ARTICLE pubs.acs.org/IECR
Rapid Pressure Swing Adsorption for Reduction of Bed Size Factor of a Medical Oxygen Concentrator Siew Wah Chai, Mayuresh V. Kothare, and Shivaji Sircar* Department of Chemical Engineering, Lehigh University, Bethlehem, Pennsylvania 18015, United States ABSTRACT: Reduction of the adsorber size of a medical oxygen concentrator (MOC) employing a generic pressure swing adsorption (PSA) technology is an ongoing research and development activity. The MOC typically produces a 9093% O2-enriched product gas from ambient air at a rate of e10 L/minute (LPM) for individual use. A common practice is to reduce the total cycle time (tc, seconds) of the PSA process in order to decrease the bed size factor [BSF, pounds of adsorbent in the PSA unit per ton of contained O2 per day production rate (TPDc)]. Adsorbent columns packed with very small adsorbent particles are used to enhance the adsorption kinetics for rapid PSA cycle operation. An experimental mini-PSA set up was used to measure the performance of a simulated Skarstrom-like PSA cycle using a commercial sample of LiX zeolite as the air separation sorbent. Different adsorbent particle sizes, adsorption pressures, and cycle times were tested. It was experimentally demonstrated that BSF cannot be indefinitely reduced by lowering tc because of finite adsorbate mass transfer and gassolid heat transfer resistances, as well as column pressure drop during the desorption step. A BSF of ∼2550 lbs/TPDc with an O2 recovery of ∼2535% for production of ∼90% O2 could be achieved by the PSA process using a dry, CO2-free air feed at a pressure of 34 atm, an adsorbent particle size of ∼0.35 mm, and a total cycle time of 35 s. A novel ‘snap on’ concept of a truly compact and portable MOC was proposed.
’ INTRODUCTION Many pressure swing adsorption (PSA) processes have been patented and commercialized for production of ∼9095% O2enriched gas from ambient air by employing a zeolitic adsorbent.13 Figure 1 shows the result of a patent search under the topic “air separation by adsorption” which plots a year by year tally of the number of U.S. Patents on the subject during the period of 1980 to 2005. An overwhelming number of 452 patents were issued. The field is very crowded but still active (∼7 patents issued per year). In principle, a generic PSA process for air separation cyclically carries out an adsorption step at a relatively higher pressure, where N2 is selectively (thermodynamic) adsorbed from a dry and CO2-free air stream on the zeolite, thereby producing an O2-enriched product gas, followed by N2 desorption at relatively lower pressures where the zeolite is regenerated for reuse. Various complementary steps are often included in the process cycle to improve the product specifications and the process performances. Three subclasses of the generic PSA technology based on the pressure levels of the adsorption step (PA) and the ultimate desorption step (PD) are common in the literature. They are pressure swing adsorption (PSA) where PA is superatmospheric and PD is ambient, vacuum swing adsorption (VSA) where PA is near ambient and PD is subatmospheric, and pressurevacuum swing adsorption (PVSA), which is a combination of the two. A VSA process employing a zeolite has also been commercialized where a N2-rich byproduct gas (99þ % N2) is simultaneously produced.4 Many zeolites including type A (CaNa exchanged 5 A, high Ca exchanged A), type X (Na, Li, Ca, Sr, Ba exchanged), low silica X (LSX) (Li, LiZn exchanged), and mordenite (Na, Ca exchanged) have been commercially used for air separation.13 They are all highly polar materials, and H2O and CO2 are generally removed from the feed air prior to air separation by using a layer of a desiccant (NaX zeolite or alumina) at the feed gas end of the adsorber. The r 2011 American Chemical Society
selectivities of adsorption exhibited by these zeolites for the components of air are in the order N2 > O2 g Ar, but they exhibit significantly different thermodynamic sorptive properties (equilibrium isotherms and heats) for sorption of these gases.13 The currently favored zeolites for air separation are LiX, LiLSX, and LiZnX because they provide higher selectivity of adsorption of N2 over O2 and Ar, higher PSA cyclic working capacity for N2, and moderate heats of adsorption for N2, which translate into superior process performance for air separation.3,5 Figure 2 is a block diagram depicting the overall performance of a generic PSA air separation system producing 92% O2-enriched product gas. The total amount of zeolite adsorbent in the system is w (lbs), and the total cycle time of the PSA process is tc (seconds). The variables F, P, and W are, respectively, the amounts (lb moles) of air feed, O2-enriched product gas, and N2-rich waste gas from the system per cycle of operation. According to Figure 2, the net tonnage rate of production of contained O2 per day (TPDc) is equal to 1.272 103 P/tc, and bed size factor ðBSF; lbs=TPDc Þ ¼ ð7:862 104 Þwtc =P ð1Þ O2 recovery from air feed ðRÞ ¼ 0:92P=0:2096F
ð2Þ
The above-described unit of BSF is an industry standard. For a given PSA process cycle (process steps and sequence, operating conditions, vessel design, system void, individual and total cycle times) and a given sorbent (sorptive and physical properties), one can define a specific O2 productivity (N, lb mol/lb of Received: March 14, 2011 Accepted: May 26, 2011 Revised: May 12, 2011 Published: May 26, 2011 8703
dx.doi.org/10.1021/ie2005093 | Ind. Eng. Chem. Res. 2011, 50, 8703–8710
Industrial & Engineering Chemistry Research
ARTICLE
zeolite/cycle = P/w) by the process. Consequently, BSF ¼ ð7:862 104 Þt c =N
ð3Þ
It can also be shown by rearranging eqs 1 3 that, total amount of adsorbent in PSA system; w ¼ BSF TPDc air feed rate to PSA system ðlb mol=sÞ ¼ F=tc ¼ ð3:451 103 ÞTPDc =R
ð4Þ ð5Þ
Equations 4 and 5 show that BSF and R are two important performance variables for a generic PSA process for air separation because, for a given plant size (TPDc), the total amount of adsorbent inventory in the system is directly proportional to BSF, while the feed air flow rate for the process is inversely proportional to R. Obviously lower BSF and higher R are preferred characteristics of a generic PSA O2 concentrator. Equation 3 shows that BSF can be decreased by lowering tc provided that N is not a function (or a weak function) of tc. In other words, rapid cycles permit the adsorbent to be used more frequently and, therefore, increase the O2 productivity rate provided that the performance is not affected by fast cycling of the process. This behavior is expected to be valid at relatively larger cycle times where the time scales of local adsorption kinetics and kinetics of heat transfer between the adsorbent and the gas phase, which are normally fast, are much smaller than the individual step times and the overall process cycle time. On the other hand, these kinetic impediments can be detrimental to the process performance if the characteristic time constants of these processes are
Figure 1. Result of patent search for 19802005.
comparable to the PSA process cycle time as can be the case for a very rapid cycle. A simplified isothermal model of an idealized differential PSA concept using a single pellet of an adsorbent for removal of a single adsorbate from an inert gas, where the gas-phase partial pressure of the adsorbate was periodically changed differentially, showed that the net rate of the adsorbate removal (mol/g of adsorbent/time) by the adsorbent increased as the cycle time was decreased.6 This was equivalent to decreasing BSF with decreasing cycle time. In the absence of an adsorbate mass transfer resistance, the rate approached infinity when the cycle time approached zero (BSF approached zero). However, the rate of adsorbate removal approached a finite value at the limit of zero cycle time when the adsorbate had a finite mass transfer coefficient (k) for sorption. The limiting rate was a strong function of k.6 A nonisothermal extension of the model, where the effect of a finite gassolid heat transfer coefficient (h) on the performance of a rapid cycle was investigated, showed that the relative efficiency of adsorbate removal by the nonisothermal process relative to the isothermal performance decreased substantially at smaller cycle times.7 Adsorbent particles of small diameter (dp) are deliberately used in a rapid PSA cycle in order to decrease the abovementioned kinetic resistances. The transfer coefficients k and h are typically proportional to dp2 and dp1, respectively, in the absence of axial dispersion. However, the column pressure drop may be significant for the steps of a rapid PSA cycle using very small dp, which can lower the overall separation performance of the process due to increased axial dispersion particularly in a shallow adsorber as discussed by Zhong, Rankin, and Ackley.28 The BSF can also increase and R can decrease due to increase in the quantity of product O2 back purge needed for desorption of N2 from a zeolite column with a large pressure drop.8 The column pressure drop in a RPSA adsorber can be minimized by employing a shallow adsorber having a relatively low length (L) to diameter (D) ratio, but L/D < 0.51.0 can cause poor gas distribution and channeling, and inadequate contact between gas and solid particles.9,10 The problem may be alleviated by adding elaborate gas distributors at two ends of the adsorber, but that option may not be appropriate for small scale RPSA systems because it will increase the specific void gas volume of the system. A very small column length can also lower the separation efficiency by creating a relatively large ratio of adsorbate mass transfer zone length to column length.28 Use of very small adsorbent particles (dp < 500 μm) can also cause agglomeration in packed beds which can impede gassolid mass and heat transfer rates.11
Figure 2. Block diagram of a generic PSA air separation system. 8704
dx.doi.org/10.1021/ie2005093 |Ind. Eng. Chem. Res. 2011, 50, 8703–8710
Industrial & Engineering Chemistry Research
ARTICLE
Figure 3, therefore, realistically depicts the BSF vs tc profile for a PSA O2 concentrator. For a given particle size, the BSF decreases approximately linearly with decreasing cycle time, according to eq 3, when the process cycle time is large. However, this pattern is not continued indefinitely. The BSF levels off or even increases with decreasing tc for very rapid cycles because of impediments introduced by mass and heat transfer resistances and column pressure drop. The progressive increase of BSF at low cycle time is expected because the separation efficiency suffers when the PSA cycle time is faster than the characteristic time constants for the heat and mass transfer process. Reduction of total cycle time to decrease the BSF of a PSA O2 concentrator, however, has been implemented in the design of many tonnage scale (∼1 to 150 TPDc) commercial plants. Table 1 provides several examples of progressive reduction of total cycle time and BSF over the years.1216 The PSA process developed by Batta (Table 1) is one of the earlier patented and commercialized processes for production of ∼90% O2 from air. The PVSA processes described in the second and third columns are more recent developments. The improvement in BSF due to faster cycle times is evident. The evolution of the preferred zeolite for the air separation application is also notable in the table. These processes typically employ adsorbent particle diameters of 12 mm for low pressure drops during the process steps. Tonnage scale production of ∼90% O2 from air by a PVSA process appears to have become the ‘state of the art technology’. Some of the applications of such a process include wastewater treatment, ozone generation, bioreactors, glass furnaces, etc. The
minimum individual step and total cycle times of these PSA processes may be limited by hydrodynamics of large volume gas flow through the packed column adsorbers and gas headers, valve switching time, vacuum pump size, etc.
’ RAPID PSA CYCLE OXYGEN CONCENTRATOR A drastic reduction of total cycle time (tc ∼ 10 s or less) for PSA O2 concentrators has been achieved under the banner of ‘rapid PSA (RPSA) cycle’. Table 2 provides a few examples of patented RPSA processes1722 which have been experimentally demonstrated. These RPSA processes employ very small particles of the adsorbent (dp e 0.5 mm) to enhance gassolid adsorption and heat transfer kinetics. The first entry in Table 2 describes the performance of a RPSA process for production of 2560% O2enriched air from ambient air.6,17 The process uses a novel design of two shallow adsorber beds packed with NaX zeolite, each sequentially undergoing two steps: (a) simultaneous pressurization with compressed air and production of O2-enriched product gas, and (b) simultaneous depressurization and back purging with a part of the O2-enriched gas. The reported data is from a pilot test rig using dry air feed. The process was designed for tonnage scale production of medium purity O2 for enhanced combustion application.23 The other RPSA O2 concentrators described in Table 2 were developed for a special application of this technology, to provide breathing oxygen to patients suffering from chronic obstructive pulmonary disease (COPD) caused by emphysema, chronic bronchitis, pulmonary fibrosis, etc. These medical oxygen concentrators (MOC) are designed for individual use at home or for travel. They produce 8596% O2 at a rate of e10 LPM (∼0.02 TPDc). They are relatively small units, and compactness of design and lightweight are critical requirements for portability. The process steps for these MOC designs are commonly used variations of the steps for the Skarstrom cycle.24 They include pressurization with air and or product gas, pressure equalization between columns, depressurization to ambient or subambient pressure level, and back purge with product gas at ambient or subatmospheric pressure. A large volume of work on model-simulated performance of different RPSA processes using different zeolites for MOC application has also been published in recent years.2529 Table 3 is a list of selected works on this topic. It describes adsorbent particle size, cycle time, and process performance. Different models for adsorption isotherms of N2 and O2 and different models for their sorption kinetics have been used in these simulations. All models include axial dispersion in the gas phase, but some of them ignore column pressure drop and nonisothermal operation of the PSA process.
Figure 3. Schematic drawing of BSF vs cycle time for a RPSA process.
Table 1. Performance of Commercial Tonnage Scale O2 Concentrators U.S. patents 3,636,679 (1972) 3,973,931 (1976) 5,711,787 (1998)
U.S. patent 5,415,683 (1995)
U.S. patent 6,146,447 (2000)
authors
Batta12 Collins13 Neill, Leavitt, Figueiredo14
Leavitt15
company
Union Carbide Corp., Praxair Technology
Praxair Technology
Sircar, Naheiri, Fischer16 Air Products
process mode
PSA
PVSA
PVSA
adsorbent
5A zeolite
LiX zeolite
LiZnX zeolite
PA:PD (atm)
36.3:1
1.43:0.34
∼1.3:0.45
parallel adsorbers
3, 4
2
1
O2 product purity
9093%
90%
90%
O2 recovery total cycle time (seconds)
3346% 240
53% 70
4075% ∼45
BSF (lbs/TPDc O2)
∼56003000
830
777
8705
dx.doi.org/10.1021/ie2005093 |Ind. Eng. Chem. Res. 2011, 50, 8703–8710
Industrial & Engineering Chemistry Research
ARTICLE
Table 2. Experimental Performance of Various RPSA O2 Concentrators product O2 process
cycle mode and time (s)
zeolite dp (mm)
Sircar17 U.S. patent 5,071,449 (1991) Kulish and Swank18 U.S. patent 5,827,358 (1998) Ackley and Zhong19 U.S. patent 6,551,384 (2003)
PSA 812 PSA 6 PSA 9 PVSA 4
NaX 0.5 zeolite (not disclosed) LiX C.D. 0.5
Jagger, Van Brunt, Kivisto, Lonnes20 U.S. patent 7,121,276 (2006) McCombs, Bosinski, Casey, Valvo21 U.S. patent application 0117957 (2006) Chai, Sircar, Kothare22 U.S. patent application 0300285 (2010)
VSA 0.545.6
purity (mole %)
PA:PD (atm)
40 9096 8595
recovery (%)
BSF (lbs/TPDc) 263 ∼110125 90 (PSA), 50 (PVSA) ∼20230
low silica LiX
3:1 3:1 3:1 (PSA) 1.5:0.5 (PVSA) 1:0.3
8595
46 not disclosed 50 (PSA), 60 (PVSA) 60
PSA 11
zeolite (not disclosed)
1.92.7:1
∼90
∼28
∼274
PSA ∼ 35
low silica LiX ∼ 0.35
34:1
∼90
∼2535
∼2550
Table 3. Model Simulation Studies of Various PSA O2 Concentrators performance authors
% O2
rec %
Kopaygorodsky et al., 85 56 200425 process: RPSA; PA = 1.6 atm; PD = 1.0 atm Santos et al., 200426 process: 94.5 22 PSA; PA = 3.0 atm; PD = 1.0 atm Santos et al., 200627 process: 94.5 32 PSA; PA = 3.0 atm; PD = 1.0 atm 90 2550 Zhong et al., 201028 process: RPVSA; PA = 1.5 atm; PD = 0.5 atm ∼90 24 Rama Rao et al., 201029 process: pulsed PSA; PA = 3.5 atm; PD = 1.0 atm
BSF lb/ TPDc
mass transfer kinetics
14.2
no resistance
no
yes
yes
linear
528
LDF model
no
no
yes
Langmuir Oxysiv 5 (NaX zeolite) Freundlich dp= ∼600 μm
18
455
LDF model
no
no
yes
Langmuir Oxysiv 5 (NaX zeolite) Freundlich dp= ∼600 μm
16
