481

A publication of

CHEMICAL ENGINEERING TRANSACTIONS

VOL. 94, 2022 The Italian Association
of Chemical Engineering
Online at www.cetjournal.it

Guest Editors: Petar S. Varbanov, Yee Van Fan, Jiří J. Klemeš, Sandro Nižetić
Copyright © 2022, AIDIC Servizi S.r.l.
DOI: 10.3303/CET2294080
ISBN 978-88-95608-93-8; ISSN 2283-9216

Experimental Study of Oxygen Separation in Oxygen-

Pressure Swing Adsorption Unit
Radek Šulc*, Miroslav Kos
Czech Technical University in Prague, Faculty of Mechanical Engineering, Department of Process Engineering,
Technicka 4, 160 00, Prague 6, Czech Republic
[email protected]

The pressure swing adsorption (PSA) units are widely used as oxygen sources where oxygen is produced in a
gaseous form. The start-up time of minutes is an undeniable advantage of PSA technology compared to
cryogenic air separation having start-up time taking hours or days. The purity of oxygen produced by PSA using
nitrogen selective zeolites (type A and type X zeolites) is limited to 95 % oxygen. The pilot-plant adsorption unit
utilizes two-bed pressure swing adsorption technology. The nominal capacity of this unit is 1.4 kg h-1 of gaseous
oxygen with a purity of 95 % oxygen. The paper deals to analyze process characteristics of oxygen separation
from the air in the pilot-plant adsorption unit for the adsorption pressure of 5.5 bar and the defined adsorption
cycle. The effect of the number of cycles needed to obtain relevant results was also investigated.

1. Introduction
Skarstrom (1960) designed the first PSA technology for oxygen enrichment using nitrogen-selective zeolite 5A.
The adsorbent was regenerated by blowdown and low-pressure purge using a part of the oxygen that was
produced. Later, Guerin and Domine (1964) modified the Skarstrom cycle by regenerating the adsorbent under
a vacuum. Adsorption at elevated pressure and desorption under vacuum are the principles of the hybrid
pressure-vacuum PSA system (PSA/VPSA system). A comprehensive overview and insight into the science
and technology of pressure swing adsorption are presented, e.g., by Ruthven et al. (1994).
The purity of oxygen produced by PSA using nitrogen selective zeolites (type A and type X zeolites) is limited
to 95 % oxygen. The rest of the product stream is mainly argon. Argon-oxygen selective silver modified X-
zeolites or carbon molecular sieves are used as adsorbents (Santos et al., 2007) to obtain higher oxygen purity,
e.g. for medical purposes or welding. Šulc and Ditl (2021) analysed the potential of consumed energy-saving
by integrating PSA technology as an oxygen source for the oxy-fuel combustion unit to reach an ecologically
more sustainable process.
There are three basic process characteristics of PSA technology: i) product purity, i) adsorbent productivity, and
iii) recovery. Recovery is strongly affected by the purge flow. For a given type of PSA equipment, the optimum
purge flow depends on the purity of the product and the adsorption pressure. Some manufacturers of oxygen
PSA generators use the term ‘air factor’. The air factor is the air ratio defined as the ratio of the air flowrate at
the PSA inlet to the gaseous oxygen stream that is produced.
The paper deals to analyze process characteristics of oxygen separation from the air in the pilot-plant adsorption
unit for the adsorption pressure of 5.5 bar and defined adsorption cycle. The effect of the number of cycles
needed to obtain relevant results was also investigated.

2. Methods
2.1 Experimental set-up
The pilot plant adsorption unit O2 (manufactured by OXYWISE Ltd., Slovakia) was used to carry out all the
experiments. The adsorption unit utilizes two-bed pressure swing adsorption technology. The nominal capacity
of this unit is 1.4 kg h-1 of gaseous oxygen with a purity of 95 % oxygen. The PSA units require an inlet air quality
of class 1.4.1 (solid particles, air humidity, oil). It is necessary to treat compressed air to achieve entering air

Paper Received: 15 May 2022; Revised: 30 May 2022; Accepted: 05 June 2022
Please cite this article as: Šulc R., Kos M., 2022, Experimental Study of Oxygen Separation in Oxygen- Pressure Swing Adsorption Unit,
Chemical Engineering Transactions, 94, 481-486 DOI:10.3303/CET2294080
482

quality. A refrigeration air drying produces an air of the required pressure dew point of +3 °C. The adsorption
unit is plotted in Figure 1 and the unit is shown in Figure 2.

Figure 1: The technological schema of the pilot-plant adsorption unit

Figure 2: The pilot-plant adsorption unit: 1) compressor, 2) air dryer, 3) air receiver, 4) PSA columns, 5) GOX
receiver, 6) exhaust pipe, 7) control unit, 8) mass flowmeter of inlet air, 9) mass flowmeter of product gas

The experiments were carried out for porous zeolite molecular sieve UOP MOLSIV™ PSAO2 XP (UOP LLC,
Honeywell) with enhanced N2 /O2 selectivity. Adsorbent particles are homogeneous spheres of 2 mm diameter
(8 ×12 mesh) with a bulk density of 657 kg/m3. The adsorbent properties are presented in Table 1. The mass
gas flow rate, the volumetric concentrations of oxygen and CO2, and the dew point were measured in the air
and product gas streams. The standard cycle consists of these sequential steps: 1) equalization, 2)
pressurization, 3) production, 4) equalization, and 5) purging. At present, the industrially used cycles combine
pressurization and production steps, as demonstrated in Figure 3. The following time intervals were used for
experiment: i) 48 s for the pressurization, ii) 19 s for the equalization, and iii) 24 s for the production step.
 483

Table 1: Properties of UOP MOLSIV™ PSAO2 XP

Particle BET surface Outer Micropore Pore Micropore Particle Bulk Particle
diameter area surface area area volume volume porosity density density
[m] [m2 g-1] [m2 g-1] [m2 g-1] [m3 g-1] [m3 g-1] [-] [kg m-3 ] [kg m-3 ]

0.002 523 40.2 247 3.47 × 10- 4 2.47 × 10- 4 0.38 657 1,095

Figure 3: The scheme of adsorption cycle used for columns C1 and C2: E-equalization, P-pressurization, Pr-
production, Pu-purging

2.2 Data analysis

The mean mass flowrate per cycle was calculated for each cycle: i) by the trapezoidal integration of
instantaneous mass flowrate shown by the flowmeter used, and ii) using internal integration built in the
flowmeter. The mean values of the oxygen content and the dew point in the air entering the PSA unit and the
product gas were calculated by averaging instantaneous values recorded by sensors. The mean oxygen content
in the exhaust stream per cycle was calculated for each cycle by the trapezoidal integration of instantaneous
oxygen content measured by the oxygen sensor.

3. Results and Discussion

The analysis was executed for 10 cycles. The time courses of instantaneous values of the mass flow and the
dew point are presented for the inlet air and the product gas in Figures 4 and 5. The time course of the oxygen
content in the exhaust gas is shown in Figure 6.

Figure 4: The inlet air: a) mass flowrate (upper), b) dew point (bottom)
484

Figure 5: The product gas: a) mass flowrate (upper), b) dew point (bottom)

Figure 6: The exhaust gas: oxygen content

Figure 7: The adsorption cycle: a) column pressure (upper), b) column temperature (bottom)

The time courses of column pressure and the temperature at the top and bottom of the columns are presented
for both PSA columns in Figure 7. The oxygen content, the dew point and flowrate evaluated for each cycle are
presented in Table 2. Using averaged data for 10 cycles, the mass balance was calculated assuming no nitrogen
 485

in the product gas and argon content of 0.94 mol % in inlet air. Water and CO2 were not taken into account. The
recovery and air factor were calculated. The calculated data are summarized in Table 3. The mass flowrates
determined by the external trapezoidal integration are 1.3 % and 2.6 % lower compared to the internal integration
for inlet air and product gas. This leads to a lower recovery value of 1.4 % and a higher value of the air factor
by 1.4 % compared to the results based on internal integration. Therefore, data based on the internal integration
function built in the flowmeter is recommended to be used.

Table 2: Stream composition and flowrate – evaluated data

Cycle No. Inlet air Product gas Exhaust
gas
mass flowrate oxygen dew mass flowrate oxygen dew oxygen
(kg/h) content point (kg/h) content point content
integration (% vol.) (°C) integration (% vol.) (°C) (% vol.)
-external -internal -external -internal
1 18.085 17.477 21.10 -1.37 1.236 1.240 94.40 -78.42 16.58
2 17.611 17.188 21.07 -1.30 1.202 1.240 94.41 -78.49 17.45
3 16.453 16.973 21.09 -1.20 1.198 1.241 94.41 -78.60 16.02
4 16.762 17.159 21.11 -1.11 1.198 1.209 94.43 -78.74 17.60
5 16.716 17.052 21.08 -1.02 1.198 1.243 94.44 -78.83 16.19
6 16.889 16.843 21.09 -0.93 1.199 1.246 94.44 -78.94 17.65
7 16.916 17.228 21.09 -0.89 1.197 1.238 94.44 -79.08 16.34
8 17.118 16.902 21.06 -0.87 1.198 1.235 94.46 -79.38 18.12
9 15.803 16.833 21.09 -0.87 1.198 1.241 94.47 -79.37 16.49
10 16.220 17.082 21.11 -0.85 1.199 1.210 94.47 -79.23 18.23
Average 16.857 17.074 21.090 -1.04 1.202 1.234 94.44 -78.93 17.07
bcalc -0.041 -0.011 -0.01 -0.004
t 3.4 2.8 3.3 0.7
Hypothesis*1 No No No Yes
Note: *1 Hypothesis is acceptable (Yes)/not acceptable (No). The critical t-distribution t(8,0.05) = 2.306.

Table 3: Mass balance

Cycle No. Inlet air Product gas Exhaust gas
External Internal External Internal External Internal
integration integration integration integration integration integration
Composition – oxygen 21.09 21.09 94.44 94.44 16.10 16.03
( mol %) – nitrogen 77.97 77.97 0 *2 0 *2 83.27 83.35
– argon 0.94 *1 0.94 *1 5.56 5.56 0.63 0.62
Mass flowrate (kg h-1) 16.857 17.074 1.202 1.234 15.655 15.839
Volume flowrate (Nm3 h-1) 13.044 13.211 0.831 0.853 12.213 12.358
Oxygen content – experiment 17.07 17.07
(mass %) – calculated 17.93 17.85
– relative error (%) 5.06 4.61
Recovery (-) 0.285 0.289
Air factor (Nm3 Nm-3) 15.70 15.49
*1 *2
Note: estimated according to dry air. no nitrogen assumed.

The effect of the number of cycles on the mass flowrate of inlet air and product gas was tested by hypothesis
testing (Bowerman and O´Connell, 1997) for each flowrate separately. In this case, the independence of flowrate
on the number of cycles was tested as the hypothesis, i.e. m = B.(Nocycles)0 = const., that is pred = 0. The critical
value of the t-distribution for eight degrees of freedom and significance level  = 0.05 is 2.306. The results of
the hypothesis test are presented in Table 2. For illustration, the values of the calculated t value and the
evaluated power-law exponent calc are presented here also. The independency of the mass flowrate on the
number of cycles was found only for the product gas flowrate obtained by the internal integration function.
Therefore, the effect of the number of cycles needed to obtain relevant results was investigated. Using the data
obtained for each cycle, the cycle-averaged mean value of the flowrate was calculated for the averaging interval
of various numbers of cycles. The number of cycles of the averaging interval that is necessary to stabilize the
flowrate line (i.e. to reach the stable flowrate value) is the minimum number of cycles needed to obtain relevant
486

results. Using this approach, the minimum number of cycles was found at the time when the volatility of the
cycle dependence of cycle-averaged flowrate is less than the defined percentage of stable value. The maximum
measured error and repeatability of the air mass flow meter are 0.35 % and  0.25 % of reading. Therefore, the
volatility range of  0.5 % was applied. The course of the mean value of flowrate on the number of cycles used
to calculate the averaging is presented in Figure 8 including the volatility limit for air and internal integration.
Using this approach and the applied volatility limit, a minimum of six cycles is needed to obtain relevant results.

Figure 8: The minimum number of cycles – flowrate of inlet air by internal integration function

4. Conclusions
The process characteristics of oxygen separation from the air in the pilot-plant adsorption unit utilizing two-bed
pressure swing adsorption technology were investigated for the adsorption pressure of 5.5 bar and the
industrially used adsorption cycle, which combines pressurization and production steps. The experiments were
carried out for porous zeolite molecular sieve UOP MOLSIV™ PSAO2 XP (UOP LLC, Honeywell) with enhanced
N2 /O2 selectivity. Data based on the internal integration function built in the flowmeter are recommended to be
used for data treatment. Using averaged data for 10 cycles, the recovery of 0.289 and the air factor of 15.49
were evaluated for given conditions. Furthermore, the effect of the number of cycles needed to obtain relevant
results was investigated. Using the proposed approach, the number of six cycles was found to be a minimum
needed to obtain relevant results for the applied volatility limit of 0.5 % for the flowrate obtained by the internal
integration function.

Acknowledgements
This work was supported by Ministry of Education, Youth and Sports of the Czech Republic under OP RDE
grant number CZ.02.1.01/0.0/0.0/16_019/0000753 "Research centre for low-carbon energy technologies". The
authors thanks Karel Soukup's Lab (ICPF, Czech Academy of Science) for the adsorbent sample analysis.

References
Bowerman B.L., O´Connell R.T., 1997, Applied statistics: improving business processes, Richard D. Irwin, USA.
Guerin de Montgareuil, P. Domine, D., 1964. Process for separating a binary gaseous mixture by adsorption.
US patent 3155468.
Ruthven D.M., Farooq S., Knaebel K.S., 1994, Pressure Swing Adsorption, Wiley, New York, USA.
Santos, J.C., Cruz, P., Regala, T., Magalhães, F.D., Mendes, A., 2007. High-purity oxygen production by
pressure swing adsorption. Ind. Eng. Chem. Res. 46, 591-599.
Skarstrom C. W., 1960, Method and apparatus for fractionating gas mixtures by adsorption, US patent 2944627.
Šulc R., Ditl P., 2021, The Potential of Energy Savings in Oxygen Production by Pressure Swing Adsorption,
Chemical Engineering Transactions, 86, 313-318.