Vous êtes sur la page 1sur 8

Chemical Engineering Science, Vol. 44. No.

Printed in Gnat Britain.

I, pp. 1-8,

1989.
0

@m-2509/89
$3.00 + 0.00
1989 Pergamon Press plc

PRESSURE
SWING ADSORPTION:
DEVELOPMENT
OF AN
EQUILIBRIUM
THEORY
FOR BINARY
GAS MIXTURES
WITH NONLINEAR
ISOTHERMS
JOHN C. KAYSER
and KENT
S. KNAEBEL
Department of Chemical Engineering, The Ohio State University, Columbus, Ohio 43210, U.S.A.

(First received 27

October

1987; accepted

in revisedform

11 July

1988)

Abstract-A conventional pressure swing adsorption cycle is analyzed for an arbitrary feed composition of a
binary mixture where both components may have nonlinear equilibrium isotherms. The method of
characteristics is used to solve the equilibrium-based continuity relationships resulting in an algebraic
expression for the recovery of the less strongly adsorbed component.. For the case of linear isotherms, the
nonlinear expression reduces properly to the previbusly developed and experimentally verified binary linear
isotherm theory. A case study examines the separation of oxygen from air at 0.0-C and pressures up to
6.0atm on zeolite 13X. At these conditions, the oxygen isotherm is essentially linear, but nitrogen has a
concave downward equilibrium isotherm. That curvature reduces recovery compared with that expected
based on purely linear isotherms, which are approached at low absolute pressures. Specifically, for constant
pressure ratios the recovery decreases as either the purge pressure or feed pressure increases. Conversely,
operation in the linear isotherm domain yields no indep+dent effect of those pressures (i.e. only the pressure
ratio is of consequence). In addition, when the purge pressure is fixed, there exists an intermediate pressure
ratio at which maximum rkcovery is achieved. In contrast, recovery increases monotonically with pressure
ratio in the linear isotherm region.

The analysis presented here identifies conditions that


directly influence process performance and therefore
facilitates cycle modifications
that may significantly
improve
separation
performance.
In addition, the
effects of curved isotherms 011 product recovery are
explored using actual nonlinear isotherm data in a
case study.
The solution is limited to constant pressure feed and
purge steps, in which complete bed utilization occurs
(i.e. the composition fronts reach the end of the bed at
the end of the respective steps). Likewise, it is presumed that pure product is recovered, and that a
portion is used for pressurization as well as purge.

1. INTRODUCTION

Background
The recent development of equilibrium-based
adsorption models has provided considerable insight to the
parameters that affect the performance
of pressure
swing adsorption (PSA) processes. In some cases (i.e.
given certain assumptions
and constraints)
those
models have yielded analytical solutions to the continuity equations. Such results are appealing because
they allow simple examination of general performance
trends over broad ranges of operating conditions. In
cases where the assumptions and constraints of the
analytical model are closely approximated,
the equations may be used quite accurately to predict cycle
times and product recovery for a given adsorber
design and operating
conditions.
In addition, the
theoretical framework
of equilibrium-based
models
has proven to be a useful basis for conducting experimental research.
1.1.

1.3. Prior work


Equilibrium-based
theories of PSA have evolved
towards applicability
in increasingly
general situations. Early theories were limited to the purification
of binary mixtures in which the more strongly adsorbed component was present at trace impurity levels
(Shendalman and Mitchell, 1972; Chan et al., 1981).
Other work extended the applicability
to arbitrary
feed compositions
of a binary mixture but resorted to
numerical integration of some of the relevant equations (Flores Fernandez and Kenney, 1983). Analytical solutions to the equations governing PSA processes for an arbitrary feed composition
were subsequently developed (Knaebel and Hill, 1985). That
analysis applied for linear isotherms, constant pressure feed and purge steps, and for either pressurization
with feed or with product. The theory was validated in
bench-scale air separation experiments (Kayser and
Knaebel, 1986) and is referred to here as the binary
linear isotherm, or BLI, theory.

1.2. Object and approach of this work


This paper presents an analytical solution to the
continuity equations applied to a specific PSA cycle
for the case of a binary mixture that exhibits nonlinear
adsorption
isotherm behavior for one or both adsorbing species. The inclusion of nonlinear isotherm
relationships
in the governing
equations is needed
since many actual (and potential)
industrial PSA
cycles
operate
under conditions
in which
the
components exhibit nonlinear adsorption equilibria.

Author

to whom correspondence

should be addressed.
1

JOHN C. KAYSER

and KENT

S. KNAEBEL

An extension of the BLI theory relaxed the constraint of constant pressure during the feed step
(Kayser and Knaebel, 1988). Cycles that allow pressure to vary during the feed step were predicted to have

Feed
NW YF

significantly different, and under certain conditions


superior, separation performance than those having a
constant pressure feed step. Following in that vein, the
case of a varying pressure feed step with nonlinear
isotherms is a logical extension of this work and will be
the subject of a future report.
Prior developments of equilibrium-based theories
for PSA with nonlinear isotherms have been reported
by Flores Fernandez and Kenney (1983), Doong and
Yang (1986) and Underwood (1986). In each case the
governing equations were integrated numerically,
hence little advantage was gained over comparable
simulations that incorporated
dissipative effects.
Mathematical models of PSA for several classes of
effects have been fully reviewed by Wankat (1986) and
Yang (1987).
The equations derived in the subsequent section are
applicable, in general, to isothermal sorption steps and
are not restricted to PSA cycles. In that vein, the
current development resembles the recent work on
equilibrium sorption processes by Basmadjian and
Coroyannakis (1987) and Davis and LeVan (1987).
Both employ hodograph transformations for adsorption step and cycle analysis. In the former, a nonadsorbing carrier bearing two solutes is treated. That
permits veloeity variation to be neglected, whereas it is
taken into account in the present work. Furthermore,
accumulation of material in the interstitial fluid is
neglected in the former, while it is included here. That
term is not negligible when adsorbent selectivity is low
to moderate, as in separation of oxygen from air with
zeolite 13X, which is considered as a case study later in
this paper. Finally, Davis and LeVan (1987) have
examined gas separations by thermal swing cycles
comprised of adsorption, heating and (optionally)
-cooling steps. Their analysis of adiabatic cycles incorporated wave phenomena that were discerned by
mapping between hodograph and physical planes.
That approach may improve PSA simulations when
both thermal effects and nonlinear isotherms are
relevant.

2.

DEVELOPMENT

OF THE

ISOTHERM

BINARY

NONLINEAR

THEORY

2.1. Basic concepts


The underlying principles of the present development are virtually identical to those presented in the
earlier papers (Knaebel and Hill, 1985; Kayser and
Knaebel, 1988). Consequently, the similar details are
only summarized here, but the distinctions are emphasized in the subsequent derivation of the working
equations.
The basic steps of the PSA cycle studied in this
paper as well as conventions of position and direction
are shown in Fig. 1. The pressures during the respective steps are shown in Fig. 2.

STEP

Blowdown
la FL

@
YW

Vent
top,

r=L

1
Zf

z=o

Pressurization
with Product
NPR

Fig. 1. Basic steps in the pressure swing adsorption cycle,


flow orientation convention, and molar flows that are pertinent to the process performance.

Step
Fig. 2. Pressureswing patternduring a typical cycle: 1 refers
to the pressurizationstep, 2 refersto the feed step, 3 refersto
the blowdown step, and 4 refersto the purge step (cf. Fig. 1).

The assumptions and constraints of the theory are


listed below.

(1) Two component, ideal gas mixture.


(2) Local equilibrium between the gas and solid
(3)
(4)
(5)
(6)
(7)
(8)
2.2.

phases.
Uncoupled adsorption equilibrium isotherms.
Negligible axial dispersion.
Negligible axial pressure gradients.
Constant pressure during feed and purge steps.
Isothermal operation.
No radial dependence of velocity or composition.
Material

balance

relationships

The analysis begins with the following component


molar balances:

&--

d@Y*
3PYA
at +
dz
>

+RT(l-E)+

(1)

Pressure swing adsorption


E-

am, + auph

( at

__

az >

+RT(l-s)fg=O. (2)

The first simplification of the balance equations is to


each
relationships
for
adsorption
substitute
component. For generality, independent but arbitrary
functions will be assumed as:
(3)

readily recast as two ordinary differential equations


(which must be simultaneously
satisfied) by the
method of characteristics. The technique involves the
elimination
of ay,/dt
and ay,/az using the total
differential of y,:
dY,=$$dt+gdz.
Solving eqs (10) and (11) simultaneously
leads
certain relationships between the terms, viz.
dz
-=
dt

There are, of course, physical constraints on these


functions. For example, as the adsorbate approaches
zero coverage, the relationship must follow Henrys
law. By convention,
the more strongly
adsorbed
component
is identified as component
A. At fixed
temperature, the partial derivatives with time of eqs (3)
and (4) are of the form:

Therefore, substitution of eqs (3) and (4) into eqs (1)


and (2), respectively, yields:
apYA
aupy,
-+pa-=
az

at

and

aph
-+A-

aupyl3 =
aZ

at

(8)

where P = PA l&3, which is also locally defined.

at.

between eqs (6) and (8) gives

aupyA
;g+tB-l)-+l
B
which will be of considerable

(6)and

aup

aYA
BAU
at+[l+(,?-l)~Al

aYA

az

=o

(9)

use later. In addition, eqs

(8) may also be combined

Cl +(P-

I)Y,I

(12)

and

dy,

(B- l)(l -YY,)Y,

df

C~+~P-~)YAIP'

(13)

The former yields characteristic trajectories in the z, tplane. The latter dictates the interdependence of composition and pressure along those characteristics. It is
clear from eq. (13) that, when pressure is constant,
composition is also constant along each characteristic.
Although eqs (6j(13) are identical in form to those
in the BLI theory, the present relations are inherently
more complicated
because of the composition
(and
pressure) dependence of p. Furthermore, at constant
pressure it is possible to rearrange eq. (9) to evaluate
the composition
dependence of the velocity which is
valid when y, and u are continuous differentiable
functions of t and z. The result is simply

1-B
dy, .
1 +(B- l)YA

Pg+(I-B)f$+8*$=0
dPY*
-

BAU

to

the componentpartial
pressure. Note that in the BLI
theory this parameter was constant, while here it is
defined locally.
The remaining
derivation
revolves around manipulation of eqs (6) and (7). First of all, adding eqs (6)
and (7) generates:

Eliminating

(11)

by eliminating

au
-:
aZ

W- l)(l -YA)YA 1 dp
1 +(B--

l)y,

P dt
(10)

Equilibrium-based
theories have been successful
and facile because eq. (10) is of a form that can be

(14)

Analytic integrals of the right-hand-side exist only for


some rather simple isotherm forms. A subsequent
relation is valid when discontinuities are present.
2.3. Wave phenomena
Before proceeding with analysis of the steps in a
PSA cycle, it is essential to establish the basis by which
the more subtle steps can be understood. Specifically,
the feed step and purge step are critical to the analysis
because of their impact on performance. Furthermore,
it has been learned that the wave phenomena associated with these steps provide keys to experimental
evaluation and operation of PSA systems (Kayser and
Knaebel, 1986; Matz and Knaebel, 1987).
Rhee et al. (1986) have discussed wave phenomena
at length. The only drawback is that, in their examples
that focus on chromatographic
applications, velocity
is presumed to be independent of composition, which
is not the case here. Despite that, the elementary
concepts that are employed here are explained more
fully there.
Formation
and propagation
of a composition
shock wave dominates the feed step. The condition
necessary for this occurrence is that the feed be
enriched in the more strongly adsorbed component as
compared with the initial column contents. For systems described by linear isotherms, the analysis is

JOHN C. KAYSER

and

simple: eq. (1) or (2) is solved in difference form by


exploiting the nil accumulation at the shock front.
For systems that have nonlinear isotherms, however, a balance of component
A around the shock
front yields a more involved expression for the velocity

where

and the subscripts 1 and 2 refer to the leading and


trailing edges of the shock front, respectively, and
&, =L(PYA,/RV,
as
specified
by
eq
(3).
Similarly, in such systems the compositions
that
bound the shock front are subject to a uniqueness
condition. Uniqueness must be determined because of
the multiplicity of solutions that could arise from the
nonlinear equations. One that is useful for the gas
adsorption
applications
described here is the generalized
entropy
condition
suggested by Oleinik
(1963). That condition is expressed as the following
inequalities based on eq. (10):

ti YA\I

(16)

y~,-y..+,

YA, -YA,

J&-yAa

where

s
Y%

J/Y-4,

BAU

dYA

(17)

l+(b-l)YA

in which j = 1, 2, or i, and y,, is within the closed


interval of y,, to yA2. Actually, the lower limit of
integratidn in eq. (17) is arbitrary and should not affect
the result. Note that the integrand is merely the
characteristic velocity defined by eq. (12). For the
simple case in which material at y,, is fed to an
initially clean column (yA, = 0), both inequalities yield
the same results, viz.

KENT

S. KNAEBW

When the influent contains more of the less strongly


adsorbed
component
than ir$tial contents of the
column, as occurs during the purge step, characteristics that describe the respective composition
paths
diverge, as given by eq. (12). This also results in a socalled simple wave, comprised
of material having
compositions
between those of the initial column
contents and the influent. The characteristics of the
influent, having constant composition, are parallel. As
a result, imminent breakthrough
of the influent depends only on its composition and velocity, not on the
initial column contents.

3. STEPWISE

CYCLE

ANALYSIS-

The performance of the cycle illustrated in Figs 1 and 2


involves, principally, eqs (8), (12), (13) and (15). Generally, these equations are manipulated
to obtain
expressions for the numbers of moles of influent
and/or efRuent during each step. The following development presumes complete bed utilization and that
the cycle is started at cyclic steady state, which always
connotes production of the pure, less strongly adsorbed component during the feed step.
3.1. Pressurization
Since the influent during this step has the same
composition as that used to purge the bed, viz. yAl. = 0
trivial. Thus, despite the
YAW ) eq. (13) becomes
Gessure change, composition
remains constant, and
eq. (8) can be integrated to determine the influent
interstitial velocity
L
UPlt =B,Pz.

dP

(1%

As a result, the number of moles of recycled product


required for pressurization is
NF% =
= A..~(~(~-~)+~l--d(ls~-~~))

(20)

W3)
YAz

Thus, yAZ = yA, only if J/,, increases monotonically


between yA = 0 and yAF. Otherwise, yA2 becomes an
intermediate
mole fraction, corresponding
to that
having the maximum characteristic velocity [as given
by eq. (12)]. In that case, the material having composition between yAl and y,, forms a so-called simple
wave. As a result, the bed cannot be completely
saturated with feed prior to imminent breakthrough of
the shock wave.
When the curvature of the isotherms is not severe,
i.e. at low partial pressures, the uniqueness condition
will be automatically satisfied, and the shock velocity
predicted by eq. (15) will be valid for yA2 = y,, . At
partial pressures sufficiently large for isotherm curvature to be significant, however, eq. (16) must be
considered.

where& is to be stipulated in a form given by eq. (4)


and the subscripts H and L refer to the high and low
pressure states, respectively.
3.2. High pressure feed
As explained in Section 2.3, this step is dominated
by the propagation
of a composition
shock wave
through the adsorbent bed. The velocity of that wave,
which is given by eq. (15), depends on the interstitial
fluid velocities at the leading and trailing edges of the
wave. These may in turn be interrelated by equating
the shock wave velocities determined from eq. (15) and
its counterpart based on component B to get

where 8 = 8,/e,

ut
-=

1 +te-

l)YA,

Ul

1 +

lb,,

and

(21)
ce -

Pressure swing adsorption

&Yi2
--Yil
pEl
1.

1-a
l+----

where&=1

A,-..&

RT

Note thatxj =A(P,yiJ /RT), as given by eq. (3) or (4),


for any state j. The parameter 8, is evaluated at a
specific (local) composition
by the tangent to the
respective isotherm. In contrast, the parameter Qi is
evaluated at a jump discontinuity by the chord of the
respective
isotherm.
For certain simple isotherm
forms, there are also simple analytical relationships
between & and pi and between 8 and /3. In fact, for
linear isotherms these terms are identical. Conversely,
under conditions
in which isotherm curvature
is
significant (i.e. yAI # y,,,,) eq. (21) cannot relate the
influent velocity to that of the product because the
composition
profile is no longer represented by an
ordinary jump condition. In that case, eq. (14) can
relate ur and u2.
The material balance equation for the feed step is
based on the time, t,, required for the shock wave to
traverse the length of the adsorber.

where, according to Fig. 1, the velocity is inherently


negative. Consequently,
when isotherm curvature is
slight (i.e. yal =yAF) and when yA, = 0, the product

moles are

Fig. 3. Graphical interpretation of the isotherm tangent and


chords that affect the material balance of a PSA cycle.

complete the entire cycle balance. For the nonlinear


isotherm case, analysis of partial purge will require a
rather involved numerical approach.
Some of the important terms from the material
balance equations are shown graphically
in Fig. 3.
Slope m, is equal to filyA= o, slope mp is equal to
f*(PH~AFIRT)I(PW~*FIRT),

N,,=N,Cl+(e--l)~,,rl.

(23)

fB (P,y,,,lRT)

-fB

and

PH yBF lR T

r*)-(!&)

3.3. Low pressure purge


During the constant pressure purge step, eq. (13)
indicates that y, is constant along each characteristic.
The characteristic velocity, given by eq. (12) is minimized when yA1 = 0 so that other characteristics
diverge from those of the influent. Therefore, a simple
wave is formed, as discussed earlier, and the analysis of
the purge step involves the integration of eq. (12) to
obtain the purge step duration:
t,=

and subsequently

the number of moles required:

s
1L

N,=

(24)

WAoUL

PL

.z A,,uLFTdt

~4s PL

= ~
BAA RT

(25)

them

m3

equals

the linear iso_

case: m1 = m2 = K,,

while m3 = KB.

3.4. Product recovery


The most significant scale of performance of a PSA
system is the recovery of the less strongly adsorbed
component. Notwithstanding,
the purity of that product is also significant, along with the productivity (i.e.
the flow rate per unit mass of adsorbent). These,
however, depend heavily on dissipative effects that
may influence the separation, e.g. at high flow rates or
short cycle times. Therefore, the latter scales of performance must generally be evaluated experimentally.
The pure component B recovery is:
R=

where PA0= PAlyA= o.

Note that the purge requirement is dictated by the


Henrys law slope of the isotherm of the more strongly
adsorbed component. For many adsorbates (e.g. those
represented by Freundlich isotherms) this term can be
extremely large and may be sufficient to preclude
complete purging. In fact, a shift to a higher temperature may then be necessary, leading to a combined
temperature and pressure swing cycle.
As implied earlier, the analysis of the purge step is
independent of the initial column contents as long as
y, > 0 at the purge inlet prior to purge. Consequently,
detailed analysis of the depressurization
step is not
needed. If partial purge is desired, however, analysis of
the depressurization
step may be needed in order to

slope
For

HP-NL-NPR

c1 -YYAF)NH

For conditions in which isotherm curvature is slight, it


is possible to combine eqs (20), (22), (23), (25) and (26)
to get an explicit expression for product recovery. For
linear isotherms, those equations reduce to the experimentally verified recovery equation of the BLI theory
that applies for pressurization with product.
4. CASE
4.1.

STUDYl

SEPARATION

OF OXYGEN

FROM

AIR

Basis and approach

To gain insight into the properties of eq. (26), the


system of nitrogen (A) and oxygen (B) on zeolite 13X at
O.OC is considered. For simplicity, the feed is assumed
to be free of argon and other minor components. As a

JOHN

and KENT

CKAYSER

result, the feed is comprised of 79 mole% nitrogen and


21 mole% oxygen.
Both isotherms have been measured using a manometric technique and are presented in Fig. 4. The
nitrogen isotherm data were correlated with a second
order polynomial that satisfies Henrys law at the
origin and fits the data well over the entire range of
measurement (i.e. up to about 6 atm). The oxygen
isotherm data were fitted to a linear equation. Neither
equation should be extrapolated much beyond the
region of measurement since neither exhibits the saturation plateau known to exist for these gases on
zeolites. The isotherm relationships are:

SKNAEBEL

velocities. Specifically, Kayser and Knaebel(l986) and


Matz and Knaebel (1988) found close agreement for
separation of oxygen from air with zeolite 5A. They
used adsorbent productvities of up to about 0.04 m3
(0, at STP)/kg (adsorbent) h and column L/d ratios of
about 25 to 1.
Before estimating the product recovery, it is necessary to ensure that yAF = 0.79 = yAL2and yA,_ = 0 = yAl
satisfy eq. (16) or (18). Based on eqs. (9), (12), (27) and
(28), it is possible to evaluate the mole fraction, yX , at
which the characteristic velocity is maximized.
yx

(K.;K;)RT+

= f
[.

(27)

nB =

PYB

KB=

with Tin K, P in atm, R = 82.057 mol cm3/atm K, and


ni has units of mol/cm3 (solid). At O.OC, we have found
Ke = 4.7542
K, = 15.013,
M, = - 2.457
and
x lo4 cm3/mol. In addition, to obtain the values of
K*, K, and M,, a particle density of 1.50 g/cm (solid)
and a bulk density of 0.780 g/cm were used. Consequently, the bed voidage used in the calculations was
0.480, consistent with the particle and bulk densities.
Note that the bed voidage incorporates the interstitial
voids and at least some of the intraparticle pores. It
can be shown that, when consistent values of particle
density and porosity are employed in the material
balance equations and in evaluating the isotherm
parameters, these individual values have no ultimate
effect on the results. Only the value of bulk density,
which is easily measured, affects the outcome.
It should be noted that previous experimental
studies have shown that the equilibrium theory appears to be valid for low to moderate interstitial

P
0.0

I.

12

i=
e
$

(30)

If the maximum allowable pressure is 6.0 atm due to


limited equilibrium data, then yX = 0.854 which is
greater than yAF = 0.79. A pressure of 6.84 atm is
necessary to maximize the characteristic velocity for
yX = 0.79. Therefore, pressures below 6.84 atm (with
y,, = 0.79) satisfy the restriction of moderate nonlinearity for these components, and uniqueness of the
shock wave is assured.
In this case study both the purge step pressure, P,,
and the feed step pressure, P,, have been varied
systematically in order to examine their effect on
recovery. As either P, or P, decreases, the operation
moves toward the linear region of the isotherms and
the recovery approaches the limit given by the BLI
theory. Consequently, recovery based on BLI theory
for the Henrys law slopes of the isotherms is also
considered. The calculated recoveries as a function of
pressure ratio are presented in Figs 5 and 6.
Discussion of case study results
The effects of curvature of the nitrogen isotherm
have been examined for the PSA cycle described in
Figs 1 and 2. Since the nitrogen adsorption isotherm
exhibits significant nonlinearity under these conditions, the effects on the performance of the PSA

4.2.

(atm)

2.24

3.36

4.48

5.60

6.72

0.003

x
E
s
4
;;
9

I-O
1
BLI
P

0.6

0.002

THEORY

1.0

0.001

i*

101

0.000
0.0000

0.0001
P

vi/

0.0002
T

0.0003

lo2
=

PL

[mcl/cucm(s))

Fig. 4. Isotherm data and correlations for oxygen and


nitrogen on zeolite 13X at O.OC.

Fig. 5. Recovery of oxygen from air at O.OC as a function of


pressure ratio for various values of Pa, with P, approaching
zero as a limit.

Pressureswing adsorption

0.2

0.0

10

lOI
P

lo2
-

lo3

PH / PL

Fig. 6. Recovery of oxygen from air at O.OCas a function of


pressureratio for various values of P,, with P, up to 6.8 atm.
process are dramatic. Based on the resvlts shown in
Figs 5 and 6, several observations are made that are in
contrast to observations based on linear isotherms:

(4 At a fixed pressure ratio, recovery of light product

(b)

w
(d)
(e)

invariably decreases as either the purge step pressure or the feed step pressure is increased, whereas
for linear isotherms recovery remains constant.
Larger pressure ratios are required to ensure net
positive recovery as the feed step pressure is
increased, while for linear isotherms the minimum
pressure ratio that yields positive recovery is
constant.
The maximum
recovery obtained as P,
approaches 0 decreases with increasing feed step
pressure, though for linear isotherms it is constant.
There is a maximum purge step pressure beyond
which positive recovery is impossible, regardless
of the pressure ratio.
For a low to moderate purge step pressure, there is
a pressure ratio at which recovery is maximized
that corresponds to a feed step pressure well below
the saturation limit.

maximum recoveries are obtained at an infinite pressure ratio (i.e. P, = 0), but the maxima decrease as P,
increases.
An analytical tool has been developed for understanding simple PSA cycles with nonlinear isotherms.
The insight gained may substantially affect the way
the cycles are designed and operated. Notwithstanding, the case study is based on a simple PSA cycle
that is probably not the optimum for product recovery. It is recognized, therefore, that the observations derived from the case study cannot be directly
transferred to other PSA cycles incorporating more
complicated steps (e.g. partial purge and/or varying
pressure during the feed step), or to applications with
significantly different adsorbent/adsorbate
interactions. Despite that, it is hoped that the methods and
approach will be applicable in diverse conditions and
adaptable to even more complex situations.

NOTATION

-43

fi
9

Ki
L

MA
ni

Nj
P

PH
PL
R
R
S
t
T

5.

CONCLUSIONS

The effects of isotherm nonlinearity in an equilibrium-based model of a specific PSA cycle have been
examined. The resultant theoretical equations are
algebraic and explicit. In addition, the equations
reduce to the experimentally verified equations of the
binary linear isotherm theory (Knaebel and Hill, 1985;
Kayser and Knaebel, 1986).
The theory is applied in a case study to the separation of oxygen from air with zeolite 13X at O.OC. The
structure of Figs 5 and 6 demonstrates that the farther
the purge step pressure (PL) or the feed step pressure
(P,) is removed from the linear region, the lower the
recovery for a given pressure ratio. Moreover, when
P, is fixed there is a pressure ratio that maximizes
recovery. Conversely, when P, is fixed, higher pressure ratios are required to obtain positive recovery, and

IL
uj

Yi
Yij

Y.2
Z
Greek

cross-sectional area of bed, cm


function of concentration of species i, mol/cm3
(solid)
first derivative of J with respect to concentration of species i
pertains to gas phase
slope of equilibrium isotherm of component i in
Henrys Law region
length of packed bed, cm
second order coefficient for equilibrium isotherm of component A, cm3/mol
moles of component i adsorbed per volume of
solid, mol/cm3 (solid)
moles of gas in stream j during a step
pressure, atm
pressure of feed step, atm
pressure of purge step, atm
gas constant, 82.057 mol cm3/atm K
recovery of light product, ratio of moles of B
produced to moles of B fed
pertains to solid phase
time, s
temperature, K
interstitial velocity, cm/s
interstitial velocity at location j, cm/s
shock-front velocity, cm/s
mole fraction of component i
mole fraction of component i at location j
mole fraction of gas with maximum characteristic velocity
axial position within adsorbent bed, cm
letters

separation

Bi

therms, flA/BB
column isotherm parameter of component i, E/(E

&

factor based

on tangents

of iso-

+(I --E)fi)
void fraction of adsorbent bed
separation factor based on chords of isotherms,
8, iea

JOHN C. KAYSER and KENT S. KNAEBEL

8
ei
*
y*r

1
l+
I[
integral
(17)

l--E.&-&

RT

& Yi*-Yil
pH
characteristic
velocity,

defined

by eq.

REFERENCES
Basmadjian, D., and Coroyannakis, P., 1987, Equilibrium
theory revisted. isotherm fixed-bed sorption of binary
systems-I.
Solutes obeying the binary langmuir isotherm.
Chem. Engng Sci. 42, 1723-1735.
Chart, Y. N. I., Hill, F. B. and Wong, Y. W., 1981, Equilibrium
theory of a oressure swing adsorntion process.
Chem.
Engng Sci. 36; 243-251.
. _
Davis. M. M. and LeVan. M. D.. 1987, Eauilibrium theory
for complete adiabatic gdsorpGon c&es. A.I.Ch.E.
J. 35,
470-479.
Doong, S. J. and Yang, R. T., 1986, Bulk tieparation of
multicomponent gas mixtures by pressure swing adsorption: pore/surface diffusion and equilibrium models.
A.1.Ch.E.
J. 32, 397-410.
Flores Fernandez, G. and Kenney, C. N., 1983, Modelling the
pressure swing air separation process. Chem. Engng Sci. 38,
827-834.
Kayser, J. C. and Knaebel, K. S., 1986, Pressure swing
adsorption: experimental study of an equilibrium theory.
Chem. Engng Sci. 41,2931-2938.

Kayser, J. C. and Knaebel, K. S., 1988, Integrated steps in


pressure swing adsorption cycles. Chem. Engng Sci. 43,
3015-3022.
Knaebel, K. S. and Hill, F. B., 1985, Pressure swing adsorption: development of an equilibrium theory for gas separations. Chem. Engng Sci. 40, 2351-2360.
Matz, M. J. and Knaebel, K. S., 1987, Temperature front
sensing for feed step control in pressure swing adsorption.
Ind. Engng Chem. kes. 26, 1638-1645.
Matz, M. J. and Knaebel, K. S., 1988, Pressure swing
adsorption: effects of incomplete purge. A.I.Ch.E.
J. (in
press).
Oleinik, 0. A., 1963, Uniqueness and stability of the generalized solution of the Cauchy problem for a quasi-linear
equation. Am. Math. Sot. Transl. 33, 285290.
Rhee, H. K., Aris, R. and Amundson, N. R., 1986, First Order
Partial D@rential
Equations.
I. Prentice-Hall, Englewood
Cliffs, NJ.
Shendalman, L. H. and Mitchell, J. E., 1972, A study of
heatless adsorption in the model system CO, in He, I.
Chem. Engng Sci. 27, 1449-1458.
Underwood,
R. P., 1986, A model of a pressure swing
adsorption process for nonlinear adsorption equilibrium.
Chem. Engng Sci. 41,409--4ll.
Wankat, P. C., 1986, Large-Scale
Adsorption
and Chromatography,
I. CRC Press, Boca Raton, FL.
Yang, R. T., 1987, Gas Separation
by Adsorption
Processes.
Butterworths, Boston, MA.

Vous aimerez peut-être aussi