Process Engineering and Chemical Plant Design 2011
Editors: Günter Wozny and Łukasz Hady
Universitätsverlag der TU Berlin Berlin 2011
Editors:
Günter Wozny and Łukasz Hady Fachgebiet Dynamik und Betrieb Technischer Anlagen Technische Universität Berlin Sekretariat KWT 9 Straße des 17. Juni 135 D-10623 Berlin http://www.dbta.tu-berlin.de
Umschlaggestaltung: Umschlagfoto:
Łukasz Hady Gasbehandlungsanlage zur Kokereigasentschwefelung, Uhde GmbH
ISBN 978-3-7983-2361-2 (Druckausgabe) ISBN 978-3-7983-2362-9 (Online-Version) Berlin 2011 * Gedruckt auf säurefreiem alterungsbeständigem Papier Druck/ Printing:
Endformat, Ges. für gute Druckerzeugnisse mbH Köpenicker Str. 187-188, 10997 Berlin
Vertrieb/ Publisher:
Universitätsverlag der TU Berlin Universitätsbibliothek Fasanenstr. 88 (im VOLKSWAGEN-Haus) D-10623 Berlin Tel.: (030)314-76131; Fax.: (030)314-76133 E-Mail: [emailprotected] http://www.univerlag.tu-berlin.de
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
PREAMBLE
The 18th International conference in “Process Engineering and Chemical Plant Design” is taking place in Berlin from september 19th to september 23rd 2011. We are pleased with the successful collaboration which is the result of a meanwhile 30 years continual international cooperation between the Cracow University of Technology and the Berlin Institute of Technology. This relationship has also been intensified by student exchange programs and international transfer of knowledge between the participating universities during the last years. This book contains the abstracts of all contributions and lectures which are presented by the miscellaneous participants within the scope of the conference. Different topics are addressed, concerning industrial problems as well as forward-looking questions and fundamental investigation of special phenomena for the chemical and the power generation industry. Thereby special attention is paid to fundamental research of complex correlations, modelling and simulation, process control and operation, sustainable and efficient energy generation as well as troubleshooting and problems within the operation and control of chemical processes. We want to appreciate all participants individually, especially our partners from Cracow University of Technology and also from the Warsaw and the West Pomeranian University of Technology. Special thanks go to the office for foreign relations (ABZ) and the DAAD for their financial support. All contributions have been peer-reviewed. Therefore we also want to thank the reviewers for their work. The authors are responsible for the contents of their articles.
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
BARBARA TAL-FIGIEL*
SOLID-LIQUID EXTRACTION FROM PLANTS WITH A BIDISPERSE POROUS STRUCTURE – EXPERIMENTAL KINETICS AND MODELLING Abstract This paper deals with the mathematical model of extraction from a capillary porous particle with bidisperse structure. Capillary porous particles with bidisperse structure, possessing capillaries of two, strongly different sizes, occur frequently in nature and technology. Neglecting the polydispersity of capillary sizes in capillary porous particles makes the obtained results less accurate and prevents elucidation of some physical mechanisms of substance transfer inside the particles. The bidisperse model of a capillary porous particle is analytically the simplest variant of the polydisperse model, which makes it possible to reveal fundamental aspects of mass transfer in real polydisperse particles. The mathematical planar model of extraction process from the particle can be described by the following set of equations: the diffusion equation for the porous body and the convective diffusion equation for the transport channel, with initial and boundary conditions for these equations, and the velocity profile in the transport channel. To solve the set of equations the grid and finite element methods were used. A numerical analysis of the model demonstrates that, for liquid extraction of a desired substance from the particle, there exists an optimum range of oscillation frequencies of the liquid in large pores. Results of numerical simulations are presented together with the criterial equation for calculating the effective diffusion coefficient, obtained in the course of processing these results. The experiments consisted of different types of liquid-solid extraction of active ingredients from plants. Keywords: solid-liquid extraction, bidisperse porous model.
*
Institute of Chemical and Process Engineering, Cracow University of Technology, Warszawska 24, 31-155 Kraków, Poland.
Solid-liquid extraction from plants with a bidisperse porous structure
3
1. Introduction Mathematical models of extraction processes, that are different from conventional diffusion models [1-4], were proposed recently. The reason for this is the need to make the description of the process more realistic, in particular, to improve the way in which the structure of a porous solid, from which the target component has to be extracted, is taken into account. New models take into consideration the bidispersed structure of a porous material and the presence of convective motion in large pores [5-9]. Convective mass transfer is taken into account within the diffusion models both by introducing coefficients of effective diffusion and by directly introducing the convective term [5,6]. The model is proposed for the process of extraction from a semi-infinite solid containing two types of pores: large pores that extend to the surface and small pores that are connected to large pores. Effective transport coefficients in two types of pores are assumed to be different. Theoretical results are compared with experimental data related to the kinetics of the extraction of the target components from plant materials obtained in apparatuses with an intensive hydrodynamic mode (mixing, ultrasounds). In this model, it is shown that the mass flux at the boundary of a semi-infinite solid with the branched system of pores at each time point t>0 is greater than that for a system, in which there are no branches from the main channel. The processing of experimental data on the kinetics of the extraction of active substances from plants has shown that the process is equally well described by both the model of extraction from a porous solid with semi-infinite transport pores and the model of extraction from a solid with the pores of a finite length. The problem of studying the regularities of solvent extraction from porous media is a matter of concern, because this process is widely used in the industry, especially in the extraction of medicinal components from plants. The pore space in actual media can have a very complicated structure (fig.1). There are pores with variable diameter, branched pores, isolated cavities, and the like. Therefore, the behavior of the diffusion process in such media can be much different from that in a single pore.
Fig. 1. Structure of plant material
Conventional methods for extracting active substances from plants are, as a rule, ineffective, since they do not ensure a sufficient degree of depletion of plants and are characterized by a long duration and nonproductive expenditures of input energy. At the same time, a constant increase in the volumes of production dictates a necessity for
B. Tal-Figiel
4 development of implementation.
new
intensive
extraction
methods
and
apparatuses
for their
2. Mass transfer model The model of solid-liquid extraction [5], depicted on fig.2, was used. A solution of the desired component from the porous body, is largely transported through small capillaries which branch off in large pores, either dead-end or through. It is assumed, that there is no liquid motion in the capillaries and active substance is transported there by molecular diffusion. However external-pressure pulses of some amplitude can induce liquid oscillations in large pores because of compression of the gas contained in capillaries [9].
Figure 2. Planar model of a particle with bidisperse structure; 1) porous block, 2) transport channel Thus, large pores can be regarded as transport channels where active substance is transferred by convection. Compared to molecular diffusion, convection can ensure a rate of solute extraction from particle, that is many times higher. Assuming, that the diffusion coefficient DM is independent of the active substance concentration, this process can be described by the following set of equations: diffusion equation for the porous body: ∂C1 ∂ 2 C1 , (1) = −D M ∂τ ∂y 2 where: C1 is the active substance concentration in the porous block, DM – molecular diffusion coefficient [m2s-1], τ – time [s], and the convective diffusion equation for the transport channel ⎛ ∂2C ∂2C ⎞ ∂C2 ∂C + u 2 = −DM ⎜ 22 + 22 ⎟ , (2) ∂τ ∂x ∂y ⎠ ⎝ ∂x where: C2 – the active substance concentration in the transport channel, u – the longitudal (along the x-axis) fluid velocity. The initial and boundary conditions for these equations are the following: C1 ( x, y, τ = 0 ) = C10 , 0 ≤ x ≤ L , h 2 < y ≤ h1 + h 2 ;
(3)
Solid-liquid extraction from plants with a bidisperse porous structure
5
C2 ( x, y, τ = 0 ) = C20 , 0 ≤ x ≤ L , 0 < y ≤ h 2 ;
(4)
C2 ( x, y = h 2 , τ ) = C1 ( x, y = h 2 , τ ) ;
(5)
q1 ( x, y = h 2 , τ ) = q 2 ( x, y = h 2 , τ ) ;
(6)
C2 ( x = 0, y, τ ) = C2 ( x = L, y, τ ) = C 20 , 0 < y < h 2 ;
(7)
q1 ( x, y = h1 + h 2 , τ ) = 0;
(8)
q 2 ( x, y = 0, τ ) = 0 ,
(9)
where: h1 and h2 are the half widths of the porous block and the transport channel, respectively L – the particle length, q1, q2 – the substance fluxes. The velocity profile in the transport channel is given by equation: ⎡ ⎛ y ⎞2 ⎤ (10) u ( x, τ ) = u max ⎢1 − ⎜ ⎟ ⎥ sin ( ωτ ) , ⎢⎣ ⎝ h 2 ⎠ ⎥⎦ where: ω – angular velocity of oscillation. The substance fluxes through the porous body-channel boundary (at y=h2) are expressed as: ∂C (11) q1 = −εD M 1 , ∂y q 2 = −DM
∂C2 , ∂y
(12)
where: ε – specifies the fraction of the channel surface through which active substance is transferred from the porous body. Since the liquid velocity in the channel is low and the channel diameter is small, the liquid flow can be assumed to be laminar and the liquid profile to be parabolic in space and harmonic in time. To solve the set of equations (1)-(12) the finite element and finite difference methods were used. Equation (1) was approximated using the absolutely stable Crank-Nicholson method of second order accuracy and equation alternating direction implicit (ADI) method of first order accuracy, with the approximation error O [Δτ, Δx2, Δy2], which is unconditionally stable in the linear case [6,7,10]. The boundary conditions were approximated by expressions of second-order accuracy. To ensure stability, the values of the Fourier number for the porous body and the Courant and Peclet numbers for the channel were checked. The program enables to monitor the concentration fields at mesh points and the relative linear concentrations of the extractable substance in the porous body and the channel:
B. Tal-Figiel
6 CL1 ( x, τ ) =
1 h1C10
CL2 ( x, τ ) =
h1 + h 2
1 h 2 C10
Pe =
C1 ( x, τ ) dy ,
(13)
∫ C ( x, τ ) dy , 2
(14)
1.
(15)
∫
h2
h2
u max L D
At Peclet numbers Pe =
uL ≤ 1, D
(16)
where: π
ω h2
u = ωπ−1 ∫ ∫ u ( y,τ)dydτ = 0 0
4umax
3π
. –fluid motion velocity averaged over the half-
period and the transport channel cross-section. Since the convective transfer in the channel has virtually no effect on the extraction rate, the main calculations were done for uL > 1. (17) Pe = D As the extraction efficiency criterion, the time τ required to attain a specified value for the residual concentration of the extractable substance in the porous body, was used: h1 +h2 L
∫ ∫ C dxdy 1
R=
h2
. (18) h1LC10 For the convective transfer of substance from a particle with a bidisperse capillary structure to occur at the highest rate, it is necessary that, at the maximum liquid velocity, the substance concentration at the channel outlet (x=L) also be maximum, because the convective flux of the substance is expressed as qconv=uC2. The velocity obtains its maximum in a time equal to a quarter of the oscillation period T: τ0=T/4= π/2ω. Within the suggested model, the instantaneous velocity of the liquid front is u fr = 2 u max sin (ω t ) and in time τ0, the front will travel the distance: 3 τ0
2 u max u av , (19) = ω 3 ω 0 which must be of the order of L. Therefore, the optimum oscillation frequency of the liquid can be found from the condition: ωL ≈ 1. (20) Sr = u av x fr = ∫ u fr dτ =
Solid-liquid extraction from plants with a bidisperse porous structure
7
In practice, it is important to know how the accelerating effect of oscillations on the extraction rate depends on the particle size. The bigger the particles, the weaker the effect of oscillations. The fluid volume contained within the transport channel is described by the balance equation expressing the law of mass conservation with respect to the substance being extracted: ∂M 1 (21) = G1 − G 0 , ∂τ where ∂M1/∂τ is the rate at which the mass of the substance being extracted grows in the transport channel, G1 is the substance flux at the boundary between the porous block and the transport channel and G0 is the outward flux from the transport channel: h ∂M1 ∂ 2 L (22) C2 dxdy , = ∂τ ∂τ ∫0 ∫0 ∂C2 ∂y 0
L
G1 = D M ∫
dx ,
h
2 ⎛ ∂C G 0 = DM ∫ ⎜ 2 0 ⎝ ∂x
− x =0
(23)
y = h2
∂C2 ∂x
⎞ ⎟dy . x =L ⎠
(24)
3. Numerical simulations Numerical simulations were conducted for the several parameter sets. One of them is shown below: ε = 0.29 h1 = h 2 = 1 ⋅ 10 −6 m C10 = 10 kg / m 3 umax = 0,002 m/ s ω = 1, 100, 1000, 10000 rad / s L = 1 ⋅ 10 −5 m D = 4.5 ⋅ 10−11 m 2 / s Numerical experiments demonstrated that, all other conditions being the same, there exist the optimal frequency of fluid oscillations in the channel ωopt, at which the minimum process duration can be achieved. Analysis of numerical calculations shows that two stages exist at any oscillation frequency (fig.3). The first is the attainment of regularity, with the substance content in the porous block falling fast because of high concentration gradients. In the process, one part of molecules of the substance being extracted goes from the transport channel outwards, and the other is accumulated in the transport channel. The second stage depends on the oscillation frequency. At low frequencies, the role played by convective substance transfer is insignificant. At higher frequencies, curve exhibits a weakly pronounced maximum. At the same time the intensity of substance transfer is still rather low. This is due to the fact that the molecules of the substance being extracted have to pass a longer way toward the opening at the transport channel virtually by means of diffusion only.
B. Tal-Figiel
8 C1
C2
0 C1
C2
C1
C2
Figure 3. Plots of C1 and C2 vs length for the frequency 1000 [s-1], and process times 0, 0.036, 0.056 [s]. At a nearly optimal frequency all curves show clearly pronounced extrema virtually coinciding in phase. The outflow of the substance to the transport channel is immediately followed by its transfer from the particle outwards, into the fluid surrounding the particle. The transfer from the porous block into the transport channel is enhanced, since the fluid in
Solid-liquid extraction from plants with a bidisperse porous structure
9
the transport channel rapidly “releases” the substance outwards, with all lines having large amplitudes.
Figure 4. Function R, eq (18) at the ω=1000 This type of substance transfer resembles exchange of potential and kinetic energies in mechanical and ultrasound oscillatory systems and therefore quite deserves being termed the “resonance” in mass transfer. Due to small particle size used for the simulation, the extraction progress is very fast what can be seen on fig.4. For longer channels the positive effect of oscillation gradually vanishes. Experimental The experimental part consisted of different types of liquid-solid extraction of active ingredients from plants (chlorophyll from stinging nettle leaves, menthol from peppermint leaves, inuline from dandelion root), obtained from Herbapol, Poland. These extraction methods included maceration, Soxhlet extraction, ultrasound-assisted extraction and microwave-assisted extraction. Solvents used in experiments were distilled water and 99,8% ethanol. Long time (48h) Soxhlet extraction was used to establish total amount of desired substance in plant material [11]. Obtained extracts were analyzed using 6715 UV/Vis Jenway Spectrophotometer. Theoretical results were compared with experimental data on the kinetics of the extraction of the target components from plant material in apparatuses with an intensive hydrodynamic mode (mixing, ultrasounds).
B. Tal-Figiel
100
100
80
80
Efficiency [%]
Efficiency [%]
10
60
0.32-0.5 1.2-1.6 3.0-4.5
40
60
40
20
0.32-0.5 1.2-1.6 3.0-4.5
20
0 0
50
100
150
200
250
300
0 10 30 40 In this model, it 20is shown that the 50matter60 Time [min] Figure 5. Comparison of water extraction efficiency of inulin from the dandelion root, with low speed mechanical mixing (6 s-1) and an ultrasonic field (P=220 W) for various grain size.
Time [min]
The processing of experimental data on the kinetics of the extraction of active substances from plants has shown that the process can be qualitatively well described by presented model. However this description is applicable only for the very beginning of process, when the dimensions of the internal channels can be regarded constant. Since both ultrasounds and microwaves deteriorate internal structure of plant material, which leads to changes in the channel size and also another mechanisms, apart from diffusion convection may appear (cavitation). Conclusions Base on the results of the conducted experiments, the following conclusions can be derived: • the obtained relations make possible to provide recommendations for determining the optimal extraction conditions, when the pore sizes in real particles are known; • for each fluid oscillation amplitude there exists the optimal oscillation frequency, at which the extraction has the highest intensity; • at a certain Strouchal number, approximately equal to unity, the extraction proceeds the most intensely; • in particles with long transport channels practically no effect was observed on the optimal oscillation frequency; • both in the case of ultrasound and microwaves the intensification of extraction process was observed. Greater efficiency was obtained with microwave extraction;
Solid-liquid extraction from plants with a bidisperse porous structure
11
• Soxhlet extraction is a good method in the case of organic solvent with low boiling temperature, and small amount of material, but both microwave and ultrasonic extractions are more energetically economic. Symbols C – active substance concentration [kg m-3] CL – relative active substance concentration, [--] DM – molecular diffusion coefficient, [m2s-1] h1 – half-thickness of the porous body, [m] h2 – half-width of the transport channel, [m] L – particle length q – active substance flux, [kg m-2s-1] R – relative amount of the desired substance that remains in the porous body, T – oscillation period, [s] ε – porosity τ – extraction time [s] ω – angular oscillation frequency [rad s-1] Pe – Peclet number Sr – Strouchal number Literature [1] Crank J.: The Mathematics of Diffusion, Clarendon Press, Oxford, 1975. [2] Aksielrud G.A., Łysianski W.M.: Ekstrakcja w układzie stało ciałe ciecz (Extraction. Solid-Liquid System), WNT, Warszawa 1978. [3] Aksielrud G.A., Altszuler M.A.: Ruch masy w ciałach porowatych (Mass Transport in Porous Media), WNT, Warszawa 1987. [4] Shewmon P.: Diffusion in Solids, Wiley, New York, 1989. [5] Abiev R.Sh.: Zh. Prikl. Khim. 74(5), (2000), 754-761. [6] Abiev R.Sh., Ostrovski G.M.: Teor.Osn. Khim. Teknol., 35(3), (2001), 270-275. [7] Malyshev R.M., et al.: Dokl. Akad. Nauk. 381(6) (2001), 800. [8] Babenko Yu., Ivanov E.V.: Teor.Osn. Khim. Teknol., 39(6), (2005), 644. [9] Ivanov E.V., Babenko Yu.I.: Zh. Prikl. Khim. 78(9), (2005), 1478. [10] Fletcher C.A.J.: Computational Techniques for Fluid Dynamics Vol.1. Springer Verlag, Berlin, 1988. [11] Quality control methods for medicinal plant materials, World Health Organization, Geneva, 1998.
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
Z. GUETTA*, J. C. SCHÖNEBERGER**, H. ARELLANO-GARCIA*, H. THIELERT**, G. WOZNY*
DEVELOPMENT OF A CLAUS PROCESS COMBUSTION CHAMBER MODEL Abstract Claus processes are widely used for the recovery of sulphur from hydrogen sulphide containing sour gas streams. The hydrogen sulphide is partially oxidized to sulphur, sulphur dioxide, and water in a combustion chamber. The hot gases coming from the furnace is cooled down and fed to the so called Claus reactors where the formation of sulphur from hydrogen sulphide and sulphur dioxide is supported by a catalyst. The scope of this work is the physical and chemical modelling of the reaction furnace in the combustion chamber. Due to the high number of microkinetic reactions and active species the numerical implementation of the mathematical model represented by a highly nonlinear algebraic equation system is a challenging task. Several numerical techniques are presented in order to reach high convergence qualities and robustness combined with a low computation time, thus enabling the use of process analysis and optimization tools. Keywords: Claus process, Combustion modeling, Sulphur, Model based control. 1. Introduction The original Claus process was developed by C.F. Claus in 1883 [11]. Currently there are plenty of different process implementations which have been developed in order to increase the rate of sulphur recovery or to reduce the capital and the operational expenditures. However, all these processes are based on the same principles, the partial oxidation of hydrogen sulphide at high temperatures (thermal region) and the formation of sulphur at low temperatures (catalytic region). The process can be characterized by the following main reactions:
*
TU-Berlin, Institut für Prozess- und Verfahrenstechnik, Sekr. KWT 9, Straße des 17. Juni 135, 10623 Berlin ** Uhde GmbH Dortmund
Development and Experimental Verification of a Claus Process Combustion Chamber Model
13
Thermal region: Catalytic region:
2H2S + 3O2 Æ 2SO2 + 2H2O 2H2S + SO2 ÅÆ 1½S2 + 2H2O
(1) (2)
2H2S + SO2 ÅÆ 1½S2 + 2H2O
(3)
The rate of sulphur recovery can be increased by adding intermediate condensation steps (multiple stage Claus processes, e.g. Superclaus) or by recycling the tail gas (e.g. SCOT process). In order to maximize the conversion of hydrogen sulphide to sulphur the concentration of hydrogen sulphide must be twice the concentration of sulphur dioxide in the gas stream leaving the reaction furnace as can be seen in equation (3). The difficulty in handling the process is attributed to the side reactions, which can take place in the reaction furnace. These reactions occur depending on the composition of the sour gas stream. Sour gas streams coming from a refinery have to be treated differently than those coming from a coal gasification or the by-product plant of a coking plant.
Tailgas
Al2O3 Secondery Air
NiO
SourGas
Primary Air Sulfur
Fig. 1. MONOCLAUS process.
The MONOCLAUS process is an emission-free sulphur recovery process as the tail gas can be fed completely to the untreated coke oven gas, which is entering the by-product plant, see Fig. 2.
Z. Guetta et al.
14 Anthracite Coal Combustion
Coke oven battery
Coke
Raw gas Draft
Water recirculation
Gas recirculation
Tar
Tar separation
Pre-cooling
Elektrofilter
Post-cooling
NH3-H2 S Absorber
H2S Desorber
Wastewater treatment absorption oil
Iron oxide
BTX Absorber
BTX Desorber
Claus Process
Sulfur
Wastewater
Benzene, Toluene and Xylene
Fine desulphurization COG
COG for Export
Fig. 2. By-product plant [13].
Sour gas coming from an amine based scrubbing process (e.g. CYCLASULF) commonly contain large amounts of ammonia, carbon dioxide, and hydrogen cyanide beside hydrogen
Development and Experimental Verification of a Claus Process Combustion 15 Chamber Model sulphide and water. The ammonia must be destroyed before the gas is fed to the Claus reactors as it leads to catalyst poisoning. Commonly, the decomposition to hydrogen and nitrogen in the furnace is not sufficient, so that a catalytic fixed bed for the decomposition is included in the combustion chamber. Due to the fact that the operation temperature of the catalyst is limited, it is important to control properly the furnace temperature. Furthermore, the minimum temperature of the furnace is also a safety relevant constraint of the operation conditions. At too low temperatures an explosive atmosphere can be created, as the combustion process is operated under oxygen deficiency conditions. Here, an accurate model is required during the process design phase in order to make arrangements e.g. for a feed charge of natural gas or quenching steam. During the operation phase of a Claus plant the model can be used for an online process optimization (e.g. GASCONTROL) or for the prediction of the concentration of undesired substances such as organic sulphur compounds in the tail gas. 2. Development of the Claus process combustion model The Claus process combustion model is based on three combustion mechanisms, which were combined together to one. The sources in the open literature of those mechanisms are referred bellow under the main combusted components. 1) 2) 3)
Combustion of NH3 - J.A. Miller et al [7]. Combustion of CH4, C2H6 and HCN - ‘‘GRI-Mech Version 3.0‘‘ [9]. Combustion of SO2 and H2S - P. Glarborg et al [2], [3], [5] & [6].
Changes in these mechanisms were done in order to make the combining feasible. Some reactions were eliminated so as to avoid duplication. In addition, some reaction rates were changed for fitting the simulated results to the measurements. The model art chosen to describe the combustion is a continuous stirred-tank reactor model. Moreover, in order to simulate the whole combustion chamber, small segments was defined where each of them is described by a continuous stirred-tank reactor model.
Fig. 3. PFR modeling by using a CSTR equations.
The mass balances in the model are formulated as follow:
0 = x j ,in Fin − x j ,out Fout + VR
i= N R
∑ν 1
r&
j ,i i
(4)
Z. Guetta et al.
16
Where x j and VR are the mole fraction of component j and the reaction volume, respectively. The reactants are defined as ideally mixed. r&i represents the reaction rate of reaction i. The reaction rates ( r&i ) is calculated as followed: NC
NC
j =1
j =1
ν' ν '' r&i = K f ,i ⋅ ∏ C j j ,i − K b ,i ∏ C j j ,i
(5)
Where, C j denotes the concentration of component j. C j is calculated under the assumption of an ideal gas behavior as followed:
Cj =
P ⋅ xj
(6)
R ⋅T
K f , i , K b , i represent the forward and backward reaction rate coefficients, respectively. K f , i is calculated by the following Arrhenius equation: βi
− Eai
K f ,i = Ai ⋅ T e R⋅T
(7)
K b ,i is depended on the chemical equilibrium constant ( K eq ) defined as followed:
K b ,i =
K f ,i K eq
(8)
The chemical equilibrium constant is calculated from the Gibbs free energy for a chemical reaction ( Δ G i0 ), which is calculated by using enthalpy and entropy polynomials taken from [1].
K eq ,i = exp(
− ΔGi0 ΔS 0 ΔH i0 ) = exp( i − ) RT R RT
(9)
The standard-state molar enthalpy and entropy of the reaction i are obtained with the relationship:
Development and Experimental Verification of a Claus Process Combustion Chamber Model
17
S 0j ΔS i0 N c = ∑ν j ,i R R j =1
(10)
H 0j ΔH i0 Nc = ∑ν j ,i RT RT j =1
(11)
The standard-state molar enthalpy and entropy for the component j are written as:
H 0j RT S 0j R
= a j ,1 +
a j , 2T 2
+
a j , 3T 2
= a j ,1 ln T + a j , 2T +
3
+
a j ,3T 2 2
a j , 4T 3 4 +
+
a j , 4T 3 3
a j , 5T 4
+
5 a j , 5T 4 4
+
a j ,6 T
+ a j ,7
(12)
(13)
3. The combustion model solver In order to solve the equation system presented in section 2, a novel tailored solver has been proposed. The main challenge represents the development of a robust solver, which demands also low computation time. Commercial solvers were considered but they showed a low performance in solving this type of equation system. The model contains 78 material balances, which include 418 reaction rates, and one energy balance. Due to the high correlation between the reactor temperature and the material balances, the equations system had to be decoupled into to two tasks. The first one solves the reactor material balances for a given temperature and the second one represents an external loop function, which solves then the energy balance. 3.1 The material balance solver The solver shown bellow is based on the Newton-Raphson approach, which was improved by adding additional functions so as to increase its performances. The proposed solver performs systematically. So for instance, the additional functions are described as followed: 1) Multiple sets of initial values - in case that no solution is found a new set of initial values is proposed and a new run starts. 2) Detect the validity of new iteration variables - the function checks whether the new iteration variables are within the boundaries defined, or if the computer memory has been exceeded (arithmetic overflow). In case of irregularity, a penalty function will be implemented. 3) Flexible convergence criteria - The convergence criteria is set to be a function of the iteration number.
Z. Guetta et al.
18
4) Saving the best solution found - Corresponding to a flexible convergence criterion, if a previous convergence criterion is not fulfilled, still a previous solution set may fulfill the new convergence criteria.
Savedinitialvalues Constant Initialvalues
Didlastsimulationconverge ? Yes:Usesavedinitialvalues No:Useconstant initialvalues
Finished Xmem ,Fmem yes
xj J=1..Ne Initialvalues
Changesconvergencecriteria no 0
CalculateF 0 andJAC
yes
If(n>D2 )
no
If(sum(|Fmem |)Δpresentor
, if Δnew Qc, as was presented in figure 8, exemplified by the two RT-6 turbines. In order to make a comparison, the magnitude of Kp sum was added to the graph. The numerical values were collected in table 5. The following graph illustrates interchangeably the scale of a phenomenon – the difference between the KC KOR and Kp sum. It proofs, in the cases of the complex flow structure and the geometry of impellers, that the real flow rates in the vessel are many times higher than it seemed to result from the study of the pumping capacity exclusively for the impellers.
J. Kamieński et al.
154
T heflownumbers K
7 .0 6 .0
K C K O R Kc K ps um
5 .0 4 .0 3 .0 2 .0 1 .0 0 0 .5
1 .0
2 .0 Δh /d
1 .5
Figure 8. The confrontation of the Kp sum, Kc and KC KOR for the set of two RT-6 turbines, versus different values of ∆h/d.
Table5 The confrontation of the Kp sum , Kc and KC KOR for the set of two RT-6 turbines.
∆h /d
Kp sum
Kc
KC KOR
0.5 1.0 1.5 2.0
0.80 0.94 1.03 0.84
2.67 2.23 3.18 3.06
4.78 4.25 6.85 6.53
6. Resumé The presented study draws attention to the aspect of the liquid flow rates in the dual mixing vessel. The knowledge of the flow structure and the flow rates is the key-issue, influencing the proper design of the vessel, impeller as well as the processes which are going to be conducted. The correct assumption of the flow rates, based on the corrected values of QCKOR and KC KOR, assures the correct geometry and dimensions of the device, including the sufficient safety and reserve factor. The list of more important symbols D - tank diameter, [m] d - impeller diameter, [m] H - liquid height, [m] Δh - impeller spacing, [m] K - flow number, general, [-] Kc - circulation flow number, [-] KC KOR - corrected circulation flow number, [-] KE- inter-stage flow number, [-] Kp - pumping flow number, [-] Kp sum- total pumping flow number for dual impeller, [-] n - rotational speed, [s-1] Q - flow rate, general, [m3/s]
The dual impeller capacities in the light of CMA model
155
Qc - circulation flow rate, [m3/s] QC KOR - corrected circulation flow rate, [m3/s] QE- inter-stage flow rate, [m3/s] Qp - pumping flow rate, [m3/s] Qp sum- total pumping flow rate for dual impeller, [m3/s] Qr - maximal flow rate in radial profile, [m3/s] Qz - maximal flow rate in axial profile, [m3/s] r - vessel’s radius, [m] ūn - mean velocity component, normal to impeller plane, [m/s] ūr - radial component of the mean velocity in vessel, [m/s] ūz - axial component of the mean velocity in vessel, [m/s] z - vessel’s axis, [m] Literature [1] Nienow A.W.: Chemical Engineering Science, 52, (1997), 2557-2565. [2] Kamieński J.: Mieszanie układów wielofazowych, Wydawnictwa NaukowoTechniczne, Warszawa 2004. [3] Jaworski Z., Nienow A.W., Dyster K.W.: Canadian Journal of Chemical Engineering, 74, (1996), 3-15. [4] Jaworski Z., Nienow A.W., Koutsakos E., Dyster K.W., Bujalski W.: Trans IChemE, Part A, 69, (1991), 313-320. [5] Stręk F.: Mieszanie i mieszalniki, Wydawnictwa Naukowo-Techniczne, Warszawa 1981. [6] Jaworski Z., Nienow A.W.: Eighth European Conference on Mixing, Cambridge, (1994), No 136. [7] Vrábel P., van der Lans R.G.J.M., Luyben K.Ch.A.M., Boon L., Nienow A.W.: Chemical Engineering Science, 55, (2000), 5881-5896. [8] Vasconcelos J.M.T., Alves S.S., Barata J.M.: Chemical Engineering Science, 50, (1995), 2343-2354. [9] Duda A., Kamieński J., Talaga J.: Inżynieria i Aparatura Chemiczna, Nr 2, (2010), 2728. [10] Duda A., Talaga J.: 16th International Conference “Chemical Engineering and Plant Design”, Berlin 2006, 159-168. [11] Duda A., Kamieński J., Talaga J.: Czasopismo Techniczne, 5-M, (2008), (105), 67-78. [12] Duda A., Kamieński J., Talaga J.: Czasopismo Techniczne, 2-M, (2008), (105), 67-74. [13] Otomo N., Bujalski W., Nienow A.W.: The 1995 IchemE Research Event / First European Conference, (1995), 829-831.
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
R.M. GÜNTHER*, J.C. SCHÖNEBERGER**, H. ARELLANO-GARCIA*, H. THIELERT**, G. WOZNY*
DESIGN AND MODELLING OF A NEW PERIODICAL STEADY-STATE PROCESS FOR THE OXIDATION OF SULFUR DIOXIDE IN THE CONTEXT OF AN EMISSION FREE SULFURIC ACID PLANT Abstract The oxidation of sulfur dioxide over vanadium pentoxide catalysts in a fixed bed represents a basic step in the sulfuric acid production process. In kinetic investigations, an effect in the dynamic operation of the catalyst was observed which leads to much higher reaction rates in a limited time period. The SMP (Saturated Metal Phase) effect can be observed by charging the catalyst with oxygen at the operating temperature and then exposing it to sulfur dioxide. This causes a reaction between the chemisorbt oxygen and the sulfur dioxide. This effect leads to several advantages in the sulfuric acid production process. In addition to a drastical increase in the efficiency of the whole process, possibilities for new process concepts can lead to an emission free sulfuric acid process in the context of coking plant gas treatment. Therefore, a mathematical model is used that can describe the dynamic effects and takes into account the varying active vanadium species. The developed approaches are based on experimental investigations in a miniplant for the oxidation of sulfur dioxide with commercial catalyst pellets. Keywords: sulfur dioxide oxidation, unsteady state, sulfuric acid production 1. Introduction Vanadium pentoxide based catalysts for the oxidation of sulfur dioxide to sulfur trioxide are some of the best investigated catalysts in heterogenous catalysis. Due to the fact that sulfuric acid is one of the most important chemicals and the sulfurdioxide oxidation is the basic step of the production process, research on the oxidation catalyst has *
Institut of Process Engineering, TU Berlin, Straße des 17. Juni 135, 10623 Berlin, Germany; Uhde GmbH Dortmund, Friedrich Uhde Straße 15, 44141 Dortmund, Germany;
**
Design and modeling of a new periodical-steady state process for the oxidation 157 of sulfur dioxide in the context of an emission free sulfuric acid plant been conducted for some 40 years [1,3,9,15,16]. Because of the equilibrium limited character of the reaction SO2 + ½ O2 SO3 it is imposible to reduce the SO2 emissions down to zero at standard operating temperatures [9]. There have been great efforts to shift the equilibrium to the product side. On the one hand, there are air quenches or gas heat exchangers to reduce the reaction temperature of the higly exothermic reaction and increase the oxygen concentration between the reactor beds. Therefore, new vanadium based catalysts were invented to work in lower reaction temperatures down to 390°C with acceptable reaction rates. On the other hand, intermediate absorptions of SO3 are installed to remove the product between the reactor beds. Investigations about the dynamic behavior of the reaction by Boreskov and Matros [6] observed a change of the catalyst structure depending on the exposed gas mixture. Also, several authors observed different behavior of the reaction between dynamic and steady state operation [2,5-8,9,11-14,21-23]. In experimental kinetic studies, Schöneberger [17] observed a peak of the reaction rate if the catalyst is first exposed to oxygen and then to reaction mixture that contained oxygen and sulfur dioxide. A reaction network that describes the whole catalytic cycle can explain this behavior [2]. The usage of the additional dimension of time gives the possibility to achieve a super-equilibrium [10] which is limited by the equilibrium of one specific partial reaction in the catalytic cycle. In the context of coking plant gas treatment, a new sulfuric acid process concept without any emissions is proposed. In this process, the sulfur dioxide in the sulfuric acid plant rest gas is hydrogenated to hydrogen sulfide and then recycled in the coking plant gas treatment. The usage of a transient working oxidation is a promising process alternative to meet the restrictions of sulfur dioxide and oxygen content in the hydrogenation reactor and the recycle gas. 2. Emission free sulfuric acid process The sulfuric acid process in the context of coking plant conditions consists of three basic process steps. The H2S containing feed gas stream is oxidized in a combustion unit to SO2 under excess air conditions to decrease temperature and minimize NOx formation. In a second step, the SO2 is oxidized to SO3 over a vanadium pentoxide (V2O5) catalyst in a strongly exothermic reaction. The fixed bed reactor is divided into 4 or 5 beds and in each bed a defined temperature is set to maximize conversion. The temperature adjustment is realized with air quenches which, in addition to the temperature decrease, also lead to high oxygen content in the process gas. The last bed typically has the lowest temperature of all reactor beds to shift the reaction equilibrium to the product side and minimize the SO2 content in the process gas stream. After the contact reactor, the gas passes through a sulfuric acid absorber, where the SO3 is absorbed with water and converted to sulfuric acid. To prevent sulfuric acid mist, wire mesh filters are installed. After this last process step, the restgas is released out of a stack to the environment. Due to the fact that the SO2 oxidation is an equilibrium reaction, there are always SO2 emissions. To prevent these emissions, a new process concept is recommended that recycles the sulfuric acid plant restgas back to the coking plant gas treatment. Thus, the SO2 emissions can be reduced down to zero. A very low SO2 and O2 content is required to feed a gas stream into the coking plant gas treatment. These requirements are only met if the SO2 content is reduced, for example, by the hydrogenation of SO2 to H2S
R.Günther et al.
158
(3H2+SO2H2S+2H2O) in an exothermal reaction with the equilibrium far to the product side. Therefore, a fixed bed with hydrogenation catalyst has to be installed behind the sulfuric acid absorber. To meet the requirements of the hydrogenation unit there has to be an oxygen content of less than xO2=0.5 mol%. A transient working reactor using SMP could be used instead of the steady state contact reactor. The transient working reactor consists of two fixed beds fed cyclically with oxygen and sulfur dioxide. SO2 oxidation with a nearly oxygen free reactor outlet stream is possible with the temporarily shifted oxygen charging of the catalyst. In the oxygen charging step, an oxygen reduced stream can be used for combustion to reduce temperature and limit NOx production. Figure 1 compares the process concepts of a conventional and an emission free sulfuric acid plant.
Stack
1 2 Sour gas
H2 S combustion
Heat recovery boiler
3 4 Sulfuric acid absorber
Contact reactor
Air
COG SMPreactor
Sour gas
H2S combustion
Heat recovery boiler
Coking plant gas treatment Sulfuric acid absorber
Hydrogenation reactor
Air
Fig. 1. Sulfuric acid plant and emission free sulfuric acid plant with SMP (Saturated Metal Phase)
3. Kinetic, model and numerical methods
Design and modeling of a new periodical-steady state process for the oxidation 159 of sulfur dioxide in the context of an emission free sulfuric acid plant For the simulation of the SMP process, the model needs to represent the dynamic behavior of the reaction. A model based on the overall reaction of the SO2 oxidation (SO2 + ½ O2 SO3) does not consider the intermediate forms of the vanadium catalyst. The SO2 oxidation takes place in a liquid metal phase [5] with the active species. Hence, besides mass transfer over the phase border, the reactions inside the metal phase have to be taken into account. Balzhinimaev [5] postulates the catalytic cycle (Eq. 1, 2, 3, 5) with vanadium intermediates as follows:
V25+ O22− + SO2 ↔ V25+ O 2− + SO3 5+ 2
2−
5+ 2
2− 3
V O + SO2 ↔ V SO 5+ 2
2− 3
V SO
5+ 2
2− 2
+ O2 ↔ V O
+ SO3
ΔhR = 42 kJ / mol ΔhR = 80 kJ / mol
(1)
Δ = 63 kJ / mol
(3)
2 SO 2 + O2 ↔ 2 SO 3 5+ 2
2− 3
V SO
4+ 2
↔V
(2)
h R
(4)
+ SO3
Δ = −42 kJ / mol h R
(5)
The capacity for the absorption of SO2, O2 and SO3 in the melt is described by Henry’s Law because of the high adsorption velocity. The reaction rates are described by the mass action law recommended by Bunimovich [8]. The reactor model is realized as a heterogeneous and transient process model of a fixed bed reactor, which considers the gas phase and the catalyst phase separately. This modeling approach is necessary to display the components of the mobile phase (SO2, O2, SO3) and of the immobile phase (SO2, O2, SO3, V25 + SO32 − , V25+ O22− , V24+ , V25+O 2− ). Furthermore, it makes sense for dynamic simulation to consider the delaying effects of the interphase mass and energy transfer. According to changes in the steady state solution, they have great impact on the dynamic solution. The following assumptions were made: 1) No axial heat conduction of the catalyst phase 2) Plug flow. No radial heat or mass transport 3) No heat conductivity between catalyst and the wall (for calculation of heat losses) 4) No temperature dependence of the heat and mass transfer coefficients 5) No temperature and concentration gradients in the catalyst pellet 6) Reaction only in the catalyst phase With the given assumptions, the partial differential equation system displayed in Eq. 6 – 10 results. It consists of an energy balance for the gas phase with a convective term for the gas flow, a heat transfer term between the phases and a heat transfer term that shows the heat losses to the environment. 3
dTg
i=1
dt
∑ci (cp,i − R)
=a
1− ε
ε
α(Tk − Tg ) −
∂Tg & 3 m 4 ci cp,i − kRW (Tg − TRW ) ∑ Aρε i=1 ∂z Dr
(6)
R.Günther et al.
160
Additionally there is a mass balance for the gas phase, which has a term for axial dispersion, a convective mass transport term through the reactor and a term for interphase mass transfer.
dcg,i dt
= Dax,i
∂ 2cg,i ∂z 2
−
& ∂cg,i m − aβ (ck,i − cg,i ) Aρε ∂z
(7)
The energy balance contains one term for the heat transfer between the phases and one term for the heat resulting from the sulfur trioxide absorption.
(1 − ε )ρk c p,k
4 dTk ∂c L = aα(Tg − Tk ) − CV ε L ∑ΔhR r&i − ε LQD 3 dt ∂t i =1
(8)
The component balance of the catalyst (melt) contains the reaction term and the interphase mass transfer.
εL
4 dciL = ε LCV ∑ν i, j r&j + aβ (cg,i − ck ,i ) dt j =1
(9)
The component balance for the immobile components in the catalyst consider only the reaction term because the components stay in the melt and simply change from one species to another.
dμm 4 = ∑ν m, j r&j dt j =1
(10)
For the boundary and initial conditions
t = 0 : ciL = ciL,0 (z) ci = ci , 0
μm = μm,0 (z ) Tg = Tg,0 (z) Tk = Tk ,0 ( z )
z = 0 : ci = ci,in Tg = Tg ,in
(11)(12) (13)(14) (15) (16) (17)
Design and modeling of a new periodical-steady state process for the oxidation 161 of sulfur dioxide in the context of an emission free sulfuric acid plant were implemented. The parameters assumed by Balzhinimaev et al [5] could not be used because of working with real catalyst and not powdered catalyst like in common kinetic investigation. Transport limitations and modified catalyst compositions were assumed. Also different experimental conditions could lead to the deviation. The kinetic parameters are estimated with the measurement data from the miniplant as per description in [18]. This leads to results which can be used for simulations of industrial processes. The effort has to be made for each catalyst which is used for dynamic processes if reliable predictions shall be achieved. For the numerical calculation the system is discretized with finite differences along the spatial coordinate. This discretization method has to be used because of the steep gradients in time and along the reactor. A discretization method like, the orthogonal collocation [4, 18], could not be used because of the affinity of the solution to oszillate. The temporal coordinate is solved by a multistep solver using Backward Differential Formulas (BDF). Because of the finite differences discretization in space the jacobian matrix is very sparse and has a bandstructur. Therefore the ODE solver works with a sparse jacobian matrix which makes it much more efficient. 3. Experimental setup The experimental setup is a fixed bed reactor which can be used for the investigation of a variety of commercial catalyst pellets. It contains a gas mix section, a heating section and the reactor with several measurement points. It is displayed in Figure 2.
FIC
SO2
TIC TI TI FIC
TI TI TI TI
N2/ O2
TI TI FIC
N2
QI
Wastegas treatment
Fig. 2. Experimental setup for research on industrial catalysts [18]
R.Günther et al.
162
The waste gas treatment is not shown in Figure 2. It contains a scrubber to remove SO3 and SO2 before process gas is released into the environment. The wall temperatures of the reactor are measured to quantify heat loss and the reactor has a heated jacket to reduce heat loss. The temperature measurement inside of the reactor show the temperature profile in the reactor and the quality indicator at the end of the reactor measures the oxygen content of the process gas. 4. Results and discussion Figure 3 shows the comparison between the temperature measurement behind each reactor bed and the simulation results of the reactor. Each reactor bed has the same height so measurement points are equidistant distributed along the reactor. Figure 4 shows the relation of the temperature over reactor length from the simulation of the same experiment. 470 460 450
] C °[ s a g er ut ar e p m et
440 430 420 410 400
reactor bed 1 - 4
390
simulation experiment
380 370 0
500
1000
1500
time [s]
2000
2500
3000
Fig. 3. Temperatures at four equidistant points in the reactor and simulation results of the model
Till t=0 the catalyst is heated with air to operating temperature and the heat loss leads to a decreasing temperature along the reactor. At t>=0 reaction mixture of SO2, O2 and inert gas N2 is fed into the reactor. The reaction starts and high temperature gradients appear which lead to a strong increase in temperature along the reactor. At t=3000s a steady state has been established. During the accelerated reaction rate most of the SO2 is converted to SO3 so that a nearly SO2 free reactor outlet stream can be achieved [23] which is in line with the simulation results in Figure 5.
Design and modeling of a new periodical-steady state process for the oxidation of sulfur dioxide in the context of an emission free sulfuric acid plant
163
480 460 ] C °[ 440 s a g er 420 ut ar e p 400 m et 380
t=540 s t=360 s t=180 s
t=60 s
t=0 s
360 0
0.1
0.2 length [m]
0.3
0.4
Fig. 4. Simulation results of the temperature profiles along the reactor 0.9 0.8 ] 3 m /l o m [ n oti ar t n e c n o c
0.7 0.6
time cSO
2
0.5 0.4 0.3 0.2 0.1 0 0
cO
2
0.1
0.2 length reactor [m]
0.3
0.4
Fig. 5: SO2 and O2 along the reactor. Cyclic operation.
Considering that the SO2 oxidation is an equilibrium limited reaction this is a great increase in efficiency. If this effect is used in the proposed emission free sulfuric acid plant the SO2 conversion is less important compared with the O2 concentration in the off gas stream. In order to reach the required SO2 concentrations together with the maximal allowed O2 the amount of catalyst required must be increased approximately by a factor of five if the conventional steady state process is choosen [18, 20]. Figure 5 shows simulation results with an oxygen free reactor inlet stream. Over time the SO2 front shifts through the reactor until it reaches the reactor outlet. This is the point
R.Günther et al.
164
where the reactor switch could take place. The oxygen amount at the reactor outlet is close to zero, because no oxygen is fed into the reactor. However, some oxygen might be generated in the gas phase by the reverse of reaction (Eq. 4). The fact that the oxygen is fed temporarily decoupled into reactor realizes this operation. Figure 6 shows the distribution of the vanadium intermediates along the reactor in a periodical operated regime. In the adsorption half cycle the oxygen chemisorbs on 5+ 2 − V25+ SO32 − and builds a reservoir of V2 O2 . Thus, the in the steady state process rate limiting step of the catalytic cycle is finished and the oxygen enriched vanadium intermediate is available for the much faster step of SO2 conversion. 1
Adsorption
Oxidation
0.8 ] -[ 0.6 n oti c ar f el 0.4 o m 0.2
V5+ SO22 3 5+
2-
V2 O2
V5+ O22
0 time [s]
Fig. 6: Vanadium intermediates along the reactor. Periodical operation.
In the oxidation half cycle the vanadium species V25+O 2− and V25 + SO 32 − were produced sequentially in the reactions (Eq. 1) and (Eq. 2), each with one SO2 molecule. When the oxygen carrier V 25 + O 22 − is completely consumed the oxidation from SO2 to SO3 stops and the catalyst has to be recharged with oxygen. 5. Conclusions A new process concept is shown, that can realize an emission free sulfuric acid plant in the context of coking plant wastegas treatment. With the use of this concept requirements of the sulfuric acid plant tailgas recycle in the coke oven gas are met. Moreover the efficiency of a sulfuric acid plant can be considerably increased. On one hand a smaller contact reactor can be built which leads to lower investment costs. On the other hand smaller downstream units can also be realized because of a decreased process gas stream. With the use of heat integration the efficiency could be further increased. With the oxygen reduced process gas stream an advantageous operation of the combustion is possible because of lower combustion temperatures which leads to lower NOx production. The process displayed in Figure 1 can be realized at new sulfuric acid plants in coking plant gas treatment. There is also the possibility to upgrade an existing sulfuric acid plant by
Design and modeling of a new periodical-steady state process for the oxidation 165 of sulfur dioxide in the context of an emission free sulfuric acid plant changing the operating conditions of the contact reactor an add as a last contact unit an SMP based reactor to realize a higher SO2 conversion and remove oxygen from the process gas stream. If the process is used in the context of coking plant gas treatment the last process step is a hydrogenation unit that has to be installed to meet the reqirements for the recycle into the cokeoven gas treatment. 6. Acknowledgement The present work was supported by the Max-Buchner-Reserch Foundation. Symbols
A a CV ci ciL cik c p ,i
Cross section area Catalyst solid surface
m² 1/m
Total vanadium concentration
mol/m³
Concentration gasphase component i
mol/m³
Concentration melt component i
mol/m³
Concentration catalyst component i
mol/m³
Heat capacity gas component i
kJ/(kg K)
c p ,k
Heat capacity catalyst
kJ/(kg K)
Dax ,i Dr ΔhR ,i k RW m& QD R r&j Tk Tg TW t z α
Axial dispersion coefficient
m²/s
Radius reactor
m
Heat of reaction
kJ/mol
Thermal transmission coefficient
W/(K m²)
Mass flow
kg/s
Heat of SO3 dissolution
J/mol
Gas constant
J/(mol K)
Reaction rate
mol/(s kgcat)
Temperature catalyst
°C
Temperature gas
°C
Temperature wall
°C
Time Length reactor Heat transfer coefficient
s m W/(m² K)
R.Günther et al.
166
β ε εL μ m,i νj νj ρk ρ
Mass transfer coefficient Void fraction
m/s -
Void fraction liquid metal
-
Concentration vanadium species
-
Stoichiometric coefficient
-
Stoichiometric coefficient
-
Density of the catalyst
kg/m³
Density of the gas
kg/m³ Literature
book: [1] Bartholomew C.H., Farrauto R.J.: Fundamentals of Industrial Catalytic Processes, Wiley-AIChE, 2 edition, 2005. [2] Bunimovich G.A, Strots V.O., Goldman O.V.: Theory and industrial application of SO2 oxidation reverse process for sulfuric acid production, Unsteady state process in catalysis, VNU Science Press, (1990), 7-24 [3] Näumann F., Schulz M.: Oxidation of sulfur dioxide, Handbook of heterogeneous catalysis, (2008), 2623-2635 [4] Finlayson, Bruce A.: Nonlinear analysis in chemical engineering, McGraw-Hill, New York, (1980), ISBN 0070209154 paper: [5] Balzhinimaev B.S., Ivanov A.A., Lapina O.B., Mastikhin V.M., Zamaraev K.I.: Mechanism of sulfur dioxide oxidation over supported vanadium catalysts, Faraday Discuss. Chem. Soc., 87, (1989), 133-147 [6] Boreskov G.K., Matros Yu.Sh.: Unsteady-state performance of heterogeneous catalytic reactions, Catalysis Reviews-Science and Engineering, 25, (1983), 551-590 [7] Briggs J.P., Hudgins R.R., Silveston P.L.: Composition cycling of an SO2 oxidation reactor, Chemical Engineering Science, 32, (1977), 1087-1092 [8] Bunimovich G.A., Vernikovskaya N.V., Strots V.O., Balzhinimaev B.S.: SO2 oxidation in a reverse- flow reactor: influence of a vanadium catalyst dynamic properties, Chemical Engineering Science, 50, (1995), 565-580 [9] Dunn J.P., Stenger H.G., Wachs I.E.: Oxidation of sulfur dioxide over supported vanadia catalysts: molecular structure - reactivity relationships and reaction kinetics, Catalysis Today, 51, (1999), 301-318 [10] Goldman, O.V., Bunimovich, G.A., Zagoruiko, A.N., Lakhmostov, V.S., Vernikovskaya, N.V., Noskov, A.S., Kostenko, O. V.: Sulphur dioxide oxidation method. RF Patent 2085481, C01B17/76, B, (1997), N21. [11] Gosiewski K.: Dynamic modelling of industrial SO2 oxidation reactors. Part II. Model of a reverse-flow reactor, Chemical Engineering and Processing, 32, (1993), 233-244
Design and modeling of a new periodical-steady state process for the oxidation 167 of sulfur dioxide in the context of an emission free sulfuric acid plant [12] Holroyd F.P.B., Kenney C.N.: Sulphur dioxide oxidation kinetics: the absorption of oxygen in V2O5-potassium pyrosulphate melts, Chemical Engineering Science, 26, (1971), 1971-1975 [13] Hong R., Li X., Li H., Yuan W.: Modeling and simulation of SO2 oxidation in a fixedbed reactor with periodic flow reversal, Catalysis Today, 38, (1997), 47-58 [14] Ivanov A.A., Balzhinimaev B.S.: New data on kinetics and reaction mechanism for SO2 oxidation over vanadium catalysts, React. Kinet. Catal. Lett., 35, (1987), 413-424 [15] Mars P., Maessen J.G.H.: The mechanism and the kinetics of sulfur dioxide on catalysts containing vanadium and alkali oxides, Journal of Catalysis, 10, (1968), 1-12 [16] Mezaki R., Kadlec B.: Remarks on the reduction-oxidation mechanism of sulfur dioxide on vanadium catalyst, Journal of Catalysis, 25, 1972, 454-459 [17] Oruzheinikov A.I., Chumachenko V.A., Matros Yu.Sh.: Analysis of a nonsteady-state kinetic model for SO2 oxidation, Reaction Kinetics and Catalysis Letters, 21, 97-102 [18] Schöneberger J.C.: Entwicklung und Analyse katalytischer Abgasbehandlungsprozesse am Beispiel der emissionsfreien Schwefelsäureanlage, Shaker Verlag Aachen, (2010) [19] Schöneberger J.C., Arellano-Garcia H., Thielert H., Wozny G.: Identification of reaction mechanism with a dynamic PFR model, Advanced Control of Chemical Processes, 7, (2009) [20] Schöneberger J.C., Arellano-Garcia H., Thielert H., Wozny G.: Ein systematischer Ansatz zur Entwicklung und Analyse verfahrenstechnischer Prozesse am Beispiel einer Kohlenwertstoffanlage, Vortrag auf dem Jahrestreffen der Dechema Fachgemeinschaft Prozess-, Apparate- und Anlagentechnik (2009) [21] Silveston P.L., Hudgins R.R., Bogdashev S., Vernikovskaya N.V., Matros Yu.Sh.: Modelling of a periodically operating packed-bed SO2 oxidation reactor at high conversion, Cemical Engineering Science, 49, (1994), 335-341 [22] Snyder J.D., Subramaniam B.: Numerical simulation of a periodic flow reversal reactor for sulfur dioxide oxidation, Chemical Engineering Science, 48, (1993), 4051-4046 [23] Vernikovskaya N.V., Zagoruiko A.N., Noskov A.S.: SO2 oxidation method. Mathematical modeling taking into account dynamic properties of the catalyst, Chemical Engineering Science, 54, (1999), 4475-4482
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
JANUSZ MAGIERA*
ENERGIE – REALE MÖGLICHKEITEN ZUR GEWINNUNG SAUBERER ENERGIE FÜR WOHNHÄUSER Allgemein Der Beitrag stellt die derzeitige Struktur der Energienachfrage im Maßstab der Welt und der Länder der EU dar. Die Struktur der erzeugten und verbrauchten Energie in Polen und Deutschland wurde detailliert erklärt. Es wurden Hybridsysteme zur Wärmeerzeugung durch erneuerbare Energien gezeigt, die im Versuch in einer Produktionsanlage in Südpolen laufen. Es wurde ein neues Steuerungs-und Energiebilanzierungssystem vorgestellt, das auch online arbeiten kann. Schlüsselwörter: Primärenergie, Solarenergie, Solarkollektor, Hybridanlage, Steuerungssystem Energie wird immer knapper und immer teurer und damit gleichzeitig das kostbarste Allgemeingut für den statistischen Einwohner der Erde. Es gibt Schätzungen, die auf den letzten Konferenzen der UNESCO präsentiert wurden, dass die Preise für Energie und sauberes Wasser in diesem Jahrhundert am schnellsten von allen Gütern wachsen werden. Am Anfang des zwanzigsten Jahrhundert betrug die Bevölkerung der Erde rund 1,7 Milliarden Menschen. Bis zum Ende des Jahrhunderts im Jahr 2000 überschritt die Bevölkerung 6,0 Mrd. Das bedeutet einen 3,2-fachen Bevölkerungsanstieg. Gleichzeitig stieg der Energieverbrauch auf der Erde um mehr als das17-fache [1]. Es gibt Schätzungen, dass im letzten Jahrhundert, mehr Energie auf der Erde hergestellt und verwendet wurde als in der gesamten Dauer des menschlichen Lebens auf der Erde [2]. Derzeit ist die Wachstumsrate des Energieverbrauchs nicht so hoch wie im vergangenen Jahrhundert, aber der Verbrauch wächst weiter. Traditionelle Energieträger wie Kohle, Erdöl oder Erdgas werden in absehbarer Zeit erschöpft sein. Ihre derzeitige Nutzung ist mit erheblichen negativen Auswirkungen auf die Umwelt durch die Emission von CO2, SO2, NOx-und Partikelemissionen verbunden. Daher sucht man intensiv neue, saubere Energiequellen, möglichst ohne negative Umweltauswirkungen. Obwohl Kernkraftwerke, per Saldo, als relativ "sauber" betrachtet werden, sind sie wegen der ungelösten Probleme der Endlagerung radioaktiver Abfälle noch immer nicht völlig akzeptabel. Man erinnert sich noch an die Reaktorkatastrophe von Tschernobyl in *
Technische Universität Krakau, 31-155 Krakau, Polen
Energie – reale Möglichkeiten zur Gewinnung sauberer Energie für Wohnhäuser
169
der Ukraine, sowie die neuesten Ereignisse im März 2011 in Japan. Daher sucht man intensiv, sowohl in der Welt als auch in Europa nach neuen, sauberen Energiequellen, die die erneuerbaren Energien (RES), umfassen. Derzeit überwiegen traditionelle Energiequellen in Form von Brennstoffen in der Struktur der Energieversorgung der Welt, wie in Abbildung 1 dargestellt. The share of the major sources of energy generation in the world in 2009 8% 26% 66%
nuclear energy renewable energy fossil fuels
Source: Global Status Report, Renewables 2010
Abb.1. Anteil der wichtigsten Quellen der Primärenergieerzeugung in der Welt
Aus Abbildung Nr. 1 kann man erkennen, dass in der Welt die fossilen Brennstoffe noch etwa 66% der Primärenergie betragen. Eine interessante Information ist, dass der Anteil erneuerbarer Energien am Gesamtenergieverbrauch 26% beträgt. Das ist relativ viel, aber der größte Teil davon wird als Biomasse in Verbrennungsprozessen verwendet. Die größten Energieverbraucher der Welt im Jahr 2009 waren China und die USA, wie in Abb.2 dargestellt. Major consumers of energy in 2009 1 toe = 41,89 GJ China: 64 GJ/M.a U.S.: 229 GJ/M.a India: 23,5 GJ/M.a Russia: 354 GJ/M.a Germany: 161,2 GJ/M.a Poland: 102,8 GJ/M.a
2500
Mtoe
2000 1500 1000 500
Ko re a
Br az il
So ut h
an ad a C
Fr an ce
an y er m
G
Ja pa n
us si a R
In di a
.S . U
C
hi na
Source: http://yearbook.enerdata.net/
Abb.2. Die größten Energieverbraucher im Jahr 2009
Es ist erwähnenswert, dass der Verbrauch pro Person und Jahr sehr unterschiedlich ist, wie in der nächsten Abbildung angezeigt. Die Struktur des Energieverbrauchs in den Industrieländern ist etwa so, dass ca. 30% im Verkehr, etwa 30% in der Industrie und die restlichen 40% in Haushalten, Dienstleistungen, Landwirtschaft und anderen Bereichen anfallen. Abb.3 zeigt den jährlichen Energieverbrauch in den 27 Ländern der Europäischen Union (EU) nach Bereichen. Im Jahr 2007 entfällt 32,6% der Energie der EU-Länder auf
J. Magiera
170
den Verkehr, 27,9% auf die Industrie, 24,6% auf die Haushalte und 15% auf andere Bereiche. Final energy consumption by sector in the EU-27, 1990-2007
Shares in 2007
1200
3.7%
1100
11.2%
Million tonnes of oil equivalen
1000 900 24.6%
800
Agriculture and other sectors Services
700
Households
600 27.9%
500 400
Industry
300 32.6%
200
Transport
100 2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
Source: Eurostat http://www.eea.europa.eu
Abb.3. Der jährliche Energieverbrauch in der EU-27 nach Bereichen
Die bequemste Energieform im Einsatz (außer im Bereich des Verkehrs) ist die Elektroenergie, obwohl ihre Kosten die höchsten im Vergleich zu anderen Energieformen auf dem Markt sind. Abb.4 zeigt die Energiepreise für die Haushalte in den 22 Ländern der Europäischen Union und Abbildung 5 die geltenden Preise in der Industrie, ohne Steuer, in diesen Ländern. Die Analyse dieser Daten zeigt, dass im Jahr 2010 die niedrigsten Strompreise, ohne Steuern, bei Haushalten in den 10 neuen EU-Ländern, die der EU am 01.05.2004 beigetreten sind, existierten. Die höchsten Preise wurden in Ländern wie Belgien, Irland, Zypern und Luxemburg erreicht. In der Industrie hingegen sind die Preise am höchsten in Ländern wie Irland, Spanien, Zypern und in der Slowakei. Für Endverbraucher werden die Strompreise inklusive Steuer interessant sein. Tabelle 1 zeigt dies. Die niedrigsten Preise für Strom findet man zurzeit in Bulgarien und Litauen. Die höchsten Preise für den Endverbraucher gibt es in Ländern wie Dänemark und Deutschland. Tabelle 2 zeigt die Struktur der Primärenergie in zwei Ländern: in Deutschland und in Polen, sowie die Menge der erzeugten Energie unter Berücksichtigung seiner Quellen und der Bevölkerungszahl in diesen Ländern im Jahr 2010. Tabelle 3 hingegen zeigt, wie viel Energie aus erneuerbaren Energiequellen in diesen Ländern erzeugt wird. Aus Tabelle 2 und Tabelle 3 kann man erkennen, dass Deutschland eindeutig der Spitzenreiter beim Einsatz erneuerbarer Energiequellen ist. Es ist bekannt, dass
Energie – reale Möglichkeiten zur Gewinnung sauberer Energie für Wohnhäuser
171
Deutschland, insbesondere bei der Nutzung von Wind- und Solarenergie, nicht nur in Europa führend ist, sondern auch eine führende Rolle im Weltmaßstab einnimmt. Price of electricity for the first half of 2010 for households (excluding tax) 2 500 kWh/a < Consumption < 5 000 kWh/a 0,18 0,16 0,14
EUR/kWh
0,12 0,10 0,08 0,06 0,04 0,02
ed en
Ki ng do m
U
ni te d
Fin la nd
Sw
om an ia
Sl ov ak ia
R
Po la nd
Po r tu ga l
ia
Li th ua ni a Lu xe m bo ur g N et he rl a nd s
yp ru s
La tv
C
Sp a in
Fr an ce
re ec e
Ir e la nd
G
an y er m
Es to ni a
G
ep ub lic
en m ar k
R
D
Bu lg ar ia
U
E
C
(2 7
ze ch
co un t
rie
Be lg iu
s)
m
0,00
Source: Eurostat
Abb.4. Die Strompreise für Haushalte in den EU-Ländern Price of electricity for the first half of 2010 for industrial consumers (excluding tax) 500 MWh/a < Consumption < 2 000 MWh/a 0,16 0,14
EUR/kWh
0,12 0,10 0,08 0,06 0,04 0,02
gd om Ki n
Un i te d
Fin la nd Sw ed en
ia
va kia Sl o
Ro m an
tu ga l
Po la nd
Po r
Li th ua ni a Lu xe m bo ur g Ne th er la nd s
La tv ia
Cy pr us
ce Fr an
Sp a in
d
ce re e
Ire lan
G
on ia Es t
Re pu bli c De nm ar k G er m an y
ga ria Bu l
Cz ec h
iu Be lg
UE
(2
7
co un trie
s)
m
0,00
Source: Eurostat
Abb.5. Die Strompreise für die Industrie in den EU-Ländern
Tabelle 1 Strompreise in der EU für Haushalte inklusive Steuer in € / kWh 2009
J. Magiera
172
[Europe’s Energy Portal http://www.energy.eu/#Domestic] Tabelle 2 Primärenergie für Deutschland und Polen im Jahr 2008 und Bevölkerungszahl in diesen Ländern im Jahr 2010 [ktoe] Steinkohle und Braunkohle Rohöl Erdgas Nukleare Energie RES GESAMT Bevölkerung im Jahr 2010
ENERGY POLEN 6 036 762 3 690 0 5 457 15 945 38 186 860
DEUTSCHLAND 50 040 3 087 11 314 38 305 29 744 132 490 81 742 000 Tabelle 3
Erneuerbare Energien in Deutschland und Polen [ktoe]
POLEN
DEUTSCHLAND
Energie – reale Möglichkeiten zur Gewinnung sauberer Energie für Wohnhäuser Solarenergie Biomasse Geothermie Wasserkraft Windenergie Source: Eurostat 2011
1 5186 13 185 72
173
735 23473 246 1801 3489
Bei der Gewinnung erneuerbarer Energien schneidet Polen schlechter ab, als andere Länder der Europäischen Union, vor allem der Länder Westeuropas. Erst in den letzten Jahren erfolgte in dieser Hinsicht ein deutlicheres Handeln. Zur Erfüllung der im EUBeitrittsvertrag unterzeichneten Verpflichtungen sollte in diese Maßnahmen wesentlich mehr investiert werden. Die zuvor diskutierten Faktoren und das Bewusstsein, dass die Energie im Laufe des Jahrhunderts die größte Wachstumsdynamik zeigen wird, ermutigt uns zum Handeln sowie zum Entwerfen und Aufbauen von Anlagen zur Wärmeenergiegewinnung auf Grundlage der Nutzung erneuerbarer Energieträger. Eine dieser Anlagen arbeitet bereits seit dem Jahr 2004. Seitens der Energieerzeugung gibt es Solarkollektoren, einen Biomassekessel mit Wärmetauscher, einen Gaskessel und auch eine Fotovoltaikanlage [3]. Das Diagramm der Anlage ist in Abb.6 dargestellt.
Abb.6. Schema der Hybridanlage für ein Zweifamilienhaus
Die erzeugte thermische Energie wird für Raumwärme und Warmwasserbereitung verbraucht. Die Solaranlage generiert Strom für die Notstromversorgung der Umwälzpumpen und für das Aufladen des Akkus mit einer Kapazität von 250 Ah. Die Anlage wurde auf einer kontinuierlichen Basis in den Jahren 2005 und 2006 betrieben. In Abb.7 werden in Form von einem Kurvenverlauf die im Gesamtjahr erzeugten Energiebilanzen und die thermische Energie, die verbraucht wurde, dargestellt.
J. Magiera
174
Energy [kWh]
Gas boiler Fireplace Solar collectors
1
2
3
4
5
6
7
8
9
10
11
12
Months of 2006
Abb.7. Energiebilanz der verbrauchten Energie eines Wohnhauses im Jahr 2006, das mit einer Anlage aus erneuerbarer Energiequellen ausgerüstet ist
Eine vergleichbare Anlage, die 50km südlich von Krakau in einer viel kleineren Ferienwohnung von ca. 110 m² montiert ist, ist in Abbildung 7 dargestellt. Für die Wärmeerzeugung wurden Solar-, Biomasse-Kessel und Elektroboiler installiert. Das Ferienhaus ist nicht in das Gasversorgungsnetz eingebunden. Zur Nutzung von Wärme und Warmwasser gibt es jedoch zwei Heizsysteme: Bodenheizung und Heizkörper, sowie Heizungsunterstützung des Swimmingpools. In diesem System werden aufgrund der periodischen Nutzung der Eigentümer keine kontinuierlichen Forschungen über erzeugte und verbrauchte Menge an Energie durchgeführt. In diesem Fall ist die Installation einer Solar-Kollektorfläche von ca. 14 m² für den Energiebedarf im Sommer deutlich überdimensioniert. Aber in dieser Zeit sind die Kollektoren nicht durch übermäßige Hitze und Stagnation belastet, da die überschüssige Wärmeenergie in den Swimmingpool geleitet wird.
Energie – reale Möglichkeiten zur Gewinnung sauberer Energie für Wohnhäuser
175
1 – solar collector 2 – fireplace with water jacket 3 – mixing thermostatic could water valve 4 – electric water heater 5 – buffer tank 6 – radiator heating 7 – floor heating 8 – divider 9 – plate heat exchanger 10 – swimming pool 11für – electric flow heater Abb.8. Hybridanlage für Heizung und Erzeugung von Warmwasser das Ferienhaus
In Abb.9 wurden teilweise die Ergebnisse der Arbeit der Anlage im Jahr 2010 sowie aktuelle Daten dargestellt. Beide Anlagen wurden im Jahr 2008 mit einem speziellen hybriden Steuer- und Energiebilanzsystem mit erneuerbaren Energien ausgestattet. [4]. Yields from solar collectors, biomass boiler and electric boiler
Abb.9. Energiebilanz für das Ferienhaus
Der Kern dieses Systems beruht darauf, einen kontinuierlichen Abgleich der Menge an Energie aus verschiedenen Quellen auf einen beliebigen Zeitraum zu ermöglichen. Es erlaubt ebenfalls die Darstellung einer Kostenbilanz unter Berücksichtigung der unterschiedlichen Kosten je Energieeinheit. Die Echtzeit-Grafiken sind sowohl Wärme-als auch Energieströmen, die aus verschiedenen Quellen generiert wurden, abgeleitet. Das System ermöglicht es, die gesamte Installation zu steuern, um eine wirtschaftliche Nutzung der thermischen Energie zu ermöglichen. Es besteht auch die Möglichkeit, die richtige
J. Magiera
176
Arbeitsweise des Systems in Echtzeit über das Internet zu beobachten und zu steuern. Die Abb.10 zeigt die Veränderungen der erzeugten Energie im Jahr 2010. Obtained energy: 1 – boiler 2 – solar collectors Consumed energy: 3 – heating 4 – electronic control 5 – swimming pool pump
11
outdoor temperature
3
2
5 4
Abb.10. Verlauf der Veränderung von Energie und Außentemperatur des Ferienhauses fortlaufend im Jahr 2010
Energie – reale Möglichkeiten zur Gewinnung sauberer Energie für Wohnhäuser
177
Schlussfolgerungen 1. Energie ist einer der wichtigsten Allgemeingüter für das menschliche Leben, jedoch werden die Kosten für ihre Herstellung und Nutzung in absehbarer Zeit trotz Bemühungen um Einsparung und Rationalisierung des Verbrauchs steigen. 2. Erneuerbare Energien können die Energienachfrage vermindern, wenn auch die Probleme der Energienachfrage der Welt und Europa in den kommenden Jahren nicht lösen. 3. Eine zur Wärmeerzeugung für Häuser und Wohnobjekte ausgelegte Anlage, die sich zur Wärmeenergieerzeugung mehr als einer Quelle bedient, einschließlich erneuerbarer Energiequellen hat den Praxistest gut bestanden. Mit einer Kostensenkung der Einrichtung wird eine weitere Verbreitung stattfinden. 4. Die Verwendung eines ausgebauten fernbedienbaren Systems der Energiebilanzierung und Steuerung, obwohl nicht in jeder Installation notwendig, ermöglicht jedoch eine wirksame Überwachung und Beaufsichtigung der Arbeit der Anlage durch das Internet. Anlagen dieses Typs, die auch als Fernlaboratorien dienen, erlauben, die Arbeit und energetisch erzielten Ergebnisse solcher Anlagen zu verfolgen. Literaturverzeichnis [1] Quaschning V.: Regenerative Energiesysteme, Carl Hauser Verlag , München Wien 1998. [2] Magiera .: Efficiency of Solar Collectors and Air Heat Pump under Real Conditions, Polish Journal of Environmental Studies, 15, (2004), 118-121. [3] Magiera .J, Wojtaś K., Turoń M.: Renewable sources of energy for house heating and usable warm water production, Environmental Protection Engineering, 32, (2006), 7179. [4] Neupauer K., Głuszek A., Magiera J.: Sterowanie nowego typu dla instalacji hybrydowych z odnawialnymi źródłami energii, Inżynieria i Aparatura Chemiczna, 49 (41), (2010), 87-88.
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
KRZYSZTOF NEUPAUER*
MÖGLICHKEIT DER BILANZIERUNG EINER HEIZUNGSANLAGE MIT DREI ENERGIEQUELLEN UND IHRE STEUERUNG IN DEM ONLINE-SYSTEM Zusammenfassung In die Arbeit wurden die Ergebnisse des Funktionierens von einer HybridHeizungsanlage, in der die Quelle der Wärmeenergie für Heizung und Warmwassererzeugung Solarkollektoren, Biomasse-Kessel in Form eines Kamins mit einem Wassermantel und elektrischer Boiler mit modulierter Leistung präsentiert. Die Anlage wird von einem neuen Steuersystem gesteuert, die das Datenerhebung, Beobachtung aller gemessenen Parameter und Steuerung der Anlage mit verschiedenen Energiequellen und von mehreren unabhängigen Empfängern ermöglicht. Das dargestellte System erlaubt die Bilanzierung der Energieflüsse sowohl während ihrer Herstellung als auch ihrer Verwendung, was die Bestimmung des Wirkungsgrades und der Kosten der Energiegewinnung aus diesen Quellen ermöglicht. Schlüsselwörter: Solarenergie, solarthermische Anlage, erneuerbare Energiequellen. 1. Einführung In letzter Zeit erfreuen sich integrierte Systeme mit mehr als einer Energiequelle, einschließlich erneuerbarer Energiequellen wie Solarkollektor, Biomasse-Heizkessel oder Wärmepumpe großer Beliebtheit[1]. Diese Anlagen benötigen ein entsprechendes Steuerungssystem, welches sie so steuern könnte, dass erneuerbare Energiequelle an Energiegewinnung Vorrang hat und konventionelle Energiequellen als Reserve betrachtet werden. Es gibt schon solche Steuersysteme, die die Arbeit der einzelnen Geräte und Heizungsanlagen steuern und bilanzieren können[2,3,4]. Das beschriebene Steuersystem erlaubt als einziger das Steuern von einigen Energiequellen on-line durchzuführen und mehrere Zahlen von Geräten und Anlagen zu leiten. Ein zusätzlicher Vorteil dieses Systems ist die Möglichkeit der Kaskadenverbindung von einigen Steuergeräten, was unbegrenzte Möglichkeiten der Steuerung von Anlagen gibt.
*
Technische Universität Krakau, Warszawska Str. 24, 31-155 Krakau, Polen
Möglichkeit der Bilanzierung einer Heizungsanlage mit drei Energiequellen 179 und ihre Steuerung in dem Online-System. 1.1. Heizungsanlage Das erste Steuersystem, das in Polen auf der Basis von dem DigiENERGY Gerät arbeitet, wurde im Dezember 2009 in einem Rekreationsgebäude 60 km südlich von Krakau (49° 47′ 6″ N, 20° 10′ 49″ E) installiert. Es steuert die Hybridanlage (Abb. 1), die aus drei Wärmequellen besteht: Solarkollektoren mit der aktiven Gesamtfläche von zirka 14 m2, Biomasse-Kessel in Form eines Kamins mit Wassermantel mit der Heizleistung von 9 kW, der mit Holz beheizt wird und aus einem Elektro-Durchlaufwasserheizer mit der Leistung von 18 kW, mit einer flexiblen Steuerung der Heizleistung [5]. Jede Wärmequelle übergibt die Wärmeenergie an einen Pufferspeicher Typ „SISS”5 mit dem Gesamtvolumen von 550 Liter, innerhalb dessen sich ein Gefäß mit dem Volumen von 150 l für die Erzeugung des warmen Nutzwassers befindet. In der Anlage befinden sich zwei Heizungskreisläufe: Fußbodenheizung im Erdgeschoß und Konvektionsheizung im Dachgeschoß. Außer der Fußbodenheizung wurde in den Badezimmern die Körperheizung für Badezimmer montiert. In der Urlaubszeit, wenn man kein warmes Nutzwasser braucht und die Solarenergie größer ist, ist der Überschuss von der Wärmeenergie in ein Schwimmbad mit der Größe von 3,15 x 6 x 1,5 m übergeben. Ein zusätzlicher Behälter des warmen Nutzwassers mit dem Volumen von 80 l und Heizleistung von 2 kW wurde in dem Kreislauf des warmen Nutzwassers installiert. Seine Aufgabe besteht in der Vorbereitung des warmen Nutzwassers ohne den Pufferspeicher zu heizen.
Abb. 1. Schema des Hybridsystems von erneuerbaren Energiequellen mit der Anordnung von Meßpunkten Elemente der Anlage: 1 – Solarkollektor, 2 – Kamin mit Wassermantel, 3 – mischendes thermostatisches Ventil, 4 – elektrischer Wassererhitzer, 5 – Pufferspeicher, 6 –Körperheizung, 7 – Fußbodenheizung, 8 – Verteiler, 9 – Plattenwärmetauscher, 10 – Schwimmbecken, 11 – Elektro-Durchlaufwasserheizer 5
SISS – Speicher in Speicher System
K. Neupauer
180 2. Das DigiENERGY -Steuersystem
2.1. Hardware Die Hardware des Steuersystems in seiner Grundversion besteht maximal aus 8 Meßmodulen und/oder Ausführungsmodulen. Die Modulmenge hängt von der Anzahl der Meßpunkte im System ab. Es besteht keine Möglichkeit der Modifizierung oder des Ausbaus von dem schon fertigen System. Das grundsätzliche und wichtigste Modul in jedem DigiENERGY System ist das Verwaltungsmodul mit dem Namen „Web Serwer”. Es füllt zwei Funktionen aus: es verbindet alle anderen Module miteinander, sowie ermöglicht die Kommunikation des Systems von der Entfernung. Jedes Modul für die Temperaturmessung bedient maximal 9 Widerstandsfühler PT1000. Das Ablesen von den digitalen logischen Signalen z.B. Impulse aus dem Durchflussmesser erfolgt durch die Karte von Ein- und Ausgängen. Jeder Randstecker wird für das Ablesen oder Aufgeben von logischen Zuständen gebraucht (0-1), also zum Zusammenrechnen von Impulsen oder zum Steuern von anderen Anlagen mit Hilfe des +24V Signals. In solcher Karte befinden sich 8 Ein/Ausgänge. Das Modul mit analogen Ein/Ausgängen besitzt Randstecker in der Version mit der Spannung 0 – 10 V oder mit Strom 4 – 20 mA. Es dient der fließenden Steuerung z.B. von der Leistung des Kessels oder zum Ablesen der Werte von dem Pyranometer oder Hygrometer. Das Relaismodul besitzt 15 Halbleiterrelais für die Spannung von 230 V, die z.B. für das Einschalten der Pumpen, der Dreiwegeventile oder Elektroventile benutzt werden können. Der Durchsatz der Umlaufpumpen kann dadurch gesteuert werden, dass sich jedes Relais für die Arbeit im PWM –Modus (Pulse Width Modulation) eignet. 2.2. Software Die Kommunikation mit dem DigiENERGY Steuersystem erfolgt durch einen Webbrowser [6]. Es ist der einzige Kommunikationsweg für die Aufsicht des Betriebs, für die Datenvisualisierung, Programmierung und Änderung der Daten. Das System verfügt über das Vierstufenmenu. In der Kartei „Informationen” - „Übersicht” befindet sich eine Ansicht der untersuchten Solaranlage (Abb. 2), die die Beobachtung der momentanen Temperaturwerte, Leistung, Volumen vom Durchsatz, Pumpendurchsatz sowie atmosphärische Bedingungen d.h. die Stärke der Sonnenbestrahlung sowie relative Feuchtigkeit ermöglicht. Das Schema wird abhängig von der Zahl der installierten Fühler und Empfänger durch das System automatisch modifiziert. Nicht alle programmierten Funktionen werden in der Visualisierung dargestellt. Das Symbol des Heizungskreislaufs wird zum Beispiel nur dann dargestellt, wenn mindestens ein Fühler diesem Kreislauf zugeordnet sein wird. Die Software des Steuersystems erlaubt Schema für beliebige Anlagen zu erstellen zu konfigurieren, was durch die Konfiguration von konkreten Anlagen wie: Kamin, Kessel, Wärmepumpe, Solarkollektor, Speicher erreicht wird.
Möglichkeit der Bilanzierung einer Heizungsanlage mit drei Energiequellen und ihre Steuerung in dem Online-System.
181
Abb. 2. Visualisierung der Betriebsdaten der Anlage, Zustand vom 23.05.2011. 11:26 Uhr
Das Steuersystem verfügt zusätzlich über die „Zählerstände” - Funktion (Abb. 3) für alle Heizungskreisläufe. Diese Funktion erlaubt aktuelle Wärmeleistung und elektrische Leistung für jeden Kreislauf sowie das Nutzwasservolumen zu verfolgen. Die energetische Bilanz wird nach dem Einstellen eines Zeitabschnittes generiert. Wenn die Einheitspreise für 1 kWh und 1 m3 des Wassers eingetragen werden, können Kostenbilanzen für den gewählten Zeitraum generiert werden.
182
K. Neupauer
Abb. 3. Beispiel der energetischen und Kostenbilanz im Zeitraum vom 1.05.2011 bis zum 10.05.2011
In der Kartei „Temperaturverlauf” und „Kollektorverlauf” können Diagramme der Tagestemperatur gleichzeitig für einige Meßpunkte am Kollektor und Speicher generiert werden. Diese Diagramme sind dann nützlich, wenn Probleme mit dem Betrieb der Solaranlage auftauchen oder wenn er optimiert werden soll. Das Diagramm auf der Abb. 4 zeigt, dass die Umlaufpumpe des Kollektors (Kurve 4) zwischen 9:30 und 17:00 Uhr automatisch eingeschaltet wurde, denn die Temperatur des Kollektors (Kurve 1) größer als die Temperatur in dem Speicher (Kurve 2) war. In den Stunden 9:30 – 12:00 sowie 15:00 – 15:30 wurde das Dreiwegeventil (Kurve 5) geöffnet und die Wärme von dem Kollektor wurde zum Speicher (Puffer) und nicht zum Schwimmbecken weitergeleitet.
Möglichkeit der Bilanzierung einer Heizungsanlage mit drei Energiequellen und ihre Steuerung in dem Online-System.
183
Abb. 4. Ausgewählter Tagestemperaturverlauf am 23.05.2011 Legende: 1 – Temp. beim Auslauf des Kollektors, 2 – Temp. im oberen Teil des Speichers, 3 – Aussentemp.,4 – Betriebszeit der Umlaufpumpe des Kollektors, 5 – Zeitraum der Öffnung von dem Dreiwegeventil (Durchfluss zum Speicher)
Die Diagramme der momentanen Leistungen in den einzelnen Kreisläufen sind in der Kartei „Energieverlauf” (Abb. 5) abrufbar. Aus dieser Abbildung lässt es sich ablesen, dass der Kamin die Wärmeströmung zum Speicher (Puffer) in der Zeit von 17:00 bis 24:00 übergeben hat. Die Wärmeströmung aus dem Kollektor wurde an diesem Tag sowohl zum Speicher als auch zum Schwimmbecken übergeben.
184
K. Neupauer
Abb. 5. Ausgewählte Tagesverläufe der Leistung am 2.05.2011 Legende: 1 – Wärmekapazität des Speichers, 2 – Wärmeleistung des Kamins, 3 – Wärmeleistung des Kollektors, die zum Schwimmbad übergeben wird, 4 – Wärmeleistung des Kollektors, die zum Speicher übergeben wird, 5 – Betriebszeit der Umlaufpumpe des Kamins, 6 – Betriebszeit der Umlaufpumpe im Kreislauf des warmen Nutzwassers, 7 – Betriebszeit der Pumpe im Heizungskreislauf, 8 – Verfügung über warmes Nutzwasser, 9 – Höhe der Innentemperatur – am Tag 20°C, in der Nacht 18°C
Die Software erlaubt Diagramme von einem Jahr zu generieren, wodurch der Verlauf von der Außentemperatur, Herstellung und Verbrauch der Energie für einzelne Kreisläufe sowie Daten aus den letzten zwei Jahren verglichen werden können. Diese Diagramme befinden sich in der Kartei „Jahresverlauf”. Bei dem Vergleich der Kurven auf der Abb. 6 für das Jahr 2010 und 2011 sieht man, dass es im April und Mai 2010 schlechtere atmosphärische Bedingungen gab und dadurch zweimal kleinere Gewinne der Wärmeenergie aus dem Kollektor erhalten wurden.
Möglichkeit der Bilanzierung einer Heizungsanlage mit drei Energiequellen und ihre Steuerung in dem Online-System.
185
Abb. 6. Jährliches Diagramm der Wärmeenergie, die aus dem Solarkollektor erhalten wurde.
Das besprochene System erlaubt Zeiträume für die Heizung des Gebäudes zu definieren und die Menge des warmen Nutzwassers sowie die Priorität für den Energiegewinn aus ausgewählten Wärmequellen zu bestimmen. In dem Zugangsmenu kann der Bediener in der Kartei „Konfiguration” Kenndaten für den Betrieb der Anlage einstellen, die Kalibration der Messgeräte durchzuführen sowie „manuellen” Betrieb jedes Systems und Geräts durchzuführen. Das System gibt breite Möglichkeiten der Registrierung von Betriebsdaten der Solaranlage. Es ermöglicht nicht nur laufende Registrierung der Systemdaten, sonder auch Einführung von zusätzlichen Wechseldaten und Speichern dieser Daten in *.csv Dateien, die dann leicht z.B. in MS Excel zu bearbeiten sind. 3. Ergebnisse des Hybridbetriebs der Heizungsanlage Auf der Abb. 7 wurden in Form eines Diagramms Beispielergebnisse des Hybridbetriebs im Monat Mai dargestellt. Die erzeugte Wärmeenergie stammte in 78% von dem Solarkollektor und in 22% aus dem Kamin. Dritte Reservequelle der Wärmenergie ist der elektrische Kessel, der in dem analysierten Monat nicht arbeitete. Gegen 2/3 der in
K. Neupauer
186
dieser Zeit erzeugten Energie wurde für die Heizung des Gebäudes und des Schwimmbads sowie für warmes Nutzwasser genutzt. Die Summen von Energiezufuhr und Abfuhr sind nicht gleich, weil der Speicherzustand am Anfang und Ende war verschieden.
Abb. 7. Im Mai 2011 erzeugte und genutzte Wärmeenergie.
Das analysierte Wohnobjekt ist nicht typisch für konstantes Bewohnen und Benutzen der Energie. Die dargestellten Ergebnisse zeigen nur einen Teil der Möglichkeiten des besprochenen Steuer- und Bilanzierungssystems. Mit Hilfe von diesem System arbeiten zur Zeit in Polen auch andere außer der hier dargestellten Anlagen: ein Zweifamilienhaus in Mszana Dolna, großes Krankenhausgebäude in Krakau, eine Versuchsanlage mit Solarkollektoren in dem Labor der Solartechnik von Krakauer Technische Universität. Dieses System wird gerade auch in dem Zentrum für erneuerbare Energiequellen an der Bergbau und Hüttenakademie in Krakau – Miękinia sowie in Połtawska Wissenschaftsakademie in Ukraine montiert. Das System wird da Anlagen mit mehreren erneuerbaren Energiequellen steuern.
Möglichkeit der Bilanzierung einer Heizungsanlage mit drei Energiequellen und ihre Steuerung in dem Online-System.
187
4. Zusammenfassung Das dargestellte Steuersystem ist imstande, den Betrieb von jeder Heizungsanlage zu kontrollieren. Es kann auch die Hybridheizungsanlagen steuern, die aus mehreren Heizungskreisläufen bestehen und für die Energieerzeugung mehr als eine Energiequelle benutzen. Es gibt die Möglichkeit der Einstellung und der „on-line” – Kontrolle von den momentanen Betriebsdaten der ganzen Anlage. Dank der „Zahlerstände” Funktion kann der Bediener jederzeit kurz- und langzeitige energetische und Kostenbilanzen erstellen. Es gibt auch die Möglichkeit des graphischen Vergleichs der Werte mit den analogischen Werten vom letzten Jahr. Das System kann sehr schnell den Verlauf von ausgewählten Werten erstellen, was für den Benutzer für die Optimierung des Betriebs der Anlage und die Diagnostik von eventuellen Fehlern wichtig ist. Korrekt installiertes und eingestelltes System wird die Anlage rationell steuern, der Verbrauch der Energie wird optimaler sein, was in der Konsequenz Kostensparen für den Benutzer der Anlage bringen wird. Literatur [1] Dissertation, Michalak P. „Badania efektywności energetycznej budynku użyteczności publicznej wykorzystującego odnawialne źródła energii“, Akademia Górniczo-Hutnicza Im. Stanisława Staszica w Krakowie, 2008. [2] www.suntime.pl/innowacje.html, Zustand vom 14.07.2011 [3] Podlejski K., FELCENLOBEN Ł.: „Zastosowanie środowiska LabVIEW do sterowania pracą instalacji grzewczej z odnawialnymi źródłami energii”, Prace Naukowe Instytutu Maszyn, Napędów i Pomiarów Elektrycznych Politechniki Wrocławskiej Nr63, Studia i Materiały Nr 29, 2009r. [4] www.vaillant.de/Produkte/Gasheizung/Regelung/produkt_vaillant/vrnetDIALOG.html, Zustand vom 14.07.2011 [5] Neupauer K., Magiera J., Głuszek A.: „Efektywność wykorzystania kolektorów w konwersji energii słonecznej do energii cieplnej”, Inżynieria procesowa w ochronie środowiska, Wydawnictwo i Drukarnia Świętego Krzyża, Opole 2009, 71-80. [6] Neupauer K., Głuszek A., Magiera J.: Inżynieria i Aparatura Chemiczna, Nr 3/2010, (2010), 87-88.
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
STEFFEN STUENKEL, GUENTER WOZNY*
CO2 CAPTURE FOR THE OXIDATIVE COUPLING OF METHANE PROCESS Abstract The gas treatment of reaction products is a crucial step in the process chain from the raw material to the product. CO2 is produced by the oxidative coupling of methane (OCM) as an unwanted by-products or waste product and should be removed in an early stage. For the gas treatment are the energy requirement, the operation and investment cost a curial factor, that effect the overall process performance. Beside economic aspects, designing a sustainable process is the aim by utilization of CO2 in a dry-reforming reactor. Due to the concurrent engineering procedure, the OCM process concept was developed and the process units were investigated parallel. The membrane reactor and the fluidized bed reactor concept were investigated for the oxidative coupling of methane and a C2+ yield of nearly 17% was achieved with standard catalysts. Beside the ethylene up to 25 mol% CO2 were produced and could remove by a state of the art chemical absorption process using 5 MJ/kgCO2 specific thermal energy. This was halved by using a polyimide membrane for pre separation. In this article the technical feasibility of the general process concept for the OCM, including down streaming is discussed and presented. Keywords: Oxidative Coupling of Methane, CO2 capture, CO2 utilization 1. Motivation Oxidative Coupling of Methane – Miniplant scale In the wake of shortages and price rises for crude oil in recent decades also, the search for alternative raw materials in the chemical industry is of great importance. Hence opens the oxidative coupling of methane (OCM), a new route for the petrochemical industry in order to produce ethylene from methane natural gas or biogas [1]. Thus, this process forms the base to activate a large new feedstock for a variety of chemical products [2]. The OCM reaction is a heterogeneously catalyzed gas phase reaction at temperatures of up to 800° C, in which unwanted by-products such as carbon dioxide produced additionally to the desired product ethylene and must remove from the reaction product [3]. Although several alternatives can be found for OCM processes in the literature [4], an efficient OCM process
*
TU Berlin, Straße des 17. Juni 135, Berlin 10623, Germany
CO2 capture and utilization for the oxidative coupling of methane process
189
is not installed in the industry yet. So far all proposed processes associated with high cost and energy demand, as well as cost-intensive gas processing part. To overcome this limitation, in this work, the novel approach of the Concurrent Process Engineering is used to investigate the whole process from raw material to the product. After extensive preliminary investigations, a flexible mini-plant system was designed, built and operated. To evaluate effects of removal and the OCM catalyst life, the whole process is also being studied in a long time study. Due to the moderate ethylene yield for currently available OCM catalysts of 30% is the goal, besides the improving of the catalysts, the downstream gas conditioning process and make the process economical. This can only be achieved through the development of integrated process concepts that are based on energetically and economically enhanced processes and the usage of side products. The development process is part of the DFG Cluster of Excellence "Unifying Concepts in Catalysis" (UniCat).
Figure 1 Process flow diagram of the OCM process, including processing
The challenge is to develop a flexible process in mini-plant scale to respond flexibly to changing one process requirements. As example here, the catalyst development is mentioned. By using different catalysts, a different amount of CO2 is produced, which are separated in the downstream gas processing needs. Due to time savings in the process of development, the entire OCM process is set up in parallel in a mini-plant, and process synthesis for optimal OCM process is carried out in parallel. So the whole process for concurrent engineering has been divided into three units (see Figure 1): the reaction unit, the gas cleaning unit and the product separation unit, which are examined in parallel. Based on a holistic view and approach of the simultaneous process of development, the requirements relating to the separation task and purities for each unit process having regard to their interactions with each defined. Here were the CO2 separation, the maximum limit for the CO2 produced fixed at 25mol%, this border can be lowered down to 15mol% CO2 by the latest developments. Therefore, the CO2 separation is investigated in the concentration range of 15mol% to 25 mol%. Based on the process schematic of the flow chart in Figure 1 as well as studies on the macro-and micro-kinetics of the catalysts, the process conditions and process requirements for CO2 separation were determined which are given in Table 1. Here, the maximum expected CO2 concentration of 25 mol% is set as the basis for the design task. Table 1 Process conditions and requirements for CO2 separation Gas temperature 40 ° C
F-Faktor 0.7Pa 0.5
Raw gas pressure 32 bar
CO2
C2H4
CH4
N2
22mol%
15mol%
15mol%
48mol%
CO2 removal 90 %
The removal of acid gas components from the reaction product stream is an important process step in the chain of gas processing, to achieve the goal in the purity of the product.
S. Stuenkel et al.
190
For the CO2 capture process, several process concepts are available. The chemical absorption, however, preferred for selective amine scrubbing and washing are state of the art in this field [4]. Moreover, new detergents such as hyper branched polymers have a promising object of research and are examined parallel in a vapor-liquid equilibrium apparatus and later in the miniplant. Alternative separation methods such as membrane processes for CO2 capture are also currently being investigated in parallel. 1.1. Design of a Miniplant for the OCM for technical feasibility demonstration The miniplant technique is well-known in process synthesis, to obtain fundamental information experimentally in an early process synthesis stage. In this approach, a topdown process synthesis is applied and the process has been developed in miniplant scale to demonstrate its technical feasibility. Due to the concurrent engineering, the downstream requirements affect the reaction unit and the catalyst, especially the C2 yield, the C2- and CO2 selectivity, the methane conversation and dilution concentration of the reaction gas. The reaction products that have to be separated in the downstream of the OCM process and their chemical families, properties and the feed conditions are given in Table 2, 3 and 4. Table 2: Defined feed gas concentration (in mol%) for the process section O2 C2H4 C2H6 CO2 H2O Unit CH4 Reaction 60 – 70 20 Purification 45 10 10 25 Separation 60 13 13 Chemical Alkan Alken/ Alkan Acid family: olefine gas Treated Raw material/ Raw Product Byproduct Remove Remove recycle material Unit Reaction Purification Separation
Table 3: Process conditions for each unit Pressure range 1 to 5 bar 1 to 32 bar Up to 32 bar
N2 20 – 10 20 -10 14
Inertgas
Temperature range 30 to 900 °C 30 to 100 °C Down to -100 °C
The defined ranges of the process- and stream conditions for the units are presented in Table 3. Those conditions are defined of a literature study for the state of the art processes for each unit operation for a starting value of the design procedure. 1.2. The downstream process The reaction gases have to be conditioned in the downstream process, that consists of a phase separation unit, a carbon dioxide removal unit, and a product separation unit, as recommended by various authors [5] . Concerning the simultaneous design and construction in the miniplant, with the state of the art separation processes were started as a base case and was improved during the project. The base case purification unit consists of an amine based absorption process for the carbon dioxide separation. For the separation unit, a cryogenic distillation for the product separation was found to be the state of the art solution and is investigated only theoretically.
CO2 capture and utilization for the oxidative coupling of methane process Components
State 25°C, 1atm
Methane Ethylene Ethane Nitrogen Carbon Dioxide
vapour vapour vapour vapour vapour
Table 4: Component properties [6] kinetic Melting Boiling Molar Diameter Point Point mass σ [A] 1atm[°C] 1atm[°C] [g/mol] 16 28 30 28 44
-162 -182 -103,72 -169,18 -89 -183 -195 -210 Over critical
3,8 4,228 4,388 3,667 3,996
Critical Temperatur e Tc [K] 191,1 282,4 305,4 126,2 304,2
191 Critical Pressure Pc [bar] 45,8 50 48,2 33,5 72,8
It shows that an increase of the downstream operating pressure up to the lowest critical pressure of the components (32 bar) results in a higher boiling point of the hydrocarbons and decreases the energy demand in the cryogenic distillation. The high operation pressure increases the absorption efficiency as well and is therefore applied in the purification unit too. Due to the high absorption pressure and a high CO2 partial pressure a physical solvent could be used. Nevertheless, to improve the selectivity was a chemical solvent chosen for the base case and the state of the art absorbent Monoethanolamine is used. Optimal thermal energy demands of 3-4 MJ/kgCO2 of solvent rgeneration are reported for standard absorption processes using Monoethanolamine (MEA) [7]. As base case for the acid gas removal unit a chemical absorption process with 30 wt% of MEA was found and modeled with a detailed rate base model and an electrolyte NRTL approach, described below. Thermal energy for solvent regeneration was found with 3.6 MJ/kgCO2 for the design and base case and was improved by developing a hybrid separation process. 2. Process synthesis for alternative acid gas removal unit The design task of this unit is to decreases the carbon dioxide concentration from the product stream by 90% from the initial CO2 concentration as shown in Table 1. Several strategies are known for process synthesis. In this research the Knowledge-Based Separation System Synthesis approach purposed by Barnicki and Fair [8] was followed. The general separation tasks can be classified by: 1. enrichment, 2. sharp separation and 3. Purification. The enrichment means the increase in one concentration of a species in one of the product stream [8]. Whereas in this context two high-purity product stream results by sharp separation. The classification of this design task is done with the ration of the key components in the product streams that has to be higher than 9 or less than 0.1 respectively. The key components in the acid gas removal part are the carbon dioxide (CO2), that has to remove and the ethylene (C2H4) as the product. The separation ration for Carbon Dioxide results to: c Rawgas (1) S CO = CO Puregas cCO 2
2
2
Purification, in this context represents the separation task for the removal of low concentration of one component. For this case the removal of CO2 of less then 10 mol%. For the OCM design case, the sharp separation has to be considered in general. The unit operation for the separation tasks applicable for all three kinds of separation tasks are presented in Table 5, and discussed in detail in the next following sections.
S. Stuenkel et al.
192
Table 5: Process alternatives and indicators for separation processes [8] Process alternative Indicator Cryogenic Distillation Relative volatility Physical Absorption Separation factor for gas solubility using Henry’s approach Molecular Sieve Adsorption Difference in shape size and kinetic diameter Equilibrium limited Adsorption Ratio of the equilibrium loading for the key components Membrane Separation Separation factor, ration of the component permeability Chemical Absorption Chemical family of the components Condensation Difference in normal boiling point Catalytically conversion Product Worth
2.1. Cryogenic Distillation The application of cryogenic distillation is economically only for high throughputs and a volatility for the key components higher than two [8]. The relative volatility for the key components at 32 bar is αCO2/C2H4 ≈ 1. For a system pressure of 1 bar results the relative volatility to αCO2/C2H4 ≈ 3. Therefore the low temperature range requires high-grade materials and high energy demand for the low temperature production, what makes the process uneconomically. 2.2. Physical Absorption For the physical absorption a selectivity of the key components of Sabs CO2/C2H4 > 4 is recommended [8]. The selectivity of the key components can be calculated using equation 2. For two common physical absorbents, methanol and water the Henry approach, equation 3 and 4 respectively can be used to obtain the selectivity. For methanol the selectivity results to SabsCO2/C2H4=1.11 and for water the selectivity results to SabsCO2/C2H4=1.03. abs SCO = 2 / C2 H 4
X CO2 X C2 H 4
(2)
X CO 2 He CO 2 = Y CO 2 P (3) X C 2 H 4 He C 2 H 4 = YC 2 H 4 P (4) 2.3. Chemical absorption Chemical absorption is favored for species that contains acid-based functional groups, like the Carbon Dioxide and for those components with low partial pressure in the gas stream. 2.3. Molecular sieve and equilibrium adsorption To consider molecular sieve adsorption, the species have to classify by shape size and their kinetic diameters. The kinetic diameters of the components from Table 4 can be arranged according to size: σC2H6>σC2H4>σCO2> σCH4>σN2. The physical size properties of the commercial available adsorbents are listed in Table 6. It can be seen, that the kinetic
CO2 capture and utilization for the oxidative coupling of methane process
193
diameter of the Carbon Dioxide is in the middle of the components and all diameters are very close to each other. Table 6: Commercial available adsorbents [9]
category 5 4 3 2 1
nominal aperture size [A°] 3 4 5 8 10
Zeolite Type 3A Linde 3ADavison 4A Linde, 4A Davison 5A Linde, 5A Davison 10X Linde 13 X Linde, 13X Davison
Equilibrium based adsorption is only suitable for species concentration less than 10 mol% and for a selectivity of the key components higher than 2. It was found, that the selectivity for equilibrium loading of the key components results to SadsCO2/C2H4=1.74, for a 5A molecular sieve [9]. 2.4. Membrane processes Considering membrane as an economical feasible separation technique, the selectivity of the key components should be lager than 15 [8]. The selectivity for the key components can be obtained by the ratio of the permeability with equation 5.
SCO2 / C 2 H 4 =
PCO2 PC 2 H 4
=
DCO2 SCO2 DC 2 H 4 SC 2 H 4
(5)
The applications of membranes in gas processing are rare and therefore it is not astonishing, that no selectivity for the key components could be found in the literature. Provisionally the carbon dioxide/methane selectivity was taken into account, which was found to be higher than 15 [10]. Therefore, membrane separation was considered and new measurements were performed for the design specification by the Helmholtz – Zentrum Geehstacht, Germany. 2.6. Condensation and catalytically conversion The separation by condensation should be considered when the difference in normal boiling points of the components is higher than 40 K. The separation by catalytic conversion is only suitable for impurities. 3. Alternative separation Process Synthesis The recommended unit operations for sharp separation are physical separation, cryogenic distillation, molecular sieve and equilibrium adsorption. It shows that none of the process alternatives of Table 5 could fulfill the design task in a single step for sharp separation. Therefore, a two-step process was developed: An enrichment step for the reduction of the Carbon Dioxide down to 10 mol% and a purification step for last 10 mol%. The processes alternatives and their indicators for choosing the alternative are shown in Table 5 and discussed in section 2. It was found that the best process alternative is the design of a hybrid separation process, consists of a membrane and a chemical absorption unit, which is described and discussed in the next sections.
S. Stuenkel et al.
194
3.1. Absorption processes The absorption technique for Carbon Dioxide separation is well developed and industrially available. Physical absorption processes like the UOP Selexol® Process or the Lurgi Rectisol® Process are known, which are using dimethyl ether and cold methanol, respectively. Those physical absorption processes cause high product losses of more than 30 %, due to a nearly similar solubility of the Ethylene and the Carbon Dioxide in the absorbent. Therefore, only chemical absorption liquids like Monoethanolamine (MEA), Diethanolamine (DEA) and Methyldiethanolamine (MDEA) or a mixture of them are applicable for the purpose of the OCM miniplant. Those chemicals are used in amine scrubbing processes like the aMDEA® Process in different concentration ranges. Standalone rigorous simulations for the absorption process of the miniplant were carried out in Aspen Plus®. The in-built ELECNRTL model, with activity coefficients of the electrolyte NRTL approach for the liquid phase was used and the Redlich-Kwong equation of state (EoS) was applied for the gas phase. Furthermore, concentration-based reaction kinetics was used and a rigorous rate-based model for absorption in packed column was applied. Chemical absorption processes are high selective on the one hand, but suffer on the other hand on a high-energy rates, which is required for regeneration of the loaded solvent in the desorption part. A parameter study was done to evaluate the influence of solvent regeneration, reboiler duty, solvent flow rate and solvent concentration. Parameterstidy -solvent flow and degree of regeneration
Reboiler duty
30
CO2 Capture [%] /
40
Degree of regeneration 20 10
Degree of Regeneration [%]
50
Reboiler Duty [kW] / Regeneration [%]
Parameterstudy on CO2 Capture - for constant solventflow
100
60
80
60
CO2 Capture
40
Degree of Regeneration 20
60
65
70
75
80
10
20
30
40
50
Reboiler duty [KW]
Solvent flow rate [kg/h]
Figure 2: Parameter study – left: effect of flow rate for a constant design task, right: effect of reboiler duty of regeneration and carbon capture
At first, a simulation study with theoretical stages was carried out to investigate the effect of the flow rate on the energy demand. The range of the flow rate were varied and it was found, that increasing flow rate lowers the reboiler energy demand for absorbent regeneration. The influences of absorbent regeneration and reboiler duty on carbon capture were investigated with the more detailed model, including rate-based calculations with chemical reactions. It was found, that a regeneration of 50 to 60% is optimal to reach the an energy demand of 3.15 MJ/kgCO2 for 90 % carbon capture with 30 wt% MEA solution, which fulfill the design task of CO2 reduction from 10 to 1 mol% for the process alternative.
CO2 capture and utilization for the oxidative coupling of methane process
195
3.2. Membrane unit The advantages of a membrane unit are the easy operation and a short start up and shut down time caused by their small size. Those units are very flexible in use, due to the modular design. There is no solvent needed and no further treatment, if the selectivity is high enough. For vapor/gas membrane separation, different kinds of materials are available differs in physical mechanism: • Polymeric membranes: Rubbery or glassy polymers, with different solubility and diffusion properties for carbon dioxide and hydrocarbons. • Molecular sieves: Adsorption effects, separation by different molecule dimensions. Glassy and rubbery polymeric membranes are preferred for the carbon dioxide separation and hydrocarbon recovery, but they are not used in olefin production. For the membrane unit, a glassy and a rubbery carbon dioxide selective membrane is investigated. The membrane unit was modeled with the solubility-diffusion model and an Aspen Custom Modeler® unit was developed for a one dimensional, dense membrane. The PengRobinson EoS is used for the fugacity, and the free-volume theory for the calculation of the permeability was applied. As further non-ideal effects, concentration polarization, the Joule-Thomson effect and pressure loss for low Reynolds numbers are considered. The membrane unit is calculated using geometry and free-volume parameter for an envelope type membrane module of the Helmholtz-Zentrum Geesthacht, Germany. The carbon dioxide concentration could be reduced from 22 to 10 mol % with an one-stage membrane unit. The product losses are calculated as the ratio of the molar ethylene purge to the incoming ethylene mole flow. These losses were found in the range of 30 % for the rubbery membrane material and of 10% for the glassy membrane. While the glassy membrane has a higher carbon dioxide/ethylene selectivity, with a low trans-membrane flux. This causes higher membrane area, in comparison with the rubbery membrane material. The product loss has to be evaluated economical in comparison to the energy saving, when using an absorption process. The product loss can be reduced by a two-stage membrane process and is described in the next section. 3.3. The two stage membrane process The application of a two-stage membrane unit can reduce the product loss, whereas the carbon dioxide reduction fulfills the design task for the alternative process design. The combination of rubbery and glassy membrane material was studied and it was found that using the rubbery membrane material in the first unit and the glassy material in the second unit could decrease the product loss. The results are presented in Table 7. Resulting by the simulation study, optimal feed temperature and operating pressure was found and the ethylene losses were reduced down to 2%. The electrical energy demand for the recompression unit of the second stage was found with 2.7 MJ/kgCO2. Table 7: Technical requirements of the two-stage membrane process Membrane area Feed Feed Membrane area Feed 2nd stage 1st Stage Temperature Temperature Pressure 2nd 1st Stage 2nd stage stage 1.75 m² 20 °C 1 m² 15°C 32 bar
Product losses 2 vol%
S. Stuenkel et al.
196
Thus, the energy demand for the entire sharp separation increases, while the product loss decreases. The economical optimum of energy demand and product losses has to be found in this project. 4. Miniplant application Simultaneously to the process synthesis, a miniplant was designed and builds for model validation and investigation of the entire process. The experimental studies showed long term effects like regeneration effects of the absorbent, deactivation of the catalyst and allows validating the theoretical models. The units are studied at first stand-alone, to optimize them. Table 8 summarizes the basic engineering details of the purification part, especially the absorption column design. Table 8: Technical and hydrodynamic operation conditions of the absorption/desorption process
Packing height [m] 5
Column diameter [mm] 40
Max. F- Gas Packing Packing feed section capacity factor [-] [Pa0,5] [kg/h] [m²/m³] 0.6 21 50 450
Maximum liquid load [m³/m²h] 55
Top pressure [bar] 32
The simultaneous design and construction requires high flexible units, to handle the changes in carbon dioxide concentration. This target can be reached by using membrane units for enlargement or reduction the capacity easily. The miniplant is shown in figure 3, left side. The results of comparison of the concentration profile for the experiments with the simulation presented in figure 3 right side. This experiment is based on the reference absorbent 30 wt% Monoethanolamine. Absorption CO2 concentration profile
Absorption column height [m]
6
5
4
Aspen Plus results Experimental results
3
2
1
0 0
5
10
CO2 gas concentration [mol%]
15
20
CO2 capture and utilization for the oxidative coupling of methane process
197
Figure 3: left side: Photo of the installed absorption and desorption process in a miniplant scale, right side: imulation results and comparison with experiments of the concentration profile
5. Conclusion – The hybrid separation process Facing the required purity of the product stream and the lack of high thermal energy demand of 3.6 MJ/kgCO2 for the regeneration step in an absorption process, the use of a membrane unit for enrichment can reduce the overall thermal energy demand to 1.6 MJ/kgCO2 for the entire sharp separation of carbon dioxide from the OCM Product stream. To reduce the ethylene loss from 9 % down to 2%, a second membrane stage can be applied with different membrane materials. For the combination of a rubbery membrane and a glassy membrane with optimal process conditions of Table 7 an electrical energy demand of 2.7 MJ/kgCO2 can be reached. Carbon dioxide utilization alternatives are given and have to evaluate in further research. Acknowledgements The authors acknowledge the support from the Cluster of Excellence "Unifying Concepts in Catalysis”, coordinated by the Berlin Institute of Technology and funded by the German Research Foundation (DFG). Literature [1] A. Behr, A. Kleyensteiber, U. Hartge, Chem.Ing.-Tech. 2010, 82, 201. DOI: 10.1002/cite.200900122 [2] K. R. Hall, Catalysis Today 2005, 106, 243. DOI: 10.1016/j.cattod.2005.01.176 [3] S. Jaso, H. R. Godini, H. Arellano-Garcia, G. Wozny, Computer Aided Chemical Engineering 2010, 28, 781. DOI: 10.1016/S1570-7946(10)28131-2 [4] Kohl A. L., Nielsen R., „Gas Purification“, Gulf Pub Co, 5th edition, 1997 [5] E.E. Wolf (Ed), Methane Conversion by Oxidative Processes, Fundamental and Engineering Aspects. Van Nostrand Reinhold, New York, 1992 [6 ]Bird, Stewart, Lightfoot, Transport Phenomena, 2. Edition, John Wiley&Son, 2006 [7] H. P. Mangalapally, R. Notz, S. Hoch, N. Asprion, G. Sieder, H. Garcia, H. Hasse, Energy Procedia 1, 2009, 963-970, DOI: 10.1016/j.egypro.2009.01.128 [8] Barnicki and Fair, Separation System Synthesis: A Knowledge-Based Approach. 2. Gas/Vapor Mixtures, Ind. Eng. Chem. Res. 1992, 31, S. 1679-1694 [9] Pakseresht S., Equilibrium Isotherms for CO, CO2, CH4, C2H4 on the 5A molecular sieve by a simple volumetric apparatus, Separation and Purification Technology, 2002, Vol. 28, iss. 1, p. 53-60 [10] A. Car, Pebax®/polyethylene glycol blend thin film composite membranes for CO2 separation: Performance with mixed gases, 2008, Separation and Purification Technology, vol. 62, p. 110–117 [11] O. K. Varghese, M. Paulose, T. J. LaTempe, C.A. Grimes, High-rate solar photocatalytic conversion of CO2 and water vapour to hydrocarbon fuel, 2009, Nanoletters, Vol. 9. No. 2. 731-737; [12] V. Abidin, C. Bouallou, D. Clodic, Valorization of CO2 Emissions into Ethanol by an Innovative Process, 2011, Proceeding for 14th conference on Process Integration, Modelling and Optimisation for Energy Saving and Pollution Reduction, Florence, Italien [13] A. Egbedi, J.J. Spivey, Effect of H2/CO ratio and temperature on methane selectivity in the synthesis of ethanol on Rh-based catalysts, Catalysis Communications, 2008, (9)
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
WALTER MARTINI*, GÜNTER WOZNY*
HYPERBRANCHED POLYMERS FOR CO2 CAPTURE: DATA ESTIMATION AND PROCESS SIMULATION Abstract The oxidative coupling of methane (OCM) is currently being intensively investigated at the Technical University of Berlin. However, during the process CO2 is being produced, which then has to be separated in a downstreaming process. This is usually done by an absorption and desorption step using aqueous amine solutions and accompanied by high energy demands for regeneration. However, hyperbranched polymers have recently attracted attention as promising candidates for gas absorbents with a high capacity for CO2 and with large selectivities. Though, only little physical data is available in literature. Therefore, missing parameters were estimated for the hyperbranched polymer Boltorn U3000 based on its structure only and used afterwards for the simulation of the CO2 separation process within Aspen Plus. Keywords: Hyperbranched Polymers, Carbon Capture, Absorption, Parameter Estimation 1. Introduction As the main constituent of natural gas, the development of processes that allow the conversion of methane to more valued products is of strong economic interest. A chemical of particular importance is ethylene, which can be obtained via oxidative coupling of methane (OCM). It is a surface induced gas phase reaction and its overall yield is still limited up to 30%. Extensive research has resulted in a reasonable understanding of the elementary reactions that occur within OCM [1]. Besides ethylene, unwanted by-products such as carbon dioxide are produced and have to be cleaned in a downstreaming process subsequent to the reaction.
*
Chair of Process Dynamics and Operation, Berlin Institute of Technology, Sekr.KWT-9, Str. des 17.Juni 135, D-10623 Berlin, Germany.
Hyperbranched Polymers for CO2 Capture: Data Estimation and Process Simulation 199
Fig. 1. Process flow diagram of the OCM process
Various alternatives for the OCM process have been proposed so far, such as the OXCO process, the UCC process, the ARCO process, the Suzuki process, the TurekSchwittay process or the Co-Generation process [2]. All of them have in common that the product separation under high pressure and the recycling of unreacted methane is of most importance for the process economics [3]. Because of its current low yield and challenges for an efficient downstream process, OCM has not been applied in the industry yet and is currently being focused on by the German Cluster of Excellence UniCat [1]. Therefore, a miniplant has been built at the Chair of Process Dynamics and Operation of the Technical University of Berlin, where fundamental studies of process alternatives and the effect of recycles and efficiencies of each process unit are being investigated [4]. Figure 2 shows its generalized layout.
Fig. 2. Simplified process flow diagram of the OCM
The whole OCM process is divided into three general sections: subsequent to the reaction, a purification step takes place, where carbon dioxide has to be removed from the raw gas completely. Finally, in the separation section, the desired product, ethylene, is being separated from all other components, which may be recycled to the reaction section again. The removal of carbon dioxide herein plays an important role for the economic efficiency of the overall process, and therefore, is being focused on in this work. The UOP Selexol ® or the Lurgi Rectisol ® processes are examples of known physical absorption processes, which use dimethyl ether and cold methanol, respectively. Due to a nearly
200
W. Martini, G. Wozny
similar solubility of the product and the carbon dioxide in the liquid, those physical absorption processes cause high product losses of more than 30 vol%. On the other hand, chemical absorption using aqueous amine solutions has been extensively used for the removal of CO2 from gas streams in many industries and was therefore chosen for the miniplant. It involves one or more reversible chemical reactions between CO2 and an amine (such as MEA) in an absorption column. Desorption of the absorbed carbon dioxide then proceeds via a thermal regeneration process, which is, in general, very energy intensive. The amine solution has a limited lifetime due to degradation through oxidation of the amine. In addition, corrosion problems are usually observed for the aqueous amine process. The high energy demand for regeneration and solvent losses are the main disadvantages of using aqueous amine solutions as washing fluids. In this work, hyperbranched polymers are being investigated as an alternative for absorbents for carbon dioxide, since they have been shown to be promising candidates for gas absorbents with a high capacity for CO2 and with large selectivities [5]. Thermodynamic data had to be generated to be able to model and simulate absorption and a desorption process to compare the use of hyperbranched polymers to MEA. 2. Hyperbranched Polymers Currently, four major classes of polymers can be distinguished in accordance with their properties and polymeric architecture. Dendritic polymers herein represent the fourth class and are highly branched globular macromolecules. They can be subdivided into four subsets that are related to the degree of structural control, i.e. random hyperbranched polymers, dendrigraft polymers, dendrons, and dendrimers [6].
Fig. 3. Representation of the four major classes of macromolecular architectures [6]
Dendrimers show a well defined, monodisperse, perfectly branched structure and have a large number of functional end groups [7]. Their generation requires absolute control of all synthesis steps and makes large-scale production difficult and hence expensive. However, many applications do not require structural perfection. Therefore, using hyperbranched polymers can circumvent this major drawback of dendrimers [8]. Unlike dendrimers, randomly branched hyperbranched polymers with similar properties can be easily synthesized via one-step reactions and therefore represent economically promising products also for large-scale industrial applications. At room temperature, many branched polymers exhibit low viscosities in the pure state as well as in solution due to the absence of chain entanglement.
Hyperbranched Polymers for CO2 Capture: Data Estimation and Process Simulation 201 The structure of dendritic polymers resembles that of a treetop. They possess a core from which multiple branches are extended. Each branch can be source for more branches and hence different generations, thus giving exponential growth, in both end-group functionalities and molecular weights. The properties of dendritic polymers differ strongly from those of linear polymers of the same molar mass (less flexibility, lower entanglement degree, a significant chain-end effect, lower viscosity in solution and in the molten state, high solubility in common solvents, a different relationship between hydrodynamic volume and molar mass, and a different origin of the glass transition temperature) [9]. Potential applications of hyperbranched polymers range from the use as selective solvents in distillation of azeotropic mixtures or extraction to the control of flow characteristics, as well as the use as drug delivery system and many more.
Fig. 4. Boltorn® U3000 molecule by Perstorp AB (Sweden)
In this work, the dendritic polymer Boltorn® U3000 by Perstorp AB was applied to carbon dioxide absorption. Because of its polydispersity, the molecular structure shown in Figure 4 is only one possible occurrence, which has been used for the simulation. The core consists of a polyalcohol, i.e. pentaerythritol (PE). The hyperbranched structure is built from 2,2-Dimethylol propionic acid (Bis-MPA), which has the unique functionality of one COOH-group and two OH-groups. The shell consists of a large number of OH-groups, which are possible locations for CO2 to attach to. Previous measurements of Henry coefficients already have shown the potential of Boltorn U3000 for the absorption of carbon dioxide [10]. 3. Estimation of Thermodynamic Data In order to be able to simulate the absorption and desorption process using Aspen Plus, several physical and thermodynamic data is needed. Only very little data is available for
W. Martini, G. Wozny
202
Boltorn U3000 in literature (such as density or viscosity at standard conditions). Therefore, a data estimation step had to be carried out first. To do this, several components that are part of the Boltorn molecule or consist of similar functional groups and where detailed data is available (see Figure 5), were chosen for a prior study of different estimation methods.
b)
a)
c)
d) Fig. 5. Applied components: a) PE b) Bis-MPA c) Glycerol d) Triethylene-Glycol-Dioctanate
The results in Table 1 show the mean deviation for each parameter of all reference components between estimation and literature, based on the best method available. Table1 Deviation of estimated thermodynamic data from literature
TB TC pC νC ΔhLV Δhf Δgf ω
Method Gani [11] Ambrose [12] Ambrose Joback [13] Gani Gani Gani Definition
mean deviation -5,9% 8,9% 14,6% 1,2% 5,7% 0,6% 2,2% 1,1%
This set of estimation methods was then applied to the parameter estimation for the hyperbranched polymer Boltorn® U3000. For parameters that are not included in Table 1, such as the relative van der Waals volumes and surface areas of the pure chemicals used for instance in UNIFAC, there was only one method available and hence no choice had to be taken. The vapor pressure was estimated based on the observation that, in general, it is very low or not existent for hyperbranched polymers. Additionally, molar liquid volumes and viscosities for a temperature range of 280 to 350 K were provided through earlier measurements [14]. To evaluate the quality of the estimations, Henry coefficients were then calculated based on the estimations and compared to measured values. As can be seen in Figure 6,
Hyperbranched Polymers for CO2 Capture: Data Estimation and Process Simulation 203 both agree rather well for both, methane and carbon dioxide in Boltorn U3000. Especially the latter is of most importance for the simulation of the absorption process.
MassrelatedHenrycoefficient[bar]
8000 7000 6000 5000 4000
CH4‐ measured CH4‐ estimated
3000
CO2‐ measured CO2‐ estimated
2000 1000 0 300
310
320 330 Temperature[K]
340
Fig. 6. Comparison of mass related Henry coefficients of CO2 and CH4 in Boltorn U3000 based on measurements and derived from data estimation
4. Simulation of the absorption and desorption process
A model for the simulation of the CO2 separation was implemented in Aspen Plus as shown in Figure 7. The same conditions as in the miniplant were used with the same packings, 50 theoretical stages and a maximum pressure of 32 bar in the absorption column, 42 theoretical stages and 2 bar in the desorption column, respectively.
MAKEUP
GASOUT
DESCO2 HEATER1
HEATER2
ABSORBER LEANFEED
1
HEATER3
2 3
MIXER
GASFEED
RICHFEED
GAS 5
PUM P
RICHOUT
DESORBER
4
FLASH
HEATX FLASHOUT
LEANOUT
Fig. 7. Aspen Plus model of the absorption an desorption process
The gas feed with a flow of 10 kg/h consisted of 41.6% CO2, 37.9% CH4, 7.1% C2H6, 13.2% C2H4 and 0.2% H2O. The aim of the absorption was to remove 90% of the carbon dioxide from the raw gas. As solvents, both the hyperbranched polymer Boltorn U3000 with small amounts of water as well as, for comparison, an aqueous solution of
W. Martini, G. Wozny
204
monoethanolamine (MEA) was applied. For MEA, the built-in ELECNRTL model with the Redlich Kwong equation of state was used. Reactions were included in the Rate-Sep model by the MEA-REA package, where kinetic parameters were adjusted to literature [15]. For Boltorn U3000, the UNIFAC Dortmund model was used and no reaction was assumed. The simulation results obtained show that for the polymer a mass flow of 220 kg/h was needed to meet the 90% CO2 removal requirement compared to 55 kg/h of the chemical absorbent MEA. This is due to its lower capacity for carbon dioxide. Because of its relatively high viscosity and for an easier operation of the desorption column (strip-steam), small amounts of water were added to the solvent Boltorn U3000. The effects of this on the separation degree of different gases in the absorption column are shown in Table 2. Table2 Separation degree of the gaseous components in the absorption column
Water content CO2 CH4 C2H4 C2H6
none
1 weight%
5 weight%
90.07% 3.58% 28.80% 27.27%
90,13% 3,80% 32,58% 29,38%
90,08% 4,72% 50,22% 36,81%
Apparently, the added water has a negative effect especially on the selectivity concerning CO2 and the main product ethylene. Therefore, it should be minimized to a degree, where operation of the desorption column is still feasible. Table3 Heat duties per kg CO2 required for the regeneration of the solvent
MEA Heat Duty [kJ/kg CO2] Pumping [W]
2980 44
Boltorn U3000 202 284
The main advantage of using a physical absorbent comes into play, when looking at the heat duty that is required for regeneration of the solvent, which usually accounts for the highest part of the whole energy demand. For chemical absorbents, typically absorption is exothermal and, hence, this energy has to be spent in the desorption column. In contrast to that, physical absorbents, such as the applied hyperbranched polymer, only need to be heated to a certain temperature for the CO2 desorption process to take place. This effect can be seen in Table 3, where the energy requirement for Boltorn U3000 is a lot smaller than that for the aqueous amine solution. On the other hand, it has to be noted that, in case of hyperbranched polymers, for instance pumping requires more effort.
Hyperbranched Polymers for CO2 Capture: Data Estimation and Process Simulation 205 5. Conclusions Thermodynamic and physical data of the hyperbranched polymer Boltorn U3000 were estimated based on its structure only. The results in terms of Henry coefficients regarding CO2 and CH4 showed good agreement with measured values. A model was built within Aspen Plus to simulate the absorption and desorption process on miniplant scale. The comparison of the physical absorbent Boltorn U3000 and the chemical absorbent MEA showed a very low energy demand for the former but some drawbacks concerning selectivities and operability because of its high viscosity. Therefore, further research is required to validate simulation results and to investigate additional hyperbranched polymers. 6. Acknowledgements The authors acknowledge the support from the Cluster of Excellence „Unifying Concepts in Catalysis“ coordinated by the Technical University of Berlin (TU Berlin) and funded by the German research foundation – Deutsche Forschungsgesellschaft – DFG. Symbols TB boiling point temperature TC critical temperature pC critical pressure νC critical molar volume ΔhLV heat of vaporization Δhf heat of formation at 298.15 K Δgf Gibbs free energy of formation at 298.15 K ω Pitzer azentric factor Literature [1] Unifying Concepts in Catalysis, http://www.unicat.tu-berlin.de/. [2] Wolf E.E., Methane conversion by oxidative processes, Van Nostrand Reinhold 1992. [3] Salerno D., Arellano-Garcia H., Wozny G., AICHE Annual Meeting, Salt Lake City, USA, (2010). [4] Stünkel S., Repke J.-U., Schomäcker R., Wozny G., Chemie Ingenieur Technik, 82, (2010), 1416-1417. [5] Rolker J., Seiler M., Mokrushina L., Arlt W., Ind. Eng. Chem. Res., 46, (2007) 65726583. [6] Tomalia D.A., Fréchet J.M.J., Journal of Polymer Science: Part A: Polymer Chemistry, 40, (2002) 2719-2728. [7] Tomalia D.A., Polymer Journal, 17, (1985) 117-132. [8] Seiler M., Fluid Phase Equilibria, 241, (2006) 155-174. [9] Zagar E., Zigon M., Macromolecules, 35, (2002) 9913-9925. [10] Martini W., Arellano-Garcia H., Wozny G., AIChE Annual Meeting, Salt Lake City, USA, (2010). [11] Constantinou L., Gani R., O’Connel J.P., Fluid Phase Equilibria, 103, (1995), 11-22. [12] Ambrose D., Journal of Applied Chemistry and Biotechnology, 26, (1976), 711-714.
206
W. Martini, G. Wozny
[13] Joback K.G., Ph. D. dissertation, The Massachusetts Institute of Technology, Cambridge, Mass., (1984). [14] Martini W., Arellano-Garcia H., Wozny G., Jahrestreffen Fachausschuss Fluidverfahrenstechnik, Fulda, Germany, (2011). [15] Kuckaa L., Mueller I., Kenig E. Y., Gorak A., Chemical Engineering Science, 58, (2003), 3571-3578.
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
VERENA LÖFFLER*, MATTHIAS KRAUME*
DEVELOPMENT OF MIXED-MATRIX MEMBRANES FOR SEPARATION OF GASEOUS HYDROCARBONS Abstract In this work mixed-matrix gas separation membranes were produced and tested for CH4 / n-C4H10 separation. The fundamental principle of mixed-matrix membranes is described on the basis of the Maxwell model to illustrate the requirements for membrane material selection. The rubbery polymer PDMS and a micro porous carbon adsorbent have been chosen as membrane components and mixed-matrix composite membranes of different phase fractions have been produced. The permeance and selectivity has been measured with a gas mixture of 95(vol.)% CH4 methane and 5(vol.)% n-C4H10. Keywords: gas permeation, mixed-matrix membranes, Maxwell model 1. Motivation Membrane based gas separation is a fast growing area of separation technology, that has drawn a great deal of interest in research and industry in the past 25 years1,2. Applications for gas permeation are for example air separation, CO2 removal, dew point adjustment or reclaiming of gasoline vapour. Gas permeation competes with conventional gas separation processes like pressure swing adsorption or gas scrubbing. In comparison to these techniques gas permeation plants offer advantageously small footprints, simple assembly, inherent steady state operation and absence of separation agents. In some applications, like e.g. air, H2 or CO2 separation, the technique is accepted as an alternative to conventional gas separation processes. In other applications further development of membrane material is desirable to reduce costs and energy demand of the plants and tap the full market potential of gas permeation2. Separation of hydrocarbons is one of these applications. In this work new mixed-matrix membrane material is produced and tested for separation of a methane-butane mixture. State of the art for membrane separation of hydrocarbons from lighter gases is dense rubbery polymer membrane materials, where the separation mechanism works according to *
Technische Universität Berlin, Fachgebiet Verfahrenstechnik, Sekr. Ma 5-7, Straße des 17. Juni 136, 10623 Berlin, Germany
208
V. Löffler et al.
the solution-diffusion model3,4. This material features solution selectivity and high fluxes. In Fig. 1, a gas permeation plant for hydrocarbon dew point adjustment of natural gas with such membranes is illustrated: The feed contains methane and different higher hydrocarbons. It is compressed to high pressure and cooled down subsequently to separate condensed components. In the membrane unit higher hydrocarbons are enriched on the permeate side whilst the lighter ones like methane are concentrated in the retentate. The recycling of permeate prevents methane loss through the permeate flow. The main energy demand of such a plant is the compressor power, the cooling power of the condenser and the power of the vacuum pump. An increase of membrane permeability and selectivity would reduce both energy demand and material costs3: The pressure difference and membrane area required decreases with increasing permeability. An increase of membrane selectivity would decrease gas slip of the lighter components and reduce the recycled permeate flow. A decrease of the permeate flow would again reduce compressor power, cooling power and power of the vacuum pump.
Fig. 1: Gas permeation process for dew point adjustment
However, further improvement of polymeric membranes is difficult, because the performance of polymeric membrane material is limited: The dependency of permeability and selectivity of all polymeric membrane materials underlies an upper bound, which depends on the components that are to be separated and is named Robeson Upper Bound after L.M. Robeson who first reported about this behavior in 1991. Fig. 2 shows the upper bound for CO2/CH4 separation5. It illustrates that rubbery polymers tend to have higher permeabilities with lower selectivities while glassy polymers use to have higher selectivities with lower permeabilities. But to reach the most attractive region of high selectivity and high permeability another membrane material is needed. Accordingly inorganic materials were investigated as possible membrane material. In a number of research projects it was shown, that porous inorganic membranes, like zeolites or activated carbons, can have selectivities and permeabilities above the Robeson bound6-8. However these inorganic materials are difficult to produce, very brittle and expensive9,10. To our knowledge, no module with inorganic gas permeation membranes has come to market so far. The next generation of gas separation membranes could be some hybrid material combining the advantages of polymeric and inorganic materials.
Development of mixed-matrix membranes for separation of gaseous hydrocarbons
209
Fig. 2: Robeson Upper Bound for CO2/CH4 separation [5]
2. Mixed-matrix membranes An innovative concept for membrane material is the combination of polymeric and inorganic materials in mixed-matrix membranes2,11. These consist of a dense polymeric matrix, in which inorganic solid material is dispersed. Accordingly mixed-matrix membranes are supposed to be easily produced and mechanically stable like polymeric membranes. They can also have separation properties close to or even above the Robeson Bound like inorganic membranes. Material selection is mostly important to produce a successful mixed-matrix membrane7,8,11, because it is necessary that transport mechanisms through both phases prefer the same component. Generally the total permeability Ptot depends on the ones of the single phases Pd and Pc and the dispersed phase volume fraction φd. To achieve an improvement in total permeability with respect to the pure polymer, it is necessary that the dispersed phase has got a higher permeability than the polymer. At the same time the dispersed phase needs to have a higher selectivity than the polymer to enhance total selectivity. This fundamental character of mixed-matrix membranes can be described very clearly by the Maxwell model, which is a simple, but generally accepted model for mixed-matrix membranes7,8. It assumes an uniform film, in which a second phase is dispersed homogeneously, as shown in Fig. 3. Originally the model was developed to calculate the
V. Löffler et al.
210
conductivity of composite materials, but it is also usable to calculate gas permeation through mixed-matrix membranes even though it underlies some limitations.
Fig. 3: Mixed-matrix membrane according to the Maxwell model
According to the Maxwell model the total permeability of a gas component A through a mixed matrix layer is P A + 2 PcA − 2 ϕ d PcA − PdA A =PcA d A Ptot (1) Pd + 2 PcA + ϕ d PcA − PdA
(
(
)
)
The total selectivity for a mixture of A and B can be calculated as follows: A Ptot S AB (2) tot = B Ptot As shown in Fig. 4 the Maxwell model predicts that highest selectivity improvements are possible, when the permeabilities of both phases are similar. In contrast, no selectivity improvement is seen when permeability in the dispersed phase is significantly higher or lower than in the polymer phase. Therefore, an adequate combination of polymer and solid material is crucial for successful preparation of mixed-matrix membranes. Furthermore, the effect increases with the dispersed phase volume fraction. So it will be advantageous to realize high dispersed phase volume fractions as long as the membrane structure is not damaged. 1,4 1,35
pφ= 0,2 d =0,2 pφ= 0,4 d =0,4
1,3
Stot/Sc
1,25 Sc/Sd = 0,5
1,2 1,15
0,5
1,1 0,9
1,05 1 0,0001
0,9
0,001
0,01
0,1
1 Pc/Pd
10
100
1000
Fig. 4: Prediction of mixed-matrix membrane selectivity according to the Maxwell model
Development of mixed-matrix membranes for separation of gaseous hydrocarbons
211
3. Material selection Since rubbery polymeric membranes are state of the art for separation of hydrocarbons from lighter gases a rubbery polymer, Polydimethylsiloxan (PDMS), was chosen as continuous phase of the mixed-matrix membranes in this work. PDMS is one of the most common rubbery polymers used not only as membrane material but also in paper, food or cosmetic industry. Like other rubbery polymers PDMS is solubility selective, so higher hydrocarbons permeate better than lighter gases. Accordingly, an inorganic phase with preferential adsorption and diffusion of higher hydrocarbons is needed. Fillers which are diffusion selective and prefer permeation of lighter gases are not suitable. Therefore, most molecular sieves are excluded. Some micro porous materials have been studied for solubility based separation in the last 15 years12-15. The transport mechanism is a selective surface flow of the adsorbed molecules where high selectivities are possible. If capillary condensation of the adsorbed species occurs, even higher selectivities are possible. In that case the other components are completely excluded from the pores. In this work a micro porous carbon adsorbent was chosen as dispersed phase, produced by the Blücher GmbH, Erkrath. The main characteristics of the chosen carbon are listed in Table 1. However development of sorption selective mixed-matrix membranes does not only require a careful material selection but also an optimization of process parameters, because surface diffusion and capillary condensation depend on parameters like temperature, pressure and concentration. Table1 Characteristic parameters of the carbon adsorbent Average pore diameter Total pore volume BET surface area Particle size distrubution Average particle size
19 0,64 1361 1 - 30 9
A cm³/g m²/g µm µm
4. Membrane preparation In this work mixed-matrix composite membranes were produced. The mixed-matrix separation layer was coated on a highly permeable support structure consisting of a polymeric non-woven with a microporous, polymeric toplayer. The coating solution was prepared by suspending a certain amount of carbon particles in polymer solution. The suspension was applied manually with a stainless steel roll. After preparation the membranes were dried. In Fig. 5 the structure is illustrated schematically and two pictures taken with scanning electron microscopy (SEM) are shown. The pictures illustrate that particles are embedded completely in the polymer matrix. Further no gaps have been seen between polymer and carbon like they were found by other groups who produced mixed-matrix membranes with glassy polymers9. However, the membrane surface is not completely even.
V. Löffler et al.
212
Fig. 5: Mixed-matrix composite membrane: a) schematic; b) SEM picture; c) magnification of separation layer
5. Measurement of permeance and selectivity The separation properties of the mixed-matrix membranes were tested at the HelmholtzZentrum Geesthacht, Centre for Membranes and Structured Materials. A simplified flowchart of the setup is shown in Fig. 6. The feed consisted of 95 (vol.)% methane and 5(vol.)% n-butane. Feed pressure was varied between 10 bar and 40 bar and permeate pressure was 1 bar. Composition of feed and permeate has been measured online by a gas chromatograph. The feed temperature was set to 25 °C. circulator
P
T
F
Q
F
P
Q
mixing vessel heat exchanger
membrane cell
compressor
Fig. 6: Setup for permeability measurements
Development of mixed-matrix membranes for separation of gaseous hydrocarbons
213
6. Results The permeances L of both components methane and n-butane were calculated & , feed pressure pF, according to equation (3) from the detected values of permeate flow V P i
permeate pressure pP, feed composition x iF , permeate composition x iP and membrane area AM. Fugacity coefficients φi have been considered on the high pressure side. & xi V P P Li = i (3) i ϕ p F x F − p P x iF A M Selectivity S was calculated as the following ratio Lbu tan e S = methane (4) L For calculation of permeability the membrane thickness was measured with SEM. Pi = L i δ (5) Dispersed phase volume fraction φd was calculated from the dispersed phase mass fraction in the coating suspension ξd, particle density and density of the cross-linked and dried polymer. Mean values of the results are given in Table 2. Table2
(
)
Experimental results
φδ
ξδ
δ
L butane
P butane
%
%
µm
Nm³/(h bar m²)
10 barrer
PDMS
15,4
3,81
MMM 1
31
3,8
15,1
2,80
MMM 2
43
6,2
14,8
3,58
-1,05 +1,12 -0,59 +0,46 -0,78 +0,49
4
1,41 1,24 1,59
-0,39 +0,42 -0,26 +0,21 -0,34 +0,22
S [-]
17,2 18,3 15,9
-1,36 +1,18 -0,59 +0,6 -0,4 +0,7
The experimental values have been used to calculate the permeability of the dispersed phase according to the Maxwell model. Corresponding to equation (1) Pd of a component A is A A A Pc ⋅ 2 (1 − ϕd ) − Ptot (2 + ϕd ) PdA =Ptot (6) A Ptot ⋅ (1 − ϕd ) − (1 + 2 ϕd ) PcA The following two figures show the results for mixed-matrix membranes with dispersed phase fraction φd = 0,31 and φd = 0,43. Fig. 7 shows, that a slightly improvement in selectivity has been reached with respect to the pure PDMS. The experimental values scatter in the range of Pc/Pd = 10 and Sc/Sd = 0,7. Both values show, that material selection needs further improvement. A carbon with both higher permeability and selectivity is needed. On the other hand Fig. 8 shows, that improvement of phase fraction to φd = 0,43 did not improve selectivity as the Maxwell model predicts. Instead selectivity has decreased below the one of PDMS and permeability is around the value of PDMS. This behaviour can be explained by the microscopic structure of the membrane. The higher dispersed phase fraction led to a more irregular coating. The SEM analyses showed that thickness of the separation layer differed between 10 and 30 µm and that the surface was
V. Löffler et al.
214
very rough. Therefore it is possible, that the coating has got defects, which lower selectivity and enhance permeability with respect to the membrane with φd = 0,31. Moreover the determination of the mean thickness was most probably less exact for φd = 0,43. 1,3 MMM 1
Stot / Sc
1,2
φ = 0,31
Maxwell model Experiments
Sc/Sd = 0,5
0,7
1,1
0,9
1
1
0,9
0,8 0,0001
0,001 0,001
0,01
0,1
1
10
100
1000
Pc / Pd
Fig. 7: Experimental results for φd = 0,31 and comparison with the Maxwell model 1,3 MMM 2
1,2
Stot / Sc
φ = 0,43
Maxwell model Experiments
1,1 Sc/Sd = 1
1
1,1 1,3
0,9
0,8 0,0001
1,5
0,001 0,001
0,01
0,1
1
10
100
1000
P c / Pd
Fig. 8: Experimental results for φd = 0,43 and comparison with the Maxwell model
7. Conclusions Preparation of mixed-matrix membranes and their optimization for a gas separation problem is a both promising and challenging task. It was shown that these hybrid membranes with an organic matrix and inorganic fillers could be the next generation of gas separation membranes because they combine the advantages of polymeric and inorganic
Development of mixed-matrix membranes for separation of gaseous hydrocarbons
215
material. However, material selection is crucial for their separation performance. In this work mixed-matrix membranes of PDMS and carbon adsorbent were successfully produced and tested for CH4/n-C4H10 separation. Analysis of the experimental results with the Maxwell model showed, that further investigations and improvement of the dispersed phase is needed. But at the same time the process parameters at permeability measurement need to be varied and investigated carefully, because performance of solubility selective mixed-matrix membranes is not only depending on membrane material but also on process conditions as pressure, temperature and concentration. 8. Acknowledgements The authors whish to thank their cooperation partners Helmholtz-Zentrum Geesthacht, Centre for Membranes and Structured Material, where permeability measurements have been carried out, and Blücher GmbH, who produced and supplied the carbon adsorbents. Symbols L permeance P permeability p pressure S selectivity x molar fraction δ thickness of separation layer φi fugacity coefficient φd dispersed phase fraction ξd mass fraction of carbon in coating suspension
Indices: A,B any gaseous component c continuous phase d dispersed phase F feed I component (methane, n-butane) P permeate tot total
Literature [1] Baker, R. W.: Membrane Technology and applications; 2. ed.; John Wiley and Sons: Hoboken, 2004. [2] Baker, R. W.: Industrial & Engineering Chemistry Research, 41, (2002), 1393-1411. [3] Ohlrogge, K.; Wind, J.; Brinkmann, T. In Comprehensive membrane science and engineering; Drioli, E., Giorno, L., Eds.; Academic Press (Elsevier): Oxford, 2010; Vol. 2. [4] Ohlrogge, K.; Wind, J.; Peinemann, K. V.; Stegger, J. In Membranen; 1 ed.; Ohlrogge, K., Ebert, K., Eds.; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, 2006. [5] Robeson, L. M.: Journal of Membrane Science, 62, (1991), 165-185. [6] Singh, A.; Koros, W. J.: Industrial & Engineering Chemistry Research, 35, (1996), 1231-1234. [7] Aroon, M. A.; Ismail, A. F.; Matsuura, T.; Montazer-Rahmati, M. M.: Separation and Purification Technology, 75, (2010), 229-242. [8] Chung, T. S.; Jiang, L. Y.; Li, Y.; Kulprathipanja, S.: Progress in Polymer Science, 32, (2007), 483-507. [9] Vu, D. Q.; Koros, W. J.; Miller, S. J.: Journal of Membrane Science, 211, (2003), 311334.
216
V. Löffler et al.
[10] Vu, D. Q.; Koros, W. J.; Miller, S. J.: Journal of Membrane Science, 211, (2003), 335-348. [11] Bernardo, P.; Drioli, E.; Golemme, G.: Industrial & Engineering Chemistry Research, 48, (2009), 4638-4663. [12] Javaid, A.: Chemical Engineering Journal, 112, (2005), 219-226. [13] Sircar, S.; Rao, M. B.; Thaeron, C. M. A.: Separation Science and Technology, 34, (1999), 2081-2093. [14] Anand, M.; Langsam, M.; Rao, M. B.; Sircar, S.: Journal of Membrane Science, 123, (1997), 17-25. [15] Rao, M. B.; Sircar, S.: Journal of Membrane Science, 110, (1996), 109-118.
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
MATAN BEERY*, JI JUNG LEE**, JOON HA KIM**, JENS-UWE REPKE***, GÜNTER WOZNY*
NOVEL AND INTENSIFIED PROCESS DESIGN FOR SEAWATER RO DESALINATION PRE-TREATMENT Abstract In this work a combined theoretical-experimental approach was applied to improve the current state of the art of seawater reverse osmosis (SWRO) pre-treatment technologies. On the one hand, a new process design software tool was developed based on steady-state simulation and evaluation of all the common state-of-the-art pre-treatment process units such as media filtration, membrane filtration, coagulation, flotation, etc. On the other hand, experimental work was carried out in order to both calibrate and validate the models as well as to explore the feasibility of new solutions that specifically address the drawbacks of current technologies. Such solutions include process integration and intensification. A novel lab-scale experimental system was set up and tested for its feasibility as a SWRO pre-treatment. The results indicate that such an integrated approach to process design in SWRO pre-treatment show a positive potential in improving the sustainability of desalination technology. Keywords: Desalination, pre-treatment, process design. 1. Introduction Due to fast urbanization, population growth, increase in standard of living and climate change, water stress has become a major global cause for concern in recent years. As a result, securing new water sources, for example by means of seawater desalination, is becoming a common solution in many areas of the world. Seawater reverse osmosis (SWRO) has seen a great rise in popularity as the preferred method of choice for * **
Chair of Process Dynamics and Operation, TU-Berlin, Strasse des 17 Juni 135, 10623, Berlin, Germany
Department of Environmental Science and Engineering, Gwangju Institute of Science and Technology (GIST), Gwangju, 500-712, Republic of Korea *** Institute of Thermal, Environmental and Natural Products Process Engineering, TU Bergakademie Freiberg, Leipziger Strasse 28, 09596 Freiberg, Germany
218
M. Beery et al.
desalination technology recently reaching annual growth rates of 15-20% [1] and a record breaking 6.6 million m³/d of new capacity (approx. 2700 olympic swimming pools) being brought online in 2010 alone [2]. Despite showing major technical and economical improvements in the last decade (especially in the area of membrane performance and energy recovery solutions), membrane based desalination still has a long way to go in order for it to be considered a sustainable technology. The reason for that being mostly its substantial energy consumption which makes it both expensive and less “green” compared to other water treatment processes. This was confirmed in a life-cycle based sustainability assessment of desalination processes [3]. The most major step in increasing SWRO’s sustainability would therefore be a reduction of its carbon footprint. RO membranes used for seawater desalination are highly susceptible to fouling due to organic/inorganic, biological and particulate matter often present in the sea. Fouling of the RO membranes has several negative effects which decrease the plant’s economical and environmental efficiencies. Such effects include reduction in production rate, higher energy and chemical consumptions, frequent membrane cleanings and replacements, increase in the plant’s downtime etc. An effective, location-specific pre-treatment of the seawater is therefore a key issue in maintaining long term operational success of an SWRO plant. As a result, the design of such systems can no longer be purely done in the traditional rule of thumb and trial and error methods (as is currently still the case). The assessment of the process sustainability aspects should be predicted in a systematic and analytical way. Reaching an optimal design thus requires a multidisciplinary approach which must be performed in the planning phase of the process and not during its operation. The goal of this work is to apply a systematic approach to the process design of SWRO pre-treatment which combines theory (in the form of modeling, simulation, evaluation) and practice (lab experiments for validation and testing of new process configurations). 2. Theoretical work The introduction of system theories and methods to chemical engineering can be traced back to the 1960’s in the UK and the US. It incorporates different methodologies which assist in decision making for the creation and operation of chemical supply chains [4]. In process design the common steps include process synthesis, simulation and evaluation, which are repeated iteratively until an optimal design is found. Dudley et al. [5] indicate that PSE, in the form of process simulators and flowsheeting tools, has seen very low to practically no acceptance in the drinking water community. They suggest that the reasons for that lay in the industry’s misconception of water treatment models being not accurate enough (or hard to calibrate) and the fact that water treatment plants are traditionally designed with conservative rules-of-thumb aiming at water quality rather than economic efficiency. In seawater reverse osmosis on the other hand, the economic performance of the plant and the total cost of ownership play a major role in the design considerations. As a result, several major companies in the membrane desalination market (DOW, Hydronautics, Toray, ERI etc.) offer, often free-of-charge, process simulation tools to assist the engineer with the design, assessment and optimization of SWRO processes. These tools are stand-alone easy to use graphic user interfaces (GUI) based on steady state simulations which are helpful in achieving fast and basic design solutions for the RO and energy recovery systems. However, there are currently no such tools to carry out process design
Novel and Intensified Process Design for Seawater RO Desalination Pre-treatment
219
for the pre-treatment system. Its design is left for the traditional rule-of-thumb approach, relying on past experience, pilot testings and over-design. This results in sub-optimal process designs which often negatively affect the performance of the RO stage and the sustainability of the entire desalination plant [6]. Using Matlab, a GUI process systems engineering design tool was created to help fill-in this gap in the pre-treatment process design stage. It follows the synthesis-simulationevaluation concept and relies on basic steady-state models and process parameters found in literature or laboratory/pilot experiments (figure 1).
Fig. 1. Screen shot of the computer-based process design tool
The most common unit operations seen in SWRO pre-treatment are: Pumps, mixers, coagulators, sedimentation, flotation, strainers, media (carbon/sand) filtration, membrane (MF/UF) filtration and cartridge filtration. The question of which units to use and how to interconnect them corresponds to process synthesis and is highly depended on the feed seawater characteristics as well as on the desired product water specifications. The guidelines given by RO membrane manufacturers and practiced by SWRO plant operators are fairly clear: RO feed water should have silt density index (SDI) of less than 3, turbidity of less than 1NTU and total organic carbon (TOC) levels of less than 2mg/l. Using the tool, the user first gets several flowsheet suggestions that will meet these specifications based on industrial heuristics [7] and feed water quality. Once one of the possible flowsheets has been selected, a steady state simulation takes place. Based on state of the art basic models which incorporate steady state mass and energy balances as well as physical separation mechanisms (such as sedimentation or flotation). The physical and chemical properties relevant to the models (most important are viscosity, density and particle size) are either calculated by correlations [8] or given as an input by the user. This level of modeling is sufficient in acquiring basic design information with relatively low computational effort. The model parameters are either literature based, experimentally identified or freely defined by the user. The user first defines an input vector describing the seawater intake feed (flow, pressure, temperature, pH, DOC,
220
M. Beery et al.
suspended solids, turbidity, and salinity). The units then get solved one by one with the tool prompting the user to input some additional required parameters if necessary (like filtration flux, media depth, coagulant dose etc.). After the flowsheet simulation is completed the processes can be analyzed. Following the principles of sustainability, the evaluation takes place in both economical and environmental forms. Economically, a net present value assessment is made on counts of the simulation sizing and operation results. Environmentally, a basic carbon footprint assessment is performed based on the forecasted process electricity and chemical consumptions. The user can then choose the best flowsheet and/or try to continue and improve the simulation and evaluation results until finally reaching an optimal process design which can then be taken into the next engineering step. 3. Experimental work The experimental work performed in this project has two distinct goals. First, it must compliment the theoretical work in the form of parameter identification and model validation. Second, it should explore the technical feasibility of previously untested/undocumented new pre-treatment process configurations including new ideas for process integration and intensification. One of the main challenges that current pre-treatment systems have a hard time of coping with is the removal of bio-fouling causing materials, namely organic fractions ranging from microalgae to small organic molecules (transparent exopolymer particles) [9]. Since coagulation tends to improve the removal of organics by causing them to flocculate into larger particles it is often used by plant operators in excess without substantial technical merits but with higher economic and environmental burdens (caused by the production of the coagulant as well as of that of the sludge). As a result, solutions which will eliminate the need for coagulation but still produce low bio-fouling RO-feed water are needed [10]. This was the motivation behind the experimental testing of a new, coagulantfree intensified filtration system for SWRO pre-treatment. The system was composed of a deep-bed rapid granular activated carbon (GAC) 50cm filtration column followed by deadend membrane microfiltration. The system is depicted in figure 2. More information about the GAC and the membrane module is given in tables 1 and 2, accordingly. The water used was collected from the shore of Mokpo, at the shore of the Yellow Sea and the experiments took place at Gwangju, Korea. The water quality parameters are given in table 3. It should be pointed out that the water is rich in organics and suspended solids which pose a difficult task to a potential desalination plant.
Novel and Intensified Process Design for Seawater RO Desalination Pre-treatment
221
Fig. 2. The system used for the combined active carbon-microfiltration experiments
Table1 Granular activated carbon properties Parameter Particle size Bulk density Specific area Suspended solids reduction potential Turbidity reduction potential Fixed carbon Maximum ash content
Value 12*30 Mesh 0.51g/l 100m²/g 92.9% 93% 95% 7%
Deviation ±0.03 ±0.5 ±0.5% ±5% ±5% ±3%
Table2 Technical data of the membrane module Type hollow fiber (Kolon)
Pore size 0.1μ
Material PVDF
Fiber length 0.25m
Diameter (in/out) 0.8/2.0 mm
Fiber num. 10
Effective area 135cm²
Mechanical strength >25kgf/fiber
Operation Dead-end (const. flux)
Table3 Raw seawater quality parameters Temp 18.5°C
pH 7.86
DO 5.97mg/l
Turbidity 4.78NTU
TOC 9.2mg/l
TSS 49mg/l
M. Beery et al.
222
The GAC filter was operated at a moderate, common rapid filtration rate of 7m/h and the membrane was tested with different fluxes to determine which flux produces controllable fouling that can be sufficiently removed using backwashes (at 20 minutes intervals) with a minimal need for chemical cleanings. Each experiment was conducted using 20L samples and the water quality parameters were measured after the GAC and the MF membrane. Flows were measured before both filters and pressure was measured on the feed side of the membrane. The measured flux, J, was corrected for temperature, T, using equation 1 [11]. (1) J corrected = J ⋅ (1.784 − 0.0575 ⋅ T + 0.0011⋅ T 2 − 10 −5 ⋅ T 3 ) The water quality results are shown in figure 3. As it is seen in the figure, the active carbon was already very efficient in removing most of the turbidity and a large amount of the organic content. The values in the effluent stream were 0.51NTU and 4.34mg/l accordingly. Compared to the usual 10-20% TOC removal usually seen in pre-treatment systems (even with coagulation), the 53% removal of organics seen here can be considered very promising. The MF membrane has shown a further decrease in turbidity to levels of 0.19NTU. Unfortunately the TOC measurements of the membrane filtrate could not be performed due to technical contingencies, however it could be speculated that a further decrease took place. 12
TOC [mg/l] , Turbidity [NTU]
TOC
Turbidity
10
8
6
4
2
0 Raw Seawater
GAC effluent
MF permeate
Fig. 3. Organic carbon content and turbidity of the seawater at different process points
The operation of the membrane filtration proved to be fairly stable when using fluxes of 300 L/m²h or lower. Under such conditions the increase in trans membrane pressure due to fouling was moderate and could be somewhat well controlled using backwashes. Higher fluxes resulted in fast increases in operating pressures and irreversible fouling (figure 4). It should be noted that even 300 L/m²h is a fairly high flux compared to the typical 60-120 L/m²h fluxes usually seen in SWRO pre-treatment. This could be explained by the high
Novel and Intensified Process Design for Seawater RO Desalination Pre-treatment
223
quality water coming out of the GAC filter. After every experiment the permeability of the membrane could be completely restored by soaking the module in a 0.5% NaOCl solution overnight. The parameters identified in these experiments were used in the process design tool GUI developed in the theoretical work. Such parameters include membrane resistance, cake resistance, carbon attachment efficiencies, permeabilities etc. MF after GAC filtration
700
0,5 0,45
600
0,4 0,35 0,3
400
0,25 300
0,2
pressure (bar)
flux (lmh)
500
0,15
200 flux pressure
100
0,1 0,05 0
0 0
50
100 time (min)
150
Fig. 4. Flux and pressure development during membrane filtration of GAC-filtered seawater
3. Conclusion and Outlook In this work a combined theoretical-experimental approach was deployed in order to improve the sustainability of SWRO pre-treatment process design. An independent GUI flowsheeting tool was programmed to assist engineers in basic design of pre-treatment process including synthesis, simulation and evaluation concerning water costs and carbon emissions. Additionally, experimental work was done to both identify model parameters and test out new process configurations, for example a coagulation-free intensified GACMF filtration process. The results show that such a process both technically feasible as well as beneficial in waters containing a large organics fraction. Literature [1] Pankratz T.: Water Desalination Report, 44 (2008), 1. [2] Gasson C.: International Desalination Association Newsletter 11-12 (2010), 7.
224
M. Beery et al.
[3] Beery M., Repke J.-U.: Desalination and Water Treatment, 16 (2010), 218–228. [4] Westerberg A.W.: Proc. Chemical Engineering Centennial Symposium (2004), 53-60. [5] Dudley J., Dillon G., Rietveld L.C.: Journal of Water Supply: Research and Technology, 57 (2008), 13-21. [6] Henthorne L.: International Desalination Association Journal, 3 (2010), 12-13. [7] Beery M., Wozny, G., Repke J.-U.: Computer-Aided Chemical Engineering, 29 (2011), 1286-1290. [8] El-Dessouky H.T., Ettouney H.M : Fundamentals of Salt Water Desalination, Elsevier 2002. [9] Voutchkov N.: Pretreatment Technologies for Membrane Seawater Desalination, Australian Water Association, Sydney 2008. [10] Busch M., Chu R., Rosenberg S.: International Desalination Association Journal, 2 (2010), 56–71. [11] Allgeier S., Alspach B., Vickers J.: Membrane Filtration Guidance Manual, United States Environmental Protection Agency, 2005
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
MICHAŁ DYLĄG*, JERZY ROSIŃSKI∗∗, JERZY KAMIEŃSKI∗∗
ANWENDUNG DES MITTLEREN GESCHWINDIGKEITSGRADIENTEN AUF DIE MODELLIERUNG VON FLOCKUNGSPROZESSEN Abstract In der Arbeit ist das Ergebnis der Vergleichsrechungen zwischen einem als ideal durchmischt angenommenen Reaktor und einem ortsdiskret betrachteten Flockungsapparat dargestellt. Die Wiedergabe der realen Flockungsgrößenverteilungen ist in jedem Fall bei der Simulation eines ortsdiskreten Reaktors besser als für einen als ideal durchmischt angenommenen Flockungsapparat. Keywords: Koagulation, Aggregation, Partikelwechselwirkungen, Populationsbilanzen, Primärpartikel und Flockengrößenverteilungen. 1. Einleitung Die Modellierung und Simulation von Strömungsvorgängen in mehrphasigen Stoffsystemen war Gegenstand vieler Forschungsaktivitäten in den vergangenen Jahren. Zugängliche Methoden der Modellierung gaben keine Möglichkeit die Prozesse zu beschreiben, bei denen mit einer Veränderung von Partikeleigenschaftsverteilungen, z. B. Größenverteilungen, zu rechnen war. Erst in den 60er Jahren wurde die so genannte Populationstheorie veröffentlicht [1] und nach Überwindung der „Kinderkrankheiten“ mit großem Erfolg angewandt [2, 3]. Obwohl der Einfluss der Hydrodynamik auf die Partikelwechselwirkungen jedoch nur sehr vereinfacht formuliert wurde, konnte ab der ersten Dekade des 21. Jahrhunderts ein erheblicher Fortschritt festgestellt werden [4, 5, 6, 7]. Ein aus wirtschaftlich und umwelttechnisch relevanten Gesichtspunkten wichtiger Prozess, bei dem Partikelwechselwirkungen durch Teilchenagglomeration, Bruch und Erosion auftreten, ist der Vorgang der Flockung. Der Anwendungsbereich zieht sich von der Bioverfahrenstechnik (Proteingewinnung) über die Lebensmitteltechnik bis vor allem in den Bereich der Abwasserreinigung hin.
*
Institute of Advanced Manufacturing Technology, ul. Wrocławska 37a, 30-011 Kraków, Polska Cracow University of Technology, ul. Warszawska 24, 31-155 Kraków, Polska
∗∗
226
M. Dyląg et al.
2. Formulierung des Problems Eine wichtige Rolle spielt der Einfluss der Hydrodynamik im Flockungsreaktor, die bisher bei der Berechnung von Flockungsprozessen gar nicht oder oft nur in einer sehr untergeordneten Weise berücksichtigt wurde. Aufgrund des hohen Rechenaufwandes bei der numerischen Lösung der Populationsbilanzen wurden die Strömungs- und populationsdynamischen Simulationsrechnungen getrennt durchgeführt. Für den in dieser Arbeit betrachteten Rührkesselreaktor wurde zunächst in einer reinen Strömungssimulation das Strömungsfeld bestimmt [8, 9]. Die gewonnenen Ergebnisse dienten als Grundlage zur Berechung des Flockungsprozesses. Die angenommenen Voraussetzungen nach Patankar, d. h. dass der Reaktor als ideal durchmischt angesehen werden kann, folgt die Anwendung des Camp und Stein Konzeptes [13] zur Lösung der Populationsbilanzen. Sowohl intuitiv als auch nach dem Stand des Wissens muss diese Vereinfachung mit großem Zweifel betrachtet werden. Dem Vergleich des ortsdiskret betrachteten Reaktors mit dem als ideal durchmischt angenommenen Reaktor wird in dieser Arbeit besondere Beachtung geschenkt. Aufgrund der Komplexität der Modellgleichungen [5] hat sich gezeigt, dass in diesem Fall eine geschlossene Transformation der Populationsbilanzen zu Momentengleichungen nicht möglich ist. Stattdessen wird eine Art Hybrid-Verfahren angewandt, bei welchem ebenfalls nur noch die Momente der Partikelgrößenverteilungen von Primärpartikeln und Flocken als Unbekannte bestimmt werden, jedoch die kinetischen Ansätze der Partikelwechselwirkungen übernommen werden können, ohne die expliziete Durchführung der Integraltransformation. Wie bei den Populationsbilanzen handelt es sich um Integro-Differentialgleichungen, aber der Vorteil liegt in der wesentlich geringeren Anzahl der Unbekannten. Diese Vorgehensweise erfordert während des Simulationsablaufes die ständige Rekonstruktion der Partikelgrößenverteilungen aus den Momenten. Hierzu hat sich die Methode der Rücktransformation über eine vorgegebene Verteilungsfunktion als praktikabel erwiesen. Dabei werden beispielsweise für eine Normalverteilung die relevanten Parameter der Verteilungsfunktion (Mittelwert, Standardabweichung,…) aus den berechneten Momenten bestimmt, sodass die Verteilungsfunktion eindeutig definiert ist. Zur numerischen Behandlung der Populationsbilanzen in ihrer ursprünglichen Form sowie der Momentengleichungen wurde die Methode der Finiten Volumen angewandt. Auf dieser Basis erfolgte sowohl die Ortsdiskretisierung als auch die Eigenschaftsdiskretisierung der Bilanzgleichungen [6, 11, 12]. Dadurch konnte einer Simulationsrechnung jede beliebige Verteilung von Primärpartikeln und der Flocken als Startverteilung vorgegeben werden. Weitere Simulationsrechnungen zeigten, dass die tatsächliche Flockungsgrößenverteilung näherungsweise bestimmt werden konnte. Für die Genauigkeit des Rechenmodells spielt von allem die Wahl der Verteilungsfunktion, die der Rekonstruktion der Flockengrößenverteilung zugrunde liegt, eine wichtige Rolle. Prinzipiell ist festzuhalten, dass die Anwendung der Momentenmethode eine schnelle Abschätzung der sich einstellenden Flockengrößenverteilung erlaubt, die Rechenzeit liegt um etwa einen Faktor Fünf niedriger als bei Verwendung von anderen Methoden. Es ist jedoch zu beachten, dass dieses sogenannte Hybrid-Verfahren bei niedrigen Primärpartikelkonzentrationen und
Anwendung des mittleren Geschwindigkeitsgradienten-Konzeptes 227 auf die Modellierung von Flockungsprozessen hohen Reynoldszahlen versagt, wenn die resultierenden Flockengrößenverteilung sehr eng sein wird. 3. Zielsetzung Die praktische Auslegung von Flockungsreaktoren erfolgt auf der Basis des so genannten G-Wert-Konzepts, bei dem schweigend angenommen wird, dass ein Flockungsreaktor als ideal durchmischt betrachtet werden kann. Die getroffene Voraussetzung liegt trotz der nichtrealistischen Betrachtung weit vom realen Ablauf des Prozesses, bei dem zusätzlich ein Energieeintrag homogener gewährleistet werden soll. Diese unreale Betrachtung begründet die Vergleichsuntersuchungen zwischen einem als ideal durchmischt angenommenen und einem ortsdiskret betrachteten Reaktor. Um dieser Zielsetzung entgegen zu kommen, werden sowohl experimentelle als auch numerische Prozeduren der Rechnungen verwendet. 4. Numerische Formulierung und Lösung Die numerische Lösung der Populationsbilanzen und Momentengleichungen erfolgte mit Hilfe der Methode der Finiten Volumen. Zur vollständigen Diskretisierung von Populationsgleichungen wurde neben der Ortsdiskretisierung die Diskretisierung der Eigenschaftskoordinaten notwendig [6, 12]. 4.1. Diskretisierung des Rührkesselreaktors Betrachtet wurde ein Rührkesselreaktor mit 6-Blatt-Turbinenrührer mit vier Strombrechern. Dazu wird die Hälfte des Reaktorquerschnittes mit einem Rechengitter diskretisiert, welches sieben Bilanzelemente in axialer Richtung und fünf in radialer Richtung aufweist und damit wesentlich gröber ist, als das Rechengitter für die Strömungsberechnungen. Abb. 1 zeigt in der linken Hälfte den diskretisierten Reaktor, in den Abb. 2 und 3 das auf dem feinen Gitter berechnete [9] und gemessene [8] Strömungsfeld. Zusätzlich wurden die Ergebnisse der Strömungssimulationen [9] und LDE-Messungen [8] quantitativ verglichen. z
r
1 8
15 22 29
2 9
16 23 30
3 4 5 6 7
10 11 12 13 14
17 18 19 20 21
24 31 25 32 26 33 27 34 28 35
Abb. 1. Rührkesselreaktor mit diskreten Bilanzelementen zur Lösung der Populationsgleichungen
228
M. Dyląg et al.
Die Ergebnisse der detalierten Strömungssimulationen wurden unter Berücksichtigung der turbulenten Dispersionsströme für die neu generierten Bilanzelemente interpoliert und zusammengefasst und in Form eines minimal notwendigen Datensatzes dargestellt, auf welchen das Simulationsprogramm mit den Populationsbilanzen zugreift. Dieser enthält ● die Elementkoordinaten in radialer und axialer Richtung, ● die in ein Element tatsächlich eintretenden Volumenströme über die Grenzflächen in bei den Richtungen, ● die über das Element gemittelten Werte der turbulenten kinetischen Energie κ und der turbulenten Energiedissipation ε sowie, ● den Druck in jedem Bilanzelement. Für ein Bilanzelement, wie es in Abb. 1 dargestellt ist, werden in radialer und axialer Richtung Ein- und Austrittseiten definiert. Als Eintrittseite wird bezüglich jeder Koordinate die Seite bezeichnet, die in positiver Koordinatenrichtung als erste durchstoßen wird. Die konvektiven Terme in den Bilanzgleichungen werden mit Hilfe der über die Elementgrenzen fließenden Volumenströme formuliert. Für jedes Element wird aus der Summe der eintretenden Volumenströme, unter der Voraussetzung einer inkompressiblen Strömung, der aus Kontinuitätsgründen insgesamt austretende Volumenström berechnet. Die austretenden Flocken und Primärpartikeln entsprechen in ihrer Konzentration und Größenverteilung den Partikeln im Bilanzelement selbst. Diese Vorstellung entstammt dem sogenannten UPWIND-Schema vom Patankar, bei welchem jedes Bilanzelement als ein ideal durchmischter kleiner Rührkessel angenommen wird [10].
Abb. 2. Rührkesselreaktor – berechnetes Geschwindigkeitsfeld für Re = 104; der farbige Teil des Bildes beschreibt die berechnete Geschwindigkeitswerte (auch nach der Farbe erkennbar)
Anwendung des mittleren Geschwindigkeitsgradienten-Konzeptes auf die Modellierung von Flockungsprozessen
229
H [mm] 181 180 170 160 150 140 130 120 110 100 90 80 70 60 50 40 30 20 10 0 -10 -20 -30 -40 -50 -60 -70 -80 -90 -100 -105 0 5 15 25 35 45 55 65 75 85 95 105 115 125 135 143
r [mm] u [m/s] 0
0,5
Abb. 3. Rührkesselreaktor – gemessenes Strömungsfeld für Re = 104
4.2. Eigenschaftsdiskretisierung Die Diskretisierung der Eigenschaftskoordinate (Masse m) kann äquidistant oder nicht äquidistant erfolgen. Bei Diskretisierung wird aber der Bereich kleiner Partikeln nur durch wenige Intervalle beschrieben. Gerade in diesem Größenbereich liegen jedoch zumindest zu Beginn des Flockungsprozesses die meisten Partikeln vor. Diese Tatsache erzwingt eine weitgehende Umformung der Populationsgleichungen von Primärpartikeln und Flocken sowie Voraussetzung, dass die Eigenschaftsintervalle (Größenklassen) bei beiden Partikelarten im gemeinsamen Größenbereich jeweils identisch sind. Zusätzlich, da es sich aufgrund der Struktur der Quell- und Senkenterme um Integro-Differentialgleichungen handelt, müssen weiterhin die Integrationen durch entsprechende algebraische Formulierungen ersetzt werden. Leitgedanke und Problemlösung ist in der Arbeit [9] ausführlich dargestellt. 5. Experimentelle Untersuchungen Zur Validierung der Simulationsergebnisse wurden experimentelle Untersuchungen zur Flockulation von Quarzpartikeln in einem Rührkesselreaktor durchgeführt. Die Untersuchung der entstehenden Flocken hinsichtlich ihrer Größenverteilung erfolgte mit einem rechnergestützten Bildanalysesystem.
M. Dyląg et al.
230
Dieses Verfahren ist bezüglich der Flockenbeanspruchung wesentlich schonender als beispielsweise eine laserspektroskopische Untersuchung mit zusätzlichen Pump- und Rühreinrichtungen zur Aufwirbelung der Flocken in der Suspension. Der Versuchsaufbau, die Versuchsdurchführung und die Auswertung sind nachfolgend beschrieben. 5.1. Versuchsaufbau 9 8
analog Sygnal
digital Sygnal
7
10
11 6
5 4 16 Flokulant Feststoff
n Ms N entmineralisierte Wasser
15 14
12
Probe 1
17
13
2
1 - Reaktorkessel, 2 - Turbinenrührer, 3 - pH-meter, 4 - Messeinrichtung N, Ms i n, 5 - Motor, 6 - Mikroskop, 7 - CCD-Kamera, 8 - Bildverarbeitungsgerät, 9 - Rechner, 10 - Steuergerät, 11 - Enkoder, 12 - LDA-Sonde, 13 - Lichtleiter, 14 - optisches System, 15 - Ar-Ion Laser, 16 - Signalprozessor, 17 - Verschiebungssteuergerät
3
Abb. 4. Aufbau des Labor-Prüfstandes-Schema der Versuchsanlage
Die Koagulation der Quarzpartikeln wurde in einem Rührkesselreaktor aus Plexiglas mit dem Durchmesser ø300×6 und mit vier Strombrechern, welche jeweils um 90 Grad zueinander versetzt an der Reaktorwand angeordnet waren, untersucht. Die Strömung im Reaktor wurde durch einen 6-Blatt-Turbinenrühren mit einem Rührerdurchmesser dR = 1/3DR. Die angebrachte Messeinrichtung erlaubt sowohl die Drehzahl des Rührens als auch die eingebrachte Leistung online zu messen. Die Anordnung ist mit der Rührkesselkonfiguration identisch, für welche die Simulationen des Strömungsfeldes durchgeführt wurden, die den populationsdynamischen Rechnungen zugrunde liegen. Damit ist eine direkte Vergleichbarkeit (Ähnlichkeit) von Rechen- und Messergebnissen gewährleistet. 5.2. Versuchsdurchführung Zur Durchführung der Flockung wurde ein handelsüblicher Quarzstaub verwendet. Dessen Größenverteilung wurde mit einem Laserbeugungsspektroskop (Fa. Fritsch) bestimmt und bei den Simulationsrechnungen entsprechend als vorgegebene Primärpartikelverteilung angenommen. Zur Durchführung der Koagulation wurde dieser Quarz zunächst dem Reaktor zugegeben, der mit vollentsalztem Wasser gefüllt war. Mit
Anwendung des mittleren Geschwindigkeitsgradienten-Konzeptes 231 auf die Modellierung von Flockungsprozessen Hilfe des Rührers wurden die Quarzpartikeln ca. fünf Minuten dispergiert, sodass zu Beginn der Koagulation keine undefinierten Agglomerate vorlagen und eine gleichmäßige Verteilung der Quarzpartikeln in der Suspension sichergestellt war. Anschließend wurden 40 g Calciumchlorid CaCl2 und 40 g Natriumhydroxid NaOH in jeweils 2 l voll entsalztem Wasser gelöst, und die Lösungen wurden an zwei gegenüberliegenden Stellen an der Flüssigkeitsoberfläche nacheinander zugegeben. Danach wurde das entsalzte Wasser dem Behälter zugegeben, bis die Einfüllhöhe den Wert 288 mm erreicht hatte. Die notwendige Zeit bis zur vollständigen Vermischung der Salze mit der Suspension ließ sich aus Ähnlichkeitsgesetzen abschätzen und betrug einige Sekunden. Zusätzlich wurde die Grenze der Löslichkeit bestimmt und festgestellt, dass der Grenzwert nicht überschritten ist, sodass keine Fällungsflockung auftritt. Durch Chemikalienzugabe stellt sich ein pH-Wert 12,1 ein. Durch hohe Dosierung des Koagulationsmittels ist die vorher getroffene Modellannahme einer Agglomerationseffektivität α’ = 1 gerechtfertigt. Sonst müsste zusätzlich ein Modell zur Bestimmung der Agglomerationseffektivität in das Simulationsprogramm implementiert werden. Dem Flockungsreaktor wurde bei laufendem Rührer vorsichtig eine Suspensionsprobe von ca. 7 µl entnommen. 5.3. Versuchsauswertung Das Funktionsschema der Bildanalyse zur Bestimmung von Partikelgrößenverteilungen ist aus Abb. 4 abzulesen. Das im Mikroskop Olympus sichtbare Probenbild wird von einer am Phototubus des Mikroskops angebrachten CCD Kamera (Sony) mit 256 Graustufen aufgenommen und in entsprechende Analogsignale umgesetzt. Diese werden von einer Bildverarbeitungskarte im angeschlossenen PC in digitale Signale umgewandelt und von einem Bildverarbeitungsprogramm eingelesen. Zur Vereinfachung des Analyseablaufs und zur besseren Reproduzierbarkeit der Messungen wurden ausgefeilte Bildauslesesysteme angewandt. Um die teilweise porösen Flocken als jeweils einzelne Partikeln erfassen zu können, wird noch der Vorgang des „Schließens“ auf das binarisierte Bild angewandt. Dabei werden die Hohlräume innerhalb einer Flocke ausgefüllt, sodass einzelne Flocken eindeutig erkennbar sind und ausgewertet werden können. In Abb. 5 sind die unbearbeiteten Photos einer Flockenprobe zu sehen. Schließlich wird die Größenverteilung der Partikeln und Bezug in einem charakteristischer Durchmesser ermittelt.
Abb. 5. Unbearbeitete Photographien einer Flockenprobe
M. Dyląg et al.
232
6. Auswertung und Diskussion der Ergebnisse – Vergleich mit dem Modell des ideal durchmischten Reaktors Simulationsrechungen wurden für einen Rührkesselreaktor bei Reynoldszahlen von Re = 104 und Re = 2,0·104 durchgeführt. Zur Lösung wurde die früher beschriebene Methode der Finiten Volumen angewandt. Nachfolgend sind die Ergebnisse der berechneten und gemessenen Partikelgrößenverteilungen von Primärpartikeln und Flocken in Reaktor im stationären Zustand dargestellt. Die Partikelwechselverteilungen werden als massenbezogene Durchgangssumenwerte in Abhängigkeit eines kugeläquivalenten Partikeldurchmesser aufgetragen. Die Nummerierung der Bilanzelemente bezieht sich auf die in Abb. 1 dargestellte Reaktordiskretisierung. Zur Auslegung von Flockungsreaktoren wird häufig das von Camp und Stein [13] vorgeschlagene G-Wert-Konzept angewandt. Die durch ein Rührwerk (oder eine Pumpe) in einem Flockungsreaktor eingebrachte Leistung P wird dabei durch den resultierenden mittleren Geschwindigkeitsgradienten G nach Gl. (1) charakterisiert, wobei ε = P/mR mit der Masse mR des Reaktorinhalts ist. G=
ρ⋅ε η
(1)
Dabei stellt G sowohl ein Maß für die Häufigkeit von Partikelkollisionen als auch für die Flockenzerstörung durch Erosion und Bruch dar. Der Reaktor wird auf dieser Grundlage als ideal durchmischt vorausgesetzt, obwohl in vielen Arbeiten [2, 6] auf die Bedeutung der im Flockungsreaktor vorliegenden unterschiedlichen lokalen Strömungsverhältnisse hingewiesen wird. In Abb. 6 ist der Vergleich zweier Rechenergebnisse mit unterschiedlichen Modellvoraussetzungen dargestellt. Die durchgezogene Linie links zeigt die gemittelte Flockengrößenverteilung im Rührkesselreaktor wie in Abb. 4 der Literatur [15], unter der Berücksichtigung von 35 Bilanzelementen mit ihrem jeweiligen lokalen Energieeintrag und der anschließenden Mittelung der lokalen Flockengrößenverteilungen entsprechend der Volumenanteile der einzelnen Bilanzelemente. Die gestrichelte Linie zeigt die mittlere Flockengrößenverteilung im Reaktor unter der Annahme idealer Durchmischung. Vor der Lösung der Populationsgleichungen wurden die lokalen Energieeinträge in den einzelnen Bilanzelementen entsprechend ihrer Volumenanteile zu einem für den gesamten Reaktor konstanten mittleren Energieeintrag gemittelt. Dieser liegt für Re = 104 bei ε = 0,0225 W/kg. Die Rechnung zeigt, dass sich unter der Annahme des ideal durchmischten Reaktors eine Flockengrößenverteilung ergibt, die bei wesentlich größeren Flocken liegt. Ursache ist die Vernachlässigung des signifikanten Einflusses hoher Energiedissipationsraten in wenigen Elementen in Rührernähe. In diesen Bereichen ist die Zerstörung der Flocken so stark ausgeprägt, dass sich die Flockengrößenverteilung für den gesamten Reaktor bei wesentlich kleineren Werten einstellt, ein Ergebnis, welches auch durch die zuvor beschriebenen Experimente bestätigt wurde [15]. Diese lokal auftretende Zerkleinerung der Agglomerate wird unter der Voraussetzung des ideal durchmischten Reaktors nicht berücksichtigt. Insbesondere können bei dieser vereinfachten Betrachtungsweise auch lokale Unterschiede in der Flockungsgrößenverteilung, nicht wiedergegeben werden [14].
Anwendung des mittleren Geschwindigkeitsgradienten-Konzeptes 233 auf die Modellierung von Flockungsprozessen Dieses Ergebnis zeigt die Problematik bei der Anwendung des G-Wert-Konzepts, da Inhomogenitäten im (turbulenten) Strömungsfeld und deren Folgen durch diese empirische Gesetzmäßigkeit nicht berücksichtigt werden. 100
Durchgang [%]
80 60 40 Simulation εlok ideale Durchmischung
20 0
50
100 150 200 Flockendurchmesser [μm]
250
300
Abb. 6. Vergleich der über der Reaktor gemittelten stationären Flockungsgrößenverteilung bei ortsdiskreten und ideal durchmischt angenommenen Reaktor mit Re = 104, Primärpartikelkonzentration cp = 1 g/l
Führt man denselben Vergleich bei einer geringeren Primärpartikelkonzentration durch (cp = 0,25 g/l), so könnte unter Berücksichtigung der in Abb. 4 der Literatur [15] gezeigten Ergebnisse, dass Flockengrößenverteilung in den verschiedenen Elementen in stationären Zustand zusammenfällt, der Eindruck entstehen, dass der Reaktor für diesen Fall tatsächlich als ideal durchmischt betrachtet werden darf. Anhand des in Abb. 7 gezeigten Ergebnisses wird jedoch deutlich, dass dies nicht der Fall ist. 100
Durchgang [%]
80 60 40 Simulation εlok ideale Durchmischung
20 0
50
100
150
200
250
300
Flockendurchmesser [μm]
Abb. 7. Vergleich der über den Reaktor gemittelten stationären Flockengrößenverteilung bei ortsdiskreten und ideal durchmischt angenommenen Reaktor mit Re = 104; Primärpartikelkonzentration cp = 0,25 g/l
M. Dyląg et al.
234
Die durchgezogen dargestellte Durchgangssummenkurve zeigt die für den Reaktor berechnete mittlere Partikelgrößenverteilung, welche wieder durch Mittelung der Partikelgrößenverteilungen in den 35 diskreten Bilanzelementen bestimmt wurde. Die gestrichelte Kurve stellt die berechnete Flockengrößenverteilung unter der Annahme eines mittleren Energieeintrags im gesamten Reaktor dar. Auch für diesen zweiten vereinfachten Fall liegt die Flockengrößenverteilung insgesamt bei kleineren Flocken mit dem Ergebnis bei höherer Primärpartikelkonzentration nach Abb. 6. Ursache für die unterschiedlichen Ergebnisse ist nach wie vor die Annahme eines gemittelten Energieeintrags im Reaktor, sodass der Einfluss der hohen Energiedissipationsrate in einigen wenigen Elementen im Bereich des Rührers und die dort auftretende Flockenzerstörung nicht mehr berücksichtigt werden. Der Reaktor kann also nur unter der Voraussetzung als ideal durchmischt betrachtet werden, wenn auch die Turbulenzstruktur möglichst homogen ist! Bei der Berechnung der Flockungsprozesse im Rührkesselreaktor bei Re = 2,2·104 wurde neben der Anwendbarkeit des Rechenmodells auf der Basis der Finite-VolumenMethode wieder der Einfluss der Primärpartikelkonzentration sowie die Folge der Modellannahme eines ideal durchmischten Reaktors untersucht. In Abb. 8 sind die berechneten und gemessenen Primärpartikel- und Flockengrößenverteilungen im stationären Zustand dargestellt. 100
Durchgang [%]
80 60 40
Primärpartikeln 18 Bilanzelement 1 Bilanzelement Simulation
20 0
20
40 60 80 100 120 Partikel- / Flockendurchmesser [μm]
140
Abb. 8. Gemessene stationäre Flockengrößenverteilungen in dem Elementen 18 und 1 des Rührkesselreaktor bei Re = 2,2·104 im Vergleich mit der berechneten mittleren Flockengrößenverteilung im Reaktor; Primärpartikelkonzentration cp = 1,0 g/l
Anhand der Messung ist festzustellen, dass bei den sich im stationären Zustand einstellenden Flockengrößenverteilungen im Reaktor ebenfalls lokale Unterschiede auftreten. Durch das Rechenmodell können diese lokalen Unterschiede nicht mehr wiedergegeben werden. Die berechneten Flockengrößenverteilungen sind für alle Bilanzelemente identisch, in Abb. 8 ist daher als Simulationsergebnis die für den gesamten Reaktor berechnete mittlere Flockengrößenverteilung dargestellt. Offensichtlich werden
Anwendung des mittleren Geschwindigkeitsgradienten-Konzeptes 235 auf die Modellierung von Flockungsprozessen vom Rechenmodell die Partikelwechselwirkungen als zu langsam beschrieben, so dass die Konzentrationsunterschiede durch die Konvektion vollständig ausgeglichen werden. Trotzdem ergibt sich durch die Simulationsrechnung eine mittlere Flockengrößenverteilung, die die realen Verhältnisse im Reaktor zufriedenstellend wiederspiegelt. Insgesamt sind die Flocken kleiner als bei einer Reynoldszahl Re = 104 bei derselben Primärpartikelkonzentration. Durch den erhöhten Energieeintrag werden größere und damit instabilere Flocken in Strömungsfeld jetzt nicht nur in unmittelbarer Rührernähe, sondern im gesamten Reaktor schneller wieder zerstört. 100
Durchgang [%]
80 60 40
Primärpartikeln 18 Bilanzelement 1 Bilanzelement Simulation
20 0
10
30 40 50 20 60 Partikel- / Flockendurchmesser [μm]
70
Abb. 9. Gemessene stationäre Flockungsgrößenverteilungen in den Elementen 18 und 1 des Rührkesselreaktors bei Re = 2,2·104 im Vergleich mit den beachteten mittleren Flockungsgrößenverteilung im Reaktor, Primärpartikelkonzentration cp = 0,5 g/l
Ein entsprechendes Resultat erhält man wieder bei einer geringeren Primärpartikelkonzentration von 0,5 g/l, das in Abb. 9 aufgetragen ist. Die Messung zeigt, dass die Flockungsgrößenverteilungen in den verschiedenen Bereichen des Reaktors eng beieinander liegen, die Ursuche wurde bereits bei Untersuchungen zur Re = 104 ausführlich diskutiert. Wie schon bei der höheren Primärpartikelkonzetrationen können auch hier mit dem Rechenmodell die lokalen Unterschiede nicht mehr wiedergegeben werden, für alle Bilanzelemente im Reaktor wird die gleiche Verteilung berechnet, welche jedoch die mittlere Größenverteilung wieder gut beschreibt. Schließlich soll auch hier ein Vergleich mit dem Modell des ideal durchströmten Rührkesselreaktors erfolgen. Abb. 10 zeigt das entsprechende Ergebnis.
M. Dyląg et al.
236 100
Durchgang [%]
80 60 40 Simulation ideale Durchmischung
20 0
10
20
30 40 50 60 70 Flockendurchmesser [μm]
80
90
100
Abb. 10. Vergleich der über den Reaktor gemittelten stationären Flockungsgrößenverteilung bei ortsdiskreten und ideal durchmischt angenommenem Reaktor mit Re = 2,2·104; Primärpartikelkonzentration cp = 1,0 g/l
Überaschenderweise stellt sich durch diese Vereinfachung eine Flockengrößenverteilung ein, die von der mit dem ortsdiskreten Rechenmodell bestimmten zu kleineren Flocken hin abweicht. Die mittlere Energiedissipationsrate im Reaktor ist zu hoch, um die in Bilanzelementen geringerer lokaler Energiedissipation überwiegenden Koagulationsvorgänge ausreichend zu beschreiben. Die Annahme eines ideal durchmischten Reaktors erweist sich auch hier als eine zu starke Vereinfachung! 7. Zusammenfassung und Schlussfolgerungen Ingesamt kann durch die Ergebnisse festgestellt werden, dass die Bildung enger Flockengrößenverteilungen bei Koagulationsvorgängen wie z. B. in einem Strömungsfeld mit höheren Reynoldszahlen auftreten, vom vorliegenden Rechenmodell nur noch näherungsweise wiedergegeben werden können. Jedoch liegen in Flockungsreaktoren selten Strömungsbedingungen mit Re > 2,0·104 vor, so dass die aufgezeigten Grenzen des Rechenmodells für praktische Anwendungsfälle kaum relevant sind. Dagegen ist bei der Simulation von Flockulationsprozessen eine gute Wiedergabe der sich einstellenden Flockengrößenverteilungen zu erwarten, da bei der Flockung mit Polymeren erfahrungsgemäß deutlich größere Flocken und damit breitere Flockengrößenverteilungen entstehen als bei Koagulationsvorgängen. Die Annahme eines ideal durchmischten Rührreaktors ist jedoch für alle hier besprochenen Anwendungen eine zu große Vereinfachung. Die Wiedergabe der realen Flockengrößenverteilungen ist in jedem Fall bei der Simulation eines ortsdiskreten Reaktors besser als für einem als ideal durchmischt angenommenen Reaktor. Die praktische Auslegung von Flockungsreaktoren erfolgt seit vielen Jahren auf Basis des sogenannten G-Wert-Konzepts von Camp und Stein, bei dieser Vorstellung wird ein Flockungsreaktor als ideal durchmischt mit einem homogenen Energieeintrag betrachtet, beispielweise ein Rührwerk. Vergleichsrechnungen zwischen einem als ideal durchmischt
Anwendung des mittleren Geschwindigkeitsgradienten-Konzeptes 237 auf die Modellierung von Flockungsprozessen angenommen und einem ortdiskret betrachteten Rührkesselreaktor bei der Reynoldszahl Re = 104 zeigen jedoch, dass bei der Annahme idealer Durchmischung die entstehende Flockengrößenverteilung bei viel zu großen Flockendurchmessern liegen. Die tatsächlichen Größenverteilungen können nur unter Anwendung des ortsdiskreten Reaktorsmodells in Übereinstimung mit Messergebnissen gut wiedergegeben werden. Ursache ist die Vernachlässigung des hohen lokalen Energieeintrags im Nahbereich des Rührers, welche zu hohen lokalen Bruch- und Erosionsraten führt. Eine Übertragung des Rechenmodells auf einen Rührkesselreaktor bei einer Reynoldszahl Re = 2,2·104 zeigt dessen Extrapolierbarkeit. Die gemessenen lokalen Unterschiede der Flockengrößenverteilung im Reaktor zeigen aber wieder eine gute Übereinstimmung mit Messergebnissen, auch bei unterschiedlichen Feststoffkonzentrationen. Insbesondere stellt sich auch hier die Annahme eines ideal durchmischten Reaktors zur Auslegung des Flockenprozesses als ungeeignet heraus.
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]
Literatur Hulburth H. M., Katz S.: Chem. Eng. Sci., 19, (1964), 555-574. Bałdyga J., Boorne J. R.: Turbulent Mixing and Chemical Reactions, John Wiley & Sons, Chichester 1999. Bałdyga J., Orciuch W.: Chem. Eng. Sci., 56, (2001), 2435-2444. Marchisio D. L.,Vigil R. D., Fox R. O.: Chem. Eng. Sci., 58, (2003), 3337-3351. Thomas D. N., Judd S. J., Fawcet N.: Wat. Res., 7, (1999), 1579-1992. Jaworski Z.: Numeryczna mechanika płynów w inżynierii chemicznej i procesowej, Akademicka Oficyna Wydawnicza EXIT, Warszawa 2005. Gierczycki A.: Powstawanie i rozpad agregatów ciała stałego zawieszonych w cieczy, Zeszyty Naukowe Politechniki Śląskiej, Gliwice 2005. Talaga J., Wójtowicz R.: Inż. i Ap. Chem., 4, (2010), 80-81. Wójtowicz R.: Inż. i Ap. Chem., 4, (2010), 92-93. Patankar S. V.: Numerical Heat Transfer and Fluid Flow, New York, Hemispheres Publishing Corporation, 1980. Dyląg M., Kamieński J., Rosiński J.: Inż. i Ap. Chem., 2, (2010), 35-36. Dyląg M., Kamieński J., Rosiński J., Szatko W.: Inż. i Ap. Chem., 4, (2010), 22-23. Camp T. R., Stein R. C.: J. Boston Soc. Civ. Eng., 4, (1943), 219-237. Dyląg M., J. Rosiński J., Kamieński J.: Czasopismo Techniczne, 5-M, (2008), 89-96. Dyląg M., J. Rosiński J., Kamieński J.: Czasopismo Techniczne, 9-M, (2008), 3-18.
18th International Conference Process Engineering and Chemical Plant Design Günter Wozny and Łukasz Hady, Editors Copyright © 2011, Berlin Institute of Technology. All rights reserved.
CONTENTS Barbara Tal-Figiel Solid-liquid extraction from plants with a bidisperse porous structure – experimental kinetics and modeling
2
Z. Guetta, J.C. Schöneberger, H. Arellano-Garcia, H. Thielert, G. Wozny Development and Experimental Verification of a Claus Process Combustion Chamber Model
12
M. Gula, J. Steinbach, K. Holtappels, A. Acikalin; H.-P. Schildberg; A. Kobiera Safety related characteristic of chemically unstable gases
22
Niklas Paul, Matthias Kraume Influence of surfactants on fluid dynamics and mass transfer of single droplets
28
Jan Talaga Untersuchungen zur Fluiddynamik von ein- und zweiphasigen Rührwerkströmungen
38
Philipp Schrader, Udo Dorn, Andres Kulaguin-Chicaroux, Sabine Enders Phase Equilibria of Surfactant Containing Systems
49
Zdzisław Jaworski, Eugeniusz Cydzik 59 Customization of thermodynamic models of electrolytes for incorporation in CFD codes Kai Langenbach, Sabine Enders Cross-Association of Multi-Component Systems
67
Aneta Głuszek Electric energy use and ecological analysis for production of flat-plate solar collectors
77
George Tsatsaronis, Tatiana Morosuk Advanced Exergetic Analysis of Energy-Intensive Processes
83
M. N. Cruz Bournazou, D. Domashk, T. Barz, G. Wozny, H. Arellano-Garcia Evaluation of Integration Approaches to DAE Systems in Engineering Applications
93
Wiesław Szatko, Walerian Bliniczew, Janusz Krawczyk Comparison of mathematical models describing changes of the suspension absorption capacity and thermal resistance of the sludge
103
V. A. Merchan, S. Kuntsche, H. Arellano-Garcia, G. Wozny Symbolic generation of higher-order derivatives with MOSAIC
114
Contents
239
Ryszard Wójtowicz The vibromixers – a current state of research and trends of further investigations
124
Jerzy Bałdyga, Magdalena Jasińska Reactive mixing and dispersion processes in rotor-stator devices
135
Jerzy Kamieński, Andrzej Duda The dual impeller capacities in the light of CMA model
145
R. Günther, J.C. Schöneberger, H. Arellano-Garcia, H. Thielert, G. Wozny Design and modelling of a new periodical-steady state process for the oxidation of sulfur dioxide in the context of an emission free sulfuric acid plant
156
Janusz Magiera Mensch-Energie-Umwelt, reale Möglichkeiten der Beschaffung ökologisch sauberer Wärmeenergie für private Haushalte
168
Krzysztof Neupauer Die Arbeitsergebnisse einer Heizungsanlage mit drei Energiequellen und ihre Steuerung in dem Online-System.
178
Steffen Stünkel, Günter Wozny CO2 capture and utilization for the oxidative coupling of methane process
188
Walter Martini, Günter Wozny Hyperbranched Polymers for CO2 Capture: Data Estimation and Process Simulation
198
Verena Löffler, Matthias Kraume Development of mixed-matrix-membranes for separation of gaseous hydrocarbons
207
Matan Beery, Ji Jung Lee, Joon Ha Kim, Jens-Uwe Repke, Günter Wozny Novel and Intensified Process Design for Seawater RO Desalination Pre-treatment
217
Michał Dyląg, Jerzy Rosiński, Jerzy Kamieński Anwendung des mittleren Geschwindigkeitsgradienten-Konzeptes auf die Modellierung von Flockungsprozessen
225