and Dual-Stage High-Purity Oxygen Production ... - ACS Publications

Jun 15, 2018 - ple single- and dual-stage vacuum swing adsorption (VSA) cycle configurations was investigated through process simulation, and optimiza...
0 downloads 0 Views 2MB Size
Article Cite This: Ind. Eng. Chem. Res. 2018, 57, 8997−9008

pubs.acs.org/IECR

Single- and Dual-Stage High-Purity Oxygen Production Using SilverExchanged Titanosilicates (Ag-ETS-10) Sayed Alireza Hosseinzadeh Hejazi,† Libardo Estupiñan Perez, Kasturi Nagesh Pai, Arvind Rajendran,* and Steven M. Kuznicki Department of Chemical and Materials Engineering, University of Alberta, 12th Floor, Donadeo Innovation Centre for Engineering (ICE), 9211−116 Street, Edmonton, Alberta, Canada T6G 1H9

Ind. Eng. Chem. Res. 2018.57:8997-9008. Downloaded from pubs.acs.org by DURHAM UNIV on 08/08/18. For personal use only.

S Supporting Information *

ABSTRACT: High-purity (>99.5%) O2 production from air with silver-exchanged titanosilicates (Ag-ETS-10) using multiple single- and dual-stage vacuum swing adsorption (VSA) cycle configurations was investigated through process simulation, and optimization. Model predictions for the Skarstrom cycle were validated through experiments and an O2 purity of 98.3 ± 0.5% with a corresponding recovery of 10.6% was achieved. Oxygen purity-recovery Pareto fronts for the Skarstrom cycle and 6-step cycle with pressure equalization (PE) and heavy product pressurization (HPP) with dry air as the feed were obtained. The optimization predicted 82.0% O2 recovery with a product purity of 99.5% for a 6-step cycle with PE and HPP and dry air feed stream. The effect of nitrogen content in the feed on the performance indicators was also studied. Operating conditions for various cycle configurations were optimized through nondominated sorting genetic algorithm technique to achieve low total energy consumption (592.4 kWh/tonne O2) and high overall productivity (1.30 tonne O2/m3 Ag-ETS-10 day). The Pareto fronts of the single- and dual-stage configurations were compared against each other in order to choose the best possible design. The results indicated that the single-stage 6-step cycle with PE and HPP presents a better performance compared to the other singleand dual-stage approaches. A simple graphical scheduling study was also conducted in order to calculate the number of columns required for a continuous process using the better performing configurations.

1. INTRODUCTION Adsorptive separation processes are well-suited for small and medium-scale air separation units for both oxygen and nitrogen production.1−4 The main advantages of this technology lies in the fact that it operates at ambient temperatures; can be operated on electrical or battery power; and can be miniaturized. Hence, it is ideal for applications where these gases can be generated at the point-of-requirement. Adsorptive air-separation units find applications in portable oxygen concentrators, on-board nitrogen-generation systems, medical-oxygen plants, etc. Although high-purity nitrogen is produced by kinetic separation on carbon molecular sieves,1 oxygen production is an equilibrium-based separation.5,6 Most adsorbents that are used for oxygen production, e.g., Zeolite 5A, Zeolite 13X, do not show equilibrium selectivity for argon and oxygen. Hence, the maximum O2 purity that can be achieved in these processes is 95% (with the balance being Ar). Many niche applications in medical, semiconductor, military, and aerospace industries require O2 purities in excess of 95%.4,7−10 For such applications, special sorbents that show oxygen/argon selectivity are required. Very few adsorbents such as carbon molecular sieves,11,12 Ag-mordenite,13,14 AgZSM,14 and AgLiLSX15 have been reported to possess this selectivity. Because the properties of O2 and Ar are very © 2018 American Chemical Society

similar, the reported O2/Ar selectivities are typically less than 1.29. Accordingly, the generation of high-purity O2 is very challenging and only a few process studies exist in the literature.13,16−20 Most studies report oxygen purity in excess of 99% with a corresponding maximum recovery of ∼14%. A summary of the studies is provided in our previous publication.21 Silver-exchanged titanosilicates (Ag-ETS-10), synthesized by exchanging Ag with the traditionally prepared Na-ETS-10 has shown one of the highest O2/Ar equilibrium selectivity (1.49) reported in the literature.22,23 The adsorption affinity on AgETS-10 decreases in the following order N2 > Ar > O2. Further studies with the particulate form, i.e., crystal+binder demonstrated that the sorbent showed high-stability, and minimal loss in selectivity.21,24 Breakthrough studies also confirmed the fact that the mass transfer was fast and that the breakthrough profiles can be adequately described by a one-dimensional mathematical model.24 In our earlier publication, multiple vacuum swing adsorption (VSA) cycles were studied for O2 Received: Revised: Accepted: Published: 8997

May 26, 2018 June 14, 2018 June 14, 2018 June 15, 2018 DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research

to understand the purity-recovery performance of the various cycles. An optimization to minimize total energy consumption and maximize O2 productivity for high-purity O2 generation is considered and the performance of the single-stage process is compared to that of the dual-stage process. Finally, a simple scheduling study is performed in order to estimate the number of columns required to make these configurations into a continuous process.

purification from a N2-free feed that consisted of 95% O2 and 5% Ar.21 Three cycle configurations, namely 4-step Skarstrom, 6-step with pressure equalization (PE), and light product pressurization (LPP), and 6-step with PE and heavy product pressurization (HPP), were considered and it was concluded that the two 6-step cycles provide a higher O2 recovery compared to the Skarstrom cycle. The 6-step with PE and LPP showed a 27.3% O2 recovery for a product containing 99.5% O2. This value was significantly improved to 91.7% for the 6step with PE and HPP. The energy-productivity optimizations in our earlier publication indicated a lower energy consumption for the same level of productivity for the two 6-step cycles compared to the Skarstrom cycle and it was concluded that for a better comparison between the configurations, N2 removal stage should be also explored. The purpose of this study is to explore the possibility of O2 purification from air using a single-stage separation process through experiments and simulations and later compare the performance indicators with the dual-stage configurations. The strategy of this work in exploring multiple options is summarized as a block diagram in Figure 1. The first

2. MATERIALS AND METHODS 2.1. Materials and Isotherms. The Ag-ETS-10 extrudates used in this study, similar to the ones reported earlier, were made by using the hydrothermal method.22,24,25 The Na-ETS-10 form was first synthesized and later exchanged with silver nitrate to form Ag-ETS10. The physical properties of the samples used in the simulations were reported in our earlier publication24 and are reported in the Supporting Information. Adsorption isotherms of N2, O2, and Ar at 303.15, 323.15, and 343.15 K up to 115 kPa, measured using the volumetric technique, along with the Langmuir model parameters, are reported in the Supporting Information. It is worth noting that the extrudates used in this study have been repeatedly used for over four years and they show reproducible isotherms confirming their excellent stability. In this study, the Skarstrom cycle using zeolite 5A was considered for the first stage. The adsorption equilibria was obtained from Mofarahi et al.26 The fitted dual-site Langmuir adsorption isotherm model parameters for N2 and O2 adsorption on zeolite 5A are shown in the Supporting Information. It is worth noting that adsorption isotherm model parameters for Ar were considered to be identical to the ones for O2 because zeolite 5A has no thermodynamic selectivity for Ar over O2. 2.2. Experimental Apparatus. The experimental setup used in this study is shown in the Supporting Information. A single column (32 cm long and 3.6 cm internal diameter), packed with 247 g of AgETS-10 extrudates was used for the experiments. The flow rates of the gases were controlled using two flow controllers (Alicat, Tucson, AZ, U.S.) at the inlet. Three mass flow meters (Alicat, Tucson, AZ, U.S.) were mounted at different positions in the system to measure the flow rate of the streams leaving the column during adsorption, light reflux, and evacuation steps. Three pressure transducers (Omega Engineering, Stamford, CT, U.S.) were used to monitor the pressure at the inlet and outlet of the column in addition to the pressure inside the reflux tank. A two-stage vacuum pump (Pfeiffer Vacuum, MVP 040−2 Asslar, Germany) was used to evacuate the column. The temperature in the column was measured using two thermocouples (Omega Engineering, Stamford, CT, U.S.) located at 8 and 24 cm from the column inlet. A tank of volume 300 cc was used to collect a portion of the raffinate product to be used as a reflux. A metering valve is introduced in the line connecting the column and the tank. This valve reduces the abrupt pressure changes that occur during when the column is switched to the light reflux step. The composition of the raffinate stream was detected using a mass-spectrometer (Pfeiffer Vacuum OmniStar GSD 320, Asslar, Germany) that was calibrated prior to the start of the experiment. The experimental rig was fully automated with electrically activated solenoid valves (ASCO, Florham Park, NJ, U.S.) that provided the switching mechanism to implement the desired cycle configurations. A Labview-based data acquisition interface was used to control all the instruments and record relevant process data. All gases in this study were obtained from Praxair Canada Inc.

Figure 1. Process configurations explored in this study for high-purity O2 production: (a) single-stage O2 production. (b) dual-stage O2 production.

configuration considers single-stage O2 purification directly from dry air. In this configuration, two cycles, viz., the Skarstrom and the 6-step with PE and HPP cycles were studied. The second configuration considers a dual-stage purification where the first-stage is used for nitrogen removal, whereas the second-stage focuses on oxygen purification. In this configuration a Skarstrom cycle is considered for the first stage, while three different cycles, viz., the Skarstrom, the 6step with PE and LPP and the 6-step with PE and HPP are studied for the second stage. Simple single-column Skarstrom experiments with air feed were compared with the model predictions. Later a detailed process optimization is performed

3. PROCESS MODELING AND CYCLE CONFIGURATIONS 3.1. Process Modeling and Optimization. The 1D axially dispersed plug flow model used for the simulations has been explained in detail in the literature.21,24,27 The model assumes that the gas is ideal; heat transfer is possible across the 8998

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research

Figure 2. Cycle configuration and pressure profiles of the three VSA cycles: (a) Skarstrom, (b) 6-step with PE and LPP, (c) 6-step with PE and HPP.

walls; the solid and gas are at thermal equilibrium; the specific heat capacity of the adsorbed phase is that of the gas phase; the binary and ternary equilibria between the three gases, O2, N2, and Ar are adequately represented by the competitive Langmuir isotherm; the mass transfer between the solid and the gas phase is described by the linear driving force (LDF) model and finally the mass transfer resistance is controlled by the diffusion in the macropores. The set of equations used for describing the column dynamics is provided in the Supporting Information. As it is common in PSA simulations, the dynamics of single column that progresses from one step to the other is simulated. Data buffers are used to store gas streams that are used either for reflux or for pressurization steps. This approach is described in our previous publication.27 The model equations were solved using an in-house MATLAB-based simulation and optimization program. The in-built MATLAB functions ode23s were used to solve the differential equations in the time domain, while the spatial domain was discretized using a finite-volume technique. Similar to our previous study, the purity-recovery and

energy-productivity optimizations were performed by implementing the nondominated sorting genetic algorithm II (GA) in MATLAB global optimization toolbox. Fifty generations, with a population size equal to 24 times of the number of decision variables, were created by the GA. The initial population was created using latin hypercube sampling (LHS) to prevent sampling bias. A desktop workstation with two 12-core INTEL Xeon 2.5 GHz processors and 128 GB RAM was used to conduct the parallelized optimizations. 3.2. Cycle Configurations. The three cycles used in this work are shown in Figure 2. The Skarstrom cycle, shown in Figure 2a, is a simple four-step cycle that comprises of an adsorption step at high pressure PH, evacuation at low pressure PL, light reflux at PL, and pressurization from PL to PH. The Skarstrom cycle has its advantages such as the simplicity and availability of the technology in the market. However, as discussed in our earlier publication, to enhance O2 recovery, more advanced cycles should be considered. The second cycle, shown in Figure 2b, is the 6-step cycle with pressure equalization (PE) and light-product pressurization (LPP). 8999

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research Table 1. Comparison of Experimental and Modelling Results for a Skarstrom Cyclea experiment

tADS (s)

tLR (s)

Pevac (bar)

PuO2,sim([%)

A B C D E

5 5 5 8 12

15 15 15 12 8

0.1 0.14 0.2 0.14 0.14

99.51 99.49 99.1 98.6 97.4

PuO2,exp (%)

ReO2,sim (%)

ReO2,exp (%)

± ± ± ± ±

10.13 11.53 12.57 18.2 28.84

9.75 10.6 11.77 17.95 28.1

98.2 98.3 97.7 97.7 96.9

0.5 0.5 0.5 0.5 0.5

In the experiments, the flow rate during the adsorption and pressurization steps was 900 sccm.

a

Figure 3. Skarstrom VSA experiments corresponding to cycle B listed in Table 1. Symbols and lines correspond to experimental and calculated histories, respectively. (a) Pressure history, (b) temperature history for thermocouple located at 8 cm from the column inlet; (c) flow rates of the streams leaving the column during the adsorption and light reflux (collection) step (d) flow rate of the gas stream leaving the column during the evacuation step.

This cycle consists of an adsorption step operated at PH; a PE step in which the gas from the raffinate product end is used to equalize with the receiver column; a counter-current blow evacuation step carried out at PL; a light reflux operated at PL, where the gas from the raffinate product of the adsorption step is used as a reflux; and a LPP step where part of the raffinate product is used to pressurize the column. The third configuration used, namely, the 6-step cycle with PE and heavy product pressurization (HPP), is shown in Figure 2c.

This cycle involves an adsorption step at PH; a PE step; a counter-current evacuation step operated at PL; a PE receiver step and the a pressurization with the extract product; and a feed pressurization step to return the pressure to PH.

4. RESULTS AND DISCUSSION 4.1. Single-Stage Production of High-Purity Oxygen. Single-stage O2 purification from air is preferred over a dual9000

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research

magnitude and the rate of change of the temperature swing are well predicted. The rate of pressurization and evacuation were fitted to the experimental results and the comparison of the experimental and calculated column pressures is shown in Figure 3a. The following function was used to describe the transient pressure profile during pressurization and evacuation steps:

stage process due to its simplicity and the reduction in the number of fluid movers that are required in the process. In this study, the possibility of producing high-purity O2 in a single stage using Ag-ETS-10 was explored experimentally and later two cycle configurations, namely the Skarstrom and 6-step with PE and HPP cycles were optimized both for O2 purity-recovery and energy-productivity values and were compared against each other. 4.1.1. Experimental Demonstration. The main goal of the experimental runs were 2-fold. First, to demonstrate that AgETS-10 can produce, under cyclic conditions, an O2 product in excess of 95%. Second, to validate the model predictions, both overall process performance and transient operation, by comparing them to experimental results. The simple 4-step Skarstrom cycle experiments were conducted and the results were compared with the model predictions. Adsorption (ADS) and evacuation (EVAC) steps were performed at PH and PL, respectively. The light reflux step (LR) was conducted at PL and the pressurization (PRESS) step with the feed stream was performed to pressurize the column from PL to PH. A onesecond idle step, in which all valves remain closed, was introduced in between two steps of the cycle. This allowed the pressure profiles to stabilize and ensured that all product was positively collected. Five experiments (labeled A to E) with the operating conditions reported in Table 1 were performed up to 70 cycles. All experiments were started by saturating the bed with pure O2. The transient product composition and temperature history were monitored in order to ensure cyclic steady-state has been reached. The effect of three different durations of the light reflux step (tLR) on O2 purity was investigated (experiments B, D, and E) while the other operating conditions were kept constant between the three experiments. As summarized in Table 1, experimental O2 purity was above 95% for all three cases and both experimental purity and recovery was closely predicted by the model. As expected, increasing light reflux ratio (i.e., increasing tLR while keeping tADS constant) improved O2 recovery at the cost of purity. In another parametric study, low pressure during evacuation step (PL) was changed while maintaining the other operating conditions constant (experiments A, B, and C). Similar to the previous case study, experimental O2 recovery is well predicted by the model while experimental O2 purity is under predicted. The absolute error in O2 purity between simulation and experiment changes from 1.2% in experiment B (tADS = 5s) to 0.5% in experiment E (tADS = 12s). The underprediction of purity values might be due to the difficulty in measuring O2 concentration during short adsorption steps or the disturbance caused by the solenoid valves. However, it should be noted the high purities obtained using the simple Skarstrom cycle verified the capability of Ag-ETS-10 for highpurity O2 production from air in a single-stage process. Further, the simulations do predict the correct trend of purity and recovery when the evacuation pressure and duration of the light reflux step were varied. In addition to this, O2 recoveries are well predicted and are, to the best of our knowledge, the highest reported O2 recovery for a product in excess of 95% using a simple single-stage adsorption process. Apart from the performance indicators such as O2 purity and recovery, transient pressure, temperature, and flow rate profiles were also recorded and compared with the predictions from the model. The predicted and experimental histories of pressure, temperature and flows from experiment B at cyclic steady-state are shown in Figure 3. It can be seen that both the

PZ = 0 = P1 + (P2 − P1)e−λ1t + (PH − P2 − P1)e−λ2t

(1)

where t is the time during the step and λ1, λ2, P1, and P2 are the fitting parameters and were fitted to the experimental pressure profiles during the pressurization and evacuation steps. This function was later used in the simulations to predict the pressure profiles at each of the two steps as shown in Figure 3a. The small pressure increase seen during the LR step arises from the pressure difference between the tank and the column. These changes were not accounted for in the model. It is important to note that the values of λ1, λ2, P1, and P2 were the only set of fitting parameters used in the simulations. All other parameters describing adsorption equilibrium, kinetics and heat transfer were identical to those reported in the previous publications.21,24 As shown in Figure 3b−d the simulations capture the temperature and exit flow rates from the adsorption and evacuation steps well. The close prediction of the transients and the oxygen purity and recovery confirms the capability of the model to describe dynamics of this separation process. 4.1.2. Purity-Recovery Optimization. As stated in the introduction, in this study a dry air feed (78.0% N2, 21.0% O2, 1.0% Ar) was introduced to the column and the operating conditions were optimized for higher purities and recoveries. The bounds for the decision variables are listed in Table 2 and Table 2. Range of Decision Variables Used in Optimization of the VSA Cycles

a

decision variable

range

tADS (s) tEVAC (s) tLR (s)a tHPP (s)b vADS (m s−1) PL (kPa)

2−10 2−10 2−10 0.02−9.9 0.005−0.2 3.0−60.0

Skarstrom and 6-step with PE and LPP. b6-step with PE and HPP.

all the other parameters are provided in the Supporting Information. Bed length was fixed at 0.32 m similar to the column available in our laboratory and PH was fixed at 100 kPa. Two cycle configurations, the Skarstrom and the 6-step with PE and HPP were considered for the single-stage O 2 purifications. Purity and recovery of the Skarstrom cycle is defined as PuO2 = ReO2 =

nOADSout − nOLRin 2 2 ADSout LRin ntotal − ntotal

nOADSout − nOLRin 2 2 nOADSin + nOFPin 2 2

(2)

where n is the number of moles collected during the corresponding step. Figure 4 shows the purity-recovery Pareto front for the Skarstrom cycle which indicates that 99.5% O2 9001

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research

in excess of 99.5 and 99.8% can be achieved at a recovery of 82.0 and 79.5%, respectively. It is worth mentioning that for on-site applications equipment size is an important design parameter and although the feed for this separation, i.e., atmospheric air, is abundant, higher recovery values reduces the size of the unit. 4.1.3. Energy-Productivity Optimization. A second set of optimizations were conducted to minimize total energy consumption and maximize the productivity of the cycles. These optimizations were used to compare different possible approaches to purify O2. The range of decision variables used were identical to the purity-recovery optimizations reported in Table 2. The constraint on O2 purity was implemented by introducing a penalty function. Productivity is defined as the total tonnes of O2 in the product stream divided by the volume of adsorbent per day: Figure 4. Pareto fronts from process optimization to maximize O2 purity and recovery for the single-stage 6-step cycle with PE and HPP with air feed.

Pr =

adsorption step: EADS γ 1 = ϵπrin2 (γ − 1) η

∫t =0

ÄÅ ÉÑ γ − 1/ γ ÅÅi ÑÑ ÅÅjj P(t ) zyz Ñ − 1ÑÑÑÑdt vADSP(t )ÅÅÅjj z z ÅÅjk Pfeed z{ ÑÑ ÑÑÖ ÅÇÅ ÄÅ ÉÑ γ − 1/ γ ÅÅi ÑÑ ÅÅjj Patm yzz Ñ − 1ÑÑÑÑdt v(t )P(t )ÅÅÅjj zz ÅÅk P(t ) { ÑÑ ÅÇ ÑÖ

evacuation and light reflux steps: E γ 1 = ϵπrin2 η (γ − 1)

t = tstep

∫t =0

(6)

The total energy consumption for the processes is calculated as

ADS

ET =

EADS + E EVAC + E LR total tonnes of O2 in the raffinate product

(7)

Note that for the 6-step cycle with PE and HPP ELR = 0. A rigorous optimization was conducted to minimize the total energy consumption and maximize O2 in a single-stage Skarstrom and 6-step with PE and HPP cycles and the energy-productivity Pareto fronts are shown in Figure 5. This figure indicates that the single-stage 6-step cycle with PE and HPP showed a significantly lower energy consumption for the same level of O2 productivity compared to the single-stage Skarstrom cycle. For instance, at identical productivity of 0.94 tonne O2/m3 Ag-ETS-10.day, the Skarstrom cycle consumes 2815 kWh/tonne O2, while the 6-step cycle with PE and HPP consumes 515.5 kWh/tonne O2. As explained earlier, this configuration enhances O2 recovery since part of the extract product is used to pressurize the column and saves power with the addition of the pressure equalization (PE) step. 4.2. Dual-Stage Production of High-Purity Oxygen. The possibility of producing high-purity O2 in a dual-stage process was also studied in this work. In this approach, as shown in Figure 1b, the majority of N2 is removed in stage 1. The product stream enriched in Ar and O2 is introduced to the O2 purification stage (stage 2) to reach the desired purity of

nOADSout 2 ADSout ntotal

nOADSout 2 nOADSin 2

t = tADS

(5)

Considering the cyclic nature of the process, calculating PI for each cycle can be time-consuming, especially considering the fact large-scale optimization has to be performed. In order to overcome this challenge, a large set of operating conditions were considered and PI was calculated using a rigorous procedure. Once this was done, the value of PI can be interpolated for given values PL and θ. This procedure significantly reduced the time required for optimization without significantly compromising the description of the process. For this cycle, the O2 purity and recovery were calculated as the following:

ReO2 =

(4)

Total energy consumption for one cycle is calculated depending on the type of steps in each cycle and is reported in terms of kilowatt hours per tonne of O2 produced per cycle. The energy consumption of the various steps are calculated as follows:

purity can be obtained at 11.7% recovery. This shows the ability of Ag-ETS-10 to produce high-purity O2 from air in a single-stage due to its higher Ar/O2 selectivity compared to other selective sorbents. However, it should be noted that according to Figure 4, it is not possible to obtain O2 purity in excess of 99.9% with the Skarstrom cycle. As presented in our previous publication, the 6-step VSA cycle with PE and HPP improved O2 recovery since part of the stream leaving the evacuation step is used to pressurize the column. This cycle showed a promising O2 recovery of 91.7% for a 99.5% O2 purity for a feed stream containing 95% O2/5% Ar. In this study single-stage O2 purification using the same cycle was studied. The decision variables and their bounds are shown in Table 2. In order to calculate the intermediate pressure PI, the amount of mass leaving the donor column during the PE step should be the same as the amount introduced to the receiver column. This value depends on the t PL and the ratio of θ = tHPP chosen by the optimizer.

PuO2 =

tonnes of O2 in the raffinate product (total volume of Ag‐ETS‐10)(cycle time)

(3)

The HPP step was followed with a feed pressurization step if the pressure in the column did not reach PH during the HPP step and the O2 introduced to the column during this step should be also added to the denominator of the definition of recovery in eq 3. Figure 4 shows the purity-recovery Pareto curve obtained for this cycle. As can be observed, purity values 9002

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research

performance indicators of the process. The results are then compared with the performance of the single-stage processes. 4.2.1. Optimization of Nitrogen Removal Stage (Stage 1). Since air consists of 78% N2 and the aim of the process is to obtain high-purity oxygen, a natural choice would be to remove the bulk of nitrogen in the first stage. With this aim, it is sufficient to employ a rather inexpensive adsorbent. Accordingly, Zeolite 5A was chosen as the adsorbent of choice for stage-1. The maximum O2 purity that can be obtained from dry air using zeolite 5A is 95.0% because zeolite 5A has no selectivity for Ar/O2. In addition to the product containing 95.0% O2/5.0% Ar, three product streams with different nitrogen content varying between 30.0% to 70.0% and the rest balanced with (O2 + Ar) were considered. The ratio of yO2/yAr was kept equal to 21.0, i.e. same ratio of O2/Ar in air, for all the product streams. Each of these four N2 purity values in the product stream was set as a constraint for the simulation and the corresponding energy-productivity Pareto curves were obtained by optimizing the operating parameters within the bounds shown in Table 2. The results shown in Figure 6a indicate that the total energy consumption of the N2 stage (En1) increases while the productivity (Pr1) drops by tightening the purity constraint on the N2 content in the product stream of this stage, i.e. smaller N2 concentration. The

Figure 5. Comparison of energy-productivity Pareto fronts for singleand dual-stage configurations for obtaining O2 purity ≥99.5%.

O2. Optimizing integrated two-stage processes can be very challenging. In this section, instead of optimizing the twostages together, each of these two stages were optimized independently and were later combined to calculate the overall

Figure 6. Results of process optimization of a Skarstrom cycle to maximize O2 productivity and minimize energy consumption for dry air feed delivering 99.5% O2 product. (a) First stage with zeolite 5A delivering product with different levels of N2 content; (b) second stage with Ag-ETS-10 with different levels of N2 content in the feed; (c) comparison of multiple dual-stage O2 purification approaches with single-stage configuration. 9003

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research

Figure 8. In this figure, gas phase axial concentration profiles of N2 at the end of the evacuation step at cyclic steady state normalized against yN2 in the feed, and O2 solid phase axial concentration profiles for the same step is plotted for three different compositions of N2 in the feed. As it can be observed in Figure 8a, increasing N2 content in the feed stream saturates the feed end of the column with nitrogen because it is the heavier component, and N2 front moves further toward the product end for higher values of yN2 in the feed. This results in pushing the O2 front to the product end and avoids losing O2 during the evacuation step. This effect is clearly visible in Figure 8b and can explain the improvement in O2 recovery for higher N2 content as discussed earlier in Figure 7. However, additional N2 content in the feed stream results in higher energy consumption during the evacuation step and lower oxygen productivity for that cycle. Hence, a detailed comparison should be made in order to compare a singlestage with a dual-stage oxygen separation process. A similar approach to the one used for the first-stage was applied for the energy-productivity optimization of the second stage. Multiple feed streams with N2 concentrations, similar to those reported in Figures 7 and 6a, were introduced to the second stage. The purity constraint in the product stream from stage-2 was set to be ≥99.5% O2. The results shown in Figure 6b show that energy consumption increases with increase in N2 composition in the feed to the second stage. However, the productivity of the second stage (Pr2) stays within the same range in spite of N2 concentration level in the feed stream (product of the first stage). The results presented in Figure 6a, b were combined using an analytical approach to calculate the overall energy-productivity values and are further discussed in the next section. 4.2.3. Optimal Design of Dual-Stage Process. An analytical approach was used in order to calculate the overall productivity and total energy consumption of the dual-stage separation processes introduced in the previous sections. Productivity is defined as the total tonnes of O2 in the product stream divided by the total volume of adsorbent per day. Therefore, overall productivity for a dual-stage process can be calculated as

product of this stage is directed to the O2 purification stage discussed in the next section. 4.2.2. Optimization of Oxygen Purification Stage (Stage 2). Similar to the previous section, in addition to the feed containing 95.0% O2/5.0% Ar, four feed streams with different nitrogen content varied between 30.0 and 78.0% and the rest balanced with O2 and Ar were introduced to the column. Each of these five different feed compositions were set as an input for the simulation and the corresponding purity-recovery Pareto curves were obtained by optimizing the operating parameters within the bounds shown in Table 2. The Pareto curves for purity-recovery optimization are shown in Figure 7.

Figure 7. Pareto fronts from process optimization to maximize O2 purity and recovery for Skarstrom VSA cycle with different N2 content in the feed.

Interestingly, it can be observed that except for the ultra highpurity region (PuO2 > 99.9%), the recovery of O2 in the product stream for a particular value of oxygen purity increases with an increase in the N2 content in the feed stream. For instance, O2 recovery for a raffinate stream containing 99.5% O2 reaches 11.7% when air (78.0% N2) is fed into the column during the adsorption step compared to a 7.0% recovery when the feed contains no N2. This effect can be further explained by comparing the axial concentration profiles of N2 and O2 for the conditions at which a purity of 99.5% is achieved, as shown in

Prtotal =

Pr2 1+

Pr2β Pr1Re 2

(8)

Figure 8. Effect of N2 content in the feed on the concentration profiles for stage-2 using Skarstrom cycle: (a) Normalized N2 gas phase axial concentration profiles of the evacuation step (normalized against yN2 in the feed); (b) solid phase O2 concentration profiles of the evacuation step in the Skarstrom cycle. 9004

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research where 1 and 2 correspond to the first and second stages, respectively. Because productivity is defined per unit volume of the adsorbent and different materials were used in each of the two stages, the volume of the sorbent used in the first stage was correlated to the volume of the sorbent used in the second stage through parameter β. This constant is defined as the ratio of the cost of a unit volume of the sorbent used in the first stage (N2 removal) over the cost of a unit volume of the material used in the second stage (O2 purification). This assisted in normalizing the volumes based on the price of the adsorbent in the market and offered a proxy for the capital cost involved using different configurations. In this study, β for zeolite 5A over Ag-ETS-10 was considered to be equal to 0.05 based on the prices available in the market. Therefore, the overall productivity values reported are in tonnes of O2 produced per cost normalized volume (CNV) of Ag-ETS-10 per day. Total energy consumption is reported in terms of kilowatt hours per tonne of O2 produced per cycle. Hence, overall energy consumption of a dual-stage process is defined as Entotal =

En1 + En2 Re 2

5. The values of energy-productivity optimization for the two 6-step cycles reported in our earlier publication were used for Pr2 and En2 and the values for the Skarstrom cycle delivering 95.0% O2 and 5.0% Ar with zeolite 5A was considered for Pr1 and En1 in eqs 8 and 9. Figure 5 indicates that compared to the single-stage Skarstrom design, higher productivity was obtained at the same level of energy consumption using the two-stage configuration that employs 6-step with PE and LPP or the 6step with PE and HPP cycles. The energy consumption was lower for the 6-step cycle with PE and HPP which was predominantly due to higher recoveries achieved in this configuration according to eq 9. On the one hand, the singlestage O2 purification using the 6-step cycle with PE and HPP presents a better performance compared to the other approaches with significantly lower energy consumption for the same range of productivity. On the other hand, the dualstage configuration offers a slightly higher productivity, although at a higher energy consumption. 4.3. Cycle Scheduling. Scheduling of an adsorptive process is an important step for practical implementation of various cycle configuration. In this study, we consider a simple scheduling strategy that is suited for a process where a continuous feed is desired. If the continuous feed is not required, a single-column process with product tanks would suffice for each stage. As shown in the definition of the productivity, i.e., eq 4, the total cycle time is simply the sum of all individual step-times, which does not necessarily guarantee that a process will result in a continuous feed. In scheduling a particular cycle, it is also important to reduce the number of columns that are required. Therefore, the scheduling of the stages should be performed such that the optimal idle time and minimum number of columns are obtained. This is often achieved by considering the simulation results and introducing idle steps. For the cycle configurations that are discussed in this study, more constraints should be considered, such as the pressure equalization donor and receiver columns must align and the light reflux (LR) step should start while another column is undergoing an adsorption (ADS) step. A simple algorithm, based on the graphical methodology proposed by Mehrotra et al.,28 was developed. The set of decision variables corresponding to the Pareto points in Figure 6 were considered. This scheduling algorithm introduced idle times at certain positions in between the steps to make the cycle continuous. For each point on the Pareto front, the number of columns were calculated and examples of configurations that resulted in the fewest number of columns are shown in Figures 9 and 10. Figure 9 shows the schedules of two cycles that employ AgETS-10 for high-purity oxygen purification from an air feed. The Skarstrom cycle that requires four columns is shown in Figure 9a, whereas the 6-step cycle with PE and HPP, that requires 8 columns, is shown in Figure 9b. In both cases, the duration of the adsorption step is shown at the top-left corner of each cycles and the duration of the other steps are proportional to their lengths reported in the figure. Note that for the Skarstrom cycle, the start of the adsorption and the LR steps are well-aligned. In a similar manner, from Figure 9b it can be seen that the pressure-equalization donor of a particular column is well aligned with the receiver and the EVAC step is aligned with the HPP step. Figure 10 shows the schedules for the dual-stage configuration. Figure 10a shows a schedule for a 4-column

(9)

where En1 and En2 correspond to the total energy consumption of the first and the second stage, respectively. It is worth noting from eqs 8 and 9 that the recovery of the second stage plays an important role in the overall productivity and total energy consumption. In other words, the higher the recovery of the second stage, the lower the energy that should be spent in the first stage since the product of the N2 removal stage is introduced to the O2 purification stage. The approach introduced thorough eqs 8 and 9 was used to calculate the overall productivity and energy consumption of the dual-sage process by combining the values from Figure 6a, b. The values used for En2 and Pr2 in eqs 8 and 9 correspond to the points with lowest energy consumption in Figure 6b. Figure 6c shows the overall energy-productivity fronts for different concentrations of N2 entering the second stage and the comparison with single-stage Skarstrom O2 purification with dry air as the feed discussed earlier. The overall energy consumption decreases by increasing the N2 content in the product stream form the nitrogen removal stage. Further, this figure clearly presents the advantage of the single-stage process over the dual-stage options in terms of total energy consumption. According to eq 9, this could be due to the low O2 recovery in the Skarstrom cycle and therefore, it was concluded that there was no advantage in using Skarstrom cycle configuration for the O2 purification stage in a dual-stage process. As discussed in our earlier publication, the 6-step with PE and LPP and the 6-step with PE and HPP cycles could deliver 99.5% O2 at 27.3 and 91.7% recovery, respectively and considering these two cycles as the second stage might be more beneficial compared to the Skarstrom cycle.24 It is worth noting that because recovery was not imposed as a constraint for the energy-productivity optimizations, the points on the energy-productivity Pareto fronts do not necessarily have the same recovery as the ones on the purity-recovery Pareto fronts. However, the chances of achieving a higher recovery in the energy-productivity optimization for a cycle which performed better in the purity-recovery optimization, are still higher. Therefore, the possibility of using the other two 6-step cycles for stage-2 was investigated and the results are shown in Figure 9005

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research

with purity of 98.3 ± 0.5% with a recovery of 10.6% was obtained from a dry air feed stream. The numerical simulations were able to predict the experimental oxygen purity and recovery within an average value of 1.2 and 0.6 percentage points, respectively. The numerical simulations were also able to predict the transient profiles of the column pressure, temperature, and flow rates. The effect of N2 content in the feed on the product purity and recovery was studied using numerical simulations combined with multiobjective optimization, and it was concluded that increasing N2 content in the feed stream saturates the feed end of the column with nitrogen. This results in pushing the O2 front to the product end and avoids losing O2 during the evacuation step. Thus, O2 recovery was slightly higher when N2 concentration in the feed was larger. The simulations predicted 82.0% recovery for a product with 99.5% O2 purity for a 6-step cycle with PE and HPP and dry air feed stream, which verified the advantage of this cycle over the other possible designs. The second set of optimizations were conducted in order to minimize total energy consumption and maximize overall productivity of multiple single-stage and dual-stage O2 purification approaches. A detailed comparison was performed between a group of cycles with zeolite 5A as the adsorbent used for N2 removal stage while delivering various levels of N2 concentration in the product. This product stream was later fed to the O2 purification stage where Ag-ETS-10 was used and the energy-productivity optimization results were compared with a single-stage Skarstrom cycle using Ag-ETS-10. The comparison presented no advantage in performing a dual-stage O2 purification when the Skarstrom cycle was used for both of the stages. However, using the 6-step cycle with PE and LPP or the 6-step cycle with PE and HPP for the O2 purification stage showed a higher productivity compared to the single-stage Skarstrom cycle. A rigorous energy-productivity optimization was conducted for the single-stage O2 purification using the 6step cycle with PE and HPP and it was concluded that this approach presents a better performance compared to the other cycle configurations with significantly lower energy consumption for the same values of productivity. A simple graphical scheduling was also conducted for the optimal points for the better performing designs. The minimum number of columns required for a continuous process was 4 for the single-stage Skarstrom and 8 for the single-stage 6-step cycle with PE and HPP. This value was 8 (i.e., 4 + 4) for a dual-stage process with the Skarstrom cycle for the N2 removal stage plus the 6-step cycle with PE and LPP for the O2 purification stage. When the 6-step cycle with PE and HPP was used as the second stage, the minimum number of columns were 7 (i.e., 4 + 3). This study clearly demonstrates the potential of adsorption processes, using Ag-ETS-10, for high-purity O2 generation. It also demonstrates how process optimization can be used effectively to compare various processes and to develop cycles that can significantly improve the recovery; reduce energy consumption and increase process productivity. Considering the fast kinetics and stability that has been exhibited by the adsorbent, future studies will focus on experimental demonstration of complex and faster cycles that can be implemented using fewer columns.

Figure 9. Cycle scheduling for the optimal operating conditions corresponding to the single-stage configuration. Idle and HPP steps are indicated by gray and black blocks, respectively. (a) Single-stage Skarstrom cycle using Ag-ETS-10 for O2 purification. Energy consumption for this cycle is 2250 kWh/tonne O2. (b) Single-stage 6-step cycle with PE and HPP using Ag-ETS-10 for O2 purification. Energy consumption for this cycle is 598.71 kWh/tonne O2.

Figure 10. Cycle scheduling for the optimal operating conditions for the dual-stage configuration.: (a) Skarstrom cycle using zeolite 5A for N2 removal (stage 1). Energy consumption for this cycle is 493.87 kWh/tonne O2. (b) 6-step cycle with PE and LPP using Ag-ETS-10 for O2 purification (stage 2). Energy consumption for this cycle is 193.24 kWh/tonne O2. (c) 6-step cycle with PE and HPP using AgETS-10 for O2 purification (stage 2). Energy consumption for this cycle is 251.19 kWh/tonne O2.

Skarstrom cycle that employs Zeolite 5A for stage-1. Figures 10b, c show the schedules for the 6-step process with PE+LPP and the 6-step process with PE+HPP, respectively. These correspond to the second stage that employs Ag-ETS-10, performing the oxygen purification. It can be seen that the entire two-stage process can be completed using 7 columns. It is worth pointing out that a more rigorous optimization that takes into account the number of columns, vacuum pumps and operating costs have to be considered in order to select the best configuration for practical implementation. Further, the requirement of having a continuous feed may vary according to the scale. Smaller systems, could be operated under intermittent feed, whereas larger systems are typically operated under continuous feed. This is another aspect that needs to be considered in future studies.



5. CONCLUSIONS In this study, the possibility of producing O2 with 99.5% purity from dry air in a single-stage VSA cycle was explored. Singlecolumn Skarstrom cycle experiments were performed and O2

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.iecr.8b02345. 9006

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research



LR LPP out

Schematic diagram of experimental VSA apparatus; single-component, low-pressure adsorption isotherms of N2, Ar, and O2 on Ag-ETS-10 extrudates and their fitted Langmuir adsorption parameters; equations for modeling adsorption column dynamics; parameters used in cycle simulations and optimizations; operating conditions for selected points from Pareto fronts shown in Figures 5 and 6c (PDF)



AUTHOR INFORMATION

*E-mail: [email protected]. ORCID

Arvind Rajendran: 0000-0003-4367-4892 Present Address †

S.A.H.H. is currently at Department of Chemical Engineering, Imperial College London, London SW7 2AZ, U.K. Notes

The authors declare the following competing financial interest(s): S.M.K. has a financial interest in Extraordinary Adsorbents, Edmonton, which has commercialized Ag-ETS-10.



ACKNOWLEDGMENTS The authors thank Lan Wu, Tong Qiu, Ashwin Kumar Rajagopalan, and Nicholas Wilkins, University of Alberta, for their assistance. This research was enabled in part by support provided by WestGrid (www.westgrid.ca and Compute Canada (www.computecanada.ca. Financial supports from Helmholtz-Alberta initiative (HAI) and NSERC for their sponsorship of the industrial research chair in molecular sieve nanomaterials are gratefully acknowledged. The authors thank Innotech Alberta for providing access to the VSA experimental system. F P PH PI PL P1 P2 t v

NOMENCLATURE volumetric flow rate [ccm] pressure [Pa] high pressure in cycle simulation [Pa] intermediate pressure in cycle simulation [Pa] low pressure in cycle simulation [Pa] fitting parameter used to describe pressure transient [Pa] fitting parameter used to describe pressure transient [Pa] time [s] interstitial velocity [m s−1]

β ϵ η λ1 λ2 γ

GREEK SYMBOLS price factor to normalize volume of adsorbent bed voidage vacuum pump efficiency fitting parameter used to describe pressure transient [s−1] fitting parameter used to describe pressure transient [s−1] adiabatic constant

■ ■

REFERENCES

(1) Ruthven, D. Principles of Adsorption and Adsorption Processes; Wiley−Interscience: New York, 1984. (2) Yang, R. T. Adsorbents: Fundamentals and Applications; Wiley− Interscience: Hoboken, NJ, 2003. (3) Agrawal, R. Separations: Perspective of a Process Developer/ Designer. AIChE J. 2001, 47, 967. (4) Kroschwitz, J. Kirk-Othmer Encyclopedia of Chemical Technology; Wiley−Interscience; Hoboken, NJ, 2004. (5) Kumar, R. Vacuum Swing Adsorption Process for Oxygen Production-A Historical Perspective. Sep. Sci. Technol. 1996, 31, 877. (6) Sircar, S.; Myers, A. L. Gas Separation by Zeolites. In Handbook of Zeolite Science and Technology; CRC Press: Boca Raton, FL, 2003; Vol. 1063. DOI: 10.1201/9780203911167.ch22 (7) Office of the Federal Register (US). Public Health Parts 1 to 399; Code of Federal Regulations; U.S. Government Printing Office, Washintgon, DC, 2010. (8) European Industrial Gases Association Homepage. http://www. eiga.eu (accessed March 2016). (9) Performance Specification, Oxygen, Aviator’s Breathing, Liquid and Gas; MIL-PRF-27210G ; U.S. Department of Defense: Arlington, VA, 2009. (10) Reynolds, T. L.; Bailey, D. B.; Lewinski, D. F.; Roseburg, C. M.; Palaszewski, B. Onboard Inert Gas Generation System/Onboard Oxygen Gas Generation System (OBIGGS/OBOGS) Study. Part 1; Aircraft System Requirements; NASA/CR-2001−210903/PT1 ; National Aeronautics and Space Administration: Washington, D.C., 2001. (11) Jin, X.; Malek, A.; Farooq, S. Production of Argon from an Oxygen-Argon Mixture by Pressure Swing Adsorption. Ind. Eng. Chem. Res. 2006, 45, 5775. (12) Mohammadi, N.; Hossain, M. I.; Ebner, A. D.; Ritter, J. A. New Pressure Swing Adsorption Cycle Schedules for Producing High Purity Oxygen Using Carbon Molecular Sieve. Ind. Eng. Chem. Res. 2016, 55, 10758. (13) Knaebel, K. S.; Kandybin, A. I. Pressure Swing Adsorption System to Purify Oxygen. U.S. Patent 5 226 933, 1993. (14) Sebastian, J.; Jasra, R. V. Sorption of Nitrogen, Oxygen, and Argon in Silver-Exchanged Zeolites. Ind. Eng. Chem. Res. 2005, 44, 8014. (15) Ferreira, D.; Magalhães, R.; Bessa, J.; Taveira, P.; Sousa, J.; Whitley, R. D.; Mendes, A. Study of AgLiLSX for Single-Stage HighPurity Oxygen Production. Ind. Eng. Chem. Res. 2014, 53, 15508. (16) 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. (17) 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. (18) Santos, J. C.; Cruz, P.; Regala, T.; Magalhães, F. D.; Mendes, A. High-Purity Oxygen Production by Pressure Swing Adsorption. Ind. Eng. Chem. Res. 2007, 46, 591. (19) Ferreira, D.; Bárcia, P.; Whitley, R. D.; Mendes, A. Single-Stage Vacuum Pressure Swing Adsorption for Producing High-Purity Oxygen from Air. Ind. Eng. Chem. Res. 2015, 54, 9591. (20) Ferreira, D.; Boaventura, M.; Bárcia, P.; Whitley, R. D.; Mendes, A. Two-Stage Vacuum Pressure Swing Adsorption Using AgLiLSX Zeolite for Producing 99.5+% Oxygen from Air. Ind. Eng. Chem. Res. 2016, 55, 722. (21) Hosseinzadeh Hejazi, S. A.; Estupiñan Perez, L.; Rajendran, A.; Kuznicki, S. Cycle Development and Process Optimization of HighPurity Oxygen Production Using Silver-Exchanged Titanosilicates. Ind. Eng. Chem. Res. 2017, 56, 5679.

Corresponding Author



light reflux step light product pressurization step outlet stream

SUBSCRIPTS ADS adsorption step EVAC evacuation step FP pressurization step HPP heavy product pressurization step I intermediate in inlet stream L low 9007

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008

Article

Industrial & Engineering Chemistry Research (22) Shi, M.; Kim, J.; Sawada, J. A.; Lam, J.; Sarabadan, S.; Kuznicki, T. M.; Kuznicki, S. M. Production of Argon Free Oxygen by Adsorptive Air Separation on Ag-ETS-10. AIChE J. 2013, 59, 982. (23) Anson, A.; Kuznicki, S. M.; Kuznicki, T.; Haastrup, T.; Wang, Y.; Lin, C. C.; Sawada, J. A.; Eyring, E. M.; Hunter, D. Adsorption of Argon, Oxygen, and Nitrogen on Silver Exchanged ETS-10 Molecular Sieve. Microporous Mesoporous Mater. 2008, 109, 577. (24) Hosseinzadeh Hejazi, S. A.; Rajendran, A.; Sawada, J.; Kuznicki, S. M. Dynamic Column Breakthrough and Process Studies of HighPurity Oxygen Production Using Silver-Exchanged Titanosilicates. Ind. Eng. Chem. Res. 2016, 55, 5993. (25) Kuznicki, S. M. Large-Pored Crystalline Titanium Molecular Sieve Zeolites. US Patent 4,853,202, 1989. (26) Mofarahi, M.; Towfighi, J.; Fathi, L. Oxygen Separation from Air by Four-Bed Pressure Swing Adsorption. Ind. Eng. Chem. Res. 2009, 48, 5439. (27) Haghpanah, R.; Majumder, A.; Nilam, R.; Rajendran, A.; Farooq, S.; Karimi, I. A.; Amanullah, M. Multiobjective Optimization of a Four-Step Adsorption Process for Postcombustion CO2 Capture Via Finite Volume Simulation. Ind. Eng. Chem. Res. 2013, 52, 4249. (28) Mehrotra, A.; Ebner, A. D.; Ritter, J. A. Simplified Graphical Approach for Complex PSA Cycle Scheduling. Adsorption 2011, 17, 337.

9008

DOI: 10.1021/acs.iecr.8b02345 Ind. Eng. Chem. Res. 2018, 57, 8997−9008