour at the liquid/liquid
interface. A microemulsion layer is created at the interface. This change results in an
additional mass transfer resistance.
This work focuses on the formation of an additional mass transfer resistance caused by
a change in phase behaviour. In micellar systems there are only a few results available in the
literature. Further, these results are not consistent. While in some works a constant mass
transfer is predicted after exceeding certain concentrations other works predict an additional
resistance and in some works an enhancement of the mass transfer rate was found for high
surfactant concentrations. In this work a systematic investigation is started. The influence of
the non-ionic surfactant Triton X-100 on the fluid dynamics and the mass transfer of single
droplets were observed. As dispersed phase a homologous row of long chain alcohols were
observed.
143
EXPERIMENTAL SECTION
The experimental investigations of the fluid dynamics and the liquid/liquid mass transfer at
single droplets were carried out in a special glass column. This setup is explained in detail in
the work of Wegener et al. [9]. Figure 1 gives a schematic flow sheet of the test cell. The test
cell consists out of 1000 mm heigh glass column which is surrounded by an acrylic glass
jacket which is filled with glycerine, due to optical accessibility for the high speed camera.
Furthermore, the jacket which is filled with glycerine offers the ability to temper the
experimental setup. A droplet which is produced at the nozzle (6) by a Hamilton PSD-2
module (4) is released by the solenoid device (5). The droplet will rise freely in the
continuous phase until it is caught by the glass funnel (8). The funnel is adjustable in it’s
height; hence various contact times can be realized. The droplet is pumped out of the system
with another Hamilton PSD-2 module (4) and can be analyzed.
Figure 1: Experimental setup: 1 glass column with acrylic glass jacket; 2 high speed camera; 3 storage
dispersed phase; 4 Hamilton PSD-2 modules; 5 solenoid device; 6 nozzle; 7 illumination, 8 glass funnel, 9
thermostat.
The test systems used in the work consisted of the following components: Water, aqueous
surfactant solutions, respectively were used as the continuous phases. Various long chained
alcohols were used as the dispersed phase. Triton X-100 was applied as a non-ionic
surfactant. Pyridine-2-azo-dimethyl-anline was used as the transferred component. This is an
1
2
3
3
4
4
5
6
7
8
9
144
azo dye. The concentration of the dye was analyzed by applying a Specord 210 photmeter by
Jena Analytik.
RESULTS AND DISCUSSION
The track of the single droplets were captured by the high speed camera and analysed. The
drop velocity and the position of the droplets were the results of this analysis. For the test
system water/1-octanol the drop rise velocity is shown as function of the droplet diameter for
various Triton X-100 concentrations in Figure 2. For spherical droplets which rose in pure
water the drop rise velocity is well described by the correlation of Feng and Michaelides [10].
With an increase of the surfactant concentration the velocity approaches the value of rigid
sphere. This behaviour is well described by the correlation of Martin [11]. Exceeding a
specific diameter (dp > 3 mm) the droplets start to deform the correlation by Thorsen et al.
was derived for low viscosity values; hence it does not agree well with the experimentally
determined velocities. Exceeding the critical micelle concentration (0.2 mmol/L) the
liquid/liquid interface is completely covered with surfactant molecules the fluid particle
behaves like a rigid sphere. Therefore, a constant mass transfer coefficient is assumed with
exceeding the critical micelle concentration. For the test system water/decanol the results are
similar.
Figure 2: Drop rise velocity of 1-octanol droplets as a function of the droplet diameter for various
surfactant concentrations compared with calculated values [13].
0
20
40
60
80
100
120
140
0 2 4 6 8
Droplet diameter dp [mm]
Dro
p r
ise
vel
oci
tyv
[mm
/s]
water/octanol
TX-100
cTX [mmol/L]
00,01
0,11
[11] Martin (1980)[12] Thorsen et al. (1968)
[10]Feng & Michaelides (2001)
145
The mass transfer was determined in the rising test cell shown in Figure 1. Two different test
systems were applied. For a better understanding the dynamic mean concentration of the
transferred component is shown in Figure 3. The concentrations were referred to the
calculated values of the mass transfer for particles with rigid interfaces [14-15]. Furthermore,
the surfactant concentration was varied. The results given in Figure 3 show the two limiting
cases, on the one hand the results for the pure system (mobile interface) and on the other hand
the results for a contaminated system (rigid interface). In pure liquid/liquid systems the
interface is mobile. Due to the effects mentioned above the mass transfer is enhanced
compared to the contaminated system. Hence, the mass transfer for the pure octanol/water
system 1,8 times faster than predicted by assuming a rigid interface. For the other test system
(water/decanol) the ratio between the mass transfer predicted by the calculated values is 1,2.
The viscosity of decanol is slightly higher; hence the mass transfer is slower.
Figure 3: Dynamic Mean concentration for the pure test systems and for high Triton X-100
concentrations; compared with the calculated values for rigid interfaces [14-15].
The Triton X-100 concentration given in Figure 3 is 50 times higher than the critical micelle
concentration [16]. Therefore, a complete coverage of surfactant molecules of the
liquid/liquid interface is assumed; the mass transfer should be calculated with the applied
correlation. Nevertheless, by using the correlation for rigid interfaces the mass transfer rate is
over estimated for both test systems by a factor of two. This additional mass transfer
resistance was explained by the change of the phase behaviour. At the liquid/liquid interface a
microemulsion layer was formed which could be determined by oscillating drop
measurements [7] and colloidal probe atomic force microscopy measurements [8]. From these
0
0,5
1
1,5
2
0 2 4 6 8 10 12
Contact time t [s]
Mea
nco
nce
ntr
ati
on
c*=
c(t)
/cr[-
]
rigid interface
water (c)
Triton X-100 (s)
dp= 2,0 mm
cPADA,c= 0,25 mg/L
dispersed phases
octanol decanolcTX-100
[mmol/L]
cTX-100
[mmol/L]
0
10 10
0
146
results the situation shown in Figure 4 at the liquid/liquid interface was derived. Due to the
formation of the microemulsion layer the viscosity increases and further the interface itself
increased. Therefore, the deceleration of the mass transfer was explained.
Figure 4: Schematic formation of the microemulsion layer at the Liquid/liquid interface [13].
CONCLUSIONS
The results of this work show that for both systems regarded the change of the phase
behaviour must be taken into consideration for high surfactant concentrations. The fluid
dynamics of single droplets are a useful tool to quantify the characteristics of the liquid/liquid
interface. For both systems it was possible to derive that with exceeding the critical micelle
concentration the fluid particles behave like rigid spheres. Nevertheless, besides the
adsorption process the formation of the microemuslion layer at the liquid/liquid interface can
not be neglected, otherwise the mass transfer is overestimated.
Next to the homologous row of long chain alcohols other test systems should be investigated
to gain a fundamental understanding of the formation of these additional mass transfer
resistances. Otherwise, the design of extraction columns or multiphase reactors in which
micellar systems might occur can not be accomplished satisfactorily.
ACKNOWLEDGMENTS
This work is part of the Collaborative Research Center “Integrated Chemical Processes in
Liquid Multiphase Systems” cooridanted by the Technische Universtitat Berlin. Financial
support by the Deutsche Forschungsgemeinschaft (DFG) is gratefully acknowledged (TRR
63)
org. phase
aq. phase
org. phase
aq. phase
147
REFERENCES
[1] Anastas, P., Warner, J., "Green Chmistry: Theory and Practice", 1998, New York:
Oxford Universtiy Press.
[2] Dwars, T., Petzhold, E. Oehme, G., "Reaktionen in mizellaren Systemen (in German),
Angewandte Chemie, vol. 44, pp.7174-7199, 2005.
[3] Lee, Y.-L. "Surfactants effects on mass transfer at single droplets during drop
formation and drop falling stages, " AIChE-J., vol. 49, pp.1859-1869, 2003.
[4] Wegener, M., Paschedag, A.,"The effect of soluble anioinc surfactants on rise velocity
and mass transfer at single droplets", Int. J. Heat and Mass Trans., vol. 55, pp. 1561-
1573, 2012.
[5] Lindland K.P., Terjesen, S. G., "The effect of a surface-active agent on mass transfer in
Falling drop, Chem. Eng. Sci., vol. 5, pp. 1-12, 1956,
[6] Beitel, A., Heideger,W. J., "Surfactants effects on mass transfer from drops subject to
interfacial instability. Chem. Eng. Sci. vol. 26, pp. 711-717, 1972.
[7] Paul, N., Schrader, P., Enders, E. Kraume, M., "Effects of phase behaviour on mass
transfer in micellar liquid/liquid systems, Chem. Eng. Sci. vol 115, pp. 148-156, 2014.
[8] Paul, N. Schön, S., v. Klitzing, R., Kraume, M., "Transport processes in micellar
systems" submitted for publication to AIChE-J..
[9] Wegener, M., Grünig, J., Stüber, J., Paschedag, A., Kraume, M. " Transient drop rise
velocity and mass transfer of a single droplet with initerfacial instabilities -
experimental investigations. Chem. Eng. Sci. vol. 62, pp 2067-2078, 2007.
[10] Feng, Z.-G., Michaelides, E. E., Drag coefficients of viscous spheres at intermediate
and high Reynolds numbers. Journal of Fluids Engineering, AIChE Journal, Vol. 7,
2001, pp. 841-849.
[11] Martin, H., Wärme und Stoffübertragung in der Wirbelschicht (in German). Chemie
Ingenieur Technik, Vol. 52, 1980, pp. 199-209.
[12] Thorsen, G., Stordalen, R. M., Terjesen, S. G., On the terminal velocity of circulating
and oscillating liquid drops. Chemical Engineering Science, Vol. 23, 1963, pp. 413-
426.
[13] Paul, N. "Theoretische und experimentelle Untersuchungen von Transport- und
Grenzflächenphänomenen in mizellaren Flüssig/flüssig-Systemen" (in German), PhD
thesis, TU Berlin, 2014.
[14] Clift, R., Grace, J. R. "Bubbles, drops and particles", Academic Press, New York,
1978,
[15] Lochiel, A.C., Calerbank, P.H., Mass transfer in the continous phase around
axissymmetric bodies of revolution. Chem. Eng. Sci. vol.19 471-484. 1964.
148
[16] Saien, Javad and Simin Asadabadi, “Adsorption and Interfacial Properties of Individual
and Mixtures of Cationic/Nonionic Surfactants in Toluene + Water Chemical
Systems”, J. Chem. Eng. Data, vol. 27, 3817-3824, 2010.
149
3. HEAT TRANSFER
151
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
HEAT TRANSFER IN HORIZONTAL GROUND HEAT EXCHANGER – ONE-
DIMENSIONAL MODEL
B. LARWA; K. KUPIEC; M. GWADERA; T. KOMOROWICZ
Cracow University of Technology, Faculty of Chemical
Engineering and Technology
Abstract. A mathematical model based on one-dimensional heat conduction
equation with an internal source of a ground heat exchanger cooperating with a
heat pump was presented. The model has been verified by a comparison of
computational results with the results of measurements presented in the literature.
Thermal calculations concerning the ground under natural conditions are also
presented. The model can be useful for simulation and design of ground heat
exchangers.
Keywords. Horizontal ground heat exchangers, renewable energy sources, transient heat
conduction
INTRODUCTION
Ground heat exchangers are essential parts of the ground-source heat pumps. The
accurate prediction of their performance is very important. The ground heat exchanger is used
for extraction or injection of heat from/into the ground by a heat transfer fluid which
circulates in a closed cycle. The heat transfer fluid is water or an organic liquid of a low
freezing point (e.g., aqueous solution of ethylene glycol).
The heat exchangers for houses are embedded either horizontally or vertically in the
ground. Horizontal ground heat exchangers can be made in various configurations: serial or
coil systems. The exchangers are made mostly of plastic pipes (PVC, polyethylene,
polypropylene) arranged at a depth of 1.2 ÷ 2.0 m. The average temperature of heating
medium depends on the depth of heat exchanger location, on the arrangement of pipes and on
the ground type.
152
Horizontal ground heat exchangers have been widely used in many countries as a heat
source for ground-source heat pump systems. Therefore, ground heat exchangers are the
subject of many studies that are both experimental and numerical. Wu et. al [1, 2] investigated
the thermal performance of slinky heat exchangers for Ground Source Heat Pump (GSHP)
systems for the UK climate. The authors presented results of experimental measurements as
well as of numerical simulation using a 3D model.
Computational fluid dynamics CFD is frequently used for digital simulations. For
cooling purposes in a continuous and cyclic operating mode a three-dimensional numerical
simulation was carried out by Benazza et al. [3]. The study was focused on a horizontal
configuration based on heat exchanger pipes laid out concentrically The variation of the
climatic conditions in transient simulation was taken into account by modeling the air and
ground temperature using a simple harmonic function. The effects of thermal conductivities
and geometrical parameters on the heat exchanger efficiency were studied. Condego et al. [4]
performed calculations using Fluent and the simulations covered one year of system operation
both in summer and winter for typical climate conditions of the South of Italy.
An overview of the applications of ground heat exchangers and numerical models
related to them was presented by Florides and Kalogirou [5]. Several calculation models are
presented for ground heat exchangers and the main input data were discussed. The present
models are further refined and can accommodate for any type of grid geometry that may give
greater detail of the temperature variation around the pipes and in the ground. Furthermore,
monitoring systems, which have been set up to test various prototype constructions with
satisfactory results were discussed.
In this work a simple, one-dimensional model of a horizontal ground heat exchanger
based on the transient heat conduction equation with an internal source is proposed. The
results obtained on the base of the model are compared with the results of measurements
presented in literature. In addition, the effect of the thermal resistance of heat transfer between
the ground surface and the environment on the temperature profiles in sub-surface layers of
the ground is analyzed theoretically.
HEAT CONDUCTION IN THE GROUND
The equation of heat conduction has the form:
2
2
x
Ta
t
T
(1)
The boundary condition for the surface of the ground is as follows:
153
aTThdx
Tkx
000 (2)
The ambient temperature Ta changes periodically according to the relationship:
maxcos ttBTT ba (3)
where the cycle time tc = 365 days, hence the frequency ω = 2π/tc = 0.199·10-6
s-1
.
The ground should be considered as a semi-infinite body. Hence, the second boundary
condition related to a constant temperature of the ground at a great depth is following:
bTTx (4)
The solution of equation (1) with boundary conditions (2-3) and (4) leads to the following
relationship for the cyclic steady state [6]:
2max1 cosexp CXttXBCTT b (5)
where X is a dimensionless position coordinate. The constants C1 and C2 are dependent on the
Biot number characterizing the relation between the internal (in the ground) and external (on
the ground surface) resistance to heat transfer:
112
1
Bi
BiC (6)
1
1tan 1
2Bi
C (7)
In definitions of X and Bi the quantity L is a characteristic linear dimension defined as
follows:
aL
2 (8)
The heat flux on the surface of the ground can be determined by calculation of the
temperature gradient at the ground surface (based on the Fourier equation and (5)). One gets:
4cos2 2max
1 Ctt
L
kBCF (9)
The maximum heat flux was determined. One obtained:
L
BkCF 1
max 2 (10)
The time, after the heat flux reaches the maximum value is as follows:
8
1
2
2max
Cttt c (11)
154
For a time higher by tc/2 then the time specified above (i.e. after a half-year period) the heat
flux reaches the value according to the formula (10), but with the opposite sign (minimum
function) - this applies to the transport of heat from the ground to the environment. The
integration of the heat flux over the time leads to the determination of the amount of heat q
transported to/from the ground per unit area. The limits of integration for heating the ground
(including the half-year period) result from the condition 00
xxT . Therefore, the values
of cosine function (9) are positive. In the next half-year period the temperature gradient at the
surface of the ground is positive, the values of cosine function are negative and the ground
transfers the heat to the environment (q < 0). The resulting lower limit of integration (when
the ground is heated) is tmax+(C2−3π/4)/ω and the upper: tmax+(C2+π/4)/ω. As a result of
integration one gets:
42max
432max
12
Ct
Ct
DBCdtFq (12)
where:
kcD (13)
Knowing the temperature profiles in the ground one can determine the average temperature of
the sub-surface layer of the ground. This quantity has an apparent sense because its value
depends on the depth of the ground layer which is considered to be the sub-surface layer
(below the sub-surface layer the ground has a temperature approximately independent of the
time and position). In calculations it was assumed that the thickness of the sub-surface layer is
2π-fold higher than the characteristic linear dimension (2πL is the wavelength). The average
temperature of the sub-surface layer of the ground was calculated according to the formula:
2
02
1dXTTT b (14)
where: ΔT = T−Tb.. When substituting (5) to (14) one gets after integration:
2max2max1 cossin
4CttCtt
BCTT b
(15)
In the particular case of absence of the thermal resistance of heat transfer between the surface
of the ground and the environment is h0→∞ and Bi→∞. Since then C1 = 1, C2 = 0 and the
above relationships simplify.
In the Figs.1a and b the temperature profiles in the ground, determined according to
relationship (5) are presented. The ground temperature profiles are shown for 3-month
155
intervals (January, April, July, October). The calculations were performed for Bi = 2 and Bi
→ ∞. As one can see from the Figs.1a and b, for X > 5 the temperatures change slightly. In
addition, for smaller values of Bi the temporal variations of temperatures in the sub-surface
layers of the ground are insignificant.
a) b)
Fig.1. Ground-temperature profiles for a) Bi → ∞, b) Bi = 2
The external thermal resistance is not only a result of the thermal resistance of heat transfer
between the ground and the air. The thermal resistance can be additionally increased by the
value of Rs caused for example by the presence of snow on the ground surface. For this case
the Biot number should be generalized to the form:
sRhk
L
Rk
LBi
01
11 (16)
For example, for a = 0.5·10-6
m2/s the following value of a linear dimension was calculated
according to (8): L = 2.24 m. For h0 = 10 W/(m2K), k = 1.5 W/(mK) and Rs = 0 the Biot
number is equal to 15. If there is a layer of snow (ks = 0.18 W/(mK)) 0.12 m thick on the
ground surface, then Rs = 0.12/0.18 = 0.67 m2K/W and the Biot number is about 2. Such a
modification causes significant changes in temperature in the sub-surface layers of the
ground.
Temporal changes in the heat flux between the ground surface and the environment are
determined from relationship (9) and are shown in Fig.2. The effect of the Biot number,
inversely proportional to the external thermal resistance was analyzed. The greater the
external thermal resistance, the smaller the heat flux between the ground and the environment
for certain values of B, k and L.
156
Fig.2. Temporal changes of the heat flux to/from the surface of the ground
The value of the external thermal resistance affects not only the heat flux, but also the date of
occurrence of the maximum heat flux. For B = 10 K, k = 1.5 W/(mK), L = 2.24 m and
tmax = 182 days (the warmest day of the year − 1st of July) exemplary values were determined.
For Bi → ∞ is C1 = 1 and from the formula (10) FmaxL/(Bk) = 1.414 was obtained; this value
corresponds to the maximum heat flux Fmax = 9.5 W/m2. The date, when this value of flux is
reached, was determined from relationship (11): t = 182−365·0.125=182−46 = 136 (16th
of
May). For Bi = 2 is C1 = 0.632 and C2 = 0.322 (according to (6) and (7)). Therefore,
FmaxL/(Bk) = 0.894; this value corresponds to Fmax = 6.0 W/m2. The date, when maximum heat
flux is reached for Bi = 2, was also calculated: t = 182−365·0.074 = 182−27 = 155 (4th of
June).
Fig.3. The effects of Bi and k on the amount of heat taken over by the ground
157
Fig.3 refers to the total amount of heat per unit area of the ground q, for the half of the year
when the direction of heat transfer does not change (in the warmer half of the year). The
relationship between q/B and the Biot number (formula (12)) for various values of heat
conduction coefficient k are presented. The calculations were carried out for
cρ = 6.5 MJ/(m3K)) and cρ = 1.3 MJ/(m
3K)). With the increase in the values of k and Bi the
amount of heat taken over by the ground during the warmer half of the year increases.
Exemplary, for B = 10 K, k = 1.5 W/(mK), cρ = 6.5 MJ/(m3K)) and Bi = 2 one obtains
q = 88.2 MJ/m2, whereas for Bi→∞ a higher value was obtained: q = 140 MJ/m
2.
In Fig.4 the temporal courses of the average ground temperature throughout the year are
shown. Also in this case there is a major impact of the Biot number on the obtained
calculation results. For Bi→∞ the lowest ground temperature appears on the 14th
of February
and the highest – on the 16th
of August. For Bi = 2 the average temperature reaches the
extreme value 18-19 days later. The extremes of temperatures in Fig.4 correspond to temporal
changes of the direction of heat flux (F = 0) in Fig.2.
Fig.4. Courses of average temperature of top layers of the ground
GROUND HEAT EXCHANGER MODEL
A system which consists of the lower and upper heat exchanger, where a working fluid
(glycol solution) circulates between them, has been considered. The lower heat exchanger is
located under the ground, while the upper is a part of the heat pump.
158
The heat is transferred between the environment and the ground to the ground surface. Air
temperature periodically changes in time (on an annual basis).
In this model the flow through parallel arrangement of heat exchanger pipes was replaced by a
flow through a horizontal cuboid channel of small thickness. The heat is transferred into
(from) the ground symmetrically by both the lower and the upper surfaces of the heat
exchanger.
In the case of ground heat exchangers the thermal resistance occurs in working fluid and in
walls of pipes but mainly in the ground. Individual resistances are related to areas through
which the heat is transported. The presented model includes only the resistance to heat
transfer in the ground. The consequence of this is that the surface of the pipes is not used in
the model. For the modeled exchanger it can be assumed that the heat transfer surface is the
surface of the ground Ag where exchanger pipes are installed. The quantity of Ag should be
treated as an adjustable parameter of the model that depends on the way of pipes arrangement
and the depth of their location under the ground surface.
In this model, the ground is treated as a semi-infinite body. The heat conduction equation for
the plate with an internal heat source placed at a some distance from the ground surface was
used. Due to the small thickness of the heat exchanger one includes heat transfer in the
ground/plate only in the vertical direction. In this model, the equation for an infinite plate, for
which one surface is the surface of ground, and the other is located at a great depth providing
the ground temperature stability is used. For transient conduction in an infinite plate with an
internal heat source the following relationship is valid:
c
q
x
Ta
t
Tv
2
2
(17)
The quantity of qV is a rate of heat generation of the heat source per unit volume: qv =Q /V
where V is the volume of horizontal cuboid. Q is related to the transport of heat between the
working fluid flowing through the ground heat exchanger and the ground.
Fig. 5. Heat exchanger as a cascade of perfect mixing tanks
159
In Fig.5 the circulation of working fluid in the heat exchanger is presented. The ground
exchanger, considered in the model as a m-stage cascade of perfect-mixing tanks, is coupled
with the upper exchanger, wherein heat is transported to (or received from) the
thermodynamic medium undergoing a phase transition (condensation or evaporation). The
more stages in the cascade, the model more precisely reflects the change in temperature of the
working fluid flowing through the heat exchanger pipes. When m → ∞, the temperature
profile of the fluid in exchanger becomes continuous.
A fluid from the upper exchanger flows to the tank 1 at a temperature of Tin = TL0. Fluid
leaving the tank 1 flows into the tank 2, then into the tank 3, etc. The fluid from the last mth
tank is transported at a temperature of Tout = TLm to the upper exchanger. The fluid
temperature TLj, the ground temperature Tj and the rate of heat transfer j
Q are different in
various stages of the cascade j = 1, 2,…, m. Heat transfer area for each stage is equal to
Agj = Ag/m, while the volume: Vj = V/m. The rate of heat transfer between the ground and fluid
in the jth
stage is equal to:
1,
jLLjLLj
TTcmQ (18)
and the total rate of heat transfer in the exchanger equals:
m
j
jQQ1
(19)
Because the thermal resistances to heat transfer in a fluid and in pipe wall were neglected,
hence the temperature of the ground in contact with the outer surface of the heat exchanger
pipes is equal to the temperature of the working fluid.
As an initial condition the temperature profile for cyclic steady state under natural conditions
was assumed:
XttXBTTt b maxcosexp;0 (20)
The boundary condition takes the form (2-3) and (4). Model equations have been solved with
the finite difference method using the Crank-Nicolson scheme.
COMPARISON WITH THE EXPERIMENTAL RESULTS
In order to verify the presented mathematical model of the horizontal ground heat exchanger,
the temperatures of working fluid at the inlet and outlet of the ground exchanger were
generated computationally in period of 53 days during heating season as well as were
compared with the measurement results presented by Wu et al [1, 2]. The cited authors
researched the ground heat exchanger used as a lower heat source of a heat pump for space
160
heating. Horizontal coils of exchanger pipes were arranged in four rows with a width of 1 m
and a length of 80 m each, at a depth of 1.14 m. The flow rate of an aqueous solution of
glycol was 0.57 kg/s. The exchanger operated continuously. The thermal diffusivity of the
ground (averaged over the depth) was 0.533·10-6
m2/s.
Characteristic values for the climate in which the research was conducted (around London)
were assumed as follows: Tb = 9.5°C, B = 9.0 K. In the calculations based on the model the
following values of parameters were used: m = 4, n = 300 (number of calculation nodes),
time step Δt = 1 h. The necessary value of the averaged external thermal resistance depending
on i.a., the heat transfer coefficient (including radiation) between the air and the ground was
assumed h0 = 20 W/(m2K).
The measurement results together with the results of calculations according to the
presented model are depicted in Fig.6. The symbols represent the experimental values read
out from the drawings shown in [1, 2] that were obtained between 6th
of November and 30th
of
December. As one can see, the predictions based on the model are correct, although there are
small differences in temperatures determined experimentally and computationally.
Fig. 6. The comparison of inlet and outlet temperatures of the working fluid in the ground heat exchanger
It should be noted that the compatibility of temperature profiles in Fig.6 depends largely
on the consistency of the actual air temperature in the research and the preceding periods with
the values used in the calculations. The averaging the physical properties of the ground
(which determine the thermal diffusivity) and the roughly assumed value of the heat transfer
coefficient h0 negatively affect the accuracy of the results obtained from the digital simulation.
The increase of the experimental liquid-temperature (for a certain period) indicates that the
weather conditions were not the ‘model conditions’. It means that under autumn conditions
161
the ambient temperature increased. It was not taken into account in the presented model. The
model is based on the regular sinusoidal temperature changes. To generate data for Fig.6
Ag = 525 m2 was assumed.
CONCLUSIONS
The resistance of heat transfer between the surface of the ground and the environment
(external thermal resistance of heat transfer) strongly affects the heat flux and the
amount of heat transferred between the ground and the environment as well as on the
temperature distribution of the sub-surface layer of ground. Reduction of this resistance
is beneficial during receiving of heat by the ground exchangers.
It is not recommended to place any objects on the ground above the exchanger because
they hinder the thermal regeneration of the ground.
Heat transfer in a horizontal ground heat exchanger can be described by a model based
on the one-dimensional transient heat conduction equation with the internal source of
heat.
The variability in the temperature of a fluid flowing through the heat exchanger can be
taken into account by treating the ground heat exchanger as a cascade of perfect-mixing
tanks.
The fluid temperatures at the inlet and the outlet of ground heat exchangers determined
with the presented mathematical model are consistent with the measurement results
presented in the literature.
Calculations based on the presented model can be useful in predicting the impact of
different process parameters on the heat pump heating power.
LITERATURE
[1] Wu Y., Gan G., Gonzalez R.G., Verhoef A., Vidale P.L., 2011, Prediction of the thermal performance
of horizontal-coupled ground-source heat exchangers, International Journal of Low-Carbon
Technologies 0, 1-9.
[2] Wu Y., Gan G., Verhoef A., Vidale P.L., Gonzalez R.G., 2010, Experimental measurement and
numerical simulation of horizontal-coupled slinky ground source heat exchangers, Applied Thermal
Engineering, 30, 2574-2583.
[3] Benazza A., Blanco E., Aichouba M., Rio J.L., Laouedj S., 2011, Numerical investigation of
horizontal ground coupled heat exchanger, Energy Procedia, 6, 29-35.
162
[4] Congedo P.M., Colangelo G., Starace G., 2012, CFD simulations of horizontal ground heat
exchangers: A comparison among different configurations, Applied Thermal Engineering 33-34,
24-32.
[5] Florides G., Kalogirou S., 2007, Ground heat exchangers – A review of systems, models and
applications, Renewable Energy, 32, 2461-2478.
[6] Carslaw H.S., Jaeger J.C., 1959. Conduction of Heat in Solids, second ed., Clarendon Press, Oxford,
65-74.
NOMENCLATURE
a thermal diffusivity of the ground, m2/s,
A − surface area, m2,
B − half of the annual maximum temperature range, K,
Bi (= h0L/k) − Biot number,
C1, C2 − constants dependent on the Biot number,
c − heat capacity of the ground, J/(kg·K),
F − heat flux, W/m2,
h – the distance between the heat exchanger and the ground surface, m,
h0 – heat transfer coefficient, W/(m2·K),
k – thermal conductivity of the ground, W/(m·K),
m − mass flow rate, kg/s,
qv – rate of heat generation per unit of volume, J/m2,
Q - rate of heat transfer, power of a heat pump, W,
t − time, days,
tc − cycle time, days,
tmax − time from the beginning of the year until the maximum ambient temperature is reached, days,
T – temperature, oC,
Tb − temperature of the ground at a great depth, oC,
x position coordinate, m,
X (= x/L) dimensionless position coordinate,
ρ ground density, kg/m3,
Indices:
163
0 − ground surface,
a – ambient (environment),
g − ground,
j − number of stage in a cascade,
L − working fluid,
− − average value.
165
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
OPTIMIZATION OF THE LUCID DETECTOR HEAT SHIELD COOLING
R. KANTOR
Cracow University of Technology, ul. Warszawska 24, 31-155 Krakow,
Poland
Abstract. LUCID (LUminosity Cherenkov Integrating Detector) is an array of
Cerenkov tubes made of aluminum, located at the both ends of the ATLAS
detector as a part of LHC which is one of CERN experiments. The purpose of the
LUCID is to monitor the luminosity in ATLAS experiment. The requirements of
quality, force a Beam Pipe to be exposed to elevated temperature after its
assembly to outgas all volatile surface contaminants. This baking process
temperature peaks at 250°C, and the LUCID detector would be exposed to
unacceptable high temperature, exceeding the maximum temperature 50°C. The
water-cooled aluminum Heat Shield have been used as a protecting system. The
aim of present analysis is to find an optimal solution of the cooling system in
terms of cooling loops layout and dimensions, to maintain, as much as possible,
uniform temperature distribution along the Heat Shield, with aid of CFD method.
Keywords. Heat Shield, CFD, optimization, cooling, temperature stabilization.
INTRODUCTION
LUminosity measurement using a Cherenkov Integrating Detector (LUCID) [1] is an
array of Cherenkov tubes [2] (length ~1500mm), made of polished aluminum filled with
C4F10 as a radiator. It is located at the both ends of the ATLAS detector [3] as a part of Large
Hadron Collider which is one of CERN experiments located near Geneva.
The main purpose of the LUCID is to count the number of particles hitting the detector and it
is used as a measurement of a luminosity [2].
The extremely high standard of a Beam Pipe [3] cleanliness requires the pipe to be warmed up
at elevated temperature after its assembly to outgas all volatile surface contaminants. The
166
baking temperature peaks at 250°C for typically up to 24 hours. Without any special
protection, the LUCID detector would be temporarily exposed to temperatures high enough to
compromise its more vulnerable components, like o-ring seals, etc., what force ambient
temperature to be kept below 50°C.
PROBLEM DESCRIPTION
To protect all sensitive components there was one 8 mm layer of an Aerogel jacket used,
which was however no more than a passive insulator. Therefore a water-cooled aluminum
screen was designed, as a heat shield, to counter-act thermal effects. Additionally this could
be the way to carry out the unnecessary heat load. The idea of the Heat Shield had to be
verified on the protective ability. A series of 2-D and 3-D CFD studies with ANSYS Fluent
software have been performed. Several different approaches to the problem allowed to select
and introduce the most efficient variant of a cooling strategy.
Geometry of the Thermal Shield; thermal requirements
The main part of the Heat Shield makes a 1800 mm long, 2 mm thick aluminum pipe of
diameter 160 mm. The Heat Shield is actively cooled by six U-shaped axial loops made of a 4
mm diameter copper tubes, attached to the outer surface of the pipe, which are circulated
with chilled water. A general 3D view is shown on the Figure. 1.
Figure 1. Geometry of the LUCID with embedded Heat Shield in between the LUCID and the Beam Pipe.
In addition to an Aerogel insulating jacket on the Beam Pipe, which consists of 8 mm layer of
Pyrogel 6350 [1], there is an Air Gap between the outer surface of the Aerogel and the inner
Cooling loops
Thermal Shield
Beam
Pipe
Cherenkov
tubes
167
surface of the Heat Shield. A relatively low heat conductivity of air allows to treat it as an
additional layer of the insulating material and consequently, to lower an installation cost.
Such an assumption is allowed only, if there is no convection of air in the Air Gap volume.
MATERIAL PROPERTIES
All essential properties of materials used to build the Heat Shield are gathered in the Table 1.
Table 1. Physical properties of materials used in CFD simulation.
Material Density ρ
[kg/m3]
Conductivity k
[W/mK]
Heat capacity C
[J/kgK]
Heat shield Aluminum
pipe: 6061-T6 2719 237 871
Aerogel: PYROGEL 6350 170 0,0282 N/A
Air 1,205 (or Ideal Gas) 0,0257 1006
Cu tubes 8978 410 381
Cooling water 998,3 0,6 4182
Determination of the cooling method
Assuming the heat load 250 W/m, generated along the Beam Pipe and its outer surface
temperature 250 ºC, there is 650 W/m2 of the total out-flowing heat flux to carry out. If we
choose a flow of 20 kg/h in each cooling loop, and six loops to remove the total heat, as a
result temperature of the cooling water is expected to raise no more than 2 ºC.
Cooling loops are attached to the shield pipe with aluminum-loaded heat-conductive epoxy
adhesive. Its special conductive properties provides a high thermal conductivity of over 1,5
W/mK. However it is assumed, that in the most pessimistic case, under unpredictable
circumstances, the adhesive bond may possibly get broken. So in result there could an air gap
appear. Therefore, in the CFD calculations, the adhesive layer will be treated as an air gap.
Moreover an external surface of the Heat Shield has been defined as an adiabatic wall
boundary condition, what practically supports the conservative approach as well.
CFD MODEL AND BOUNDARY CONDITIONS
The CFD modeling of the heat removal allows to determine the minimal cooling water flow,
mean temperature and other parameters to keep temperature of the Heat Shield below 50 ºC,
what had been determined as a critical condition.
As mentioned in the Problem Description, ANSYS Fluent has been used for modelling and
solving the problem. The simulation studies were made both, in 2-D and 3-D domain. Each
168
simulation had its distinct advantages for the comprehension of the problem and, on the other
hand, a totally different mesh.
As a first approach, the 2-D study have been performed (Figure 2). Assuming in the first
approach that the model can be divided into six identical pieces as it consist of six identical
cooling loops distributed evenly along a perimeter of the Heat Shield, there only the 60-
degree segment out of the total 360° of its cylindrical geometry section had to be simulated.
On account of an irregular profile the 2-D mesh consists of 9086 quadrangular and, near
critical point located around Al/Cu bonds, nonstructural cells.
Figure 2. 2-D simplified geometry, material properties and boundary conditions.
This kind of simplification excludes an air convection, which is possible to be modeled as a
real phenomenon only in the full cross section domain. Moreover the 2-D approach does not
allow to simulate a fully developed cooling fluid flow. In consequence temperatures of the
inner walls of Cu pipes have to be defined a priori. Taking into account working parameters
of the assigned water chiller, temperature of circulating water was set to 20 ºC at the water
inlet and 22 ºC at the outlet. Consequently the simulated model was simplified to a heat
transfer effect between two elements – the Beam Pipe and the Cu cooling piping. However
the 2-D model, as in many other cases, gives a huge advantage on account of a model
Aerogel k=0.015 W/mK
Beam Pipe surface:
temp. 523K
Al shell: 2mm k=237 W/mK
Air Gap k=0.0242 W/mK Symmetry
Symmetry
Cu pipe: k=410 W/mK Surface temp. 293 K
Adiabatic wall
Adhesive k=1.5 W/mK
169
preparation effort and time consuming. Regardless of a complexity of mathematical models,
time of simulation, using a standard PC, was no longer than 1 min. For comparison, the 3-D
model consists of 1.353.575 cells (Figure 3).
Figure 3. 3-D geometry and boundary conditions (b.c); all remaining b.c. as in the 2-D case.
On the other hand the 3-D approach allows to develop the model and apply to it both
temperature distribution along the entire Heat Shield length, and the transfer of heat between
the Heat Shield and the cooling water. As an advantageous outcome of the 3-D modeling
there could be found:
1. Temperature difference resulting of the water flow rate and its final temperature,
2. Temperature distribution on the whole surface of the Heat Shield, what is important,
taking into account, that temperature of cooling water increases along with Cu tube.
3. The minimal water flow rate, to keep temperature of the Heat Shield below 50ºC.
This information is useful for selecting the most suitable water chiller.
In addition to boundary conditions shown on the Figure 2, the 3-D modeling requires also
parameters of water at the inlet [3]:
- Inlet velocity w = 0,768 m/s arises from assumed mass flow rate 20 kg/s and Cu tube
diameter d = 3 mm,
- Inlet water Temperature: 20°C,
- Dynamic fluid viscosity m = 0,001 Pa·s.
RESULTS
The solution results are expressed in the forms of Node Values plots along predefined paths.
For the aim of the analysis, an attention is focused on the temperature distribution on the Heat
Shield external surface, so in the 2-D case it is the outer circumference of the Shield, and in 3-
Water Inlet
20 kg/h; 293 K
Water Outlet
170
D the path is drawn from cross-section located at 0,1 m from the Inlet/Outlet plane.The graphs
below (Figure 4) show no significant difference between temperature distributions of 2-D and
3-D.
Figure 4. Profiles of temperatures on the aluminum Heat Shield surface – comparison between 2-D and 3-
D approach.
Above shown diagrams confirm that 2-D and 3-D solutions are, within numerical errors
limits, comparable. However attempts of comparison are justified in so far, as it concerns the
Inlet/Outlet zone. The wave-shaped profiles of temperature are an effect of uneven
temperature distribution according to positions of cooling pipes. There is no information
about the second end of the Heat Shield, where on the one hand, we expect the temperature of
water to be raised, on the other hand the cooling loop does not cover whole length of the Heat
Shield (Figure 3). The importance of the full 3-D approach is shown on the Figure 5.
Figure 5. Profile of the temperature along an axial cross-section of the Thermal Shield surface (3-D) and
the location of the path (right side).
Curve length [m]
Tem
pera
ture
[K]
Sta
tic
Tem
per
atu
re [
K]
Curve length [mm]
3D Temperature of Heat Shield, z=0,1m
2D Temperature of Heat Shield
171
The temperature profile along the axisymmetric cross-section shows that lack of any heat sink
at one end causes of significantly raised the Heat Shield temperature. The temperature peak at
around 333K (60°C) exceeds considerably the maximaum value limited to 50°C.
Moreover this 20 degrees difference generates unnecessary thermal stress, exposing
particularly the adhesive bond, to a mechanical destruction.
AIR CONVECTION EFFECT AND QUALITY OF ADHESIVE
The natural convection in the air gap between Aerogel and the Heat Shield is the effect which
typically should be restricted by the geometric parameters or uniform temperature
distribution. In case of the LUCID Heat Shield the natural convection can be restricted by
narrowing the air by means of increasing thickness of the Aerogel layer. This however causes
rise of the total cost of the LUCID assembly.
In the present chapter discussion on the air natural convection phenomena in the LUCID Heat
Shield assembly and its influence on the temperature distribution along the Heat Shield
perimeter was carried out. There were three cases considered:
1. Case I - simulation without natural convection effect,
2. Case II - simulation with natural convection effect,
3. Case III - simulation with natural convection effect and unbroken adhesive bonds between Cu
cooling pipes and the Thermal Shield.
Case I - simulation without natural convection effect
In addition to CFD models and parameters used earlier to simulate the heat transfer by
thermal conduction, the Ideal Gas model and Gravitational Force had to be activated. The
natural convection is driven by the temperature difference between the Aerogel external
surface and the Heat Shield inner surface.
The Case I, which Temperature Profile is shown on the Figure 6, is the simulation without
modeling of natural convection effect. The simplified 2-D CFD simulation on a cross section
of the Heat Shield and the Beam Pipe has been carried out.
172
Figure 6. Case I - the Temperature Profile along the Heat Shield’s inner wall - simulation without
modeling of natural convection effect.
The Temperature Profile in the Case I (Figure 6) shows no effect of variation along the
perimeter and fits within narrow range between 40.6 °C and 40.7 °C, what clearly is the
effect of superposition of numerical errors and the Cu cooling loops layout. If there is no
natural convection and other non-symmetrical effects considered, a simplification of geometry
is allowed, and results of simulations are similar. The difference may be evaluated by
comparison of temperature profiles shown on the Figure 4 and Figure 6.
Case II - simulation with natural convection effect
The Case II, which Temperature Contours are shown on the Figure 7, is the simulation
including modeling of natural convection effect in the Air Gap volume.
Sta
tic
Tem
per
atu
re [
K]
Position [mm]
173
Figure 7. The Temperature Contours on the cross section of the Heat Shield.
The natural convection effect is significant for chosen Air Gap thickness and temperature
difference between the Aerogel and the Heat Shield surfaces. Velocity Vectors within the Air
Gap are shown on the Figure 8. The Temperature Contours and the Velocity Vectors confirm
the expectations mentioned above.
174
Figure 8. The Velocity Vectors on the cross section of the Heat Shield.
The Temperature Profile shown on the Figure 6, is drawn on a base of the simulation
including natural convection effect. The horizontal axis named “Position” indicates distance
from the symmetry axis of the cross section.
Figure 9. Case II - the Temperature Profile along the Heat Shield inner wall - simulation with modeling of
natural convection effect.
Aerogel
Air gap
Heat Shield Beam Pipe
Sta
tic
Tem
per
atu
re [
K]
Position [mm]
175
The Temperature Profile in this case indicates significant influence of the natural convection
on the Temperature Profile, which varies from 42.6 °C to 49.5 °C.
The lowest temperature is read at the bottom of the Heat Shield and it is slightly higher than
the temperature at the same location in the Case I. The highest temperature, located on top of
the Heat Shield, reaches temperature close to the limit of 50 °C.
The temperature difference of 7 °C is acceptable only if the maximum temperature is
considered and there is no specific requirements for uniformity of temperature of the whole
Heat Shield surface.
Case III - simulation with natural convection effect and a proper adhesive bonds
Typically, if there is no break in the adhesive bonds between Cu cooling pipes and the Thermal Shield,
temperature of the Heat Shield stays at maximum 26 °C, what is shown on the Figure 10.
Figure 10. Case III - the Temperature Profile along the Heat Shield inner wall - simulation with modeling
of natural convection effect and a proper adhesive bonds.
The temperature difference of 3,6 °C is almost a half of the same difference in the Case II. The cause
for such a considerable difference lies in the sort of the chosen adhesive, which Thermal Conductivity
is at least 1,5 W/mK, while in case of air it amounts to 0,023 W/mK.
SUMMARY AND CONCLUSIONS
Heat transport by natural air convection around the Heat Shield external surface, on account
of geometrical limitations, is rather negligible. The space is occupied by the array of
Cerenkov tubes and other instrumentation. The results of the CFD obtained with given input
parameters, indicate significant effect of the forced convective cooling by water, moreover the
Sta
tic
Tem
per
atu
re [
K]
Position [mm]
176
cooling system is able to remove the total amount of unnecessary heat generated by Beam
Pipe around the LUCID area.
3-D simulation allows to consider longitudinal effects and to find the Temperature Profile
along the axis of the Heat Shield. One of problems discovered by 3-D analysis is the area at
the far end of the Heat Shield, which requires stronger attention (Figure 11). Temperature of
the fragment of the Heat Shield without cooling pipes reaches 60°C and is unacceptable, as
mentioned in the Introduction.
Figure 11. Geometrical view of the optimized Cu loop path and appropriate temperature profiles along
the cross-section (see Figure 5).
There was considered two ways solving the problem. First possibility is to increase the mass
flow rate and second, to modify the geometry by extending the Cu loops up to the end of the
Heat Shield. The first method seems to be the easiest, and does not require any mechanical
changes, but due to potential problems described in the Results, it has been rejected. The
second one, although requires some modifications of LUCID assembly design and its adjacent
elements, proved itself to be a much better solution and finally have been chosen. As a result,
Te
mp
era
ture
[K
]
Distance [m]
Te
mp
era
ture
[K
]
Distance [m]
First approach design Optimized design
Temperature of the Heat Shield along Z-coordinate between Cu pipes of the loop
177
the Temperature Profile along the Heat Shield turned into almost the uniform 40°C (Figure
11). Having in mind that such results have been obtained on very conservative assumptions,
e.g. lack of physical contact between Cu pipes and the Heat Shield, the presented results,
bring a strong degree of confidence that the LUCID vulnerable elements should be safely
protected from reaching excessive temperature.
Considering more strict requirements of temperature level and uniformity than requirements
for the LUCID Detector, it is advisable to protest the air domain against developing the
natural convection effect, e.g. by using baffles or by increasing thickness of the Aerogel layer.
ABBREVIATIONS
LUCID - LUminosity Cherenkov Integrating Detector
C4F10 - Perfluorobutane (R610)
ATLAS - A Toroidal LHC Apparatus (experiment at the CERN physics laboratory)
CERN - Conseil Européen pour la Recherche Nucléaire
LHC - Large Hadron Collider
REFERENCES
[1] JELLEY J. V.: Cerenkov Radiation and Its Applications. In: London, Pergamon, 1958.
[2] ATLAS Collaboration, Atlas Detector and Physics Performance, Technical Design
Report, In: ATLAS TDR 14, CERN/LHCC 99-14, 25 May 1999.
[3] ATLAS Collaboration, ATLAS Forward Detectors for Luminosity Measurement and
Monitoring, Letter of Intent. In: CERN/LHCC/2004-010, 22 March 2004.
[4] ANSYS Fluent Documentation.
[5] ASPEN Aerogel Product Overview. In: http://www.aerogel.com/products/overview.
[6] PINFOLD J.: Plans for the Very Forward Region of ATLAS - The LUCID Luminosity
Monitor. In: Astroparticle, Particle and Space Physics, Detectors and Medical Physics
Applications, Proceedings of the 9th Conference, Villa Olmo, Como, Italy, 17 - 21
October 2005, 379-388
[7] STANISZEWSKI B., Wymiana ciepła, podstawy teoretyczne, wydanie drugie
poprawione In: Warszawa, PWN, 1979.
179
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
HEAT TRANSFER ENHANCEMENT IN NATURAL CONVECTION IN MICROPOLAR
NANOFLUIDS
K. NERING; K. RUP
Institute of Thermal and Process Engineering, Cracow University of
Technology al. Jana Pawła II 37, 31-864 Kraków, Poland
Abstract. This paper presents an analysis of momentum, angular momentum and
heat transfer during the unsteady natural convection in micropolar nanofluids.
Selected nanofluids treated as single phase fluids contain small particles with
diameter size d = 10 nm and d= 38.4 nm. In particular, two ethylene glycol based
nanofluids and two water-based nanofluids were analyzed. Volume fraction of
these solutions was 6%. First ethylene glycol solution contained Al2O3
nanoparticles (d = 38.4 nm), and the second ethylene glycol solution contained Cu
nanoparticles (d = 10 nm). For water–based nanofluids, the first one contained
Al2O3 nanoparticles (d = 38.4 nm), and the second one contained Cu nanoparticles
(d = 10 nm).
Keywords. Micropolar fluid, Nanofluid, Heat transfer enhancement
INTRODUCTION
Conventional fluids, such as water, oil, alcohol, ethylene glycol, widely used in heat
exchange devices, has relatively low thermal conductivity coefficient. Recently, a new
generation of heat carriers known as nanofluids has been developed [1-4]. These types of
fluids consist of conventional fluid and nanoparticles with diameter of particle between 10
and 100 nm mixed uniformly with fluid. Generally, they contain particles of substances such
as Al2O3, TiO2, CuO and Cu [1, 4]. The discussed nanofluids are characterized by increased
effective thermal conductivity and dynamic viscosity. During experimental studies nanofluids
behave like a single phase Newtonian fluid in convectional heat exchange process [2-5].
180
Recently in literature [4, 5] was presented develop of methods based on large number of
experimental data, uses to determine nanofluids thermophysical parameters. These
correlations provide theoretical and practical analysis of heat exchange due to natural
convection. Paper [1] analyzes the process of steady natural convection in nanofluid in the
vicinity of a vertical plate heated by constant heat flux. In particular water suspension of
Al2O3 and CuO was analyzed. Volume fraction of these suspensions did not exceed 10%.
Similar work [2] described natural convection in water suspension of Al2O3 with the same
thermodynamic conditions. Another paper [3] describes numerical solution of equations of
conservation of mass, momentum and energy in natural convection process in water
suspensions of Al2O3 and CuO placed in six different closed areas. Increased heat exchange
was observed only in triangle-shaped area. The amount of increase was only 5% compared to
water without nanoparticles [3].
Due to miniaturisation of heat exchange devices, micropolar fluids as refrigerant or
heating media are also analysed in literature [6-8]. A useful model of micropolar fluid is a
model proposed by Eringen. This model takes into account fluid microrotation [6-8].
The aim of work described in this paper is the analysis of increased heat exchange due
to natural convection in water and ethylene glycol solutions of Al2O3 and Cu with properties
of micropolar nanofluids in the vicinity of vertical plate heated by heat flux of q0 that rises
suddenly.
ESTIMATING PROPERTIES OF NANOFLUIDS
The typical approach used to study thermodynamic properties of nanofluids is based on
assumption that nanofluids behave like single phase fluids. There are empirical equations
proposed by authors used to determine different features of nanofluids such as thermal
conductivity, viscosity, density and thermal expansion [4, 5]. It is worth mentioning that all
models are applicable only in specific range of nanofluid parameters.
Several authors are proposing different methods to estimate heat conductivity of nanofluid.
This parameter is the most important with respect to the heat transfer process [4, 5]. Based on
large amount of data presented in [4] a method of heat conductivity calculation was proposed:
10 0.03
0.4 0.66 0.66
4.4 Re Prs
f f f
fr f
T
T
(1)
Equation (1) is suggested especially when nanofluid is based on water and ethylene glycol
with Al2O3, TiO2, CuO or Cu nanoparticles. In equation (1) Reynolds number is given by
equation:
181
2
Re2
f b
f p
k T
d
(2)
and Prandtl number is:
Pr( )c
(3)
Equation (3) shows that Prandtl number increases after adding nanoparticles to base fluid.
Recently, there are many models determining dynamic viscosity have been developed [5]. For
example classic models such as Einstein (4) or Brinkman (5) models:
1 2, 5
f
(4)
2,5
1
f
(5)
For water suspension of nanoparticles of Al2O3, authors recommend the following
relationship [4]:
2
123 7, 3 1f
(6)
Recently, using large amount of experimental data from many authors, an empirical equation
to determine dynamic viscosity has been proposed [4]:
0.3 1.031 34.87
f
p fd d
(7)
To calculate the equivalent diameter of a base fluid molecule from equation (7), an equation
proposed by [4] was used:
0
6f
f
dM
N
(8)
One of the methods to determine density, heat capacity and thermal expansion coefficient is
the conventional approach [4, 5]. It can be assumed that nanofluid is a single phase fluid.
Thus those parameters can be calculated as in case of mixtures. It is given by:
1
f s
(9)
1
f sc c c
(10)
1
f s
(11)
In energy equations, heat capacity and thermal expansion coefficient are always considered
with fluid density, thus equations (10) and (11) will be used.
182
PROBLEM FORMULATION
In this paper unsteady laminar heat and momentum exchange in nanofluids in terms of natural
convection will be considered. Nanofluid is in the vicinity of a vertical plate. The heat flux
through the plate rises suddenly to the value of q0.
Figure 1. Considered fluid schema
Problem presented in this work will be solved using the following assumptions:
Oberbeck-Boussinesq approximation is assumed.
The analysed flows geometry justifies the use of the boundary layer theory.
Viscous dissipation and pressure work are neglected.
Eringen’s theory of thermomicrofluid is assumed.
Taking into account the simplification resulting from the boundary layer the boundary layer
theory and fluid density changes according to the Oberbeck-Boussinesq approximation the
following system of equations can be obtained:
Taking into account the simplification resulting from the boundary layer theory and fluid
density changes according to the Oberbeck-Boussinesq approximation, the following system
of equations can be obtained:
u v0
x y
(12)
2
2
u u u 1 u κ Nu v μ κ g t t
τ x y ρ y ρ y
(13)
2
2
N N N γ N κ uu v 2N
τ x y ρj y ρj y
(14)
2
2
t t t tu v a
τ x y y
(15)
The above system of partial differential equations together with the following boundary
conditions:
183
,0 0vu , tt (16)
,0 0x , 0vu , tt (17)
0qt uy 0, u v 0, , N n
y y
(18)
y , u 0, N 0, t t
(19)
formulates the mathematical description of momentum, angular momentum and heat transfer
driven by the unsteady convection in micropolar nanofluids.
In equations (12) ÷ (15), u and v being the velocity components in the x and y directions, N
the microrotation component in the xy-plane, τ the time, ρ the density, μ the dynamic
viscosity, κ the rotational viscosity coefficient, γ the spin-gradient viscosity, j the microinertia
density, a the thermal diffusivity, β the coefficient of volumetric expansion and t the fluid
temperature. In the present analysis, the spin gradient viscosity is assumed to be [7,8]:
γ μ j2
(20)
In the last condition listed in (18) we have assumed that the microcirculation on the boundary
layer is equal to the angular velocity, namely, , 0,u
N x ny
. As the suspended particle
cannot get closer than its radius to the wall, the microstructure effect must be negligible on the
boundary. Therefore, in the vicinity of the boundary, the rotation is due to fluid shear and thus
the microrotation must be equal to the angular velocity of the boundary.
In condition (18), the parameter n is a number between 0 and 1 and that relates microgyration
vector to the shear stress. The value n = 0 corresponds to the case of the high density of liquid
microparticles that prevents them from performing rotational movements in the vicinity of the
wall. The value n = 0.5 is indicative of weak concentrations, at n = 1 flows are believed to
represent turbulent boundary layers [7,8].
The fluid differential equations are recast in a dimensionless form by introducing:
01 1
2 3 4 2
0
t t τT , τ
q 1 λ 1[ν ( ) ] [( ) ]
λ gβ q gβ
(21)
1 1
2 20 04 4
u vU , V
q q[ν g ] [ν g ]
λ λ
(22)
1
4x1 1
2 24 4
0 0
x y yX , Y (Gr )
xλ 1 λ 1[ν ] [ν ]
q gβ q gβ
(23)
184
1
40 0 2
x 2
q qgβGr x , N N[gβ ]
ν λ λ
(24)
12
2
x1
0 2
νκ 1 xΔ , P (Gr )
ν ρ j jq( g )λ
(25)
The set of partial differential equations (12) – (15) together with initial and boundary
conditions (16) – (19) in dimensionless form will be solved numerically using finite
difference method [8, 9].
SOLUTION OF A PROBLEM
Equations (12) – (15) will be solved using explicit finite difference scheme. Spatial
distribution grid contains M x N points in the X and Y directions respectively, is the time
step. Due to the intensive heat, momentum, angular momentum and mass transfer, only in the
direct vicinity of the considered vertical surface, the maximum values of dimensionless
coordinates X = 100 and Y = 30 were assumed [8]. A characteristic feature of the difference
equations was to determine the temperature field, the velocity field components and the
microrotation component N at time 1n depending on certain parameters, but determined at
time n . Convection terms of balance equations comprising time derivatives and spatial Y
coordinate derivatives were approximated by “forward” formulas whereas spatial X
coordinate derivatives were approximated by “backward” formulas. Diffusion terms were
approximated by central differences. Derivatives appearing in the boundary conditions (18)
were approximated by higher order difference formulas taken in the form [9]:
3
ij i,j 1 i,j 2 i,j 3
|ij
T 1( 11T 18T 9T +2T ) O[ Y]
Y 6 Y
(26)
4
|i,j i,j i,j 1 i,j 2 i,j 3 i,j 4
|ij
1 U 1N ( 25U 48U 36U 16U 3U ) O[ Y]
n Y 12 Y+
(27)
These difference formulas are statically stable and exhibit characteristics of conservation [9].
Before performing basic calculations for the established, non-zero values of parameters Δ and
P describing the properties of micropolar fluid, calculation tests were done similarly to [8]. In
the process of steady natural convection in a Newtonian fluid, exact analytical solutions are
known [10], and were compared to the corresponding calculation results. On the basis of trial
calculations, further ones, taking into account the non-zero values of Δ and P parameters,
were performed with the following spatial area division: M x N = 250 x 150, the set size of
time step = 0.002. Assumed area division is smaller than area division in work [8] and the
185
time step is two times greater. This change of area division and time step needs to be done to
obtain greater accuracy of the applied differential forms.
RESULTS AND DISCUSSION
The set of equations (12) – (15) with initial condition (16) and boundary conditions (17) –
(19) were integrated for the selected values of parameters Pr∞, P, ∆ and n. Ethylene glycol
(G) and pure water (W) in temperature of 60°C ware base fluids. The Prandtl number of
ethylene glycol was Pr∞=56.310 and Prandtl number of water was Pr∞=3.000. In the next
stage of analysis it was assumed that base fluid has micropolar features with the following
parameters: ∆ = 5.0, P = 1.0 and n = 0.5. These parameters were assumed based on literature
and previous own work [8].
Main analysis was focused on the effects occurring in nanofluids. In this work, the following
homogeneous water and ethylene glycol solutions of nanoparticles were analysed:
water solution of Al2O3 nanoparticles with mean diameter of 38.4 nm (W+Al2O3),
ethylene glycol solution of Al2O3 nanoparticles with mean diameter of 38.4 nm
(G+Al2O3),
ethylene glycol solution of Cu nanoparticles with mean diameter of 10 nm (G+Cu).
Nanoparticle volume fraction for the above solutions was φ = 6%. Parameters describing
these solutions for temperature 60°C calculated using equations (1) – (11) were presented in
Table 1.
Table 1. Thermophysical parameters of nanofluids in temperature t = 60C
Fluid
den
sity
ρ [
kg/m
3]
dyn
amic
vis
cosi
ty
μ [
kg
/(m
·s)]
ther
mal
co
ndu
ctiv
ity
λ [
W/(
m·K
)]
hea
t ca
pac
ity
ρc p
[J·
m3/K
]
ther
mal
ex
pan
sio
n
ρβ
[k
g/(
K·m
3)]
Pra
ndtl
nu
mb
er
Pr ∞
no
rmal
ised
coo
rdin
ates
X/X
f; Y
/Yf
no
rmal
ised
par
amet
er
Δ/Δ
f
no
rmal
ised
par
amet
er
P/P
f
Water (W) 983.24 4.688·10-4 0.6590 4111221.4 0.4956 3.000 1.000 1.000 1.000
W + Al2O3
(38.4 nm) 1157.05 0.9089·10-3 0.9757 4042640.1 0.4678 3.253 0.668 0.516 2.238
Ethylene glycol (G) 1088.1 5.706·10-3 0.2598 2789779.6 0.6202 56.310 1.000 1.000 1.000
G + Al2O3
(38.4 nm) 1255.61 1.247·10-2 0.4630 2800485.0 0.5850 60.052 0.598 0.458 2.798
G + Cu
(10 nm) 1559.21 3.039·10-2 0.633 2828907.0 0.6104 87.185 0.378 0.188 6.999
Fig. 2 (left) shows the dimensionless U velocity component in the X-axis direction of the
ethylene glycol (Pr∞ = 56.310) and ethylene glycol based nanofluids with Prandtl number Pr∞
186
= 60.052 and Pr∞ = 87.180 at fixed times of the process = 30 and 255. In order to simplify
the analysis of the thermophysical parameters values of ∆ and P were considered to equal
zero.
For the assumed Grashof number (Grx=108), dimensionless coordinate X adopts values from
Table 1. For ethylene glycol nanofluids: (G+Cu) X value equals XG+Cu
= 0.378Xf G;
(G+Al2O3) X value equals XG+Al2O3
= 0.598Xf G.
Figure 2. Profiles of the velocity component U (left) and fluid temperature changes (right) at selected
moments of the process
Fig. 2 (right) presents the temperature profiles in the considered liquids at certain moments of
the process = 10, and 80 for ethylene glycol based nanofluids. Similarly as for the U
velocity component, proper values of parameters describing the thermophysical properties of
fluids were assumed. The temperature of heated plate is lower for fluid with nanoparticles
(∆=0; P=0). High temperature of the heated vertical plate indicates significantly smaller
intensity of heat interception by the analyzed micropolar fluid than other fluids. Fig. 2 (right)
shows larger rate of heat interception intensity in the entire range of time in comparison with
base fluid.
On the basis of temperature profile specified changes of local Nusselt number Nux in the
analyzed fluid on heated vertical plate:
ox
w
q xNu
t t
(28)
Using dimensionless equations (20) – (24) with Nusselt number (28) we obtain:
1
5
1
5
1x
wx
NuX
TGr
(29)
where
187
ww 1
2 30 4
t tT
q 1[ ( ) ]
g
(30)
Figure 3. Transient changes of the local Nusselt number.
The relationship (29) is shown graphically in Fig. 3 (left side for water based fluid and the
right side for ethylene glycol based nanofluids). For the sake of comparison, Fig. 3 comprises
the corresponding curve obtained for the selected values of parameters Pr∞, P, ∆. Curves from
Fig. 3 represent local Nusselt number with respect to local Grashof number (Nux/(Grx)1/5
)
specific for value Grx = 108. It is worth mentioning that corresponding lines of parameter
Nux/(Grx)1/5
have different dimensionless X coordinate. For pure liquid there is no change of X
coordinate, but for nanofluids this coordinate changes due to equations (23) and (24) with
Grashof number Grx = 108 similarly as for the U velocity component. As indicated in Fig. 3,
intensity of heat exchange in micropolar fluid is significantly lower than in corresponding
nanofluids.
On the basis of calculated velocity field, a shear stress on vertical plate was determined.
Taking into account constitutive equations for micropolar fluid [1, 2]:
w |y 0
uτ [(μ κ) κN]
y
(31)
After adding dimensionless equations (21) – (24) to above equation (31) we obtain:
ww
2 2/5|y 02/5 3/5
x2
1 Uτ
(5X) Y5 Gr 1 n
x
(32)
188
Table 2. A comparison of results
Pr∞ Δ P 15
x xNu Gr w
3.0
W
0.0 0.0 0.7108* 0.5054*
0.0 0.0 0.72091 0.47181
5.0 1.0 0.58602 0.22234
3.253
W + Al2O3
(38.4 nm)
0.0 0.0 0.7252* 0.4918*
0.0 0.0 0.73755 0.45634
2.580 2.238 0.64371 0.27699
56.310
G
0.0 0.0 1.3795* 0.16189*
0.0 0.0 1.4117 0.15302
5.0 1.0 1.1311 0.0691
60.052
G + Al2O3
(38.4 nm)
0.0 0.0 1.39844* 0.15778*
0.0 0.0 1.4351 0.14886
2.290 2.798 1.2402 0.09164
87.180
G + Cu
(10 nm)
0.0 0.0 1.5126* 0.13592*
0.0 0.0 1.5632 0.12753
0.940 6.999 1.45053 0.09924
In order to make a comparative analysis, Table 2 summarizes the Nusselt number values
according to the (29) formula and the dimensionless shear stress in accordance with (32)
formula, obtained from the numerical calculations performed for the variable parameters ∆, P,
Pr∞ and constant n parameter (n = 0.5). The summarized results relate to the steady state with
Grashof number 108, which is reached for nanofluid when dimensionless coordinate X is
lower than X coordinate for pure liquid (Xf = 100). This coordinate is measured along the
vertical plate. Exact values of quotient X/Xf and Y/Yf (Table 1) according to relationships (1),
(7), (9), (10) and (22) taking into account respective values of thermophysical parameters of
pure liquid and considered nanofluids. In Table 2 values with * was taken from [10]. Result
from [10] was calculated with exact analytical solution of conservation equations for a
Newtonian fluid.
Figure 4. The changes of the profiles of dimensionless microrotation at certain moments of the process.
189
Fig. 4 presents the dimensionless component profiles of microrotation N in selected moments
of the heating process in nanofluids with micropolar properties. The greatest microrotation
changes are observed in analyzed nanofluids in the vicinity of vertical plate (Y<5) heated with
constant heat flux q0. Left side of Fig. 4. shows microrotation profiles for water based
nanofluids and the right side presents microrotation for ethylene glycol based nanofluids.
Heat transfer enhancement during natural convection in considered nanofluids is represented
by the following equation:
1/5
x x
1/5
x x f
Nu GrE 1
Nu Gr (33)
Table 3 presents values of E parameters calculated with relationship (33) for the considered
nanofluid in stationary case. Calculating these values corresponding results from Table 2 were
used. Maximum value of E parameter is for nanofluid with Cu nanoparticles with mean
diameter of 10 nm.
Table 3. Obtained values of the E parameter with respect to equation (33)
fluid Δ P 15
x xNu Gr E [%]
W 0.0 0.0 0.72091f -
W + Al2O3 (38.4 nm) 0.0 0.0 0.73755 2.31
W (micropolar) 5.0 1.0 0.58602f -
W + Al2O3 (38.4 nm) 2.580 2.238 0.64371 9.8
G 0.0 0.0 1.4117f -
G + Al2O3 (38.4 nm) 0.0 0.0 1.4351 1.66
G + Cu (10 nm) 0.0 0.0 1.5632 10.7
G (micropolar) 5.0 1.0 1.1311f -
G + Al2O3 (38.4 nm) 2.290 2.792 1.2502 9.6
G + Cu (10 nm) 0.94 6.999 1.45053 28.2
CONCLUDING REMARKS AND CONCLUSION
In this paper, a process of heat and momentum exchange during natural convection in
nanofluids with micropolar properties was analyzed. To describe the analyzed phenomena of
exchange, equations of hydrodynamic and thermal boundary layer were used.
Parameter E describing heat transfer enhancement between heated plate and the nanofluid
showed in Table 3. Maximum value of E parameter for nanofluid without micropolar
properties appears for ethylene glycol based nanofluid with Cu nanoparticles (E = 10.7%). On
the other hand, analyzing nanofluids with Al2O3 nanoparticles, greater value of heat transfer
enhancement can be observed in water based nanofluid (E = 2.31%).
190
Micropolar fluids are fluids with non-zero values of ∆ and P parameters. These fluids are
characterized by different behaviour during natural convection. Maximum value of E
parameter for nanofluid with micropolar properties appears for ethylene glycol based
nanofluid with Cu nanoparticles (E = 28.2%). Significantly higher temperature value of
heated plate after initial time of the process in the vicinity of micropolar fluid indicates lower
intensity of heat exchange by the analyzed micropolar fluid compared to Newtonian fluid.
The highest changes of microrotation component N are observed before time = 10 at the
beginning of the process with both type of nanofluids.
In order to perform a comparative analysis of results presented in this work, the exact result
for Newtonian fluids was quoted from literature.
REFERENCES
[1] Popa C., Fohanno S., Nguyen C. T., Polidori G.: On heat transfer in external natural
convection flows using two nanofluids. Int. J. of Thermal Science 49, (2010) 901-908.
[2] Polidori G., Fohanno S., Nguyen C. T.: A note on heat transfer modeling of
Newtonian nanofluids in laminar free convection. Int. J. of Thermal Science 46,
(2007) 739-744.
[3] Abuali O., Ahmadi G.: Computer simulations of natural convection of single phase
nanofluids in simple enclosures: A critical review. Applied Thermal Engineering 36
(2012), 1-13.
[4] Coricone M.: Empirical correlating equations for predicting the effective thermal
conuctivity and dynamic viscosity of nanofluids. Energy Convers. Manage. 52(2011)
789-793.
[5] Kakaç S., Pramuanjaroenkij A.: Review of convective heat transfer enhancement with
nanofluids. International Journal of Heat and Mass Transfer 52 (2009), pp. 3187-3196.
[6] Eringen A. C.: Theory of Micropolar Fluids, J. Math. Mech. 1966, 16, pp. 1 – 18.
[7] Mohammedan A. A., Gorla R. S. R.: Heat transfer in a micropolar fluid over a
stretching sheet with viscous dissipation and internal heat generation. Int. J. of
Numerical Methods for Heat and Fluid Flow, 2001, Vol. 11, No. 1, pp. 50-58.
[8] Rup K., Dróżdż A.: The effect of reduced heat transfer in miclropolar fluid in natural
convection. Archives of Thermodynamics 34 (2013), 45-59.
[9] Tannehill J.C., Anderson A. D., Pletcher R. H.: Computational Fluid Mechanics and
Heat Transfer. Taylor & Francis, Washington 1997.
[10] Martynenko O. G., Sokowiskyn J. A.: Svobodnokonvektiwnyj teploobmen na
vertikalnoj poverchnosi, Nauka, Mińsk, 1977.
191
4. SEPARATION PROCESSES
193
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
FILTERABILITY OF W/O PICKERING EMULSIONS IN MULTIPHASE REACTIONS
A. DREWS; L. SCHUMACHER; T. SKALE; D. ZEDEL
Hochschule für Technik und Wirtschaft Berlin, FB II, Process
Engineering in Life Science Engineering, Wilhelminenhofstr. 75A,
12459 Berlin, Germany, [email protected]
Abstract. The use of Pickering emulsions has recently received increased
attention in multiphase reactions for both chemical and biocatalysis. For a
continuous process, an efficient separation step is required to retain the aqueous
phase with the (bio)catalyst inside the reactor. In this study, it could be shown for
the first time that membrane filtration is feasible for such a w/o Pickering
emulsion separation which makes these emulsions a promising technology for
multiphase catalysis. In addition, the effects of emulsion properties and filtration
parameters on filterability were investigated. Solid particle content influenced
drop sizes and thus permeability. Flux was largely unaffected by L/L phase
fraction. The influence of transmembrane pressure on flux was unexpected with
fluxes increasing disproportionally with pressure, which might be attributed to
coalescence in the deposition layer.
Keywords. Pickering emulsion, ultrafiltration, L/L systems, (bio)catalysis
INTRODUCTION
The use of emulsions stabilised by solid micro- or nanoparticles, so-called Pickering
emulsions (PE), has recently received increased attention in multiphase reactions for both
chemical and biocatalysis [1] where the catalyst requires water but substrates and/or products
are mainly soluble in an organic solvent. Common dispersions or surfactant stabilised
emulsions fail in some such cases because of, e.g., interfacial inactivation of enzymes [3] or
the difficulty in separating the reverse micelles that contain the catalyst from the organic
194
phase [4]. To design a continuous PE process, an efficient separation step is required to retain
the aqueous phase with the (bio)catalyst inside the reactor while products can be withdrawn
continuously (Fig. 1). To our knowledge, there is no literature on the use of membranes for
w/o PE separation or on their filterability. The aim of this study thus is to investigate the
effects of emulsion properties and operating parameters on filterability.
Fig. 1: Principle of continuous (bio)catalysis using w/o PE
MATERIALS AND METHODS
w/o PE made from two different organic solvents (toluene and 1-dodecene) and using
different silica nanoparticles (a) synthesised acc. to [2] and functionalised with TMODS
(typically 170 nm), and b) commercial ones by Wacker Chemie AG (20 nm)) were prepared
using an Ultraturrax T-18 at 25000 min-1
for 120 s or by ultrasonication. PE were filtered in a
stirred cell (XFUF-04701, Merck Millipore, Aeff. = 13.7 cm²). Initial L/L phase fractions
(waq0 = 6.4-50%), solid particle contents (0.1-1 wt% silica) and transmembrane pressure
(TMP = 1-5 bar) were varied.
RESULTS
(Bio)catalysed reactions were successfully carried out in both solvent systems. As an
example, Fig. 2 shows that an enzyme loaded PE yields significantly higher product
concentrations than those resulting from auto hydrolysis of the substrate [5].
(bio-) cat.
substrate
product
substrate product
195
Fig. 2: Benzoin production from benzoin acetate using lipase TL in a water/toluene PE with 170 nm
silica/TMODS particles (Ultraturrax)
Both 1-dodecene and toluene w/o PE could successfully be separated by ultrafiltration at
reasonably stable permeabilities of around 3-50 L/(m² h bar), depending on particle content as
shown in Fig. 3. A higher amount of particles in relation to the dispersed (aqueous) phase
volume s can stabilise a larger interfacial area [6, 7]. Thus, droplet size decreases which
gives rise to a higher cake resistance during filtration. In fact, droplet sizes of around
25-50 μm (at s = 40 g/L) and 45-70 μm (at s = 30 g/L) were observed in the toluene PE [8].
Fig. 3: Effect of silica particle content on filterability: water/toluene PE with silica/TMODS particles
(Ultraturrax), waq0 = 6.4%, PES 10 kD (Alfa Laval)
Surprisingly, flux was found to be largely unaffected by L/L phase fraction. In most cases it
remained almost constant during batch filtration although the dispersed phase fraction
0
0.05
0.1
0.15
0.2
0 10 20 30 40P
rod
uct
co
nce
ntr
atio
n [g
/L]
time [min]
with enzyme
without enzyme
0
50
100
150
200
250
300
0 5 10 15 20 25 30 35 40
Flu
x [L
/(m
²h)]
Filtered volume [mL]
S p [bar] 3.5 5[g/L]
30
40
196
increased to up to around 80 vol.% (cf. Fig. 4). Refilling the stirred cell with solvent, i.e.,
continuous phase after a batch filtration did not yield an increase in flux.
Fig. 4 shows the maximum permeate yield of a 70 vol.% dodecene PE before water started to
pass through the membrane as indicated by the sudden drop in flux (apart from visual
observation of water drops in the permeate). In all cases, around 90 % of the feed dodecene
had been collected before water began to permeate. At this stage, the remaining dodecene
volume was smaller than the spherical packing void volume, so apparently the droplets in the
deposit layer were deformed before they leaked into the permeate. This would increase the
resistance due to reduced hydraulic diameters.
Fig. 4: Filtration of PEs after different membrane pre-treatment steps, VPE = 30 mL, aq0 = 30 vol.%,
0.5 wt% 20 nm silica (ultrasonication), PVDF 1 kD (ETNA01PP, Alfa Laval), TMP = 2.5 bar, wtip = 1 m/s
[9]
The different curves resulted from differently pre-treated membranes. Filtering a PE through
an untreated fresh membrane resulted in a steady decrease in flux over time. This, however,
was apparently not due to the increased dispersed phase fraction, as membranes that had been
washed in pure dodecene for 90-120 min at 2.5 bar prior to PE filtration showed a constant
flux after an initial drop during the filtration of the first approx. 10% (see above). These
constant fluxes were only slightly lower than the pure dodecene fluxes during the washing
step which were around 10 L/(m² h) for PE2 and 15 L/(m² h) for PE3. The different pre-
treatments did not influence the maximum yield.
0
5
10
15
20
25
30
35
40
45
0 0.2 0.4 0.6 0.8 1
Flu
x J
[L/(
m2
h)]
Vpermeate/Vdodecene,0 [-]
PE 1, no pre-treatment
PE 2, washed in dodecene for 90 min
PE 3, washed in dodecene for 120 min
197
First investigations on the influence of TMP showed a disproportionate increase of fluxes
with TMP (see also Fig. 3). After filtration at higher pressures (> 3 bar), drops were found to
have increased in size from the approx. 10 µm before filtration. Apparently, coalescence had
occurred, and thus increased hydraulic diameters might explain the observed increased
permeability.
CONCLUSIONS
It could be shown for the first time that membrane filtration is feasible for w/o Pickering
emulsion separation which makes these emulsions a promising technology for multiphase
catalysis. Solid particle content influenced drop sizes and thus permeability. Flux was largely
unaffected by dispersed phase fraction. The influence of transmembrane pressure on flux was
unexpected with fluxes increasing disproportionally with pressure, which might be attributed
to coalescence in the deposit layer. The effects of additional substrate, product or surfactant
compounds as well as pH and ionic strength on droplet size and filterability will be
investigated in the future.
REFERENCES
[1] Wu, C., Bai, S., Ansorge-Schumacher, M.B., Wang, D.: Nanoparticle Cages for
Enzyme Catalysis in Organic Media, Advanced Materials, 23 (2011) 5694-5699.
[2] Stöber, W., Fink, A., Bohn, E.: Controlled growth of monodisperse silica spheres in
the micron size range, Journal of Colloid and Interface Science, 26 (1968) 62-69.
[3] Baldascini, H., Janssen D. B.: Interfacial inactivation of epoxide hydrolase in a two-
liquid-phase system, Enz. Mic. Tech., 36 (2005) 285-293.
[4] Nguyen, L.A.T., Minding, M., Schwarze, M., Drews, A., Schomäcker, R., Kraume,
M.: Adsorption and filtration behaviour of non-ionic surfactants during reverse
micellar-enhanced ultrafiltration, Journal of Membrane Science, 433 (2013) 80-87.
[5] Schumacher, L., Plikat, C., Ansorge-Schumacher, M.B., Drews, A.: Macro kinetics of
biocatalysis in Pickering emulsions, Chemie Ingenieur Technik (2014) in press.
[6] Binks, B.P., Whitby, C.P.: Silica particle-stabilized emulsions of silicone oil and
water: aspects of emulsification, Langmuir, 20 (2004) 1130-1137.
[7] Chevalier, Y., Bolzinger, M.A.: Emulsions stabilized with solid nanoparticles:
Pickering emulsions. Colloids and Surfaces A: Physicochem. Eng. Aspects, 439
(2013) 23-34.
198
[8] Drews, A., Schumacher, L., Skale, T.: Innovative biokatalytische
Produktionsverfahren – Erschließung neuer pharmazeutischer Produktklassen und
Ressourcenschonung, in: Gesundheit: Vielfältige Lösungen aus Technik und
Wirtschaft (ed. Matthias Knaut), Berlin 2014.
[9] Skale, T., Zedel, D., Carl, A., Kraume, M., von Klitzing, R., Drews A.: Einfluss von
transmembraner Druckdifferenz, Partikelgehalt und Phasenanteil auf das
Filtrationsverhalten von Pickering Emulsionen, Chemie Ingenieur Technik (2014) in
press.
ACKNOWLEDGEMENTS
Financial support by the Federal Ministry of Education and Research (BioPICK 031A163A)
and the German Research Foundation DFG (collaborative research centre "Integrated
Chemical Processes in Liquid Multiphase Systems" InPROMPT TRR63) is gratefully
acknowledged. We thank Alfa Laval for kindly providing free membrane sample
199
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
SEPARATION OF CO2 FROM THE OCM PRODUCT STREAM USING
HYPERBRANCHED POLYMERS SOLUTIONS
C. WALOWSKI; S. ENDERS
Technische Universität Berlin, Chair of Thermodynamics and Thermal
Separation Processes, Sekr. BH 7-1, Ernst-Reuter-Platz 1, D-10587
Berlin, Germany
Abstract. To improve the gas washing of carbon dioxide the use of
hyperbranched polymers solutions as selective solvents is adopted. Due to their
structure hyperbranched polymers cannot adequately be described by common
thermodynamic models. As possible approach, the lattice-cluster-theory equation-
of-state, which directly takes into account the molecular architecture, is sketched
briefly and exemplarily applied to the binary systems CO2 + Boltorn U3000 and
propane + Boltorn H3200. The results are compared with experimental data taken
from the literature.
Keywords. CO2 washing, hyperbranched polymers, Boltorn U3000, Boltorn
H3200, lattice-cluster-theory equation-of-state (LCT EOS).
INTRODUCTION
The separation of carbon dioxide from gas stream mixtures plays an important role in
many technical applications [1]. The oxidative coupling of methane (OCM, [2,3]) shall be
offered here as example. Within the OCM process, methane reacts to ethylene which is the
main product while other components, such as water or CO2, have to be separated from the
product stream. A well-known method for this task is the two-column gas washing process
with aqueous monoethanolamine (MEA) solutions serving as solvents [4]. To further improve
this unit operation the idea of replacing the conventional amine-based solvents by
hyperbranched polymers solutions [5] shall be put forward.
200
Hyperbranched polymers have attracted much attention in the last decade(s) for their
potential use in different technical applications [6]. Due to their structure they exhibit an
enormous capacity for the incorporation of functional (chemical) groups. Both their structure
as well as their chemical nature control their physical properties and make them predestines
for tailor-made solutions not only in the chemical industry.
The design of every process, however, requires adequate thermodynamic models and
reliable data on the components involved. In this paper, the lattice-cluster-theory (LCT)
equation of state (EOS) by Langenbach et al. [7-9], which is based on the LCT by Dudowicz
and Freed [10,11], is applied to overcome these shortcomings. The theory in sketched briefly
and used to calculate the fluid-phase equilibria of the binary systems CO2 + Boltorn U3000
and propane + Boltorn H3200.
THEORY
Lattice Model and Description of the Molecular Architecture
The LCT EOS is based on a Sanchez-Lacombe-like lattice [12] with the coordination number
z that is made of lN lattice sites which may be either occupied by molecule segments of
species i ( iN ) or being left empty. The latter “void” sites or “holes” ( iN ) account for the
compressibility and the number of lattice sites are related to each other by
i
i
ivl NMNN (1)
where iM is the segment number of species i.
It shall be assumed that each component exhibits an individual occupied site volume as well
as an individual void site volume given by
3
,, iocciiocc Mv (2)
3
,, ivivv (3)
In Eqs. 2 and 3 ,occ i and ,v i are the length of an occupied and void lattice site, respectively.
These are adjustable pure-component parameters which, together with Eq. 1, can be used to
express the volume V as follows,
,( )v v i i i occ i
i
V n v n M v (4)
when following abbreviations are introduced:
,v i v i
i
v v (5)
201
,occ i occ i
i
v v (6)
1 i
i iM M
(7)
The averaged lattice void volume, occupied volume and segment number in Eqs. 5-7 are
dependent on the composition which is expressed by the segment fractions:
i ii
j j
j
n M
n M
(8)
Within the LCT the segment number is not the only property to a priori describe the nature of
molecules. All required combinatorial numbers are compiled in Table 1.
Table 1: Combinatorial numbers and their meanings [9].
Combinatorial
number
Description Example (2,3-
Dimethylbutane)
M Number of united atom groups 6
1N Number of distinguishable ways to choose bonds
between united atom groups
5
2N Number of distinguishable ways to choose two
consecutive bonds between united atom groups
6
3N Number of distinguishable ways to choose three
consecutive bonds between united atom groups
4
N Number of ways finding three bonds meeting at one
segment
2
'N Number of ways that three bonds meet at one segment
and one of them is adjacent to another bond not
connected to the branching segment
4
N Number of ways to find four bonds meeting at one
segment
0
1,1N Number of ways to find two bonds not meeting in one
segment but being on the same molecule
4
1,2N Number of ways to find one bond not adjacent to two
consecutive bonds on the same molecule
4
2,2N Number of ways to find two consecutive bonds not
adjacent to two further consecutive bonds in the same
molecule
1
202
Actually, the number of combinatorial numbers required can be reduced in some special
cases. Langenbach et al. [7-9] used following relations by Nemirovsky et al. [10] to further
simplify the LCT summation contributions:
1 2 1,1 1 12 2 0N N N N N (9)
2 3 1,2 1 22 2 3 0N N N N N N (10)
While the combinatorial coefficients of small molecules are easily accessible, they can be
derived for hyperbranched polymers by knowledge of a few characteristics (Figure 1) such as
the number of core segments, the separator length and the generation number g. As example
the description of the polymer Boltorn U3000 is presented in Section 3.
Figure 1: Schematic presentation of a hyperbranched polymer of generation number 3g [8].
Helmholtz Free Energy
The Helmholtz free energy F of any system is known to be dependent on volume, temperature
and composition. Within the lattice model the volume is represented by the void and occupied
volume fractions, respectively:
vv
l
N
N (11)
1occ v (12)
The LCT EOS yields an expression for the Helmholtz free energy which consists of two parts
- first, on the left-hand side of Eq. 13, the mean-filed (MF) contribution and second, on the
right-hand side, the extended mean-field (EMF) contributions abcC [9].
203
;
2 2 62
;
0 0 1
( , , , , , ) ( , , , )
( , , , )
MFv i i g i v i i
l B l B
a a b c
v occ abc i g i ij
a b c
F T z M N S z M
N k T N k
T z C z N
(13)
While the well-known first contribution takes into account the molecular size only ( iM ), the
latter – the essence of the LCT EOS – is dependent on both molecular size ( iM ) and
molecular structure ( ;g iN ). The mean-field contribution is known to be given by:
( , , , )ln ln
MF
v i i iv v i
i vl B i
S z M
N k M
(14)
To express the EMF contributions in a compact way the exchange interaction energy
parameter ij shall be introduced:
2ij jj ijii
B B B Bk k k k
(15)
Furthermore, a notation for the products of the segment fractions is applied for simplification:
0 1 1
0 1 1
, , , ,
; ; ; ;
:n n
n n
i i i i
i i i i
(16)
The athermal-limit contributions ( 0a ) are then given by [9]:
vi
i
icC 001001 (17)
vi vj
ji
ijcC ,002002 (18)
vi vj vk
kji
ijkcC ,,003003 (19)
vi vj vk vl
lkji
ijklcC ,,,004004 (20)
105 0C (21)
0106 C (22)
where c is a function of molecular structure determined by the combinatorial numbers in
Table 1. The first-order contributions ( 1a ) read [9]:
0101 C (23)
vi vj
jiij
ijcC ,102102 (24)
204
vi vj vk
kjijk
ijkcC ,,003103 (25)
vi vj vk vl
lkjikl
ijkcC ,,,104104 (26)
vi vj vk vl vm
mlkjilm
ijkcC ,,,,105105 (27)
0106 C (28)
vi
iii
icC 111111
(29)
vi vj
jijj
ijcC ,112112
(30)
vi vj vk
kjikk
ijkcC ,,113113
(31)
vi vj vk vl
lkjill
ijkcC ,,,114114
(32)
0115 C
(33)
0116 C
(34)
Finally, the second-order contributions ( 2a ) can be calculated via [9]:
0201 C
(35)
vi vj
jiij
ijcC ,
2
2202 )1,1,2,1(
(36)
vi vj vk
kji
ikij
iij
jkik
i
jkjkij
ij
cc
ccC ,,0
22
0
2
2
2
203)0,1,2,0()0,0,0,2(
)0,0,1,0()2,2,4,4(
(37)
vi vj vk vl
lkjiij
jlik
ij
ilik
ij
jkkl
ij
ijkl
cc
ccC ,,,
22
22
204)0,0,0,4()0,0,0,2(
)0,1,4,4()1,2,4,2(
(38)
2 , , , ,
205
2 , , , ,
( 8,3, 1,0)
( 2,0,0,0)
ij
jk lm i j k l m
m v
iji v j v k v l v kl lm i j k l m
m v
c
Cc
(39)
vi vj vk vl vm
nmlkji
vn
mnkl
ijcC ,,,,,2206 )0,0,0,3( (40)
2
211 2 (0, 2,1,2)ij
ii i
i v
C c
(41)
205
2 0
2 2
212 ,0
2 2
( 4,4, 2, 2) (0,2,0,0)
( 2,4, 2, 2) (0,4, 2, 2)
ij i
ij jj jj
i jij ii v j v ii jj ij ii
c cC
c c
(42)
vi vj vk
kjiij
ii
ij
jj
ij
kkjk
ij
jjkk
ij
ijkk
ccc
ccC ,,
222
22
213)2,2,4,8()0,1,4,4()0,1,4,4(
)0,1,4,4()2,4,8,4(
(43)
lkji
vi vj vk vl llkk
ij
llkl
ij
kljj
ij
lljk
ij
cc
ccC ,,,
22
22
214
)0,0,0,2()0,0,0,4(
)0,2,6,8()0,2,6,16(
(44)
vi vj vk vl vm
mlkjimmkl
ijcC ,,,,2215 )0,0,0,12( (45)
216 0C (46)
0 2
221 2 (0,2, 1, 2)i
ii i
i v
C c
(47)
vi vj
ji
ij
jj
ij
jjii ccC ,2
2
2222 )0,1,4,2()4,4,8,4( (48)
vi vj vk
kjikk
ij
vi vj vk
kjikkjj
ij ccC ,,
2
2,,2223 )0,0,0,2()0,4,12,16( (49)
vi vj vk vl
lkjillkk
ijcC ,,,2224 )0,0,0,12( (50)
0225 C (51)
0226 C (52)
Based on the Helmholtz free energy all further thermodynamic properties can be derived
using the framework of standard thermodynamics. The pressure, for example, is given by the
equation:
42
00 ;
1
1 51 1 1
1 ;
0 1
2 62 1
2 ;
0 1
ln (1 ) ( , , , , )
( ) ((1 ) ) ( , , , , )
( ) ((1 ) ) ( , , , ,
cv occv occ occ c i i g i ij
c
b c b c
B v occ v occ bc i i g i ij
b c
b c b c
B v occ v occ bc i i g i
b c
Pvz c C z M N
RT M
z k T b c b C z M N
k T b c b C z M N
)ij
(53)
For calculation of the phase equilibrium the chemical potential is commonly applied, here
given as segment-molar property:
206
, ,
1 1
; ;2 2 6
2
10 0 1
, ,
1
( )
(( 1) (1 ) )
( )
( )
MF
B B T V n M n M
b c b c b c
v v occ v v occ v occ abc
a a
B b c abca b c v occ occ occ
T V n M n M
S
k T k n M
n b c bn c C
k T z Cn M
n M
(54)
where
; ; ,
, ,
1: ( )
( ) ( )
v v vv v i v occ
v i occ occT V n M n M
nn v v v
n M v
(55)
For further information about the required derivatives the reader is referred to [9].
RESULTS
The LCT EOS is applied to the two binary systems CO2 + Boltorn U3000 and propane +
Boltorn H3200. The first system is of special relevance for the OCM process. Although
propane is also a side-product of the reaction, the second system is primarily presented here to
show the capability of the LCT EOS.
For the pure-component parameters it is assumed that the lengths of the void and occupied
lattice sites are the same ( , , :v i occ i i ); the energy parameter is proposed to have the
form
0 0 1/ / /ii B ii B ii ii Bk k k T (56)
so that there are three pure-component parameter for each component ( 0 1, ,ii ii ii ).
Concerning the molecular structure, propane is suggested to consist of three segments (
1 2 33; 2; 1; 0M N N N ). While this choice is quite intuitive, CO2, however, could be
modeled as consisting of one, two or three segments. Here the first case is assumed (
1 2 31; 0; 0; 0M N N N ).
Figure 2 shows the saturation densities and vapor pressure of CO2. Although the LCT EOS
demonstrates its strengths for rather large molecules it is also able to adequately calculate the
CO2 saturation properties.
207
220 240 260 280 300 320
10-4
10-3
/k
B = 106.72 K
= 428.57
= 4.15667 Å
Exp. [13]
LCT EOS [8]
vo
lum
e / m
3/m
ol
temperature / K
220 240 260 280 300 320
1
10
Exp. [13]
LCT EOS [8]
/k
B = 106.72 K
= 428.57
= 4.15667 Å
pre
ssu
re / M
Pa
temperature / K
Figure 2: Experimental [13] and LCT EOS calculated [8] saturation properties of CO2.
The molecular structure of the hyperbranched polymers is modeled as depicted in Figure 1
with the generation number 3g . For Boltorn U3000 the core, the separator groups A and
the end groups B are given by following chemical formula:
Core: C(CH2O-)4
12x A: COC(CH3)(CH2O-)2
16x B: COC(CH3)(CH2OH)(CH2OR)
where R=CH3-(CH2)14CO-. The different groups are divided into segments again – the core
into 5, group A into 4 and group B into 20 segments. Considering the branching points (28
branching points of degree 3; one of degree 4) a total of 373 segments per molecule can be
derived. The second polymer, Boltorn H3200, is modeled in a similar fashion [8].
The adjustment of the hyperbranched polymer pure-component parameters is not
straightforward as there is a lack of experimental data available. For Boltorn U3000 and
Boltorn H3200 they are fitted to binary phase equilibria; the reader is referred to [8].
Figure 3 shows the LCT EOS calculated bubble point curves for the binary system Boltorn
U3000 + CO2 which are compared to the experimental results by Kozlowska et al. (2009)
[14]. The theory is in good agreement with the measurements in a broad pressure range from
2-13 MPa.
208
280 300 320 340 360 380 400 420
0
2
4
6
8
10
12
14
Exp. [14]
LCT EOS [8]
w = 0.95
w = 0.9
pre
ssu
re /
MP
a
temperature / K
w = 0.85
Figure 3: Experimental [14] and LCT EOS calculated [8] bubble point curves of the binary system
Boltorn U3000 + CO2 in dependence of the polymer mass fraction w.
The second binary system, Boltorn H3200 + propane, is presented in Figure 4 and
demonstrates the capability of the LCT EOS. Although the calculated values are not in perfect
agreement with the experimental data by Portela et al. (2009) [15], only one set of parameters
is necessary to describe the vapor-liquid-liquid equilibrium (VLLE) up to a pressure of 11
MPa.
320 340 360 380
0
2
4
6
8
10
Exp. [15]
LCT EOS [8]
pre
ssure
/ M
Pa
temperature / K
Figure 4: Experimental [15] and LCT EOS calculated [8] VLLE of the binary system Boltorn H3200 +
propane.
209
CONCLUSION
The LCT EOS has briefly been presented as thermodynamic model that is capable to directly
incorporate molecular structure. By comparison with experimental data taken from the
literature the theory has proven an adequate tool for the calculation of binary phase equilibria
with hyperbranched polymers involved. For the design of the CO2 washing process, however,
aqueous hyperbranched solutions may be used. Thus, ternary systems water + hyperbranched
polymer + CO2 have to be described in the future. For parameter adjustment and validation
not only theoretical but also experimental work is necessary due to the lack of experimental
data available for this kind of systems.
Acknowledgment
The authors thank the Cluster of Excellence UNICAT for financial support.
LITERATURE
[1] Olajire, A.A.: CO2 capture and separation technologies for end-of-pipe applications –
A review. Energy 35 (2010), 2610-2628.
[2] Stansch, Z., Mleczko, L., Baerns, M.: Comprehensive Kinetics of Oxidative Coupling
of Methane over the La2O3/CaO Catalyst. Ind. Eng. Chem. Res. 36 (1997), 2568-2579.
[3] Stünkel, S., Martini, W., Arellano-Garcia, H., Wozny, G: Entwicklung eines optimalen
CO2-Abtrennungsprozesses für die oxidative Kopplung von Methan im Miniplant-
Maßstab. CIT 83 (2011), 488-495.
[4] Aaron, D., Tsouris, C.: Separation of CO2 from Flue Gas: A Review. Separ. Sci.
Technol. 40 (2005), 321-348.
[5] Rolker, J., Seiler, M., Mokrushina, L., Arlt, W.: Potential of Branched Polymers in the
Field of Gas Absorption: Experimental Gas Solubilities and Modeling. Ind. Eng.
Chem. Res. 46 (2007), 6572-6583.
[6] Gao, C., Yan, D.: Hyperbranched polymers: from synthesis to applications. Prog.
Polym. Sci. 29 (2004), 183-275.
[7] Langenbach, K., Enders, S.: Development of an EOS based on lattice cluster theory
for pure components. Fluid Phase Equilibr. 331 (2012), 58-79.
[8] Langenbach, K., Enders, S., Browarzik, C., Browarzik, D.: Calculation of the high
pressure phase equilibrium in hyperbranched polymer systems with the lattice-cluster
theory. J. Chem. Thermodynamics 59 (2013), 107-113.
210
[9] Langenbach, K., Browarzik, D., Sailer, J., Enders, S.: New formulation of the lattice
cluster theory equation of state for multi-component systems. Fluid Phase Equilibr.
362 (2014), 196-212.
[10] Nemirovsky, A.M., Dudowicz, J., Freed., K.F.: Dense self-interacting lattice trees with
specified topologies: From light to dense branching. Phys. Rev. A 45 (1992), 7111-
7127.
[11] Dudowicz, J., Freed, K.F.: Effect of Monomer Structure and Compressibility on the
Properties of Multicomponent Polymer Blends and Solutions: 1. Lattice Cluster
Theory of Compressible Systems. Macromolecules 24 (1991), 5076-5095.
[12] Sanchez, I.C., Lacombe, R.H.: An elementary molecular theory of classical
fluids. Pure fluids. J. Phys. Org. Chem. 80 (1976), 2352-2362.
[13] Duschek, W., Kleinrahm, R., Wagner, W.: Measurement and correlation of the
(pressure, density, temperature) relation of carbon dioxide. II. Saturated-liquid and
saturated-vapour densities and the vapour pressure along the entire coexistence curve.
J. Chem. Thermodynamics 22 (1990), 841-864.
[14] Kozlowska, M.K., Jürgens, B.F., Schacht, C.S., Gross, J., de Loos, T.W.: Phase
Behavior of Hyperbranched Polymer Systems: Experiments and Application of the
Perturbed-Chain Polar SAFT Equation of State. J. Phys. Chem. B 113 (2009), 1022-
1029.
[15] Portela, V.M., Straver, E.J.M., de Loos Th.W.: High-Pressure Phase Behavior of the
System Propane−Boltorn H3200. J. Chem. Eng. Data 54 (2009), 2593-2598.
211
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
DETERMINATION OF RHEOLOGICAL DATA FROM MIXING EXPERIMENTS
F. RIEGER; J. MORAVEC
Czech Technical University in Prague, Faculty of Mechanical
Engineering, Technická 4, 166 07 Prague 6, Czech Republic
Abstract. Sludge suspensions and highly concentrated fine-grained suspensions
behave mostly like non-Newtonian power law liquids. When designing mixing
equipment for suspensions of this kind, the Metzner - Otto concept is often used
for calculating the power consumption required for mixing. In this paper, the
concept is used for describing the flow behaviour of several types of sludge.
Rheological experiments were carried out in a vessel mixed using a curved blade
turbine or alternatively an anchor agitator. The sludge exhibited strongly non-
Newtonian behaviour corresponding to the power-law model with the flow index
in the range from 0.07 to 0.22 and the coefficient of consistency in the range from
2 to 84 Pa.sm
. The Metzner - Otto constant of the anchor agitator obtained from
the literature has been validated for suspensions having such a low flow index.
Our work has also proved that sludges can be substituted by model sawdust
suspensions, thus facilitating experimental work.
Keywords. Mixing, suspension, rheology, power law, Metzner - Otto, sludge.
INTRODUCTION
The way of determination of the rheological properties of suspensions is not as easy as
in case of pure liquids. Classical rotational rheometry experimental methods can be used for
suspensions with fine-grained particles having a low settling velocity. However, the particles
are very often larger and their density is higher, which causes many problems in experimental
flow property measurements. Non-standard methods are used in these cases.
212
This paper is focused on determining the flow properties of real sludge, and on
replacing these suspensions by model suspensions with similar flow properties. The reason
for carrying out the work was obtained from demands of cooperating company producing
mixing equipment. Recently, the company recorded a growth in demands of equipment for
mixing of various types of sludges, which should be used for further treatment in
biotechnologies. The company produces own types of impellers and needed to determine
experimentally process characteristics of the impellers when used in such sludge suspensions.
However, the experimental work with sludge suspension is not convenient. Thus a
requirement on finding a model suspension having similar flow properties and facilitating the
experimental work arose.
Sludge suspensions contain a large number of fine-grained particles, and also bigger
particles or fibrous material. This causes problems when making measurements using
classical rheometry. It is proven below that the flow behaviour of suspensions of this kind is
non-Newtonian and can be described in most cases by the Ostwald - de Waele rheological
model
mK . (1)
In this paper, the parameters of the model were obtained using a non-standard method
of rheological measurements – in a vessel mixed by a curved blade turbine or, alternatively,
by an anchor agitator.
THEORETICAL BACKGROUND
Measurements of the rheological properties of pseudoplastic fluids are based on determining
the power consumption P that is required to mix the suspension at a specific impeller speed n.
The theory for determining the power consumption for mixing pseudoplastic power-law fluids
was published in [1], where it is shown that the power number
53
PoDn
P
(2)
is dependent on the flow index m and the Reynolds number
K
Dne
m
m
22
R
. (3)
In the region of creeping flow, the power number is inversely proportional to the Reynolds
number
m
mC
RePo . (4)
213
The function C(m) is thus described as
31DKn
PmC
m . (5)
This method, however, is not useful for practical measurements, because the power
characteristics of the impeller have to be measured for different values of the flow index. For
non-Newtonian fluids, the Metzner - Otto concept [2] is mostly used in practice. It uses the
power characteristics of impellers measured for pure Newtonian liquids, i.e. the relation
RePo f , (6)
where the Reynolds number is defined using the effective viscosity
ef
nD
2
Re . (7)
The effective viscosity ef is determined from a rheogram of the measured fluid as the
apparent viscosity at the effective shear rate ef . Metzner and Otto assumed that, for the given
geometrical layout of the vessel and impeller, the effective shear rate is proportional to the
impeller speed n, and it can be determined from the equation
kn . (8)
Constant k is dependent only on the type of impeller and on the geometrical layout of the
vessel-impeller system.
If we compare the Metzner - Otto concept with the general criterion equation of creeping flow
for a Newtonian liquid, we get the following: for a Newtonian liquid, the power characteristic
of an impeller is described as
2Re
PonD
AA
. (9)
The effective viscosity of a power-law fluid can be expressed from the expression of the
apparent viscosity, in which the shear rate is substituted from equation (8), thus
11 mm
efef knKK . (10)
When the viscosity from equation (10) is substituted for the viscosity in equation (9), we get
m
mmAk
nD
knAK
RePo
1
2
1
. (11)
From a comparison of this equation with relation (5), it follows
1 mAkmC . (12)
The Metzner - Otto concept is thus a special case of a general method for calculating the
power consumption of a power-law fluid, which assumes that function C(m) can be expressed
in the form of equation (12).
214
This theory is used in our case for measuring the rheological properties of sludge suspensions.
In the experiments, we measured the values of impeller rotations n and torque Mk used for
mixing. From these values, the power number was calculated using the equation
52
2Po
Dn
M k
. (13)
This power number was compared with the power number obtained from the power
characteristics of the impeller, measured in a Newtonian liquid of known viscosity and
density. From this comparison, the value of Reynolds number Re defined by equation (7) was
determined, and subsequently the effective viscosity value ef was obtained. This viscosity
was plotted in a graph against values of the effective shear rate calculated from equation (8).
EXPERIMENTAL AND DISCUSSION
Real sludge suspensions
Several different types of sludge were measured in the experiments. Each tested sludge
suspension was mixed in a cylindrical flat-bottomed vessel of diameter T = 100 mm by a
curved blade turbine (CVS 69 1027b) with diameter D = 69 mm (Figure 1). The height of the
suspension level in the vessel was equal to the vessel diameter (H = T).
Figure 1. Experimental layout for measuring the rheological properties of sludge.
215
The Metzner - Otto constant k = 11.3 was taken from [3] for this configuration. The value of
the constant was determined from data valid for power-law fluids with flow indexes in the
range from 0.2 to 1. A Rheotest R2 viscometer was used as a drive for mixing and for torque
and impeller speed measurement.
The rheograms of all the tested sludge samples obtained from the experiments have already
been presented in [4], but for easy reference they are shown again here in Figures 2 to 6, and
the flow properties are summarized in Table 1.
Table 1. Tested real sludge suspensions and their flow properties.
Sludge description (kg∙m3) K (Pa∙s
m) m (-)
Concentrated raw sludge (6.5% b.w., 15°C) 1015 13.5 0.18
Concentrated raw sludge (6.5% b.w., 70°C) 1015 4.9 0.21
Sludge (6.5% b.w., 70°C) after pasteurization 1015 4.4 0.22
Neutralization sludge 1105 2.0 0.11
Suspension of poultry droppings (18.3% b.w.) 1043 84.5 0.10
Sludge from a food processing line – sample 1B 1033 46.6 0.06
Sludge from a food processing line – sample 2B 1008 33.1 0.07
Sludge from a food processing line – sample 4B 1021 18.5 0.10
Figure 2. Rheograms of concentrated raw sludge (6.5% b.w.) from a wastewater treatment plant.
216
Figure 3. Rheogram of sludge from a wastewater treatment plant (6.5% b.w.) after pasteurization.
Figure 4. Rheogram of neutralization sludge.
Figure 5. Rheogram of a suspension of poultry droppings (18.3 % b.w.).
217
Figure 6. Rheograms of sludge suspension samples from a food processing line.
It can be seen that all the tested sludge suspensions exhibited strongly non-Newtonian
behaviour corresponding to the power-law model with flow index m in the range from 0.07 to
0.22 and coefficient of consistency K in the range from 2 to 84 Pasm
. In our work, it was
necessary to find a model suspension which would exhibit similar flow properties.
Model sludge suspensions
A suspension of sawdust was hypothesized to be a possible substitute for real sludge, from the
viewpoint of flow behaviour. The experiments with sawdust suspensions were carried out
using the same concept as in the case of real sludge suspensions. An RC20 viscometer was
used for measuring instead of the Rheotest R2, and the configuration of the vessel and the
impeller was also changed. The model suspensions were mixed in a flat-bottomed cylindrical
vessel of diameter T = 150 mm (H = T), using an anchor agitator of diameter D = 135 mm
(Figure 7).
The dimensions of the impeller corresponded to CVS 69 1014. The value of the Metzner -
Otto constant for this agitator was published in [1]. However, there is an ambiguous result at
this point. The source data used in that paper include only measurements with liquids having a
flow index in the range from 0.5 to 1. As is shown in Figure 8, describing the measurements,
there are two ways to express the Metzner - Otto constant from the data. The first way
corresponds to the results from other reviewed literature showing that k is constant and the
value is k = 16.2, according to [1]. In the second way, the source data are approximated by the
function
1
21.2
mmk (14)
218
which better fits the data. However, this would denote that k is not a constant but a function of
the flow index m. This ambiguity needed to be resolved before performing the measurements
with a model suspension using the anchor agitator.
Figure 7. Experimental layout for measuring the rheological properties of a model suspension.
Figure 8. The dependence of function C(m) on flow index m (n in the original paper) [1].
219
Water solutions of polyacrylamide with concentrations of 1.5 and 2.0% b.w. were used to
prepare a power-law fluid with a low flow index. The rheological properties of the solutions
were measured in vessels of diameters T = 100 mm and T = 150 mm mixed using an anchor
agitator. The experimental layout corresponded to Figure 7. The properties that were obtained
are listed in Table 2. The values were used to supplement the data in Figure 8, taken from [1].
Two graphs were made, corresponding to the two theories: the first, that k does not depend on
flow index m (Figure 9), and the second, that k is a function of m (Figure 10). As shown by
the results, the added points (the points at low values of the flow index) proved that k is
constant, because points in Figure 9 better fit to the line representing the theory of constant k,
than in case of Figure 10 in which the line corresponds to the theory of k as a function of m.
The line in Figure 9 is represents eq. (12) with parameters A = 180 and k = 15.5. Thus the old
value of the constant k = 16.2 in the original paper was corrected using the new points to
value k = 15.5. This value was then used to re-evaluate the flow properties of a sawdust
suspension in measurements of the model suspensions.
Two different types of sawdust produced by wood sanding of different roughnesses were used
in the experiments. The first sample was obtained from coarse sanding, and rose in the
suspension to the liquid level without mixing. The second sample was produced from fine
sanding, and the particles settled to the bottom of the vessel in the suspension. Rheograms of
two the suspensions, which differed in the concentration of coarse sawdust, are presented in
Figure 11. The flow curve of the fine sawdust suspension is then shown in Figure 12. Values
of the parameters of the power-law rheological model and also the densities of all the
suspensions measured by a pycnometer are listed in Table 3.
Figure 9. Dependence of function C(m) on flow index m updated with points of PAA (in semi-logarithmic
coordinates proving constant k – line according to eq. (12) with parameters A = 180, k = 15.5).
220
Figure 10. Dependence of function C(m) on flow index m actualized with points of PAA (in logarithmic
coordinates proving dependence of k = f(m) in form of eq. (14)).
Figure 11. Rheograms of a coarse sawdust suspension for two different concentrations of sawdust.
Figure 12. Rheograms of a fine sawdust suspension (9.5% b.w.) with dry and 48-hour-wetted sawdust.
221
Table 2. Flow properties of water solutions of polyacrylamide.
Solution description K (Pa∙sm
) m (-)
PAA – 1.5 % b.w. 6.1 0.33
PAA – 2.0 % b.w. 10.9 0.30
Table 3. The tested model sawdust suspensions and their flow properties.
Sludge description (kg∙m3) K (Pa∙s
m) m (-)
Suspension of sawdust from coarse sanding (6% b.w.) 1020 0.014 0.69
Suspension of sawdust from coarse sanding (9% b.w.) 1027 0.28 0.45
Suspension of sawdust from fine sanding (9.5% b.w.) 1028 6.18 0.12
Water suspension of sawdust from fine sanding after 48 hours
of sawdust wetting (9.5% b.w.) 1028 5.26 0.21
It can be seen that a fine sawdust suspension at a suitable concentration has similar flow
properties to those of a real sludge suspension, and may be used in place of sludge in a model
suspension in laboratory experiments.
CONCLUSIONS
Real sludge suspensions behave like non-Newtonian power-law fluids with a very low flow
index. The flow index of the suspensions tested in this work ranged from 0.07 to 0.22, and the
coefficient of consistency ranged from 2 to 84 Pa.sm
. The rheological properties of such a
suspension can be measured in a vessel mixed by an impeller using the Metzner - Otto
concept of power consumption determination. In the paper, the Metzner - Otto constant k of
an anchor agitator was corrected to be 15.5, adjusting the information published in [1]. It has
also been proved that suspensions of sawdust, obtained from fine sanding, have a flow
behaviour similar to that of a real sludge suspension, and can be used advantageously as
model suspensions in laboratory experiments.
ACKNOWLEDGEMENT
This work has been supported by research project (TA02010243, Mixing Equipment for
Sludge Processing) of the Technology Agency of the Czech Republic.
222
LIST OF SYMBOLS
A constant in equation (9) (m)
C impeller off-bottom clearance (m)
C(m) function in equation (4) (m)
D impeller diameter (m)
h width of the impeller blade (m)
H liquid level height (m)
k the Metzner - Otto constant (-)
K coefficient of consistency (Pa∙sm
)
m flow index (-)
n impeller speed (s-1
)
P power consumption (W)
Po power number (-)
Re Reynolds number (-)
Rem Reynolds number for a power law fluid (-)
T vessel diameter (m)
shear rate (s-1
)
ef effective shear rate (s-1
)
dynamic viscosity (Pas)
ef effective dynamic viscosity (Pas)
density (kgm-3
)
shear stress (Pa)
REFERENCES
[1] Rieger F., Novák V.: Power consumption of agitators in highly viscous non-
Newtonian liquids. Trans. Instn. Chem. Engrs. 51 (1973), p. 105-111.
[2] Metzner A. B., Otto R. E.: Agitation of non-Newtonian fluids. AICHE J. 3 (1957), p.
3-10.
[3] Novák V., Rieger F.: Výpočet příkonu rychloběžných míchadel při míchání
pseudoplastických kapalin. Chem. prům. 29 (1979), p. 285-290.
[4] Moravec J., Rieger F., Jirout T., Pešl L.: Model suspension for measurements of
process characteristics of impellers used for mixing of sludge suspensions. In: XXI
Ogólnopolska Konferencja Inżynierii Chemicznej i Procesowej. Kołobrzeg (2013).
223
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
IMPROVED DESORPTION CONTROL VIA RAMAN SPECTROSCOPY
E. ESCHE (1); B. KRAEMER (1); D. MÜLLER (1); K. MEYER (2); N. ZIENTEK (2); M.
MAIWALD (2); G. WOZNY (1)
(1) Technische Universität Berlin, Chair of Process Dynamics and
Operations, Sekr. KWT-9, Straße des 17. Juni 135, D-10623 Berlin,
Germany
(2) BAM Federal Institute for Materials Research and Testing, Richard-
Willstätter-Straße 11, D-12489 Berlin, Germany
Abstract. In this contribution a Raman spectrometer based control structure for
the heating of a desorption column is proposed. For this purpose calibration
experiments for the absorption of carbon dioxide using monoethanolamine
solutions are carried out and calibration models are developed to measure both
carbon dioxide liquid loads and monoethanolamine mass fractions. The
calibration experiments are supported by online NMR spectroscopy to accurately
measure the appearance of all species in the electrolyte system. Both models are
tested during the plant operation of a mini-plant for the oxidative coupling of
methane and the proof of concept for the control structure is given. The Raman
spectroscopy implemented in the ATEX conform mini-plant shows a reliable and
robust performance being even indifferent to impurities hindering the GC
analysis.
Keywords. Absorption, Carbon Capture, Raman spectroscopy, desorption control
MOTIVATION AND INTRODUCTION
Given the high interest in separating carbon dioxide (CO2) from industrial product
streams and exhaust gases to limit greenhouse gas emissions, the amine-based absorption and
subsequent desorption has become the center of attention for many research groups. Most
groups focus on designing structural improvements to the process via additional or modified
224
equipment or on the application of new and structurally complex scrubbing liquids [1]. At the
same time, however, the control and operation of the absorption systems remain a challenge,
which shall be tackled here.
Naturally the heating of the desorption unit is the largest energy sink in the whole
process concept. This is especially true for the absorption/desorption of CO2 from exhaust
gases, as these are usually not compressed in any meaningful way. Thus, no pressure swing
between absorption and desorption can be exploited. Up to now, there are three fundamentally
different approaches to control the heating of the desorption column. The first approach is the
direct temperature control, the second is the quality control of the carbon dioxide
concentration in the desorption gas outlet, and the third is via the gas outlet concentration of
absorption. All three have a couple of downsides, which basically are caused by dead time
between measurements and control action. For a meaningful temperature measurement the
composition of the scrubbing liquid needs to be known, and the propagation of changing gas
compositions to the desorption bottom can take up to several hours.
Hence, in this contribution, we propose a Raman spectroscopy-based control of the
desorption bottom.
PROPOSED CONTROL STRUCTURE
The oxidative coupling of methane (OCM) mini-plant at Technische Universität Berlin
features, among others, an absorption desorption process to remove carbon dioxide (CO2)
from the outlet gas stream. The mini-plant consists of an absorption column operated at 5 to
32 bar, a flash, and an electrically heated desorption column. As a base-case scenario the
mini-plant is operated with a 30wt.-% monoethanolamine (MEA) solution to remove 90% of
the carbon dioxide from the OCM product gas. A sketch of the flow sheet of the mini-plant is
shown in Figure 1. Up to now the liquid cycle of the process is evaluated by taking samples
on an hourly basis and performing titrations and GC analysis ex situ to determine the carbon
dioxide load and the MEA concentration. The required heating duty to desorb the MEA
solution is determined iteratively by continuously measuring the gas outlet concentration of
the absorption column and adjusting the heating duty accordingly.
There are two issues with the current control structure. First of all, there is a considerable
dead-time between the desorption bottom and the liquid feed to the absorption, which makes
it difficult to react to minor fluctuations in the gas composition. Secondly, the iterative
process is slow and prone to errors given that it might take up to two hours to reach a new
steady-state, which is required to properly evaluate the current operation point. At the same
225
time both MEA and water are partially lost in the gas outlets, changing the performance of the
whole process. Hence, a new control structure is proposed in this contribution.
A Raman spectrometer probe is set into the outlet stream of the desorption column (QIC II in
Figure 1). The inflow measurement allows for the online determination of both the amine
concentration in the aqueous solution and the carbon dioxide load [2].
Figure 1. Sketch of the absorption desorption process and the Raman-based desorption control.
Thus, a PID controller can be implemented in the process control system to control the outlet
CO2 load of the absorbent flow after the desorption column to a maximally desired value. The
actual value will depend on the application, meaning the partial pressure of carbon dioxide in
the feed gas and the purity required for the gas outlet flow.
The gas outlet measurement of the absorption column can in turn be used to control the liquid
feed to the absorption column (QIC I in Figure 1). This way, fluctuations in the feed gas can
be quickly compensated. Of course, the window for the compensation is relatively small
given, that the absorption column is usually operated close to the flooding point.
In addition, the Raman measurement can be used to estimate the loss of water and MEA. As
has been mentioned before, there is a certain, inestimable loss with the gas outlets of the
absorption column, the flash, and the desorption column and over time both water and MEA
amounts in the entire plant need to be replenished. Ideally, of course, this will be done
continuously to minimize the offset to the desired MEA concentration in the process. With the
proposed control structure this can directly be implemented as part of QIC II.
226
CALIBRATION OF THE RAMAN-SPECTROMETER
In this section the preparation of the Raman spectroscopy for the described application is
outlined. Before presenting details on the calibration test bed, the basics of Raman and NMR
spectroscopy will be revisited briefly with regards to MEA solutions and carbon dioxide.
Raman-Spectroscopy
According to [3] nine most important species are observed by the chemical absorption of
carbon dioxide in MEA solutions: MEA, CO2, H2O, MEACOO-, H3O
+, OH
-, HCO3
-, CO3
2-,
and MEAH+. Several functional groups appearing within these species show raman-active
modes. The vibrational bands of various groups have been mapped quite extensively. [4]
contains a conclusive list of organic groups. For the species listed above characteristic
frequencies at wave numbers of 400 to 3,400 cm-1
can be expected. The intensity of each
oscillation is directly proportional to the amount of the species containing the respective
functional group and can hence be used for a concentration measurement. However, a
chemometrical model is proposed in this work, which allows for a multivariate
(multiparameter) calibration based on the whole spectroscopic data set. In this contribution a
Raman spectrometer of type RXN1 of Kaiser Optical Systems inc. with an InvictusTM
-Laser
(785 nm) is applied. The inflow probe has a diameter of half an inch and can be used at up to
450 °C and 220 bar.
NMR Spectroscopy
To calibrate the Raman spectroscopy and to support gravimetric data, nuclear magnetic
resonance (NMR) spectroscopy is used in this contribution. Nuclei possessing magnetic spins
appear within NMR spectra [5] and the peak area directly correlates with the number of nuclei
appearing in the sample. Once the peaks appearing in the spectra and their stoichiometric
properties are identified, NMR spectroscopy represents an absolute comparison method for
amount of substance quantification and hence a calibration free method. In accordance with
[6] 1H and
13C spectra will be taken to determine MEA, MEACOO
-, H2O, and HCO3
-. The
NMR spectrometer of Varian (Agilent Technologies inc.) is operating at 500 MHz proton
frequency and uses an interchangeable flow cell for flow-through measurements up to 130°C
and 30 bar. [10] NMR spectroscopy was successfully applied to determine the exact species
distribution under operation condition before [11, 12]. At present, this information was not
used for the proposed chemometrical models presented here. Species distributions and
formation of by-products from amine degradation are of importance for future work.
227
Calibration test bed
The calibration test bed set-up is sketched in Figure 2. The MEA solution is sucked via a pre-
filled line into the autoclave from the scales. The autoclave itself can be operated at up to
30bar and has a jacket for heating or cooling using a liquid thermostat. The total volume is
roughly one liter, which is magnetically stirred. For sampling an HPLC pump supplies liquid
from the autoclave over the Raman probe and through the NMR flow probe. Helium feed and
vacuum pump have been added to fully flush and inertize the whole system. In addition, the
vacuum supports the liquid filling process. The Pt100 after the Raman probe is the
temperature measurement for the thermostat controller. To minimize the heat loss of the pipe
going to the NMR, it is inserted into the heating hose of the thermostat. The CO2 injection is
carried out via sample cylinders, which in turn are weighted before and after injection.
Figure 2. P&ID of the calibration test bed.
Calibration Experiments
The calibration has two primary goals. First of all, a calibration model is to be developed,
which returns the carbon dioxide load of a liquid sample in mol CO2 per mol MEA. Secondly,
the mass fraction of MEA of the whole liquid sample is of interest. Hence, experiments are
carried out in the calibration test bed, varying the MEA concentration from 10 to 50wt.-% and
at 25 to 35wt.-% MEA successively loading the solution starting at 0 to up to 0.7 mol
CO2/mol MEA. The experiments are performed with a liquid flow rate through the HPLC
pump of 8mL/min, a tenth of which passes the NMR flow cell. The temperature is set to
40°C, which is controlled by the thermostat. For safety reasons the laser output should never
surpass 150mW in the mini-plant. Hence, the same is enforced during the calibration
Scales
Vacuum
Thermostat
Raman
probe
Low field NMR
228
experiments. Measuring each spectrum with 4 accumulations over 15s exposure time. Each
measurement is repeated 4 times over a period of 8min.
In addition, experiments were carried out to investigate the influence of ethene (C2H4)
appearing in the liquid as a dissolved gas. Ethene is the main product of the OCM reaction
and has a certain, undesired solubility in amine-solutions [7, primary data in 8]. The ethene is
found to have no consequence on the Raman spectra in the range of interest for the
calibration.
The temperature influence on the Raman spectra was also investigated. While there is of
course a clear correlation between the temperature and the equilibrium load of CO2 in the
amine solution, the temperature appears to show little difference in the Raman spectra at a
constant CO2 load, which correlates with previously published data [2].
Calibration Model – CO2 Load
Several different calibration models are developed differing in the pre-treatment of the
calibration data. At this point only one of them will be discussed.
The spectral data is reduced to a range from 400 to 1,520 cm-1
as the most significant peak
changes appear therein. The spectra are normalized with respect to a peak at 418cm-1
, which
is inherent to the Raman spectrometer and independent of the actual application. This reduces
the differences in intensity. The spectra are numerically differentiated using the Savitzky-
Golay differentiation [9] based on 31 data points for each differential.
The actual model is then developed based on PLS regression in accordance with [9]. The
resulting model shows a root mean square error of cross-validation (RMSECV) of 0.0269 mol
CO2/mol MEA. Figure 3 shows the parity plot comparing predicted and measured (NMR)
carbon dioxide loads.
229
Figure 3. Parity plot for the carbon dioxide load calibration model.
Calibration Model – Amine Mass Fraction
The calibration model for the MEA mass fraction on the other hand is based on the spectral
range from 2,804 to 3,136 cm-1
. In this range, there seems to be no significant difference
between loaded and unloaded amine solution and a steady intensity increase for increasing
MEA mass fractions. Here, the Savitzky-Golay differentiation is first of all carried out based
on 21 data points and afterwards range normalization is performed. The PLS-based model
shows an RMSECV of 0.0099g MEA/g solution and the according parity plot is shown in
Figure 4.
230
Figure 4. Chemometrical model prediction for the MEA mass fraction calibration model.
The MEA mass fraction model is additionally validated against the calibration data for the
CO2 liquid load. At almost every single measurement point an underestimation of the actual
MEA content is observed. Thereof it is quite obvious that a more comprehensive set of data
for the MEA mass fraction calibration is required including CO2 loading.
CASE STUDY: OCM MINI-PLANT
In order to test the calibration and to proof the feasibility of the proposed control structure the
Raman spectrometer is implemented in the OCM mini-plant. To cover a large range of
different CO2 liquid loads and MEA mass fractions, the Raman probe is inserted at the liquid
outlet of the absorption column (A), of the flash (F), and of the desorption column (D)
respectively. The latter is the position at which the Raman probe would also be positioned for
the actual implementation of the new control. At all three positions liquid samples are taken
on an hourly basis to measure CO2 load and MEA mass fractions offline. The gas feed to the
absorption column contains methane (CH4), ethene (C2H4), carbon dioxide (CO2), and
nitrogen (N2). The feed pressure of carbon dioxide, the liquid cycle flow, as well as the heat
of desorption are varied and samples are taken continuously for more than 140 hours of
operation time.
231
Robustness of Raman Spectroscopy
Whilst both titration and GC analyses are prone to user errors during sampling, sample
preparation, and examination in titrator and GC respectively, the Raman spectroscopy shows
a robust and reliable performance throughout the entire plant operation. Even for cases at
which impurities and overloading in the liquid samples prevent reliable GC results, the
Raman spectra stay consistent and evaluable. Fig. 5 shows the proof of concept for the new
control structure for the desorption heating duty. The increase in the heating duty (red line)
directly correlates with decrease in the CO2 liquid load (black line).
Figure 5. Raman-based CO2 liquid load for changes in the desorption heating duty. The desorption
heating duty is given in percent of the maximum duty possible of 30kW electrical heating supplied to the
bottom of the desorption column. The black line shows the result of the model called B at measurement
positions flash outlet (F) and desorption outlet (D). The discontinuity shows the switch of the Raman
spectrometer from flash to desorption.
All other process parameters are held constant at the same time. In contrast CO2 liquid load at
flash and desorption outlet measured offline show an erratic behavior (red triangles and blue
diamonds) which are mostly caused by failures of the GC analysis to measure the MEA mass
fraction required for the proper evaluation of CO2 load titrations.
CO
2 li
qu
id lo
ad [
mo
l CO
2/m
ol M
EA
]
Des
orp
tio
n h
eati
ng
du
ty [
%]
232
Calibration Model Evaluation
Given preparation and other errors of the offline analyses, no proper validation of the Raman-
based results for the mini-plant operation is possible. Based on the calibration data, the CO2
load model shows a maximum variance of 0.091 mol CO2/mol MEA and the MEA mass
fraction model a variance of 0.05 g MEA/g solution.
CONCLUSIONS AND OUTLOOK
The calibration models for CO2 load and MEA mass fraction presented in the contribution are
robust and facilitate the proposed control structure for the absorption desorption process. It is
shown that there is a direct correlation between desorption heating and the measurement data
received from the respective outlet stream with a short time constant. Hence, the requirements
for a successful implementation of the desorption control is fulfilled. Also, it is feasible to
measure the MEA mass fractions at the same time and use the data to compensate for liquid
losses.
To further validate the derived models further analysis of mini-plant samples, e.g. using NMR
offline measurements, will be carried out next. In addition, further calibration experiments
will be performed to broaden the basis for the MEA mass fraction model. As a next step, the
implementation of the whole control structure based on the Raman spectrometer is envisioned
and an extension to other amine-systems, such as Methyldiethanolamine (MDEA), is desired.
ACKNOWLEDGEMENTS
This work is part of the Cluster of Excellence “Unifying Concepts in Catalysis” coordinated
by the Berlin Institute of Technology (Technische Universität Berlin). Financial support by
the German Research Foundation (DFG) within the framework of the German Initiative for
Excellence is gratefully acknowledged.
REFERENCES
[1] Yeh, J.T.; Pennline, H.W.; Resnik, K.P.: Study of CO2 Absorption and Desorption in
a Packed Column. Energy Fuels 2001. 15 (2), 274-278, DOI: 10.1021/ef0002389
[2] Vogt, M.; Pasel, C.; Bathen, D.: Characterisation of CO2 absorption in various
solvents for PCC applications by Raman spectroscopy. Energy Procedia 2011. 4,
1520-1525, DOI: 10.1016/j.egypro.2011.02.020.
233
[3] Freguia, S.; Rochelle, G.T.: Modeling of CO2 capture by aqueous monoethanolamine.
AIChE Journal 49, 2003, 7, 1676-1686.
[4] Lin-Vien, D.: The Handbook of infrared and raman characteristic frequencies of
organic molecules. Boston: Academic Press, 1991.
[5] Schwedt, G.: Analytische Chemie: Grundlagen, Methoden und Praxis. Weinheim:
Wiley-VCH, 2008, 2.
[6] Böttinger, W.: Fortschritt-Berichte / 03 / VDI. Bd. 851: NMR-spektroskopische
Untersuchung der Reaktivabsorption von Kohlendioxid in wässrigen Aminlösungen.
Düsseldorf: VDI-Verl., 2006.
[7] Sada, E.; Kmazawa, H.; Butt, M.A.: Solubilities of Gases in Aqueous Solutions of
Amine. Journal of Chemical Engineering Data, Vol. 22, No. 3, 1977, 277-278.
[8] Sada, E.; Kito, S.: Kagaku Kogaku, 36, 1972, 218.
[9] Kessler, W.: Multivariate Datenanalyse: Für die Pharma-, Bio- und Prozessanalytik;
ein Lehrbuch. Weinheim. Wiley-VCH, 2007.
[10] Maiwald, M.; Fischer, H. H.; Kim Y.-K.; Ott, M.; Albert, K.; Hasse, H.: Quantitative
High-resolution On-line NMR Spectroscopy in Reaction and Process Monitoring, J.
Magn. Reson. 166 (2004) 135–146
[11] Böttinger, W.; Hasse, H.; Maiwald, M.: Online NMR Spectroscopic Study of Species
Distribution in MDEA-N2O-CO2 and MDEA-PIP-H2O-CO2, Ind. Eng. Chem. Res.
47 (2008) 7917–7926
[12] Böttinger, W.; Hasse, H.; Maiwald, M.: Online NMR Spectroscopic Study of Species
Distribution in MEA-N2O-CO2 and MEA-H2O-CO2, Fluid Phase Equilib. 263 (2008)
131–143
235
20th International Conference
Process Engineering and Chemical Plant Design
October 15-17, 2014, Berlin, Germany
COMPARISON OF DIFFERENT MEMBRANES FOR THE REMOVAL OF
SURFACTANTS FROM ORGANIC SOLVENTS BY ORGANIC SOLVENT
NANOFILTRATION
D. ZEDEL (1,2); A. DREWS (1); M. KRAUME (2)
(1) Hochschule für Technik und Wirtschaft Berlin, FB II, Process
Engineering in Life Science Engineering, Wilhelminenhofstraße 75A,
12459 Berlin, Germany, [email protected]
(2) Technische Universität Berlin, Berlin, Chair of Chemical and Process
Engineering, Straße des 17. Juni 136, 10623 Berlin, Germany
Abstract. Organic solvent nanofiltration proves to be a promising, energy
efficient separation technology for the removal of solutes from organic solvents.
In this study, the membrane performance characteristics of three different PDMS
based membranes were analysed and compared for the application in surfactant
separation from organic solvents. It could be shown that an increase in thickness
of the dense separation layer decreases the pure solvent fluxes at different
temperatures and that membrane compaction decreases the slope of flux increase
with increasing pressure. All membranes showed considerable surfactant retention
of up to 70% and fluxes of 25L/(m²h) with the highest retentions for membranes
with thicker active layers.
Keywords. Organic solvent nanofiltration, 1-Dodecene, Marlipal 24/70,
surfactants, membrane comparison, PDMS, membrane compaction
INTRODUCTION
Surfactants can be utilised in various chemical reactions as solution mediators to
overcome the miscibility gap between aqueous and organic solutions by creating
microemulsions [1]. Some of the potential reactions include chemical reduction and oxidation
236
reactions, as well as C-C coupling reactions [1]. Example reactions for the refinement of long-
chain olefins are the hydroformylation of the higher alkenes 1-octene [2] and 1-dodecene [3].
Although the solubilisation by surfactants speeds up the chemical reaction, it makes the phase
separation as well as the purification of the product more difficult. When compared to a
process that only utilises a catalyst in an aqueous phase (without addition of surfactant),
additional process steps have to be implemented to remove the surfactant after the reaction.
Membrane processes for industrial applications in organic solvents separation (organic
solvent nanofiltration, OSN) make separations down to a molecular level possible [4].
Polymeric membranes made from polydimethylsiloxane (PDMS) or polyimide (PI) and
ceramic membranes are two basic types typically used for OSN.
Among the many industrial applications of organic solvent nanofiltration, the separation of
surfactants from organic solvents has not been mentioned in the literature so far. The aim of
this study is to compare the membrane performance of three PDMS membranes at different
operation conditions (pressure, temperature, surfactant mass fraction).
MATERIALS AND METHODS
Mixtures of 1-dodecene (synthesis quality, acquired from Merck Darmstadt KGaA) as the
nonpolar solvent and the technical grade non-ionic surfactant Marlipal 24/70 (fatty alcohol
polyethylene glycol ether), as well as pure 1-dodecene were used as feeds in a magnetically
stirred high pressure test cell (GH400 by Berghof Membrane Technology GmbH & Co. KG
Germany, effective membrane area of 3.5 x 10-3
m2) in dead end filtration mode. Polymeric
thin film composite flat sheet nanofiltration membranes oNF-1 and oNF-2 (manufactured by
Borsig Membrane Technology GmbH) as well as 05/070 (obtained from Helmholtz-Zentrum
Geesthacht), all made from PDMS on polyacrylonitrile support, were tested at transmembrane
pressures between 1 and 35bar, at 20°C to 45°C and 3wt.-% initial surfactant feed
concentration. All experiments were conducted at a stirrer speed of 400 min-1
. Permeate
samples were taken at regular intervals and the permeate surfactant concentrations were
determined by high accuracy density measurements (DSA 5000 M by Anton Paar Germany
GmbH). Additionally membranes were analysed by scanning electron microscopy (SEM)
using a Hitachi S4000 SEM with cold cathode field emission electron source. The retention
value 𝑅 used in this work is calculated with Equation (1) with ��𝑃 as the surfactant weight
fraction in the complete permeate and 𝑤𝐹 as the surfactant weight fraction in the feed
solution.
𝑅 = 1 −��𝑃
𝑤𝐹 (1)
237
RESULTS
All tested membranes proved to be able to separate the surfactant Marlipal 24/70 from the
organic solvent 1-dodecene with all tested membranes, though the membrane performance is
significantly different. Figure 1 shows the permeate flux of pure 1-dodecene against the
transmembrane pressure for the membranes oNF-1, 05/070 at 20°C and 45°C and oNF-2 at
20°C. For all membranes, an increase in transmembrane pressure results in a flux increase
with a decreasing slope. It can be seen that oNF-2 shows the highest flux at any given TMP,
followed by oNF-1 and 05/070. As shown previously [5] and [6], [7], this non-linear flux
behaviour at increasing TMP could be explained by polymer swelling in 1-dodecene and
subsequent pressure dependent compaction.
Figure 1: Pressure influence on pure 1-dodecene flux, lines from model equation [5] for better visual
comparison.
The order of the fluxes can be explained, when considering the membrane morphology, for
example the thickness of the dense PDMS layer. New oNF-1 membranes were analysed using
SEM and the results are displayed in Figure 2. It is apparent that the membrane oNF-2 (which
had the highest flux) has the thinnest dense layer of all the tested membranes, followed by
oNF-1 and 50/070. The membrane producer, however, states that oNF-1 has a higher
permeability than oNF-2 for three pure solvent test systems (toluene, hexane and 2-propanol)
at 30bar and 30°C. It is assumed that the flux difference between the membranes can at least
partly be attributed to the difference in dense layer thickness. The fluxes for oNF-2 at 35bar
are 1.3 times higher than the fluxes of oNF-1 and 4.5 times higher than the fluxes of 05/070.
0 5 10 15 20 25 30 350
10
20
30
40
50
60
p in bar
Flu
x in
L/(
m²
h)
20°C 45°C
oNF-1
05/070
oNF-2
238
The dense layer thickness for oNF-2, however, is half that of oNF-1 and a quarter of that of
05/070. The factors do not exactly match, which might be attributed to a membrane
compaction effect. The membrane compaction effect can be seen in Figure 3, where SEM
images of both, new and used oNF-1 membranes are compared. It can be seen, that the dense
layer thickness is halved at a maximum transmembrane pressure difference of 35bar. Since
the membranes had to be depressurised in order to prepare them for electron microscopy, it
can be assumed, that the pressure influence is at least partly irreversible. Information about
the additional reversible compaction under pressure cannot be quantified using that method.
It was seen, that the membrane compaction effect reduces the flux, while a reduced
uncompressed dense layer thickness at the beginning increases flux.
The apparent might be solved, when considering that pressure induced compaction
compresses the complete dense layer structure and increases the transport resistance, while the
dense layer at the dry, uncompressed state might have a similar transport resistance per unit of
length for every membrane.
It is also evident from Figure 1 that at a higher temperature of 45°C the fluxes of oNF-1 and
05/070 are higher than at 20°C. Both membranes show a similar flux increase for increased
temperature of a factor of 1.6 to 1.7. This flux increase can be contributed to a decrease in
solvent viscosity by a similar factor of 1.5 from 1.32mPa s to 0.88mPa s. This behaviour was
further verified and explained by [8] and shown for oNF-1 in [5].
Figure 2: Dense layer thickness for different membrane samples, determined with SEM.
239
Figure 3: Dense layer thickness comparison of new and of pressurised oNF-1 membranes, determined
with SEM
Preliminary experimental results for surfactant mixtures of 3wt.-% Marlipal 24/70 in 1-
dodecene at room temperature and a transmembrane pressure of 20bar show that the
membrane with the thinnest dense layer (oNF-2) has the lowest surfactant retention of 49%,
while oNF-1 with a thicker dense layer has a higher retention of 63%. Since 05/070 showed
by far the lowest fluxes, a higher pressure was used in the surfactant retention test. At 35bar, a
surfactant retention of 70% was measured for 05/070, compared to a value of 62% for oNF-1
at this pressure level. So the membrane with the thickest dense layer shows the highest
retention value. It can be assumed that the thickness of the dense layer only plays a partial
role in determining the surfactant retention. Apart from the thickness, surfactant adsorption to
the membrane, as well as a different structure of the dense layer could influence the surfactant
retention behaviour. It was seen before that the flux difference between the membranes in
Figure 1 at constant temperature cannot be completely attributed to the difference in dense
layer thickness and no detailed information is available on the membrane composition and
dense layer “recipe” of the commercially available oNF-1 and oNF-2 membranes, as well as
on the method of cross-linking.
CONCLUSIONS
The organic solvent nanofiltration membrane performance of three different PDMS based
membranes was compared here for the first time for the separation of surfactants from organic
240
solvents. It could be seen that the pure solvent fluxes could be partly correlated with the dry
state dense layer thickness of the tested membranes. Additionally, an irreversible membrane
compaction caused by the transmembrane pressure difference could be observed with SEM.
This membrane compaction is assumed to cause the disproportionate increase of the fluxes
with pressure (decreasing slope behaviour at increasing pressures). The highest retention and
the lowest flux was seen for the membrane with the thickest dense layer.
REFERENCES
[1] Torsten Dwars, Eckhard Paetzold, and Günther Oehme. Reaktionen in micellaren
systemen. Angewandte Chemie, 117(44):7338–7364, 2005.
[2] Hesna Ünveren and Reinhard Schomäcker. Rhodium catalyzed hydroformylation of 1-
octene in microemulsion: comparison with various catalytic systems. Catalysis Letters,
110:195–201, 2006.
[3] Marco Haumann, Herbert Koch, Peter Hugo, and Reinhard Schomäcker.
Hydroformylation of 1-dodecene using rh-tppts in a microemulsion. Applied Catalysis
A: General, 225(1–2):239 – 249, 2002.
[4] Pieter Vandezande, Lieven E. M. Gevers, and Ivo F. J. Vankelecom. Solvent resistant
nanofiltration: separating on a molecular level. Chemical Society reviews, 37(2):365,
2008.
[5] Daniel Zedel, Anja Drews, and Matthias Kraume. Feasibility of surfactant removal
from organic solvents by organic solvent nanofiltration. Journal of Membrane Science.
(in preparation)
[6] Darıo R. Machado, David Hasson, and Raphael Semiat. Effect of solvent properties on
permeate flow through nanofiltration membranes. part i: investigation of parameters
affecting solvent flux. Journal of Membrane Science, 163(1):93 – 102, 1999.
[7] Loïc Leitner, Christelle Harscoat–Schiavo, and Cécile Vallières. Experimental
contribution to the understanding of transport through polydimethylsiloxane
nanofiltration membranes: Influence of swelling, compaction and solvent on
permeation properties. Polymer Testing, 33(0):88 – 96, 2014.
[8] Tadashi Uragami, Tadanori Yono, and Mizuho Sugihara. Studies on syntheses and
permeabilities of special polymer membranes. xx. permeabilities of alcohols and
hydrocarbons through acrylonitrile-butadiene-styrene terpolymer membranes. Die
Angewandte Makromolekulare Chemie, 82(1):89–102, 1979.
241
ACKNOWLEDGEMENTS
Financial support by the German Research Foundation DFG (collaborative research centre
"Integrated Chemical Processes in Liquid Multiphase Systems" InPROMPT TRR63) is
gratefully acknowledged. We thank the Helmholtz-Zentrum Geesthacht for kindly providing
free membrane samples.
243
LIST OF AUTHORS
Bałdyga, J. Warsaw University of Technology
Faculty of Chemical and Process Engineering
69
Barz, T. Technische Universität Berlin
Chair of Process Dynamics and Operations
15, 25
Ditl, P. Czech Technical University in Prague
Faculty of Mechanical Engineering
Department of Process Engineering
107
Drews, A. Hochschule für Technik und Wirtschaft Berlin
Process Engineering in Life Science Engineering
193, 235
Dyląg, M. The Institute of Advanced Manufacturing Technology
37, 85
Enders, S. Technische Universität Berlin
Chair of Thermodynamics and Separation Processes
199
Esche, E. Technische Universität Berlin
Chair of Process Dynamics and Operations
15, 25, 223
Fleck, A. Technische Universität Berlin
Chair of Process Dynamics and Operations
25
Gwadera, M. Cracow University of Technology
Faculty of Chemical Engineering and Technology
151
Hamerla, T. Technische Universität Berlin
Department for Chemistry
15, 47
Hohl, L. Technische Universität Berlin
Chair of Chemical & Process Engineering
97
Illner, M. Technische Universität Berlin
Chair of Process Dynamics and Operations
25
Janczewski, L. The Institute of Advanced Manufacturing Technology
85
Jasińska, M. Warsaw University of Technology
Faculty of Chemical and Process Engineering
69
Kamieński, J. Cracow University of Technology
Institute of Thermal and Process Engineering
Faculty of Mechanical Engineering
37
244
Kamp, J. Technische Universität Berlin
Chair of Chemical & Process Engineering
119
Kantor, R. Cracow University of Technology
165
Kim, S.-J. Technische Universität Berlin
Chair of Chemical & Process Engineering
141
Komorowicz, T. Cracow University of Technology
Faculty of Chemical Engineering and Technology
151
Kopiczak, B. Cracow University of Technology
Institute of Thermal and Process Engineering
127
Kowaliński, W. Warsaw University of Technology
Faculty of Chemical and Process Engineering
69
Kraemer, B. Technische Universität Berlin
Chair of Process Dynamics and Operations
223
Kraume, M. Technische Universität Berlin
Chair of Chemical & Process Engineering
97, 119, 141, 235
Kupiec, K. Cracow University of Technology
Faculty of Chemical Engineering and Technology
151
Larwa, B. Cracow University of Technology
Faculty of Chemical Engineering and Technology
151
Maiwald, M. BAM
Federal Institute for Materials Research and Testing
223
Matras, Z. Cracow University of Technology
Institute of Thermal and Process Engineering
127
Meyer, K. BAM
Federal Institute for Materials Research and Testing
223
Moravec, J. Czech Technical University in Prague
Faculty of Mechanical Engineering
213
Müller, D. Technische Universität Berlin
Chair of Process Dynamics and Operations
15, 25, 223
Nering, K. Cracow University of Technology
Institute of Thermal and Process Engineering
179
Paul, N. Technische Universität Berlin
Chair of Chemical & Process Engineering
97, 141
245
Pešava, V. Czech Technical University in Prague
Faculty of Mechanical Engineering
Department of Process Engineering
107
Pogrzeba, T. Technische Universität Berlin
Department for Chemistry
15, 49
Rieger, F. Czech Technical University in Prague
Faculty of Mechanical Engineering
213
Rieger, M. Thyssen Krupp Industrial Solutions
Coke Plant Technologies
3
Rosiński, J. Cracow University of Technology
Institute of Thermal and Process Engineering
Faculty of Mechanical Engineering
37
Rup, K. Cracow University of Technology
Institute of Thermal and Process Engineering
179
Schmidt, M. Technische Universität Berlin
Department for Chemistry
59
Schomäcker, R. Technische Universität Berlin
Department for Chemistry
15, 25, 49, 59
Schumacher, L. Hochschule für Technik und Wirtschaft Berlin
Process Engineering in Life Science Engineering
193
Schwarze, M. Technische Universität Berlin
Department for Chemistry
59
Skale, T. Hochschule für Technik und Wirtschaft Berlin
Process Engineering in Life Science Engineering
193
Skřivánek, J. Czech Technical University in Prague
Faculty of Mechanical Engineering
Department of Process Engineering
107
Speelmanns, E. Technische Universität Berlin
Chair of Process Dynamics and Operations
3
Szatko, W. Cracow University of Technology
Institute of Thermal and Process Engineering
Faculty of Mechanical Engineering
37
Talaga, J. Cracow University of Technology
Institute of Thermal and Process Engineering
Faculty of Mechanical Engineering
37
Villwock, J. Technische Universität Berlin
Chair of Chemical & Process Engineering
119
246
Walowski, C. Technische Universität Berlin
Chair of Thermodynamics and Separation Processes
199
Wójtowicz, R. Cracow University of Technology
Institute of Thermal and Process Engineering
Faculty of Mechanical Engineering
37
Wozny, G. Technische Universität Berlin
Chair of Process Dynamics and Operations
3, 15, 25, 223
Zedel, D. Hochschule für Technik und Wirtschaft Berlin
Process Engineering in Life Science Engineering
193, 235
Zientek, N. BAM
Federal Institute for Materials Research and Testing
223
Berlin 2014
ISBN 978-3-00-047364-7