Page 199 - Pressure Swing Adsorption
P. 199
176
PRESSURE SWING ADSORPTION DYNAMIC MODELING OF A PSA SYSTEM 177
Table 5.2. (Co11tm11ed)
, oressurizatmn and blowctown, especially if the seoarat10n 1s equilibrium
8 9 12 5 17 21 and Shin and Knaebel 20 have
cohc_c,l1mt1on of lht! bed undergoing purge m terms of raffinate product com.-entrat,on and is not controlled. Sang and co~workers • · · 1 · ·
applicable for a self-purging cycle. used the exoenmenta\ly measured pressure-time history of the column (via a
Velocity boundary coriditions:
best fil eoualion) to account for !he changmg column pressure. The reai
pressunzat1on situation 1S obviously best represented by this approach. in 'the absence of
( !Oa)
=t·ou, high-pressure adsorption expenmental data, however, an approx1rnation 1s necessary. One approach is
(!Ob) to consider that the column pressure vanes either linearly or exoonentiallv
= 0, blowdown 22
(toe) over the penod of- the oressurizat10n or blowctown step. An alternative
= Gt·oH, purge; G = 0 for self-purging cvcle approach is to assume that the column pressure changes instantaneously with
(!Od)
the pressure change followed by mass transfer (at constant high or low
(II) pressure) between the gas and solid ohases. 26 The former:1s a good approxi-
., The additional vel?clly boundary condition at z = L allows the convenience of using the matmn for an equilibrium-controlled seoaratmn, while the latter 1s more
sc11~1e C{)l!ocat1011 · coeffic1ents for the velocuv gradient as for the concentration gradient m the appropriate for kinetic separations. Exoerimentally measured pressure pro-
fluid phase.
files for equilibrium~controlled air separation on 5A zeolite and kinetical-
Initial conditions: dean bed
iy controlled air separation on RS-10 (4A) molecular sieve are shown in
c,(z,O)~O; ii,(z,0)~0
(12) Figure 5.1.
saturated bed
(13) 5.1.2 Equilibrium Isotherms
"The sum of 1he m~le fractions of 11 components 1n the gas phase at every pomt in the bed is equal The pressure range of PSA operat10n often exceeds the pressure ran~e- over
10
one. Therefore solving for (n - I) components m the gas phase 1s sufficienl; the concermation of the which expenmental eQuilibnum data are avaiiable. Moreover, the mult1com-
r_emmmng component m the gas phase 1s otuamcd bv difference. Thili sci of equations applies 1or flow
lrom _ 7- 0 to L For flow from z - L to O the 0/ffz 1ermli hecomc ncguuvc, and the houndarv ponent equiHbna are commonly predicted from smgle-comoonent isotherm
cond111om; arc mterchanged. data. Reli-able models to represent both srngie and multicomoonent actsorp-
t10n equilibria are therefore an essential requirement.
Linear, Freundlich, and Langmuir isotherms have been used to define the
smgie-component adsorption in PSA purification processes. Although the
linear isotherm is the simplest eauilibnum model, even a slight curvature of
mtegration·. However, a full solution of the eauations retaining the second the isotherm influences the cyclic steady state of a PSA separation and
term 1~- Eq. 5.12 is very difficult and has not yet been attempted except in the should be considered. Since a PSA process involves both adsorot1on and
30
work of Munkvold et al. Several approximations have therefore been desorption at the same temperature, stmole oualitative· reasoning suggests
prooosect to simplify the solution m an acceptable way. that th.e fonn of the isothenn should not deviate too greatly from Iineanty,
3 7
The early PSA modeis - assume that the column pressure remains otherwlse either adsorption or desorption will become- unacceptably slow.
c_on~tant durmg the high- and low-pressure steps. It is further assumed that Moderate cmvature of the isotherm (either type I or type II of Brunauer··s
cturmg pressu~1zation and blowdown the solid phase remams frozen while the ciassificatton) 1s acceptable, but it is obviously important that the oortion of
gas phase undergoes a_ sauare wave change in pressure. (Dunng blowdown the isotherm over which the process operates should be completely re-
the mole fractions in the gas ohase remains the same as at the end of the versible. Any hysteresis, as occurs for example in the alum.ma-water system, 33
oreceding hig~-oressure step while the pressure 1s reduced; durmg pressur- (see Figure 2.5) will lead to an unacceptable buildup of the residuai conc~n-
1zat1~n the ~es1dual gas profile ts compressed so that it extends only through a tration m the adsorbed phase. In such a system PSA operat10n should be
fractional distance equal to the pressure ratio from the product end while the confined to the region below the point of inflectJOn, where the isothenn 1s
remamder o~ the bed is filled with feed.) These approxtmatmns are accept-
reversible.
able_ for punfication Processes operated on a Skarstrom cycle. The approxi- The Langmuir model provides a reasonably good fit for most type I
mation of constant column pressure cturmg the adsorption and desorption isotherms over a wide concentration range and for type 11 isotherms up to
s!eps aI_so holds for many bulk separation process cycles. However, the the inflectlon pomt. The Freundlich isotherm is also sometimes used, but,
change m pressure is not instantaneous, and in a bulk separation process it slnce 1t does not reduce to Henris Law. tt 1s likely to be less reliable 1n the
becomes important to allow for mass transfer between fluid and soiid during
low-concentration reg1on. In the simulation of PSA ourificatton processes
(