7th International Chemical Engineering Congress & Exhibition Kish, Iran, 21-24 November, 2011

Numerical Simulation of a Pressure Swing Adsorption for Air Separation

Masoud Mofarahi, Ehsan Javadi Shokroo
Corresponding Authors Address: Chemical Engineering Department, Faculty of Engineering, Persian Gulf University, Iran-Bushehr Corresponding Author Email:
Mofarahi@pgu.ac.ir

Abstract
A two-bed four step pressure swing adsorption (PSA) using zeolite 5A adsorbent for oxygen separation from air studied by dynamic mathematical simulation. The mathematical model contains partial differential equations corresponding to the bulk gas phase mass, energy and momentum balances. The effects of operational variables such as purge to feed ratio, high operating to low operating pressure and feed flow rate on oxygen purity and recovery were investigated. The results show that in a constant feed flow rate, increasing purge to feed ratio can lead to a reduction the oxygen recovery, but instead will cause to increase the oxygen purity. In the same conditions, increase the feed flow rate will result in a reduction the oxygen purity while the oxygen recovery increases. Results of simulation indicated a very good agreement with some current literature experimental work. Keywords: Pressure swing adsorption; Simulation; Oxygen production; Mathematical modeling; zeolite 5A

Introduction Air separation process can be done based on the cryogenic system and non-cryogenic. Recent systems (non-cryogenic) based on cyclic batch adsorption processes, and membrane technologies. Other systems in this field, such as separation systems based on chemical processes are available but compared to these processes are not very important. More information is available elsewhere [1]. Cyclic batch adsorption processes differ from each other, mainly in the methods by which the adsorbent is regenerated during the desorption cycle. Adsorbent regeneration can be done by increasing the temperature (in the thermal swing system), pressure drop (in the pressure swing and vacuum swing system), purging with a non-adsorbing inert gas (in the purge gas stripping system with constant T and P.) and or using a stream containing a competitively adsorbed species. Choice of regeneration methods for each system depends on

Numerical Simulation of a Pressure Swing Adsorption

economic factors and also desired technical characteristics. The pressure swing system (PSA) is well suited to rapid cycling, and this has the advantage of minimizing the absorbent inventory and therefore, the capital costs of the system [2]. PSA process is a wide operating unit to separation and purification of gases, which is acts based on the capability of solids adsorption and selective separation of gases. The most important operational parameter in this system is pressure, and most industrial units operate at\or vicinity of surrounding temperature. In the recent years, the use of this method was followed by researchers as a more important issue in the air separation, because in generally the PSA process is more economical than the other separation processes. Sometimes, in order to justify economically, this process can be replaced the other separation processes. The PSA process evolution around the worldwide is still continuing and to achieve the best economic conditions each day newer acts are done for this important process. Use of this process to oxygen and nitrogen production from the air took for the first time in 1958 by skarstrom. He provided his recommended PSA cycles to enrich the oxygen and nitrogen in the air under a subject of heatless drier [3]. The main reasons for the success of this technology are many reforms that achieved in this field and that also is the new, designed and configured for cycles and devices [4,5,6,7,8]. In overall, the PSA process performance strongly influenced by design parameters (such as: bed size, adsorbent physical properties, configuration and number of beds) and operational variables (such as: pressurization time, production time, purge time, feed flow rate, purge flow rate, production flow rate, temperature and/or pressure variations). This could be an optimum amount of process variables to achieve maximum possible performance. Therefore, it is important to review the behavior of the PSA operating variables to knowing the optimum operating conditions. Zeolite adsorbent used in this separation process is usually 5A or 13X. In this process also argon with oxygen is removed as the product from the system due to having almost the same adsorption behavior near the oxygen. During the progress made on the PSA process, zeolites studies in order to improve their quality (capacity and selectivity) continually be looking away from the years. Including improvements in this area has been set to reduce the inert inorganic material named. More comprehensive information on zeolite developments can be found in other sources [9,10]. The most important theoretical models to describe the PSA behavior based on equilibrium between gas phase, and adsorbed phase can be founded are cited to Shendalman & Mitchell (1972), Chan et al (1981), Kenney & Florez-Fernandez (1983) and Knaebel & Hill (1985) [11]. In 1989, Farooq, Ruthven & Boniface, as the same models of inter-particle diffusion which were provided to nitrogen separation from air previously, used a diffusion model for a two-bed PSA system with zeolite 5A adsorbent to oxygen production from air [12]. The model also in 2001 by Mendes et al [13] to separate oxygen from air was presented. They were studied a system of two bed system with zeolite 5A, and they compared the results of the basic Skarstrom cycle with cycle that includes the pressure equalization step. They used LDF-DG model to describe the inter-particle mass transfer. The model gave very well results for steady state and unsteady state. In the model presented by Jee et al. [14] also LDF model used to survey the effects of co-pressure and pressure variable steps on the PSA performance. They also considered temperature variation, and they were considered thermal equilibrium between the adsorbent and fluid bulk flow. In this case, the system was used by those also that is a two-bed PSA system with zeolite 5A.

7th International Chemical Engineering Congress & Exhibition Kish, Iran, 21-24 November, 2011

In this work, we were studied the effect of operational variables on the Skarstrom PSA cycle to survey the optimal conditions based on mathematical modeling and simulation under operational conditions that use in this study. Mathematical Modeling To develop a mathematical model for a PSA system the main assumptions that have been applied include: 1. Gas behaves an ideal gas. 2. The flow pattern is described by the axially dispersed plug-flow model. 3. Absorbing properties throughout the tower would remain constant and unchanged. 4. Radial gradient is to be negligible. 5. Equilibrium equations for the components of the air by two-component Langmuir isotherm can be expressed. 6. Mass transfer rate by a linear driving force equation is expressed. 7. Thermal equilibrium between gas phase and solid is assumed. 8. Pressure drop along the bed is calculated by the Erguns equation. And other common assumptions in the simulation of adsorption processes. Overall and component mass balances for the bulk phase in the adsorption bed to form the following equations are written 2 ( ) 1 (1) + + + . =0 2 (2) 2 () 1 2 + + + . =0
=1

When the ideal gas law ( = and = ) is applied to eqs 1 and 2, the component and overall mass balances can be represented as follows: 2 1 (3) + + + + . =0 2 (4) 2 2 1 1 1 2 + + + + 2 + + 1 1 2 +

=1

=0

Another characteristic of adsorption process is temperature variations caused by heat of adsorption and desorption occur. In this system, energy balance for the gas phase and also heat transfer to the bed wall is included.

2 2

+ + 2

+ +

=1

(5)

= 0

To evaluate heat loss through the walls and the accumulation of energy, corresponding to an energy balance has also been used. (6) = 2 2

Numerical Simulation of a Pressure Swing Adsorption

Where (7) = 2 2 The well-known Danckwerts boundary conditions are applied Pressurization and production step (8-1) | =0 = | =0 | =0+ ; | =0 = (8-2) | =0 = | =0 | =0+ ; | =0 = Where yi |z=0 means the feed composition for the component i. Counter current purge step (9-1) | = = | =+ | = ; | =0 =0 (9-2) | = = | =+ | = ; | =0 =0 Where | =+ means a volume-averaged composition of the effluent stream during the adsorption step for the purge step. Counter current blowdown step (10-1) | =0 = | =0 = (10-2) | =0 = | =0 = Boundary conditions for the interstitial velocity Pressurization and counter current blodown step (11-1) = = 0 Pressurization and production step =0 = (11-2) Counter current purge step = = . (11-3) The initial conditions for feed flow , 0 = 0; , 0 = 0; (, 0) = 0 (12) (13)

, 0 = ; (0) =

In this study, the pressure time function is assumed as an exponential function which is adapted to the literature [21]. (14) = . 1 . ( ) + . In the above equation a, b and f(t) parameteres defined regared to duration and initial and final pressures of each step. To consider the pressure drop effect across the bed, Erguns equation was introduced as a momentum balance [15]. (15-1) = + 150 (1 )2 (1 ) = , = 1.75 (15-2) 2 2 2 4

7th International Chemical Engineering Congress & Exhibition Kish, Iran, 21-24 November, 2011

Where u is the interstitial velocity. The multi-component adsorption equilibrium was predicted by the following LDF model. = 1 + =1 Where, = 1 + 2 , = 3 exp 4

(16)

(17)

The sorption rate into an adsorbent pellet is described by the LDF model with a single lumped mass-transfer parameter[17]. 15 (18) = , = 2 Where [18], 15 0.5 (19) = (1 + )2 2 The adsorption isotherm parameters and diffusion rate constant of N2 and O2 over zeolite 5A are also shown in Table 1. Table 2 and Table 3 shows the adsorbent and bed characteristics, respectively [14].
Table 1 Equilibrium\Rate parameters and heat of adsorption of N2 and O2 on zeolite 5A [14] N2 O2 ( ) ( ( . ) ) 6.210 -1.270 1.986 1970 5470 0.0066
-1

7.252 -1.820 54.19 662.6 3160 0.0267

() Heat of adsorption, -Hi (cal/mol) Diffusion rate constant(s ) [22]

Results and Discussion The implicit finite difference scheme was used to solve a mathematical model that considered of coupled partial differential equations. The central first order difference is used to discretize the first order space derivatives and the second order derivatives discretized by using a second order central difference. The forward finite difference is used to time spacial. Solving of algebraic equations were done by MATLAB software. In order to validate the simulation results, first the results of this work were compared with the other experimental data in the literature. In an experimental study, Adelio M. M. Mendes et al [19] were simulated a PSA commercial unit performance to evaluate the effects of some operational variables. They concluded that to affect of pressure rising in the adsorption step, in a constant feed flow rate, increased pressure is leading to decrease both purity and recovery of oxygen. The experimental results by these authors together with the simulation in this work has come in Fig. 1. As obvious in this figure, the simulation and presented the model in this work make predict the results of the other experimentally work with relatively high accuracy.

Numerical Simulation of a Pressure Swing Adsorption

Table 2 Characteristics of adsorbent [14] Zeolite 5A Adsorbent Type Average pellet size, RP (cm) Pellet density, P (g/cm ) Heat capacity, Cps (cal/g.k) Particle porosity, Bed density, B (g/cm3) Table 3 characteristics of adsorption bed Length, L (cm) [in this work] Inside radius, RBi (cm) [in this work] Outside radius, RBo (cm) [in this work] Heat capacity of the column, Cpw (cal/g.K) [14] Density of column, w (g/cm ) [14] Internal heat-transfer coefficient, hi (cal/cm .K.s) [14] External heat-transfer coefficient, ho (cal/ cm .K.s) [14] Axial thermal conductivity, KL (cal/cm.s.K) [14] Axial dispersion coefficient, DL (cm2/s) [14]
2 2 3 3

Sphere 0.157 1.16 0.32 0.65 0.795

76 2.138 2.415 0.12 7.83 . . .

Another work which was done by Adelio M. M. Mendes et al [13] the effects of some other operational variables on the PSA unit performance were studied by experiments and simulations. They concluded that to affect of the purge flow rate on oxygen purity and recovery, increasing the purge flow rate led to decreased recovery but instead the oxygen purity will increase. The experimental results of these authors together with presented results of modeling in this work are shown in Fig. 2. In this consideration also can be seen that the results of simulation indicate a very good agreement with same current literature experimental work.

Figure 1: O2 purity a recovery as a function of production pressure, compare the model prediction in this work and experimental data by Mendes et al [19].

7th International Chemical Engineering Congress & Exhibition Kish, Iran, 21-24 November, 2011

Figure 2: O2 purity and recovery as a function of purge flow rate, compare the model prediction in this work and experimental data by Mendes et al [13].

In a PSA process, the duration of the adsorption step is determined by studying the breakthrough curve. The term breakthrough curve refers to the response of the initially clean bed to an influent with a constant composition. It can be seen by monitoring the concentration of the effluent. Breakthrough occurs when the effluent concentration reaches a specific value. The adsorbate concentration in the flow at any given point in a bed is a function of time, resulting from the movement of concentration front in the bed. The breakthrough curve for a gas containing a single adsorbate can be obtained by the solution of the mass balance equations for both the bed and adsorbent particles, along with the equilibrium isotherm. The duration of the adsorption step is the time period needed for breakthrough to occur. After this time, the product purity will decline, and before this time the full bed capacity will not be employed. Thus the adsorption time should

Figure 3: O2 effluent mole fraction at the end of the bed during AD step

be near the breakthrough time. This time depends upon isotherm, diffusivity and residence time of the feed in the bed [20]. In this work, we have considered AD time based on start time the release of absorbed nitrogen (t=35 seconds). Fig. 3 shows oxygen concentration versus time at the top of the bed during AD step. The temperature profile during the AD step at the cyclic steady state illustrated in Fig. 4. It can be seen that the changes in temperature are too small. In Fig. 5 it also shows the changes in temperature in the cyclic steady state for successive cycles at the top

Numerical Simulation of a Pressure Swing Adsorption

the bed. As it implies, the temperature variation is very low for the conditions studied in this figure. Fig. 6.a represents the effect of ratio on the oxygen purity in the AD step effluent for different ratios. As indicated, at all the ratios, increase the will lead to increasing oxygen purity. Reason of this is due to the increased purge flow rate that ultimately led to a better desorption bed in the low pressure step. Therefore, at higher ratio with a cleaner bed it can be achieved the higher levels of oxygen purity. In this figure also it can be seen that for ratios of = and = , than the other operating pressures ratios, the highest purity achieved for oxygen in the product. In order to better compare the = and = , in Fig. 6.b changes in oxygen purity are given for these two cases separately. Obviously, the system gives the best pure oxygen in = . Fig. 7 shows oxygen recovery profile, for the same ratios of as previous, in terms of ratio. Clearly, in any ratio of , unlike the behavior of purity, increased ratio lead to decrease the levels of oxygen recovery. It is evident the reduction of product flow rate causes by increased purge flow rate will be led to reduce the amount of oxygen recovery. In addition, this figure shows for = . , despite being the least amount of oxygen purity, the oxygen recovery has the highest quantity instead than the other values. The reversal behavior of the purity and recovery is an evident effect and is also seen in the other sources. Fig. 8 indices the variation of oxygen purity versus to the feed flow rate for different values of ratios. As indicated, for all ratios, increased feed flow rate will be led to decrease the oxygen purity. Increasing the feed flow rate led to rises the AD step pressure that eventually causes the co-absorption of nitrogen with the oxygen and finally decreases oxygen purity in the product output. In Fig.9 it is clear that increasing feed flow rate will increase the oxygen recovery in the system. This effect is an evident treat for the system, because in the same conditions increased feed flow causes the enhanced product flow and ultimately goes up the amount of oxygen to be recovered.

Figure 4: Temperature dependency in terms of time and length during AD step at cyclic steady state, ( = . ; = . ; feed flow rate=5lit(STP)/min)

7th International Chemical Engineering Congress & Exhibition Kish, Iran, 21-24 November, 2011

Figure 5: Steady state temperature profile at the top of the column, ( = . ; = . ; feed flow rate=11 lit(STP)/min; cycle time=110 s)

Figure 6: O2 purity as a function of P/F ratio for various ratios of PH/PL, (feed flow rate=5 lit. (STP)/min; cycle time=110s)

Figure 7: O2 recovery as a function of P/F ratio, for various ratios of PH/PL, (feed flow rate=5 lit. (STP)/min; cycle time=110s)

Numerical Simulation of a Pressure Swing Adsorption

Figure 8: O2 purity as a function of feed flow rate, for various ratios of PH/PL, (P/F=0.3; cycle time=110)

Figure 9: O2 recovery as a function of feed flow rate, for various ratios of PH/PL, (P/F=0.3; cycle time=110)

Conclusions In this work, a two-bed four step PSA process under laboratory scale using zeolite 5A adsorbent was studied by mathematical simulation. First, to determine the accuracy of predictions, the results of this work have been compared with the other simulations and other experiments. Simulation results indicated a satisfactory compliance with the current sources. The effect of operational variables such as ratio, ratio and feed flow rate on product purity and recovery in a oxygen production PSA unit was investigated. In a constant feed flow rate, increase ratio causes the reduction oxygen recovery, but instead will cause to increase the oxygen purity. In the same conditions, increase the feed flow rate will result in the reduction oxygen purity, but against the oxygen recovery rises. In this study, values of equal to . , , . . , . and side the effects of ratio and feed flow rate were studied. Based on the area that studied in this work, in terms of oxygen purity, the best ratio of equal to was found. Furthermore, in terms of oxygen purity and recovery, in lower amounts of feed flow rate, the best ratio of equal to was seen.

7th International Chemical Engineering Congress & Exhibition Kish, Iran, 21-24 November, 2011

Nomenclature Aw cross-sectional area of the wall (cm2) AD adsorption step B equilibrium parameter for the Langmuir model (atm-1) BD blowdown step Cpg, Cps, Cpw gas, pellet, and wall heat capacities, respectively (cal/g.K) 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 average heat of adsorption (cal/mol) k parameter for the LDF model KL axial thermal conductivity (cal/cm.s.K) L bed length (cm) P total pressure (atm) PG purge step PR pressurization step P/F ratio of purge flow rate to feed flow rate PH/PL ratio of operating pressures q, q*, q amount adsorbed, equilibrium amount adsorbed, and average amount adsorbed, respectively (mol/g) qm equilibrium parameter for the Langmuir model (mol/g) R gas constant (cal/mol.K) Rp radius of the pellet (cm) RBi, RBo inside and outside radii of the bed, respectively (cm) T time (s) Tatm temperature of the atmosphere (K) T, Tw pellet or bed temperature and wall temperature, respectively (K) u interstitial velocity (cm/s) yi mole fraction of species i in the gas phase z axial distance in the bed from the inlet (cm) Greek Letters particle porosity , t voidage of the adsorbent bed and total void fraction, respectively g, p, B, w gas density, pellet density, bulk density, and bed wall density, respectively (g/cm3) Subscripts B H i L p g s w bed higher operating pressure component i lower operating pressure pellet gas phase solid wall

Numerical Simulation of a Pressure Swing Adsorption

References [1] W.F.Castle, air separation and liquefaction: recent developments and prospects for the beginning of the new millennium, international journal of refrigeration, 25 (2002) 158-172. [2] Douglas M.Ruthven, Principles of adsorption and adsorption process, John Wiley & Sons, 1984. [3] C. W. Skarstrom, U.S. Patent No. 2,944,627 (1960), to Exxon research and engineering. [4] C. W. Skarstrom, U.S. Patent No. 3,237,377 (1963), to Exxon research and engineering. [5] Fuderer A, Rudelstorfer E., Selective adsorption process, U.S. Patent No. 3,846,849 (1976). [6] Malek A, Farooq S., Study of a six-bed pressure swing adsorption process, AIChE J. 43 (1997) 2509-2523. [7] Malek A, Farooq S., Hydrogen purification from refinery fuel gas by pressure swing adsorption, AIChE J. 44 (1998) 1985-1992. [8] Sircar S, Golden TC., Purification of hydrogen by pressure swing adsorption, Sep Sci Technol. 35 (2000) 667-687. [9] Ralph, T. Yang., Air separation by pressure swing adsorption using superior adsorbents, Final Technical Report, Department of Chemical Engineering University of Michigan, 2001. [10] Santos, J. C., Magalhaes, F.D., Mendens, A., Contamination of zeolites used in oxygen production by PSA: Effects of water and carbon dioxide, Ind. Eng. Chem. Res. 47 (2008) 6197 6203. [11] Flores-Fernandez G., Kenney C.N., Modelling of the pressure swing air separation process, Chem. Eng. Sci. 38 (1983) No.6, 827-834. [12] Farooq S., Ruthven D.M., Boniface H.A., Numerical simulation of a pressure swing adsorption oxygen unit, Chem. Eng. Sci. 44 (1989) No.12, 2809-2816. [13] 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. and Purif. Technol. 24 (2001) 173-188. [14] Jee J.G, Park H.J, Haam S.J., Lee Ch.H, Effects of Nonisobaric and Isobaric Steps on O2 Presure Swing Adsorption for an Aerator, Ind. Eng. Chem. Res., 41 (2002) 4383-4392. [15] Yang, J., Lee, C.-H., Adsorption dynamics of a layered bed PSA for H2 recovery from coke oven gas, AIChEJ. 44 (1998) 1325. [16] Ruthven, D. M., Farooq, S., Knaebel, K. S., Pressure swing adsorption, VCH Publishers, New York, 1994 [17] Wever, T. W., Chakravort, R. K., Pore and Solid diffusion Models for Fixed-bed Adsorbers, AIChE J. 20 (1974) 228. [18] Jeong-Geun Jee, Min-Bae Kim, Chang-Ha Lee, Pressure swing adsorption processes to purify oxygen using a carbon molecular sieve, Chem. Eng. Sci. 60 (2005) 869 882 [19] Ade lio M. M. Mendes, Carlos A. V. Costa, Alrio E. Rodrigues, Analysis of Nonisobaric Steps in Nonlinear Bicomponent Pressure Swing Adsorption Systems. Application to Air Separation, Ind. Eng. Chem. Res. 39 (2000) 138-145. [20] S. Jain, A.S. Moharir, P. Li, G. Wozny, Heuristic design of pressure swing adsorption: a preliminary study, Sep. and Purif. Technol. 33 (2003) 25-43 [21] Salil U. Rege, Ralph T. Yang, Kangyi Qian, Mark A. Buzanowski, Air-prepurification by pressure swing adsorption using single/layered beds,Chem. Eng. Sci. 56 (2001) 2745-2759 [22] Jeong-Geun Jee, Sang-Jin Lee, Chang-Ha Lee, Comparison of the Adsorption Dynamics of Air on Zeolite 5A and Carbon Molecular Sieve Beds, Korean J. Chem. Eng. 21(2004)1183-1192