Tariq Shamim

Energy Conversion and Management 86 (2014) 1010–1022

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier .com/locate /enconman

CFD analysis of bubble hydrodynamics in a fuel reactorfor a hydrogen-fueled chemical looping combustion system

http://dx.doi.org/10.1016/j.enconman.2014.06.0270196-8904/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +971 (2) 810 9158; fax: +971 (2) 810 9901.E-mail address: [email protected] (T. Shamim).

Atal Bihari Harichandan, Tariq Shamim ⇑Institute Center for Energy (iEnergy), Department of Mechanical and Materials Engineering, Masdar Institute of Science and Technology, P.O. Box 54224, Abu Dhabi, UnitedArab Emirates

a r t i c l e i n f o a b s t r a c t

Article history:Received 4 September 2013Accepted 10 June 2014Available online 10 July 2014

Keywords:CFD simulationChemical looping combustionFuel reactorKinetic theory of granular flowEulerian method

This study investigates the temporal development of bubble hydrodynamics in the fuel reactor of ahydrogen-fueled chemical looping combustion (CLC) system by using a computational model. The modelalso investigates the molar fraction of products in gas and solid phases. The study assists in developing abetter understanding of the CLC process, which has many advantages such as being a potentially prom-ising candidate for an efficient carbon dioxide capture technology.

The study employs the kinetic theory of granular flow. The reactive fluid dynamic system of the fuelreactor is customized by incorporating the kinetics of an oxygen carrier reduction into a commercial com-putational fluid dynamics (CFD) code. An Eulerian multiphase treatment is used to describe the contin-uum two-fluid model for both gas and solid phases. CaSO4 and H2 are used as an oxygen carrier and a fuel,respectively. The computational results are validated with the experimental and numerical results avail-able in the open literature. The CFD simulations are found to capture the features of the bubble formation,rise and burst in unsteady and quasi-steady states very well. The results show a significant increase in theconversion rate with higher dense bed height, lower bed width, higher free board height and smaller oxy-gen carrier particles which upsurge an overall performance of the CLC plant.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The concern about the effect of greenhouse gases on the globalclimate changes is growing worldwide. Anthropogenic carbondioxide (CO2) emissions are the major contributor and a significantpart of it is due to the combustion of various carbonaceous fuels inpower generating plants. Accordingly, the development of noveltechnologies to control CO2 emissions with minimum energylosses has become a topic of strategic interest for researchers. Inaddition to developing renewable energy conversion technologies,the use of CO2 capture and storage (CCS) with the use of fossil fuelswill make a significant contribution to the reduction of CO2 emis-sions. Many CO2 capture methods such as amine scrubbing, oxy-fuel combustion, and decarbonization of fuels, are currently underinvestigation. Chemical looping combustion (CLC) introduced byRichter and Knoche [1] is an efficient technology in which CO2 isinherently separated during the combustion process.

A CLC system consists of two separated reactors (air and fuelreactors). A suitable oxygen carrier is oxidized (storing oxygen)in the air reactor and is transported to the fuel reactor using a

cyclone. In the fuel reactor, the metal oxide is reduced and thereleased oxygen reacts with the fuel. Carbon-dioxide and water(H2O) vapors are the two flue gases generated in the fuel reactor.Afterwards, H2O is condensed from the flue gases and an uncon-taminated stream of CO2 is captured and stored for later use.Fig. 1 shows a schematic of CLC process. The abbreviationsMe–O/Me represent the metal oxide in oxidized/reduced forms,respectively. In the recent past, this novel technology has beenstudied extensively [2–13] focusing on different distinct aspectssuch as technical feasibility, economic assessments, incorporationof the CLC system in power generation cycles, numerical simula-tion of the system with solid and gas fuels, and the experimentalimprovement of the CLC system.

In a CLC system, the distribution of solid–gas particles in thereactors (which are generally based on the fluidized bed reactordesign) is highly complex and can be better understood by usingcomputational fluid dynamics (CFD) models based on the fluiddynamics and reaction kinetic mechanisms. The CFD models canbe used to simulate the behavior of the reactants and the productsin the reactors during unsteady and quasi-steady states by usingthe conservation principles of mass, momentum, energy and spe-cies. The difficulties of capturing the flow physics in a relativelycomplex circulating fluidized bed (CFB) require significantly large

http://crossmark.crossref.org/dialog/?doi=10.1016/j.enconman.2014.06.027&domain=pdf

http://dx.doi.org/10.1016/j.enconman.2014.06.027

mailto:[email protected]

http://dx.doi.org/10.1016/j.enconman.2014.06.027

http://www.sciencedirect.com/science/journal/01968904

http://www.elsevier.com/locate/enconman

Nomenclature

CD drag functionDi,mix diffusion coefficient of the mixtureDturb turbulent diffusivityegg coefficient of restitution for particle collisionF factor that depends on CD and Reg~Fa external body force~Flift;a lift force~Fvm;a virtual mass forceg0,gg radial distribution functionG acceleration due to gravityI2S second invariant of the stress tensor~Ja

i diffusion flux of species iKgl multi-phase exchange coefficient_mab mass transfer from phase a to phase b_mab mass transfer source between phases

P pressure~Rab interaction force between phasesRa

i net rate of production of homogeneous species i bychemical reaction for phase a

Reg relative Reynolds numberSa source term in continuity equationSa

i rate of creationScturb turbulent Schmidt numbertg particulate relaxation timeu0g fluctuating velocity of the particlesvr,g terminal velocity correlation for the granular phase~va velocity of phase a~vab inter-phase velocity

Yai the local mass fraction and net rate of production of

homogeneous species i

Greek Symbolsa volume fractionbgd inter-phase momentum transfer for calculating drag be-

tween solid phase and gas phasecg dissipation of fluctuating energy due to inelastic colli-

sionHg granular temperatureka bulk viscosity of phase ala shear viscosity of phase alg granular shear viscositylg,col granular collisional shear viscositylg,fr granular frictional viscositylg,kinetic granular kinetic viscositylturb turbulent viscosityq density��sa ath phase stress–strain tensor/f internal friction angle

Subscriptsa,b phases (solid and gas)dg diameter of particles in solid phaseg granular phasegas gas phasei, j speciesl liquid phase

A.B. Harichandan, T. Shamim / Energy Conversion and Management 86 (2014) 1010–1022 1011

computational efforts for complete modeling. With more advancedcomputing facilities and better numerical methods, CFD is nowbecoming a popular means for the CFB modeling. Jung and Gamwo[14], Deng et al. [15], Kruggel-Emden et al. [16], Shuai et al. [17]and Mahalatkar et al. [8] have performed the CFD studies of fuelreactors using gaseous fuels. Numerical studies for fuel reactorsusing solid fuels were reported by Wang et al. [18], Mahalatkaret al. [19] and Garcia-Labiano et al. [20].

There have been a limited number of investigations on theunsteady aspects of bubble formation, rise and burst in the fuelreactor of a hydrogen-fueled CLC system. One relevant work wasperformed by Deng et al. [15], who numerically simulated the bub-bling fluidized bed reactor of a hydrogen-fueled CLC system tostudy the quasi-steady features of bubble dynamics. However,the temporal development of bubbles inside the reactor, which is

Fig. 1. Schematic view of chemical looping combustion process.

very important particularly at the beginning of reaction and thesubsequent stages before attaining the quasi-steady state, hasnot been investigated. Hence, the primary objective of the presentstudy is to investigate the temporal development of bubble hydro-dynamics of reactive flow in a fuel reactor of a hydrogen-fueledCLC system by employing a two-dimensional multiphase CFDmodel.

The present study is carried out by using calcium sulfide (CaS)as an oxygen carrier and hydrogen (H2) as a fuel. To keep the focuson the investigation of the bubble hydrodynamics, H2was selectedas the fuel gas due to its simpler reaction kinetics. The hydrogenfueled CLC process produces pure steam in the fuel reactor. Thissteam can be expanded to a very low pressure before being usedin the gas turbine cycle of a power plant for power production. Ithas an excellent performance without the formation of NOx [21].The understanding of the bubble hydrodynamics in a hydrogen-fueled reactor will also be useful for the analysis of the higherhydrocarbon-fueled fuel reactors.

Fig. 2 represents the schematic of a CLC system with two inter-connected reactors in which the reduction and the oxidation pro-cesses take place. Calcium sulfate (CaSO4) is reduced to CaS andit is then oxidized back to CaSO4. The present study does not con-sider the formation of SO2 and H2S. The CLC process described inthe present study has two chemical reactions as follows:

CaSO4 þ 4H2 ! CaSþ 4H2Oðreduction of oxygen carrier in the fuel reactorÞ ð1Þ

CaSþ 2O2 ! CaSO4

ðoxidation of oxygen carrier in the air reactorÞ ð2Þ

The reactions described in Eqs. (1) and (2) are both exothermic innature with low level energy releases at low-temperature region

Fig. 2. CLC with two interconnected fluidized bed reactors.

1012 A.B. Harichandan, T. Shamim / Energy Conversion and Management 86 (2014) 1010–1022

(�600–1200 K) for the first reaction and high-temperature region(�800–1700 K) for the second reaction, respectively. The presentstudy considers the simulations of the fuel reactor only (it is shownwith the dotted line in Fig. 2). In this study, the reaction kineticfor CaSO4 in Eq. (1), which is a first-order reaction with respectto the hydrogen partial pressure, has been calculated on the basisof a shrinking-core model [22] with the activation energy of151,000 J/mol and the reaction kinetic constant of 4300. In addi-tion to the investigation of bubble hydrodynamics, the paper alsopresents the effects of some important physical parameters on theperformance of a CLC system.

2. Mathematical model

The study employs the kinetic theory of granular flow with anEulerian approach. The continuity, momentum and energy equa-tions are solved for each phase whereas a single pressure is sharedby all the phases.

2.1. Hydrodynamic model

The general governing equations considered for the unsteadyand multiphase flow are given as follows:

2.1.1. Continuity equationsThe continuity equation for phase a is:

@

@tðaaqaÞ þ r � ðaaqa~maÞ ¼

Xn

b¼1

ð _mba � _mabÞ þ Sa ð3Þ

where aa, qa and~ma are the density and velocity of phase a, respec-tively. _mba represents the mass transfer from the bth to ath phase,and _mab symbolizes the mass transfer from phase a to phase b.The last term Sa is the source term. In a heterogeneous reaction,the exchange of mass, momentum and heat between the gas–solidphases are considered in the continuity equation.

2.1.2. Momentum equationsThe momentum equation for phase a is:

@

@tðaaqa~maÞ þ r � ðaaqa~ma~maÞ ¼ �aarpþr � ��sa þ aaqa~g

þXn

b¼1

~Rba þ _mba~mba � _mab~mab

� �

þ ~Fa þ~Flift;a þ~Fmm;a

� �ð4Þ

where ��sa is the ath phase stress–strain tensor.

��sa ¼ aala r~ma þr~mTa

� �þ aa ka �

23la

� �r �~ma

��I ð5Þ

Here, la and ka are the shear and bulk viscosities of phase a,~Fa is

an external body force, ~Flift;a is a lift force, ~Fmm;a is a virtual mass

force, ~Rba is an interactive force between phases, p is the pressureshared by all the phases, and~mba is the interphase velocity. If masstransfer takes place from phase b to phase a (i.e., _mba > 0), ~mba =~ma;whereas for mass transfer from phase a to phase b (i.e., _mba < 0),~mba =~mb. Similarly, if _mab > 0 then ~mab =~ma, if _mab < 0 then ~mab = ~mb .

The value of~Rba is influenced by friction, pressure, cohesion, and

other effects. Based on the physical conditions, ~Rba ¼ ~Rab and~Raa = 0. The velocity gradients in the gaseous (primary) phase flow

field in a multiphase flow generate lift forces (~Flift;a) that act onsolid particles. These forces are more substantial for bigger parti-cles. In the present model, the lift forces on solid particles are insig-nificant due to their smaller resisting surface area and incomparison to larger drag forces. So, these forces are not included

in the model. The virtual mass force (~Fvm;a) is considered when asolid phase (secondary, b) moves faster relative to the gas phase(primary, a). The solid particles are likely to experience a virtualmass force due to the inertia of gas phase caused by the fast mov-ing particles (bubbles or droplets). As the gas phase is denser thanthe solid phase, this force is not very important and is not incorpo-rated in the present model. Furthermore, the adhesive forcesbetween particles were not considered for the present simulations.

2.1.3. Species transport equationsThe local mass fraction (Ya

i ) of ith species in a multiphase flow ispredicted by the species conservation equation, which is describedas:

@

@tqaaaYa

i

� �þr � qaaa~maYa

i

� �¼ �r � aa~Ja

i þ aaRai þ aaSa

i

þXn

b¼1

_mbiaj � _majbi

� �þR ð6Þ

where Rai is the rate of formation of identical species i by chemical

reaction for phase a, _majbi is the mass transfer source between speciesi and j from phase a to b, and R is the heterogeneous reaction rate. Inaddition, aa is the volume fraction for phase a and Sa

i is the rate of for-

mation.~Jai is the diffusion flux of species i which is present due to dif-

ferential concentration and temperature of species. The massdiffusion due to differential concentration is modeled by using thedilute approximation (also called Fick’s Law) as defined in Eq. (7):

~Jai ¼ � qDi;mix þ

lturb

Scturb

� �rYa

i ð7Þ

Scturb is selected to be 0.7 for calculations.

2.1.4. Energy equationsThe energy equation for phase a is:

@

@tðaaqahaÞ þ r � ðaaqauahaÞ ¼ rðkarTaÞ þ Q ab þ Sabhab ð8Þ


where h, k and Q are enthalpy, thermal conductivity of mixture andheat exchange between gas phase and solid phase, respectively. Qab

is the heat transfer when phase b changes to phase a. The heatexchange between two phases is a function of the temperature dif-ference and is expressed as:

Q ab ¼ habðTa � TbÞ ð9Þ

where hab is the heat transfer coefficient. Also, Qab = �Qba.

2.1.5. Kinetic theory of granular flowA random granular motion of particles is likely to occur due to

the collisions of solid particles inside the reactor. The transportequation for the solid phase [23] is described by the followingequation:

@

@tðagqgHgÞ þr:ðagqgHgugÞ ¼ �

23

pg��Iþag

��sg

� �:r~ug þr:ðkgrHgÞ

� c�3bggasHg ð10Þ

Here, Hg is the granular temperature and is given by: Hg ¼ 13 ðu0gu0gÞ,

where u0g is the fluctuating velocity of the particles and can bederived from u0g ¼ Vg � ug , and pg is the solid pressure. bggas is theinterphase momentum transfer for calculating the drag betweentwo phases. The well-known Ergun equation [24] is considered foragas < 0.8 to describe the dense region in the reactor:

bggas ¼ 150ð1� agasÞrglgas

agasd2g

þ 1:75qgasag jugas � ug j

dgð11Þ

If agas > 0.8, the drag coefficient is calculated by the equation pro-posed by Wen and Yu [25]:

bggas ¼34

Cdjugas � ug j

dga�2:65

gas ð12Þ

where

Cd ¼24Re ð1þ 0:15Re0:687Þ; Re 6 10000:44; Re > 1000

(ð13Þ

and Re ¼jugas � ug jagasqgasdg

lgasð14Þ

2.2. Closure models

2.2.1. Interphase exchange coefficientThe constitutive closure models suitable for the above men-

tioned governing equations are described below:The solid–fluid interchange exchange coefficient in its general

form is expressed as:

Kgl ¼agqgf

tgð15Þ

f ¼ CDRegal

24m2r;g

ð16Þ

where f is a factor that depends on the drag function (CD) and rela-tive Reynolds number (Reg) [26].

andCD ¼ 0:63þ 4:8ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiReg=mr;g

p !2

Reg ¼qldg j~mg �~mlj

llð17Þ

mr;g ¼ 0:5 A� 0:06Reg þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið0:06RegÞ2 þ 0:12Regð2B� AÞ þ A2

q� �ð18Þ

where

A ¼ a4:14l ð19Þ

andB ¼0:8a1:28

l ; al 6 0:85a2:65

l ; al > 0:85

(ð20Þ

tg ¼qgd2

g

18llð21Þ

where tg is the particulate relaxation time.

2.2.2. Fluid–solids drag forceThe motion of solid particles through a viscous fluid inside the

fuel reactor often experiences a resistance due to interphase dragforces between the solid–liquid phases. The solid–gas interactionsare defined by the interphase momentum exchange and the dragcorrelation established on the settling of beds and the terminalvelocity of fluids as:

Fgasg ¼ Fggas ¼3lgasagagas

4mr;gd2p

0:63ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiReg=mr;g

qþ 4:8

� �2

ð22Þ

2.2.3. Solids pressureFor the cases of granular flows having smaller solid volume

fractions than the maximum possible value (compressible regime),a solid pressure is considered for the pressure gradient term in themomentum equation, which is composed of a kinetic term and aparticle collision term as:

pg ¼ agqgHg þ 2qgð1þ eggÞa2g g0;ggHg ð23Þ

where egg is the coefficient of restitution for particle collisions, g0;gg

is the radial distribution function, and Hg is the granular tempera-ture that is proportionate to the kinetic energy of the fluctuatingparticle motion.

2.2.4. Radial distribution functionThe prospect of collisions between granules in a dense solid

granular phase medium is described by the radial distributionfunction. This function describes the non-dimensional distancebetween granules (assumed to be spherical).

g0 ¼35

1� ag

ag;max

� �1=3" #�1

ð24Þ

2.2.5. Solid shear stressesThe momentum exchange due to particle translation and colli-

sion inside the reactor gives rise to the solid shear stress tensorthat has components of the shear and the bulk viscosities. For gran-ular particles with the maximum solid-volume-fraction, plastictransition of particles occurs and an additional component of fric-tional-viscosity is considered to account for it. Thus, the solid shearviscosity is expressed as:

lg ¼ lg;col þ lg;kinetic þ lg;fr ð25Þ

lg;col ¼45agqgdgg0;ggð1þ eggÞ

Hp

� �1=2

ag ð26Þ

lg;kinetic ¼agdgqg

ffiffiffiffiffiffiffiffiffiffiHgp

p6ð3� eggÞ

1þ 25ð1þ eggÞð3egg � 1Þagg0;gg

ð27Þ

lg;fr ¼pg sin /f

2ffiffiffiffiffiffiI2Sp ð28Þ


/ and I2D are the angle of internal friction and second invariant ofthe stress tensor respectively.

2.2.6. Granular temperatureThe diffusion coefficient for granular energy kg is given by:

kg ¼150qgdg

ffiffiffiffiffiffiffiffiffiffiHgp

p384ð1þ eÞg0

1þ 65agg0ð1þ eÞ

2

þ 2qgc2g dgg0ð1þ eÞ

�ffiffiffiffiffiffiffiHg

p

rð29Þ

cg ¼ 3ð1� e2Þa2gqgdgg0Hg

4dg

ffiffiffiffiffiffiffiHg

p

r !�rug

" #ð30Þ

where cg is the dissipation of fluctuating energy due to inelasticcollision.

2.3. Numerical considerations

A finite volume method based on phase-coupled SIMPLE (PC-SIMPLE) algorithm [27] has been used to solve the unsteady mul-tiphase flow problems. The commercial CFD software code FLUENTwas used for solving the coupled equations by adopting a secondorder upwind differencing scheme. The numerical solutions wereobtained by using a convergence criterion of 10�5 for each scaledresidual component and a time step of 5 � 10�4 s.

Fig. 3 shows the schematic and grids used for the fuel reactor.The width of the reactor was 0.25 m. The computational domainof the reactor was discretized by 2500 quadrilateral cells. Thisnumerical grid was selected based on the findings reported bymany researchers that a numerical cell size of 10 times larger thanthe particle-size can capture the bubble hydrodynamics in fluid-ized bed reactors accurately [28]. The grid independence test per-formed in the present work also has the similar observations likethat of Gelderbloom et al. [28]. The fuel reactor was filled withmetal oxide particles up to a height of 0.4 m. Only H2 fuel gaswas fed to the reactor at the bottom section. The dispersed k� eturbulence model was used. The heat transfer coefficient betweenthe gas phase and the solid phase used for the present simulationwere proposed by Gunn [24]. The restitution coefficient betweenthe solid particles was 0.9. The boundary conditions were fixedby specifying the inlet velocity and the outlet pressure values at

Fig. 3. Schematic and grid of the fuel reactor.

the reactor inlet and outlet respectively. In the CFB reactor, thesolid–fluid mixture behaves as a fluid medium due to introductionof pressurized fluid through the particulate medium. For simplic-ity, no-slip condition at wall was used for both the phases. How-ever, the case of solid particulate having some finite slip alongthe wall can also be simulated by employing different values ofspecularity coefficient in the numerical model. The convectiveterms are evaluated by using a second order QUICK scheme. Themodel parameters used for the base case in the present study aresimilar to those used by Deng et al. [15].

3. Results and discussions

A multiphase CFD model based on the kinetic theory of granularflow has been numerically simulated to describe the hydrodynam-ics and chemical reactions of dense gas–solid flows. The solid par-ticles are considered to be smooth, spherical, inelastic andundergoing binary collisions. The shrinking-core model [22] forthe grain geometry is considered for calculating the reaction kinet-ics in the fuel reactor. 100 wt.% of H2 is fed as the feed fuel gas intothe reactor through a distributor. The oxygen carrier present in thereactor’s static bed and the supplied fuel gas at the reactor inlet arehighly mixed by the momentum of upward moving gas bubbles.The reaction takes place between the gaseous reactant (H2) andthe desorbed oxygen (O2) of the oxygen carrier. In the process,H2O in the gaseous form is produced and oxygen carrier is reduced.The simulations are carried out for different dimensions of thereactor and for different particle sizes of the granules.

Fig. 4 shows the temporal profiles of gas phase mole fractions ofH2 (reactant) and H2O (product) along the centerline of the reactor(x = 0 cm) and (a) at a height of 30 cm from the inlet (y = 30 cm)(dense bed region) and at (b) the outlet. For validation of the sim-ulation results, the results reported by Deng et al. [15] with similarsimulation parameters are also shown in the figure. One of the dif-ferences in this work and Deng et al. [15] is the use of more accu-rate discretization scheme (second order) in the presentsimulations. Another difference is the use of pressure outletboundary condition at the reactor outlet in the present simulationsas opposed to the use of outflow boundary condition by Deng et al.[15]. The temporal profiles of reactant and product molar fractionsare found to be in good agreement with the results of Deng et al.[15]. The oscillations of reactant and product observed inFig. 4(a) are due to the movement of bubbles and high reactionrates in the dense bed region. The reactant (H2) and product(H2O) molar fractions oscillate around 0.68 and 0.32, respectively.In the initial phase (up to 1.0 s), the reactant mole fraction (H2)decreases rapidly from unity. Beyond 1.0 s, the reactant mole frac-tion does not decrease and it oscillates around 0.7. Similar varia-tion of the product mole fraction is observed, which increasesrapidly from zero to 0.32 at around 1.0 s. The result shows thatthe reaction reached quasi-steady state after 1.0 s. At the reactoroutlet, the reactant and product mole fraction profiles achievesteady values of 0.65 and 0.35 respectively after 3.8 s. The constantprofiles of H2 and H2O at the outlet are expected because ofunavailability of metal oxides in the upper portion of the reactor.Thus, the conversion rate of the fuel gas is 35% under the simula-tion condition.

To develop a better understanding of the hydrodynamic andreacting environment in the fuel reactor, Fig. 5 shows the develop-ment of solid volume fraction profiles in the fuel reactor from 0 to1.0 s. As the reaction starts, smaller bubbles grow larger by merg-ing with the adjacent smaller bubbles. Merging of bubbles is alsoobserved at the upper portion of the dense fluidized bed, justbelow the interface of static bed and the free board regions. Inthe dense bed region, the bubbles formed near the distributer tendto move upward and, as a result, two columns of vertically offset

Fig. 4. Mole fractions in gas phase along the centerline of the reactor (x = 0 cm) and (a) at a height of 30 cm from the inlet (y = 30 cm) (dense bed region) and at (b) the outlet.

1 For interpretation of color in Fig. 7, the reader is referred to the web version othis article.


bubbles are formed. The smaller bubbles are noticed to trail behindthe larger ones. Due to a low pressure zone created at the wake ofthe larger bubbles, the trailing smaller bubbles accelerate upwardand coalesce with the larger bubbles. Also, the upward moving lar-ger bubbles, having higher velocity than the smaller ones, coalescewith the neighboring smaller bubbles. In the process, the bubblesizes grow quickly in the narrow flow passage area of the reactorand slugs are formed. Thus, the solid particles are pushed upwardby the rising slugs. Later, the solid particles are pushed down intothe bed along the central core of the reactor as well as along thewalls of the reactor due to gravity and the differential density ofgas–solid particles. These phenomena are noticeable from 0.5 s to1.0 s. At 1.0 s, a core-annulus region is formed similar to thatobserved experimentally by Clift and Grace [29] for slug flow reac-tors. This study used the reactors of different heights and widthsfor different superficial velocities. The cross-section of the reactorused by Clift and Grace [29] was also rectangular (similar to thepresent simulation) where solid particles were pushed upwardby the rising slugs which subsequently moved back down alongthe walls of the reactor forming a core-annulus section.

Fig. 6 shows the development of solid volume fraction profilesbeyond 1.0 s. The bubbles and slugs have low solid volume frac-tions (nearly equal to zero). Due to the difference in gas velocitiesinside the bubbles and the slugs, the reaction rates are different inthe respective regions. They are significantly higher in the emul-sion region due to larger concentration of solid particles. The reac-tion rates are lower in the fast moving bubbles. This transientbehavior of bubble dynamics is observed within the first 1.0 s afterwhich the quasi-steady state condition is attained. The continuous

supply of the feed fuel gas through the distributer ensures the glo-bal mixing of gas–solid particles in the reactor with the continuousbubble formation, rise and burst. However, the core-annulus fea-ture as observed during the unsteady period does not prevail forlonger time. This salient feature of bubble dynamics is clearlynoticeable up to 1.6 s after which the core-annulus structure burstsinto intimate mixing of gas–solid particles and the quasi-steadystate condition is achieved.

Fig. 7 shows the quasi-steady state distribution of (a) solid vol-ume fraction, and the gas phase mole fractions of (b) reactants and(c) products at 10 s. A global mixing of the gas and particles isobserved. The qualitative representation of the bubble hydrody-namics is shown in Fig. 7(a) where the blue color1 represents thepure gas and the red color denotes the dense gas–solid mixture.The reactant (H2) profile, shown in Fig. 7(b), implies that the H2 molefraction decreases linearly from about 1.0 around the distributor(inlet) to 0.65 at the interface between the dense fluidized and thefree-board regions. A sudden increase in its magnitude to a constantvalue of 0.7 has been observed in the free-board region. As expected,the reverse characteristics are observed for the product (H2O) profileas shown in Fig. 7(c). Similar observations were also observed byDeng et al. [15] and the profiles presented in Fig. 7 are in qualitativeagreement with their results.

The study also investigated the effects of reactor geometricdimensions and particle size on the bubble hydrodynamics and

f

Fig. 5. Unsteady variation of solid volume fraction from time 0 to 1.0 s.


the reactor fuel conversion performance. The following sectionsdescribe the results of these investigations.

3.1. Effect of dense bed heights

Fig. 8 shows the variations of solid volume fraction profiles fordifferent dense bed heights at the quasi-steady state (t = 10.0 s).The static bed heights in the fluidized bed region are selected as10, 20, 30, 40, 50 and 60 cm with a constant total bed height(100 cm). The figure shows a qualitative representation of the bub-ble dynamics in the form of formation, rise and burst. In the densebed region, characterized by a roughly constant porosity, thephenomena of bubble hydrodynamics are prominent in the lowerone-third part whereas in the upper two-third part, the reactionefficiency decreases [30]. Thus, the fuel gas conversion rate in theupper zone of the dense bed is limited by the gas transfer betweenbubbles and emulsion. The growth of bubble size has beenobserved to decrease considerably with the increase of static bedheights thereby increasing the reaction rate in the reactor. How-ever, the bubble size shows an increasing trend for the reactorswith lower (<30 cm) and higher (>50 cm) static bed heights.

Fig. 9 depicts the temporal variation of mole fractions of H2 atthe outlet and the conversion rate of the feed fuel gas at 10.0 sfor different dense bed heights. It is observed from Fig. 9(a) thatthe H2 mole fractions at the outlet decrease with an increase inthe static bed height. For small dense bed heights, the H2 molefraction at the outlet remains almost constant till the bed heightof 40 cm. However, for higher bed heights such as 50 and 60 cm,the H2 mole fraction has been observed to vary with time due torelatively higher concentration of oxygen carrier. Also, the timespan over which the H2 molar fraction decreases at the outlet(before attaining a certain constant value) is bigger for reactorswith smaller dense bed heights. Fig. 9(b) implies the increase ofthe conversion rate with the static bed height which is alsoreflected from solid volume fraction contours.

3.2. Effect of bed widths

Fig. 10 shows the variation of solid volume fraction profiles fordifferent reactor bed widths at the quasi-steady state (t = 10.0 s).For all the cases, the static bed height and the total height of thereactor are kept fixed at 40 cm and 100 cm, respectively. The bed

Fig. 6. Unsteady development of solid volume fraction contour with quasi-steady reaction rate.

Fig. 7. (a) Solid volume fraction, (b) mole fraction of H2 and (c) mole fraction of H2O at the quasi-steady state conditions (t = 10 s).


widths (w) for different cases are selected as 10, 15, 20, 25 and30 cm. The flow Reynolds numbers in all the cases are kept similarto the base case. The global mixing of gas–solid particles isobserved to be more prominent for the reactors with smaller

widths. The velocities of gas and emulsion in these reactors are rel-atively higher than the reactors with the higher widths. Thus, thegas and the emulsion tend to grow over the entire span of the reac-tor. Alternate layers of gas and emulsion are noticed in the smaller

Fig. 8. Solid volume fraction contour with different dense bed heights (H) at the quasi-steady state conditions (t = 10 s).

Fig. 9. (a) Mole fraction of H2 and (b) conversion rate for different dense bed heights of the reactor at the quasi-steady state conditions (t = 10 s).


width reactors. However, for a bed width of 10 cm, the character-istics of volume fraction contours are somewhat different as nocore-annulus region was formed and many bubbles were observedwith most of the oxygen carrier particles accumulated near thewall. Moreover, the upper portions of the reactors with high widthsare predominantly covered by the gas particles. This causes the netreaction rate to be lower due to unavailability of oxygen carrierparticles.

Fig. 11 shows the temporal variations of H2 mole fractions at theoutlet and the conversion rate of the feed fuel gas at the quasi-steady state (t = 10.0 s) for different reactor bed widths. The resultsdepict that the H2 mole fraction at the outlet increases and conse-quently the fuel conversion rate decreases with an increase in thereactor bed width. The H2 molar fraction at the outlet and the fuelconversion rate become nearly constant for reactors with bedwidths P20 cm.

3.3. Effect of free board heights

Fig. 12 shows the variations of solid volume fraction contoursfor different heights of free board region at the quasi-steady state

(t = 10.0 s). For these simulations, the static bed height and thebed width of the reactor are kept fixed at 40 cm and 25 cm, respec-tively. The heights of the free board region for different cases areselected as 40, 60, 80, 100 and 120 cm. The results show that anincrease in the height of the free board region increases the mixingof gas–solid mixture and reduces the bubble size. This increasesthe H2 conversion rate in the reactor. The reactor free board heightalso affects the gas concentration profile in the fuel reactor mainlydue to differences in the fluid dynamics of the dense bed and thefree board regions. The metal oxide particle concentration in thefree board region decreases with an increase in the reactor height.However, this increases the reaction rate in this region due to abetter gas–solid contact than in the dense bed.

Fig. 13 shows the temporal variations of H2 mole fractions at theoutlet and the conversion rate of the feed fuel gas at the quasi-steady state (t = 10.0 s) for different reactor free board heights.The results show that the H2 mole fraction at the outlet decreaseswith an increase in the reactor free board region height. Asexpected, the H2 molar fraction decreases rapidly from unity tocertain constant values. These observations are also similar to theexperimental results reported by Kolbitsch et al. [31] and Abad

Fig. 10. Solid volume fraction contour with different bed widths: (a) w = 10 cm, (b) w = 15 cm, (c) w = 20 cm, (d) w = 25 cm, (e) w = 30 cm at the quasi-steady state conditions(t = 10 s).

Fig. 11. (a) Mole fraction of H2 and (b) conversion rate at for different bed widths of the reactor at the quasi-steady state conditions (t = 10 s).


et al. [30] for the chemical-looping combustion of methane. Theresults show that the reactors with lower free board heights expe-rience faster decrease of molar fraction of feed fuel gas at the startof the simulation. The conversion rate of H2 increases with anincrease in the reactor free board heights.

3.4. Effect of particle sizes

Fig. 14 shows the variation of solid volume fraction contour fordifferent particle sizes at the quasi-steady state (t = 8.0 s). For allthe cases, the static bed height, free board height and the widthof the reactor are kept fixed at 40 cm, 60 cm and 25 cm, respec-tively. The particle sizes for different cases are selected as 0.1,0.2, 0.3, 0.4 and 0.5 lm. The results show that the global mixingof gas–solid particles is significantly higher for the reactors withsmaller particle sizes. Furthermore, the gas and emulsion velocities

in the reactors with smaller particles are also relatively higher.Thus, the gas and emulsion tend to spread over a large part ofthe reactor. However, the reactors with bigger particles, in spiteof having more oxygen carriers in the dense bed region, experiencevery low reaction rates due to improper mixing of gas–solidparticles.

Fig. 15 shows the temporal variations of H2 mole fraction at theoutlet and the conversion rate of the feed fuel gas at the quasi-steady state (t = 8.0 s) for different particle sizes. The results showthat the H2 mole fraction at the outlet is higher for cases with lar-ger particle sizes. Hence, the conversion rates for these cases arelower as shown in Fig. 15(b). This is due to a relatively lower mix-ing of the gas–solid particles which gives rise to lower reactionrates. The results also indicate that the feed fuel gas conversionrate can further be increased by using nano-size oxygen carrierparticles.

Fig. 12. Solid volume fraction contour with constant bed width (w = 25 cm) and different free board heights: (a) h = 40 cm, (b) h = 60 cm, (c) h = 80 cm, (d) h = 100 cm, (e)h = 120 cm at the quasi-steady state conditions (t = 10 s).

Fig. 13. (a) Mole fraction of H2 and (b) conversion rate for different free board heights of the reactor at the quasi-steady state conditions (t = 10 s).

Fig. 14. Solid volume fraction contour with different particle sizes: (a) Dp = 0.1 lm, (b) Dp = 0.2 lm, (c) Dp = 0.3 lm, (d) Dp = 0.4 lm, (e) Dp = 0.5 lm at the quasi-steady stateconditions (t = 8 s).


Fig. 15. (a) Mole fraction of H2 and (b) conversion rate for different particle sizes in the reactor at the quasi-steady state conditions (t = 8 s).


4. Conclusions

A CFD model has been developed and employed for the simula-tions of the fuel reactor in a chemical looping combustion (CLC)system. The model included the reaction kinetics of fuel and metaloxide. The governing equations were solved by using a commercialCFD code. The numerical model was employed to investigate thetemporal development of the bubble hydrodynamics and theunsteady aspects of bubble formation, rise and burst in the fuelreactor of a hydrogen-fueled CLC system. The molar fractions ofthe product in gas and solid phases were also investigated by thenumerical model.

The computational results were validated qualitatively andquantitatively with the numerical results of Deng et al. [15] andthe experimental results of Clift and Grace [29]. The bubbledynamics and the flow patterns were in good agreement withthe literature results. A low conversion rate of the feed fuel gaswas obtained which is due to combined effect of low bed temper-ature, bigger particle size, smaller dense bed height and larger bedwidth of the reactor. All these parameters led to fast and large bub-bles in the reactor which reduced the reaction rate and hence theconversion rate of the feed fuel gas. The low fuel conversion canbe improved by (i) the use of an oxygen carrier with small particlesizes and high reactor temperature; (ii) the use of a different oxy-gen carrier material, for example Ni; and (iii) the recirculation ofthe unutilized hydrogen to the fuel reactor.

The paper also presents the results of a parametric study basedon the reactor dimensions and particle sizes. The results show thata significant increase in the fuel conversion rate can be obtainedwith a higher dense bed height, a lower bed width, a higher freeboard height and smaller oxygen carrier particles. However, addi-tional cost must be considered for larger dense bed and higher freeboard heights. The smaller oxygen carrier particles (preferably ofnano-size) are a preferred option for enhancing the overall perfor-mance of CLC plant as they do not generate large bubbles and aproper mixing of gas–solid particles can very well be established.

The direct numerical simulation of the present multiphase reac-tive model will be considered for the future work in order to getbetter bubble hydrodynamics.

Acknowledgements

The work is financially supported by the MIT/Masdar InstituteCollaborative Research funds (grant # 10MAMA1).

References

[1] Richter HJ, Knoche KF. Reversibility of combustion processes. ACS SymposiumSeries 1983;235:71–85.

[2] Abad A, Mattison T, Lyngfelt A, Ryden M. Chemical looping combustion in a300 W continuously operating reactor system using a manganese basedoxygen carrier. Fuel 2006;85:1174–85.

[3] Johansson E, Mattisson T, Lyngfelt A, Thunman H. A 300 W laboratory reactorsystem for chemical-looping combustion with particle circulation. Fuel2006;85:1428–38.

[4] Mattisson T, Garcia-Labaiano F, Kronberger B, Lyngfelt A, Adanez J, Hofbauer H.Chemical-looping combustion using syngas as fuel. Int J Greenhouse GasControl 2007;1(2):158–69.

[5] Song Q, Xiao R, Deng Z, Zhang H, Shen L, Xiao J, et al. Chemical-loopingcombustion of methane with CaSO4 oxygen carrier in a fixed bed reactor.Energy Convers Manage 2008;49(11):3178–87.

[6] Forero CR, Gayan P, de Diego LF, Abad A, Garcia-Labiano F, Adanez J. Syngascombustion in a 500 Wth chemical-looping combustion system using animpregnated Cu-based oxygen carrier. Fuel Process Technol 2009;90(12):1471–9.

[7] Dennis JS, Scott SA. In situ gasification of a lignite coal and CO2 separationusing chemical looping with a Cu-based oxygen carrier. Fuel 2010;7(89):1623–40.

[8] Mahalatkar K, Kuhlman J, Huckaby ED, O’Brien T. Computational fluid dynamicsimulation of chemical looping fuel reactors utilizing gaseous fuels. Chem EngSci 2011;66:469–79.

[9] Hassan B, Shamim T, Ghoniem AF. A parametric study of multi-stage chemicallooping combustion for CO2 capture power plant. In: Proceedings of the ASME2012 Summer Heat Transfer Conference, Puerto Rico; 2012. Paper ASME2012-58597.

[10] Moldenhauer P, Ryden M, Mattisson T, Lyngfelt A. Chemical-loopingcombustion and chemical-looping with oxygen uncoupling of kerosene withMn- and Cu- based oxygen carriers in a circulating fluidized bed 300 Wlaboratory reactor. Fuel Processing Technol. 2012;104:378–89.

[11] Wang B, Zhao H, Zheng Y, Liu Z, Yan R, Zheng C. Chemical looping combustionof a Chinese anthracite with Fe2O3-based and CuO-based oxygen carriers. FuelProcessing Technol 2012;96:104–15.

[12] Abad A, Gayan P, de Diego LF, Garcia-Labiano F, Adanez J. Fuel reactormodelling in chemical-looping combustion of coal: 1. model formulation.Chem Eng Sci 2013;87:277–93.

[13] Peltola P, Ritvanen J, Tynjälä T, Hyppänen T. Model-based evaluation of achemical looping combustion plant for energy generation at a pre-commercialscale of 100MWth. Energy Convers Manage 2013;76:323–31.

[14] Jung J, Gamwo IK. Multiphase CFD based models for chemical loopingcombustion process: fuel reactor modeling. Powder Technol 2008;183:401–9.

[15] Deng Z, Xiao R, Jin B, Song Q. Numerical simulation of chemical loopingcombustion process with CaSO4 oxygen carrier. Int J Greenhouse Gas Control2009;3:368–75.

[16] Kruggel-Emden H, Rickelt S, Stepanek F, Munjiza A. Development and testingof an interconnected multiphase CFD-model for chemical looping combustion.Chem Eng Sci 2010;65:4732–45.

[17] Shuai W, Guodong L, Huilin L, Juhui C, Yurong H, Jiaxing W. Fluid dynamicsimulation in a chemical looping process with two interconnected fluidizedbeds. Fuel Processing Technol 2011;92:385–93.

[18] Wang X, Jin B, Zhong W, Zhang Y, So Min. Three-dimensional simulation of acoal gas fueled chemical looping combustion process. Int J Greenhouse GasControl 2011;5:1498–506.

http://refhub.elsevier.com/S0196-8904(14)00552-4/h0160






























































[19] Mahalatkar K, Kuhlman J, Huckaby ED, O’Brien T. CFD simulation of achemical-looping fuel reactor utilizing solid fuel. Chem Eng Sci 2011;66:3617–27.

[20] Garcia-Labiano F, de Diego LF, Gayan P, Abad A, Adanez J. Fuel reactormodelling in chemical-looping combustion of coal: 2. Simulation Optim ChemEng Sci 2013;87:173–82.

[21] Jin H, Ishida M. A novel gas turbine cycle with hydrogen-fueled chemicallooping combustion. Int J Hydrogen Energy 2000;25:1209–15.

[22] Zafar Q, Abad A, Mattisson T, Gevert B. Reaction kinetics of freeze granulatedNiO/MgAl2O4 oxygen carrier particles for chemical-looping combustion.Energy Fuels 2007;21:610–8.

[23] Patil DJ, Annaland MS, Kuipers JAM. Critical comparison ofhydrodynamic models for gas-solid fluidized beds-Part I: bubblinggas-solid fluidized beds operated with a jet. Chem Eng Sci 2005;60:57–72.

[24] Gunn DJ. Transfer of heat or mass to particles in fixed and fluidized beds. Int JHeat Mass Transfer 1978;21:467–76.

[25] Wen CY, Yu HY. Mechanics of fluidization. Chem Eng Prog Symp Ser1966;62:100.

[26] Syamlal M, O’Brien TJ. Computer simulation of bubbles in a fluidized bed.AlChE Symp Ser 1989;85:22–31.

[27] Vasquez SA, Ivanov VA. A phase coupled method for solving multiphaseproblems on unstructured meshes. In: Proceedings of ASME FEDSM’00: ASME2000 Fluids Engineering Division Summer Meeting, Boston; 2000

[28] Gelderbloom SJ, Gidaspow D, Lyczkowski RW. CFD simulations of bubbling/collapsing fluidized beds for three geldart groups. AlChE J. 2003;49:844–58.

[29] Clift R, Grace JR. Continuous Bubbling and Slugging. London: Academic Press;1985.

[30] Abad A, Adanez J, Garcia-Labiano F, de Diego LF, Plata A, Gayan P. Modeling ofthe chemical-looping combustion of methane using a Cu-based oxygen-carrier. Combust Flame 2010;157:602–15.

[31] Kolbitsch P, Proll T, Hofbauer H. Modeling of a 120 kW chemical loopingcombustion reactor system using a Ni-based oxygen carrier. Chem Eng Sci2009;64:99–108.



































Documents

Tariq Shamim