modeling and assessment of propylene/propane …
TRANSCRIPT
MODELING AND ASSESSMENT OF PROPYLENE/PROPANE
SEPARATION BY PRESSURE SWING ADSORPTION ON SiCHA
MONA KHALIGHI
(B. Eng, M.Sc, Sharif University of Technology)
A THESIS SUBMITTED
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
DEPARTMENT OF CHEMICAL AND BIOMOLECULAR ENGINEERING
NATIONAL UNIVERSITY OF SINGAPORE
2013
I
Acknowledgement
In the name of Allah, the Beneficent, the Merciful. This research would not have been
possible if not for the help of numerous people. Most importantly is the support and
understanding that my husband and my family gave me throughout the PhD candidature.
Without their support I would not have sustained through the stress and frustration of this
research.
I would like to express my deepest gratitude to my supervisors, Professor Karimi, and
Professor Farooq for their advice, support, patience, and encouragement throughout the
course of this research. It is not often that one finds advisors who are always energetic
and active in academic field. I am also grateful that despite their busy schedule they
managed to allocate time to read and comment critically on my papers and thesis.
I would like to express my special thanks to Ms. Shilpi Aggarwal and Mr. Sadegh
Tavallali, who spent a lot of time and energy to discuss many research topics on
optimization and programming. I am also grateful to my officemates Ms. Hanifah
Widiastuti, and Mr. Susarla Naresh who provided several valuable suggestions for my
research.
My genuine acknowledgement is given to National University of Singapore for
providing research scholarship. I also hope that the ideas proposed in this thesis will
bring about improvement to the Propylene/Propane separation. I dedicate this thesis to
my husband Alireza, who has brought much love into my life, my mother, father, and
brothers for their unwavering support.
b
I here
been writt
sourc
This t
eby declar
ten by me
es of info
thesis has
re that the
e in its en
ormation w
s also not
unive
___
MO
II
Declara
e thesis is
tirety. I h
which ha
t been sub
ersity prev
________
ONA KH
ation
s my orig
have duly
ave been u
bmitted fo
viously.
________
HALIGHI
ginal work
acknowl
used in th
or any de
_
k and it h
ledged all
he thesis.
gree in an
has
l the
ny
C
A
S
N
L
L
C
C
Contents
Acknowledge
ummary……
Nomenclature
List of Figure
List of Tables
CHAPTER 1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
CHAPTER 2
2.1
2.2
ement I
………………
e………….
es………….
s………… .
INTR
Selectivity
Zeolites .....
Adsorption
Pressure sw
PSA cycles
Objectives a
Structure of
LITER
Propylene/p
Different ad
…. ...............
....................
...................
....................
ODUCTION
....................
....................
mechanism
wing adsorpti
...................
and scopes ..
f this thesis ..
RATURE RE
propane sepa
dsorbents for
III
....................
....................
....................
....................
N .................
....................
....................
s ..................
ion (PSA) ....
....................
....................
....................
EVIEW ......
aration .........
r propylene/
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
/propane sep
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
paration ........
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
.. VIII
.... XI
... XV
. XXI
.......1
...... 2
...... 4
...... 7
...... 8
.... 10
.... 13
.... 15
.....16
.... 16
.... 17
C
2.3
2.3.1
2.4
2.4.1
2.5
2.6
2.7
CHAPTER 3
PSA Proce
3.1
3.1.1
3.2
3.3
3.3.1
3.3.2
3.4
3.5
3.5.1
Characterist
Synthe1
SiCHA stru
Diffusi1
Pressure sw
Simulation
Optimizatio
Non-i
ess 42
Pore diffusi
Bounda1
Bi-LDF mo
Propylene/p
Adsorp1
Contro2
Process per
Model solut
Accura1
tics of SiCH
esis of SiCHA
ucture ...........
ion of propy
wing adsorpti
of PSA proc
on of a PSA
sothermal P
ion model ....
ary condition
odel ..............
propane syst
ption data.....
olling transpo
rformance ....
tions ............
acy of mass a
IV
HA ................
A .................
....................
lene/propane
ion processe
cess .............
process .......
Pore Diffusi
....................
ns for a 5-ste
....................
tem ..............
....................
ort mechanis
....................
....................
and energy b
....................
....................
....................
e in SiCHA .
es for propyl
....................
....................
ion Model f
....................
ep PSA proc
....................
....................
....................
sm ...............
....................
....................
balances ......
....................
....................
....................
....................
lene/propane
....................
....................
for a Kineti
....................
cess .............
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
e separation .
....................
....................
ically Contr
....................
....................
....................
....................
....................
....................
....................
....................
....................
.... 19
.... 19
.... 21
.... 23
.... 29
.... 34
.... 36
rolled
.... 42
.... 48
.... 52
.... 55
.... 56
.... 58
.... 59
.... 60
.... 61
C
3.6
3.7
3.8
CHAPTER 4
4.1
4.1.1
4.1.2
4.1.3
4.2
4.3
4.4
4.5
4.6
4.6.1
4.6.2
4.6.3
4.6.4
Breakthroug
PSA results
Chapter con
Propy
Adsorption
Equilib1
Molecu2
Kinetic3
Kinetic and
PVSA Proc
Model Equa
Numerical S
PVSA Proc
Effect o1
Effect o2
Effect o3
Effect o4
gh results ....
s ...................
nclusion.......
ylene/Propan
Parameters
brium Param
ular dynamic
c Parameters
d Equilibrium
cess Model ...
ations ..........
Simulation ..
cess Perform
of Length to
of Pressuriza
of High Pres
of Rinse Tim
V
....................
....................
....................
ne Separation
for SiCHA .
meters ...........
c simulation
s ...................
m Selectivity
....................
....................
....................
mance ............
o Feed Veloc
ation Time ..
ssure Adsorp
me................
....................
....................
....................
n Using SiCH
....................
....................
n ...................
....................
y ...................
....................
....................
....................
....................
city Ratio ....
....................
ption Time ..
....................
....................
....................
....................
HA ..............
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
.... 62
.... 64
.... 80
.....81
.... 82
.... 83
.... 86
.... 90
.... 96
.... 98
.. 101
.. 107
.. 108
.. 113
.. 114
.. 114
.. 115
C
C
4.6.5
4.6.6
4.6.7
4.6.8
4.6.9
4.7
CHAPTER 5
using a Su
5.1
5.2
5.3
5.4
5.5
5.5.1
5.6
5.7
5.8
CHAPTER 6
Effect o5
Effect o6
Effect o7
Effect o8
Effect o9
Chapter con
Comp
urrogate-base
Introduction
Optimizatio
Assessment
Implementa
Optimizatio
ANFIS1
Comparison
Comparison
Chapter con
Concl
of Evacuatio
of Adsorptio
of Evacuatio
of Reflux ra
of Temperat
nclusion.......
paring SiCHA
ed SimOpt A
n ..................
on of PVSA
t Approach ..
ation of Simu
on Algorithm
S Model .......
n Based on E
n Based on T
nclusion.......
lusion and Fu
VI
on Time .......
on Pressure ..
on Pressure ..
atio ...............
ture ..............
....................
A and 4A Z
Approach .....
....................
Processes ...
....................
ulation Mod
m ..................
....................
Energy Cons
Total Annua
....................
uture work ..
....................
....................
....................
....................
....................
....................
Zeolite for P
....................
....................
....................
....................
del ................
....................
....................
sumption .....
alized Cost (T
....................
....................
....................
....................
....................
....................
....................
....................
ropylene/Pro
....................
....................
....................
....................
....................
....................
....................
....................
TAC) ..........
....................
....................
....................
....................
....................
....................
....................
....................
opane Separ
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
.. 115
.. 116
.. 116
.. 117
.. 117
.. 120
ration
...121
.. 122
.. 124
.. 128
.. 130
.. 133
.. 134
.. 137
.. 139
.. 149
...151
L
6.1
6.2
List of public
Conclusion
Future work
cations ........
...................
k ..................
....................
VII
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
....................
.. 151
.. 153
...167
VIII
Summary
The separation of light olefins such as ethylene/ethane and propylene/propane from
the off-gas of catalytic crackers is a key step in the petrochemical industry. The current
method for these separations involves cryogenics. The US DOE has identified
propylene/propane separation as the most energy-intensive single distillation process
practiced commercially (Jarvelin and Fair, 1993). Thus, low-energy alternatives for these
separations are highly desirable. Adsorption offers an attractive option due to its low
energy demands. Pressure/Vacuum Swing Adsorption (PVSA) is a well-established
technology for gas separation. Since commercial inception in 1950 (Ruthven et al., 1994),
it has progressed much in size, versatility, and complexity. It can handle multicomponent
separation and purification, and offers great flexibility in design and operation.
In this study, first a non-isothermal micropore diffusion model has been developed to
simulate kinetically controlled pressure swing adsorption (PSA) processes. In this model,
micropore diffusivity depends on adsorbate concentration in the solid phase according to
the chemical potential gradient as the driving force for diffusion. The model has been
validated with published experimental data for the kinetically controlled separation of
propylene/propane on 4A zeolite. Its performance has also been extensively compared
with that of a bi-LDF model for the same system. The results clearly show that a non-
isothermal micropore diffusion model with concentration-dependent diffusivity is
comprehensive and complete for kinetically selective systems. The conditions under
which the bi-LDF model predictions may significantly deviate from those of the pore
diffusion model have also been discussed.
IX
Second, separation of propylene/propane mixture with new 8-ring zeolite, pure silica
chabazite (SiCHA), has been studied in this work. Since the diffusion of propane
molecules in SiCHA is extremely slow, thus equilibrium information for propane has
been indirectly estimated using available uptake data at 80 °C and 600 torr. Moreover,
molecular simulation has been used to obtain equilibrium information of propylene and
propane and verify our estimation. The ideal kinetic selectivity of propylene/propane
mixture is ~28 at 80 °C, which increases with decreasing temperature. A 4-step,
kinetically controlled pressure swing adsorption process has been suggested for this
separation and studied in detail using the non-isothermal micropore diffusion model,
developed and verified earlier. In this model, Langmuir isotherm represents adsorption
equilibrium and micropore diffusivity depends on adsorbate concentration in the
micropores according to chemical potential gradient as the driving force for diffusion.
Finally this work compares 4A zeolite and a new 8-ring silica chabazite zeolite
(SiCHA) for separating these in a simple pressure vacuum swing adsorption (PVSA)
process. Our assessment is based on the simulation of a simple 4-step PVSA cycle with
heavy reflux using a non-isothermal isobaric micropore diffusion model with
concentration-dependent diffusivity developed by Khalighi et al. (2012). For both
adsorbents, surrogate neuro-fuzzy models are developed using this rigorous simulation
model and minimize energy consumptions and total annualized costs of the processes via
a genetic algorithm (GA). If one neglects capital cost and bases the comparison of the
two adsorbents on minimum energy consumption per tonne of propylene feed, then 4A
zeolite seems better than SiCHA. However, this superiority of 4A zeolite comes at the
cost of lower feed rate. Thus, if one bases the comparison on total annualized cost, then
X
this conclusion is surprisingly reversed, and SiCHA proves better than 4A zeolite. This
clearly suggests that total annualized cost is a more reliable basis for comparing two
adsorbents.
XI
Nomenclature
specific surface area of the pellet, cm-1
isotherm constant of DSL isotherm for site and component , kPa-1
pre-exponential constant of DSL isotherm for site and component , kPa-1
concentration in the bulk gas phase of component , mol/cm3
concentration in the macropore gas phase of component , mol/cm3
molar specific heat capacity of the gas mixture, J/mol-K
specific heat of the column wall, J/g-K
specific heat capacity of the adsorbent, J/g-K
total concentration in the bulk gas phase, mol/cm3
particle diameter, cm
axial dispersion coefficient, cm2/s
micropore diffusivity coefficient, cm2/s
temperature-dependent limiting (concentration independent) micropore
diffusivity for component i in a binary system, cm2/s
temperature-independent pre-exponential constant for component i, cm2/s
molecular diffusivity, cm2/s
macropore diffusivity, cm2/s
wall thickness, cm
'a
ijb i j
0ijb i j
ic i
pic i
pgc
pwc
psc
C
pd
LD
cD
0 0,c c iD D
ciD∞
mD
pD
e
XII
activation energy of diffusion for component , J/mol
total gas flow gone to the compressor or vacuum pump, L/s
reflux ratio for rinse step
film heat transfer coefficient, W/cm2-K
convection heat transfer coefficient between wall and surrounding, W/cm2-K
isosteric heat of adsorption for component , J/mol
diffusive flux, mol/cm2-s
gas thermal conductivity, W/g-K
extend film mass transfer coefficient, cm/s
wall conduction heat transfer coefficient, W/g-K
dimensionless Henry's law constant
inlet pressure of compressor or vacuum pump, Pa
outlet pressure of compressor or vacuum pump, Pa
imaginary partial pressure of component
equilibrium adsorbate concentration of component , mol/g
equilibrium adsorbate concentration, mol/ cm3
temperature-independent saturation capacity of adsorbate , mol/g
average adsorbed concentration of component per unit adsorbent particle
volume, mol/ cm3
average adsorbed concentration of component per unit crystal mass, mol/g
iE i
F
G
wh
0h
iHΔ i
iJ
gk
fk
wK
K
inP
outP
imip i
* ciq i
* pq
siq i
piq i
ciq i
XIII
adsorbed concentration of component , mol/g
micropore radius, cm
macropore radius, cm
universal gas constant, J/K-mol
column (inside) radius, cm
adsorbent particle radius, cm
constant ambient temperature, K
temperature of the gas phase, K
adsorbent (solid) temperature, K
wall temperature, K
velocity, cm/s
energy consumption for compressor or vacuum pump, kWh/tonne of propylene
mole fraction of component
Greek Letters
ratio of the internal surface area to the volume of the column wall, cm-1
ratio of the external surface area to the volume of the column wall, cm-1
ratio of the convection area to the conduction area
heat capacity ratio (=1.4)
compression efficiency
gas density, g/cm3
crystal adsorbent density, g/cm3
ciq i
cr
pr
gR
wR
pR
T∞
gT
sT
wT
v
W
iy i
wiα
woα
β
γ
η
gρ
cρ
XIV
solid density, g/cm3
density of the column wall, g/cm3
gas viscosity, g/cm-s
tortuosity factor
axial heat dispersion coefficient, W/cm-K
bed porosity
adsorbent particle porosity
sρ
wρ
gμ
τ
λ
ε
pε
XV
List of Figures
Figure 1.1: An illustration of the zeolite A structure (LTA). .............................................. 5
Figure 1.2: The zeolite mineral mordenite (MOR): SiO4 polyhedra are represented as
yellow tetrahedra; AlO4 polyhedra are aqua colored ones. .......................................... 6
Figure 1.3: The ZSM-5 zeolite (MFI). The framework is represented by tiles assembly
showing straight channels in the structure. ................................................................... 7
Figure 1.4: Change in equilibrium loading by pressure in PSA process and by
temperature in TSA process. ....................................................................................... 10
Figure 1.5: The basic two-bed pressure swing adsorption system. ................................... 12
Figure 1.6: The steps in the basic Skarstrom PSA cycle. ................................................. 13
Figure 2.1: Diffusivity ratio of propylene/propane for various adsorbents at different
temperatures. ............................................................................................................... 19
Figure 2.2: 29Si MAS NMR spectrum of calcined pure silica chabazite structure (Díaz-
Cabañas et al., 1998) ................................................................................................... 21
Figure 2.3: Left: T-atom diagram of a chabazite cage. The black bold part of the cage is
illustrated on the right. Right: one 8-membered-ring and one of the 6-membered rings
with the four proton positions are presented. “Symmetry-equivalent positions cause
the positions to be represented with two, three, or four protons. The sizes of the
proton spheres represent half the van der Waals radius. Protons attached to O(3)
(cyan) are not exposed to the eight-membered-ring window and are thus distinctive
from those attached to O(1) (purple), O(2) (pink), and O(4) (blue).” (Bordiga et al.,
2005) ........................................................................................................................... 22
XVI
Figure 2.4: SEM of SiCHA (Olson et al., 2004). .............................................................. 23
Figure 2.5: Potential energy, Ep, for propylene (PE) and propane (PA) vs. normal
distance of the mass center to the plane of the ring (ter Horst et al., 2002). ............... 25
Figure 2.6: Propylene molecule in the 8-memebered oxygen ring (ter Horst et al., 2002).
..................................................................................................................................... 25
Figure 2.7: Propane molecule in the 8-memebered oxygen ring (ter Horst et al., 2002). 26
Figure 2.8: The bond angle for the propane (PA) and propylene (PE) molecules vs. the
normal distance of the mass center to the ring plane (ter Horst et al., 2002). ............ 27
Figure 2.9: Propylene and propane uptake data in SiCHA at 353 K and 600 torr (Olson et
al., 2004). .................................................................................................................... 28
Figure 2.10: Equilibrium measurement of propylene in SiCHA. Langmuir isotherm is
fitted to the experimental data (Olson et al., 2004). .................................................... 29
Figure 2.11: Four different types of optimization strategies, (a) Simplified approach, (b)
Black-box approach, (c) Equation-oriented, (d) Simultaneous tailored. .................... 41
Figure 3.1: Schematic of the 5-step PSA process including pressure-time history. PR =
feed pressurization, HPA = high pressure adsorption, RI = rinse, BD = blowdown and
EV = evacuation. ......................................................................................................... 51
Figure 3.2: Experimental data for the adsorption equilibrium of propylene (a) and
propane (b) are well fitted by the dual-site Langmuir isotherm. ................................ 58
Figure 3.3: Schematic of equation domain. ...................................................................... 60
Figure 3.4: Experimental measurements and simulated breakthrough responses for
propylene and propane at 423 K and 250 kPa. The MSEs for model predictions are
XVII
1.05E-04 (C3H6) and 4.81E-05 (C3H8) for the pore model and 2.10E-04 (C3H6) and
3.45E-05 (C3H8) for the bi-LDF model. .................................................................... 63
Figure 3.5: Temperature profiles for the breakthrough experiments at the top (68 cm),
middle (43 cm) and bottom (18 cm) of the column. The distances are measured from
the feed end. The MSEs for model predictions are 0.059 (top), 0.156 (middle) and
0.207 (bottom) for the pore model and 0.250 (top), 0.829 (middle) and 0.717 (bottom)
for the bi-LDF model. ................................................................................................. 64
Figure 3.6: Experimentally measured pressure profiles and their linear or exponential fits
used in the simulation (value in blowdown and evacuation step is 6 s-1 and 0.15 s-1,
respectively). For experimental details, see run 4 in Table 3.3. ................................. 65
Figure 3.7: Prediction of the effect of nitrogen in the feed on the purity and recovery of
propylene compared with experimental results. For experimental conditions, see run
1-3 in Table 3.3. .......................................................................................................... 67
Figure 3.8: Prediction of the effect of feed temperature on the purity and recovery of
propylene compared with experimental results. For experimental conditions, see run
3-5 in Table 3.3. PN is perfect positive correlation and PP is perfect negative
correlation. .................................................................................................................. 68
Figure 3.9: Comparison of experimentally measured molar flow rates with model
predictions over a cycle after reaching cyclic steady state. The results are from two
different experimental runs, run 6 in (a) and run 4 in (b) and (c). For experimental
details, see Table 3.3. .................................................................................................. 71
XVIII
Figure 3.10: Temperatures measured at three different locations in the column over a
cycle after reaching cyclic steady state in run 4, See Table 3.3 for experimental
details. ......................................................................................................................... 71
Figure 3.11: Concentration profiles of propylene and propane inside the crystal at z/L=
0.1 at the end of the high pressure adsorption (step 2) and the end of the evacuation
(step 5) after reaching cyclic steady state in run 4 detailed in Table 3.3. ................... 72
Figure 3.12: Effect of propylene/propane diffusivity ratio on the purity and recovery
predicted by the pore and bi-LDF models. The propane diffusivity was gradually
increased while holding the propylene diffusivity constant. The experimental
conditions are same as in run 4 in Table 3.3. .............................................................. 73
Figure 3.13: Dimensionless adsorbate phase concentration of propylene and propane in
five step of PSA run 4 are shown in Figs. a-j. ............................................................ 78
Figure 4.1: Adsorption isotherms for propylene on SiCHA. Points represent the
experimental data by Olson (2004) and solid lines represent the Langmuir isotherm.
..................................................................................................................................... 84
Figure 4.2: Propylene and propane equilibrium isotherm in SiCHA at 80 ˚C obtained
from MC simulation are compared with experimental data and Langmuir model
estimates, respectively. The Langmuir model parameters were obtained indirectly
from the uptake data of Olson et al. (2004). ............................................................... 85
Figure 4.3: Illustration of CHA structure. The cages are indicted in blue and the windows
in green........................................................................................................................ 86
Figure 4.4: Density contours of C3H6 and C3H8 in CHA at 700 torr. The unit of density
scale............................................................................................................................. 90
XIX
Figure 4.5: Experimental and simulated uptake data for propylene in SiCHA at 30 ˚C and
600 Torr. ..................................................................................................................... 93
Figure 4.6: Experimental and simulated uptake data for propylene in SiCHA at 80 ˚C and
600 torr. ....................................................................................................................... 93
Figure 4.7: Experimental and simulated uptake data for propane in SiCHA at 80 ˚C and
600 Torr. ..................................................................................................................... 95
Figure 4.8: Effective kinetic selectivity of Propylene over propane in (a) SiCHA at 353 K
and 266 kPa and in (b) 4A at 353 K and 10 kPa. The selectivity at t = 0 is a small
nonzero value. ............................................................................................................. 98
Figure 4.9: Schematic diagram of the PSA cycle. 1) Pressurization 2) high-pressure
adsorption 3) rinse 4) countercurrent evacuation. ..................................................... 100
Figure 4.10: Recovery vs. purity plots show the effects of different parameters on the
performance of a PVSA process. The arrows indicate the increasing directions of
operating parameters. a) propylene b) propane and c) propane purity vs. propylene
purity for the feed composition of 50/50 propylene/propane. Each parameter increases
in the direction of arrow. Table 4 shows the range of the parameters. ..................... 110
Figure 4.11: Recovery vs. purity plots show the effects of different parameters on the
performance of a PVSA process. The arrows indicate the increasing directions of
operating parameters. a) propylene, b) propane, and c) propane purity vs. propylene
purity for the feed composition of 85/15 propylene/propane. Each parameter increases
in the direction of arrow. Table 4 shows the range of the parameters. ..................... 112
Figure 5.1: (a) 5-step PVSA process, (b) 4-step PVSA process with 4A zeolite and
SiCHA. ...................................................................................................................... 130
XX
Figure 5.2: Testing points vs. ANFIS results for 50/50 propylene/propane in SiCHA.
...................................................................................... Error! Bookmark not defined.
Figure 5.3: Number of iteration vs. difference of ANFIS and COMSOL results for
optimum parameters. ................................................................................................. 137
Figure 5.4: Optimization algorithm used in this work. ................................................... 139
Figure 5.5: Effect of bed length on the minimum energy for SiCHA and 4A for 50/50 and
85/15. ........................................................................................................................ 141
Figure 5.6: Effect of bed diameter on the minimum energy for SiCHA and 4A for 50/50
and 85/15. .................................................................................................................. 142
Figure 5.7: Scale up procedure for optimum results. ......... Error! Bookmark not defined.
XXI
List of Tables
Table 2.1: Bond lengths and angles for propene and propane molecules in gas phase (ter
Horst et al., 2002)........................................................................................................ 24
Table 2.2: Summary of PSA processes for the separation of propane/propylene mixtures.
..................................................................................................................................... 31
Table 3.1: Mass and heat transport parameters used in simulating the breakthrough
experiment with propylene/propane feed at 423 K, 250 kPa, and 7.5 cm/s. .............. 56
Table 3.2: Parameters of the Dual-site Langmuir isotherms for propylene and propane on
4A zeolite. ................................................................................................................... 57
Table 3.3: Operating conditions of the PSA experiments taken from Grande and
Rodrigues (2005). ....................................................................................................... 69
Table 4.1: Summary of Henry constant and diffusivity coefficient for propylene/propane
in available adsorbents at 80˚C. .................................................................................. 82
Table 4.2: Force field parameters for SiCHA, propylene, and propane. .......................... 88
Table 4.3: Equilibrium and diffusivity information obtained from the uptake of propylene
and propane in SiCHA at 600 torr. ............................................................................. 94
Table 4.4: The PVSA operating parameters, their ranges used in the parametric study, and
their comparison for the 50/50 and 85/15 propylene/propane feed mixtures. .......... 113
Table 4.5: The PVSA operating parameters of six points with desired product purities,
where points 1-3 are for the 50/50 and points 4-6 are for the 85/15 propylene/propane
feed mixture. ............................................................................................................. 118
XXII
Table 5.1: Best PVSA processes for 4A zeolite and SiCHA and two industrially relevant
feed compositions. .................................................................................................... 136
Table 5.2: Minimum-cost processes for the four adsorbent-feed combinations. ............ 147
Table 5.3: Comparison of simulation results and ANFIS model for scale up results. .... 149
XXIII
1
CHAPTER 1 INTRODUCTION
The separation of olefin/paraffin mixtures resulting from the thermal or catalytic
cracking of hydrocarbons is a crucial operation in the petrochemical industry. A
practically relevant example is the separation of propylene/propane mixtures, which is of
immense economic significance owing to the wide use of the separated propylene and
propane. A major application of propylene is its use as the monomer feedstock for
polypropylene elastomer, while applications of propane include recycling to the cracking
step or being used separately for various purposes, such as fuel for engines, oxy-gas
torches, barbecues, portable stoves and residential central heating. There are two main
sources of propane/propylene mixtures: (1) 50/50 propylene/propane mixture from steam
cracking of liquid feedstock and (2) 85/15 propylene/propane mixture from off-gases
produced by the fluid catalytic cracking (FCC) units in refineries. The temperature of
both streams is about 600-800 K and the pressure is 2-3 atm.
The conventional method for separating a propylene/propane mixture is cryogenic
distillation (Eldrige, 1993). However, as the relative volatility of the mixture is close to
unity (1.09-1.15), the process requires many (> 100) contacting stages and large energy
input for maintaining high reflux ratios (Ruthven and Reyes, 2007). Cryogenic
distillation consumes over 20 GJ of energy per tonne of propylene produced (Imtex,
2009). It uses non-renewable energy resources and emits significant greenhouse gases
(GHGs) and criteria air contaminants (CACs). The U.S. Department of Energy has
reported that propylene/propane separation is the most energy-intensive single distillation
pr
d
p
(K
m
se
1
pr
re
se
ad
se
rocess pract
evelop econ
otential for c
1- Absor
and El
2- Memb
3- Pressu
Of the ab
Knaebel et a
This stud
mixtures bec
eparate more
Select.1
A pressu
ropylene/pro
equirement f
electivity, c
dsorption k
eparation fac
AB
X
α =
ticed comme
nomical alter
commerciali
rption/strippi
ldridge, 199
brane separa
ure swing ad
bove three,
al., 2005), an
dy suggests
cause by us
e easily and
ivity
ure swing a
opane separa
for an econ
capacity and
kinetics or
ctor is define
A
B
A
B
XX
YY
ercially (Jar
rnatives for
ization are as
ing using aq
8).
ation (Stoitsa
dsorption (PS
the last (PSA
nd hence has
s adsorption
ing high se
in a more en
adsorption (
ation becaus
omic separa
d life. The
adsorption
ed as:
2
rvelin and F
this separat
s follows:
queous silve
as et al., 2005
SA) using ze
A) exhibits
the potentia
n-based proc
elective adso
nergy-efficie
PSA) proce
se of its exp
ation proces
selectivity
equilibrium
Fair, 1993).
tion. Some a
er nitrate so
5).
eolite molecu
high selecti
al to offer a l
cess to sepa
orbent prop
ent manner.
ess can be
ected low en
s is an adso
may depend
m (Ruthven
Therefore,
alternative t
lution (Brya
ular sieves.
ivity (separa
low energy o
arate propyl
pylene/propa
an attractiv
nergy deman
orbent with
d on a diff
n, 1984). T
it is desirab
echnologies
an, 2004; Sa
ation factor >
option.
lene/propane
ane mixtures
ve alternativ
nd. The prin
sufficiently
ference in e
The equilib
(1.1)
ble to
with
afarik
> 10)
e gas
s can
ve for
ncipal
high
either
brium
3
where Xi and Yi are the mole fraction of component i in adsorbent and fluid phase at
equilibrium, respectively. The search for a suitable adsorbent is the first step in the
development an adsorption separation process. Since, the separation factor usually varies
with temperature and sometimes with composition; the choice of proper conditions to
maximize the separation factor is important consideration in process design. The
separation factor for an ideal Langmuir system is independent of composition and equal
to the ratio of the Henry’s law constants of the two components (Eq. 1.2).
propyleneE
propane
K
Kη
=
(1.2)
The kinetic separations are usually possible with molecular sieve adsorbents such as
zeolites or carbon sieves (Ruthven, 1984). The kinetic selectivity is measured by the ratio
of micropore diffusivities for the relevant components. The definition of the separation
factor in kinetically controlled process is given by Eq.1.3. It is clear that kinetic
selectivity is time-dependent (Majumdar et al., 2011).
*
*0
*
*0
( )
( )
c c
c A AAB
c c
c B B
q t q
q c
q t q
q c
η
=
(1.3)
where qc and qc* are adsorbed phase concentration and equilibrium adsorbed phase
concentration in micropore and c0 gas concentration in the external fluid phase. Eq.1.3
can be reduced to Eq.1.4 by the following assumptions: (1) short contact times, (2)
uncoupled diffusion, (3) linear or Langmuir isotherm. Moreover, the ideal kinetic
selectivity can be calculated using Eq.1.4 and it only accounts for the loading in the
micropores and ignores the nonselective storage capacity of the macropores.
w
d
w
se
se
1
pr
fr
ar
by
m
si
o
o
,Id AB
K
Kη =
where Ki is
iffusivity. Th
,Ef ABη
=
where qp is ad
The equ
eparation of
electivity for
Zeolit.2
Acceptab
ractical sepa
rom a few Å
re such adso
y a crystal st
mean microp
ignificant di
lefin/paraffin
Zeolites
f SiO4 and A
( )( )
0
0
cA A
B c B
DK
K D
Henry’s la
he effective
0
0
( )
( )
p
A
p
B
q t
c
q t
c
dsorbed phas
ilibrium and
f specific ga
r separation
es
ble adsorptiv
aration proce
Å to a few ten
orbents. Amo
tructure. By
pore diamete
ifferences in
n separation
are porous c
AlO4 tetrahed
aw constant
kinetic selec
se concentra
d kinetic se
ases. Next,
of olefin/par
ve capacity
esses to only
ns of Å (Rut
ong these ad
contrast, the
er is control
n the adsorp
n. Therefore,
crystalline al
dra joined to
4
for compo
ctivity is:
ation per part
electivity rep
a few adso
raffin mixtur
y requiremen
y those micr
thven, 1984)
dsorbents, th
e others have
lled by the m
ptive proper
it is interest
luminosilica
ogether in di
onent i and
ticle volume
present the
orbents are i
res.
nt limits th
roporous ads
). Silica gel,
he micropore
e a distributi
manufacturi
rties of zeol
ting to consi
ated structure
ifferent regu
d Dc0 is lim
e.
potential o
introduced w
he choice o
sorbents wit
activated ca
e size of zeol
ion of micro
ing condition
lites which
ider differen
es. Their fra
ular arrangem
(1.4)
miting micro
(1.5)
of adsorbent
which have
f adsorbent
th pore diam
arbon and ze
lites is contr
opore size an
ns. This lea
can be use
nt type of zeo
amework con
ments with sh
opore
ts for
high
ts for
meters
olites
rolled
nd the
ads to
ed for
olites.
nsists
hared
ox
m
d
4
al
co
w
A
xygen atom
molecular dim
etermines th
As of Oc
0 naturally o
The atla
llocates a th
omposition
were discove
Al/Si ratio in
1- Low-s
≈ 1). Z
and w
labora
and ch
ms. This str
mensions in
he micropore
ctober 2011,
occurring zeo
s of zeolite
hree letter co
(Baerlocher
red from 19
their framew
silica or alum
Zeolites A (
were disco
atories. They
hannel struct
Figure
ructure form
nto which g
e structure, it
, 201 unique
olite framew
e structure t
ode to be us
r et al., 200
50 to 1970 a
works (Ribei
minium rich
Figure 1.1)
vered by
y represent
ture.
e 1.1: An illus
5
ms an open
guest molecu
t is uniform
e zeolite fram
works are kno
types publis
sed for a kno
07).The fam
and may be
iro et al., 19
h zeolites: ze
and X are th
R. M. Milto
a fortunate
stration of the
crystal lat
ules can pen
without any
meworks ha
own.
shed by the
own framew
mous and ind
classified in
84):
eolite A (LT
he most com
on at the
optimum in
e zeolite A str
ttice that co
netrate. Sin
y pore size di
ave been ide
e IZA struc
work topolog
dustrially im
nto three gro
TA) and X (F
mmon comm
union carb
n compositio
ructure (LTA
ontains pore
ce crystal l
istribution.
entified, and
ture commi
gy irrespecti
mportant ze
oups accordi
FAU) (ratio
mercial adsor
bide corpor
on, pore vol
).
es of
attice
d over
ission
ive of
olites
ing to
Si/Al
rbents
ration
lume,
2- Interm
5). Th
Si/Al
next c
a large
Figure 1.2
3- High
In con
hydro
is mor
Gilson
only w
the ea
mediate silic
he third com
ratio from 1
commercially
e pore mord
2: The zeolitetetra
silica zeolite
ntrast to the
philic surfac
re homogen
n, 2002). Th
weakly inter
arly 1970's,
a zeolites: z
mmercially
1.5 to 3.0 we
y successful
denite (Figure
mineral morahedra; AlO4
es: ZSM-5 (M
low and inte
ces within a
eous with an
hey more str
ract with wat
the request
6
zeolite Y (FA
important m
ere made by
l synthetic z
e 1.2) with r
denite (MOR4 polyhedra ar
MFI), Silica-
ermediate si
porous crys
n organophi
rongly adso
ter and othe
t and attract
AU), morden
molecular si
y D. W. Brec
eolite introd
ratio Si/Al ≈
R): SiO4 polyhre aqua colore
-CHA (ratio
ilica zeolites
stal, the surfa
ilic-hydroph
orb the less p
er polar mole
tion for mo
nite (MOR)
eve zeolites
ck (Ribeiro e
duced in the
5.
hedra are repred ones.
o Si/Al ≥ 10)
s, representin
face of the hi
obic selectiv
polar organi
ecules. In th
re siliceous
(ratio Si/Al
s type Y, w
et al., 1984)
early 1960's
resented as ye
).
ng heterogen
igh silica ze
vity (Guisne
ic molecules
he late 1960'
molecular
l = 2-
with a
). The
s was
ellow
neous
olites
et and
s and
s and
sieve
ar
1
p
eq
ad
1
1
compo
Devel
Figure
Adsorptiv
re explained
Adsor.3
The sepa
ossible sepa
quilibrium,
dsorption str
3X) show h
993; Ruthv
ositions wa
lopment Lab
1.3: The ZSM
ve separatio
d next.
rption mec
aration of g
aration mech
kinetic, and
rengths of co
high equilibr
en and Rey
as achieved
boratories. Fi
M-5 zeolite (Mshowing stra
on of a gaseo
chanisms
gaseous mixt
hanisms (R
d steric. Eq
omponent ga
rium selectiv
yes, 2007).
7
d by the s
igure 1.3 sho
MFI). The fraaight channels
ous mixture
tures by a P
ege and Ya
quilibrium s
ases. For exa
vity for prop
Propylene
synthesis at
ows the ZSM
amework is res in the structu
may take pl
PSA proces
ang, 2002;
separation o
ample, alum
pylene over
is adsorbed
t the Mobi
M-5 structure
epresented byture.
lace in diffe
ss is usually
Ruthven an
occurs due
mina-rich zeo
r propane (J
d much mo
il Research
e.
y tiles assemb
rent ways, w
y based on
nd Reyes, 2
to the diff
lites (e.g. 5A
Jarvelin and
ore strongly
and
ly
which
three
2007):
fering
A and
Fair,
than
pr
se
4A
2
sm
an
se
ex
cr
pr
th
ad
pr
R
b
ad
1
fo
d
en
(P
ropane due
eparation rel
A zeolite is
003; Padin e
maller size.
nd Rodrigue
eparation de
xample, AlP
rystalline str
ropylene to
he equilibriu
dsorption-ba
ropylene/pro
Rodrigues, 19
etter than eq
dsorbent for
Pressu.4
There ha
or separation
ecades. Th
nvironmenta
PSA) is not
to the elec
lies on the d
well known
et al., 2000)
In fact, the
es, 2005) am
pends on the
PO4-14 (Re
ructures, exc
adsorb due
um selective
ased separat
opane mixtu
999; Grande
quilibrium s
r a separation
ure swing
as been a phe
n and purifi
hese techno
al, pharmace
a new proce
ctrostatic for
differing ads
n for the sepa
), where pro
4A zeolite
mong the avai
e molecular
ege and Yan
cludes propa
to its linear
e processes
tion process
ure, it has
e et al., 2010
eparation ba
n, a proper p
adsorptio
enomenal gr
fication of m
ologies are
eutical and e
ess. Early PS
8
rces exerted
sorption rate
aration of pr
opylene diffu
has one of
ilable adsorb
sieving prop
ng, 2002), a
ane due to i
r shape. Am
are the easi
ses exploit
been show
0b) that kine
ased on 5A
pressure swin
on (PSA)
rowth in the
multicompon
being us
electronic ga
SA patents w
d by the ex
es of compon
ropylene/pro
uses much fa
the highest
bents in the l
perties of cry
an aluminop
its molecula
mong the abo
iest to oper
equilibrium
wn in publis
etic separatio
or 13X zeol
ng adsorptio
developmen
nent gas mix
sed in the
as industries
were issued
xchangeable
nent gases.
opane mixtu
faster than p
kinetic sele
literature so
ystalline mic
phosphate w
ar shape and
ove three se
rate; hence,
m selectivity
shed studies
on based on
lites. After f
n process is
nt of adsorp
xtures durin
e chemical,
s. Pressure s
to Finlayson
cations. Ki
For example
ure (Grande e
ropane due
ectivities (Gr
far. Lastly,
crostructures
with variatio
d size, but a
eparation opt
most large-
y. However
s (Da Silva
the 4A zeol
finding a sui
necessary.
ptive technol
ng the past
, petrochem
swing adsor
n and Sharp
inetic
e, the
et al.,
to its
rande
steric
s. For
ns in
llows
tions,
-scale
r, for
a and
lite is
itable
logies
three
mical,
rption
p (UK
9
365092), Hasche and Dargan (US 1794377), and Perley (US 1896916) from 1930 to
1933. Some of the major commercial PSA processes are:
1- Air fractionation (production of O2 and N2 enriched air).
2- Production of H2 and CO2 from steam-methane re-former (SMR) off-gas.
3- Production of CH4 and CO2 from landfill gas.
4- Gas desulfurization (CH4 from H2S).
All adsorption separation processes involve two principal steps: 1) Adsorption:
During this step, preferentially adsorbed species are picked up from the feed. 2)
Desorption: During this step, these species are removed from the adsorbent (Ruthven et
al., 1994).
A typical PSA system involves a cyclic process where a number of connected vessels
containing adsorbent material undergo successive pressurization and depressurization
steps in order to produce a continuous stream of purified product gas. A necessary
characteristic of a PSA process is that the preferentially adsorbed species are removed by
reducing the total pressure, rather than by raising the temperature (thermal swing
adsorption) or purging with a displacing agent. Figure 1.4 shows a schematic of the basic
difference between PSA and TSA operation. The main advantage of PSA, relative to
other types of adsorption processes such as thermal swing, is that the pressure can change
more rapidly than the temperature. Thus, PSA process offers a faster cycle and thereby
increases the throughout per unit of adsorbent bed volume.
1
d
P
or
ad
is
el
F
PSA c.5
The choi
ifferent cycl
SA cycle is
r kinetic). I
dsorbed spec
s recovered
lementary st
1- Pressu
2- High-
3- Depre
4- Desor
evacu
Figure 1.4: C
cycles
ce of a suita
les has been
classified a
It can furth
cie (the raffi
at high pu
teps, the mos
urization (wi
pressure fee
essurization o
rption at th
uation, purgi
Change in equtempe
able operatin
proposed to
according to
er be group
finate) or the
urity. Any
st common o
ith feed or ra
ed with raffin
or blowdown
he lower op
ing the bed
10
ilibrium loaderature in TSA
ng cycle is cr
o optimize d
the nature o
ped accordin
e more stron
PSA cycle
of which are
affinate prod
nate withdra
n (concurren
perating pre
d with the
ding by pressuA process.
ritical in a P
different aspe
of the adsorp
ng to wheth
ngly/rapidly
can be co
:
duct);
awal;
nt or counter
essure; this
raffinate p
ure in PSA pro
PSA process.
ects of the o
ption selecti
her the less
adsorbed sp
onsidered as
rcurrent to th
may be a
product or,
ocess and by
. A wide ran
overall proce
ivity (equilib
strongly/ra
pecie (the ex
s a sequenc
he feed);
accomplishe
in a kineti
nge of
ess. A
brium
apidly
xtract)
ce of
d by
ically
11
controlled process, by slow equilibration with consequent evolution of the slower-
diffusing sorbate;
5- Pressure equalization (which is used in many cycles, prior to the blowdown step,
to conserve energy and separative work);
6- Rinse (purging with the preferentially adsorbed species at high pressure,
following the adsorption step).
One cycle of a PSA process may contain some or all of the above steps. Each step
has a duration and the cycle time is the total duration of all steps. PSA systems are
typically operated at a cyclic steady state (CSS), which means that the temperature, mole
fraction, and solid concentration of bed profiles are identical at the beginning and at the
end of each cycle (Knaebel et al., 2005). The cyclic steady state can be reached after
running some cycles. One of the most popular modes of operation is the Skarstrom cycle.
In its basic form, it utilizes two packed adsorbent beds, as shown schematically in
Figure 1.5. The following four steps involve the cycle: pressurization, adsorption,
countercurrent blowdown and countercurrent purge. Both beds undergo these four
operations and the sequence, as shown in Figure 1.6, is phased in such a way that a
continuous flow of product is preserved. In step 1, bed 2 is pressurized to the higher
operating pressure, with feed from the feed end, while bed 1 is blowdown to the
atmospheric pressure in the opposite direction. In step 2, high-pressure feed flows
through bed 2. The more strongly adsorbed component is recollected in the bed and a gas
stream enriched in the less strongly adsorbed component leaves as effluent at a pressure
only slightly below that of the feed. A fraction of the effluent stream is withdrawn as
product and the rest is used to purge bed 1 at the low operating pressure. The direction of
th
st
he purge flo
tructure but w
w is also op
with the bed
Figure 1.
pposite to th
ds interchang
.5: The basic
12
hat of the fe
ged.
two-bed pres
eed flow. Ste
sure swing ad
eps 3 and 4
dsorption syst
follow the
tem.
same
1
av
in
Objec.6
This thes
vailable indu
nto two desir
Figu
ctives and
sis presents t
ustrial feed
rable produc
ure 1.6: The st
scopes
the study of
composition
cts, namely 9
13
teps in the ba
f a pressure s
ns, 50/50 an
90% propane
asic Skarstrom
swing adsorp
nd 85/15 pro
e and 99% p
m PSA cycle.
ption proces
opylene/prop
propylene.
ss to separate
pane gas mix
e two
xture,
14
This study used a new 8-ring pure silica zeolite, SiCHA, in the PSA process.
Propylene/propane has high diffusivity ratio in this adsorbent suggesting that there is a
good potential for kinetic-based separation of propylene/propane mixture using SiCHA.
To simulate a PSA process, the kinetic and equilibrium information of the gas
component is required. The kinetic and equilibrium measurements for propylene are
available in the literature. However, equilibrium information for propane is not available,
due to its very slow diffusion in the micropores of SiCHA. Therefore, in this work,
equilibrium data of propane was extracted from available limited uptake data. Later, the
estimates were confirmed by molecular simulation.
Furthermore, a proper mass transfer model is required to simulate a PSA process. For
SiCHA used as an adsorbent in this project, the appropriate model is kinetic selective
base. A proper representative model for simulation of this system is a pore diffusion
model. Moreover, olefin/paraffin separation is a highly non-isothermal process.
Therefore, this study proposed a non-isothermal pore diffusion model to simulate the
PSA process.
Finally, this work compared 4A zeolite and SiCHA for separating propylene/propane
mixture in pressure vacuum swing adsorption (PVSA) process. For both adsorbents, this
study developed surrogate Network-based Fuzzy Inference System (ANFIS) using this
rigorous simulation model and minimizes energy consumptions per tonne of propylene
and total annualized costs of the processes via a genetic algorithm (GA). Since the non-
isothermal pore diffusion model is a complex and has highly nonlinear equations, instead
of dealing with this complex model an approximate model was used by ANFIS.
1
pr
pr
n
pr
d
d
C
av
S
pr
4A
C
Struct.7
The outl
resented on
rocesses, pro
ew adsorben
rocess also
ependent dif
iffusion and
Chapter 3. In
vailable upt
iCHA is inv
rocess for S
A zeolite. F
Chapter 6.
ture of thi
ine of this r
the availab
oper adsorbe
nt, SiCHA. I
is discussed
ffusivity acc
d the verifica
n Chapter 4
take measur
vestigated fo
iCHA is com
Finally, conc
is thesis
report is as
ble literature
ents for this
In addition,
d. A non-is
cording to th
ation of this m
4, the equil
rements and
or propylene/
mpared with
clusions and
15
follows. In
e on the se
separation, a
a review of
sothermal p
he chemical p
model using
librium info
d molecular
/propane sep
h the optimu
d future plan
n Chapter 2,
eparation of
and the synt
available op
pore diffusio
potential gra
g available li
ormation of
simulation.
paration. In C
um results fo
ns for the pr
, a compreh
f propane/pro
thesis and ki
ptimization m
on model wi
adient as the
iterature data
propane is
Moreover,
Chapter 5, th
or the comm
resent study
hensive revie
opylene by
inetic behavi
methods for
ith concentr
driving forc
a are present
calculated
the potenti
he optimum
mercial adsor
are present
ew is
PSA
ior of
r PSA
ration
ce for
ted in
from
ial of
m PSA
rbent,
ted in
C
2
in
im
fi
th
re
la
m
m
an
m
pr
p
u
al
so
CHAPTE
Propy2.1
The sepa
ndustry. The
mportant exa
irst is the by
he second
efineries.(Gr
atter has 80-8
Cryogeni
mixtures to g
much energy,
nd propane.
mixtures are h
ropane in h
olypropylen
sed for vario
lso desirable
o far, no rese
ER 2 LI
ylene/prop
aration of ole
e separation
ample. The
yproduct from
is the off
rande and R
87% propyle
ic distillatio
et 99 mol%
, and tall col
Therefore,
highly desir
high purities
ne, a polyme
ous purposes
e to design p
earch has ad
ITERAT
pane separ
efin/paraffin
n of propan
industry us
m the steam
f-gases from
Rodrigues, 2
ene.
on is the c
propylene. I
lumns due to
economical
able. Furthe
s. For insta
er with exte
s such as fue
processes th
ddressed this
16
TURE RE
ration
n mixtures is
ne/propylene
es two main
m cracking o
m the fluid
005) The fo
current met
It is a difficu
o the small d
alternatives
rmore, most
ance, 99% p
ensive applic
el for engine
at produce b
point.
EVIEW
s a crucial op
e is probab
n feed mixtu
of liquid feed
d catalytic
ormer has 5
thod for se
ult separation
difference in
for the sepa
t application
pure propyle
cations. Sim
es, oxy-gas t
both product
peration in t
bly the mo
ures for this
dstocks such
cracking
50-60% prop
eparating pr
n requiring h
n the volatili
aration of pr
ns require bo
ene is the r
milarly, 90%
torches, barb
ts in high pu
the petrochem
ost common
s separation
h as naphtha
(FCC) unit
pylene, whil
ropylene/pro
high reflux r
ities of propy
ropylene/pro
oth propylen
raw materia
pure propa
becues. Thus
urities. How
mical
n and
. The
a, and
ts in
le the
opane
ratios,
ylene
opane
e and
al for
ane is
s, it is
wever,
2
ad
gu
is
C
is
an
se
ca
h
th
m
ch
ce
in
in
m
m
tw
i.
m
Differ2.2
Selecting
dsorption-ba
uide this sel
s defined as
Chapter 1]. It
s more appro
nd diffusion
electivity by
Indeed, t
apacity to a
igher the pu
he adsorben
molecules in
harge balanc
ell of the L
ncrease the d
nside the stru
molecules to
microporous
wo importan
e. bond leng
make up the w
rent adsorb
g the right a
ased separati
ection: equil
the ratio of
t is suitable
opriate for k
are uncoupl
y the square r
the adsorbe
dsorb the fa
urity that the
nts are fund
some zeoli
cing cations
TA structur
diffusion of
ucture obstru
increase. Th
materials th
nt aspects: (1
gths and bo
windows, (2
bents for p
adsorbent is
ion process.
librium and
f Henry's co
for equilibri
kinetic separa
led, kinetic s
root of the li
nts used in
ast diffusing
separation p
damental an
ites can be m
s. For instan
e) can be g
guest molec
ucts the 8-ri
he other suit
hat have been
) their wind
ond angles o
2) they are n
17
propylene
s the first a
Two metric
kinetic selec
onstants for
ium-based se
ations. In the
selectivity is
imiting diffu
n kinetic sep
g component
process can
nd practicall
modified by
nce, the mon
radually sub
cules. Remov
ing windows
able choices
n recently co
dow sizes are
of tetrahedra
naturally non
e/propane
and most cr
cs have been
ctivity (Ruth
a Langmuir
eparations. I
e Henry's la
s obtained by
usivity ratio
paration are
ts. The large
obtain. Thu
ly importan
y changing t
novalent Na
bstituted by
ving the Na+
s which caus
s of adsorben
onsidered. T
e determined
al atoms and
n-acidic. Thi
separatio
ritical step i
n proposed in
hven et al., 1
rian system
In contrast, k
aw limit whe
y multiplyin
[See Eq. (1.4
e still need
er diffusivity
us, the diffus
nt. The diff
the type of
a+ cations in
the divalen
+ cations fro
ses the diffu
nts are catio
These kinds o
d only by the
d joining ox
s factor is v
n
in developin
n the literatu
994). The fo
[see Eq. (1.
kinetic selec
ere the adsor
ng the equilib
4) in Chapte
to have a
y coefficien
sivity rates w
fusivity of
extra-frame
NaCaA (ps
nt Ca2+ catio
om their loca
usion of the
on-free crysta
of materials
e crystal stru
xygen atoms
ery importan
ng an
ure to
ormer
.2) in
ctivity
rption
brium
er 1].
large
nt, the
within
guest
ework
seudo
ons to
ations
guest
alline
have
ucture
s that
nt for
18
separation applications where chemical reactions should be avoided. Some important
examples of these cation-free materials are pure silicates and aluminophosphates. Those
of them whose diffusion of molecules is controlled through 8-ring window apertures are
attractive for the separation of small hydrocarbons. A proper choice of a window size for
the kinetic selectivity of a separation process can be enhanced by allowing some of the
molecules to enter the structure more rapidly than the others (Hedin et al., 2008).
Recently, several new cation-free 8-ring crystalline microporous materials have been
investigated such as ITQ-3, SiCHA, DD3R, AlPO-14, ZSM-58, etc. Figure 2.1 shows
diffusivity ratio of propylene/propane for different adsorbents changing by temperature
(Grande et al., 2010a; Grande and Rodrigues, 2004; Olson et al., 2004). As seen in this
figure, new adsorbent SiCHA, in particular, exhibits the highest diffusivity ratio (~ 410 )
for propylene over propane among the known adsorbents. SiCHA is a synthetic, pure
silica zeolite having the chabazite (CHA) structure. SiCHA has higher diffusivity ratio
among the other adsorbents that can be suggested to have a potential for kinetic
separation. . Kinetic separation using this new adsorbent could be an attractive option for
separating propylene/propane that nobody has investigated so far.
2
si
m
p
n
H
pr
p
an
m
co
Figur
Chara2.3
2.3
Zeolites
ilica crystall
mineral mela
olymorphs i
ot only adso
H2 and CH4. C
rinciple offe
ores, (2) dis
nd (3) a far
microporous
ompound in
e 2.1: Diffusi
acteristics
Synthe.1
exist in nat
line phases a
anophlogite
is a scientific
orption and s
Compared to
er (1) a larg
stinct adsorp
r superior th
SiO2 polymo
n the presen
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Dpr
opyl
ene/
Dpr
opan
e
ivity ratio of p
of SiCHA
sis of SiCHA
ture as high
are dense no
(MEP) (Ca
c challenge w
separation of
o zeolites of
ger void spac
ption proper
hermal stabi
orphs involv
nce of a su
0E+00
0E+01
0E+02
0E+03
0E+04
0E+05
0E+06
2.2
19
propylene/protemperature
A
A
h-alumina m
on-porous so
amblor et a
which may r
f organic mo
f the same s
ce owing to
rties, charact
ility (Díaz-C
ves a two-ste
uitable (norm
2.7
10
SiCHAITQ-3ZSM-58AlPO-144A13X
opane for varies.
materials (Si/
olids, with th
al., 1999). T
result in pot
olecules, but
structure, pu
o the absenc
terized by th
Cabañas et
ep process: t
mally organ
3.2
000/T (K)
ious adsorben
/Al < 5), w
he only exce
The synthesi
tential applic
also storage
ure silica pol
ce of counte
their extreme
al., 1998).
the synthesi
nic) structure
3.7
nts at differen
while natural
eption of the
is of pure
cations, inclu
e of gases su
lymorphs m
r cations in
e hydrophob
The synthes
s of a host–
e-directing
7
nt
pure
e rare
silica
uding
uch as
may in
their
bicity
sis of
–guest
agent
20
(SDA), and its calcination to remove the guest organics. Apparently, the use of SDAs
affords the required kinetic pathway and/or the additional stabilization energy that makes
the synthesis feasible. Diaz-Cabanas et al. (1998) did some modification on their previous
method and they successfully decreased the framework density from 17 SiO4/2 nm-3 to
15.4 SiO2 nm-3. The new pure silica polymorph isostructural zeolite chabazite have the
lowest ever reported framework density amongst these materials (14.6 T nm-3 for the type
material, structure code CHA).
The new pure silica chabazite sample was synthesized hydrothermally using N,N,N-
trimethyladamantammonium (TMAda+) in hydroxide form as the structure-directing
agent at near to neutral pH in the presence of fluoride. In a typical synthesis 13.00 g of
tetraethylorthosilicate were hydrolysed in 31.18 g of a 1.0 m TMAdaOH aqueous
solution and the mixture was stirred to allow the ethanol and water to evaporate to a final
H2O/SiO2 molar ratio of 3.0. Then, 1.33 g of HF (aq., 46.9%) was added and the mixture,
which was homogenised by hand, was transferred to Teflon lined stainless steel 60 ml
autoclaves. The autoclaves were heated at 150 °C whilst rotated at 60 rpm. After 40 h
crystallisation time (pH = 8.5) the solid product was collected, washed and dried, and
recognized as chabazite by powder X-ray diffraction (XRD). Its chemical analysis
indicates a composition close to [C13H24NF0.5]3[Si36O72(OH)1.5] [Anal. Found: C, 17.49; H,
2.98; N, 1.56; F, 1.06. The above composition requires: C, 16.78; H, 2.60; N, 1.51; F,
1.02%]. A charge imbalance between F2 and TMAda+ suggests the presence of
connectivity defects in this material. They have included 1.5 OH2 per uc in the above
idealised composition. Figure 2.2 shows the 29Si MAS NMR spectrum of calcined pure
si
S
2
ch
pr
m
ch
in
in
in
un
fo
pr
ilica CHA
i(OSi)3OH d
Figure
SiCHA2.4
As show
habazite cag
rotons in the
membered rin
habazite top
nterconnecte
n an ABC se
nterconnecte
nique tetrah
our possible
roton is atta
which has
defect group
2.2: 29Si MA
A structur
n in Figure
ge. The blac
e right part
ng and one
pology migh
ed by units o
equence that
ed by 8-mem
hedral site b
e acid site c
ached to. A
two bands
ps and the sec
AS NMR specCa
re
2.3, the left
ck bold part
of same figu
6-membere
ht be describ
of 4-member
leads to a fr
mbered-ring
ut four diffe
configuration
As shown in
21
at δ -101.4
cond to Si(O
ctrum of calcinabañas et al.,
part of this
of the fram
ure. Thus, th
ed ring with
bed as layer
red rings. Th
ramework w
windows. T
ferent oxyge
ns, dependin
n Figure 2.3
4 and -111.
OSi)4 species
ned pure silic1998)
figure is a s
mework is h
he right part
h the four p
rs of double
he double 6
with a regular
The chabazite
en atoms in
ng on whic
, the four o
.4. The firs
s.
ca chabazite s
schematic T
highlighted w
t of Figure 2
proton posit
e 6-member
6-membered-
r array of bar
e structure c
the asymme
ch of the ox
oxygen atom
st is assigne
structure (Día
-atom image
with the diff
2.3 shows o
tions drawn.
ed rings tha
-ring layers
rrel-shaped c
contains only
etric unit, g
xygen atom
ms belong t
ed to
az-
e of a
ferent
one 8-
. The
at are
stack
cages
y one
giving
ms the
o the
fo
m
ri
m
m
on
m
b
m
th
ca
pr
fo
Filforedew(b
ollowing rin
membered rin
ings and on
membered rin
membered rin
ne 4-memb
member of tw
etween the d
membered rin
he 8-membe
age and is n
roton linked
orm an H-bo
igure 2.3: Llustrated on tour proton poepresented wier Waals rad
window and ablue).” (Bordi
ng systems:
ngs that crea
ne 8-membe
ngs, where
ngs that brid
ered ring, o
wo 4-membe
double 6-me
ngs. A mino
ered-ring, th
not part of th
d to O(3) co
ond, makes th
eft: T-atom the right. Rigositions are pith two, three,
dius. Protons are thus distiiga et al., 200
O(1) is the
ates the doub
ered ring; O
O(2) is a pa
dge the doub
one 6-memb
ered rings an
embered ring
or relevant di
at delimits t
he open wind
uld interact
his site sligh
diagram of ght: one 8-mepresented. “S, or four protoattached to O
inctive from 05)
22
e oxygen fo
ble ring unit
O(2) and O(
art of the 8-
ble 6-memb
bered ring,
nd one 6-me
gs. O(4) belo
ifference is t
the CHA ca
dow. This ch
with an ato
htly different
a chabazite embered-ring
Symmetry-equons. The sizesO(3) (cyan) athose attache
orming the
s. O(1) belo
(3) have alt
-membered
bered rings.
and one 8
embered ring
ongs to one
that O(1), O
age, while O
haracteristic,
om of oxyge
t from the ot
cage. The bg and one of uivalent posits of the protoare not exposed to O(1) (p
bridge betw
ongs to both
ternating po
rings and n
Hence, O(2)
8-membered
g, and O(4)
4-membered
O(2), and O(4
O(3) is protr
, together w
en in a 6-me
thers (Bordig
black bold pathe 6-membetions cause t
on spheres repsed to the eigpurple), O(2)
ween the tw
two 4-memb
sitions in th
not part of th
) is a memb
ring. O(3)
forms the b
d ring and tw
4) are all pa
ruding insid
with the fact t
embered-ring
ga et al., 200
art of the caered rings withe positions present half thght-membered) (pink), and
wo 6-
bered
he 6-
he 4-
ber of
is a
bridge
wo 8-
arts of
de the
that a
g and
05).
age is th the to be
he van d-ring
d O(4)
ra
m
re
pr
2
ex
6
pr
pr
d
te
fo
The crys
ange of 2-10
method (Olso
eported 8.7 µ
2.4
Recently
ropylene an
004) calcula
xperimental
00 torr. Fro
ropane in Si
ropylene an
iffusivity ra
emperature d
or propylen
tals of SiCH
0 µm, as sh
on, 2004). T
µm and 1.45
F
Diffusio4.1
y, Olson et
nd propane u
ated the dif
data to Cra
om this figu
iCHA. Olson
d propane a
atio of prop
dependent, in
e and prop
HA are pseud
hown in Figu
The effectiv
g/cc by Ols
Figure 2.4: SE
on of propy
al. (2002; 2
uptake on S
ffusivity coe
ank’s solutio
ure, it is ob
n et al. (200
as 1.1E-9 and
pylene over
ncreasing to
ane are ~1
23
do-cubes (th
ure 2.4 usin
ve average
son et al. (20
EM of SiCHA
ylene/propan
2004) and H
SiCHA by g
efficients of
on. Figure 2
vious that p
02; 2004) ha
d 5.6E-13 cm
r propane
o 46000 at 3
0 and 73 k
hree dimensi
ng scanning
crystal size
004), respect
A (Olson et al
ne in SiCHA
Hedin et al.
gravimetry m
f propylene
2.5 shows th
propylene c
ave reported
m2/s at 353 K
at 353 K
03 K. The c
kJ/mol, whi
ional) and th
electron mi
and density
tively.
l., 2004).
A
. (2008) hav
method. Ols
and propan
he uptake da
an diffuse m
the diffusiv
K, respective
is 2000 wh
calculated ac
ich effects
heir size is i
icroscopy (S
y of SiCHA
ve measured
son et al. (2
ne by fitting
ata at 353 K
much faster
vity coefficie
ely. The rep
hich is stro
ctivation ene
the diffusio
in the
SEM)
A are
d the
2002;
g the
K and
than
ent of
ported
ongly
ergies
on of
24
propylene and propane through 8-ring of SiCHA. While SiCHA and DD3R have a very
similar structure, thereby explaining why propylene diffuses faster than propane through
SiCHA, Olson et al. (2004) have used the ter-Horst et al. (2002) study for DD3R. Ter-
Horst et al. (2002) to study the transport behavior of propylene and propane in DD3R
molecules. They have reported that the minimum cross section of propene is smaller than
propane through DD3R. Table 2-1 shows that the bond lengths and angles between the
carbon atoms for propylene and propane are different.
Table 2-1: Bond lengths and angles for propene and propane molecules in gas phase (ter Horst et al., 2002).
Hydrocarbon C═C (Å) C—C (Å) (CCC) (°)
Propylene 1.34 1.506 124.3
propane 1.532 112
Figure 2.5 shows the potential energy, Ep, as a function of the normal distance to the
ring plane. When the mass center of both molecules approaches the ring plane, the higher
Ep is required. When the mass center of both molecules is at a distance of about -2 Å,
propylene can jump through the 8-membered oxygen ring with the CH3 group head on as
shown in Figure 2.6. At the same distance only one CH3 group of propane crosses the 8-
membered oxygen ring while the CH2 group of propane still is positioned before the ring
as we can see in Figure 2.7. This shows that propylene fits through the ring at lower ring
energies than propane. It might be caused by larger van der Waals interactions between
the ring and the CH2 group of propane compared to the CH group of propylene. It may
also be the case that propylene has better molecule geometry to fit through the ring. A
la
at
Fm
arge deform
toms) is nee
igure 2.5: Pomass center to
Figure 2
ation of the
ded to fit thr
otential energythe plane of t
2.6: Propylen
e propane m
rough the 8-m
y, Ep, for propthe ring (ter H
ne molecule in
25
molecule (e.g
membered o
pylene (PE) aHorst et al., 20
n the 8-meme
g., an angle
oxygen ring.
and propane (002).
ebered oxygen
change bet
(PA) vs. norm
n ring (ter Ho
tween the ca
mal distance
orst et al., 200
arbon
of the
02).
ri
h
ad
m
pr
pr
Figure
Figure 2.
ing plane. Th
igher when
djustment o
membered ox
ropane mole
ropylene hav
2.7: Propane
.8 shows the
he bond ang
the PE ente
f its bond a
xygen ring. T
ecule to the
ve a perfect
e molecule in
e bond angle
gle of PE doe
ers the 8-m
angle from
This angle ch
ring plane. I
value to mo
26
the 8-memeb
of propylen
es not vary f
membered ox
113° to 125
hange occur
It looks that
ove through
bered oxygen
ne and propa
from its orig
xygen ring.
5° before th
rs at a norma
t the angle b
the ring wh
ring (ter Hor
ane vs. norm
ginal 124° an
However, P
he molecule
al distance o
between the
hile the angle
rst et al., 2002
mal distance t
nd is just sli
PA needs a
can pass th
of -2 to 0 Å o
carbons atom
e for propan
2).
to the
ightly
large
he 8-
of the
ms of
ne has
to
v
in
an
Fdi
m
ca
sh
m
m
pr
o be first ad
alence energ
ncreases slig
nd torsions w
igure 2.8: Thistance of the
Since, S
materials is p
age and in ri
Figure 2.
hows that pr
measured by
measurement
ropane beca
djusted to a
gy of PA in
ghtly (1.7 kJ
were observe
he bond anglee mass center
iCHA struc
principally d
ing states, as
.9 shows the
ropylene dif
y Olson et
t for propyl
ause it is dif
a value near
ncreases hig
J/mol). They
ed.
e for the propto the ring pl
cture is simi
determined b
s reported by
e uptake dat
ffuses faster
al. (2004).
lene. Howev
fficult to do
27
r the value
ghly (29.3 k
y reported th
pane (PA) anane (ter Horst
ilar to DD3
by the differe
y ter-Horst e
ta measured
r than propa
Figure 2.1
ver, they di
the equilibr
of the prop
kJ/mol) whil
hat no signifi
nd propylene t et al., 2002)
3R structure
ence in pote
t al. (2002).
by Olson e
ane. Figure 2
0 shows th
id not repo
rium experim
pylene angle
le the valen
ficant change
(PE) molecu).
e, the diffus
ential energy
et al. (2004).
2.9 shows t
he equilibriu
ort any equi
ment for pro
e. Therefore
nce energy o
e in bond len
ules vs. the n
sion rate in
y between th
. This figure
the isotherm
um experim
ilibrium dat
opane due to
e, the
of PE
ngths
normal
both
he in-
e also
m data
mental
ta for
o low
d
S
eq
ad
F20
iffusion rate
iCHA is no
quilibrium d
dsorbent.
igure 2.9: Pr004).
e through th
ot available
data to stud
opylene and
he SiCHA. S
in the liter
dy the sepa
propane upta
28
So far, the
rature. There
aration proc
ake data in S
equilibrium
efore, there
ess of prop
SiCHA at 353
information
is a gap to
pylene/propa
3 K and 600
n for propan
o obtain pro
ane with Si
torr (Olson
ne on
opane
iCHA
et al.,
Fex
2
b
V
im
li
M
so
in
pu
si
p
sm
b
pr
im
b
igure 2.10: Experimental d
Pressu2.5
separa
While a
ased separat
VSA (Vacuu
mportant. Th
iterature so f
Sikavitsa
Magnetically
orbent, whic
ncluding pre
urge with h
imulations s
acked beds,
maller partic
ed voids. T
ropylene/pro
mprove the
eds.
quilibrium mdata (Olson et
ure swin
ation
highly selec
tion, a well
m Swing A
he PSA pro
far are summ
as et al. (1
Stabilized F
ch selectively
essurization
high purity
suggested th
was enhan
cle sizes, wh
Their propyl
opane feed.
performance
measurement ot al., 2004).
g adsorp
ctive adsorb
-designed p
dsorption) o
ocess studies
marized in Ta
1995) inves
Fluidized B
y forms Π-c
with feed,
product an
hat the sepa
nced due to
hile retarded
lene recover
They showe
e as compar
29
of propylene i
ption pro
bent is a key
rocess, such
or TSA (Tem
s on propyle
able 2-2
stigated the
eds (MSFBs
omplexation
high-pressu
nd countercu
aration in M
faster transp
due to highe
ry was only
ed that MSF
red to the t
in SiCHA. La
ocesses f
y first step f
h as PSA (P
mperature Sw
ene/propane
e feasibility
s). They use
n bonds with
ure adsorptio
urrent blow
MSFBs, com
port resultin
er axial disp
y 17% with
FBs in PSA p
traditional P
angmuir isoth
for propy
for an econo
Pressure Sw
wing Adsorp
separation
of a PSA
ed Ag+-exch
h olefins. A 4
on, co-curre
wdown was
mpared to t
ng from hig
ersion coeff
h 99% purit
processes co
PSA cycles b
herm is fitted
ylene/prop
omic adsorp
wing Adsorp
ption), is eq
published i
A process u
hanged resin
4-step PSA
ent high-pre
proposed.
the convent
h flow rates
ficients and l
ty from a 4
ould signific
based on pa
to the
pane
ption-
tion),
qually
n the
using
n as a
cycle
essure
Their
tional
s and
larger
42/58
cantly
acked
30
Rege et al. (1998) proposed a 4-step PSA process using a monolayer of AgNO3
dispersed on silica gel substrate. This sorbent, monolayer AgNO3/SiO2, exhibited
superior selective adsorption of propane through Π-complexation. They assumed equal
time duration for all steps in their proposed 4-step PSA process. While they obtained
99.1% propylene from an equimolar feed of propylene/propane, propylene recovery was
quite low at 43.5%. In this study, they compared their result with kinetic separation using
4A zeolite. They found that equilibrium separation of propylene/propane on AgNO3
dispersed on SiO2 substrate was superior to kinetic separation on zeolite 4A. However,
the recovery of their system was low.
Among the commercial adsorbents, zeolite 4A exhibits the highest kinetic selectivity.
Silva et al. (1999) studied the separation of propylene/propane on zeolite 13X and zeolite
4A. While the former showed higher loading capacity and lower mass transfer resistance,
the latter’s kinetic selectivity for propylene was at least one order of magnitude higher.
From their study, macropore and micropore diffusion seemed to dominate mass transfer
in zeolite 13X and zeolite 4A, respectively. Later, Da Silva and Rodrigues (2001b)
proposed a 5-step PSA process using zeolite 4A and a 5-step VSA process using Zeolite
13X. Both processes produced 97-98% pure propylene, but at only 17-26% recovery.
Padin et al. (2000) studied 4-step PSA process using ALPO4-14 which has unique pore
structure and separates propylene from propane sterically. They obtained 99% pure
propylene from a 50/50 feed with 52% recovery. They also compared the separation
results of ALPO4-14 with AgNO3/SiO2 and 4A zeolite adsorbents. Purity and recovery of
propylene for AgNO3/SiO2 were respectively, 99.05% and 43.58% and for 4A zeolite
were 99.97% and 23.59%. Therefore, ALPO4-14 showed higher recovery for 99% pure
31
propylene compared to 4A zeolite and AgNO3/SiO2. Grande et al. (2005) studied a 5-step
PSA process using zeolite 4A extrudates. They used two mixtures with different
propylene/propane ratios (54/46 and 85/15) diluted with 50% nitrogen. They assumed a
bi-LDF approximation for mass transfer and included heat balance equations in their
simulation. The 85/15 feed at 408 K gave the best performance with a simulated
propylene purity of 99.43% and recovery of 84.3%. Recently, Grande et al. (2010b)
proposed a new dual-unit VPSA technology for producing 99% pure polymer-grade
propylene (PGP) with high recovery. They proposed two VPSA units in series using
zeolite 4A with varying crystal sizes. They designed the upstream 3-column VPSA unit
to produce PGP, while the downstream 2-column unit to produce pure propane.
Propylene from the downstream unit was recycled to the upstream unit to enhance
recovery. The proposed 2-stage VPSA process produced 99% PGP with 95.9% recovery
of propylene. The power consumption of their 2-stage VPSA process was at least 20%
higher than what would be required in the traditional cryogenic distillation.
Table 2-2: Summary of PSA processes for the separation of propane/propylene mixtures.
PSA Cycle
Steps Feed Composition & Operating Conditions
Adsorbent Performance for propylene
4-step (Sikavitsas
et al., 1995)
Pressurization with feed, high pressure adsorption, high pressure purge with high purity olefin, countercurrent blowdown
58% propylene 42% propane PH = 1 atm PL = 0.03 atm T = 298 K
Ag+ Exchanged Amberlyst 15 Resin
Purity>99% Recovery 17%
4-step
(Salil U. Pressurization with feed,
50% propylene 50% propane
AgNO3/SiO2 Purity 99.05% Recovery 43%
32
Rege et al.,
1998) high pressure adsorption, high pressure purge with high purity olefin, countercurrent blowdown
PH = 1 atm PL = 0.1 atm T = 298 K
4-step
(Padin et al.,
2000)
Pressurization with feed, high-pressure adsorption with feed gas, high-pressure cocurrent purge with part of the compressed C3H6-rich product, countercurrent blowdown
50% propylene 50% propane PH = 1 atm PL = 0.1 atm T = 393 K
AlPO4-14 Purity 99.38% Recovery 52%
5-step
(Da Silva
and
Rodrigues,
2001b)
Pressurization with feed, high-pressure adsorption, cocurrent depressurization to an intermediate pressure, cocurrent purge with propylene product, and countercurrent blowdown
25% propylene 25% propane 50% nitrogen PH = 5 atm PM = 0.5 atm PL = 0.1 atm T = 423 K
4A Zeolite Purity 97% Recovery 26%
5-step
(Da Silva
and
Rodrigues,
2001a)
Pressurization with feed, high-pressure adsorption, cocurrent depressurization to an intermediate pressure, cocurrent purge with propylene product, and countercurrent blowdown
25% propylene 25% propane 50% nitrogen PH = 5 atm PM = 0.5 atm PL = 0.1 atm T = 423 K
13X Zeolite Purity 98% Recovery 19%
4-step Pressurization (85% propylene and AgNO3/SiO2 Purity 99.24%
33
(Rege and
Yang, 2002) with feed, high pressure adsorption with feed and purge product, high pressure cocurrent purge with part of the C3H6-rich product obtained in blowdown step, countercurrent blowdown to a low pressure
15% propane) (50% propylene and 50% propane) PH = 7atm PL = 0.2 atm T = 393 K
AlPO4-14 AgNO3/SiO2
AlPO4-14
Recovery 75% Purity 99.18% Recovery 72% Purity 98.52% Recovery 71% Purity 98.65% Recovery 64%
4-step 5-step
(Grande et
al., 2005)
pressurization, adsorption, rinse, co-current depressurization to intermediate pressure (for 5-step)and counter-current blowdown
50% propylene 50% propane PH = 2.46 atm PL = 0.098 atm PM = 0.49 atm (for 5-step) T = 343 K
Ag/SBA-15 Purity 91% Recovery 97% Purity 99% Recovery 63%
5-step
(Grande and
Rodrigues,
2005)
pressurization, adsorption, rinse, co-current depressurization to intermediate pressure and counter-current blowdown
PH = 2.46 atm PL = 0.098 atm PM = 0.49 atm (for 5-step) T = 408 K 54:46 propylene/propane diluted in N2 PH = 4.98 atm PL = 0.098 atm PM = 0.49 atm (for 5-step) T = 433 K 85:15 propylene/propane diluted in N2
4A Zeolite Purity 99.43% Recovery 84.3% Purity 99.31% Recovery 90.2%
2 units and 6-step
(Grande et
al., 2010b)
Unit 1: Pressurization, feed, depressurization 1, rinse, depressurization 2, evacuation.
60% propylene 40% propane PH = 1.48 atm PM = 0.49 atm PL = 0.148 atm T = 423 K
4A Zeolite Purity 99.56% Recovery 96%
n
M
re
S
to
h
2
k
re
u
A
m
th
in
m
UPfdeae
The abov
early equim
Moreover, al
egard for pr
iCHA, for th
o develop an
igh-purity pr
Simu2.6
Adequate
inetically se
esistances: e
sually the m
Assumptions
micropore dif
he intrapartic
Kapoor a
n their LD
methane/carb
Unit 2: Pressurizatiofeed, depressurizatevacuation, pand pressure equalization
ve discussio
olar mixture
ll the studie
ropane purity
his separatio
n adsorption
roducts.
lation of P
e representat
elective pro
external film
most domin
of Linear
ffusivity (Sh
cle mass tran
and Yang (1
DF rate con
bon dioxide
on,
tion, purge
on suggests
es of propyle
es focused s
y. Furtherm
on. Since 90%
n process th
PSA proce
tion of the m
cess. Adsor
m, macropor
nant in kin
Driving For
hin and Knae
nsfer resistan
989) advoca
nstant expre
separation o
34
that most a
ene/propane
solely on ob
ore, no stud
% purity is a
hat separates
ess
mass transfe
rption-based
e, and micr
etically sele
rce (LDF) (
ebel, 1988) h
nce in kineti
ated adjustin
essions to
on a carbon
adsorption-ba
. Only two c
btaining 99
dy has evalu
also required
s a propylen
er phenomen
d processes
ropore. Of th
ective proce
(Kapoor and
have been u
ically contro
ng a cycle ti
match thei
molecular s
ased separat
considered a
mol% prop
uated the ne
d for propan
ne/propane m
na is essentia
involve thre
hese, the la
esses (Lami
d Yang, 198
used in the li
olled PSA sep
ime depende
ir experime
sieve (CMS)
tion studies
an 85/15 mix
pylene with
ew 8-ring ze
ne, there is a
mixture into
al for model
ee mass tra
st (micropor
ia et al., 2
89) and con
terature to m
parations.
ent paramete
ental results
). However,
used
xture.
little
eolite,
need
o two
ling a
ansfer
re) is
2008).
nstant
model
er, Ω,
s for
, they
35
found that their experimental estimates of Ω differed considerably from the predictions of
a priori correlations developed by Nakao and Suzuki (1983) and Raghavan et al. (1986).
These correlations were developed by forcing the LDF model solution to match the
solution from the pore diffusion model based on constant diffusivity under different
boundary conditions.
Shin and Knaebel (1988) assumed constant diffusivity in their pore diffusion model
for producing nitrogen via air separation on molecular sieve RS-10, a modified form of
4A zeolite. However, the effective constant diffusivity values that gave overall best fits of
their experimental PSA performance data over a wide range were different from the
actual diffusivity values measured from low-concentration uptake experiments.
To overcome the limitations arising from the assumptions of LDF and constant
micropore diffusivity, Farooq and Ruthven (1991) developed a pore diffusion model in
which micropore diffusivity varied with adsorbed concentration according to chemical
potential gradient as the driving force for diffusion. They applied their model with
concentration-dependent pore diffusivity to simulate high-purity nitrogen production
from air on a CMS. While the models of Kapoor and Yang (1989) and Shin and Knaebel
(1988) applied some degree of data fitting to improve the agreement between
experimental and simulation results, the experimental results were predicted reasonably
well by the approach of Farooq and Ruthven (1991) that involved no parameter fitting. It
merely used the parameters established from independent unary equilibrium and uptake
experiments. Farooq et al. (1993) further demonstrated the predictive ability of this
detailed micropore diffusion model by applying it to the air separation data of Shin and
K
fr
co
p
ac
ap
m
h
cy
co
as
k
2
tr
ac
pr
ad
op
d
pr
Knaebel (198
rom a differe
The abov
ontrolled PS
ossible whe
ccount in a
pplied an is
modified 4A
eats of adso
yclic operati
oncentration
s the drivin
inetically co
Optim2.7
A PVSA
rue performa
chieves afte
roperties suc
dsorption, a
perational p
imensions o
ressurization
88) using ind
ent laborator
ve discussion
SA process,
en the conc
pore diffus
sothermal m
zeolite, this
orption in the
ions were sm
n-dependenc
ng force for
ontrolled sep
mization of
A process is
ance, and th
er many cy
ch as equilib
and capacity
parameters
of the adsorp
n, high-pres
dependently
ry.
n suggests th
using indep
centration-de
sion model.
model to air
s was possib
ese adsorben
mall. A non-
e of microp
r diffusion i
aration with
f a PSA pr
inherently tr
hus design,
ycles of co
brium isothe
y, the perfo
of a PVSA
ption beds,
ssure adsor
36
measured u
hat a comple
pendently m
ependence o
Although F
r separation
ble only bec
nts, and the
-isothermal m
ore diffusivi
is necessary
h non-negligi
rocess
ransient and
is dictated b
ntinuous op
erms, kinetic
ormance at
A process.
while the l
rption, rins
unary equilib
ete and reliab
measured equ
of micropor
Farooq et al
for nitroge
cause oxyge
net changes
micropore d
ity based on
y to reliably
ible heat effe
d cyclic, and
by the cycli
peration. In
c/equilibrium
CSS depen
The former
latter includ
e, blowdow
brium and k
ble predictio
uilibrium an
re diffusivit
l. (1993; 19
en productio
en and nitrog
s in their loa
diffusion mo
n chemical p
y assess an
fect.
d has no true
ic steady st
n addition t
m selectivity,
nds on bot
r include th
de the opera
wn, and ev
inetic param
on of a kineti
d kinetic da
ty is taken
991) success
on on CMS
gen have m
adings durin
odel includin
potential gra
adsorbent
e steady stat
ate (CSS) th
to the adso
, isosteric he
th structural
he numbers
ational steps
vacuation),
meters
ically
ata, is
n into
sfully
S and
modest
ng the
ng the
adient
for a
te. Its
hat it
orbent
eat of
l and
s and
(e.g.
their
37
sequence, pressure levels, and durations. Thus, unlike a continuous plant, one cannot
design a PVSA process without fixing or optimizing its operational details. Since the true
test of an adsorbent lies in the CSS performance of its PVSA process, one cannot assess
an adsorbent (or compare adsorbents) without finding the best PVSA process for that
adsorbent. Thus, to compare 4A zeolite and SiCHA and identify the best, first the best
PVSA process for each must be separately developed/designed. This highlights the need
for a full-fledged synthesis and optimization (Agarwal et al., 2009; Haghpanah et al.,
2013b) of the relevant PVSA processes.
The full-fledged synthesis and optimization of a PVSA process is a major challenge
for several reasons. Adsorption is a highly nonlinear phenomenon. Its modeling,
simulation, and optimization in the context of a PVSA process involves repeated solution
of complex hyperbolic partial differential and algebraic equations (PDAEs). This is
extremely time-consuming and requires efficient numerical simulators (Haghpanah et al.,
2013a) and sophisticated optimization algorithms (Agarwal et al., 2010b). Many cycles of
operation must be simulated to arrive at the cyclic steady state (CSS) describing the
actual performance of a PVSA process at each point during optimization.
Several optimization studies (Biegler et al., 2005) using a variety of approaches for
several practical separation problems (e.g. Agarwal et al. (2010a; 2010b; 2003) for CO2
capture and concentration; Lewandowski et al. (1998) and Cruz et al. (2005; 2003) for air
separation; Nikolic et al. (2009) for hydrogen recovery) exist in the literature, but none on
propylene/propane separation. As shown in Figure 2.11, Biegler et al. (2005) classified
the various optimization approaches into four groups: 1) Simplified, 2) Black-box, 3)
Equation-oriented, and 4) Simultaneous tailored as illustrated in Figure 2.11. While the
38
simplified approach of Smith IV and Westerberg (1990) assumes a sequence of bed
operations and bed design parameters such as bed length and pressure levels to find the
minimum number of beds and a cyclic operating schedule, the other approaches address
much wider and varying scopes for the design, operation, and optimization.
The black-box approach is essentially simulation-based optimization (Subramanian
et al., 2000; Varma et al., 2008), in which a series of separate (black-box) simulations of
a PVSA process guides the optimization algorithm. The simulations may involve either a
fully rigorous model of the PVSA process, or an approximate or surrogate model derived
and updated with continuous help from the rigorous model. For instance, Kapoor and
Yang (1988) used polynomial expressions to fit the outputs (product purities and
recoveries) of a rigorous simulation model in terms of the inputs (feed pressure,
depressurization pressure, and throughput) for CO-H2 separation. Lewandowski et al.
(1998) developed an Artificial Neural Network (ANN) model for the separation of
nitrogen from air, and used a nonlinear programming approach to minimize the cost of
producing nitrogen. Other surrogate models such as ANFIS (Adaptive Network-based
Fuzzy Inference System) and Kriging (Agarwal et al., 2009; Biegler and Lang, 2012;
Caballero and Grossmann, 2008; Faruque Hasan et al., 2011; Lang et al., 2011) are also
attracting increasing attention. The black-box approaches have one major disadvantage.
The details of process dynamics are not fully integrated within or transparent to the
optimization algorithm. While this does reduce the complexity of the optimization model,
it compromises the nature and progress of the optimization algorithm. If a black-box
approach uses a surrogate model, then it has one more major disadvantage. The surrogate
model being less complex than the rigorous one, does speed up the optimization
39
algorithm, but its predictions of process performance, especially in extrapolated
situations, are often inaccurate.
In contrast to the black-box approach, the equation-oriented and simultaneously
tailored approaches embed the PDAEs for the PVSA process explicitly inside the
optimization formulation. Nilchan and Pantelides (1998) proposed complete
discretization (CD) involving a third order orthogonal collocation on finite elements for
the spatial domain and a first order backward finite difference method for the temporal
domain. They imposed simple periodic boundary conditions on process variable profiles
to ensure CSS, and used SQP (Sequential Quadratic Programming) for optimization.
Agarwal et al. (2010b) presented a novel superstructure for the optimal cycle
configuration of PVSA processes. They formulated an optimal control problem, and
employed complete discretization for its solution. They used a first-order finite volume
method for the spatial domain and orthogonal collocation on finite elements for the
temporal domain. They used IPOPT (Biegler, 2010) to solve the large nonlinear program.
Nikolic et al. (2009) reported an optimization framework for complex PSA processes
with multi-bed configurations and multi-layered adsorbents, and illustrated it for
hydrogen recovery from SMR (Steam Methane Reforming) off-gas (Nikolic et al., 2008).
They used orthogonal collocation for the spatial domain, and solved the PDAEs in
gPROMS (Barton and Pantelides, 1994). They employed a state transition network (STN)
approach for efficient simulation and optimization using the gOPT toll with reduced
sequential quadratic programming (rSQP) algorithm. STN approach has simpler and
linear implementation in multi-bed PSA systems. This approach can develop a
nondeterministic finite state machine which can optimize the inputs in a more ingenious
40
way. In this approach, states are represented by operation steps (such as pressurization,
adsorption, etc.), inputs are the duration of each step, operating parameters and time
elapsed in the process. The boundary conditions and gas valve conditions are existed in
each state. They claimed that their developed network covers all of the most important
states or configuration in a PSA process.
Jiang et al. (2003) proposed the simultaneous tailored approach for PVSA process
optimization. Instead of solving the PDAEs to the full CSS condition at each iteration as
in the black-box approach, they imposed just the CSS condition as a constraint in the
optimization problem. At each iteration, they solve PDAEs in an inner loop for exactly
one cycle to obtain the values of the constraints and objective function. In other words,
the algorithm attains CSS only when it achieves the optimal solution. Initially, they used
a modified finite volume (van Leer) method with smooth flux delimiters to decrease the
oscillations for steep fronts. Then, they employed the DAE solver DASPK 3.0 to solve
and integrate the bed equations. Finally, they used reduced-space successive quadratic
programming (rSQP) for optimization.
Most studies on propylene/propane separation have not considered producing high
purity propylene and propane simultaneously with low energy consumption. In other
words, significant room exists for improving and optimizing adsorption-based processes
for this separation.
41
Figure 2.11: Four different types of optimization strategies, (a) Simplified approach, (b) Black-box approach, (c) Equation-oriented, (d) Simultaneous tailored.
Tuning Parameters
Simple model optimization
CSS + Bed Models
(a)
Optimization + CSS
Bed model
(d)
Optimization
CSS
Bed Model
(b)
Optimization
+
Bed model
+
CSS
(c)
C
m
fo
co
d
L
on
as
pr
th
m
d
k
re
3
se
al
CHAPTE
Kineti
A non-is
micropore dif
or diffusion,
ontrolled sep
evelopment
Langmuir iso
n adsorbate
s the driving
ropylene/pro
he model. M
model for th
iffusion m
inetically se
esistances in
Pore d.1
A non-is
eparation pr
long the mic
ER 3 No
ically Con
sothermal m
ffusivity, wh
, is necessar
paration wit
of such a
otherm repre
concentratio
g force for
opane on 4A
Moreover, thi
e same syst
odel with
elective syste
n a kinetically
diffusion m
othermal po
rocess. It all
cropore radiu
on-isothe
ntrolled
icropore dif
hich is based
ry. This mo
th non-negli
model is d
esents adsorp
on in the soli
diffusion. T
A zeolite me
is study com
tem. The res
concentrati
ems. The im
y controlled
model
re diffusion
lows for the
us and make
42
ermal Po
PSA Pro
ffusion mode
d on chemic
odel reliably
gible heat e
detailed in t
ption equilib
id phase acc
This study u
easured by G
mpares the p
sults clearly
ion-depende
mportance of
separation i
model is de
e concentrati
s the follow
ore Diff
ocess
el with the c
cal potential
y assesses an
effect was hi
this chapter
brium and m
cording to th
uses experim
Grande and R
erformance
y show that
ent diffusiv
f macropore
is also invest
eveloped for
ion depende
ing assumpt
fusion M
concentratio
gradient as
n adsorbent
ighlighted in
. In this m
micropore dif
he chemical p
mental data
Rodrigues (2
of our mod
a non-isoth
vity is com
and solid-fl
tigated.
r a kinetically
ence of adso
tions.
Model fo
on-dependen
the driving
for a kineti
n Chapter 2
model, a dua
ffusivity dep
potential gra
for separatio
2005) to val
el with a bi-
hermal micro
mprehensive
fluid heat tra
y controlled
orbate diffus
or a
nce of
force
ically
. The
al-site
pends
adient
on of
lidate
-LDF
opore
e for
ansfer
d PSA
sivity
43
1- The ideal gas law applies.
2- The system is isobaric.
3- Axially dispersed plug flow model describes the flow pattern.
4- The adsorbent consists of uniform microporous crystals.
5- Chemical potential gradient is the driving force for diffusion along the micropore
radius.
6- The macropore gas is in equilibrium with the bulk gas in the bed voids.
7- Temperature gradients along the radii of the column and microparticle are
negligible. As Farooq and Ruthven (1990) conducted breakthrough experiments
in stainless steel columns with and without internal Teflon lining and confirmed
that the major heat transfer resistance in the radial direction was at the inner side
of column wall. The radial temperature profiles were measured. Although radial
temperature gradient existed in the column, but the inside wall film resistance to
heat transfer was more important and hence a simple one-dimensional heat
transfer model with a lumped heat transfer coefficient confined at the wall was
sufficient to capture the experimentally measured temperature breakthrough
behavior.
8- A finite heat transfer rate is introduced between the bulk gas and adsorbent
particles.
9- Lumped coefficients account for the heat transfer between the bed and column
wall and that between the column wall and external surroundings.
44
Based on the above, the equations describing the PSA process are as follows. The
signs of terms with ( ± ) depend on the flow direction. ( + ) sign applies for flow from
0z = to L and ( − ) applies for flow from z L= to 0 .
Mass balance for component i :
( ) ( ) 1 pii i iL p
qc y c vD C
t z z z t
ε ρε
∂∂ ∂ ∂∂ − − ± = − ∂ ∂ ∂ ∂ ∂ (3.1)
Overall mass balance:
( ) ( ) 1 pip
i
qC vC
t z t
ε ρε
∂∂ ∂ −± = −∂ ∂ ∂
(3.2)
(1 )p pi p pi p c ciq c qρ ε ε ρ= + − (3.3)
where v is the interstitial velocity, LD is the axial dispersion coefficient, p is the gas
pressure, T is the gas temperature, iy is the mole fraction of component i in the bulk gas
phase, ii
pyc RT= is its concentration in the bulk gas phase, pic is the concentration in the
macropore gas phase, piq is the average adsorbed concentration of component i per unit
adsorbent particle mass calculated with Eq. 3.3 (Qinglin et al., 2004); and ciq is the
average adsorbed concentration of component i per unit crystal mass, ε is the bed
porosity, and pε is the adsorbent particle porosity. Since the bulk gas is assumed to be in
equilibrium with the macropore gas, pi ic c= is set. Note that v is computed from Eq. 3.2.
The boundary conditions for Eqs. 3.1 and 3.2 depend on the PSA cycle and vary with
each step of the cycle. Therefore, they are discussed later for a 5-step PSA process used
by Grande and Rodrigues (2005).
In this study, LD was calculated from following equation (Wakao and Funazkri, 1978):
45
( )20 0.5m
L
DD ScRe
ε= +
(3.4)
Crystal balance for component i :
2
2
1( )ci
i
qr J
t r r
∂ ∂=∂ ∂
(3.5)
where, r is the radial distance along the crystal, ciq is the adsorbed concentration of
component i at r, and iJ is the diffusive flux. Using chemical potential gradient as the
driving force for diffusion and defining an imaginary partial pressure of component i ,
imip , which is in equilibrium with the adsorbed concentration in the micropore, ciq , the
following equation is obtained (Hu and Do, 1992):
0
ln
ln
imi ci
i c ici
p qJ D
q r
∂ ∂= −∂ ∂ 0
imci i
c i imi
q pD
p r
∂= −∂
(3.6)
In the above, 0c iD is the temperature-dependent limiting micropore diffusivity at
zero adsorbate concentration. The imaginary gas phase pressure, imip , can be calculated
from an appropriate isotherm model. After substituting Eq. 3.6 in Eq. 3.5, the micro-
particle balance becomes:
202
1( )
imci ci i
c i imi
q q pr D
t r r p r
∂ ∂∂=∂ ∂ ∂
(3.7)
with boundary conditions:
0
0im
r
p
r =
∂ =∂
(3.8a)
c
im
r rp p
==
(3.8b)
46
where, cr is the micropore radius. The temperature-dependence of micropore diffusivity
follows the Eyring-type form:
/
0i gE R T
c i ciD D e−∞= (3.9)
where, iE is the activation energy of diffusion and ciD∞ is the temperature-independent
pre-exponential constant.
The average adsorbate accumulation in particles is equal to its flux into the
microparticles:
03(1 )
c
impi pi c i ci i
p p p c imc i r r
q c D q p
t t r p rρ ε ε ρ
=
∂ ∂ ∂= + −∂ ∂ ∂
(3.10)
Gas phase energy balance:
2
2
g g gpg T pg T
T T Tc C c vC
t z z
λε
∂ ∂ ∂= − +
∂ ∂ ∂
( ) ( ) ( ) ( )1 1 2pi wpg g p f g s g w
i w
q hc T a h T T T T
t R
ε ερ
ε ε ε∂− −
′− − − −∂ (3.11)
where pgc is the molar specific heat capacity of the gas mixture, 'a is the specific surface
area of the pellet that is area to volume ratio, fh is the film heat transfer coefficient
between the gas and the solid phase. gR is the universal gas constant; gT is the
temperature of the gas phase, sT is the adsorbent (solid) temperature, wT is the wall
temperature, wh is the film heat transfer coefficient between the adsorption bed and the
column wall, wR is the column (inside) radius, and λ is the axial heat dispersion
calculated from the correlation by Wakao (1978).
( )7.0 0.5gk PrReλ = + (3.12)
47
fh is calculated from Nusselt number (Wakao and Funazkri, 1978):
0 6 1 32 0 1 1f p . /
g
h dNu . . Re Pr
k= = + (3.13)
Prandtl number;
g pgw
g
cPr
k
μ= (3.14)
Solid phase energy balance:
spgi p pi s ps
i
Tc q c
tρ ρ ∂ + = ∂
( )( )(1 )pi ci
pgi s p i p c f g si i
q qc T H a h T T
t tρ ε ρ
∂ ∂ ′− + −Δ − + −∂ ∂ (3.15)
where – iHΔ is the isosteric heat of adsorption for component i . Finally, the wall heat
balance is given by:
2
02( ) ( )w w
w pw w wi w g w wo w
T Tc K h T T h T T
t zρ α α ∞
∂ ∂= + − − −∂ ∂
(3.16a)
2
(2 )
wwi
w
R
e R eα =
+ (3.16b)
2( )
(2 )w
wow
R e
e R eα +=
+ (3.16c)
where pwc and wρ are the specific heat and density of the column wall, respectively. wiα
is the ratio of the internal surface area to the volume of the column wall, e is the wall
thickness, woα is the ratio of the external surface area to the volume of the column wall,
0h is the convection heat transfer coefficient between wall and surrounding, wK is the
wall conduction heat transfer coefficient, and T∞ is the constant ambient temperature.
T
co
b
h
th
R
co
to
pr
co
ob
ob
pr
re
h
co
ar
h
d
fr
The wall hea
olumn wall,
etween the c
eat transfer
he film resi
Rodrigues, 2
olumn wall,
o include t
ropylene/pro
oefficient f
bservations
bservations.
ropose of int
esistance to
igher value
ompensating
rgues that E
eat transfer.
3.1
As menti
epend on the
rom the exp
at balance gi
, heat excha
column wall
has been ne
istances on
2005) have
but have us
the metal
opane system
fitted by G
best was 30
If the axial
troducing th
radial heat t
of the ove
g for the neg
q. 3.16a is a
Bounda.1
ioned earlier
e type of PS
perimental s
iven by Eq.
ange between
l and the am
eglected, wh
its two si
neglected th
ed an overal
wall resista
m studied la
Grande and
0-40% highe
l conduction
he overall he
transfer, then
erall heat tr
glected heat
a more appro
ary conditio
r, the bound
SA cycle and
tudy of Gra
48
3.16a accou
n the adsorb
mbient air. T
hich is a reas
ides. Some
he contribut
ll heat transf
ance to rad
ater, This w
Rodrigues
er than the h
n along the c
at transfer co
n its value sh
ransfer coef
loss due to
opriate descr
ons for a 5-s
dary conditio
d its steps. S
ande and Ro
unts for axia
bent bed an
The resistanc
sonable app
studies in
tion of the
fer coefficien
dial heat t
work finds th
(2005) to
0h that this s
column wall
oefficient is
hould be low
fficient is a
axial condu
ription for th
step PSA pr
ons for Eqs.
ince, this stu
odrigues (20
al heat cond
nd the colum
ce of the me
proximation
the literatu
axial condu
nt in place o
transfer. Ho
that the ove
match the
study fits to
l is indeed n
to account f
wer than that
a clear indi
uction. This
he role of th
rocess
. 3.1, 3.2, 3.
udy extensiv
005) on a 5
duction alon
mn wall, and
etal wall to r
in comparis
ure (Grande
uction along
of 0h , presum
owever, for
rall heat tra
eir experim
o match the
negligible an
for the metal
t of 0h . Thu
ication that
study, there
he column w
.11 and 3.16
vely uses the
5-step PSA
ng the
d that
radial
on to
e and
g the
mably
r the
ansfer
mental
same
nd the
l wall
s, the
it is
efore,
wall in
6 will
e data
cycle
49
shown in Figure 3.1 , this work presents here the boundary conditions applicable for that
process only. As shown in Figure 3.1, their PSA cycle consists of pressurization, feed,
rinse with pure propylene, blowdown to intermediate pressure, and counter-current
evacuation for bed regeneration, where propylene product is withdrawn. For this
particular PSA process, the following boundary conditions apply (Wehner and Wilhelm,
1956):
Pressurization, feed, rinse steps:
0 0 0
0
( )il i iz z z
z
yD v y y
z−= = =
=
∂ = − −∂
(3.17a)
0i
z L
y
z =
∂ =∂
(3.17b)
0 0 0( )g
pg g gz z z
TCc v T T
z
λε −= = =
∂= − −
∂ (3.18a)
0g
z L
T
z =
∂=
∂ (3.18b)
0w feedz
T T=
= (3.19a)
0 ( )ww w
z L
TK h T T
zβ ∞
=
∂− = −
∂ (3.19b)
2( )
( 2 )w
w
e R
e e Rβ +
=+
(3.19c)
where β is the ratio of the convection area to the conduction area at the column end and
we have assumed the convection area as the total cross section area of the column end.
Blowdown and evacuation steps:
0
0i i
z z L
y y
z z= =
∂ ∂= =∂ ∂
(3.20ab)
50
0
0g g
z z L
T T
z z= =
∂ ∂= =
∂ ∂
(3.21ab)
00
( )w ww w w
z z L
T TK K h T T
z zβ ∞
= =
∂ ∂= − = −
∂ ∂ (3.22ab)
Boundary conditions for velocity in steps 1, 2, 3, 4, and 5 respectively are:
0z L
v=
= (3.23a)
00zv v
==
(3.23b)
00zv Gv
== (3.23c)
0
0z
v=
= (3.23d)
0z L
v=
= (3.23e)
It is assumed that bed pressure remains constant during the adsorption and rinse steps
and linearly changes during pressurization. The following exponential form is used to
compute the pressure profiles during the blowdown and evacuation steps (Farooq and
Ruthven, 1991):
( ) ( )exp( )II I IIp t p p p at= + − − (3.24)
where Ip and IIp are the initial and final pressures in the blowdown and evacuation
steps, and a is computed by fitting the above equation to the experimental pressure
profiles of blowdown and evacuation steps.
In Eq. 3.2:
ii
c C f ( t )= = for pressurization, evacuation and blowdown (3.25a)
f ( t )≠ for high pressure adsorption and rinse (3.25b)
51
Assuming that appropriate isotherm models and adsorption and kinetic data are
available for the adsorbed species, this completes our non-isothermal pore diffusion
model for a kinetically selective PSA process. Since this work will compare performance
with that of a bi-LDF model, this study briefly describes the latter for the sake of
completeness and highlight its differences.
Figure 3.1: Schematic of the 5-step PSA process including pressure-time history. PR = feed pressurization, HPA = high pressure adsorption, RI = rinse, BD = blowdown and EV =
evacuation.
PR HP RI BD EV
Step 1 Step 2 Step 3 Step 4 Step 5
PR HP RI BD EV
PH
PL
Time
3
pr
in
m
u
et
co
K
w
re
co
ex
Bi-LD.2
Grande
ropylene/pro
nto account
micropore res
sing linear m
Since ma
t al., 2008),
ombines the
Knudsen diffu
pic
kt
∂=
∂
pi pk ε=
5
RBi
ε=
where, Bi is
esistances, R
orrected for
xternal film,
fk
ShD
=
DF model
and Rodr
opane separa
the bidisper
sistances, th
models.
acropore dif
Grande and
e effects of
fusions in the
( )pi i pik c c−
2
15 pi ip
p i
D Bi
R Bi +
p f
p pi
R k
Dε
the Biot nu
pR is the adso
tortuosity f
, is calculate
2 0 1f p
m
d.
D= +
rigues (200
ation using 4
rsity of a zeo
ey approxim
ffusion inclu
Rodrigues (
f transport
e macropore
1+
umber repres
orbent partic
for compone
ed from Sher
0 6 1 31 . /. Re Sc
52
05) propos
4A zeolite in
olite adsorbe
mated the dif
udes both Kn
(2005) used
across the
s as follows
sents the rati
cle radius, D
ent i , and fk
rwood numb
ed a bi-L
n a 5-step P
ent by distin
ffusive proce
nudsen and
an effective
external flu
:
io of interna
piD is the eff
f is mass tr
ber (Wakao a
LDF mode
SA process.
nctly treatin
esses in thes
molecular d
e LDF rate c
uid film and
al macropore
ffective macr
ransfer coeff
and Funazkri
el for stud
While they
ng macropore
se two resista
diffusions (L
constant ( pik
d molecular
(3.26)
(3.27a)
(3.27b)
e to external
ropore diffus
fficient acros
i, 1978):
(3.28)
dying
y took
e and
ances
Lamia
) that
r and
l film
sivity
ss the
53
where mD is the molecular diffusivity, which can be calculated by using the Chapman-
Enskog equation (Bird et al., 1960).
Reynolds number;
g p
g
vdRe
ρμ
= (3.29)
where pd is particle diameter, gρ is gas density, gμ is gas viscosity and v is velocity.
Schmidt number;
g
m g
ScD
μρ
= (3.30)
where, Bi is the Biot number represents the ratio of internal macropore to external film
resistances, pR is the adsorbent particle radius, piD is the effective macropore diffusivity
corrected for tortuosity for component i , and fk is mass transfer coefficient across the
external film.
Knudsen diffusivity (Ruthven, 1984) is calculated from following:
9700k p
TD r M= (cm2/s) (3.31)
where pr is macropore radius, which we take as 1E-04 cm in this work.
Macropore diffusivity equation:
1 1 1
p k mD D Dτ
= +
(3.32)
where pD is the macropore diffusivity that combines the contributions from molecular
and Knudsen diffusivity and τ is the tortuosity factor.
54
For the micropores, they used the Darken equation to describe the concentration
dependence of micropore diffusivity:
( )*2
15ci cici ci
c
q Dq q
t r
∂ = −∂
(3.33a)
0
ln
lni
ci c ici
pD D
q
∂=∂
(3.33b)
In contrast to the above, our pore diffusion model captures the strong influence of the
concentration profiles on the diffusion in the microparticle (Do, 1998; Farooq and
Ruthven, 1991). For dual-site Langmuir isotherm, i
ci
ln p
lnq
∂∂
in Eq. 3.33b is given by Eq.
3.34.
Micropore concentration-dependent expression for DSL isotherm:
i
ciq
dln p A B
dln C D
×=+
(3.34)
( ) ( )1 1 2 2 2 2 1 11 1si i i i j j si i i i j jA q b b P b P q b b P b P = + + + + +
( )( )2 2 1 11 1i i j j i i j jB b P b P b P b P = + + + +
( )( )2
1 1 1 2 21 1si i j j i i j jC q b b P b P b P= + + +
( )( )2
2 2 2 1 11 1si i j j i i j jD q b b P b P b P= + + +
By combining the macropore and micropore resistances, Grande and Rodrigues
(2005) obtained the following bi-LDF rate equation:
( ) ( )*
2
15(1 )pi ci
p pi i pi p c ci cic
q Dk c c q q
t rρ ε ρ
∂= − + − −
∂ (3.35)
w
co
an
d
d
th
(2
b
4A
sy
3
co
in
d
ca
su
where, * ciq is
orresponding
nd energy b
escribed pre
This com
escribing kin
he developed
2005), and th
efore, Grand
A zeolite. T
ystem are no
Propy.3
This stud
ontrolling in
ndicators are
ispersion c
alculations a
ummarized i
the equilibr
g to its con
balance equa
eviously.
mpletes our
netically sel
d pore diffus
hen compare
de and Rodr
Therefore, th
ow identified
ylene/prop
dy assumes
ntra-particle
e discussed
oefficient,
are calculate
in Table 3-1
rium adsorba
centration in
ations for bi
discussion o
ective PSA
sion model w
es its perform
rigues (2005
he appropria
d.
pane system
nitrogen as
transport m
here. Exte
Knudsen d
ed. The detai
.
55
ate concentr
n the macro
i-LDF mode
of the two m
processes. In
with the exp
mance with
5) studied th
ate adsorptio
m
s inert on 4A
mechanism a
ernal fluid f
diffusivity a
iled value of
ration of com
opore gas, c
el are the sa
models (por
n the follow
perimental d
that of the b
he separation
on models a
A zeolite. A
and calculat
film of mas
and axial
f mass and h
mponent i i
pic . The bou
ame as pore
re diffusion
wing, this stu
ata of Grand
bi-LDF mod
n of propylen
and data fo
Adsorption e
tion of proc
ss transfer c
heat disper
heat transpor
in the microp
undary condi
diffusion m
and bi-LDF
udy first vali
de and Rodr
del. As menti
ne/propane u
r simulating
equilibrium
cess perform
coefficient,
rsion coeffi
rt coefficient
pores
itions
model
F) for
idates
rigues
ioned
using
g this
data,
mance
axial
ficient
ts are
Tw
b
L
m
d
w
is
an
d
eq
Table 3-1: Mawith propylene
3.3
The expe
een reported
Langmuir (M
mixture equi
iffusion and
*
1ciq =+
where subscr
s the temper
nd 2 02ib b=
ependence.
The imp
quilibrium
ass and heat te/propane fee
ParameterDC0 Dm Dk Dp kf
Kg hf λ
Cp
Adsorp.1
erimental eq
d by Grande
MSL) isother
ilibrium isot
d bi-LDF).
1 1
1 1s i i i
j jj
q b p
b p+
+
ripts 1 and 2
rature-indepe
2 /i gH R T
ie− Δ a
plicit MSL
loadings in
transport parad at 423 K, 2
rs C3H5.50E
0.073.070.033.3
3.9E1.7E
1.11E84
ption data
quilibrium d
e et al. (2003
rm model to
therm (Math
2 2
21s i i i
j jj
q b p
b p+
represent th
endent satur
are the isot
model req
n the solid
56
ameters used50 kPa, and 7
H6 CE-12 2.773 078 335 02 -04 3.-02 1.
E-02 1.04
data for prop
3) Grande an
o these data
hias et al.,
he first and
ration capaci
therm const
quires a non
d phase. Th
d in simulatin7.5 cm/s.
C3H8 70E-14 0.069 3.007 0.033 3.16 .1E-04 .4E-02 04E-02
99
pylene and
and Rodrigue
a. In this st
1996) is u
second sets
ity of adsorb
tants with
nlinear equ
his comput
ng the breakth
unit cm2/s cm2/s cm2/s cm2/s cm/s
W/cm-K W/cm2-K W/cm-K J/mol-K
propane on
es (2005) fit
tudy, the fol
used for bo
of sites in th
bate i , and
Arrhenius-t
uation solve
tational bur
hrough exper
4A zeolite
tted the Mul
llowing dua
oth models
(3.36)
he adsorben
1 01
H
i ib b e−Δ=
type temper
er to obtain
rden is red
riment
have
ltisite
al-site
(pore
nt, siq
1 /i gH R T
rature
n the
duced
57
significantly by using the explicit DSL model. The parameters
1 2 01 02 1 2( , , , , , )s i s i i i i iq q b b H H−Δ −Δ for the dual-site models are obtained from
independent fits of the single-component equilibrium data to the unary form of the dual-
site isotherm model. The fits of the DSL model to the experimental equilibrium data of
propylene and propane on 4A zeolite are shown in Figure 3.2 with fitted parameters in
Table 3-2. As shown in these figures, the DSL model provides a good fit. A perfectly
positive correlation is assumed for the binary prediction (Ritter et al., 2011).
Grande and Rodrigues (2004) measured the individual transport parameters of
propylene and propane on 4A zeolite by three different methods, namely zero length
column (ZLC), column breakthrough and gravimetry. The kinetic parameters obtained
from these three techniques were in good agreement.
Table 3-2: Parameters of the Dual-site Langmuir isotherms for propylene and propane on 4A zeolite.
Gas qs1(mol/kg) qs2(mol/kg) b01(/kpa) b02(/kpa) -ΔH1(kJ/mol) -ΔH2(kJ/mol) propylene 0.7656 1.1866 4.20E-05 4.49E-05 58.01 20.38 propane 1.7527 0 4.55E-09 0 16.23 0
(a)
0 100 200 300 400 5000.0
0.5
1.0
1.5
2.0
2.5
Propylene 373K Propylene 423K Propylene 473K DSL
q (m
mol
/g)
P (kPa)
Fw
in
ze
ca
igure 3.2: Exwell fitted by t
3.3
With
ndeed the co
eolite. The f
apacities in t
c
p
D
Dγ =
1α
ε−
=
perimental dathe dual-site L
Contro3.2
the above
ontrolling ma
following tw
the micropor
2
2
/
/c
p
r
R
p
p
Kε
ε−
00.0
0.5
1.0
1.5
q (m
mol
/g)
ata for the adLangmuir isot
olling transp
data, this w
ass transfer
wo equations
res and macr
(b)
100
58
sorption equitherm.
port mechan
work can no
mechanism
represent th
ropores, resp
200
Pr Pro DS
P (kPa)
ilibrium of pr
nism
ow confirm
for propylen
he ratios of d
pectively (Si
300
ropane 423Kopane 473K
SL
ropylene (a) a
that microp
ne/propane a
diffusional ti
ilva and Rod
400
K
and propane (
pore diffusi
adsorption o
me constant
drigues, 199
(3.37)
(3.38)
(b) are
on is
on 4A
ts and
6) :
w
th
γ
in
w
4A
fo
b
3
re
P
P
P
where K is th
he controllin
( )1 10γ α+ >
nto account.
which confirm
A zeolite pa
or describing
eing in equil
Proces.4
For eval
ecovery and
Propylene Pu
Propylene Re
Propylene Pro
he dimensio
ng mechanis
0 and if 0.1
For propan
ms that mic
articles. Thu
g propylene/
librium with
ss perform
luating the p
productivity
urity (%) =
ecovery (%)
oductivity (mo
onless Henry
sm for (1γ +
( )1γ α< + <
ne and propy
ropore resis
us, our pore
/propane sys
h the bulk ga
mance
performance
y for propyle
3 6
0
0
Ev
Evacuation
t
t
C HC v
0
0
Evacuation
Pressurization
t
C
t
C
C=
1 1ol.hr .kg )− − =
59
y's law const
) 0.1α+ < , m
10< both ma
ylene on 4A
tance is the
diffusion m
stem. Furthe
as in the inter
e of the PS
ene as define
3 6
0 0
vacuation
Evacu
C H z
t
z
C v
v dt
=
=+
3 6
3 6
0
0
C H z
C H z
v dt
C v dt
=
=
−
+
0(
3600
Evacuatt
=
tant (0
limp
P P R
ρ→
macropore d
acropore and
A Zeolite, thi
dominant m
model should
rmore, our a
r-particle vo
SA process,
ed by Grand
3 8
0
0
uation
C H z
dt
C v
=
=
3 6
3 6
0
0
Rinse
Feed
t
C H z
t
C H
C v
C v+
3 6 0
tion
C H z
Total Adsorp
C v d
t V=
*p p
g
q
R T). Micro
diffusivity i
d micropore
is study find
mass transpo
be the mor
assumption o
oid spaces of
this study
de and Rodrig
0
*100dt
0
0
*100z
z
dt
dt
=
=
3 60
Rinset
C H
ption Adsorption
dt C
ρ−
opore diffusi
is controllin
e should be t
ds ( )1γ α+ <
ort mechanis
e accurate m
of macropor
f the bed is v
uses the p
gues (2005)
(3.39)
(3.40)
0)
zv dt
=
(3.41)
ion is
ng for
taken
0.1< ,
sm in
model
re gas
valid.
urity,
:
3
P
in
M
ar
re
(a
sq
to
pr
pr
up
ad
M
Model.5
Dell Opt
rocessor, 8 G
n the previo
Multiphysics,
re used to i
epresents the
at each axia
quare to rep
o overcome
roblem, and
rofiles are se
p in the s
dsorption, ri
MATLAB to
l solutions
tiplex 780
GB of RAM
ous section
, which uses
implement t
e axial direc
al position w
resent the co
the need to
reduce com
et up in the l
quare geom
inse, blowdo
execute the
r(cr
ysta
l rad
ial )
s
with Intel(R
M is used for
are written
s the finite e
the pore dif
ction ( z ) of
within the b
oncentration
simulate a
mputational in
line geometr
metry. Five
own and eva
cycling of P
Figure 3.3:
z(column
60
R) Core(TM
numerical s
n in dimens
element met
ffusion mod
f the bed and
bed), respect
n profiles wi
full sphere a
ntensity. The
ry whereas th
COMSOL
cuation step
PSA steps.
: Schematic o
n length)
Concent
M) 2 Quad
simulation. T
sionless form
thod. Two d
del. A squar
d the radial
tively as sh
ithin the part
at each posi
e PDEs that
hose for the
files, repre
ps, are solved
of equation do
tration
CPU Q940
The model e
m and solv
dimensions (
re of unit le
direction (r
hown in Figu
rticle along t
ition in the b
describe the
micropartic
esenting th
d and export
omain.
00 @ 2.66
equations det
ed in COM
(axial and ra
ength and w
r) of the par
ure 3.3. Us
the bed is he
bed, simplif
e bulk fluid p
cle profiles a
e pressuriza
ted as modu
GHz
tailed
MSOL
adial)
width,
rticles
ing a
elpful
fy the
phase
are set
ation,
les to
ph
b
tr
th
on
v
b
b
co
ca
co
at
v
th
ph
The cycl
hase concen
e saturated
ransferred as
he change in
ne-dimensio
ariations. Th
etween 30-5
i-LDF mode
3.5
Mass bal
onducting t
alculated fro
oncentration
t the inlet.
0
1tτ
=
In the ab
elocities, res
he adsorbate
hase in equ
e simulation
ntration is as
with feed a
s the initial c
n purity is le
onal geometr
he number o
50 cycles. It
el.
Accura5.1
lance and en
the simulati
om the mas
n and flow r
1 e e
f f
y vdt
y v
− =
bove equati
spectively, y
mole fractio
uilibrium wi
n sequence i
ssumed to be
at low press
conditions fo
ss than 0.01
ry is sufficie
of cycles ne
required 3-4
acy of mass
nergy balance
ion study.
ss balance o
ate are mon
11
f
L
v
εε
−= +
on, L is th
fy and fc
on in the exi
ith fc . The
61
s started wit
e that of feed
sure. The be
or the subseq
% for five c
ent, since it
eeded to rea
4 h of CPU t
and energy
e errors for t
The mean
of an initial
nitored after
p p
f
q
c
ρε
he column l
are the mole
it stream and
e left-hand
th the pressu
d gas. Initia
ed profiles
quent step. T
consecutive c
does not all
ach the cycl
time for the
y balances
the two mod
resistance
lly clean ad
introducing
length, fv
e fraction an
d pq is the c
side of Eq.
urization ste
al solid loadi
at the end
The cycling i
cycles. In th
low any rad
lic steady st
pore model
dels are comp
time, t , o
dsorption co
a concentra
and ev are
nd feed con
concentratio
. 3.42 is ob
ep. The initia
ing is assum
of each step
is continued
he bi-LDF m
dial concentr
tate (CSS) v
and 1-2 h fo
puted first b
of the adso
lumn where
ation step ch
(3.42)
inlet and o
ncentration,
n in the adso
btained from
al gas
med to
p are
d until
model,
ration
varied
or the
before
orbate
e exit
hange
outlet
ey is
orbed
m the
si
is
th
eq
w
an
m
3
b
n
pr
ex
m
S
an
imulated bre
sotherm data
he bi-LDF m
Breakthr
quation:
0
1T
T
τ∞
−
where subscr
nd 0 show
models are 1.
Break.6
To valid
inary breakt
ow. The se
rovides the
xperiment c
mol% propan
LPM, 423 K
The b
nd diffusion
eakthrough r
a. The pore
model.
ough in an a
0 0
1T vdt
T v ε∞ ∞ −+
ript ∞ repre
ws the initial
5% and 0.01
kthrough r
date our non
through expe
et of equatio
necessary m
chosen for th
ne, and 51
K, and 250 kP
reakthrough
n of propyle
response, and
diffusion m
adiabatic col
0 0
piL
v T
ρε
ε
0 0
L
v T
esents the fin
l conditions.
1%, respectiv
results
n-isothermal
eriments rep
ons describi
model for sim
his work ha
mol% nitro
Pa, respectiv
h results in F
ene and prop
62
d the right-h
model had an
lumn must a
( )(i i
pg
H q
Cc
∞−Δ
11
pρε
ε
−+
nal condition
. The heat b
vely.
pore diffus
orted by Gra
ing high-pre
mulating the
as the feed c
ogen. Feed
vely.
Figure 3.4 cl
pane on 4A
hand side is c
n error of 0.7
also satisfy t
0 )iq−=
i pgii
pg
q c
Cc
∞
+
n of the col
balance erro
sion model
ande and Ro
essure adsor
e binary bre
composition
rate, tempe
early illustra
zeolite is w
calculated fr
79% compa
the followin
1 s ps
pg
c
Cc
ρεε
−
lumn (equili
ors for the b
with experi
odrigues (200
rption from
eakthrough e
n of 25 mol
erature and
ate that the b
well represe
rom the oper
ared to 0.82%
ng energy ba
0( )T T∞ − (3.
ibrium cond
bi-LDF and
imental data
05) are simu
0z = to z
experiments
% propylen
pressure are
binary adsor
nted by the
rating
% for
alance
.43)
ition)
pore
a, the
ulated
z L=
. The
ne, 24
e 1.1
rption
pore
63
diffusion model with the parameters estimated from the single component measurements.
Moreover, the pore diffusion model has a better match with the experimental data for
propylene compared to the bi-LDF model, which is quantitatively supported by a lower
mean square error (MSE = 1.05E-04 versus 2.10E-04). It is evident from the almost
instantaneous breakthrough of propane that its uptake in the adsorbent micropores in the
observed time-scale is practically negligible. The roll-up in its experimental breakthrough
profile is a typical under-damped response of the flow meter at the column exit. Hence,
comparing the mean square errors (MSE) of the two models for propane breakthrough is
not very meaningful.
0 500 1000 1500 20000.0
0.1
0.2
0.3 Propane Propylene Pore Bi-LDF
Mol
ar F
low
(m
mol
/s)
t (s)
Figure 3.4: Experimental measurements and simulated breakthrough responses for propylene and propane at 423 K and 250 kPa. The MSEs for model predictions are 1.05E-04 (C3H6) and 4.81E-05 (C3H8) for the pore model and 2.10E-04 (C3H6) and 3.45E-05 (C3H8) for the bi-LDF model.
le
F
ev
0
th
ex
Fcmfo0
3
R
The tem
ength in the
igure 3.5. T
vident in ad
.156 (middle
he bi-LDF m
xperimental
igure 3.5: Temm) and bottomor model pred.250 (top), 0.8
PSA r.7
The P
Rodrigues (20
mperature pro
same breakt
The superio
ddition to th
e) and 0.207
model). The
data more a
422
424
426
428
430
432
T (
K)
mperature prom (18 cm) of dictions are 0.829 (middle)
results
PSA experim
005) are sim
ofiles measu
through exp
or agreemen
he pore mod
7 (bottom) vs
e MSE valu
accurate than
0 52
4
6
8
0
2
ofiles for the bthe column. T.059 (top), 0.1and 0.717 (b
ments for pro
mulated using
64
ured at thre
eriment are
nts with por
del’s signifi
s. 0.250 (top
es show tha
n bi-LDF mo
500 10
t
breakthroughThe distances156 (middle) ottom) for the
opylene/prop
g the two mo
ee different
compared w
re model pr
icantly lowe
p), 0.829 (mi
at pore diffu
odel.
Bott Midd Top Pore Bi-L
000 15
(s)
h experimentss are measureand 0.207 (boe bi-LDF mod
pane separat
odels describ
locations al
with the mod
redictions a
er MSE valu
iddle) and 0
usion model
tomn of columndle of column
p of columne
LDF
500 20
s at the top (6d from the feottom) for thedel.
tion reported
bed earlier. T
long the co
del predictio
are even vis
ues (0.059
.717 (bottom
l can predic
n
000
8 cm), middleed end. The Me pore model
d by Grande
The experim
olumn
ons in
sually
(top),
m) for
ct the
e (43 MSEs and
e and
mental
65
conditions are summarized in Table 3-3. The equilibrium, kinetic and heat transfer
parameters in Table 3-1 and Table 3-2 are used in the simulations. The purity, recovery,
and productivity of propylene are calculated using Eqs. 3.39 - 3.41. The measured
pressure profiles are appropriately fitted to linear or exponential equations and used as
inputs in the simulations for the pressure-changing steps. Representative pressure profiles
in a cycle are shown in Figure 3.6.
0 100 200 300 4000
100
200
300
Pre
ssur
e (k
Pa)
Time (s)
Experiment Simulation input
Figure 3.6: Experimentally measured pressure profiles and their linear or exponential fits used in the simulation (value in blowdown and evacuation step is 6 s-1 and 0.15 s-1, respectively). For experimental details, see run 4 in Table 3-3.
The experimentally observed effects of the nitrogen mole fraction and feed
temperature on the purity and recovery of propylene are compared with two model
predictions in
Figure 3.7 and
66
Figure 3.8, respectively. Representative propylene and propane flow rates measured over
a cycle after the cyclic steady state is reached in one experiment from each of the two sets
are similarly compared with the model predictions in Figure 3.9. The total flow rate
measured is converted to component flow rates using the measured compositions of these
streams. Representative temperature profiles measured over a cycle at three different
locations in the column after the cyclic steady state is reached are shown in Figure 3.10,
where the model predications are also included.
0.1 0.2 0.3 0.4 0.595
96
97
98
99
100
Experimental purity Pore Bi-LDF
Pu
rity%
N2 mole fraction
84
85
86
87
88
89
90
Exprimental recovery
Rec
ove
ry%
67
Figure 3.7: Prediction of the effect of nitrogen in the feed on the purity and recovery of propylene compared with experimental results. For experimental conditions, see run 1-3 in Table 3-3.
The overall observation from
Figure 3.7 - Figure 3.10 is that both the models capture the experimental trends in the
range of operating conditions investigated. In the purity-recovery plots shown in
Figure 3.7 and
Figure 3.8, clearly the pore model predictions are quantitatively closer to the
experimental results than those from the bi-LDF model.
A perfect positive correlation has been assumed in this study for binary prediction
using the DSL model. DSL constants of propylene ( ) have higher values than that
DSL constants of propane ( ), and . As a result, with perfect negative
correlation propane equilibrium is somewhat higher under binary conditions compared to
perfect positive correlation. For propylene, the effect is negligible on its binary
equilibrium for the composition and pressure range covered in this study. The effect of
using perfect negative correlations on PSA simulation is also shown in
Figure 3.8. Although the qualitative trends are similar, the predictions with perfect
negative correlation are quantitatively far removed from the experimental results.
The component flow rates over a complete cycle compared in Figure 3.9 also suggest
marginal quantitative superiority of the pore diffusion model. In case of the temperature
profiles in Figure 3.10, the pore model also seems closer to the experimental data for
most part; except the adsorption step where the temperature rise was rapid. It is important
11 21,b b
12 22,b b 22 0b = 11 21b b>
68
to note that during a rapid change in temperature, the thermocouple readings are affected
by their response times and the thermal conduction along the probe wall.
Figure 3.8: Prediction of the effect of feed temperature on the purity and recovery of propylene compared with experimental results. For experimental conditions, see run 3-5 in Table 3-3. PN is perfect positive correlation and PP is perfect negative correlation.
In order to investigate the importance of fluid-solid heat transfer resistance, the fluid-
solid heat transfer coefficient is varied over three orders of magnitude above the value
given in Table 3-1. These perturbations do not affect the purity and recovery results in
Figure 3.8 and Figure 3.9 as well as the temperature profiles in Figure 3.10, which
confirms that the adsorbent is, in fact, in thermal equilibrium with the fluid phase. With
400 420 440 46095
96
97
98
99
100
Experimental purity Pore (PP) Pore (PN) Bi-LDF
Pur
ity%
T (K)
80
84
88
92
96
Experimental recovery
Rec
ove
ry%
69
the exception of a very rapid cycle, fluid-solid thermal equilibrium is a widely accepted
assumption in adsorption process modeling (Suzuki, 1990).
The small difference between the pore and bi-LDF model predictions for the present
system may mislead to conclude that the latter model with the concentration dependence
of micropore diffusivity accounted by Eq. 3.33 will always be a good approximation of
the more detailed pore diffusion model. A closer look at the representative concentration
profiles of propylene and propane along the crystal radius shown in Figure 3.11 reveals
that propane hardly enters the micropores during the cyclic operation. This means that the
diffusion of propylene in the micropores is practically like a single-component diffusion.
Table 3-3: Operating conditions of the PSA experiments taken from Grande and Rodrigues (2005).
Run no.
Feed component
C3H6/C3H8/N2
Phigh (kPa)
Pinter
(kPa)Plow
(kPa)tpr (s)
tad
(s) tri (s)
tbd
(s) tev (s)
T (K)
1 0.27/0.23/0.51 500 50 10 54 100 25 40 180 4332 0.37/0.31/0.32 500 50 10 54 100 25 40 180 4333 0.45/0.41/0.14 500 50 10 54 100 25 40 180 4334 0.45/0.41/0.14 500 50 10 54 100 25 40 180 4085 0.45/0.41/0.14 500 50 10 54 100 25 40 180 4636 0.25/0.24/0.51 500 50 10 60 60 25 40 220 433
70
0 100 200 300 4000.0
0.2
0.4
0.6
0.8 Propylene Pore Bi-LDF
Mol
ar fl
ow (
mm
ol/s
)
Time (s)
(b)
(a)
0 100 200 300 4000.0
0.2
0.4
0.6
0.8 Propane Propylene Pore Bi-LDF
Mol
ar fl
ow (
s)
Time (s)
71
Figure 3.9: Comparison of experimentally measured molar flow rates with model predictions over a cycle after reaching cyclic steady state. The results are from two different experimental runs, run 6 in (a) and run 4 in (b) and (c). For experimental details, see Table 3-3.
0 100 200 300 400
404
408
412
416 Bottom Middle Top Pore Bi-LDF
T (
K)
Time (s)
Figure 3.10: Temperatures measured at three different locations in the column over a cycle after reaching cyclic steady state in run 4, See Table 3-3 for experimental details.
(c)
0 100 200 300 4000.0
0.2
0.4
0.6
0.8 Propane Pore Bi-LDF
Mo
lar
flow
(m
mo
l/s)
Time (s)
72
0.0 0.2 0.4 0.6 0.8 1.00.00
0.05
0.10
0.15
0.20
0.25 Propylene (step 2) Propane (step 2) Propylene (step 5) Propane (step 5)
Dim
en
sio
nle
ss a
dso
rbe
dp
ha
se c
on
cen
tra
tion
r/rc
Figure 3.11: Concentration profiles of propylene and propane inside the crystal at z/L= 0.1 at the end of the high pressure adsorption (step 2) and the end of the evacuation (step 5) after reaching cyclic steady state in run 4 detailed in Table 3-3.
As pointed out earlier, the pore diffusion model used here captures the strong
influence of the concentration profiles of the two components in the microparticle during
binary diffusion, which is not captured by Eq. 3.33 used in the bi-LDF model. In the
absence of propane in the micropore, it is therefore not surprising that the two models
give such close results. To prove this point further, PSA simulations are carried out for
the conditions of run 4 in Table 3-3 by gradually increasing the diffusivity of propane.
The results are shown in Figure 3.12. Increasing propane diffusivity increases its
diffusion into the micropores developing its concentration profile, which was previously
absent. Hence, the difference between the two models grew larger, as expected.
73
0 50 100 150 200 25094
95
96
97
98
99
100
Propylene Propane
Dpropylene/Dpropane
Pu
rity
%
80
81
82
83
84
85
86
Rec
over
y%
Figure 3.12: Effect of propylene/propane diffusivity ratio on the purity and recovery predicted by the pore and bi-LDF models. The propane diffusivity was gradually increased while holding the propylene diffusivity constant. The experimental conditions are same as in run 4 in Table 3-3.
74
a
b
75
c
d
76
e
f
g
77
h
Fof
al
(F
igure 3.13: Df PSA run 4 a
Figure
ll five PSA
Figure 3.13a
Dimensionlessare shown in F
e 3.13 show
steps at the
a-b), propane
s adsorbate phFigs. a-j.
ws the solid c
e cyclic stea
e does not h
i
78
hase concentr
concentratio
ady state of
have sufficie
ration of prop
on profiles o
run 4 in Ta
ent time to a
pylene and pr
of propylene
able 3. Durin
adsorb into t
j
ropane in fiv
e and propan
ng pressuriz
the interior o
e step
ne for
zation
of the
79
crystals. However, propylene, being faster, begins to diffuse into the crystals at the
column inlet. Even during high-pressure adsorption (Figure 3.13c-d), propane adsorption
is limited to the crystal surface only and it breaks through almost immediately. Propylene
does move into the crystal interior, but it does not have enough time to saturate the entire
column. Thus, column capacity is not utilized fully, and better performance may be
obtained by increasing the time for adsorption. During rinse (Figure 3.13e-f), propane
concentration on the crystal surface near the column inlet is almost zero. This is because
the pure propylene feed from the column inlet pushes propane out from the other end.
Here, propylene gets sufficient time to diffuse into the crystals and it saturates most of the
crystals in this step. During blowdown (Figure 3.13g-h), most of the propane desorbs and
comes out from the column outlet. Decrease in the pressure also makes propylene move
from the interior to the surface of the crystals. Furthermore, propylene is also lost from
the column outlet during this step. By the time of countercurrent evacuation
(Figure 3.13i-j), little propane is left in the column. Propylene is withdrawn as the
product, but most of it comes out from the column inlet only. In other words, the duration
is not enough to recover all of propylene. Furthermore, while propylene concentration is
nearly zero at the crystal surface, and most of it still exists in the interior. The
dimensionless concentration of propylene at the crystal center is near 0.1 and same as that
at the crystal center in the previous step. Therefore, it is clear that more recoverable
propylene remains in the micropores at the end of the evacuation step.
The above discussion clearly suggests that the column operating parameters are
far from the optimal. Process performance can be improved significantly by proper
optimization. Figure 3.13 also gives us some idea of the conditions at the cyclic steady
st
b
as
in
ph
to
cy
3
co
d
T
pu
In
d
pr
th
m
H
ap
IT
pr
tate (CSS). I
ed saturated
ssumed. The
nitial conditi
hase and fee
o reach CSS
ycles to reac
Chapt.8
A non
ontrolled PS
iffusivity ac
The model e
ublished exp
n compariso
iffusion mod
resent system
he propylene
micropore du
Hence, the
ppropriate fo
TQ3, ZSM5
ropylene/pro
In this study
d with the fe
ese seem to
ion for the
ed gas for ga
. Comparing
ch CSS, whil
ter conclu
n-isothermal
SA separation
ccording to th
quations hav
perimental s
n to the bi-L
del is quanti
m. Further a
e/propane sy
ue to its very
pore diffus
or screening
8, SiCHA an
opane separa
y, for the pr
eed at the lo
be far from
pressurizati
as phase. Suc
g the two ini
le the latter t
sion
micropore d
n process th
he chemical
ve been solv
separation da
LDF model
itatively supe
analysis has r
ystem on 4A
y low diffus
ion model
g other poten
nd DD3R, re
ation.
80
ressurization
ow pressure
m what we o
on step wou
ch an initial
itial conditio
takes 25 cyc
diffusion mo
at allows for
l potential gr
ved in COM
ata for prop
advocated i
erior althoug
revealed that
zeolite, whe
sivity and sh
developed
ntial adsorbe
eported in th
n step in the
and gas ph
observe at C
uld be to as
condition m
ons, we obse
cles.
odel has bee
r concentrati
radient as th
MSOL and t
pylene/propa
in the literat
gh the differ
at the small d
ere propane
hould not be
and valida
ents, such as
he literature
e first cycle
hase at the fe
CSS. It appe
ssume a cle
may require f
erved that the
en developed
ion depende
he driving fo
the model w
ane separatio
ture for this
rence is not v
difference is
practically d
e mistaken a
ated in this
certain 8-rin
for the indu
of simulatio
feed conditio
ears that a b
ean bed for
fewer simula
e former tak
d for a kineti
ence of micro
orce for diffu
was verified
on on 4A ze
system, the
very large fo
indeed uniq
does not ente
as a general
s study is
ng silica zeo
strially impo
ons, a
ons is
better
solid
ations
kes 35
ically
opore
usion.
with
eolite.
e pore
or the
que to
er the
rule.
more
olites,
ortant
81
CHAPTER 4 Propylene/Propane Separation Using SiCHA
The ratios of Henry's constants and those of diffusion coefficients for propylene and
propane in several adsorbents (Grande and Rodrigues, 2004; Lamia et al., 2008; Padin et
al., 2000; Salil U. Rege et al., 1998; Sikavitsas et al., 1995) studied in the literature are
compiled in
Table 4-1. It is evident from the table that pure silica chabazite (SiCHA), a new 8-ring
silica zeolite, shows high kinetic selectivity between propylene and propane (Hedin et al.,
2008; Olson et al., 2004). However, this adsorbent has received limited attention. Kinetic
separation using this new adsorbent could be an attractive option for separating
propylene/propane. In this chapter, a 4-step, kinetically controlled pressure swing
adsorption process has been suggested for propylene/propane separation on SiCHA and
studied in detail using the non-isothermal micropore diffusion model, developed and
verified in Chapter 3. The Langmuir isotherm replaces the dual site Langmuir isotherm to
represent adsorption equilibrium for propylene/propane adsorption on SiCHA. This
study estimates the equilibrium information for propane indirectly using available uptake
data at 80 °C and 600 torr. Moreover, this work uses molecular simulation to obtain
equilibrium information of propylene and propane and confirms our estimation.
Tav
*
*
4
pr
P
(2
pr
pr
la
pr
˚C
es
eq
d
Table 4-1: Suvailable adsor
Adsor4A Ze
13X Z5A Ze
AgNO3
AlPAg+ exchan
SiCH
propylene
propane
K D
K
*Diffusion r
Adsor4.1
The lack
ropane in Si
SA process.
2004). It em
ropylene an
ropylene, an
atter. Due to
ropane. How
C, and 100
stimated its
quilibrium l
ata was indi
ummary of Hrbents at 80˚C
rbent eolite eolite
eolite 3/SiO2
PO4 nged resin HA
propylene
propane
D
D
ratios are not
rption Par
k of adequa
iCHA poses
. To date, th
mployed a th
nd propane
nd only at 80
o the same r
wever, it rep
˚C. Using th
s diffusivity
loading they
irectly analy
Henry constanC.
Kpropylene/12.429.9.311.12
5620.3
t high enoug
rameters fo
ate experim
a major cha
he lone expe
hermogravim
at 600 torr
0 ˚C for prop
reason, the s
ported equilib
heir lone up
y in SiCHA
y may have
yzed to shed
82
nt and diffus
/Kpropane D43 1 2 3
2 .8 8
gh to exploit
for SiCHA
ental studie
allenge in de
erimental stu
metric meth
. It reported
pane, becaus
study could n
brium data f
ptake data fo
A, but ment
used for tha
d some light
sivity coeffic
Dpropylene/Dpr
321 0.6
0.71 1.8 3
0.3 5023
t kinetic sele
A
es on the a
eveloping an
udy on SiCH
hod to meas
d uptake da
se of the ext
not measure
for propylen
or propane a
tioned nothi
at estimation
on the equi
cient for pro
ropane Kin
ectivity.
adsorption o
nd simulating
HA is the on
ure the dyn
ata at 30 ˚C
tremely slow
e the equilib
ne at 30 ˚C,
at 80 ˚C, Ol
ing about t
n. In their s
ilibrium isot
opylene/propa
netic Selectiv223 ** ** ** ** ** 28
of propylene
g a SiCHA-b
ne by Olson
namic uptak
C and 80 ˚C
w diffusion o
brium loadin
45 ˚C, 60 ˚C
lson et al. (2
the saturatio
study, the up
therm of pro
ane in
vity*
e and
based
et al.
kes of
C for
of the
ng for
C, 80
2004)
on or
ptake
opane
on
fo
so
˚C
te
w
an
an
b
te
in
k
co
co
re
n SiCHA. In
or propane
omewhat dif
4.1
For prop
C, and 100
emperature:
1e
s
q b
q=
+
0exb b=
where, b is t
nd sq is the
nd b . Howe
e independe
emperature w
ndependentq
J/mol, and
orresponding
orresponding
equired equi
n addition, th
reported by
fferent, whic
Equilib.1
ylene, Olson
˚C, and fitt
bp
bp+
xp( )H
RT
Δ−
he Langmui
saturation l
ever, for the
ent of temp
were simulta
sq , 0b and Δ
0 4.55b = ×
g to these e
g Langmuir
librium isoth
he equilibriu
y Olson et
ch are discus
brium Param
n et al. (200
ted a separa
ir constant,
oading. Usin
Langmuir is
perature. He
aneously reg
HΔ . This pr
810−× /torr f
equilibrium
r model fit
herm parame
83
um and uptak
al. (2004)
ssed next.
meters
4) reported
ate Langmu
eq is the eq
ng Eq. 4.1, t
sotherm to b
ence, the pr
gressed with
rocedure ga
for propyle
parameters
are shown
eters for pro
ke data for p
have been
isotherm da
uir isotherm,
quilibrium lo
they reporte
be thermody
ropylene eq
h Eqs. 4.1 an
ave us sq =
ene. The M
were 5.07.
in Figure 4
opylene.
propylene an
reanalyzed.
ata at 30 ˚C,
, given by E
oading at pa
ed five separ
ynamically co
quilibrium d
nd 4.2 to ob
125.2= mg/g
Mean Squar
The experi
4.1. With th
nd the uptake
Our result
45 ˚C, 60 ˚C
Eq. 4.1 for
(4.1)
(4.2)
artial pressur
rate values f
onsistent, sq
data for diff
btain temper
g, 3HΔ = −
re Error (M
imental data
his, we hav
e data
ts are
C, 80
each
re ,p
for sq
s must
ferent
rature
33.08
MSE)
a and
e the
84
Figure 4.1: Adsorption isotherms for propylene on SiCHA. Points represent the experimental data by Olson (2004) and solid lines represent the Langmuir isotherm.
In the absence of any experimental measurement of sq for propane, we argue as
follows to assume that it is the same as that ( sq = 125.2 mg/g) for propylene on SiCHA.
SiCHA is a neutral adsorbent that interacts with propane and propylene via van der Waals
forces. The slightly larger size (4.35 Å) of propane versus that (4.05 Å) of propylene
allows propane to adsorb slightly more strongly than propylene (Ruthven and Reyes,
2007) at low to moderate pressures. At high pressures and saturation loading, however,
the smaller size and linear structure of propylene molecule may suggest slightly more
adsorption for propylene. Given that the molecule sizes are very close, it is reasonable to
assume that their saturation loadings are nearly equal. Therefore, this work takes sq =
125.2 mg/g for both propane and propylene in this study.
0
20
40
60
80
100
120
140
0 100 200 300 400 500 600 700
q(m
g/g)
P(Torr)
30 C45 C60 C80 C100 C
85
To further justify our above argument and confirm our assumption of propane’s sq ,
This study employs Monte Carlo (MC) molecular simulation described in next section.
This study first matches the theoretical prediction of propylene isotherm at 80 ˚C with
experimental results. Figure 4.2 shows that the predictions from molecular simulation
match the experimental data very well. Then, the isotherm of propane is computed using
molecular simulation. Fitting Langmuir isotherms to these simulation results, 1 27.3sq =
mg/g for propane and 127.2 mg/g for propylene are obtained, which are identical for all
practical purposes. Note that this predicted sq matches quite well with our estimated
value of 125.2sq = mg/g for propylene from Olson’s experimental equilibrium data.
Determination of the other Langmuir isotherm parameter for propane and reanalysis of
the kinetic parameters for propylene and propane are discussed next.
Figure 4.2: Propylene and propane equilibrium isotherm in SiCHA at 80 ˚C obtained from MC simulation are compared with experimental data and Langmuir model estimates, respectively. The Langmuir model parameters were obtained indirectly from the uptake data of Olson et al. (2004).
0
20
40
60
80
100
120
140
0 200 400 600
q(m
g/g)
P(Torr)
Propylene Exp.
Propylene MCsim.
to
an
S
(Y
6
ar
ca
in
4.1
The prim
o first valida
nd then to e
iCHA has a
Yakubovich
.6 × 6.6 Å3.
re formed by
ages and win
Figure
Many sim
nteractions o
Molecu.2
mary objectiv
ate the abilit
stimate the
a space group
et al., 2005
The cage i
y eight-mem
ndows in SiC
e 4.3: Illustrat
mulation stud
of Si atoms
ular dynami
ve here is to
ty of simula
saturation ca
p of R-3m w
). Each unit
s connected
mbered rings
CHA.
tion of CHA s
dies on adso
(Smit and
86
ic simulatio
o use the tech
ation to pred
apacity of p
with lattice c
cell contain
to six neigh
s with a diam
structure. Thein green.
orption in zeo
Maesen, 20
on
hnique of M
dict the satur
ropane. As
constants a =
ns an ellipso
hboring cag
meter of 3.9
e cages are in
olites neglec
008). To fu
Monte Carlo
ration capac
a naturally o
= 13.831 Å a
oidal cage wi
ges by small
9 Å. Figure 4
ndicted in blue
ct the short-r
ully and atom
(MC) simul
city of propy
occurring ze
and c = 15.0
ith a size of
windows, w
4.3 illustrate
e and the win
ranged dispe
mistically m
lation
ylene,
eolite,
023 Å
f 11 ×
which
es the
ndows
ersion
mimic
87
SiCHA framework, this study consideres the dispersion interactions of both Si and O
atoms in SiCHA under study. Table 4-2 lists the potential parameters of Si and O atoms,
which are optimized to reproduce the experimental heats of adsorption (Hirotani et al.,
1997). Two types of models are commonly used to mimic hydrocarbon molecules,
namely the united-atom model and all-atom model (Ryckaert and Bellemans, 1978). Both
models give comparable adsorption isotherms in silicalite; however, computation is faster
with the united-atom model. Consequently, the united-atom model is used in this work
with every CHx group as a single interaction site. The bond lengths are assumed to be
rigid. The nonbonded dispersive interactions are modeled by the Lennard-Jones (LJ)
potential:
12 6
LJ ( ) 4 [( / ) ( / ) ]u r r rε σ σ= − (4.3)
The bond bending is represented by a harmonic potential:
2bending 0( ) 0.5 ( )u kθθ θ θ= − (4.4)
Table 4-2 gives the force field parameters for propane and propylene, in which the
LJ parameters are optimized to reproduce the experimental vapor-liquid coexistence
curves and critical properties of pure hydrocarbons (Martin and Siepmann, 1998, 2000).
Adsorption of C3H6 and C3H8 in SiCHA is simulated by grand-canonical Monte Carlo
(GCMC) method. The conventional Metropolis technique in MC simulation is
prohibitively expensive in sampling the conformation of large molecules. To improve
efficiency, advanced configurational-bias technique is adopted in which a molecule is
grown atom-by-atom biasing energetically favorable configurations while avoiding
overlap with other atoms (de Pablo et al., 1992; Frenkel et al., 1992; Siepmann and
Frenkel, 1992). At first, five trial positions are generated with a probability proportional
88
to exp( )iinternalUβ− , where 1/ Bk Tβ = and i
internalU is the internal energy at a position i
including the intramolecular bond bending interactions. Then, one of the trial positions is
chosen for growing an atom with a probability proportional to
( ) ( )exp / expi iexternal external
i
U Uβ β− − , where iexternalU is the external energy including the
intermolecular LJ interactions.
Table 4-2: Force field parameters for SiCHA, propylene, and propane.
Site (A)σ ε /kB (K)
Lennard-
Jones
Si 0.677 18.60
O 2.708 128.21
−CH3 3.75 98.0
−CH2 − 3.95 46.0
= CH2 3.675 85.0
= CH− 3.73 47.0
Bond x yCH CH−
x yCH CH=
r0 = 1.54 Å
r0 = 1.33 Å
Bending x 2 yCH CH CH− −
x yCH CH CH= −
kθ /kΒ = 62500 Κ, θ 0 = 114.0ο
kθ /kΒ = 70420 Κ, θ 0 = 119.7ο
The SiCHA framework is treated as rigid and periodic boundary conditions are used
in three dimensions to mimic the periodicity. A spherical cutoff length of 11 Å is used to
evaluate the LJ interactions along with long-range corrections. A typical GCMC
simulation is carried out for 20000 cycles, in which the first 10000 cycles are used for
89
equilibration and the second 10000 cycles for ensemble averages. Each cycle consisted of
a number of trial moves: (a) Translation; either a randomly selected adsorbate molecule is
translated with a random displacement in x, y, or z dimension and the maximum
displacement is adjusted to an overall acceptance ratio of 50%. (b) Rotation; either a
randomly selected adsorbate molecule is rotated around x, y, or z dimension with a
random angle, and the maximum angle is adjusted to an overall acceptance ratio of 50%.
(c) Partial regrowth; part of a randomly selected adsorbate molecule is regrown locally. It
is decided at random which part of the molecule is regrown and from which bead the
regrowth was started. (d) Complete regrowth; a randomly selected adsorbate molecule is
regrown completely at a random position. (e) Swap with reservoir; a new adsorbate
molecule is created at a random position or a randomly selected adsorbate molecule is
deleted. To ensure microscopic reversibility, the creation and deletion are attempted at
random with equal probability. The simulation statistical uncertainty is estimated by
block transformation technique and found to be generally smaller than the symbol sizes
presented in the figures below.
Figure 4.2 shows the isotherms of C3H6 and C3H8 in SiCHA at 80 °C. The
experimental isotherm of C3H6 is available and thus included for comparison (Olson et
al., 2004). Good agreement is found between the predicted and experimental data for
C3H6 over the entire range of pressure under study. This validates our models and force
fields used in the simulation. The predicted adsorption of C3H8 is greater than that of
C3H6 over the pressure range. This is consistent with the potential parameters used for
C3H8 and C3H6. As listed in Table 4-2, −CH3 group possesses the strongest interaction
strength. C3H8 contains two −CH3 groups, which is more than that in C3H6.
C
C
v
sm
ad
at
d
th
in
m
as
te
Consequently
C3H6. Howev
an der Waal
maller than
dsorption at
t 700 torr. A
ensity is at t
herein. Upon
n accord with
Based
mg/g) of prop
s propylene
Figure 4.4
4.1
Olson et
emperatures
y, C3H8 has
ver, the satur
ls volume of
62.3 Å3 fo
saturation. F
All adsorbate
the cage cen
n comparison
h the greater
d on our sim
pylene, this s
(125.24 mg/
4: Density con
Kinetic.3
t al. (2004)
(30 ˚C and
a stronger i
ration loadin
f C3H6 is esti
or C3H8. Th
Figure 4.4 sh
e molecules
nters. The de
n, C3H8 exh
r adsorption
mulations and
study estima
/g).
ntours of C3H
c Parameter
) measured
80 ˚C). How
90
interaction w
ng is predom
imated to be
hus, it is ex
hows the den
are observe
ensity in the
ibits a highe
discussed ab
d using the e
ates the satur
H6 and C3H8 in
rs
the uptake
wever, they r
with the SiC
minately gov
e 56.5 Å3 fro
xpected that
nsity contou
ed to reside
windows is
er density th
bove for C3H
experimenta
ration capac
n CHA at 700
of propyle
reported upta
CHA framew
verned by ad
om Materials
at C3H6 has
urs of C3H6 a
in the cages
zero implyi
han C3H6 in t
H8.
al saturation
ity of propan
0 torr. The un
ene at 600
ake data for
work compar
dsorbate size
s Studio, wh
slightly sm
and C3H8 in
s and the hi
ing no adsor
the cages. T
capacity (12
ne to be the
nit of density
torr and at
propane at 8
red to
e. The
ich is
maller
CHA
ighest
rption
This is
25.24
same
scale
t two
80 ˚C
91
only. Pure uptakes of SiCHA were used. They used analytical solution of the diffusion
model for planer geometry subjected to a constant boundary condition (Crank, 1975) to
compute the micropore diffusivity for both propane and propylene.
2 2
2 2 20
(2 1)81 exp
(2 1) 4c
ne
n D tq
q n l
ππ
∞
=
− += − + (4.5)
However, it is not clear if they extracted both cD and eq from Eq. 4.5, or they
assumed eq and extracted cD . For propane, they reported only cD and not eq .
SiCHA used in the study has 3D crystals rather than planar sheets. Thus, Eq. 4.5 is
not the most appropriate choice for describing the uptake of propylene and propane on
SiCHA. A more appropriate and general approach would be to assume a spherical
geometry (Ruthven et al., 1994) .
22
1c
q qr D
t r r r
∂ ∂ ∂ = ∂ ∂ ∂ (4.6)
(0, )
0t
q
r
∂ =∂
(4.7)
( , )cer t
q q= (4.8)
( ,0)0
rq = (4.9)
If one assumes a concentration-independent micropore diffusivity ( cD ), then Eqs.
4.6 - 4.9 have the following analytical solution (Ruthven, 1984).
2 2
2 2 21
6 11 exp c
ne c
n D tq
q n r
ππ
∞
=
= −
(4.10)
If one does not make that assumption, and allows diffusivity coefficient to vary with
concentration as in 0 / (1 / )c c sD D q q= − , then Eqs. 4.6 – 4.9 must be solved numerically.
92
This study uses the method of orthogonal collocation. Thus, it is possible to use the above
three approaches to model the uptake data of Olson et al. (2004) for propane and
propylene at 600 torr, and estimate both eq and micropore diffusivity values ( 0orc cD D ).
Figure 4.5 and Figure 4.6 show the uptake of propylene at 30 ˚C and 80 ˚C
respectively, and the above three fitted models. The spherical model with concentration-
dependent micropore diffusivity represents the best fit with the least MSE at both 30 ˚C
and 80 ˚C. At 30 ˚C, its MSE is 15 versus 17 for constant diffusivity micropore model
and 19 for Crank’s solution. At 80 ˚C, it is 2.1 versus 3.3 for constant diffusivity
micropore model and 7.0 for Crank’s solution. The MSE for 30 ˚C is higher than at 80 ˚C
because diffusion is slower at lower temperatures and hence uptake measurements may
have more uncertainty. The eq values are very similar (~119 mg/g at 30 ˚C and ~89 mg/g
at 80 ˚C) from all three models, as listed in Table 4-3. Interestingly, in addition to the
differences in the values from the three models, some discrepancy also exists between the
values reported by Olson et al. (2004) and those computed by us from Eq. 4.5.
93
Figure 4.5: Experimental and simulated uptake data for propylene in SiCHA at 30 ˚C and 600 Torr.
Figure 4.6: Experimental and simulated uptake data for propylene in SiCHA at 80 ˚C and 600 torr.
0
20
40
60
80
100
0 2 4 6 8 10 12 14
q(m
g/g)
t0.5(min0.5)
Propylene 80˚C
Experimental
Analytical (spherical)
Numerical (spherical)
Cranc's Solution (planer)
Olson's Results
0
20
40
60
80
100
120
140
0 200 400 600
q(m
g/g)
t0.5(min0.5)
Propylene 30˚C
Experimental
Analytical (spherical)
Numerical (spherical)
Crank' Solution (planer)
Olson Results
94
Table 4-3: Equilibrium and diffusivity information obtained from the uptake of propylene and propane in SiCHA at 600 torr.
Component Model qe(mg/g) D/r2(1/s) MSE*
Propylene @ 30 °C
Analytical (spherical) 119.13 2.95E-05 17 Numerical (spherical) 119.75 6.71E-05 15
Analytical (planer) 118.66 1.03E-04 19
Analytical (planer, Olson) 120.00 4.70E-04 26
Propylene @ 80 °C
Analytical (spherical) 88.61 8.10E-05 3.34 Numerical (spherical) 89.16 1.73E-04 2.19
Analytical (planer) 88.07 2.52E-04 7.08 Analytical (planer, Olson) 90.00 1.50E-03 42.3
Propane @ 80 °C
Analytical (spherical) 116.45 6.07E-08 0.211 Numerical (spherical) 108.77 3.45E-08 0.203
Analytical (planer) 90.55 2.13E-07 0.132 Analytical (planer, Olson) 90.00 7.60E-07 0.588
* Mean Square Error.
Figure 4.7 shows the uptake of propane at 80 ˚C and our predictions from the three
models. The MSEs for all three models are in the range of 0.1-0.2, but eq values vary
significantly from 90 to 116 mg/g, as presented in Table 4-3. This variation is due to the
fact that the uptake of propane versus t is nearly linear, which implies that it may be
difficult to estimate eq reliably. eq value simulated based on MC is 103.4 mg/g, which is
in good agreement with the value obtained from micropore diffusion model including
concentration dependence of micropore diffusivity (108.7 mg/g).
95
Figure 4.7: Experimental and simulated uptake data for propane in SiCHA at 80 ˚C and 600 Torr.
The micropore diffusion model with concentration-dependent diffusivity is the most
appropriate description for the uptake of propylene and propane on SiCHA. For propane,
this study obtains eq = 108.7 mg/g at 80 ˚C and 600 torr. Using this with sq = 125.2 mg/g
estimated before, we got b = 0.011 /torr for the Langmuir isotherm. Thus, the
dimensionless Henry's constants for propylene and propane at 80 ˚C were 379.7 and
999.3, respectively. This is consistent with what has been observed with ethane and
ethylene on SiCHA (Olson et al., 2004). The equilibrium data for propylene in SiCHA at
80˚C calculated using these Langmuir isotherm parameters are compared with the
predictions from MC simulation in Figure 4.2.
0
4
8
12
16
20
0 5 10 15 20 25 30
q(m
g/g)
t0.5(min0.5)
Propane 80˚C
Expeimental
Analytical (spherical)
Numerical (spherical)
Olson's Results (planer)
Crank Solution
pr
te
pr
th
pr
et
v
Δ
th
Δ
k
al
4
th
re
se
se
The last
ropane, whi
emperature.
ropane. The
his work ob
ropane. The
thane versus
ersus olefin
33.08HΔ = −
he ratio of Δ
33.08HΔ = −
The abov
inetic prope
llows us to c
Kineti4.2
Majumda
he effective
ealistic repr
electivity. T
electivity sho
important p
ich is norm
Olson et al.
erefore, this
btains HΔ =
e higher valu
s ethylene, a
ns with SiC
kJ/mol for
HΔ values o
30.94*
29.16
− = −−
ve discussion
erties of prop
compute ads
ic and Equ
ar et al. (20
kinetic sele
resentation
The effectiv
own in Eq. 1
parameter fo
mally compu
(2004) repo
study again
29.16= − kJ
ue for propa
and is due to
CHA. Instead
propylene o
obtained from
35.5− kJ/mo
n achieves o
pane and pr
orption selec
uilibrium
11) showed
ectivity give
of an ads
ve kinetic s
1.4.
96
or adsorption
uted from t
orted HΔ = −
n employs M
J/mol for pr
ane is again
o the higher
d of using
obtained in o
m MC simu
ol for propan
ur complete
opylene, wh
ctivities as f
Selectivity
that for a k
en by Eq. 1
orbent’s se
selectivity c
n is the iso
the variation
33.5− kJ/mo
MC simulatio
ropylene an
n consistent
r van der W
these value
our reanalys
ulations as th
ne.
characteriza
hich is need
follows.
y
kinetically co
.3 is time-d
eparation po
can be simp
steric heat o
n of 0b b=
ol for propyl
ons. From th
nd 3HΔ = −
with what
Waals interact
es as is, th
sis discussed
he scaling f
ation of both
ded for PSA
ontrolled ad
dependent, a
otential com
plified to t
of adsorptio
exp( )H
RT
Δ−
lene, but non
hese simulat
30.94 kJ/mo
is observed
tions of para
is work cho
d earlier and
factor to com
h equilibrium
simulations
dsorption pro
and gives a
mpared to
the ideal ki
on for
with
ne for
tions,
ol for
with
affins
ooses
d uses
mpute
m and
s, and
ocess,
more
ideal
inetic
97
Using the independent unary equilibrium and kinetic parameters estimated before,
this study computs the selectivities for propylene/propane in SiCHA. Figure 4.8 a and b
show the time-dependent effective selectivities of propylene over propane in SiCHA and
4A zeolite according to Eq. 1.3. These figures show that the selectivity passes through a
maximum at a short contact time, and then it gradually reaches the equilibrium selectivity
limit. The maximum effective selectivity of propylene over propane in SiCHA is 32, the
equilibrium selectivity (Eq. 1.2) is 0.4, and the ideal kinetic selectivity is 28 at 80 ˚C.
Even though the equilibrium selectivity is lower than unity, the kinetic selectivity seems
sufficient for a kinetically selective PSA process, and in fact can be increased
significantly by lowering the temperature. As may be seen from
Table 4-1, the alumina rich zeolites exhibit higher equilibrium selectivity for
propylene/propane than pure SiCHA. Due to the electrostatic forces arising from the
exchangeable cations, the olefins are adsorbed more strongly than the corresponding
paraffins. The maximum effective selectivity of propylene over propane in 4A is 190, the
equilibrium selectivity is 12.43, and the ideal kinetic selectivity is 223 at 80 ˚C.
F
4
w
igure 4.8: EffkPa and i
PVSA4.3
This wor
with feed, (2)
fective kineticn (b) 4A at 35
A Process M
rk begins its
) high-pressu
0
40
80
120
160
200
0
Sel
ecti
vity
0
5
10
15
20
25
30
35
Sel
ecti
vity
c selectivity o53 K and 10 k
Model
s study with
ure adsorpti
0 50 10
0 50
98
of Propylene okPa. The sele
a 5-step PV
on with feed
00 150 20
t0.5(
100
t0.5(
over propane ectivity at t =
VSA cycle c
d, (3) cocurr
00 250 30
s0.5)
150
(s0.5)
in (a) SiCHA0 is a small n
comprising (
rent rinse w
00 350 40
(b)
200 25
(a)
A at 353 K annonzero value
(1) pressuriz
with the propy
00
50
nd 266 e.
zation
ylene
99
product from Step 5, (4) cocurrent blowdown to intermediate pressure, and (5)
countercurrent evacuation. In this cycle, propane is collected in steps 2 through 4, and
propylene in step 5. However, our simulations revealed that propane passes through the
bed virtually unadsorbed due to its low diffusivity. Thus, steps 2 and 3 deliver most of the
propane, and step 4 gives little propane. In other words, step 4 essentially produces
propylene, and thus has the same role as step 5. Clearly, step 4 in this 5-step cycle seems
redundant, and can be eliminated. Its elimination does not compromise recoveries,
because product purity specifications automatically fix product recoveries in a binary
separation when there are only two useful products and no waste stream. This is evident
from the following equations obtained via simple mass balance.
2 1 2 1 1 21
1 2 1 2 1
(1 ) 1
(1 )(1 )
z Pu Pu z Pu PuRe
Pu Pu Pu Pu z
− −= ×− − −
(4.11)
1 2 1 2 1 22
1 2 1 2 2
(1 ) 1
(1 )(1 )
z Pu Pu z Pu PuRe
Pu Pu Pu Pu z
− −= ×− − −
(4.12)
where, iRe , iz and iPu are recovery, feed mol fraction, and purity of component i .
Moreover, the 4-step cycle should also consume less energy than the 5-step cycle. Thus,
this work eliminates step 4 and study a 4-step PVSA cycle for SiCHA (Figure 4.9).
In evacuation step, propylene product is collected in a tank and a part of this product
is recycled to the rinse step as heavy reflux. Since rinse duration and amount of gas that is
recycled to this step are known, the rinse velocity can be calculated using the following
equation:
g gtank
rinserinse H
R TGMv
t A Pε= (4.13)
100
where G is reflux ratio and tankM is the molar amount of product collected in the tank, rinset
is rinse duration, A is column cross section area, gR is the universal gas constant, ε is the
bed porosity, gT is gas temperature, and HP is the operating pressure of the rinse step.
To simulate this process, this work uses the following isobaric and non-isothermal
model based on intra-particle micropore diffusion with concentration-dependent
diffusivity (Khalighi et al., 2012).
Figure 4.9: Schematic diagram of the PSA cycle. 1) Pressurization 2) high-pressure adsorption 3) rinse 4) countercurrent evacuation.
Propane
Feed Propylene
1 3 4 2
PH
PL
1 3 4
Time
2
4
T
T
h
G
M
Model4.4
This study as
1- The id
2- The sy
3- Axiall
4- The ad
5- The
compo
6- The m
7- The ch
8- Temp
neglig
9- The g
10- Lump
wall a
With the
The (+) sign
olds for the
Gas phase ma
2
LDz
∂−∂
Mass transfer
piq
tε
∂=
∂
l Equation
sumes the fo
deal gas law
ystem is isob
ly dispersed
dsorbent con
extended L
onent param
macropore re
hemical pote
erature grad
gible.
as and adsor
ped coefficie
and that betw
ese assumpti
holds for th
countercurre
ass balance f
2i ic vc
z z
∂ ∂± +∂ ∂
r into macrop
(1ip
c
tε ∂ + −
∂
ns
ollowing:
applies.
baric.
plug flow m
nsists of unif
Langmuir
meters describ
esistance is n
ential gradie
dients along
rbent particle
ents account
ween the colu
ions, the fol
he cocurrent
ent flow from
for compone
1(ic
t
εε
∂∂ −+∂
pores:
) cip c
q
tε ρ ∂
∂
101
model describ
form microp
isotherm u
bes the mixtu
negligible.
ent is the driv
g the radii
es are in ther
t for the hea
umn wall an
llowing equa
t flow from
m z L= to z
ent i ( A = p
) 0piq
t=
∂
bes the flow
porous crysta
using indep
ture equilibri
ving force fo
of the col
rmal equilib
at transfer b
nd external su
ations descr
m 0z = to z
0z = .
propylene, B
w pattern.
als.
pendently e
ium.
or micropore
lumn and m
brium everyw
between the
urroundings
ribe the 4-st
L= (colum
B = propane)
estimated s
e diffusion.
microparticle
where in the
bed and co
.
tep PSA pro
n length) an
):
(4.14)
(4.15)
single
e are
bed.
olumn
ocess.
nd (-)
102
Mass transfer into micropores:
3
c
ci cii
r rc
q qD
t r r =
∂ ∂=∂ ∂
(4.16)
Overall mass balance for steps 2 and 3:
10pi
i
qCv
Z t
εε
∂∂ −± + =∂ ∂ (4.17)
Overall mass balance for steps 1 and 4:
10pi
i
qCv C
Z t t
εε
∂∂ ∂ −± + + =∂ ∂ ∂
(4.18)
In above equations, iC c= is the total gas phase concentration, ii
g g
Pyc R T= is
the concentration of component i in the bulk gas phase, P is the total gas pressure, iy
is the mole fraction of component i in the bulk gas phase, gT is the gas temperature in
the thermal equilibrium with the adsorbed phase, v is the interstitial velocity, LD is the
axial dispersion coefficient, piq is the average adsorbed concentration of component i
per unit adsorbent particle volume, ciq is its average adsorbed concentration per unit
crystal mass, ciq is the local adsorbed concentration per unit crystal volume along the
crystal radius, cρ is the crystal density, pε is the adsorbent particle porosity, and t is the
adsorption time. Note that v is computed from Eq. 4.17 or 4.18. The boundary conditions
for Eqs.4.14, 4.17 and 4.18 vary with each step in the PSA cycle, and are discussed later.
Gas phase energy balance:
103
2
2
(1 ) ( )g g pgpa pi s ps
i g
T T c vPc q c
t z R z
ε λρε ε
∂ ∂ − ∂ + = − ∂ ∂ ∂
1
2(1 )(( ) ) ( )
npg ci w
i pa g c g wig w
c q hPH c T T T
R t t R
ε ρε ε=
∂∂ −− + −Δ − − −∂ ∂
(4.19)
where pac is the molar specific heat capacity of the adsorbed gas that this study assumes
it has the same value as molar specific heat capacity of the gas mixture, psc and pgc are
molar specific heat capacity of the adsorbent and molar specific heat capacity of the gas
mixture, respectively. λ is the axial heat dispersion calculated from the correlation by
Wakao (1978). – iHΔ is the isosteric heat of adsorption for component i , wT is the wall
temperature, wh is the film heat transfer coefficient between the adsorption bed and the
column wall, wR is the column (inside) radius.
Wall heat balance:
2
02( ) ( )w w
w pw w wi w g w wo w
T Tc K h T T h T T
t zρ α α ∞
∂ ∂= + − − −∂ ∂
(4.20a)
2
(2 )
wwi
w
R
e R eα =
+ (4.20b)
2( )
(2 )w
wow
R e
e R eα +=
+ (4.20c)
where pwc and wρ are the specific heat and density of the column wall, respectively. wiα
is the ratio of the internal surface area to the volume of the column wall, e is the wall
thickness, woα is the ratio of the external surface area to the volume of the column wall,
104
0h is the convection heat transfer coefficient between wall and surrounding, wK is the
wall conduction heat transfer coefficient, and T∞ is the constant ambient temperature.
Pressure change with time in steps 1 is represented by (Farooq et al., 1993) :
( ) [ ]1expH H LP P P P a t= − − − (4.21)
Pressure change with time in steps 4(Farooq et al., 1993):
( ) [ ]2expL H LP P P P a t= + − − (4.22)
where LP and HP are the low and high pressures in the pressurization and evacuation
steps. The constants in Eqs. 22 and 23, 1a and 2a , are assumed 0.15 and 0.05/s,
respectively.
Boundary conditions for steps 1, 2, and 3:
2
2 0 0 00
( )il i iz z z
z
cD v c c
z−= = =
=
∂ = − −∂
(4.23a)
0i
z L
c
z =
∂ =∂
(4.23b)
0 0 0
( )gpg g gz z z
TCc v T T
z
λε −= = =
∂= − −
∂ (4.23c)
0g
z L
T
z =
∂=
∂ (4.23d)
0 ( )ww w
z L
TK h T T
zβ ∞
=
∂− = −
∂ (4.23e)
2( )
( 2 )w
w
e R
e e Rβ +
=+
(4.23f)
0
0i i
z z L
y y
z z= =
∂ ∂= =∂ ∂
(4.23g)
Boundary conditions for steps 4:
105
0
0i i
z z L
c c
z z= =
∂ ∂= =∂ ∂
(4.24ab)
0
0g g
z z L
T T
z z= =
∂ ∂= =
∂ ∂ (4.24cd)
0
0
( )w ww w w
z z L
T TK K h T T
z zβ ∞
= =
∂ ∂= − = −
∂ ∂ (4.24ef)
Boundary conditions for velocity:
0z L
v=
= for step 1
(4.25a)
00zv v
== for step 2 (4.25b)
0 rinsez
v v=
= for step 3 (4.25c)
0z L
v=
= for step 4 (4.25d)
Mass balance for micro-particles:
22
1ci cii
q qr D
t r r r
∂ ∂ ∂ = ∂ ∂ ∂ (4.26)
Micro-particles boundary conditions:
0
0i
r
q
r =
∂ =∂
(4.27a)
1
c c
c
i i ir R r Ri
is i i r Ri
q b c
q b cθ= =
=
= =+
(4.27b)
This study uses the following to describe the concentration-dependence of diffusivity in a
Langmuir system with constant intrinsic mobilities (Kärger and Bülow, 1975):
0 (1 )1
A BA B A
A B A
D q rD
q rθ θ
θ θ ∂ ∂= − + − − ∂ ∂
(4.28)
106
0 (1 )1
B AB A B
A B B
D q rD
q rθ θ
θ θ ∂ ∂= − + − − ∂ ∂
(4.29)
To compute purity (%) and energy consumption (W kWh/tone propylene), we use:
3 6
3 6 3 8
001
0 00 0
100Evacuation
Evacuation Evacuation
t
C H z
t t
C H C Hz z
C v dtPu
C v dt C v dt
=
= =
=+
(4.30)
3 8 3 80 00 0
2
0 00 0
100Adsorbtion Rinse
Adsorbtion Rinse
t t
C H C Hz z
t t
total totalz z
C v dt C v dtPu
C v dt C v dt
= =
= =
+ =+
(4.31)
1
01
(1 )
it outin
in
PW F P dt
P
γγγ
η γ
− = × − −
(4.32)
where, component 1 is propylene, 2 is propane, η is the compression efficiency and γ is
the adiabatic constant. F is the total gas flow through the compressor or vacuum pump,
and inP and outP are the inlet and outlet pressures.
For recoveries, Eqs. 4.11 and 4.12 are used.
The total energy consumption for the separation is given by:
1
(kWh per tonne of propylene)3.6 42
compressor vacuum pumptotal
W WW
F
+=
× × (4.33)
where, 1F is the amount of propylene produced in the evacuation step. As shown in
Figure 4.9, the pressurization and adsorption steps may require compressors, if the high-
pressure in the PVSA process exceeds the feed stream pressure, which is 2-3 atm in
practice. Since the rinse occurs at the high pressure, a compressor is required during the
rinse to increase the pressure of the heavy reflux from the evacuation step. Finally, a
vacuum pump is needed for the evacuation step. Thus, if the high-pressure in a PVSA
cy
fo
4
G
d
fi
th
re
re
b
b
P
m
co
ex
st
th
re
co
to
re
ycle does no
or the rinse,
Nume4.5
Dell Opt
GB of RAM
imensionles
inite element
he axial dir
epresenting
espectively.
ed is helpful
ed; hence, it
DEs that be
microparticle
orrespond to
xported as s
tructures seq
he bed at va
ecorded fina
ontinuously
o ensure pro
each the cyc
ot exceed the
and the othe
erical Simu
tiplex 780 w
M is used fo
s form and
t method. Tw
rection of th
the axial
Using a squ
l in that it o
t simplifies t
elong to bu
profiles are
o pressuriza
structures to
quentially. T
acuum cond
al condition
executed un
files as clos
lic steady st
e feed pressu
er for the eva
ulation
with Intel® C
or numerica
solved usin
wo geometri
he bed, and
direction of
uare to repre
bviates the n
the problem
lk profiles a
e set up in
ation, adsorp
o MATLAB
The initial pr
dition to star
from the pre
ntil the purity
e as possible
tate (CSS) re
107
ure, then the
acuation.
Core 2 Quad
al simulation
ng COMSO
ies are used:
d the other
f the bed
esent variab
need to simu
m and reduce
are set up i
the square
ption, rinse
B that execu
essurization
rt the cyclin
evious step a
y difference
e to the true
equiring 3-4
work in Eq.
d CPU Q940
n. The mod
OL Multiphy
: one is a lin
is a square
and radial
ble profiles w
ulate a full s
es the level o
in the line
geometry. F
and evacu
utes the cyc
n structure ca
ng whereas t
as initial con
in five cycle
e steady state
h of CPU ti
. 4.32 involv
00 @ 2.66 G
del equation
ysics softwa
ne of unit len
e of unit le
direction o
within the p
sphere at eac
of computati
geometry w
Four COMS
uation steps,
cling codes
arries the ini
the other st
nditions. Th
es dwindles
e. 30-40 cyc
ime. 50-60 c
ves two parts
GHz Process
ns are writte
are that emp
ngth, represe
ength and w
of the part
particle alon
ch position i
ion required
whereas thos
SOL files, w
, are solved
and solves
itial conditio
ructures tak
he cycling lo
to less than
cles are need
cycles are ne
s: one
sor, 8
en in
ploys
enting
width,
ticles,
ng the
in the
d. The
se for
which
d and
these
ons of
ke the
oop is
0.1%
ded to
eeded
to
η
ra
4
T
so
3
sh
pr
is
pr
ad
in
ar
pr
th
d
in
F
th
o reach the c
0.72η = as
ation ( pC C
PVSA4.6
As menti
This study ap
ources are a
33-353 K b
hows that it
ropylene and
s selected a
ressurization
dsorption tim
n our study.
round its va
ropylene an
he range for
escribes the
n the direct
igure 4.10c
he same curv
cyclic steady
compressor
vC ).
A Process P
ioned before
pproximates
at about 600-
before enteri
t is possible
d 90% propa
nd a param
n time, leng
me, rinse tim
Each param
alue at the b
d propane. T
r each para
change in p
tion of its
describes th
ves for the 8
y state (CSS
efficiency (
Performan
e, two main
them as 50
-800 K and
ing the PVS
e to design
ane. To this
metric study
gth to veloc
me, evacuati
meter is var
base point,
Table 4-4 lis
ameter. Ever
purity-recove
arrow. Fig
he changes in
5/15 propyle
108
S) requiring
Van Ness et
nce
sources of p
0/50 and 85/
2-3 atm, it i
SA process
a PVSA pr
end, a base
is performe
city ratio, f
ion time, and
ried one at
and study th
sts the param
ry curve pa
ery of propy
gure 4.10b
n the purities
ene/propane
3-4 h of CP
t al., 2001) a
propylene/pro
/15 propylen
is assumed t
to increase
rocess based
point (A in
ed for each
feed gas pre
d temperatu
a time, whi
he effects o
meter values
assing throug
ylene, as one
describes t
s of both com
feed.
PU time. Th
and 1.4γ =
opane feeds
ne/propane.
that both fee
selectivity.
d on SiCHA
Figure 4.10
h feed. This
essure, evac
ure as the pr
ile keeping
on purities a
s for the bas
gh point A
e specific par
the same fo
mponents. F
his work ass
as heat cap
exist in pra
While both
eds are cool
This work
A to obtain
and Figure
study cons
cuation pres
rocess param
the others f
and recoveri
se points (A
in Figure 4
rameter incr
for propane,
Figure 4.11 s
sumes
pacity
actice.
h feed
led to
now
99%
4.11)
siders
ssure,
meters
fixed,
ies of
A) and
4.10a
reases
, and
shows
109
83
85
87
89
91
93
97 97.5 98 98.5 99 99.5 100
%R
ecov
ery
%Purity
(a)
V0 L tpr tad
tri tev PH PL
T G
97
97.5
98
98.5
99
99.5
100
86 87 88 89 90 91 92 93
%R
ecov
ery
%Purity
(b)
V0 tad tpr
tad tri tev
PH PL T
G
Point A
Point A
110
Figure 4.10: Recovery vs. purity plots show the effects of different parameters on the performance of a PVSA process. The arrows indicate the increasing directions of operating parameters. a) propylene b) propane and c) propane purity vs. propylene purity for the feed composition of 50/50 propylene/propane. Each parameter increases in the direction of arrow. Table 4 shows the range of the parameters.
85
86
87
88
89
90
91
92
93
94
96.5 97 97.5 98 98.5 99 99.5 100
% P
ropa
ne p
urity
% Propylene purity
(c)
V0 L tpr tad
tri tev PH PL
T G
Point A
Point 1
111
96
96.5
97
97.5
98
98.5
99
99.5
98 98.5 99 99.5 100
%R
ecov
ery
%Purity
(a)
V0 L tpr
tad tri tev
PH PL T
G
88
90
92
94
96
98
100
83 84 85 86 87 88 89 90 91 92 93
%R
ecov
ery
%Purity
(b)
V0 L tpr
tad tri tev
PH PL T
G
Point A
Point A
112
Figure 4.11: Recovery vs. purity plots show the effects of different parameters on the performance of a PVSA process. The arrows indicate the increasing directions of operating parameters. a) propylene, b) propane, and c) propane purity vs. propylene purity for the feed composition of 85/15 propylene/propane. Each parameter increases in the direction of arrow. Table 4 shows the range of the parameters.
83
85
87
89
91
93
95
98 98.5 99 99.5 100
% P
ropa
ne p
urity
% Propylene purity
(c)
V0 L
tpr tad
tri tev
PH PL
T G
Point 6
Point 4
Point A
Point 5
v
O
ar
d
w
th
pr
S
Table 4-4and thei
Op
Energy (kPro
PropPro
Prop
4.6
For a fix
elocity. As t
On the other
rgue that as
ecreases. Th
which contam
he high-pres
ropylene dec
horter resid
: The PVSAr comparison
perating Para
v0 (cm/s)L (cm) tpr (s) tad (s) tri (s) tev (s)
PH (kPa)PL (kPa)
G T (K)
kWh/ tonne opane Recovpylene Recovpane Purity
pylene Purity
Effect o6.1
xed column
the velocity
hand, propy
the feed vel
his allows m
minates the p
sure adsorpt
creases, but
dence time
A operating pn for the 50/
ameter
)
) )
of propyleneery (%) very (%) (mol%)
y (mol%)
of Length to
length, a d
increases, p
ylene purity
locity increa
more propyle
propane pro
tion step, wh
purity increa
restricts pro
116
parameters, th/50 and 85/1
Po50/
e)
o Feed Velo
decrease in
propane reco
increases, b
ases, the resi
ene to exit d
oduct. Meanw
hich increase
ases with the
opane adsor
heir ranges u5 propylene
oint A /50 Feed 14.2 75
310 350 60
556 110 4.5 0.5 353 88
99.4 88.2 89.4 99.4
ocity Ratio
0/L v ratio
overy increa
but its recove
idence time
during the hi
while, more
es its recover
e decreasing
rption durin
used in the pe/propane fee
Point A 85/15 Feed
6.5 125 330 420 50
555 150 14.9 0.5 353 53
96.0 98.0 89.5 99.3
means an
ases, but its
ery decrease
for adsorpti
igh-pressure
e propane is
ry. The
g length to fe
ng pressuriz
parametric sted mixtures.
d StudRang1-15
50-15101-6101-6101-6101-6101-3
1-500.3-0
333-3-----
increase in
purity decre
es. This work
on in the co
e adsorption
collected d
recovery
eed velocity
zation and
tudy,
dy ge 5 50
600 600 600 600
00 0
0.7 53
inlet
eases.
k can
olumn
step,
during
of
ratio.
high-
pr
w
to
p
th
m
d
w
br
le
pr
pr
T
T
an
ressure adso
which increas
o travel to th
4.6
Changing
erformance
he feed flo
microparticle
epends on t
what is neces
4.6
Propane
reak-through
ength of thi
ropylene rec
ropylene fro
Thus, more p
This also mea
nd hence its
orption. Hen
ses propylen
he product en
Effect o6.2
g the press
for both pro
ow into the
s. In the si
the value of
sary to reach
Effect o6.3
purity decre
h of propyle
is step incre
covery from
ont penetrate
ropane prod
ans that less
purity incre
nce, there is
ne purity. Ho
nd, which red
of Pressuriz
surization ti
oducts. In th
e column i
mulations, t
f 1a . Therefo
h HP does no
of High Pre
eases, as high
ene from the
eases, which
m the evacua
es deeper in
duct is collec
s propane is
ases.
116
s less co-ad
owever, the
duces propan
zation Time
ime does n
his step, afte
s determine
the time req
ore, for a ch
ot affect pur
essure Adsor
h-pressure a
e column. M
h contamina
ation step. H
nto the adsor
cted in the ad
left in the b
dsorbed prop
shorter bed
ne purity an
e
not have m
er the colum
ed by the
quired to re
hosen 1a , pre
rity and reco
rption Time
adsorption tim
More propyl
ates the pro
However, by
rbent leavin
dsorption ste
bed to contam
pane in the
d also causes
nd propylene
much effect
mn reaches th
rate of di
each the hig
essurization
overy.
e
me increases
ene exits th
opane produc
y increasing
ng less capa
ep, and its re
minate the p
evacuation
s more propy
recovery.
on the pr
he high pres
ffusion into
gh pressure
time longer
s. This sugge
he column, a
ct and decr
adsorption
city for pro
covery incre
propylene pro
step,
ylene
ocess
ssure,
o the
( HP )
r than
ests a
as the
reases
time,
pane.
eases.
oduct
d
pu
th
is
pu
n
co
H
d
d
co
A
ev
pr
ad
pu
re
4.6
By incre
ecrease. Ho
ushes more
he evacuatio
s necessary t
urity in the
atural gas u
omponent r
However, in
istillation op
esorption o
ontaminate t
4.6
As evacu
At the same t
vacuation in
ropylene rec
dsorption of
urity. Howe
ecovery.
Effect o6.4
easing the
wever, prop
propane out
n product is
to sufficient
subsequent e
upgrading, p
reflux in dis
this work,
peration. In
f the slowe
the propylen
Effect o6.5
uation time i
time, propan
ncreases, bot
covery but d
f propylene
ever, at the
of Rinse Tim
rinse step
pylene purity
t from the be
used to exe
ly remove th
evacuation s
purging with
stillation or
the rinse s
this step, th
er compone
ne product in
of Evacuati
increases, pr
ne purity incr
th componen
decreases its
in the high
same time,
116
me
duration, pr
y and propa
ed thus incre
ecute a cocur
he propane f
step. As Bha
h the slower
r a similar
step is analo
he faster diff
nt (propane
n the evacuat
on Time
ropylene pur
reases, but it
nts desorb m
s purity. A c
h-pressure a
, more prop
ropane puri
ane recovery
easing its rec
rrent rinse st
from the bed
adra and Far
r componen
countercurre
ogous to he
fusing comp
e) from the
tion step.
rity decrease
ts recovery
more from th
cleaner adso
adsorption st
pane is co-a
ity and pro
y increase. L
covery. In th
tep at high p
d and increa
rooq (2011)
nt is analog
ent mass-tra
eavy-compo
ponent (prop
e bed that w
es, but its re
decreases. A
he adsorbent
orbed phase
tep, thus inc
adsorbed, w
opylene reco
Longer rinse
his step, a p
pressure. The
ase the propy
explained fo
ous to the
ansfer opera
onent reflux
pylene) facil
would other
covery incre
As the durati
, which incr
allows incre
creasing pro
which reduce
overy
e step
art of
e step
ylene
or the
light-
ation.
in a
litates
rwise
eases.
ion of
reases
eased
opane
es its
co
in
w
th
fr
re
in
o
th
re
th
ev
re
T
w
4.6
If the pre
onditions co
nput flow to
when the hig
he micropore
ront penetra
eason, less p
ncreases pro
ccupying the
he evacuatio
4.6
Decreasin
ecovery. Sin
he accompa
vacuation tim
emoval of pr
The result is a
with the prop
Effect o6.6
essure of the
onstant, incl
the bed incr
h adsorption
e transport b
ates deeper i
propylene is
opylene reco
e last portion
n step. Ther
Effect o6.7
ng the evac
nce more pro
anying prop
mes, the cle
ropylene from
an increase i
ylene produ
of Adsorpti
e adsorption
luding the f
rease. The pr
n pressure is
becomes fast
into the bed
s lost in the
overy and pr
n of the colu
efore, propy
of Evacuati
cuation pres
opylene deso
pane desorpt
aner adsorbe
m the feed a
in propane p
uct.
116
on Pressure
n step is incr
feed velocity
ropylene pur
s increased.
ter due to inc
d, and becom
e high-press
ropane purity
umn, more p
ylene purity d
on Pressure
ssure increas
orbs from th
tion decrea
ent due to d
along with a
purity, but a d
e
reased while
y, then the
rity decrease
Since equili
creased conc
mes sharper
sure adsorpti
y. In additio
propane can
decreases.
e
ses both pr
he bed, its r
ases propyle
deeper evacu
small increa
decrease in r
e holding al
capacity of
es, but its re
ibrium isoth
centration. T
r at the sam
ion and rins
on, because
be adsorbed
ropane purit
recovery incr
ene purity.
uation also le
ase in propan
recovery, as
ll other oper
f the column
covery incre
herm is favor
The mass tra
me time. For
se steps and
propylene i
d and desorb
ty and propy
reases. How
As with lo
eads to incre
ne co-adsorp
propane is g
rating
n and
eases,
rable,
ansfer
r this
d this
is not
bed in
ylene
wever,
onger
eased
ption.
going
pu
fr
o
pu
d
d
k
d
pr
th
d
it
pr
en
th
co
1
4.6
As expe
urity, but de
rom the colu
f propylene
urity. Thus,
4.6
Propylen
ecreases. A
ecrease, but
inetic select
ecreases its
In Figure
ropylene an
hese zones
esired puriti
t can be conc
ropylene and
nergy consu
he points c
onsumptions
02) kWh/ton
Effect o6.8
ected, an inc
ecreases its r
umn, so prop
e breaking t
increased pr
Effect o6.9
ne purity inc
s temperatu
the diffusiv
tivity, and t
adsorbed am
e 4.10c and
d propane p
represent S
ies. Clearly,
cluded that S
d propane. T
umptions for
correspondin
s of these p
nne of propy
of Reflux ra
crease in the
recovery. In
pane recover
through with
ropane recov
of Tempera
creases as th
ure decrease
vity ratio of p
hus the pur
mount and he
Figure 4.11c
purities of ≥
iCHA-based
several oper
SiCHA can s
Table 4-5 lis
three such p
ng to the t
rocesses for
ylene. These
116
atio
e reflux rati
ncreased recy
ry increases.
h the propa
very is at the
ature
he process
s, the diffus
propylene ov
rity. Howeve
ence its recov
c, the upper
≥ 99% and ≥
d PVSA pro
rating points
successfully
sts the operat
processes fo
three proces
r the 85/15
significant v
io in the rin
ycled propy
. However, t
ane product
e expense of
temperature
sivities of b
ver propane
er, the redu
very in the e
right quadra
≥ 90%, respe
ocesses that
s exist in the
separate ind
ating parame
or each feed.
sses for th
(50/50) feed
variations su
nse step inc
ylene displac
this also incr
t, hence dec
f its purity.
e decreases,
both propyle
increases. T
uced diffusiv
evacuation st
ants represen
ectively. Op
t successful
ese zones for
dustrial feed
eters, purities
. In Figure 4
he 85/15 fe
d vary from
uggest that m
creases propy
ces more pro
reases the ch
creasing pro
but its reco
ene and pro
This increase
vity of propy
tep.
nt the zones
perating poin
ly achieve
r both feeds
s into high-p
s, recoveries
4.11c, it is sh
eed. The en
m 42 to 89 (
much room e
ylene
opane
hance
opane
overy
opane
es the
ylene
s with
nts in
these
, thus
purity
s, and
hown
nergy
64 to
exists
116
for optimization. Such optimization, however, is non-trivial due to the multitude of
parametric possibilities and complexity of their interactions. This rigorous optimization
will be addressed in next Chapter.
Table 4-5: The PVSA operating parameters of six points with desired product purities, where points 1-3 are for the 50/50 and points 4-6 are for the 85/15 propylene/propane feed mixture.
Operating Parameter Point 1 Point 2 Point 3 Point 4 Point 5 Point 6 v0 (cm/s) 14.2 14.2 15.2 6.5 6.5 6.5 L (cm) 75 84 84 125 125 125 tpr (s) 310 310 310 330 330 330 tad (s) 325 350 360 420 420 420 tri (s) 60 74 60 50 40 50 tev (s) 556 556 556 555 555 675
PH (kPa) 110 110 110 150 150 150 PL (kPa) 4.5 4.5 4.3 15.1 14.9 14.9
G 0.5 0.5 0.5 0.5 0.5 0.5 T (K) 333 333 333 333 333 333
Energy (kWh/tonne propylene) 64 83 102 42 63 89 Propane Recovery (%) 99.02 99.18 99.23 94.01 94.5 94.94
Propylene Recovery (%) 89.05 89.3 89.23 98.03 98.23 98.2 Propane Purity (mol%) 90.04 90.31 90.21 90.05 90.40 90.45
Propylene Purity (mol%) 99.01 99.10 99.15 99.0 99.02 99.1
From Table 4-5, this study can also compare the parameters and energy requirements
for the two feeds. To be accurate, the two feeds can be compared only if first the
minimum energy process is found for each feed via rigorous optimization. Since this
optimization has not been done, the processes with the least energies from the three that
is listed for each feed in Table 4-5 are compared. These are point 1 for the 50/50 feed,
and point 4 for the 85/15 feed. Point 1 is the origin in Figure 8c and point 4 is the origin
in Figure 9c. They attain the minimum desired purity targets.
116
While the pressurization times and evacuation times are quite similar for points 1 and
4, but the velocities, bed lengths, adsorption times, rinse times, low-pressure levels, and
high-pressure levels are quite different. Since the 85/15 feed has more propylene, it needs
lower velocity, longer bed length, and higher adsorption pressures, so that propylene has
sufficient time and driving force to adsorb in the column. Similarly, it needs shorter rinse
time, because it has less propane.
All energy needs for this separation are normalized as kWh per tonne of propylene
that exits the column in the evacuation step. The 50/50 feed requires more energy (64 vs.
42 kWh/tonne) than the 85/15 feed. The feed pressure of 2 atm (202.6 kPa) exceeds the
PVSA high-pressure levels, so rinse and evacuation are the main contributors to energy
consumption. Interestingly, both feeds need roughly 30% (20/64 for the 50/50 feed and
13/42 for the 85/15) of the total energy for the rinse and 70% (44/64 for the 50/50 feed
and 29/42 for the 85/15 feed) for the evacuation. This study can rationalize the higher
energies for the 50/50 feed as follows. First, the high-to-low pressure ratio for the 50/50
feed is 110/101~ 1.09 compared to 150/101 ~ 1.5 for the 85/15 feed. Since the reflux
ratio is the same for both, the energy for the rinse step is higher for the 50/50 feed.
Because the evacuation times for the two feeds are not very different, and the low-
pressure for the 50/50 feed in the evacuation step is lower (4.5 kPa vs. 15.1 kPa), the
50/50 feed needs more evacuation energy. Thus, the higher pressure ratio seems to be the
main reason for the higher energy consumption for the 50/50 feed.
4
pu
ad
fo
re
an
pr
in
st
pr
m
an
se
p
ad
th
en
Chapt4.7
A simple
urity produc
dsorbent, S
orward rinse
elevant feed
nd 90 mo
ropylene/pro
SiCHA
nformation f
tudy on SiCH
rocess using
molecular sim
nd compare
eparation of
arametric stu
dsorption tim
hose for the
nergy-intens
ter conclu
e 4-step PVS
cts from a
SiCHA. The
e, and revers
d mixtures o
ol% propan
opane separa
is a relativ
for propylene
HA, this wo
g the limited
mulation. Rig
the potentia
f two comm
udies sugge
me, longer b
50/50 feed.
sive than tha
sion
SA cycle has
propylene/p
e cycle, in
se evacuatio
f 50/50 and
ne. This
ation and ma
vely new a
e/propane se
ork has also
d experimen
gorous optim
als of variou
mon propyle
st that the P
bed length, l
However, t
at of the riche
116
been propo
propane mix
nvolving pr
on, is able to
d 85/15 prop
suggests th
ay merit furth
adsorbent, a
eparation are
estimated th
ntal uptake d
mization is
us adsorbents
ene/propane
PVSA proces
lower veloc
the separatio
er 85/15 feed
sed and simu
xture using
ressurization
o satisfactor
pylene/propa
hat SiCHA
her study.
and adequa
e not availab
he data nece
data from th
essential to
s, such as Si
e feed mixtu
ss for the 85
city, and hig
on of the lea
d.
ulated for ob
a new kine
n, high-pres
rily separate
ane into 99
A is indee
ate equilibri
ble. Being th
ssary for sim
he literature
o reliably an
iCHA and ze
tures. Howe
5/15 feed ma
gher adsorpti
aner 50/50 fe
btaining two
etically sele
ssure adsorp
e two indust
mol% propy
ed suitable
ium and ki
he first simul
mulating a P
e and approp
nd fully eva
eolite 4A, fo
ever, our lim
ay require h
ion pressure
eed may be
o high
ective
ption,
trially
ylene
e for
inetic
lation
PVSA
priate
aluate
or the
mited
higher
e than
more
116
CHAPTER 5 Comparing SiCHA and 4A Zeolite for
Propylene/Propane Separation using a Surrogate-based
SimOpt Approach
It was mentioned in Chapter 2 that most studies on propylene/propane separation
have not considered producing high purity propylene and propane simultaneously with
low energy consumption. Therefore, significant room exists for improving and
optimizing adsorption-based processes for this separation. In this chapter, we compare
4A zeolite and a new 8-ring silica chabazite zeolite (SiCHA) for separating these
mixtures in a pressure vacuum swing adsorption (PVSA) process. We base our
assessment on a 5-step PVSA cycle with concurrent pressurization, high pressure
adsorption, rinse with the heavy component (i.e., heavy reflux), forward blowdown, and
reverse evacuation, which we simulate rigorously using a non-isothermal isobaric
micropore diffusion model with concentration-dependent diffusivity developed by
Khalighi et al. (2012). We develop fast neuro-fuzzy surrogates for these simulations, and
estimate minimum energy consumptions per tonne of propylene using a genetic algorithm
(GA). We show that the blowdown step, although widely used in the literature for 4A, is
in fact redundant for both 4A and SiCHA. While 4A zeolite requires lower separation
energy per tonne of propylene due to its higher selectivity compared to SiCHA, it allows
lower throughput. However, a comparison based on approximate total annualized cost
also confirms that 4A is superior to SiCHA for this separation. Between the two
industrial propylene/propane feeds of 50:50 and 85:15, the latter requires lower energy
than the former for separating two pure products.
et
fe
o
6
pr
n
C
5
fr
cu
pr
pr
se
en
te
it
se
pr
m
Two indu
t al., 2013).
eedstocks su
f the fluid c
00-800 K, a
ressure of 2
eglected. Th
Chapter 3 and
Introd.1
The s
rom the off-
urrent metho
ropylene/pro
racticed com
eparations a
nergy dema
echnology fo
t has progres
eparation an
In this
ropylene/pro
mol% purity,
ustrially rele
. First is the
uch as naphth
catalytic crac
a low temper
atm for the
he non-isoth
d also used i
duction
separation o
-gas of catal
od for these
opane separ
mmercially (
are highly d
ands. Pressu
or gas separa
ssed much in
d purificatio
s work, we f
opane mixtu
as required
vant propyle
e 50/50 mo
ha, and seco
cking (FCC)
rature of 353
feed streams
hermal micr
n Chapter 4
f light olefi
lytic cracker
e separations
ration as the
Jarvelin and
esirable. Ad
ure/Vacuum
ation. Since c
n size, versat
on, and offer
focus on the
ures into two
d for polypro
116
ene/propane
l/mol mixtu
ond is the 85
) units. Whil
3 K is used
s. Pressure d
ropore diffu
is used for s
ins such as
rs is a key
s involves c
e most ener
d Fair, 1993)
dsorption of
Swing Ad
commercial
tility, and co
rs great flexib
e adsorption-
o high-purity
opylene prod
e feed mixtu
ure from the
5/15 mol/mo
le both sour
here to incr
drops throug
sion model
simulating v
ethylene/eth
step in the
cryogenics.
rgy-intensiv
). Thus, low-
ffers an attra
dsorption (P
inception in
omplexity. It
bility in desi
-based separ
y products.
duction. For
ures are cons
e thermal cr
ol mixture fr
rces are at ab
rease kinetic
gh the adsorp
developed
various PVSA
hane and pr
petrochemic
The US DO
ve single dis
-energy alter
active optio
VSA) is a
n 1950 (Ruth
t can handle
ign and oper
ration of ind
For propyle
propane, we
sidered (Kha
racking of l
rom the off-
bout 2-3 atm
selectivity a
ption column
and validat
A processes
ropylene/pro
cal industry
OE has iden
stillation pr
rnatives for
n due to its
well-establ
hven et al., 1
e multicompo
ration.
dustrially rel
ene, we targ
e target 90 m
alighi
liquid
gases
m and
and a
ns are
ed in
opane
. The
ntified
ocess
these
s low
lished
994),
onent
evant
get 99
mol%
116
purity, as used in engines, oxy-gas torches, barbecues, etc. In a previous work, we
(Khalighi et al., 2013) identified 4A zeolite and SiCHA as the two most promising
adsorbents for this separation from those studied in the literature. They are the two top
candidates, when all reviewed adsorbents are ranked according to the kinetic selectivity.
While 4A zeolite is commercially available and well-studied, SiCHA is not. For 4A
zeolite, Grande and Rodrigues (2005) suggested a 5-step PVSA process with
pressurization, high-pressure adsorption, rinse with propylene product (also called heavy
reflux), cocurrent blowdown, and countercurrent evacuation. Furthermore, Khalighi et al.
(2012) developed a non-isothermal micropore diffusion model with concentration-
dependent diffusivities for kinetically selective PVSA processes, which we validated with
published data (Grande and Rodrigues, 2005) on propylene/propane separation with 4A
zeolite. For SiCHA, Khalighi et al. (2013) demonstrated the power of combining limited
published data (Olson et al., 2004) with molecular simulation estimates to assess process
suitability of a new adsorbent. They showed that a 4-step PVSA cycle with
pressurization, high-pressure adsorption, rinse with propylene product, and
countercurrent evacuation can indeed yield 99% propylene and 90% propane. However,
none of these studies offered a definitive conclusion on the relative merits of 4A zeolite
and SiCHA. In fact, as we discuss later, such a conclusion is not possible without a
detailed optimization of the PVSA processes for these two adsorbents. But, such an
optimization study for propylene/propane separation does not exist in the literature.
Our aim is to compare 4A zeolite and SiCHA for propylene/propane separation
based on extensive simulations and detailed optimization, and identify the best adsorbent
along with its best PVSA process. Our assessment criteria are energy consumption per
to
co
.
as
ca
w
at
u
al
fo
5
tr
ac
pr
co
o
b
ad
du
fi
th
onne of prop
onsider two
First is the 5
s naphtha, a
atalytic crac
we assume a
tm for the fe
se the noniso
l. (2012) alo
or simulating
Optim.2
A PV
rue performa
chieves afte
roperties suc
onstants, the
f a PVSA p
eds, while t
dsorption, r
urations. Th
ixing or opti
he performa
pylene and to
industrially
50/50 mol/m
and second
cking (FCC)
low tempera
eed streams.
othermal, co
ong with the
g various PV
mization of
SA process i
ance, and th
er many cy
ch as equilib
e performanc
process. The
the latter in
rinse, blowd
hus, unlike a
mizing its op
ance of its
otal annualiz
relevant pro
mol mixture
is the 85/15
units. Whil
ature of 353
We neglect
oncentration-
equilibrium
VSA process
f PVSA Pr
is inherently
hus design,
ycles of co
brium isother
ce at CSS de
former incl
nclude the o
down, and
a continuous
perational de
PVSA pro
116
zed cost for
opylene/prop
from the the
5 mol/mol m
e both sourc
K to increa
t pressure dr
-dependent m
m and kinetic
ses.
rocesses
y transient an
is dictated b
ntinuous op
rm, isosteric
epends on bo
lude the num
operational s
evacuation)
s plant, one
etails. Becau
cess at the
a fixed prop
pane feed mi
ermal cracki
mixture from
ces are at ab
ase kinetic se
rops through
micropore di
c parameters
nd cyclic, an
by the cycli
peration. In
c heat of ads
oth structura
mbers and d
steps (e.g. p
), their sequ
cannot desi
use the true m
e CSS, one
pylene/propa
ixtures (Kha
ing of liquid
m the off-g
bout 2-3 atm
electivity an
h the adsorpt
iffusion mod
s from Khal
nd has no tru
ic steady st
n addition t
sorption, and
al and operat
dimensions o
pressurizatio
uence, pres
ign a PVSA
measure of a
cannot ass
ane feed rate
alighi et al., 2
d feedstocks
ases of the
m and 600-80
nd a pressure
ion columns
del of Khalig
lighi et al. (2
ue steady stat
ate (CSS) th
to the adso
d diffusional
tional param
of the adsor
on, high-pre
sure levels,
A process wi
an adsorbent
sess or com
e. We
2013)
such
fluid
00 K,
e of 2
s, and
ghi et
2013)
te. Its
hat it
orbent
l time
meters
rption
essure
, and
ithout
t is in
mpare
116
adsorbents without finding the best process for each. Thus, to compare 4A zeolite and
SiCHA and identify the best, we must first develop/design the best PVSA process for
each separately. This highlights the need for a full-fledged synthesis and optimization
(Agarwal and Biegler, 2012; Haghpanah et al., 2013b) of the relevant PVSA processes.
The full-fledged synthesis and optimization of a PVSA process is a major
challenge for several reasons. Adsorption is a highly nonlinear phenomenon. Its
modeling, simulation, and optimization in the context of a PVSA process involves
repeated solution of complex hyperbolic partial differential and algebraic equations
(PDAEs). This is extremely time-consuming and requires efficient numerical simulators
(Haghpanah et al., 2013a) and sophisticated optimization algorithms (Agarwal et al.,
2010b). Many cycles of operation must be simulated to arrive at the cyclic steady state
(CSS) describing the actual performance of a PVSA process at each point during
optimization.
Several optimization studies (Biegler et al., 2005) using a variety of approaches
for several practical separation problems (e.g. Agarwal et al. (2010a; 2010b; 2003) for
CO2 capture and concentration; Lewandowski et al. (1998) and Cruz et al. (2005; 2003)
for air separation; Nikolic et al. (2009) for hydrogen recovery) exist in the literature, but
none on propylene/propane separation. Biegler et al. (2005) classified the various
optimization approaches into four groups: 1) Simplified, 2) Black-box, 3) Equation-
oriented, and 4) Simultaneous tailored. While the simplified approach of Smith IV and
Westerberg (1990) assumes a sequence of bed operations and bed design parameters such
as bed length and pressure levels to find the minimum number of beds and a cyclic
116
operating schedule, the other approaches address much wider and varying scopes for the
design, operation, and optimization.
The black-box approach is essentially simulation-based optimization
(Subramanian et al., 2000; Varma et al., 2008), in which a series of separate (black-box)
simulations of a PVSA process guides the optimization algorithm. The simulations may
involve either a fully rigorous model of the PVSA process, or an approximate or
surrogate model derived and updated with continuous help from the rigorous model. For
instance, Kapoor and Yang (1988) used polynomial expressions to fit the outputs
(product purities and recoveries) of a rigorous simulation model in terms of the inputs
(feed pressure, depressurization pressure, and throughput) for CO-H2 separation.
Lewandowski et al. (1998) developed an Artificial Neural Network (ANN) model for the
separation of nitrogen from air, and used a nonlinear programming approach to minimize
the cost of producing nitrogen. Other surrogate models such as ANFIS (Adaptive
Network-based Fuzzy Inference System) and Kriging (Biegler and Lang, 2012; Caballero
and Grossmann, 2008; Faruque Hasan et al., 2011; Lang et al., 2011) are also attracting
increasing attention. The black-box approaches have one major disadvantage. The details
of process dynamics are not fully integrated within or transparent to the optimization
algorithm. While this does reduce the complexity of the optimization model, it
compromises the nature and progress of the optimization algorithm. If a black-box
approach uses a surrogate model, then it has one more major disadvantage. The surrogate
model being less complex than the rigorous one, does speed up the optimization
algorithm, but its predictions of process performance, especially in extrapolated
situations, are often inaccurate.
116
In contrast to the black-box approach, the equation-oriented and simultaneously
tailored approaches embed the PDAEs for the PVSA process explicitly inside the
optimization formulation. Nilchan and Pantelides (1998) proposed complete
discretization (CD) involving a third order orthogonal collocation on finite elements for
the spatial domain and a first order backward finite difference method for the temporal
domain. They imposed simple periodic boundary conditions on process variable profiles
to ensure CSS, and used SQP (Sequential Quadratic Programming) for optimization.
Agarwal et al. (2010b) presented a novel superstructure for the optimal cycle
configuration of PVSA processes. They formulated an optimal control problem, and
employed complete discretization for its solution. They used a first-order finite volume
method for the spatial domain and orthogonal collocation on finite elements for the
temporal domain. They used IPOPT (Biegler, 2010) to solve the large nonlinear program.
Nikolic et al. (2009) reported an optimization framework for complex PSA processes
with multi-bed configurations and multi-layered adsorbents, and illustrated it for
hydrogen recovery from SMR (Steam Methane Reforming) off-gas (Nikolic et al., 2008).
They used orthogonal collocation for the spatial domain, and solved the PDAEs in
gPROMS (Barton, 1992). They employed a state transition network (STN) approach for
efficient simulation and optimization using the gOPT toll with reduced sequential
quadratic programming (rSQP) algorithm. STN approach has simpler and linear
implementation in multi-bed PSA systems, where states are represented by operation
steps (such as pressurization, adsorption, etc.), inputs are the step durations and operating
parameters.
op
in
op
on
th
a
o
an
pr
5
4
co
G
st
as
st
h
op
st
u
Jiang
ptimization.
n the black-b
ptimization
ne cycle to
he algorithm
modified fi
scillations fo
nd integrate
rogramming
Assess.3
Based o
-step cycle
ountercurren
Grande and R
tudy of 4A z
s the base cy
tep of cocurr
ave zero du
ptimizer wo
tep cycles. S
s to compare
et al. (2003)
Instead of s
box approac
problem. At
obtain the v
m attains CSS
nite volume
for steep fron
e the bed eq
g (rSQP) for
sment App
on the argum
with pressu
nt evacuation
Rodrigues (2
zeolite. To be
ycle on whic
rent blowdo
uration, ena
ould be able
Strictly speak
e the two ad
) proposed th
solving the P
ch, they imp
t each iterat
values of the
S only when
e (van Leer)
nts. Then, th
quations. Fin
optimization
proach
ments and o
urization, hi
n to be the b
2005) allowe
e fair, we ad
ch to compar
wn is subop
abling it to
to automati
king, we mu
sorbents, as
116
he simultane
PDAEs to th
posed just th
tion, they so
e constraints
n it achieves
method with
hey employe
nally, they u
n.
bservations
igh-pressure
best cycle fo
ed an additio
dopt the 5-ste
re SiCHA an
ptimal for bo
vanish dur
ically choose
ust find the b
it is possible
eous tailored
he full CSS
he CSS con
olve PDAEs
s and objecti
the optimal
th smooth flu
ed the DAE
used reduced
of Khalighi
e adsorption
or both SiCH
onal step of
ep cycle of G
nd 4A zeolit
oth 4A zeolit
ring optimiz
e the best b
best PVSA p
e that the be
d approach f
condition at
ndition as a
in an inner
ive function
l solution. In
ux delimiter
E solver DAS
d-space succ
i et al. (2013
n, rinse with
HA and 4A z
cocurrent bl
Grande and R
te. However
te and SiCH
zation. In o
etween the
process for ea
est PVSA cyc
for PVSA pr
t each iterati
constraint i
r loop for ex
n. In other w
nitially, they
rs to decreas
SPK 3.0 to
cessive quad
3), we expec
h propylene
zeolite. How
lowdown in
Rodrigues (2
, to show tha
HA, we allow
other words
5-step and t
ach adsorben
cle for 4A ze
ocess
ion as
in the
xactly
words,
y used
se the
solve
dratic
ct the
, and
wever,
n their
2005)
at the
w it to
s, the
the 4-
nt for
eolite
116
is not the same as that for SiCHA. However, this requires the synthesis of an optimal
cycle for each adsorbent, which is a challenge in itself. Therefore, instead of full cycle
synthesis optimization, we allow limited synthesis option of 4-step versus 5-step cycle.
Thus, in this study, we optimize the 5-step cycle (Figure 5.1a) for both SiCHA
and 4A zeolite separately. It involves (1) pressurization, (2) high-pressure adsorption, (3)
rinse with recycled heavy product from step 5 (called heavy reflux), (4) cocurrent
blowdown, and (5) countercurrent evacuation. Steps 2, 3 and 4 produce propane, and step
5 produces propylene. Our assessment is purely based on the nonisothermal isobaric
micropore model of Khalighi et al. (2012), which they validated on the experimental data
(Grande and Rodrigues, 2005) of 4A zeolite. Chapter 3 and 4 lists the model equations
and boundary conditions, and summarizes the parameters used in this study. For more
details, please refer (Khalighi et al., 2013; 2012). As indicated earlier, we target 99%
pure propylene and 90% propane. Recall that recoveries are fixed by the purities in a
binary separation, when there are no waste streams such as in our chosen 5-step cycle.
Since energy consumption is a key consideration in this separation, we use energy use per
tonne of propylene as the first criterion for judging a PVSA process. As an alternate
criterion, we use total annualized cost for a given feed flow. We compare the two
adsorbents based on both these two criteria.
5
M
pr
ri
th
fi
co
su
si
ti
R
pu
Figure 5.1
Imple.4
We so
Multiphysics
repared a se
inse, blowdo
hrough these
irst cycle as
onvergence
uccessive cy
imulation ru
ime on a Del
RAM. From
urities.
(a)
P PR
PH
PL
PR HP
Step 1 Ste
Feed
: (a) 5-step PV
mentation
olved the m
software (G
parate COM
own, and e
e steps seque
sumes clean
to CSS by
ycles differs
un required a
ll Optiplex 7
the exit vel
HPA RI
PA RI
ep 2 Step 3
Propane product
VSA process,
n of Simul
model equati
Guide, 2005
MSOL file fo
evacuation. W
entially. Eac
n bed with n
monitoring
s by less tha
about 50-60
780 with 2.66
locity and c
Propyle
BD EV
BD EV
Step 4 Step 5
116
, (b) 4-step PV
ation Mod
ions in their
5). COMSO
or each of the
We program
ch cycle beg
no gaseous c
propylene
an 0.1%, we
cycles to re
6 GHz Intel®
composition
ene product
Time
VSA process
del
r dimension
OL uses the
e five steps o
mmed a MA
gins with the
components
purity. Whe
e assume tha
each the CS
® Core 2 Qu
profiles at
Feed
PR
Step 1
P PR
PH
PL
(b)
with 4A zeol
nless forms
finite elem
of pressuriza
ATLAB pro
e pressurizat
in any phas
en propylen
at the CSS
SS and need
uad Q9400 C
CSS, we co
HPA
Propane product
RI
1 Step 2 Step 3
RI
HPA
lite and SiCH
using COM
ment method
ation, adsorp
ocedure to
tion step, an
se. We judg
ne purity for
is reached.
ed 3-4 h of
CPU and 8 G
omputed pro
Propylene p
EV
3 Step 4
EV
Tim
HA.
MSOL
d. We
ption,
cycle
nd the
ge the
r five
Each
CPU
GB of
oduct
product
me
116
3 6
3 6 3 8
0
0 00 0
0Propylene Purity 0
(%)1 0
EV V
EV
E
t
C H z
t t
C H C Hz z
C v
C v dt C
dt
v dt
=
= =
=+
×
(5.1)
3 8 3 8 3 80 0 00 0 0
0 0 00 0 0
100Propane Purity%
Ad Ri BD
Ad Ri BD
t t t
C H C H C Hz z z
t t t
total total totalz z z
C v dt C v dt C v dt
C v dt C v dt C v dt
= = =
= = =
× + +=
+ +
(5.2)
For a binary separation, recoveries are fixed by specified purities (Khalighi et al., 2013),
when there is no waste stream.
(1 ) 1
(1 )(1 )j i j i i j
ii j i j i
F Pu Pu F Pu PuRe
Pu Pu Pu Pu F
− −= ×
− − − (5.3)
where, iPu , iRe , iF are the purity, recovery, and molar feed composition (mol) of
component with as the other component. For 99% pure propylene and 90% pure
propane, propylene (propane) recovery is 89% (99%) for the 50/50 propylene/propane
feed, and 98% (94%) for the 85/15 propylene/propane feed.
Let denote the final pressure in step 1 and the pressure during steps 2 and 3.
Let denote the blowdown pressure in step 4, and denote the evacuation pressure in
step 5. Figure 5.1a uses one compressor for steps 1 and 2, another for step 3, and one
common vacuum pump for steps 4-5. Since the feed is at 2 atm, the compressor must
pressurize the feed to during steps 1 and 2, if exceeds 2 atm. In step 3, it must
pressurize the rinse stream from 1 atm to . Since we are assuming an isobaric system,
pressure drops through the beds are zero and the compressor needs no additional energy.
Thus, work done by the compressor for the 5-step cycle is given by,
= + + (5.4)
where, and given by the following generic expression,
116
1
0
1( ) ( ) 1 IF 2
1 ( )
0 IF 2
Hin in H
C in
H
PF t P t dt P atm
W P t
P atm
γγτγ
η γ
− − > = −
≤
(5.5ab)
where, η = 0.72, γ = 1.4, is the step duration, ( ) is the flow of gas entering the
compressor, and ( ) is the pressure of gas entering the compressor. is computed
via the following.
1
0
1( ) ( ) 1
1ri H
C in inatm
PW F t P t dt
P
γγτγ
η γ
− = − −
(5.6)
where, atmP is atmospheric pressure.
We assume that ≤ 1 atm and the vacuum pump always delivers gas at 1 atm.
The vacuum pump will reduce the bed pressure from to in step 4, and from to
in step 5. Then, the total energy consumption by the vacuum pump is given by,
= + (5.7)
Again, each right side term in eq. 7 is computed by the following generic expression,
1
0
1( ) ( ) 1
1 ( )atm
V in inin
PW F t P t dt
P t
γγτγ
η γ
− = − −
(5.8)
With this, the total energy consumption for the cycle in Figure 5.1a is given by:
= ( + )/(3.6 × 42 × ) kWh per tonne propylene fed (5.9)
where, 1F is the total moles of propylene entering the process during pressurization and
adsorption.
fr
(
w
th
5
pr
co
w
ap
m
U
op
th
co
ab
al
A
For ri
raction ( ) f
), we comp=where, is
he universal
Optim.5
In this st
rimary reas
ontrols the tr
with concentr
ppropriate fo
more comple
Using the mi
ptimization
he best and
omparison o
As menti
bout 3-4 h o
lso not adv
ANFIS (Ada
inse, we coll
from the tank
pute the velo/(the moles o
gas constant
mization A
tudy, a surr
on for usin
ransport in 4
ration-depen
for simulatin
x than the o
cropore diff
approach se
d most expe
of two promi
ioned earlier
of CPU time.
isable. Mag
aptive Netwo
lect propyle
k as heavy r
ocity of propy( )of gas collect
t, ε is the bed
Algorithm
rogate-based
ng the black
4A zeolite an
ndent diffusiv
ng our 5-step
ones that are
fusion mode
eems intracta
edient choic
ising adsorbe
r, each simu
. Thus, a Sim
guire et al.
ork-based F
116
ne in a buff
reflux. For g
ylene enterin
ted in the tan
d porosity,
d black-box
k-box approa
nd SiCHA. T
vities develo
p process. T
based on th
l with an eq
able at this t
ce, as the p
ents.
ulation run fo
mOpt strateg
(1998) show
Fuzzy Infere
fer tank duri
given rinse d
ng the feed d
ank, is colu
is gas tem
approach is
ach is as fo
Therefore, th
oped by Kha
The discretiz
he usual line
quation orien
time. Thus,
primary aim
for the 5-step
gy using the
wed that a
ence System
ing evacuati
duration (
during step 3
umn cross se
mperature.
s used for o
follows. Mic
he micropore
alighi et al. (
zation of thi
ear driving f
nted or simu
a black-box
m for this
p process in
rigorous sim
neuro-fuzzy
m) is more
ion and recy
) and reflux
3 as follows.
(5.1
ection area,
optimization
cropore diffu
e diffusion m
2012) is the
is model is m
force assump
ultaneous tai
x approach s
study is rel
n COMSOL
mulation mo
y model suc
accurate tha
ycle a
x ratio
.
10)
is
. The
fusion
model
most
much
ption.
ilored
seems
lative
takes
del is
ch as
an an
A
op
ar
h
(S
T
g
th
(T
la
m
co
al
au
d
G
P
b
ANN (Artific
ptimization.
A neuro-
rtificial neur
idden neuron
Sugeno and
Their ANFIS
enerates fuz
he incoming
T-norm oper
ayer has sev
model outpu
omputes the
Our sele
lgorithm is
ugments the
elivers the d
GA uses an A
5.5
The follo
IP , LP , G,
lowdown tim
cal Neural N
Thus, this s
-fuzzy mode
ral network
ns for fuzzy
Kang, 1988)
S architectu
zzy members
signals from
ration). The
veral nodes,
ut based on
weighted gl
ection of G
mainly for
e objective
desired purit
ANFIS mode
ANFIS5.1
owing ten va
RIt , inlet v
me ( Bdt ), an
Network) mo
study preferr
el is the pro
(ANN) and
y inference. J
) in an adapt
ure uses the
ship values
m the previo
third layer
where node
the first-o
lobal output
GA (Genetic
its simplicit
function wi
ties of 99%
el that is buil
S Model
ariables cons
velocity ( 0v )
nd evacuati
116
odel. It redu
red to use an
oduct of a h
fuzzy logic
Jang (1993)
tive network
e following
for the inpu
us layer, and
computes th
e k compute
order Takag
of the system
Algorithm
ty, and ease
ith penalties
propylene
lt and update
stitute the in
), pressuriza
on time ( Et
uces training
n ANFIS mod
hybrid intel
c (Amin et a
implemente
k-based fuzz
g layers of
ut variables.
d computes
he normalize
es the contri
gi-Sugeno ru
m.
m, (Holland,
e / speed of
s on purities
and 90% pr
ed as follow
nput variable
ation time (
Ev ). For eac
g time, and w
del.
ligent system
al., 2010). I
ed Takagi-Su
zy inference
operations.
The second
the firing str
ed firing stre
ibution of th
ules. Lastly
1973)) as
f implementa
s to ensure
ropane. Duri
s.
es for our AN
( Prt ), adsorp
ch input var
will be bette
m that com
It uses a lay
ugeno fuzzy
system (AN
The first
d layer mult
rength of the
ength. The f
he kth rule i
, the fifth
an optimiz
ation. This
that the pr
ing optimiza
NFIS model
ption time
riable, this
er for
mbines
yer of
rules
NFIS).
layer
iplies
e rule
fourth
in the
layer
zation
study
ocess
ation,
l: HP ,
( Adt ),
study
116
selects appropriate lower and upper bounds (Table 5-1). Purity of propylene, purity of
propane, and total energy consumption of a cycle comprise the three output variables of
the ANFIS model. Recall that recoveries are fixed by the purities. To build the initial
ANFIS model, we synthesize 200 sets of input variables using Latin hypercube sampling
(LHS) (Stein, 1987). For each point, we simulate the 5-step process with a bed of 2.5 cm
diameter and 75 cm length using COMSOL and MATLAB until CSS, and compute the
three output variables. From the 200 points and their solutions, we randomly choose 150
sample points to train the ANFIS model, and the remaining 50 sample points to validate
it. Figure 5.2 shows the qualities of ANFIS predictions for the test samples. As seen in
Figure 5.2, the predictions for all 50 points are very close (within 1%) to the results from
rigorous simulations. Thus, our ANFIS model is reliable. In fact, we increase its accuracy
even further by retraining it with all the 50 validation points as well. In other words, we
have an initial ANFIS model based on 200 rigorous simulations. As we discuss later,
during optimization, we continue to retrain and improve our ANFIS model with the
solutions from our optimization procedure. Figure 5.2 shows that the accuracy of our
ANFIS model improves with optimization progresses.
Using the initial ANFIS model, we proceed to optimize via GA. Figure 5.3 shows the
schematic of our optimization algorithm. We use the ANFIS model inside GA in
MATLAB to optimize the 5-step process. Then, we simulate in COMSOL the process
corresponding to the best values for the ten optimization variables. If any of the three
outputs (two purities and energy consumption per tonne of propylene) predicted by the
ANFIS model differs by more than 0.1% from its true value from COMSOL simulation,
then we retrain the ANFIS model by including this set of input variables. We now repeat
116
GA to find the best process. We continue to repeat this procedure, until the performance
prediction by GA for the best process matches with that from rigorous simulation within
0.1% for each of the three outputs. Figure 5.2 shows how the prediction errors of ANFIS
reduce with optimization. The algorithm converges after 4-5 iterations of GA.
Table 5-1: Best PVSA processes for 4A zeolite and SiCHA and two industrially relevant feed compositions.
SiCHA 4A zeolite Variable Bounds
Decision Variable 50/50 85/15 50/50 85/15 Both feeds
v0 (cm/s) 14.43 11.72 10.41 7.41 May-50
L0 (cm) 75 75 75 75 fixed
L0/v0 5.2 6.4 7.2 10.12 -
tpr (s) 95 134 97 104 20-1000
tad (s) 195 251 145 179 20-1000
tri (s) 58 62 53 49 1-1000
tbd (s) 0 0 0 0 1-1000
tev (s) 329 403 256 312 20-1000
PH (kPa) 296.43 348.03 257.09 275.78 101.3-1013
PM (kPa) NA NA NA NA 50.65-101.3
PL (kPa) 27.09 30.02 33.91 45.38 5.065-50.65
G 0.51 0.42 0.58 0.43 0.1-1.0 W (kWh/tonne propylene) 108 101 81 72 -
Propane Recovery (%) 99.18 99.12 99.14 99.17 - Propylene Recovery (%) 88.99 89.02 89.03 88.99 - Propane Purity (mol%) 90.01 90.03 90.04 90.01 -
Propylene Purity (mol%) 99.09 99.02 99.04 99.08 -
Feed rate (mol/h) 11.08 10.56 6.93 5.29 -
5
an
co
op
fe
w
pr
T
so
Figure 5.2
Comp.6
We comp
nd 75 cm le
onsumption.
perating par
eed, by using
Minimiz
where, λ=150
ropylene pur
The optim
Table 5-1. Th
olutions from
2: Number of
parison Ba
pare the two
ength, but a
. For each
rameters that
g the followi
ze Z W λ= +
00 for SiCH
rity, and P2 i
mization res
he best valu
m our optimi
0
1
2
3
4
5
6
1
Purit
y Er
ror%
f iteration vs.
ased on En
adsorbents
allow the fe
adsorbent-
t achieve mi
ing objective
[(min 0, 99λ −
HA and λ=
is propane p
sults for the
ues for all va
ization are n
1.5 2
116
difference ofparameters
nergy Con
using a 5-ste
ed rate to v
feed combi
inimum ener
e function in
]1 min[0P− +
1000 for 4A
purity.
e four adsorb
ariables are
not limited by
2.5 3Iterati
f ANFIS and s.
nsumption
ep process w
vary for min
ination, we
rgy consump
n GA.
)20,90 ]P−
A zeolite ar
bent-feed co
away from
y the impose
3.5 4ion
50/50 SiCHA
50/50 4A
COMSOL re
with a bed of
nimizing the
e identify t
mption per to
re penalty p
ombinations
their bound
ed bounds.
4 4.5
85/15 SiCHA
85/15 4A
sults for optim
f 2.5 cm diam
e specific en
the best pr
nne of propy
(5.11)
parameters,
are present
ds. Thus, the
5
mum
meter
nergy
ocess
ylene
P1 is
ted in
e best
116
All processes yield the desired purities, but the duration for blowdown is zero for all.
Adsorption and rinse are sufficient to produce 90% propane, and blowdown is
unneceaary. This confirms our earlier assertion that the 4-step process is better than the 5-
step process for both SiCHA and 4A zeolite, as far as energy demand per tonne of
propylene feed is concerned.
The best SiCHA-based processes require 101 (108) kWh of energy per tonne of
propylene for the 85/15 (50/50) feed. This confirms the assertion of Khalighi et al. (2013)
that the separation of 85/15 feed is less energy-demanding than that of 50/50 feed. The
best 4A-based processes require 72 (81) kWh of energy per tonne of propylene for the
85/15 (50/50) feed. In other words, the relative separation energy demands for the two
feeds are similar for 4A zeolite as well. The lower concentration of propylene in the feed
necessitates a higher feed velocity, higher high pressure, higher rinse time, and lower
evacuation pressure. All these result in higher energy consumptions for the compressors
and vacuum pumps.
The energy advantage of 4A, however, comes at the cost of throughput. As we see
from Table 5-1, the best SiCHA-based processes allow 11.07 mol/h of the 50/50 feed and
10.56 mol/h of the 85/15 feed. In contrast, the 4A-based processes allow only 6.92 mol/h
of the 50/50 feed and 5.29 mol/h of the 85/15 feed. Clearly, SiCHA is superior from the
perspective of throughput. If one considers the plant cost in addition to the energy costs,
SiCHA might prove better than 4A zeolite! Thus, it is worthwhile to consider total
annualized cost of separation as the ultimate criterion for comparing these two
adsorbents.
5
m
or
9
as
W
m
o
Comp.7
We d
minimum TA
r feed rate. O
9% propylen
ssume the fo
We design a
minimum TA
r feed rate.
PS
Ke
PSA process
C
Cycl
Cper
c
Fig
parison Ba
design a sep
AC. To ensur
Our objectiv
ne and 90%
ollowing for
a separate
C. To ensure
Our objecti
SA simulation
Kinetic and equilibrium parameters
Model eand assu
COMSOL and
MATLAB
lic Steady State
Calculate the rformance and
energy consumption
gure 5.3: Optim
ased on To
arate 5-step
re a fair com
ve is a proces
% propane. S
simplicity.
5-step pro
e a fair com
ve is a proc
equations umptions
1
116
mization algo
otal Annua
process for
mparison, ea
ss that separ
Since our int
cess for ea
parison, eac
cess that sep
Data
200 rasam
150 sample for training
- 10 input paramtbd, trv, PH, PI, PL,- Purity of propyenergy consumpt
Train the
orithm used in
alized Cos
r each adso
ach process m
rates a given
terest is onl
ach adsorbe
ch process m
parates a giv
a bank
andom mples
50 sample fortesting
meters ( tpr, thpa, tr
, v0, G) ylene and propanetion
ANFIS
Add tto data ban
the A
n this work.
st (TAC)
orbent, and t
must have th
n propylene/p
ly a relative
ent, and th
must have th
ven propyle
r
ri,
e,
R
this sample nk and train ANFIS
No
then compu
he same cap
propane feed
e comparison
hen comput
he same cap
ne/propane
Optimization
Genetic Algorithm ANF�C
Optimum operating
parameters
Run the COMSOLwith optimum
parameters
Whether thesimulation resultare comparable with GA results
Results
Yes
ANFIS
ute its
pacity
d into
n, we
te its
pacity
feed
L
e s
116
into 99% propylene and 90% propane. Since our interest is only a relative comparison,
we assume the following for simplicity.
1. The monetary unit is 2012 US$.
2. Capital annualization factor is 0.1.
3. The process operates 8000 h per annum.
4. It uses ≥ 2 identical beds of length and diameter . Multiple columns are
necessary to receive the propylene/propane feed in a continuous manner.
5. A buffer with negligible cost collects the propylene product from the evacuation
step, and decouples the operations of the evacuation and rinse steps.
6. 3 ≤ ≤ 8 holds. This is based on expert observations (Agrawal, 2013;
Towler, 2013) from practice that most adsorption columns in the industry obey
these limits on L/d ratio. This is largely to limit pressure drop in a real column.
7. The annual operating expenditure (OPEX) for the 5-step process is solely from
the energy required for separation.
8. The electricity tariff is 0.0671 in 2012 US$/kWh. (EMA, 2013)
9. The total capital cost of the process is seven times the purchase cost of
columns.
Using the above, the fewest columns required for a continuous feed are,
= 1 + ( + + )/( + ) (5.12)
where, represents the integer ceiling of .
116
Figure 5.4: Effect of bed length on the minimum energy for SiCHA and 4A for 50/50 and 85/15.
68
76
84
92
100
30 60 90 120 150 180
Spec
ific
Ener
gy C
onsu
mpt
ion
(KW
h/to
nne
prop
ylen
e)
Bed Length (cm)
IsothermalNon-isothermalAdiabaticIsothermalNon-isothermaAdiabatic
85/15 SiCHA L/v0= 6.4
50/50 SiCHA L/v0= 5.2
50/50 4A L/v0= 7.2
85/15 4A L/v0= 10.1
116
Figure 5.5: Effect of bed diameter on the minimum energy for SiCHA and 4A for 50/50 and 85/15.
Recall that we used a bed with diameter = 2.5 cm and length = 75 cm in our
ANFIS model, and allowed the feed flow (or inlet interstitial velocity ) to vary. This
may be too small to achieve a desired flow of mol/s. Thus, we need a larger column
with diameter and length , which must now be additional variables in our cost
optimization along with , bed pressures, and step durations. To avoid a new ANFIS
model with and as variables, we devise a scale-up procedure based on the following
three heuristics.
Heuristic 1: The energy consumption of the 5-step PVSA process with bed length L and
inlet interstitial velocity is depends on / . In other words, changes in and do not
affect the energy consumption, as long as / remains the same.
To show the above heuristic, we tuned a separate ANFIS model for several over a
wide range of 20-200 cm. For each , we minimized the energy consumption by varying
60
80
100
120
1 2 3 4 5
Spec
ific
Ener
gy C
onsu
mpt
ion
(kw
h/to
nne
prop
pyle
ne)
Bed Diameter (cm)
SiCHA 50/50 feed SiCHA 85/15 feed 4A 50/50 feed 4A 85/15 feed
116
and other parameters. Figure 5.4 shows the optimal / ratios for the four adsorbent-
feed combinations at different s. As we can see, the optimal / is roughly constant at
5.14 for the 50/50 feed on SiCHA, 6.40 for the 85/15 feed on SiCHA, 7.21 for the 50/50
feed on 4A zeolite, and 10.11 for the 85/15 feed on 4A zeolite. Figure 5.4 also shows that
the minimum energy consumption also remains constant with for a given / . Note
that the optimal / ratios for 4A are higher than those for SiCHA. This suggests a
longer (thus larger column and higher capital cost) for 4A than SiCHA at a given feed
rate.
Heuristic 2: The energy consumption (kWh per tonne of propylene) of the 5-step PVSA
process remains practically unchanged with bed diameter as long as and other
parameters ( , bed pressures, and step durations) remain constant.
For this, we simulated the four minimum-energy processes from Table 5-1 for various
using COMSOL. Figure 5.5 shows that energy consumption is practically independent
of for each solution. In other words, for any given , the largest diameter (or minimum / ratio) will maximize the feed rate and capacity.
Lastly, the column in our ANFIS model was small and non-isothermal, i.e.
allowed heat losses. In contrast, industrial columns are large and nearly adiabatic.
Therefore, we need the following heuristic to account for the heat effects.
Heuristic 3: While our ANFIS model is for a non-adiabatic, non-isothermal, 5-step PVSA
process, it predicts very well the energy consumption of an adiabatic industrial column.
In other words, the impact of heat effects on energy consumption is negligible.
To understand the heat effects, we simulated the minimum-energy non-isothermal
116
processes reported in Figure 5.4 for various under isothermal and adiabatic conditions.
For the former, we fixed the inside and outside heat transfer coefficients to be very large,
and for the latter, we made the inside heat transfer coefficient zero. The energy
consumptions for these two limits are also shown in Figure 5.4. As we can see, the effect
of heat transfer on energy consumption is practically negligible. This validates the use of
our ANFIS model for designing systems with larger columns.
Heuristic 4: For this kinetically controlled PVSA process, adsorption during column
pressurization (Step 1) should be negligible. In other words, a frozen bed assumption
should hold for this separation.
From our various simulations, the average ratio of the actual amount of feed
entering the column during pressurization to that entering a frozen bed is 1.01 for both
SiCHA and 4A zeolite. This confirms that adsorption during pressurization can be
neglected. Then, we can compute as follows.
= ε 1 − [ + (1 − )] (5.13)
where, = /4. Eq 13 fixes , so it ceases to be an optimization variable.
The first three heuristics enable us to size a large column with length , diameter ,
and inlet interstitial velocity to accommodate a feed of mol/s based on the simulation
of our small ANFIS column with a feed of mol/s. First,
( / ) = ε × ( ) × ( / )( ) × ( )( ∙ 3∙ −1∙ −1) 0( ) (5.14a)
Heuristic 1 tells us that as long as we maintain / = / , the energy consumption
of the large column will be the same as that of the ANFIS column. Then,
116
= ε × ( ) × ( ) × ( )( ∙ ∙ ∙ ) ( ) = ( / ) (5.14b)
where, = /4 cm3, = 368.2 cm3, = 353 K is the feed temperature, and = 8314 ∙ ∙ ∙ .
From eq. 5.14, the annual OPEX for the 5-step process of capacity mol/s is, $ = 28.8 × × ( $ )
(5.15)
where, is the mol fraction of propylene in the feed, is its molecular weight (g/mol),
and is the specific energy consumption.
For computing the annual capital expenditure (CAPEX), we use the following
correlation (Turton et al., 2008) for the purchase cost ( ) of a column:
log = 3.4974 + 0.4485 log( ) + 0.1074[log( )]
(2012 $) = ×
$ = 0.7 × × × (5.16)
From eqs. 5.15 and 5.16, we get = + as the objective
function for our GA-based optimization.
= + + (min[0, 99 − ] + min[0, 90 − ]) (5.17)
where, = 15000 for SiCHA and = 10000or 4A zeolite. Note that and are
not in the objective function, and = . Thus, the variables in our optimization are
, bed pressures, and step durations except . Once we know the best value for , we
116
can compute = 4 /(3 ) and = 3 .
For = 10 mol/s (12,700 tonne/year of the 50/50 feed and 12,500 tonne/year of the
85/15 feed), our optimization gives the best 5-step processes in Table 5-2 for the four
feed-adsorbent scenarios. First, Step 3 (blowdown) again has zero duration. Thus, the 5-
step process is worse than the 4-step process for SiCHA and 4A zeolite from both energy
and cost perspectives. Second, the minimum-TACs (7.78 $/tonne propylene for the 50/50
feed and 7.04 $/tonne propylene for the 85/15 feed) for SiCHA are higher than those
(6.51 $/tonne propylene for the 50/50 feed and 5.44 $/tonne propylene for the 85/15 feed)
for 4A zeolite. Thus, separation using 4A zeolite is cheaper than that using SiCHA
zeolite, and the 85/15 feed is cheaper to separate than the 50/50 feed.
Since we did not assume frozen bed for the minimum energy results in Table 5-1,
we ran our optimizations again with the frozen bed assumption. Table 5-3 also lists the
results for the min-energy processes with the frozen bed assumption. We see that the
energy consumptions for the min-TAC processes are higher than those for the min-energy
processes, as the optimizer increases energy consumption slightly to reduce column size.
However, they are not too far away from the minimum energy consumptions, as OPEX
dominates CAPEX in this separation. Finally, as expected, the TACs for the min-energy
processes are higher than those for the min-TAC processes.
116
Table 5-2: Comparison of 4A zeolite and SiCHA based on minimum-energy PVSA processes for two industrially relevant feeds.
SiCHA 4A zeolite
Decision parameters 50/50 85/15 50/50 85/15
v0 (cm/s) 33.23 24.56 18.78 15.73
tad (s) 231 261 168 197
tri (s) 43 51 36 40
tbd (s) 0 0 0 0
tev (s) 356 411 289 363
PH(kPa) 401.32 432.69 321.35 371.82
PL (kPa) 31.03 36.82 21.54 38.93
PM (kPa) 0 0 0 0.00
G 0.64 0.47 0.67 0.39
Propane Recovery (%) 99.19 99.12 99.11 99.17
Propylene Recovery (%) 89.02 89.05 89.07 89.04
Propane Purity (%) 90.03 90.05 90.07 90.05
Propylene Purity (%) 99.1 99.02 99.01 99.08
W (kWh/tonne propylene) 110 104 83 75
L0/v0 2.26 3.05 3.99 4.77
F0 (mol/s) 9.59E-03 7.64E-03 4.34E-03 4.21E-03
tpr (s) 3.05 4.09 5.45 6.25
F (mol/s) 10 10 10 10
N 3.00 3.00 3.00 3.00
Volume (m3) 0.38 0.48 0.85 0.88
capex ($/year) 6.88E+03 7.48E+03 9.42E+03 9.55E+03
opex ($/year) 4.46E+04 7.17E+04 3.37E+04 51742.15
TAC ($/year) 5.15E+04 7.92E+04 4.31E+04 6.13E+04
TAC ($/tonne propylene) 7.78 7.04 6.51 5.44
D (cm) 54.62 58.91 71.14 71.88
L (cm) 163.85 176.73 213.41 215.65
Velocity (cm/s) 72.59 57.87 53.44 45.23
116
Table 5-3: Comparison of 4A zeolite and SiCHA based on minimum-energy PVSA processes for two industrially relevant feeds and using frozen bed assumption.
SiCHA 4A zeolite
Decision parameters 50/50 85/15 50/50 85/15
v0 (cm/s) 15.01 12.31 11.18 8.06
tad (s) 196 253 146 182
tri (s) 57 60 50 46
tbd (s) 0 0 0 0
tev (s) 333 405 259 315
PH(kPa) 300.01 345.65 300.21 277.64
PL (kPa) 28.01 31.21 35.28 46.19
PM (kPa) 0 0 0 0.00
G 0.52 0.43 0.6 0.45
Propane Recovery (%) 99.16 99.06 99.11 99.12
Propylene Recovery (%) 89.01 89.05 89.04 88.98
Propane Purity (%) 90.02 90.05 90.05 90.02
Propylene Purity (%) 99.04 99 99.05 99.07
W (kWh/tonne propylene) 108.68 101.54 81.89 72.94
L0/v0 5.00 6.09 6.71 9.31
F0 (mol/s) 3.24E-03 3.06E-03 2.41E-03 1.61E-03
tpr (s) 6.63 8.11 8.67 11.36
F (mol/s) 10 10 10 10
N 3.00 3.00 3.00 3.00
Volume (m3) 1.14 1.20 1.53 2.29
capex ($/year) 1.07E+04 1.10E+04 1.23E+04 1.52E+04
opex ($/year) 4.41E+04 7.01E+04 3.32E+04 50320.97
TAC ($/year) 5.48E+04 8.11E+04 4.56E+04 6.55E+04
TAC ($/tonne propylene) 8.28 7.20 6.88 5.82
d (cm) 78.43 79.93 86.50 99.02
L (cm) 235.29 239.78 259.51 297.06
Velocity (cm/s) 47.09 39.36 38.68 31.92
Finally, to confirm that the energy predictions remain valid through our scale up
procedure, we simulate both the small column and the large scaled-up column using both
COMSOL and ANFIS. Table 5-4 shows that the purities, recoveries, and energy
consumptions for the small columns are reasonably close. Therefore, the ANFIS model is
well-trained. In addition, the energy consumptions for the large column are close to those
pr
En
5
pr
co
le
pr
su
in
ad
se
in
se
to
redicted by C
Table 5-4
Out
nergy (kWh/to
Propane Pur
Propylene Pu
Chapt.8
The adsor
roperties su
onstant, etc.
evel. To hav
rocess. In t
urrogate AN
ndustrially re
The k
dsorption-ba
eems a good
ndicator of
electivity of
onne of pro
COMSOL fo
4: Comparisonand accu
tput
onne propylene
rity (mol%)
urity (mol%)
ter conclu
rption literat
ch as equili
with little r
ve a scientific
this work, w
NFIS model t
elevant prop
key conclusi
ased separat
d indicator o
the process
f 4A zeolite
opylene (25%
or the small
n of COMSOuracy of ANF
C
Small colu
50/50 8
e) 160.8 1
99.16 9
89.04 8
sion
ture abounds
ibrium/kineti
regard to the
c compariso
we employed
to compare 4
ylene/propan
ion of this
tion of prop
f the energy
s productivi
e (224 vs 28
% less for
116
column. Thi
OL results for FIS prediction
COMSOL resul
umn Scale
5/15 50/50
48.3 161.2
9.14 99.15
8.95 89.03
s with compa
ic selectivity
e ultimate p
on between a
d a simulati
4A zeolite a
ne feeds of 5
work is tha
pylene/propa
y need (OPEX
ity (CAPEX
8) translates
the 50/50 f
is confirms o
the ANFIS cons for the ANF
lts
up column
85/15
149.36
99.16
88.98
arisons base
y, capacity,
performance
adsorbents w
on-optimiza
and SiCHA a
50/50 mol/m
at 4A zeolit
ane mixture
X) in this se
X). In this
s well into l
feed and 29
our scale up
olumn and thFIS column.
ANFI
Small column
50/50 85/15
161.3 145.8
99.51 98.67
89.92 88.99
ed on specifi
heat of ads
of adsorben
we should co
ation based a
adsorbents fo
mol and 85/1
te is superio
es. While k
eparation pro
process, th
lower separ
9% less for
procedure.
he process col
IS prediction
n Scale up co
5 50/50 8
8 162.9 1
7 99.71 9
9 89.32 8
ic thermophy
sorption, He
nts at the pr
onsider optim
approach us
for separating
5 mol/mol.
or to SiCHA
kinetic selec
ocess, it is a
he higher ki
ration energy
the 85/15
lumn
olumn
85/15
145.8
98.84
88.65
ysical
enry’s
ocess
mized
sing a
g two
A for
ctivity
a poor
inetic
y per
feed)
116
compared to SiCHA. However, the minimum energy process for a 4A zeolite needs
larger columns (more capital cost) than SiCHA. Unless the capital costs for this
separation are comparable to the operating costs, 4A zeolite seems a better adsorbent.
Our total annualized cost optimizations based on some simple assumptions confirm this
conclusion, as the total cost of separation for 4A zeolite is lower than that for SiCHA.
The minimum-TAC processes use slightly higher energy (kWh per tonne of propylene
fed) than the minimum-energy processes to reduce capital costs. This works confirms that
4A zeolite is superior to SiCHA.
C
m
S
w
6
fo
CHAPTE
In this c
modeling and
iCHA and
work are also
Concl6.1
Based o
ollowing con
1- A non
simula
micro
accord
partia
and M
with e
4A ze
to the
the di
rarely
becom
ER 6
chapter, the
d optimizati
4A zeolites
o discussed.
usion
on the simu
nclusions are
n-isothermal
ate kinetical
pore diffusi
ding to the c
l differentia
MATLAB us
experimental
eolite. The d
bi-LDF mo
fference is n
y diffuses to
mes significa
e major fin
ion studies c
are present
ulation and o
e drawn:
l isobaric m
ly controlled
ivity depend
chemical pot
al equations
sing finite e
l data presen
etailed pore
odel [advoca
not very larg
the adsorbe
ant the penet
116
Conclus
ndings and
conducted o
ted. In addi
optimization
micropore d
d pressure sw
ds on adsor
tential gradie
of the PSA
element met
nted in the lit
diffusion m
ated by by G
e for the pre
ent. It has be
tration of pro
ion and F
conclusions
on adsorptio
ition, some
n studies car
diffusion mo
wing adsorpt
rbate conce
ent as the dri
process hav
thod. The d
terature prop
model is quan
Grande and R
esent system
een further c
opane into th
Future w
s obtained f
on and diffu
recommend
rried out in
odel has be
tion process
entration in
iving force f
ve been solv
eveloped m
pylene/propa
ntitatively su
Rodrigues (2
m because in
confirmed th
he micropore
work
from simula
usion of gas
dations for f
this project
een develope
ses. In this m
the solid p
for diffusion
ved in COM
model is vali
ane separatio
uperior comp
2005)], alth
this case pro
hat the differ
es is signific
ation,
ses in
future
t, the
ed to
model,
phase
n. The
MSOL
dated
on on
pared
hough
opane
rence
cant.
116
2- A new 8-ring zeolite, pure silica chabazite (SiCHA), has been studied in this
work. The diffusion of propane in SiCHA is extremely slow, thus making
equilibrium information for propanevery challenging. Therefore propane
equilibrium parameters have been indirectly estimated using available
experimental uptake data and validated using molecular simulation. Using a
combination of experimental and estimated equilibrium parameters, and
experimental kinetic parameters, a 4-step PVSA process with SiCHA including
pressurization, adsorption, rinse and evacuation step is developed to separate
propylene/propane mixture. Two main industrially relevant feed compositions,
such as 50/50 propylene/propane and 85/15 propylene/propane, are studied for
this separation. It has been demonstrated that the process can deliver the
industrial requirements of 99% propylene and 90% propane products.
3- In this work, the performance of two adsorbents, SiCHA and 4A zeolite which
have the highest kinetic selectivity among the available adsorbents is studied. This
comparison bases on optimization results. This study detects the best adsorbent
along with best PVSA process. Our assessment criteria are the energy
consumption per tonne of propylene of the PVSA process and total annualized
cost for a fixed propylene/propane feed rate. This work optimizes four processes,
50/50 and 85/15 propylene/propane feed using SiCHA and 50/50 and 85/15
propylene/propane feed using 4A. It is assumed a low temperature of 353 K to
increase kinetic selectivity. For both adsorbents, this study uses surrogate neuro-
fuzzy models to predict this rigorous simulation model and optimize the processes
via a genetic algorithm (GA). 85/15 feed process with 4A zeolite has the
6
co
minim
tonne
to SiC
Futur6.2
Besides th
onsideration
1- To con
more
accura
2- To co
data o
3- To m
indust
4- To c
model
discre
orthog
this se
5- A mo
orient
condu
consu
mum energy
of propylen
CHA.
e work
he simulati
ns for future
nduct an exp
temperature
ately.
onduct an ex
of propane.
modify the pr
trial purpose
onvert the
l to algebra
etization, a
gonal colloc
et of algebrai
ore robust op
ed approach
ucted in GA
ming, it cou
consumptio
ne comparing
ion and op
research inc
periment to m
e points, in o
xperiment to
roposed mod
es which is p
hyperbolic
aic equation
finite volum
ation in fini
ic equation c
ptimization
h using com
AMS softwa
uld give more
116
on per tonne
g to the othe
ptimization
clude:
measure the
order to calc
find a poss
del equation
pressure drop
partial diffe
ns using co
me method
ite element c
can be solve
algorithm c
mplete discre
are. While,
e reliable op
e of propyle
er three proc
work cond
uptake data
culate their
sible method
n to non-isob
p in the bed c
erential alge
mplete disc
can be us
can be used
ed in MATLA
can be prop
etization me
this proce
ptimum resul
ene and min
cesses and 4A
ducted in th
a of propylen
diffusivity c
d to measure
baric model
column is pr
ebraic equat
cretization m
ed for spat
d for tempora
AB software
posed based
thod. This a
edure is com
lts.
nimum TAC
A seems sup
his study,
ne and propa
coefficients
e the equilib
for pilot pla
romising.
tions of pro
method. For
tial domain,
al domain. T
e.
on an equa
algorithm ca
mplex and
C per
perior
some
ane at
more
brium
ant or
opose
r this
, and
Then,
ation-
an be
time
116
6- To find a possible procedure to synthesize a SiCHA with higher equilibrium
selectivity for propylene and propane. It should increase the kinetic selectivity as
well therefore it may reduce the energy consumption and total cost of the
separation process.
116
References
Agarwal, A., Biegler, L., Zitney, S., 2009. A trust-region algorithm for the optimization
of PSA processes using reduced-order modeling. National Energy Technology
Laboratory-In-house Research; National Energy Technology Laboratory (NETL),
Pittsburgh, PA, and Morgantown, WV (United States).
Agarwal, A., Biegler, L.T., 2012. A NOVEL FILTER TRUST-REGION ALGORITHM
FOR CONSTRAINED OPTIMIZATION USING REDUCED ORDER MODELING.
Agarwal, A., Biegler, L.T., Zitney, S.E., 2010a. Superstructure-based optimal synthesis
of pressure swing adsorption cycles for precombustion co2 capture. Industrial &
Engineering Chemistry Research 49, 5066-5079.
Agarwal, A., Biegler, L.T., Zitney, S.E., 2010b. A superstructure-based optimal synthesis
of PSA cycles for post-combustion CO2 capteffectively captureure. AIChE Journal 56,
1813-1828.
Amin, T., Lloyd, H.C.C., W., W.T.S., 2010. Evaluation of rainfall and discharge inputs
used by Adaptive Network-based Fuzzy Inference Systems (ANFIS) in rainfall-runoff
modeling. Journal of Hydrology 391, 248–262.
Baerlocher, C., McCusker, L.B., Olson, D.H., 2007. Atlas of Zeolite Framework Types,
Sixth Revised ed. Elsevier.
Barton, P., Pantelides, C., 1994. Modeling of combined discrete/continuous processes.
AIChE Journal 40, 966-979.
116
Barton, P.I., 1992. The modelling and simulation of combined discrete/continuous
processes. Imperial College London (University of London).
Bhadra, S.J., Farooq, S., 2011. Separation of Methane–Nitrogen Mixture by Pressure
Swing Adsorption for Natural Gas Upgrading. Industrial & Engineering Chemistry
Research 50, 14030-14045.
Biegler, L.T., 2010. Nonlinear programming: concepts, algorithms, and applications to
chemical processes. SIAM.
Biegler, L.T., Jiang, L., Fox, V.G., 2005. Recent Advances in Simulation and Optimal
Design of Pressure Swing Adsorption Systems. Separation & Purification Reviews 33, 1-
39.
Biegler, L.T., Lang, Y., 2012. Multi-scale Optimization for Advanced Energy Processes,
11th International Symposium on Process Systems Engineering-PSE2012. Elsevier, p.
51.
Bird, R.B., Stewart, W.E., Lightfoot, E.N., 1960. Transport Phenomena. John Wiley &
Sons, New York.
Bordiga, S., Regli, L., Cocina, D., Lamberti, C., Bjorgen, M., Lillerud, K.P., 2005.
Assessing the Acidity of High Silica Chabazite H−SSZ-13 by FTIR Using CO as
Molecular Probe: Comparison with H−SAPO-34. The Journal of Physical Chemistry B
109, 2779-2784.
Bryan, P.F., 2004. Removal of Propylene from Fuel-Grade Propane. Separation &
Purification Reviews 33, 157 - 182.
Caballero, J.A., Grossmann, I.E., 2008. An algorithm for the use of surrogate models in
modular flowsheet optimization. AIChE Journal 54, 2633-2650.
116
Camblor, M.A., Villaescusa, L.A., Díaz-Cabañas, M.J., 1999. Synthesis of all-silica and
high-silica molecular sieves in fluoride media. Topics in Catalysis 9, 59-76.
Crank, J., 1975. The Mathematics of Diffusion. OXFORD UNIVERSITY PRESS,
London.
Cruz, P., Magalhães, F.D., Mendes, A., 2005. On the optimization of cyclic adsorption
separation processes. AIChE Journal 51, 1377-1395.
Cruz, P., Santos, J.C., Magalhaes, F.D., Mendes, A., 2003. Cyclic adsorption separation
processes: analysis strategy and optimization procedure. Chemical Engineering Science
58, 3143-3158.
Da Silva, F.A., Rodrigues, A.E., 1999. Adsorption Equilibria and Kinetics for Propylene
and Propane over 13X and 4A Zeolite Pellets. Industrial & Engineering Chemistry
Research 38, 2051-2057.
Da Silva, F.A., Rodrigues, A.E., 2001a. Propylene/propane separation by vacuum swing
adsorption using 13X zeolite. AIChE Journal 47, 341-357.
Da Silva, F.A., Rodrigues, A.E., 2001b. Vacuum Swing Adsorption for
Propylene/Propane Separation with 4A Zeolite. Industrial & Engineering Chemistry
Research 40, 5758-5774.
de Pablo, J.J., Laso, M., Suter, U.W., 1992. Simulation of Polyethylene Above and
Below the Melting Point. J. Chem. Phys. 96, 2395-2403.
Díaz-Cabañas, M.-J., Barrett, P.A., Camblor, M.A., 1998. Synthesis and structure of pure
SiO2 chabazite: the SiO2 polymorph with the lowest framework density. Chem.
Commun., 1881 - 1882.
116
Do, D.D., 1998. Adsorption analysis: equilibria and kinetics. Imperial College Press,
London.
Eldrige, R.B., 1993. Olefin/paraffin separation technology: a review. Industrial &
Engineering Chemistry Research 32, 2208-2212.
EMA, 2013. Singapore Electricity Tariff http://www.ema.gov.sg/Electricity/new/.
Farooq, S., Rathor, M.N., Hidajat, K., 1993. A predictive model for a kinetically
controlled pressure swing adsorption separation process. Chemical Engineering Science
48, 4129-4141.
Farooq, S., Ruthven, D.M., 1990. Heat effects in adsorption column dynamics. 2.
Experimental validation of the one-dimensional model. Industrial & Engineering
Chemistry Research 29, 1084-1090.
Farooq, S., Ruthven, D.M., 1991. Numerical simulation of a kinetically controlled
pressure swing adsorption bulk separation process based on a diffusion model. Chemical
Engineering Science 46, 2213-2224.
Faruque Hasan, M.M., Karimi, I.A., Farooq, S., Rajendran, A., Amanullah, M., 2011.
Surrogate-based VSA Process Optimization for Post-Combustion CO2 Capture, in: E.N.
Pistikopoulos, M.C.G., Kokossis, A.C. (Eds.), Computer Aided Chemical Engineering,
vol. 29. Elsevier, pp. 402-406.
Frenkel, D., Mooij, G.C.A.M., Smit, B., 1992. Novel Scheme to Study Structural and
Thermal Properties of Continuously Deformable Molecules. J. Phys.: Condensed Matter
4, 3053-3076.
Grande, C.A., Firpo, N., Basaldella, E., Rodrigues, A.E., 2005. Propane/Propene
Separation by SBA-15 and π-Complexated Ag-SBA-15. Adsorption 11, 775-780.
116
Grande, C.A., Gascon, J., Kapteijn, F., Rodrigues, A.E., 2010a. Propane/propylene
separation with Li-exchanged zeolite 13X. Chemical Engineering Journal 160, 207-214.
Grande, C.A., Gigola, C., Rodrigues, A.E., 2003. Propane–Propylene Binary Adsorption
on Zeolite 4A. Adsorption 9, 321-329.
Grande, C.A., Poplow, F., Rodrigues, A.E., 2010b. Vacuum Pressure Swing Adsorption
to Produce Polymer-Grade Propylene. Separation Science and Technology 45, 1252 -
1259.
Grande, C.A., Rodrigues, A.E., 2004. Adsorption Kinetics of Propane and Propylene in
Zeolite 4A. Chemical Engineering Research and Design 82, 1604-1612.
Grande, C.A., Rodrigues, A.E., 2005. Propane/Propylene Separation by Pressure Swing
Adsorption Using Zeolite 4A. Industrial & Engineering Chemistry Research 44, 8815-
8829.
Guide, F.U., 2005. COMSOL Multiphysics user's Guide Version 3.2. Sweden.
Guisnet, M., Gilson, J.P., 2002. Zeolites for Cleaner Technologies. Imperial College
Press, London.
Haghpanah, R., Majumder, A., Nilam, R., Rajendran, A., Farooq, S., Karimi, I.A.,
Amanullah, M., 2013a. Multi-objective Optimization of a 4-step Adsorption Process for
Post-combustion CO2 Capture Using Finite Volume Technique. Ind Eng Chem Res 0, 0.
Haghpanah, R., Nilam, R., Rajendran, A., Farooq, S., Karimi, I.A., Ullah, M.A., 2013b.
Cycle Synthesis and Optimization of a VSA Process for Post-combustion CO2 Capture.
AIChE Journal 0, 0.
116
Hedin, N., DeMartin, G.J., Roth, W.J., Strohmaier, K.G., Reyes, S.C., 2008. PFG NMR
self-diffusion of small hydrocarbons in high silica DDR, CHA and LTA structures.
Microporous and Mesoporous Materials 109, 327-334.
Hirotani, A., Mizukami, K., Miura, R., Takaba, H., Miya, T., Fahmi, A., Stirling, A.,
Kubo, M., Miyamoto, A., 1997. Grand canonical Monte Carlo simulation of the
adsorption of CO2 on silicalite and NaZSM-5. Applied Surface Science 120, 81-84.
Holland, J.H., 1973. Genetic algorithms and the optimal allocation of trials. SIAM
Journal on Computing 2, 88-105.
Hu, X., Do, D.D., 1992. Multicomponent adsorption kinetics of hydrocarbons onto
activated carbon: effect of adsorption equilibrium equations. Chemical Engineering
Science 47, 1715-1725.
Imtex, 2009. Olefin/Parafin separation.
Jang, J.-S.R., 1993. ANFIS: Adaptive-Network-Based Fuzzy Inference System. IEEE
Transactions on Systems, Man, And Cybernetics 23, 665-685.
Jarvelin, H., Fair, J.R., 1993. Adsorptive separation of propylene-propane mixtures.
Industrial & Engineering Chemistry Research 32, 2201-2207.
Jiang, L., Biegler, L.T., Fox, G.V., 2003. Simulation and optimization of pressure-swing
adsorption systems for air separation. AIChE Journal 49, 1140-1157.
Kapoor, A., Yang, R.T., 1988. Optimization of a pressure swing adsorption cycle.
Industrial & Engineering Chemistry Research 27, 204-206.
Kapoor, A., Yang, R.T., 1989. Kinetic separation of methane--carbon dioxide mixture by
adsorption on molecular sieve carbon. Chemical Engineering Science 44, 1723-1733.
116
Kärger, J., Bülow, M., 1975. Theoretical prediction of uptake behaviour in adsorption
kinetics of binary gas mixtures using irreversible thermodynamics. Chemical Engineering
Science 30, 893-896.
Khalighi, M., Chen, Y.F., Farooq, S., Karimi, I.A., Jiang, J.W., 2013. Propylene/Propane
Separation Using SiCHA. Industrial & Engineering Chemistry Research 1, 1-2.
Khalighi, M., Farooq, S., Karimi, I.A., 2012. Nonisothermal Pore Diffusion Model for a
Kinetically Controlled Pressure Swing Adsorption Process. Industrial & Engineering
Chemistry Research 51, 10659-10670.
Knaebel, S.P., Ko, D., Biegler, L.T., 2005. Simulation and Optimization of a Pressure
Swing Adsorption System: Recovering Hydrogen from Methane. Adsorption 11, 615-
620.
Ko, D., Siriwardane, R., Biegler, L.T., 2003. Optimization of a Pressure-Swing
Adsorption Process Using Zeolite 13X for CO2 Sequestration. Industrial & Engineering
Chemistry Research 42, 339-348.
Lamia, N., Wolff, L., Leflaive, P., Leinekugel-Le-Cocq, D., Gomes, P.S., Grande, C.A.,
Rodrigues, A.r.E., 2008. Equilibrium and Fixed Bed Adsorption of 1-Butene, Propylene
and Propane Over 13X Zeolite Pellets. Separation Science and Technology 43, 1124 -
1156.
Lang, Y., Zitney, S.E., Biegler, L.T., 2011. Optimization of IGCC processes with reduced
order CFD models. Computers & Chemical Engineering 35, 1705-1717.
Lewandowski, J., Lemcoff, N.O., Palosaari, S., 1998. Use of Neural Networks in the
Simulation and Optimization of Pressure Swing Adsorption Processes. Chemical
Engineering & Technology 21, 593-597.
116
Maguire, L.P., Roche, B., McGinnity, T.M., McDaid, L., 1998. Predicting a chaotic time
series using a fuzzy neural network. Information Sciences 112, 125-136.
Majumdar, B., Bhadra, S.J., Marathe, R.P., Farooq, S., 2011. Adsorption and Diffusion of
Methane and Nitrogen in Barium Exchanged ETS-4. Industrial & Engineering Chemistry
Research 50, 3021-3034.
Martin, M.G., Siepmann, J.I., 1998. J. Phys. Chem. B 102, 2569.
Martin, M.G., Siepmann, J.I., 2000. J. Phys. Chem. B 104, 8008.
Mathias, P.M., Kumar, R., Moyer, J.D., Schork, J.M., Srinivasan, S.R., Auvil, S.R., Talu,
O., 1996. Correlation of Multicomponent Gas Adsorption by the Dual-Site Langmuir
Model. Application to Nitrogen/Oxygen Adsorption on 5A-Zeolite. Industrial &
Engineering Chemistry Research 35, 2477-2483.
Nakao, S., Suzuki, M., 1983. Mass transfer coefficient in cyclic adsorption and
desorption. J. Chem. Eng. Japan 16.
Nikolic, D., Giovanoglou, A., Georgiadis, M.C., Kikkinides, E.S., 2008. Generic
modeling framework for gas separations using multibed pressure swing adsorption
processes. Industrial & Engineering Chemistry Research 47, 3156-3169.
Nikolić, D., Kikkinides, E.S., Georgiadis, M.C., 2009. Optimization of Multibed Pressure
Swing Adsorption Processes. Industrial & Engineering Chemistry Research 48, 5388-
5398.
Nilchan, S., Pantelides, C.C., 1998. On the Optimisation of Periodic Adsorption
Processes. Adsorption 4, 113-147.
Olson, D.H., 2002. Light hydrocarbon separation using 8-member ring zeolites The
Trustess of the University of Pennsylvania (Philadelphia, PA) USA.
116
Olson, D.H., Camblor, M.A., Villaescusa, L.A., Kuehl, G.H., 2004. Light hydrocarbon
sorption properties of pure silica Si-CHA and ITQ-3 and high silica ZSM-58.
Microporous and Mesoporous Materials 67, 27-33.
Padin, J., Rege, S.U., Yang, R.T., Cheng, L.S., 2000. Molecular sieve sorbents for kinetic
separation of propane/propylene. Chemical Engineering Science 55, 4525-4535.
Qinglin, H., Farooq, S., Karimi, I.A., 2004. Prediction of binary gas diffusion in carbon
molecular sieves at high pressure. AIChE Journal 50, 351-367.
Raghavan, N.S., Hassan, M.M., Ruthven, D.M., 1986. Numerical simulation of a PSA
system using a pore diffusion model. Chemical Engineering Science 41, 2787-2793.
Rege, S.U., Yang, R.T., 2002. Propane/propylene separation by pressure swing
adsorption: sorbent comparison and multiplicity of cyclic steady states. Chemical
Engineering Science 57, 1139-1149.
Ribeiro, F.R., Rodrigues, A.E., Rollmann, L.D., Naccache, C., 1984. Zeolites: Science
and Technology. Martinus Nijhoff, Hague.
Ritter, J.A., Bhadra, S.J., Ebner, A.D., 2011. On the Use of the Dual-Process Langmuir
Model for Correlating Unary Equilibria and Predicting Mixed-Gas Adsorption Equilibria.
Langmuir 27, 4700-4712.
Ruthven, D.M., 1984. Principles of Adsorption and Adsorption Processes. John Wiley &
Sons. Inc., Fredericton.
Ruthven, D.M., Farooq, S., Knaebel, K.S., 1994. Pressure Swing Adsorption VCH, New
York.
Ruthven, D.M., Reyes, S.C., 2007. Adsorptive separation of light olefins from paraffins.
Microporous and Mesoporous Materials 104, 59-66.
116
Ryckaert, J.P., Bellemans, A., 1978. United-Atom Model. Faraday Discuss. Chem. Soc.
66, 95.
Safarik, D.J., Eldridge, R.B., 1998. Olefin/Paraffin Separations by Reactive Absorption:
A Review. Ind. Eng. Chem. Res. 37, 2571–2581.
Salil U. Rege, Joel Padin, Ralph T. Yang, 1998. Olefin/paraffin separations by
adsorption: pi-Complexation vs. kinetic separation. AIChE Journal 44, 799-809.
Shin, H.S., Knaebel, K.S., 1988. Pressure Swing Adsorption: An Experimental Study of
Diffusion-Induced Separation. AIChE J. 34, 1409-1416
Siepmann, J.I., Frenkel, D., 1992. Configurational-bias Monte Carlo: A ew sampling
scheme for flexible chains. Molecular Physics 75, 59-70.
Sikavitsas, V.I., Yang, R.T., Burns, M.A., Langenmayr, E.J., 1995. Magnetically
Stabilized Fluidized Bed for Gas Separations: Olefin-Paraffin Separations by .pi.-
Complexation. Industrial & Engineering Chemistry Research 34, 2873-2880.
Silva, J.C., Rodrigues, A.E., 1996. Analysis of ZLC technique for diffusivity
measurements in bidisperse porous adsorbent pellets. Gas Separation & Purification 10,
207-224.
Smit, B., Maesen, T.L.M., 2008. Molecular simulations of zeolites: Adsorption,
diffusion, and shape selectivity. Chem. Rev. 108, 4125-4184.
Smith Iv, O.J., Westerberg, A.W., 1990. Mixed-integer programming for pressure swing
adsorption cycle scheduling. Chemical Engineering Science 45, 2833-2842.
Stein, M., 1987. Large sample properties of simulations using Latin hypercube sampling.
Technometrics 29, 143-151.
116
Stoitsas, K.A., Gotzias, A., Kikkinides, E.S., Steriotis, T.A., Kanellopoulos, N.K.,
Stoukides, M., Zaspalis, V.T., 2005. Porous ceramic membranes for propane–propylene
separation via the p-complexation mechanism: unsupported systems. Microporous and
Mesoporous Materials 78, 235–243.
Subramanian, D., Pekny, J.F., Reklaitis, G.V., 2000. A simulation—optimization
framework for addressing combinatorial and stochastic aspects of an R&D pipeline
management problem. Computers & Chemical Engineering 24, 1005-1011.
Sugeno, M., Kang, G., 1988. Structure identification of fuzzy model. Fuzzy sets and
systems 28, 15-33.
ter Horst, J.H., Bromley, S.T., van Rosmalen, G.M., Jansen, J.C., 2002. Molecular
modelling of the transport behaviour of C3 and C4 gases through the zeolite DD3R.
Microporous and Mesoporous Materials 53, 45-57.
Van Ness, H., Smith, J., Abbott, M., 2001. Introduction to chemical engineering
thermodynamics. McGraw-Hill, Crawfordsville, IN.
Varma, V.A., Pekny, J.F., Blau, G.E., Reklaitis, G.V., 2008. A framework for addressing
stochastic and combinatorial aspects of scheduling and resource allocation in
pharmaceutical R&D pipelines. Computers & Chemical Engineering 32, 1000-1015.
Wakao, N., Funazkri, T., 1978. Effect of fluid dispersion coefficients on particle-to-fluid
mass transfer coefficients in packed beds : Correlation of sherwood numbers. Chemical
Engineering Science 33, 1375-1384.
Wehner, J.F., Wilhelm, R.H., 1956. Boundary conditions of flow reactor. Chemical
Engineering Science 6, 89-93.
116
Yakubovich, O.V., Massa, W., Gavrilenko, P.G., Pekov, I.V., 2005. Crystallography
Reports 50, 544.
116
List of publications
PUBLICATION
• "Non-isothermal Pore Diffusion Model for a Kinetically Controlled PSA
Process", Mona Khalighi, Shamsuzzaman Farooq, and Iftekhar A Karimi, accepted in
Industrial and chemical engineering research, 2012
• "Modeling and Simulation of a PSA Process for Propylene/Propane Separation on
SiCHA", Mona Khalighi, Y Chen, S Farooq1, I A Karimi, and J Jiang, submitted in
Industrial and chemical engineering research, 2012
• “Assessment of Pressure/Vacuum Swing Adsorption Process For
Propylene/Propane Separation”, Mona Khalighi, I A Karimi , S Farooq, In progress.
CONFERENCE
• 5th PBAST Conference, Singapore 2009, Poster presentation "Non-isothermal Po
Propane\Propylene Separation by PSA on SiCHA", Mona Khalighi, Shamsuzzaman
Farooq, and Iftekhar A Karimi
• 7th International Chemical Engineering Congress & Exhibition (IChEC 2011),
Iran, Oral presentation, " Separation of propylene/propane via pressure swing adsorption
using SiCHA, Mona Khalighi, Shamsuzzaman Farooq, and Iftekhar A Karimi
116
• ESCAPE 22 Conference, London 2012, Poster presentation "Modeling and
Simulation of a PSA Process using SiCHA for Propylene/Propane Separation", Mona
Khalighi, Shamsuzzaman Farooq, and Iftekhar A Karimi
• 11th International Symposium on Process Systems Engineering (PSE),
Singapore2012, Oral presentation " Optimizing the PSA process of propylene/propane
using Neuro-Fuzzy modeling", Mona Khalighi, Shamsuzzaman Farooq, and Iftekhar A
Karimi, ( will be published in Elsevier)