Upload
duongliem
View
219
Download
0
Embed Size (px)
Citation preview
10.06.2016
1
Liquid-liquid extraction design
under uncertainty
ISEC 2014
Bettina Rüngeler1, Andreas Pfennig2
1AVT.FVT – Fluid Separation Processes, RWTH Aachen University,
Germany2CEET – Institute of Chemical Engineering and Environmental Technology,
TU Graz, Austria
2
Outline
� Motivation
� Framework for the design under uncertainty
� Resource allocation
� Cascaded option trees
� Conclusion
Liquid-liquid extraction design under uncertainty
10.06.2016
2
3
� new solvent extraction processes
• high-viscous media
• fermentation broths
• aqueous two-phase systems
� low expertise in design and scale-up
� high uncertainties in knowledge and data
Goal: Optimal resource allocation in design of common
and new solvent extraction processes
Motivation
Liquid-liquid extraction design under uncertainty
4
Uncertainty as supporting information
� uncertainty of input data quantifiable
� uncertainty reducible by process development steps
� -> explicit consideration of uncertainties in all results
� analysis of sensitivities towards uncertainty levels
� systematic reduction of uncertainty
� economics as criterion for the selection of best option
Liquid-liquid extraction design under uncertainty
10.06.2016
3
5
Design under uncertainty
Liquid-liquid extraction design under uncertainty
extraction
model
uncertainty in
input parameters
uncertainty in
results
p(Y)
Yp(Z)
Z
process options
6
Example process
Liquid-liquid extraction design under uncertainty
process options:
ethylbenzene
solvent device
n-decanol
methyl isobutyl ketone (MIBK)
mixer-settler
column
10.06.2016
4
7
Economics of various options
Liquid-liquid extraction design under uncertainty
0.2 0.4 0.6 0.8 1.00
10
20
30
40
400
500
600
700
pro
bab
ility
den
sity fun
ction
cost of product [EUR/kg]
ethylbenzene, column n-decanol, column
ethylbenzene, mixer-settler MIBK, mixer-settler
8
Design under uncertainty
Liquid-liquid extraction design under uncertainty
extraction
model
uncertainty in
input parameters
uncertainty in
results
p(Y)
Yp(Z)
Z
process options
Best process
determinable?
no
yes extraction
process
10.06.2016
5
9
modKL as distinctness measure
� modified Kullback-Leibler distance
� modification for symmetry: modKL12=modKL21
� comparison of distribution p(x) and q(x):
dxxqxp
xqxqdx
xqxp
xpxp
xqxpxm
∫∫∞
∞−
∞
∞−+
++
=
+=
))()((
)(2ln)(
2
1
))()((
)(2ln)(
2
1KLmod
))()((2
1)(
Berkelmans, I., 2010: Development and Application of a Framework for Technology and Model
Selection Under Uncertainty. Massachusetts Institute of Technology
69.02lnKLmod
:)()(
0KLmod
:)()(
≈=
≠
=
=
xqxp
xqxp value of modKL distinctness
0 identical
modKL < 0.65 slightly different
≥ 0.65 strongly different
0.69 totally different
10
Comparison of processes
Liquid-liquid extraction design under uncertainty
ethylbenzene column vs.
ethylbenzene mixer-settler
ethylbenzene column vs.
n-decanol column
ethylbenzene column vs.
MIBK mixer-settler
0.0
0.2
0.4
0.6
mo
dK
L
modKL decision bound
10.06.2016
6
11
Economics of remaining options
Liquid-liquid extraction design under uncertainty
0.00 0.05 0.10 0.15 0.20 0.25 0.300
10
20
30
cost of product [EUR/kg]
column
mixer-settler
ethylbenzene
modKL: 0.12
0.00 0.05 0.10 0.15 0.20 0.25 0.300
10
20
30
pro
ba
bili
ty d
en
sity fu
nctio
n
12
Design under uncertainty
Liquid-liquid extraction design under uncertainty
extraction
model
uncertainty in
input parameters
uncertainty in
results
p(Y)
Yp(Z)
Z
process options
optimal resource
allocationexperiment
literature
modeling
Best process
determinable?
no
yes extraction
process
10.06.2016
7
13
� methods for uncertainty reduction
� reduce uncertainty for input parameters
� associated data
Definition of design steps
Liquid-liquid extraction design under uncertainty
distribution
coefficient
density
phase 1density
phase 2
viscosity
phase 1viscosity
phase 2
settling
time
interfacial
tension
…
literature experiment modeling
new uncertainty level analysis costs
14
Change of modKL values
Liquid-liquid extraction design under uncertainty
0.0 0.2 0.4 0.60
10
20
determination of
density of solvent
settling time
distribution coefficient
pro
bab
ility
de
nsity fun
ctio
n
modKL
original
modKL-
value
modKL
decision
bound
10.06.2016
8
15
New distribution after analysis
Liquid-liquid extraction design under uncertainty
0.00 0.05 0.10 0.15 0.20 0.25 0.300
20
40
60
80
100
120
140
160
180
cost of product [EUR/kg]
column
mixer-settler
ethylbenzene
modKL: 0.69
0.00 0.05 0.10 0.15 0.20 0.25 0.300
20
40
60
80
100
120
140
160
180
pro
ba
bili
ty d
en
sity fu
nctio
n
16
Resource allocation
� How to choose next design step?
cost – benefit analysis
� cost:
• cost for research (work and material)
� benefit:
• reduced uncertainty
• less risk for decision on best process
Liquid-liquid extraction design under uncertainty
10.06.2016
9
17
Clear summary of design steps
� cascaded option tree
Bednarz, A., Rüngeler, B., Pfennig, A., 2014: Use of Cascaded Option Trees in Chemical-
Engineering Process Development, Chemie Ingenieur Technik 86 (5)
18
Conclusion
� use of uncertainties as additional information to process
� support for process engineer by systematic resource allocation
� remaining uncertainties at each step of process development
visible
� framework for design of solvent extraction under uncertainties
Liquid-liquid extraction design under uncertainty
10.06.2016
10
Thank you for your attention.
Liquid-liquid extraction design under uncertainty
Bettina Rüngeler
Aachener Verfahrenstechnik
Fluid Separation Processes
Wüllnerstr. 5
52064 Aachen, Germany
20
� Bednarz, A., Rüngeler, B., Pfennig, A., 2014: Use of Cascaded Option Trees in Chemical-
Engineering Process Development, Chemie Ingenieur Technik 86 (5), DOI:
10.1002/cite.201300115
� Berkelmans, I., 2010: Development and Application of a Framework for Technology and
Model Selection Under Uncertainty. Dissertation, Massachusetts Institute of Technology.
� Biegler, L. T., Grossmann, I. E., Westerberg, A. W., 1997: Systematic methods of
chemical process design. Upper Saddle River, N.J. Prentice Hall PTR.
� Guthrie, K. M., 1969: Capital cost estimating. Chemical Engineering (March 24), 114–142.
Liquid-liquid extraction design under uncertainty