Tao Tian,Yayan Wang,Bing Liu,Zhaoyang Ding,Xinxi Xu,*,Meisheng Shi,Jun Ma,Yanjun Zhang,Donghui Zhang,*

1 Institute of Medical Support Technology,Academy of System Engineering of Academy of Military Science of Chinese PLA,Tianjin 300161,China

2 School of Chemical Engineering and Technology,Tianjin University,Tianjin 300072,China

Keywords:Pressure swing adsorption Air separation Mathematical modeling Rotating distribution valve

ABSTRACT In this work,a six-bed pressure swing adsorption(PSA)process was investigated to produce medical oxygen from air,which uses the combination of six-way rotating distribution valve and PSA and has the main advantage of effectively saving space compared to the traditional two-bed or four-bed PSA process and can obtain greater productivity.The mathematical model of adsorption beds was developed based on the separation mechanism and the interaction among different equipment.Moreover,a pilot-scale device has been constructed to verify the accuracy of mathematical model by experiment.The oxygen product conformed to the medical standard (＞93% (vol)) with a recovery of over 57%.Some related parameters were also discussed in detail,such as step time,ratio of length to the diameter,flow rate of product.

1.Introduction

Affected by the covid-19 epidemic,the demand for ventilators and medical oxygen increased dramatically.It is inconvenient to transport oxygen in remote areas,so there is a great sense to develop air separation device to produce medical oxygen onsite[1,2].Large-scale apparatus like cryogenic fractionation units is not suitable for medical institutions because of its high investment cost and large area occupation.Therefore,the research for miniaturizing oxygen plant is of practical significance.Pressure swing adsorption (PSA) is advanced on compact equipment and high degree of automation,which attracts much attention from academia [3].Because of the limited production capacity of a small PSA unit,a medium-sized PSA unit might be more attractive for actual demand,even the number and scale of beds must be increased.The research of PSA oxygen production mainly focuses on numerical simulation and lab-scale experiments,which has been frequently studied before,but there are few studies on pilot-scale experiments [4].Moreover,simple Skarstrom process[5,6] or other four-bed processes [7-9] are applied more on air separation,but there are few reports about the application of multi-bed or new integrated valve like rotating distribution valve.

Many researchers have made great innovations in the field of simulation.The mechanism of mass transfer and heat transfer has been investigated and improved [10-13].The mass transfer rate and bed pressure drop model have been revised by experiments [14-16].S.Farooq and D.M.Ruthven [17,18] summarized models of pressure swing adsorption and adopted models for simulating equilibrium separation,and zeolite was selected as adsorbent.They also discussed the effects of mass transfer resistance and axial dispersion on system performance.Teagueet al.[19]proposed a mathematical model for board military aircraft and predicted the dynamic oxygen content and energy consumption.Ahari [20,21] developed a dynamic model,and then established a laboratory-scale device to verify the accuracy.In their study,the effect of kinetic parameters on the separation efficiency was considered.Santoset al.[22] compared and evaluated three different adsorbents.Yavaryet al.[23,24] found that more pressure equalization steps could improve the product recovery rate.Shuklaet al.[25] adopted different adsorption/desorption kinetics and thermodynamics models and verified the models by hydrogen separation.Liaoet al.[26,27] adopted two-bed PSA for purifying H2from methane steam reforming off gas,and researched the effects of adsorption pressure,adsorption time and purge-to-feed ratio(P/F ratio) on process performance.Otherwise,they designed a kriging-based PSA model and combined it with the hydrogen network model to conserve hydrogen resources in refineries.Wuet al.[28] built a four-bed and six-bed VPSA model and discussed the effects of adsorption pressure,adsorption time and purge-to-feed ratio(P/F ratio) on process performance.Guanet al.[29] proposed a dual-reflux pressure swing adsorption (DR-PSA) process and evaluated the process separation performance with different operating conditions.Liuet al.[30] simulated two-stage vacuum/pressure swing adsorption (VSA/PSA) process for CO2capture and H2purification and compared with other processes in detail.Voet al.[31] developed a rigorous dynamic model for an SMR and used artificial neural network (ANN) trained through 81 datasets to predict outputs with high accuracy.

In this work,a six-bed PSA process for air separation was proposed,and zeolite 5A was selected as adsorbent.A rotating distribution valve was employed to save space and reduce mechanical failures.A rigorous mathematical model for this system has been developed.A pilot-scale(medium-sized)device was built to verify the accuracy of mathematical models.The systematic analysis was implemented,including temperature,concentration,and pressure profiles within one cycle.Moreover,several operating parameters,like step time,ratio of bed height to diameter,product flow rate,were examined to observe the effects of these factors on the process performances.

2.Process and Mathematical Model for the VPSA System

2.1.Brief introduction for six-bed PSA process

As shown in Fig.1,the apparatus established in this work is equipped with six beds.The pressure in adsorption step is maintained at approximately 600 kPa while the pressure in regeneration step is nearly atmospheric pressure,101 kPa.The system of PSA process is designed to produce oxygen-enriched product with the flow rate of 6.0 m3·h-1and purity over 93%.Unlike the traditional two-bed and four-bed process for air separation,this apparatus is compact and space-saving due to the rotating distribution valve.Besides,when the same adsorbent is used and the purity of O2is basically the same,the productivity of the sixbed process is 0.0357 m3·h-1·(kg adsorbent)-1,which is higher than the four-bed process given in the literature [32],and its productivity is 0.0231 m3·h-1·(kg adsorbent)-1.

2.2.Determination and design of six-bed oxygen flow control system

There are two methods achieving the control of the oxygen production process,one is solenoid valve control,the other is multipass rotary distribution valve control.The former is simple to realize but means high failure rate and large floor area.The structure of the latter is complex,but it overcomes the shortcomings of the solenoid valve.The six-bed multi-pass rotary distribution valve is composed by the fixed valve,rotary valve,shaft,head assembly,sealing ring components,valve frame and other components.There are holes and slots connecting with the rotary valve,which are used for step switching,such as adsorption,equalization,purification,reversing,flushing and other functions.

The six-bed multi-pass rotary distribution valve plays a controlling role.As the Fig.1 shows,the bottom of the fixed valve has some through hole,4 small holes and 1 large hole,the side of the fixed valve has 6 middle holes and 6 small holes.The rotary valve has a similar structure,12 small holes,4 middle holes and 1 large hole at the bottom,4 middle holes and 4 small holes on the side.During actual operation,the six-bed multi-pass rotary distribution valve controls different steps of PSA by adjusting the rotary valve to different positions,in which rotating to 12 small holes on the rotary valve corresponding to the 12 steps of the PSA process.For example,the air flows into the adsorption bed from the inlet hole and the high purity oxygen is sent to the oxygen tank through the product hole during adsorption step.At the regeneration stage,the nitrogen desorbed from adsorbents is discharged from nitrogen hole to atmosphere.With the cooperation of the fixed valve and cylinder,the 6-bed-12 step PSA process is completed.

2.3.Process of six-bed PSA process

In this work,cycle sequence of the PSA process consists of adsorption step (AD),pressure equalization step (ED and ER),cocurrent blowdown step (CoD),counter-current blowdown step(BD),purge step (PUR) and the final pressurization step (FR).The time schedule is listed in Table 1 and Fig.2 displayed the schematic diagram of air-production apparatus.The total time of one cycle is set to 120 s.Notably,the rotating distribution valve is disassembled into 36 valves in numerical calculation.The valve switch status is illustrated in Table S1 (supporting information).

Table 1Time schedule of the PSA process

Fig.1.PSA apparatus and six-way rotary distribution valve.

The process is operated as follows.In AD1 step,the compressed air flows into the bed with valve A1 open,and nitrogen is selectively adsorbed by adsorbents.The oxygen product desorbs from the bed and flows into product buffer tank through the valve A4.During AD2 step,part of the oxygen product is employed to raise the pressure of bed B.

In ED1 step,bed A and bed C are directly connected to carry out equalization pressurization step,which increases the pressure of bed C.The other two equalization-depressurization steps are the same as the first one,but the objects are bed A/D and bed A/E,respectively.

The CoD gas is the light-component-enriched gas phase in bed A.It purges bed F through valve A5 and F5.In BD step,the remaining gas in bed A flows out from outlet A2.In PUR step,bed A is purged with the gas from bed B for further adsorbent regeneration.

Then bed A would undergo three equalization-pressurization steps (ER1,ER2,ER3) that are coupled with the three equalization-depressurization steps (ED1,ED2,ED3).In PR step,oxygen product gas enters bed A through valve A6 and its pressure is raised to nearly 500 kPa.

2.4.Mathematical model and process for the PSA system

Fig.3 depicts the mathematical model of adsorption bed,which is crucial for the six-bed-PSA simulation.In addition,the following assumptions are made to simplify the calculation according the process characteristic:

(1) The gas phase behaves as an ideal gas.

(2) There is no radial variation in gas concentration,temperature and pressure.

(3) Pressure drop along the bed is calculated by the Ergun equation.

(4) The gas and solid phase are always in thermal equilibrium.

(5) The porosity of bed and adsorbent particle is uniform along the bed.

(6)The inter-phase mass transfer coefficient is expressed by the linear driving force (LDF) model.

(7) Extended-Langmuir model is used to describe the adsorption behaviors.

Based on the assumptions listed above,complete mathematical models of mass,energy,momentum,adsorption equilibrium as well as LDF equations are established to describe the dynamic adsorption behaviors of air on molecular sieve and listed in the following Table 2.The physical parameters of adsorption bed are presented in Table 3 and the initial and boundary conditions of adsorption column are defined in Table 4.

The adsorption isotherms,measured by the static volumetric method,are displayed in Fig.4 [35,36].The isotherms were measured at 293.15,298.15,303.15 and 308.15 K with the pressure profiles ranging from 0 to 0.75 MPa.The Langmuir equation parameters,as demonstrated in Eq.(4),are obtained from curve fitting and are listed in Table 5.To ensure the simulation results,the partial differential equations(PDEs)are discretized by Upwind Differencing Scheme1(UDS1)method and each column possesses 100 nodes in the spatial coordinate[30].PDEs are converted into a series of differential algebraic equations (DAEs),which could be solved by the implicit Euler integration method.Both the absolute and relative tolerances of mixed Newton nonlinear solver are less than 1.0×10-5[31].In addition,the Cyclic Steady State(CSS)was considered to be reached when the bed temperature change at the same position is less than 0.1 K and instantaneous concentration variation in light and heavy product streams is less than 0.1%from one cycle to the next.

Fig.3.The schematic diagram of adsorption bed for mathematical model.

Table 2Mathematical models of mass,energy,momentum,adsorption equilibrium and LDF Eqs.[33,34]

Table 3Physical parameters for system

3.Results and Discussion

3.1.Model verification

In numerical simulation,the six-bed-PSA process would achieve cyclic steady state after 25 cycles,and the experiment results shows that the system would reach cyclic steady state after 23 cycles.Fig.5 shows the comparison of pressure profiles between experiment and simulation and these two results are nearly consistent.The pressure increases rapidly at AD1 step and maintains 600 kPa during AD2 step.After AD2 step,ED1/ED2/ED3 steps are carried out in succession and the pressure drops to 0.225 MPa.The pressure equalization steps enhance oxygen recovery significantly and recover part of mechanical energy.Subsequently,an oxygenenriched stream extracted from adsorption bed during CoD step was employed to purge other adsorption bed at the PUR step and the pressure drops to 0.15 MPa.During BD step,the pressure reduces to around atmosphere,while most of nitrogen adsorbed is desorbed and the adsorbents is partially regenerated simultaneously.Then,an oxygen-enriched gas produced at CoD step is introduced into adsorption bed for completely regeneration.After that,adsorption bed undergoes three consecutive ER steps and PR step to repressurize to adsorption pressure,which is ready for another new PSA cycle sequence.The purity and recovery of each component in feed and product stream are listed in Table 6 and Table 7,respectively.A series of experiments have been performed,andthe results are listed at Table S2(Supplementary Material),and the simulation results are of good coherence to experiment data.

Table 4The initial and boundary conditions of adsorption bed used for PSA process simulations

Table 5Extended Langmuir model fitting parameter

Fig.4.Adsorption isotherms on zeolite 5A (N2,O2 and Ar).

Fig.5.The pressure change in one bed within a cycle.

3.2.Concentration distribution

The gas and solid phase concentration in adsorption bed are difficult to be directly measured by experiment.However,these datacould be obtained by numerical calculation and used to explain and analyze experiment phenomena.Gas and solid phaseconcentration of N2at different bed position are shown in Fig.6 and Fig.7,respectively.Furthermore,gas and solid phase concentration of N2at the end of each step are demonstrated in Fig.8 and Fig.9,respectively.The adsorption front moves forward to bed top as the AD step proceeds and reaches the middle of adsorption bed at the end of AD step.The oxygen concentration in the upper part of adsorption bed is still high,as shown in Fig.9.Duringequalization depressurization step,adsorption front of N2moves forward in adsorption bed gradually and O2concentration in upper part of adsorption bed decreases rapidly.Meanwhile,the N2concentration in solid phase decreases as the pressure decreases,and the purity of oxygen in equalization depressurization stream is high as listed in Table 8,which means it should be fully utilized to keep high O2recovery in product stream.Hence,the process design of three equalization steps to improve oxygen recovery is reasonable.In purge step,the nitrogen concentration in both gas phase and solid phase decreases significantly.However,a slightly increases of nitrogen concentration could be easy observed in top of adsorption bed,as indicated in the point A in the Fig.7 and Fig.9.The reason is that a nitrogen-enriched stream originated from high pressure bed at the end of CoD step flows to low pressure bed at the end of PUR step.Therefore,the quantity of cocurrent stream for purge step must be moderate.

Table 6The purity of components in each stream

Table 7The recovery of components in each stream

Table 8Flow rate of the gas during steps of equalization depressurization and co-current down

Fig.6.N2 concentration in solid phase at different time.

Fig.7.N2 concentration in gas phase at different time.

Fig.8.Axial distribution of N2 on solid phase at the end of each step.

Fig.9.Axial distribution of N2 on gas phase at the end of each step.

3.3.Temperature distribution

A tridimensional diagram of temperature variation within one cycle is presented in Fig.10.The temperature fluctuation ranges from 281 K to 300 K.The maximum temperature is observed at the end of AD step because the adsorption process is an exothermic process.In contrast,the minimum temperature appears at the end of PUR step as the desorption process is an endothermic process.Actually,the temperature variation trend inside adsorption bed is similar to the nitrogen concentration variation trend in solid phase.Considering that the feed temperature is 288 K,the maximum temperature rise in the AD step is 12 K,and the maximum temperature reduction in the desorption step is 7 K,so the difference is small,which would not pose a threat on the structure of the adsorbent.

Fig.10.The temperature distribution with time and axial distance evolution at CSS.

3.4.Effects of operating parameters on process performances

A series of process simulations have been performed to investigate the effects of key operating parameters,including step time,ratio of bed height to diameter and product flow rate.Specially,when the parameters studied vary within the selected scope,all other operating parameters are kept constant.Fig.11 shows the impact of the step time on the oxygen purity and recovery in product stream (a),and the adsorption pressure (b) at CSS.The optimal step time is set to be in a range of 9.5-10 s.The oxygen purity reaches 93.2% and the recovery exceeds 57.6% and the adsorption pressure increases with time extension,as presented in Fig.11(b).If the adsorption time is too long,the heavy components will penetrate the bed and reduce the purity of O2,while the increasing of purge time will reduce the recovery.On the contrary,the short adsorption time will lead to insufficient utilization of the adsorbent,and more light components will remain in the bed,which will merge into the waste gas in the BD step and reduce the recovery.If the purge time is too short,the regeneration of the bed will be insufficient,which will adversely affect the purity of O2.It can be seen from Fig.5 that the adsorption step is accompanied by the pressure increase of the bed,so the longer the step,the higher the adsorption pressure.

Effects caused by ratio of bed height to diameter is demonstrated in Fig.12.Both purity and recovery of oxygen increase as the ratio increases from 1 to 10.Within the studied ratio,the purity and recovery of O2in product stream are higher than 93%and 57%,respectively.Compared with the impact on purity and recovery of oxygen,the ratio of bed height to diameter do a opposite influence on adsorption pressure.Under the same adsorption time and product flow rate,the increase in ratio of bed height to diameter will delay the penetration of heavy components,so higher purity of O2can be obtained.However,when the ratio is too low,heavy components will penetrate instantly,which will affect the overall process performance.According to the equation (3),the increase of bed height to diameter will increase the bed pressure drop,so the adsorption pressure will decrease.

Fig.11.Effect of step time on purity and recovery (a),and the adsorption pressure (b) at CSS.

Fig.12.Effect of ratio of bed’s height to diameter on purity and recovery (a),and the adsorption pressure (b) at CSS.

Fig.13.Effect of flow rate of product on purity and recovery (a),and the adsorption pressure (b) at CSS.

As shown in Fig.13,effects of product flow rate on process performances is more significant than that of step time and ratio of height to diameter.With product flow rate increasing,the oxygen purity in product stream decreases but the recovery of O2increases.The product flow rate can be approximated with the step time.Under the same step time,too large flow rate will cause reduce the purity of O2,but increase the yield.If the product flow rate is too low,the adsorbent is not fully utilized,leaving more light components in the bed and reducing the recovery rate.Similarly,according to the equation(3),the pressure drop is positively correlated with the product flow rate and the higher the flow rate,the greater the pressure drop.

4.Conclusions

In this work,a pilot-scale six-bed pressure swing adsorption(PSA) process has been developed for oxygen production.This device adopted a six-way rotating distribution valve,which not only simplified the valve structure,but saved space.Besides,it can obtain greater productivity than the conventional four-bed process.The mathematical model was established to describe the special valve and PSA system.The accuracy and reliability of model were verified by experiment.The results indicated that the purity of oxygen product was higher than 93% with a recovery above 57%.Process simulations have been carried out to reveal transient behaviors in adsorption bed,and it was also employed to determine effects of operating parameters on the purity and recovery.Simulation results demonstrated that the effect of product flow rate,compared to others operating parameters,was more significant.

Nomenclature

Cpgconstant pressure specific heat of the gas mixture，kJ·kg-1·K-1

Cpsspecific heat of the adsorbent，kJ·kg-1·K-1

Cpwspecific heat of bed wall,kJ·Kg-1·K-1

Cvvalve constant，kmol·bar-1·s-1

ctotal gas phase concentration，mol·m-3

cigas phase concentration of componenti，mol·m-3

Daxaxial dispersion coefficient，m2·s-1

Dbinternal diameter of bed,m

De，ieffective diffusion coefficient of componenti，m2·s-1

Dk，iKnudsen diffusion coefficient of componenti，m2·s-1

Dmmolecular diffusion coefficient，m2·s-1

Dvmolecular diffusion volume，m2·s-1

dPparticle porediameter，m

Fmolar flow of rates，mol ·s-1

ΔHiisosteric heat of adsorption of componenti，kJ ·mol-1

hambheat transfer coefficient between column wall and ambient,W ·m-2·K-1

hfheat transfer coefficient between gas and adsorbent,W ·m-2·K-1

hwheat transfer coefficient between gas and column wall,W ·m-2·K-1

IP1 extended Langmuir model fitting parameters

IP2 extended Langmuir model fitting parameters

IP3 extended Langmuir model fitting parameters

IP4 extended Langmuir model fitting parameters

kLDFMass transfer coefficient,s-1

kgeffective axial thermal conductivity,W ·m-1·K-1

kseffective axial thermal conductivity of bulk,W·m-1·K-1

kweffective axial thermal conductivity of column wall,W·m-1·K-1

Lbedlegend of adsorptionbed，m

Lbufferlegend of buffer tank，m

Mmolecular weight of the gas，kg ·mol-1

Ntotal number of component

Ppressure，Pa

qiadsorbed phase concentration of componenti,mol·kg-1

qi*equilibrium adsorbed phase concentration of componenti,mol·kg-1

Rideal gas constant，J ·mol-1·K-1

Rbedradius of adsorptionbed，m

Rbufferradius of buffer tank，m

rPparticle radius，m

Ttemperature，K

Tambtemperature of ambient,K

Tgtemperature of gas,K

Tstemperature of adsorbent bulk,K

Twtemperature of column wall,K

Tfeedtemperature of feed gas，K

ttime，s

vgsuperficial velocity，m ·s-1

wadsmass of adsorbent，kg

Wtwall thickness of bed,m

yimolar fraction of componenti

zaxial direction

εbbulk phase porosity

εpparticle phase porosity

εmvolumetric efficiency of compressor and vacuum pump

ρgdensity of the gas phase，kg ·m-3

ρsdensity of the adsorbent，kg ·m-3

ρwdensity of the column wall,kg·m-3

ηpwork efficiency of compressor and vacuum pump

μ gasviscosity，Pa ·s

τ tortuosity factor

ψ adsorbent shape factor

Subscripts

g gas phase

in gas enters the system from the outside

ispecies

out gas is emitted from the system to the outside

s adsorbent bulk

w column wall

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This work is supported by Major military logistics research projects (AWS13Z006) and National Key Research and Development program of China (2017YFC0806404).

Supplementary Material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.cjche.2021.03.027.