7208
Ind. Eng. Chem. Res. 2005, 44, 7208-7217
SEPARATIONS Parametric Study of Pressure Swing Adsorption Process To Purify Oxygen Using Carbon Molecular Sieve Min-Bae Kim, Jeong-Geun Jee, Youn-Sang Bae, and Chang-Ha Lee* Department of Chemical Engineering, Yonsei University, 134 Shinchon-dong, Seodaemun-gu, Seoul, 120-749, Korea
To produce O2 with a high purity of 99+% and a high productivity from various oxygen-rich feeds, a parametric study was done on a six-step pressure swing adsorption (PSA) purifier using carbon molecular sieve (CMS). The cyclic performances of the PSA process such as purity, recovery, and productivity were compared under nonisothermal conditions. To study the effects of N2 amount on the PSA purifier, various feeds with 90% O2 or more were experimentally and theoretically applied for the PSA process. Since N2 plays a key role in product purity, the maximum purity of the PSA was 99% O2 with 51.5% recovery from a higher nitrogen feed (O2:Ar:N2; 90:4:6 vol %) and 99.8% O2 with 56.9% recovery from a lower nitrogen feed (O2:Ar:N2; 95:4:1 vol %) within the experimental range. The adsorption step time and feed flow rate served as key operating variables in the purification of the oxygen-rich feeds because the concentration wave fronts of minor impurities such as N2 and Ar were controlled by kinetic selectivity. To produce 99% O2 purity from feeds with various amounts of N2, the optimum operating variables were set to maximize the recovery and productivity within the experimental ranges. A high feed flow rate accompanied by a short adsorption step time could increase both purity and productivity. Without any serious loss of recovery and productivity, the process could purify the feed with higher than 91% O2 to the product with higher than 99% O2. The nonisothermal model incorporating mass, energy, and momentum balance together with a concentration-dependent rate model could accurately predict the performance results. 1. Introduction There has been a great demand for O2 with a purity level of 99% or more in industry; for example, in the fields of welding and cutting processes, plasma chemistry, ozone generator, and breathing oxygen.1 As another application, due to stringent environmental regulations, incineration processes using pure oxygen have been widely used to prevent secondary pollutants in combustion processes. Furthermore, with the expansion of semiconductor industries, the demand for a highpurity O2 of over 99.8%, as a standard, is ever increasing. To satisfy such a demand, the pressure swing adsorption (PSA) or vacuum swing adsorption (VSA) processes for producing high-purity O2 were developed.2 Most patents adopt a two-stage process consisting of zeolite bed for N2 removal and carbon molecular sieve (CMS) bed for the purification of Ar and N2.3-6 And Knaebel and Kandynin5 have suggested that O2 product with 99.6% purity and a 25% recovery from a binary feed which has a composition of 95% O2 and 5% Ar could be obtained from a five-step PSA with two consecutive blowdown (BD) steps. Recently, the VSA patent7 using certain silver-exchanged zeolites, particularly sliverexchanged X-type zeolites, was issued to produce oxygen * To whom correspondence should be addressed. Tel: +822-2123-2762. Fax: +82-2-312-6401. E-mail:
[email protected].
at purities above 95%. However, due to the extremely low O2 recoveries and potentially higher cost of the adsorbent, economic analysis should be accompanied. In addition, since a cryogenic plant for air separation generally produces 99.6% O2, it needs to purify again to satisfy the demand of semiconductor industries for 99.6% or higher O2. Oxygen produced by an on-site process has to go through an additional purifier to reach a level of high purity. It is important to optimize the O2 PSA purifier to maximize productivity because the additional purifier, as a secondary adsorption unit, works against producing high-purity O2 gas at low cost.8,9 Ahn et al.10 presented the optimum operating conditions and carbon-to-zeolite ratio to obtain H2 with a 99.99% purity from different H2 mixtures, that is, with various amounts of N2, because N2, the weakest adsorbate in the feed, acted as a key component in producing high-purity H2 in the equilibrium separation. In the previous study,11 the four different PSA cycles were suggested to purify oxygen. However, to obtain the desired O2 purity without any severe decrease of recovery and productivity from various feed compositions, the effects of operating variables on the CMS PSA purifier needed to be studied. In this study, the effects of operating variables such as step times, feed flow rate, and adsorption pressure on the performance of O2 purification were investigated in the six-step two-bed PSA packed with CMS. In addition, the optimum operating variables to maximize
10.1021/ie049032b CCC: $30.25 © 2005 American Chemical Society Published on Web 07/29/2005
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005 7209
the recovery and productivity of 99+% O2 products from feeds with various amounts of N2 were studied through theoretical and experimental works. Mass, energy, and momentum balances incorporating a concentrationdependent rate model was employed to predict the experimental results of a kinetically controlled separation system using CMS.
The multicomponent adsorption equilibrium was predicted by the following loading-ratio correlation (LRC) equation,
qmiBiPini
qi )
1+
2. Mathematical Models To understand the dynamic behaviors of the CMS bed during the PSA running, the mathematical models were developed on the basis of the following assumptions: (i) the gas phase behaves as an ideal gas mixture, (ii) radial concentration and temperature gradients are negligible, (iii) thermal equilibrium between adsorbents and bulk flow is assumed, (iv) the flow pattern is described by the axially dispersed plug flow model, and (v) the pressure drop along the bed is considered by using the Ergun’s equation. The assumption of neglecting radial gradient was widely accepted by numerous studies,12,13 and the others are also common assumptions in simulating the adsorption processes.14-17 The component and overall mass balances for the bulk phase in the adsorption column are given by
∂2ci ∂(uci) ∂ci 1 - ∂qi + + Fp )0 -DL 2 + ∂z ∂t ∂t ∂z
(
∂(uC)
+
∂C
∂z
∂t
+ Fp
( )∑ 1-
n
∂qi
i)1
∂t
)
(1)
)0
(2)
Another characteristic of the adsorption process is the temperature variation caused by the heat of adsorption and desorption. 2
-KL
∂T ∂z2
+ FgCpg
∂(uT) ∂z
∂T
+ (tFgCpg + FBCps) ∂qi
n
FB
(- ∆Hi) ∑ ∂t i)1
+
2hi RBi
where Dei/Rp2 ) Ci‚Pri0.5(1 + BiPi)2 This rate model depicts the adsorption rate as a function of sorbate concentration of solid phase. And it showed good agreement with the Maxwell-Stefan diffusion model for the single file diffusion mechanism.20 The adsorption isotherm and rate parameters of N2, Ar, and O2 on CMS are also shown in Table 1, which were presented in the previous studies.11,18,19 And these values are similar to the published data.21-23 Equations 1-4 were reduced to dimensionless forms by introducing the following dimensionless variables and parameters:
Qi )
ci qi C u P , Yi ) ,Γ) ,U) ,P) , q0 CH CH uH PH
(T - Tw) ) 0 (3) Pem )
2
(1 - ) 150 (1 - ) , b ) 1.75 2 2 2Rp 4Rp
A)
(5-2)
Tw T , Θw ) T0 T0
uHt z ,Z) L L
u HL uHLFgCpg , Peh ) DL KL
(1 - )Fpq0 tFgCpg + FBCps ,B) , CH FgCpg F Bq 0 2hiL D′ ) ,E) FgCpgTini RBiuHFgCpg F)
2πRBihiL 2πRBohoL ,G) FwCpwAwuH FwCpwAwuH
The resulting dimensionless component and overall mass balance equations are
(5-1)
(7)
p
τ)
where Aw ) π(RBo2 - RBi2) To consider the pressure drop effect across the bed, Ergun’s equation was introduced as a momentum balance.
where u is the interstitial velocity.
KDei ∂qi ) ωi(qi* - qi), ωi ) ∂t R2
-
∂Tw ) 2πRBihi(T - Tw) FwCpwAw ∂t 2πRBoho(Tw - Tatm) (4)
a)
j
where qmi ) k1 + k2 × T, Bi ) k3 exp(k4/T), ni ) k5 + k6/T. In this study, the modified LDF model based on the surface diffusion model was applied for the sorption rate by using the following concentration-dependent diffusivity combined with the Langmuir isotherm.11,18,19
∂t
dP ) aµu + bFu|u| dz
BjPjn ∑ j)1
Θ)
where t is the total void fraction () + (1 - )R), and FB is the bed density ()(1 - )Fp). To consider heat loss through a wall and heat accumulation in the wall, another energy balance for the wall of the adsorption bed was used.
-
(6)
n
2 Qi 1 ∂ Yi ∂(UYi) ∂Yi + +A )0 + 2 Pem ∂Z ∂Z ∂τ ∂τ
∂(UΓ) ∂Z
+
∂Γ ∂τ
n
∂Qi
∑ i)1 ∂τ
+A
)0
(8)
(9)
Assuming that the ideal gas law holds (i.e., C ) P/RT),
7210
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005
the above overall mass-balance equation becomes
-
U ∂Θ
+
U ∂P
Θ ∂Z
P
+
∂Z
∂U
1 ∂Θ
-
Θ ∂τ
∂Z
1 ∂P
+
P
Table 1. Equilibrium/Rate Parameters and Heats of Adsorption of O2, Ar, and N2 for CMS11 Equilibrium Parameters (Langmuir-Freundlich Model)
+
∂τ Θ
A
P
-
+
∂(UΘ)
Peh ∂Z2
∂Θ
+B
∂Z
n
∂Qi
∑ i)1 ∂τ
) 0 (9-1)
- D′
(-∆Hi) ∑ ∂τ i)1
+
E(Θ - Θw) ) 0 (10) ∂Θw ) F(Θ - Θw) - G(Θw - 1) ∂τ
(11)
The boundary and initial conditions of mass and energy balances are presented below. The well-known Danckwerts boundary conditions were applied.24 • Boundary conditions for feed pressurization (PR), adsorption (AD), and cocurrent purge (PU) steps:
-
-
( )|
) PemU(Yi|Z)0- - Yi|Z)0+);
(∂Θ ∂Z )|
) PehU(Θ|Z)0- - Θ|Z)0+);
∂Yi ∂Z
Z)0
Z)0
( )| ∂Yi ∂Z
(∂Θ ∂Z )|
Z)L
Z)1
)0
(12-1) )0 (12-2)
where Yi|Z)0- means dimensionless feed composition for component i. • Boundary conditions for idle (ID), first blowdown (BD1), and second blowdown (BD2) steps
( )| ∂Yi ∂Z
Z)0
)
( )| ∂Θ ∂Z
Z)0
) 0;
( )| ∂Yi ∂Z
)
Z)1
( )| ∂Θ ∂Z
Z)1
)0 (13)
• Initial condition for fluid flow
Yi(Z,0) ) 0; Qi(Z,0) ) 0
(14)
In this study, the adsorption beds were initially vacuumed up to 10-3 mmHg. • Initial condition for heat flow
Θ(Z,0) ) T0
Ar
N2
k1 × k2 × 105 (mol/g‚K) k3 × 104 (1/atm) k4 (K) k5 k6 (K) heat of adsorption, - ∆H h (cal/mol)
15.2675 -3.230 22.900 966.09 1.1869 -106.0 3300
20.4150 -5.30 239.7 324.55 1.646 -238.2 3400
23.6266 -6.38 361 1443.8 1.6916 -270.0 3200
Adsorption Rate Parameters
∂Qi
n
∂τ
O2
103 (mol/g)
And both energy balance equations become
1 ∂2Θ
equilibrium constants
(15)
In this study, the pressure history during a PSA experiment at the bed end was measured. These data were fitted by polynomials and used as a boundary condition for the overall mass balance. The gPROMS modeling tool developed by Process Systems Enterprise Ltd. was used to obtain the solution of dynamic simulation from the above model. Details of the simulating procedures used are described in the previous work by Jee et al.11,25 3. Experimental Section A schematic diagram of the two-bed PSA unit is shown in Figure 1. The adsorption beds were made of
constants for rate model
O2
Ar
N2
Ci (s-1)
0.024
0.000047
0.000090
stainless steel pipe with a length of 100 cm, i.d. of 2.2 cm, and thickness of 0.175 cm. The beds were packed with CMS supplied by the Takeda chemical company. The characteristics of the adsorbent and adsorption bed are listed in Table 2. Three calibrated resistance temperature detectors (RTD, Pt 100Ω) were positioned at 10, 50, and 80 cm from the feed end in order to measure temperature variations inside the bed. Two pressure transducers were located at the feed and bed ends in order to measure bed pressure variation. The feed flow rate was controlled by a mass flow controller (Hastings, 202D799). And a surge tank was equipped to prevent flow fluctuation. The total amounts of feed flow and the flow rate of each step were measured by a wet gas meter (Sinagawa Co. W-NK-1B). To keep the pressure in the adsorption bed constant, an electric back pressure regulator was installed between the adsorption bed and the product bed. The concentration variations of the effluents at the adsorption and desorption steps were analyzed by two portable oxygen analyzers (Teledyne Analytical Instruments, 320B/RC-D) and a mass spectrometer (Balzers, QME 200). The system was fully automated by personal computer with a developed control program and all measurements including pressure, temperature, and O2 purity were saved on the personal computer through the AD converter. Two different ternary mixtures, O2/Ar/N2 (lower nitrogen feed - 95:4:1 vol % and higher nitrogen feed 90:4:6 vol %), were used as feed gases for the PSA experiments. These gases were supplied by Daesung Industrial Gases Co. Ltd., Korea. And the concentration of each ternary mixture gas was confirmed by a mass spectrometer as well as an oxygen analyzer. The adsorbent was regenerated at 423 K for 12 h. The bed packed with the adsorbent was kept at 1.5 atm with pure O2 (99.99+%) to prevent contamination from outside air. Prior to each experimental run, the adsorption bed was evacuated to the level of 10-3 mmHg for 2 h. The adsorption pressure was in the range of 4-6 atm and the feed flow rate in the range of 2-6 LSTP/min. The temperature of feed and surroundings were kept in the range of 296-297 K. The more detailed operating conditions are shown in Table 3. 4. Description of the PSA Process In this study, the six-step two-bed PSA process was studied to obtain the high purity of 99+% from O2-enriched feeds in the range of 90-95%. Table 4 shows the cyclic sequence of the PSA process.6 This process consisted of conventional PSA steps such
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005 7211
Figure 1. Schematic diagram of apparatus for a two-bed PSA process. Table 2. Characteristics of Adsorbent and Adsorption Bed adsorbent
CMS
type micropore diameter pellet size pellet density, Fp [g/cm3] particle porosity, R bed density, FB [g/cm3]
pellet 3Å 6-12 mesh 0.90 0.300 0.633
Adsorption Bed length, L [cm] inside radius, RBi [cm] outside radius, RBo [cm] heat capacity of column, Cpw [cal/g‚K] density of column, Fw [g/cm3] internal heat-transfer coefficient, hi [cal/cm2‚K‚s] external heat-transfer coefficient, ho [cal/ cm2‚K‚s]
100 1.1 1.275 0.12 7.83 9.2 × 10-4 3.4 × 10-4
as pressurization (PR), adsorption (AD), blowdown (BD), and purge (PU). The idle (ID) step was included to keep cyclic symmetry because the BD step is inevitably fast. The flow direction of a PU step was cocurrent, which is different from general O2 PSA, because the AD step worked as a step to remove the Ar and N2 impurities in the feed. Furthermore, in the process, the two-stage BD step (high to medium pressure - BD1, and medium to ambient pressure - BD2) was introduced in order to control purity and recovery. The effluent from the BD1 step followed after the AD step was used as a purge gas for the other bed. Then, oxygen product was produced at the BD2 step. Since the PR step time was longer than the BD2 step time in the other bed as shown in Tables 3 and 4, one bed was operated from the middle of the PR step to AD step during the ID step in the other bed. The bed pressure at the end of the BD1 step was mid-pressure between
the adsorption pressure and the ambient pressure. However, the bed pressure at the end of the BD2 step was at ambient pressure. To calculate the mean O2 purity through the BD2 step, the information of flow rate variation during the BD2 step time was inevitably needed. In this study, the BD1 step time was fixed at the point where accurate measurement of gas could be obtained at its minimum amount. The flow rate of the BD2 step was controlled by a metering valve to achieve the mid-pressure of the AD step. The flow rate history in the O2-producing BD step was measured by a wet gas meter under the condition of each PSA experiment.11 Then the measured flow rate variation during the BD2 step was applied to the following equation.
Y h O2
t)BDsteptime V‚YO dt ∫t)0 ) t)BDsteptime V dt ∫t)0 2
where V ) flow rate of the effluent and YO2 ) mole fraction of oxygen in product. 5. Results and Discussion 5.1. Effect of Operating Variables. The effects of operating variables such as adsorption pressure, feed flow rate, and step time on the performances of the PSA process are shown in Figures 2-4. Figure 2 shows the effect of the PR and AD step times on O2 purity and recovery using the same adsorption pressure and feed flow rate with a lower nitrogen feed (O2/Ar/N2; 95:4:1 vol %). As can be seen in Figure 2a, the longer the PR step times, the higher the O2 purity and the lower the O2 recovery. However, the variation
7212
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005
Table 3. Operating Conditions for PSA Experiments
a
Lower nitrogen feed (O2/Ar/N2 - 95:4:1 vol %), higher nitrogen feed (O2/Ar/N2 - 90:4:6 vol %).
Table 4. Cyclic Sequences of PSA Cycle6 (v: Cocurrent Flow; V: Countercurrent Flow)
of O2 purity and recovery was small, about 0.2% and 2%, respectively, within the PR step time range. Unlike the above result, the AD step time had a very crucial effect on O2 purity and recovery, as shown in Figure 2b. Since the AD step worked as a step to remove the impurities, a longer AD step time led to an increment in the amount of O2 at the adsorbed phase while the shorter AD step time significantly affected the O2 purity. Therefore, high-purity O2 desorbed from the solid phase could be produced at the BD2 step followed by the BD1 step. However, an excessive AD step time of over 15 s caused a steep decrease in the recovery rate without any increase in purity because a larger amount of O2 was vented without further adsorption on the CMS. Figure 3 shows the effects of adsorption pressure and feed flow rate on O2 purity and recovery with a lower nitrogen feed. As shown in Figure 3a, the increased adsorption pressure had a weak effect on the O2 purity while it led to an increase of O2 recovery. This indicates that the higher the adsorption pressure, the larger the adsorbed amount of O2 without an increase of kinetic selectivity between the O2 and impurities. Since increased adsorption pressure caused the increased partial pressure of impurity at the bed end, it led to a slight decrease in purity. Therefore, the system was more
favorable to lower adsorption pressure with respect to product purity. The feed flow rate had a great effect on process performance, as shown in Figure 3b. The results are very similar to those showing the effect of AD step times in Figure 2b. The prolonged feed flow rate led to an improvement in the removal of impurities because of the short contact time between the impurities and the CMS and, as a consequence, there was a favorable removal of impurities from the gas phase of the bed. From the results of Figures 2 and 3, it is clear that the step time and feed flow rate at the AD step are the main operating variables to control purity and recovery within the experimental range. Moreover, in the case of feed conditions with 95% or more O2, the product with 99.8+% O2 could be simply generated with a 50+% recovery. In the case of the higher nitrogen feed (O2/Ar/N2; 90: 4:6 vol %), based on the results shown in Figures 2 and 3, the purity and recovery of O2 product were examined according to the variation of main operating variables such as step time and feed flow rate at the AD step. As can be seen in Figure 4a, when the higher nitrogen feed was used with an AD step time of 15 s, the O2 purity reached no more than 97.3%, unlike the result obtained from the lower nitrogen feed in Figure 2b. However, when the AD step time was extended to 30 s with the flow rate fixed at 4 LSTP/min, the purity increased to 98% while the recovery decreased from 55% to less than 50%. The effect of feed flow rate on O2 purity in Figure 4b was more significant than that of AD step time in
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005 7213
Figure 2. Effect of (a) PR and (b) AD step times on the O2 purity and recovery at 5 atm and 4 LSTP/min (lower nitrogen feed).
Figure 3. Effect of (a) adsorption pressure and (b) feed flow rate on the O2 purity and recovery (lower nitrogen feed).
Figure 4a, showing a similar change of recovery in both cases. At the feed flow rate of 6 LSTP/min with an AD step time of 15 s, O2 purity increased to 99% accompanied by a 50+% recovery. These results imply that the control of contact time between N2 and the CMS is more effective in improving the purity and recovery of O2 product than the elongation of the AD step time to remove N2 from the kinetic separation bed. 5.2. Adsorption Dynamics by Feed Composition. Figure 5 shows the profiles for gas-phase O2 mole fraction, adsorbed phase O2 loading, and axial temperature along the bed at various AD step times and feed flow rates under lower nitrogen feed condition. Figure 5a shows that the gas-phase O2 mass-transfer zone (MTZ) nearly reached the bed end at the AD step time of 20 s. Similarly, the adsorbed phase O2 loading increased when the AD step time was extended. However, the adsorbed phase O2 MTZ shows a broader shape than the gas-phase O2 MTZ in the AD step time range of 15-30 s because of mass-transfer resistance. As a result, though the gas phase was nearly saturated with O2 during the AD step time of 20 s, the adsorbed phase was not fully saturated with O2 until the end of that step time. The temperature profile of the AD step time of 15 s then showed a crossover with the others at the near bed end because of an increase in the adsorbed phase O2 loading at the bed end. Also, the temperature excursion at the near feed end caused the small excursion of gas and adsorbed phase O2 MTZ. The increment of the feed flow rate, as shown in Figure 5b, had a greater effect on adsorption dynamics
in comparison with the AD step time. With a 2 LSTP/ min feed flow rate, both gas and adsorbed phases were noticeably contaminated with impurities even at the end of the AD step and this coincides with the results shown in Figure 3b. However, although the partial pressure of the impurities was high, the O2 MTZ from the adsorbed phase was relatively less contaminated due to the low adsorption rate of the main impurity, Ar, in the lower nitrogen feed. That is to say, both MTZs in the adsorbed and gas phases showed a crossover at the end of the bed. In the case of low feed flow rate, the temperature sharply decreased near the bed end and the crossover of the temperature profile with the other profiles occurred at the middle of the feed end. As shown in Figure 6, the higher the amount of nitrogen in the feed, the slower the moving velocity of O2 MTZ during the AD step because of competitive adsorption. In Figure 6a, the contamination of gas and adsorbed phases is shown to be severe at an AD step time of 15 s in comparison with the phases shown in Figure 5a. Furthermore, even at an AD step time of 20 s, the impurities were not sufficiently excluded from the gas phase because the increased amount of N2 in the feed gas had a detrimental effect on O2 adsorption. Then, compared to Figure 5a, an earlier crossover of axial temperature profiles occurred due to a relatively small adsorption of O2. Figure 6b clearly shows the effect of the feed flow rate on the MTZ. The impurities, mainly N2, were not sufficiently removed from the adsorption bed even at 4 LSTP/min. In addition, the crossover of the MTZs in the
7214
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005
Figure 5. Effect of (a) AD step time and (b) feed flow rate on the gas and adsorbed phase O2 concentration and axial temperature along the bed at the end of AD step in the condition of (a) 5 atm and 4 LSTP/min and (b) 5 atm and 15 s AD step time (lower nitrogen feed). Figure 4. Effect of (a) AD step time and (b) feed flow rate on the O2 purity and recovery (higher nitrogen feed).
gas and adsorbed phases does not occur under each condition, unlike the results presented in Figure 5b. Since the kinetic selectivity between O2 and N2 was much smaller than that between O2 and Ar, the competitive adsorption between O2 and N2 by the high partial pressure of N2 led to a significant decrease of O2 MTZ. Compared to Figure 5b, the temperature increase was broader and the earlier crossover of the axial temperature was shown in Figure 6b. Therefore, a higher feed flow rate was required to purify the impurity efficiently as the amount of N2 in the feed increased. 5.3. Optimization of O2 PSA Purifier. Because AD step time, feed flow rate, and amount of N2 impurity had a significant effect on O2 purity, recovery, and productivity, the simulated results for optimizing the three main operating variables are presented in Figures 7-9. Figure 7 shows the simultaneous effect of AD step time and feed flow rate on O2 purity and recovery in the higher nitrogen feed. As shown in Figure 7a, O2 purity was drastically changed by the operating variables, which were at the condition of a feed flow rate of slower than 4 LSTP/min and an AD step time of shorter than 20 s. The effect of feed flow rate on O2 purity was prominent through all AD step times while AD step time had relatively little effect with a feed flow rate higher than 6 LSTP/min. But as the feed flow rate decreased to 1 LSTP/min, the effect of AD step time on O2 purity became more pronounced. Figure 7b shows almost linear and steep decrease of the recovery with an AD step time
of higher than 20 s and feed flow rate of higher than 5 LSTP/min. The effect of feed flow rate on O2 recovery was prominent through all AD step times while the effect of AD step time was not noticeable with feed flow rates up to 3 LSTP/min. From the above results, it is clear that the AD step time and feed flow rate showed a similar effect on performance because these two operating variables led to a significant change of O2 loading on the solid phase at the AD step. However, an increased AD step time would result in a decrease in productivity because the nonproducing step time (AD step) and total cycle time were prolonged. Also, the increased feed flow rate would lead to a decrease in recovery because of the bypassing of O2 due to an insufficient contact time between adsorbate and adsorbent. The amount of N2 in the feed has a significant effect on the operating conditions to obtain the desired level of purity from feeds with various amount of N2. Therefore, it was important to find the optimum operating variables within the experimental range to obtain maximum recovery and productivity. Figure 8 shows the AD step times and feed flow rates needed to produce O2 with a level of 99% purity from feeds with various amounts of N2. Adsorption pressure and other step times were fixed at the base operating condition presented in Table 3 and the amount of Ar in the feed was also fixed at 4 vol %. Depending on the increase of the amount of N2 in the feed, the required feed flow rate for O2 with a purity level of 99% decreased steeply with an increase in the AD step time. As the AD step time increased, the difference among the required feed flow rates for each feed decreased. However, in the case of the feed with
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005 7215
Figure 6. Effect of (a) AD step time and (b) feed flow rate on the gas and adsorbed phase O2 concentration and axial temperature along the bed at the end of AD step in the condition of (a) 5 atm and 4 LSTP/min and (b) 5 atm and 15 s AD step time (higher nitrogen feed).
6% N2, the required feed flow rate abruptly increased at the condition of short AD step time in comparison with the feed with 5% N2. This figure implies that there is a minimum feed flow rate if using a fixed AD step time to obtain O2 with 99% purity from various feeds with less than 6% N2. Figure 9 shows the effect of the amount of N2 impurity in the feed, under the condition of obtaining 99% O2 product, on recovery and productivity at each AD step time and feed flow rate as determined by the results shown in Figure 8. Figure 9a shows that an increased AD step time accompanied by a pertinently adopted feed flow rate linearly increases the O2 recovery rate under all feed conditions except 6% N2. This is because the prolonged contact time between the gas and solid phases increased the net O2 loading of adsorbent. However, as the amount of N2 impurity in the feed gas increased, recovery decreased under all operating conditions. Since a high feed flow rate from Figure 8 should be applied to the higher nitrogen feed, a decrease in recovery was more significant with 6% N2. As a consequence, a certain amount of O2 was vented with impurities at the AD step. In contrast, as shown by the steep slope in Figure 9b, productivity linearly decreased as the AD step time increased. This is because the increased AD step time with a low feed flow rate caused an increase in the total cycle time by an increase of nonproductive step time as well as small increase in the amount of feed treatment. Figures 9a-1 and 9b-1 show the maximum recovery and productivity resulting from each feed composition in the simulated range. As the amount of N2 impurity in the feed gas increased, the maximum recovery decreased, but the difference between the lower nitrogen
Figure 7. 3-D plots of (a) O2 purity and (b) O2 recovery to the variation of AD step times and feed flow rate (higher nitrogen feed).
Figure 8. AD step time and feed flow rate to obtain 99% purity oxygen at 5 atm adsorption pressure and various feed conditions. (All of the step times except AD step were the same as base condition.)
feed (N2: 1%) and the higher (N2: 6%) was less than 3%. It is because the low feed flow rate was applied for
7216
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005
the process performance with regard to purity, recovery and productivity. The PSA process could produce O2 with a purity level of 99% from the higher nitrogen feed (O2:Ar:N2; 90:4:6 vol %) and a purity level of 99.8% O2 from the lower nitrogen feed (O2:Ar:N2; 95:4:1 vol %) without any great loss in recovery. Given that the adsorption rate of the N2 on the CMS was faster than that of Ar, N2 was the main impurity regardless of the amount of nitrogen in a feed. Moreover, although the adsorption rate of N2 is much slower than that of O2, the competitive adsorption of both components occurred at the bed end during the AD step due to a high partial pressure of N2. Therefore, for O2 with a purity level of 99% from feeds with various amounts of N2, optimizing the AD step time and feed flow rate worked to maximize the recovery and productivity within the experimental ranges. A prolonged AD step time with the slow feed flow rate could maximize O2 recovery while a reduced AD step time with the high feed flow rate could maximize productivity. Although the amount of N2 impurity in the feed increased by up to 5%, the level of recovery and productivity under optimized operating conditions for each feed did not severely decrease. However, in the case of the feed with 6% N2, recovery and productivity significantly decreased in order to produce O2 with a purity level of 99%. Thus, with respect to the operating cost and efficiency, it is strongly recommended that feed with less than 6% N2 be used for the purification process to produce O2 with a purity level higher than 99%. Figure 9. Effect of N2 impurity in feed gas on the (a) O2 recovery, (a-1) maximum recovery, (b) O2 productivity, and (b-1) maximum productivity applied AD step times. (The other operating variables, used in simulation run, were fixed to those of base condition.)
the AD step time of 30 s - as shown in Figure 8 - and the net amount of product from the BD step was nearly the same under each feed condition. This means that the difference between productivity results under these conditions was small. However, as shown in Figure 9b-1, the difference between the maximum productivities for each condition was relatively large in comparison with the results for the other AD step time. Consequently, a high feed flow rate accompanied by a short AD step time can increase both purity and productivity without any serious loss of recovery. Moreover, with operating cost and process efficiency in mind, it is strongly recommended that the feed with less than 6% N2 be used in the process to produce O2 with a purity of higher than 99%. 6. Conclusions A parametric study of an O2 PSA purifier using a CMS was performed to obtain oxygen with a level of 99+% purity together with a high degree of recovery and productivity from various feeds. Since the concentration wave fronts of minor impurities such as N2 and Ar were controlled by kinetic selectivity on the CMS, the PR step time had little effect on O2 purity and recovery. In addition, the increased adsorption pressure led to a small decrease of O2 purity by about 0.2% but an increase in O2 recovery by higher than 10% in the operating range. The concentration wave front velocity and MTZ shape of the impurities were significantly affected by the AD step time and feed flow rate. As a result, the feed flow rate and AD step time worked as the main operating variables to improve
Acknowledgment This research was supported by a grant (AB2-101) from Carbon Dioxide Reduction & Sequestration Research Center, one of the 21st Century Frontier Programs funded by the Ministry of Science and Technology of Korean government. Nomenclature AW ) cross-sectional area of the wall (cm2) B ) equilibrium parameter for Langmuir-Freundlich model (atm-1) ci ) i component concentration in bulk phase (mol/cm3) Cpg, Cps, Cpw ) gas, pellet, and wall heat capacity, respectively (cal/g‚K) De ) effective diffusivity defined by solid diffusion model (cm2/s) DL ) axial dispersion coefficient (cm2/s) hi ) internal heat-transfer coefficient (cal/cm2‚K‚s) ho ) external heat transfer coefficient (cal/cm2‚K‚s) -∆H h ) average heat of adsorption (cal/mol) k ) parameter for Langmuir and LRC models K ) proportionality parameter for LDF model KL ) axial thermal conductivity (cal/cm‚s‚K) L ) bed length (cm) n ) equilibrium parameter for Langmuir-Freundlich model P ) total pressure (atm) Peh ) Peclet number for heat transfer Pem ) Peclet number for mass transfer Pr ) reduced pressure P h ) P/PH q, q*, q j ) amount adsorbed, equilibrium amount adsorbed, and average amount adsorbed, respectively (mol/g) n q0 ) ∑i)1 qmi(T0) qm ) equilibrium parameter for Langmuir-Freundlich model (mol/g)
Ind. Eng. Chem. Res., Vol. 44, No. 18, 2005 7217 Qi ) q j i/q0 R ) gas constant (cal/mol‚K) Rp ) radius of pellet (cm) RBi, RBo ) inside and outside radius of the bed, respectively (cm) t ) time (s) Tatm ) temperature of atmosphere (K) T, Tw ) pellet or bed temperature and wall temperature, respectively (K) u ) interstitial velocity (cm/s) v ) superficial velocity (cm/s) yi ) mole fraction of species i in gas phase Yi ) ci/CH z ) axial distance in bed from the inlet (cm) Z ) z/L Greek Letters R ) particle porosity ,t ) voidage of adsorbent bed and total void fraction, respectively Γ ) C/CH Θ ) T/T0 Θw ) Tw/T0 Fg, Fp, FB, Fw ) gas density, pellet density, bulk density, and bed wall density, respectively (g/cm3) τ ) uHt/L ω ) LDF coefficient (s-1) µ ) viscosity (N‚s/m2) Subscripts B ) bed i ) component i p ) pellet g ) gas phase H ) high-pressure feed step s ) solid phase w ) wall
Literature Cited (1) Zayaraman, A.; Yang, R. T.; Cho, S.-H.; Bhat, S. G.; Choudary, V. N. Adsorption of Nitrogen, Oxygen and Argon on Na-CeX Zeolites. Adsorption 2002, 8, 271. (2) Hayashi, S.; Tsuchiya, H.; Haruna, K. Process for Obtaining High Concentration Argon by Pressure Swing Adsorption. U.S. Patent, 4,529,412, 1985. (3) Miller, G. W.; Theis, C. F. Secondary Oxygen Purifier for Molecular Sieve Oxygen concentrator. U.S. Patent, 4,813,979, 1989. (4) Miller, G. W.; Theis, C. F. Molecular Sieve Oxygen Concentrator with Secondary Oxygen Purifier. U.S. Patent, 4,880,443, 1989. (5) Knaebel, K. S.; Kandynin, A. Pressure Swing Adsorption System to Purify Oxygen. U.S. Patent, 5,226,933, 1993. (6) Lee, C. H. Method for Producing Oxygen and Apparatus Thereof. Korea Patent, 0067797, pending, 2003.
(7) Dee, D. P.; Chiang, R. L.; Miller, E. J.; Whitley, R. D. High Purity Oxygen Production by Pressure Swing Adsorption. U.S. Patent, 6,544,318, 2003. (8) Wilson, S. J.; Webley, P. A. Cyclic Steady-State Axial Temperature Profiles in Multilayer, Bulk Gas PSA-The Case of Oxygen VSA. Ind. Eng. Chem. Res. 2002, 41, 2753. (9) Zhou, L.; Yu, M.; Zhong, L.; Zhou, Y. Feasibility Study on Pressure Swing Sorption for Removing H2S from Natural Gas. Chem. Eng. Sci. 2004, 59, 2401. (10) Ahn, H.; Lee, C.-H.; Seo, B.; Yang, J.; Baek, K. Backfill Cycle of a Layered Bed H2 PSA Process. Adsorption 1999, 5, 421. (11) Jee, J.-G.; Kim, M.-B.; Lee, C.-H. Pressure Swing Adsorption Processes to Purify Oxygen Using a Carbon Molecular Sieve. Chem. Eng. Sci. 2005, 60, 869. (12) Kim, M.-B.; Moon, J.-H.; Lee, C.-H.; Oh, M.; Cho, W. Effect of Heat Transfer on the Transient Dynamics of TSA Process. Korean J. Chem. Eng. 2004, 21, 703. (13) Jee, J.-G.; Park, H.; Haam, S.; Lee, C.-H. Effects of Nonisobaric and Isobaric Steps on O2 Pressure Swing Adsorption for an Aerator. Ind. Eng. Chem. Res. 2002, 41, 4383. (14) Rota, R.; Wanket, P. C. Intensification of Pressure Swing Adsorption Processes. AIChE J. 1990, 36, 1299. (15) Kikkinides, E. S.; Yang, R. T. Effects of Bed Pressure Drop on Isothermal and Adiabatic Adsorber Dynamics. Chem. Eng. Sci. 1993, 48, 1545. (16) Chou, C.; Huang, W.-C. Simulation of a Four-Bed Pressure Swing Adsorption Process for Oxygen Enrichment. Ind. Eng. Chem. Res. 1994, 33, 1250. (17) Mendes, A. M. M.; Costa, C. A. V.; Rodrigues, A. E. Oxygen Separation from Air by PSA: Modelling and Experimental Results Part I: Isothermal Operation. Sep. Purif. Technol. 2001, 24, 173. (18) Bae, Y.; Lee, C.-H. Sorption Kinetics of Eight Gases on a Carbon Molecular Sieve at Elevated Pressure. Carbon 2005, 43, 95. (19) Bae, Y.-S.; Moon, J.-H.; Ahn, H.; Lee, C.-H. Effects of Adsorbate Properties on Adsorption Mechanism in a Carbon Molecular Sieve. Korean J. Chem. Eng. 2004, 21, 712. (20) Krishna, R.; Wesselingh, J. A. The Maxwell-Stefan Approach to Mass Transfer. Chem. Eng. Sci. 1997, 52, 861. (21) Ruthven, D. M.; Raghavan, N. S.; Hassan, M. M. Adsorption and Diffusion of Nitrogen and Oxygen in a Carbon Molecular Sieve. Chem. Eng. Sci. 1986, 41, 1325. (22) Ruthven, D. M. Diffusion of Oxygen and Nitrogen in Carbon Molecular Sieve. Chem. Eng. Sci. 1992, 47, 4305. (23) Chen, Y. D.; Yang, R. T.; Uawithya, P. Diffusion of Oxygen, Nitrogen and Their Mixtures in Carbon Molecular Sieve. AIChE J. 1994, 40, 577. (24) Yang, J.; Lee, C.-H. Adsorption Dynamics of a Layered Bed PSA for H2 Recovery from Coke Oven Gas. AIChE J. 1998, 44, 1325. (25) PSE Ltd. gPROMS Introductory User Guide; Bridge Studios: London, Release 2.1.1, 2002.
Received for review October 6, 2004 Revised manuscript received February 7, 2005 Accepted February 25, 2005 IE049032B
