international journal of hydrogen energy 34 (2009) 9686–9695

Available at www.sciencedirect.com

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

CFD simulation of an expanded granular sludge

bed (EGSB) reactor for biohydrogen production

Xu Wang, Jie Ding*, Nan-Qi Ren, Bing-Feng Liu, Wan-Qian Guo

State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology, 202 Haihe Road, Nangang District,
Harbin, Heilongjiang 150090, China

article info abstract

Article history: Understanding how a bioreactor functions is a necessary precursor for successful reactor
Received 5 August 2009 design and operation. This paper describes a two-dimensional computational fluid
Received in revised form dynamics simulation of three-phase gas–liquid–solid flow in an expanded granular sludge
22 September 2009 bed (EGSB) reactor used for biohydrogen production. An Eulerian–Eulerian model was
Accepted 9 October 2009 formulated to simulate reaction zone hydrodynamics in an EGSB reactor with various
Available online 31 October 2009 hydraulic retention times (HRT). The three-phase system displayed a very heterogeneous
flow pattern especially at long HRTs. The core-annulus structure developed may lead to
Keywords: back-mixing and internal circulation behavior, which in turn gives poor velocity distribu-
Hydrogen production tion. The force balance between the solid and gas phases is a particular illustration of the
Expanded Granular Sludge Bed (EGSB) importance of the interphase rules in determining the efficiency of biohydrogen produc-
Hydraulic Retention Time (HRT) tion. The nature of gas bubble formation influences velocity distribution and hence sludge
Computational Fluid Dynamics (CFD) particle movement. The model demonstrates a qualitative relationship between hydro-
Hydrodynamics dynamics and biohydrogen production, implying that controlling hydraulic retention time
Eulerian–Eulerian model is a critical factor in biohydrogen-production.
ª 2009 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved.

1. Introduction [7–13]; however, the physical characteristics of biohydrogen

production reactors are not so well described.
Fossil fuels are a finite resource that are being used at an Understanding how a bioreactor functions is a necessary
increasing rate, and there are well known environmental precursor for successful reactor design and operation. For
problems associated with fossil fuel combustion [1]. Hydrogen a chosen bioprocess, the bioreactor should provide optimum
gas is noted as a valuable and clean energy source, and is conditions of shear stress, mass transfer, mixing, control of pH,
widely used in numerous industries [2,3]. Biological hydrogen temperature and substrate concentrations [14]. Various types
production through fermentation of organic waste has of biohydrogen production reactors have been designed and
attracted considerable attention as an effective production developed [15–20], but most of these reactors were designed by
route because of its high hydrogen production rate coupled means of semi-empirical correlations, and quantitative
with the possibility for reducing environmental organic studies on the fluid dynamics of such reactors are scarcely
pollution [4–6]. There have been many publications on both reported. Velocity fields, turbulent intensity and volume frac-
biohydrogen-producing mechanisms and microorganisms tion of multi-phases in the reactor, biomass activity, intrinsic

* Corresponding author. Tel./fax: þ86 0 451 86282193.

E-mail address: dingjie123@hit.edu.cn (J. Ding).
0360-3199/$ – see front matter ª 2009 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2009.10.027
 international journal of hydrogen energy 34 (2009) 9686–9695 9687

Notation mt turbulent viscosity, Pa s

r density, kg m3
CD drag coefficient, dimensionless
s standard deviation, mg L1
d diameter, m
sk turbulent Prandtl numbers for k, dimensionless
k turbulent kinetic energy, m2 s2
s3 turbulent Prandtl numbers for 3, dimensionless
p model evaluated parameters or pressure,
dimensionless or Pa Subscripts
Re Reynolds number, dimensionless ef effective value
u superficial velocity, m h1 L liquid
l volume fraction, dimensionless G gas
3 dissipation rate, m s3 S solid
m viscosity accounting for turbulence, Pa s k physical quantity related to phase

kinetics and size distributions, compressibility and activated sludge, based on a MLVSS of 8.49 g/L. Detailed information
sludge sedimentation, all of which influence microbial ecology, about the start-up and steady-state performance of this
have been neglected [21–25]. H2-producing EGSB reactor are described elsewhere [18].
When analyzing bioreactor fluid dynamics, the use of
single experimental techniques is highly time-consuming and
2.2. Computational fluid dynamics model
restricted by the limitations of even high-end measuring
equipment. Additionally, such measurements often cannot
2.2.1. Eulerian–Eulerian model
satisfy current research requirements understanding fluid
In this paper, a two-dimensional Eulerian–Eulerian three-
dynamics at the micro-level. Computational fluid dynamics
phase fluid model has been employed to describe the flow
(CFD) is becoming more widely available to analyze the
behavior of each phase, so the H2 gas, wastewater and sludge
characteristics of a bioreactor [26–30]. For example, Milewska
granules are all treated as different continua, with wastewater
and Molga [31] applied CFD to simulate accidents in industrial
as the primary phase, and the gas & sludge granules as the
batch stirred tank reactors. In the study, they found that even
secondary phase. This model was chosen because of the large
for a reactor operating at potentially safe conditions, a failure
number of gas bubbles and granular particulates [33–35]. Each
of the stirring system can lead to serious thermal runaway.
phase is presumed to be incompressible in this study. It is
Murthy et al. [32] used CFD simulation to predict the critical
possible to make this assumption about the gas phase, as this
impeller speed for solid suspension in a gas–liquid–solid stir-
phase is present as very small bubbles which, as a result of
red reactor.
their size do exhibit incompressible behavior. The wastewater
In the light of the above, an attempt was made to use CFD
was regarded as pure water, with a density of 1050 kg/m3 at
simulation to analyze hydrodynamics information in an
37  C. The sludge granules took up about 35% of the volume in
expanded granular sludge bed (EGSB) reactor used for bio-
the bed region and were considered to be 1 mm-diameter
hydrogen production. With the help of this model, the
spherical solid granules with a density of 1460 kg/m3. The
hydrodynamics of the EGSB reaction zone at various condi-
hydrogen gas was assumed to have a density of 1.225 kg/m3.
tions of HRT were simulated. The flow fields, i.e., velocity
The gas phase volume fraction was related to gas production
distribution and volume fraction, were simulated for different
and the gas bubbles were assumed to have a diameter of
HRTs and these results were used to improve the reactor
0.1 mm.
design and to optimize reaction conditions.

2.2.2. Governing equations

The mass and momentum conservation equations are solved
2. Materials and methods in a computational two-dimensional mesh (Fig. 2). The pres-
sure field was assumed to be shared by the three phases, in
2.1. Description of the laboratory-scale H2-producing proportion to their respective volume fractions. The motion of
EGSB each phase is governed by respective mass and momentum
conservation equations.
A sequence of H2-production experiments were carried out in The continuity equation for each phase is
a Plexiglas EGSB reactor with an internal diameter of 6 cm and
a height of 120 cm (Fig. 1). The reactor has a working volume of vðrk lk Þ
3.35 L and was operated at 35  1  C. þ Vðrk lk uk Þ ¼ 0 (1)
vt
The EGSB reactor was seeded with a mixture of activated
sludge taken from bed mud from a domestic wastewater where rk is the density of phase k, lk is the volume fraction of
discharge channel and sludge from an acidogenic reactor phase k, and uk is the velocity vector of phase k.
working as part of the wastewater treatment facilities in If it is assumed that each phase is incompressible, equation
a local pharmaceutical factory (Harbin, China). The final ratio (1) can be simplified to:
of mixed liquor volatile suspended solids (MLVSS) to mixed
liquor suspended solids (MLSS) was 0.68 in the inoculating Vðrk lk uk Þ ¼ 0 (2)
9688 international journal of hydrogen energy 34 (2009) 9686–9695

Fig. 1 – Schematic diagram of the EGSB reactor.

The momentum balance equation for each phase is 3 CD;LG

MD;LG¼ r lG juG  uL jðuG  uL Þ (7)
4 dG L

vðrL lL uL Þ    T 
þ VðrL lL uL uL Þ ¼ lL Vp þ V lL mef ;L VuL þ VuL 3 CD;LG
vt MD;LG¼ r lG juG  uL jðuG  uL Þ (8)
4 dG L
þ rL lL g  MI;LG (3)
where CD is the drag coefficient and d is the diameter of a gas
bubble (dG) or a solid particle (dS).
vðrS lS uS Þ    T  The drag coefficient CD, LG exerted by the gas phase on the
þ VðrS lS uS uS Þ ¼ lS Vp þ V lS mef ;S VuS þ VuS
vt liquid phase is obtained by the Schiller and Naumann drag
þ rS lS g  MI;LS (4) model [36], as follows:
(
24ð1þ0:15ð1lG ÞRe0:687 Þ
CD;LG ¼ ð1lG ÞRe
ð1  lG Þ2:65 ð1  lG ÞRe  1000 (9)
vðrG lG uG Þ    T  ð1  lG ÞRe > 1000
þ VðrG lG uG uG Þ ¼ lG Vp þ V lG mef ;G VuG þ VuG 0:44
vt
þ rG lG g  MI;LG where Re is the relative Reynolds number, which is obtained
(5) from:

where p is the pressure, mef is the effective viscosity, g is the rL dG juG  uL j

Re ¼ (10)
gravitational acceleration, and MI is the interphase transfer mL
force.
The drag coefficient exerted by the solid phase on the liquid
The volume fractions satisfy the compatibility conditions
phase, CD, LS, is calculated from the Wen and Yu drag model
X
n [36], which is as follows:
lk ¼ lL þ lS þ lG ¼ 1: (6)
24 h i
k¼1

CD;LS ¼ 1 þ 0:15ðlS ReÞ0:687 l0:265

S (11)
lS Re
2.2.3. Interphase momentum transfer
In this paper, the drag forces exerted by the solid phase and The relative Reynolds number for the liquid phase and the
gas phase on the liquid phase are calculated as: solid phase is estimated as follows:
 international journal of hydrogen energy 34 (2009) 9686–9695 9689

ML;LG ¼ CL rL lG ðuG  uL Þ  ðV  uL Þ (13)

ML;LS ¼ CL rL lS ðuS  uL Þ  ðV  uL Þ (14)

where CL is the lift coefficient and has a value of 0.5.

2.2.4. Turbulence closure

We assume that a standard k–3 model for single phase flows
(with extra terms that include interphase turbulent
momentum transfer [37]) can take into account the effects of
turbulence. The modeling of multiphase turbulent flows is
much more complex and computationally expensive for three
phase flows mainly because of the influence of the secondary
phases on the turbulence of the primary phase [32]. Therefore,
we have assumed for our three-phase CFD turbulence model,
that turbulence in a multiphase fluidized bed is restricted to
the primary phase.
The turbulence viscosity of the continuous phase is
obtained by the k-3 turbulence model:
 
kL
mt;L ¼ Cm rL (15)
3L

The equations of change for the turbulent kinetic energy (k)

and the energy dissipation rate for the primary phase are
Fig. 2 – Two-dimensional computational domain. given by

    Y
DlL rL kL mt;L  
¼ V lL m þ VkL þ lL rL pkL  3L þ lL rL kL (16)
Dt skL
rL dS juS  uL j
Re ¼ (12)
mL
    Y
The lift force acting perpendicular to the direction of the DlL rL 3L mt;L  
¼ V lL m þ V3L þlL rL C31 pkL C32 3L þlL rL 3L (17)
relative motion of the solid and gas phase is given by: Dt s3L

Fig. 3 – Experimental and simulated values of the water velocity in the reaction zone at different HRTs: A. HRT 1 h; B. HRT
2 h; C. HRT 4 h; D.HRT 6 h.
9690 international journal of hydrogen energy 34 (2009) 9686–9695

Fig. 4 – Water-velocity vectors of H2-production reaction zone: A. HRT 1 h; B. HRT 1.5 h; C. HRT 2 h; D.HRT 3 h; E. HRT 4 h;
F. HRT 6 h.

Q
Here, kL represents the influence of the secondary phase on Fluent 6.3 CFD Flow Modeling Software package to solve the
the primary phase and the predictions for turbulence quan- continuity and momentum equations.
Q
tities for the dispersed, and 3L represents the predictions for The simulation results vary little with grid density so
turbulence quantities for the secondary phases, which are truncation errors in the numerical simulation can be
both obtained using the Tchen theory of the dispersion of neglected. An analysis independent of grid was performed
discrete particles by homogeneous turbulence [35]. Standard to eliminate errors in simulation accuracy, numerical
values were used for the turbulence parameters: C31 ¼ 1.44, stability, convergence and computational step related to
C32 ¼ 1.92, Cm ¼ 0.09, sk ¼ 1.0, s3 ¼ 1.3. grid coarseness. The grid independent analysis was done
with a maximum cell density of 12382, through the selected
2.2.5. Numerical solution grid number of 5544 cells, to a minimum of 968 cells. When
In this paper, we have analyzed the reaction zone of our EGSB the optimum cell number was used, the difference in pres-
reactor. The simulated reaction zone was 120 cm tall and 6 cm sure drop was below 3%. Therefore, a two-dimensional
diameter. The meshes were created in the Ansys Fluent computational domain of the reaction zone of the EGSB
GAMBIT preprocessor program and exported into the Ansys reactor was devised with 5544 cells, 11314 faces, and 5771

Fig. 5 – Local enlargement – water-velocity vectors: M. enlargement E of Fig. 4; N. enlargement F of Fig. 4.

 international journal of hydrogen energy 34 (2009) 9686–9695 9691

Fig. 6 – Contours of water velocity magnitude of H2-produciton reaction zone: A. HRT 1 h; B. HRT 1.5 h; C. HRT 2 h; D. HRT 3 h;
E. HRT 4 h; F. HRT 6 h.

nodes (see Fig. 1). The initial sludge bed was packed with 2.3. Tracer experiment
granular solids with a volume fraction of 0.55. The initial
hydrogen gas phase was assumed to be accumulated into Velocity patterns at four different HRTs were obtained from
the sludge blanket, and it was further assumed that the gas tracer experiments using Particle Image Velocity measure-
would be released when force balance between gas phase ments. For the each test, a pulse micro glass beads was
and solid phase broke up. The reactor wastewater inlet was injected as a tracer into the reactor input stream. A planar
modeled with a velocity-inlet boundary condition. The cross section of the flow was illuminated with a sheet light
outlet was set as a pressure outlet boundary condition. All source, and an image was formed of the tracer particles. The
other solid surfaces were defined by wall boundary condi- measurement of velocity was based on the displacement of
tions with free slip for the biogas and no slip for the sludge the tracer particles during a given time interval.
or wastewater. In this study, the simulation was operated in
steady state conditions. The convergent solution was
defined as the solution for which the normalized residual for 3. Results and discussion
all variables was less than 1  103 and the calculated
outflow rate had reached a constant value. Convergence was 3.1. Model validation and error analysis
typically reached after 350 iterations. All the simulations
were carried on a computer with a 32 bit processor (Intel As far as the CFD model validation is concerned, Fig. 3 pres-
Core 2 Duo CPU T9300) with 2GB of random access ents a comparison between the experimentally measured and
memory, run at a clock speed of 2.50 GHz. the simulated values of the water velocity in the reaction zone

Fig. 7 – Contours of sludge volume fraction: A. HRT 1 h; B. HRT 1.5 h; C. HRT 2 h; D. HRT 3 h; E. HRT 4 h; F. HRT 6 h.
9692 international journal of hydrogen energy 34 (2009) 9686–9695

Fig. 8 – Contours of H2-gas volume fraction: A. liquid up-flow velocity of 6 m/h; B. liquid up-flow velocity of 4 m/h; C. liquid
up-flow velocity of 3 m/h; D. liquid up-flow velocity of 2 m/h; E. liquid up-flow velocity of 1.5 m/h; F. liquid up-flow velocity
of 1 m/h.

at different HRTs showing good agreement between measured 3.2. Hydrodynamics analysis
and predicted values. The relative error between measured
and simulated data was within 10% indicating that the model In a biohydrogen production reactor, liquid flow pattern,
provides a good overall description of reaction zone behavior. sludge activity, and mass balance analysis are interdepen-
The errors incurred in calculation derived from several dent. Highly reactive sludge and long hydraulic retention
sources, such as geometry meshing, numerical treatment of times (HRT) are required to obtain a good mass balance.
initial and boundary conditions, and incomplete iteration. A good mass balance results in more organics being trans-
With regard to geometry meshing, the grid selected was ferred to biogas, and a good flow regime is required for effi-
considered to be grid independent, thus related influences cient contact between microbes and wastewater, which in
such as truncation errors can be ignored. Adopting an iterative turn supplies a good growing environment for the culture. To
approach, discrete equations cannot satisfy absolute conver- obtain hydrodynamic information from the reaction zone of
gence, and the iterative process will only stop when a certain the EGSB reactor, six steady state simulations, at HRTs of 1,
condition occurs, for example, the normalized residual for all 1.5, 2, 3, 4, 6 hours (assuming uniform recycle ratio, and HRT
variables. This is why we see incomplete iterations. However, transformed to inlet up-flow velocity) were conducted.
it should be noted that in the study, the inaccuracy due to Figs. 4–6 present the water velocity in the reaction zone of
incomplete iteration was less than 1  103. The definition of an H2-production EGSB reactor with various HRTs. When the
initial values and boundary treatments are the most signifi- influent runs up through the bottom of the sludge bed at
cant (and complex) steps in the setup of the numerical a fixed up-flow speed, and bed expansion is observed due to
calculation. To obtain a high accuracy simulation, the virtual the low-density of the sludge granules. Due to variation with
conditions should be suitably simplified in addition to time of the path tortuosity and consequent length of passage
selecting the right parameters. Errors incurred at this stage are through the particle interstices of the sludge blanket, and the
difficult to analyze and quantify, and require further exami- random geometry of the granules and impurities, the flow
nation, to be described elsewhere. regime is not homogeneous in the direction of flow. In addi-
tion, it is assumed that gas bubbles produced by biodegrada-
tion or biofermentation, such as methane, carbon dioxide and
hydrogen, will flow upwards with the stream swirl. This
movement will make the bulk up-flow velocity greater than
Table 1 – Operation strategy of the EGSB reactor. the up-flow velocity of liquid alone, and the cross-sectional
Operating time (d) OLR (kg COD/m3 d) HRT (h) COD (mg/L) velocity will be non-homogeneous.
When the HRT is long, and liquid up-flow velocity is low,
1–6 8 6 2000
7–13 12 4 2000
lateral pulsating movement will take place as the fluid
14–20 24 4 4000 elements flowing through sludge blanket take the paths of
21–27 36 4 6000 least resistance around individual sludge granules. As can
28–35 48 4 8000 been seen (Fig. 5), internal circulation behavior occurs, and the
36–42 96 2 8000 wastewater is not distributed very evenly. The core-annulus
43–51 192 1 8000
structure occurs because the water velocities in the core
52–59 96 2 8000
region are much lower than those in the annulus region,
60–150 120 2 10,000
whereas because water velocities near the wall are increasing
 international journal of hydrogen energy 34 (2009) 9686–9695 9693

Fig. 9 – Hydrogen production rate, hydrogen yield and OLR variation in the EGSB reactor.

and upward, this may lead to back-mixing and internal heights were estimated as ranging between 0.41 and 0.44 m,
circulation behavior. Similar results have been obtained by indicating that the bed expansion was about 8%w10%,
other researchers [38–40]. When the liquid flows upward with reasonably agreeing with experimental results [18]. The
a moderate velocity to obtain an appropriate HRT (see simulation indicated that the gas–liquid–solid system exhibits
Fig. 4B–D), axial pulsation and disturbances between granules a more heterogeneous structure, with particle clusters form-
of sludge are small, but compared with the velocity of lower ing and dissolving dynamically. In the numerical simulation,
fluid levels, the turbulent motion and velocity distribution are clusters could be seen to fall along the wall, stacking together
greater. When the HRT is short (see Fig. 4A), an increasing into agglomerated clusters which ultimately become large
distance between individual sludge granules results due to enough to escape from the wall. Particles are dynamically
the high liquid up-flow velocity, which in turn decreases the squeezed out of these clusters and pushed upward by the up-
lateral pulsating of the liquid. Additionally because of the flowing biogas, and then these particles are further aggregated
lifting action of the biogas, the flow pattern will be in a highly into strands in an upper section of the bed. This cluster
turbulent state. formation, which may involve gas–sludge, sludge–sludge, and
Another important reactor hydrodynamic characteristic, sludge–wall interactions, is currently too complex to be well
influencing the process of biohydrogen production is the understood. To discover the rules of interphase interaction,
relative volume fraction of the sludge and the gas phases. such as force balance between sludge particles and gas
Sludge bed expansion and the existence of sludge floccs, bubbles, is crucial to the understanding of biohydrogen
greatly change the nature of the interaction between the production, because gas release from the sludge blanket will
liquid and gas phases. This interaction plays a significant role occur only when the force balance between particles and
in the expanded sludge bed reactor [38]. When the mean bubbles is disrupted. It should be noted that work described in
sludge concentration is significantly higher than local sludge this paper on solid phase simulation, only describes stable
concentrations; this indicates that the particles are aggre- conditions. Certain details about the movement of sludge
gating into clusters. Fig. 7 shows the sludge volume fraction at particles during changes in process conditions, such as wash-
six different simulation states. The values of the averaged bed out, have not been addressed; however, these problems could
no doubt be solved by further research. Fig. 8 presents the
biogas volume fraction predicted by CFD simulation; the
simulation results demonstrate that gas bubbles are an
important phenomenon. The motion of the biogas bubbles is
both upward and transverse. Sludge particle clusters were
dragged transversely, following the biogas bubbles. This
behavior could lead to wash-out, and a lower biomass with
a consequent reduction in biohydrogen yield. Bubble move-
ment, bursting, and size change are key factors affecting
particle circulation and cluster formation in the reactor. We
intend to further extend and validate our work by comparing
our modeling results with experimentation.

3.3. Hydrogen production analysis

The strategy of operation for hydrogen production using the

laboratory-scale H2-producing EGSB reactor (more completely
Fig. 10 – Hydrogen production rate, MFR and HRT variation described by Guo et al. [18]) is shown in Table 1. Fig. 9 shows
in the EGSB reactor. hydrogen production rate, hydrogen yield and Organic Loading
9694 international journal of hydrogen energy 34 (2009) 9686–9695

Rate (OLR) variation in the EGSB reactor. Fig. 10 shows a rela- Key Fund of China (No. 50638020), the Provincial Youthful
tionship between HRT and biohydrogen production rate in an Science Foundation of Heilongjiang (Grant No. QC06C036), and
experiment where inlet chemical oxygen demand was kept State Key Laboratory of Urban Water Resource and Environ-
constant and HRT was varied, and a simulated relationship ment, Harbin Institute of Technology (2008QN02) for their
between HRT and the outlet hydrogen mass flow rate (MFR). supports for this study.
From our experimental data we see that HRT, being related
to water up-flow velocity, affects biohydrogen production.
references
When the HRT exceeds 2 hours the hydrodynamic behavior
occurring is suitable for biohydrogen production. This is
because (see section 3.2), this behavior gives an appropriate
[1] Liu XM, Ren NQ, Song FN, Yang CP, Wang AJ. Recent
velocity distribution to maximize interphase interaction. By advances in fermentative biohydrogen production. Progress
integrating this information with the previous simulation in Natural Science 2008;18(3):253–8.
results, a qualitative relationship between hydrodynamics [2] Kapdan IK, Kargi F. Bio-hydrogen production from waste
and biohydrogen production can be obtained. Although materials. Enzyme and Microbial Technology 2006;38(5):
uniformity of velocity distribution in the reactor is improved 569–82.
[3] Demirbas A. Progress and recent trends in biofuels. Progress
when the HRT is reduced, the hydrogen yield does not
in Energy and Combustion Science 2007;33(1):1–18.
continually increase. Short HRTs need a high liquid up-flow
[4] Han SK, Shin HS. Biohydrogen production by anaerobic
velocity, although this generates better velocity distribution, fermentation of food waste. International Journal of
the consequent strong turbulence causes damage to the Hydrogen Energy 2004;29(6):569–77.
microorganisms and sludge-flocs, reducing biohydrogen [5] Hawkes FR, Dinsdale R, Hawkes DL, Hussy I. Sustainable
production. Despite being essentially a hydrodynamic study, fermentative hydrogen production: challenges for process
our work clearly indicates that choosing an appropriate HRT is optimisation. International Journal of Hydrogen Energy 2002;
27(11–12):1339–47.
extremely important to optimized biohydrogen production.
[6] Guo WQ, Ren NQ, Chen ZB, Liu BF, Wang XJ, Xiang WS, et al.
Simultaneous biohydrogen production and starch
wastewater treatment in an acidogenic expanded granular
4. Conclusions sludge bed reactor by mixed culture for long-term operation.
International Journal of Hydrogen Energy 2008;33(24):
7397–404.
We have performed a two-dimensional CFD simulation of
[7] Ren N, Wang X, Xiang W, Lin M, Li J, Guo W. Hydrogen
a gas–liquid–solid three-phase flow in the reaction zone of production with high evolution rate and high yield by
a laboratory-scale EGSB reactor producing biohydrogen. To immobilized cells of hydrogen-producing bacteria strain B49
evaluate the role of hydrodynamics in the design of a reactor, in a column reactor. High Technology Letters (English
CFD simulations have been carried out over a range of HRT. Language Edition) 2002;8(4):21–5.
The results showed that the gas–liquid–solid system in an [8] Khanal SK, Chen WH, Li L, Sung SW. Biohydrogen production
in continuous-flow reactor using mixed microbial culture.
EGSB reactor displayed significant variation in heterogeneous
Water Environment Research 2006;78(2):110–7.
structure particularly when the HRT was extended. The [9] Wang AJ, Ren NQ, Shi YG, Lee DJ. Bioaugmented hydrogen
typical core-annulus structure developed has lower water production from microcrystalline cellulose using co-culture –
velocities in the core region than those in the annulus. Water Clostridium acetobutylicum X-9 and Etilanoigenens harbinense
velocities near the wall are increasing and upward; this may B-49. International Journal of Hydrogen Energy 2008;33(2):
lead to the back mixing and internal circulation behavior 912–7.
[10] Ren NQ, Lin HL, Zhang K, Zheng GX, Duan ZJ, Lin M. Cloning,
which is detrimental to velocity distribution. The rules of
expression, and characterization of an acetate kinase from
interphase, such as force balance between solid and gas, are
a high rate of biohydrogen bacterial strain Ethanoligenens sp
known to have a direct bearing on biohydrogen-gas release, hit B49. Current Microbiology 2007;55(2):167–72.
and there is a need for better understanding to optimize bio- [11] Ren NQ, Liu BF, Ding J, Xie GJ. Hydrogen production with R.
hydrogen production. Bubble movement, bursting, and size faecalis RLD-53 isolated from freshwater pond sludge.
change are key factors affecting velocity distribution and Bioresource Technology 2009;100(1):484–7.
sludge particle movement in the reactor. Therefore, recording [12] Ren NQ, Xing DF, Rittmann BE, Zhao LH, Xie TH, Zhao X.
Microbial community structure of ethanol type fermentation
and analyzing bubble information in transient condition
in bio-hydrogen production. Environmental Microbiology
models are likely to be a fruitful area for further research.
2007;9(5):1112–25.
Experimental data from biohydrogen production integrated [13] Xing DF, Ren NQ, Gong ML, Li JZ, Li QB. Monitoring of
with our simulation indicate that a qualitative relationship microbial community structure and succession in the
between hydrodynamics and biohydrogen production can be biohydrogen production reactor by denaturing gradient gel
obtained. These results indicate that an appropriate HRT is electrophoresis (DGGE). Science in China Series C-Life
crucial to optimized biohydrogen production. Sciences 2005;48(2):155–62.
[14] Xia JY, Wang SJ, Zhang SL, Zhong JJ. Computational
investigation of fluid dynamics in a recently developed
centrifugal impeller bioreactor. Biochemical Engineering
Journal 2008;38(3):406–13.
Acknowledgments [15] Lee KS, Wu JF, Lo YS, Lo YC, Lin PJ, Chang JS. Anaerobic
hydrogen production with an efficient carrier-induced
The authors would like to thank the National Natural Science granular sludge bed bioreactor. Biotechnology and
Fund of China (No. 30870037), the National Natural Science Bioengineering 2004;87(5):648–57.
 international journal of hydrogen energy 34 (2009) 9686–9695 9695

[16] Chang JS, Lee KS, Lin PJ. Biohydrogen production with fixed- International Journal of Hydrogen Energy 2009;34(9):
bed bioreactors. International Journal of Hydrogen Energy 4189–97.
2002;27(11–12). PII S0360-3199(02) 00130-1. [29] Nagarajan V, Ponyavin V, Chen Y, Vernon ME, Pickard P,
[17] Chang FY, Lin CY. Biohydrogen production using an up-flow Hechanova AE. CFD modeling and experimental validation of
anaerobic sludge blanket reactor. International Journal of sulfur trioxide decomposition in bayonet type heat
Hydrogen Energy 2004;29(1):33–9. exchanger and chemical decomposer for different packed
[18] Guo WQ, Ren NQ, Wang XJ, Xiang WS, Meng ZH, Ding J, bed designs. International Journal of Hydrogen Energy 2009;
et al. Biohydrogen production from ethanol-type 34(6):2543–57.
fermentation of molasses in an expanded granular sludge [30] Rismanchi B, Akbari MH. Performance prediction of proton
bed (EGSB) reactor. International Journal of Hydrogen exchange membrane fuel cells using a three-dimensional
Energy 2008;33(19):4981–8. model. International Journal of Hydrogen Energy 2008;33(1):
[19] Mohan SV, Babu VL, Sarma PN. Anaerobic biohydrogen 439–48.
production from dairy wastewater treatment in sequencing [31] Milewska A, Molga EJ. CFD simulation of accidents in
batch reactor (AnSBR): effect of organic loading rate. Enzyme industrial batch stirred tank reactors. Chemical Engineering
and Microbial Technology 2007;41(4):506–15. Science 2007;62(18–20):4920–5.
[20] Zhang ZP, Tay JH, Show KY, Yan R, Liang DT, Lee DJ, et al. [32] Murthy BN, Ghadge RS, Joshi JB. CFD simulations of gas-
Biohydrogen production in a granular activated carbon liquid-solid stirred reactor: prediction of critical impeller
anaerobic fluidized bed reactor. International Journal of speed for solid suspension. Chemical Engineering Science
Hydrogen Energy 2007;32(2):185–91. 2007;62(24):7184–95.
[21] Chisti MY, Mooyoung M. Hydrodynamics and oxygen- [33] Bin Y, Zhang MC, Dou BL, Song YB, Wu J. Discrete particle
transfer in pneumatic bioreactor devices. Biotechnology and simulation and visualized research of the gas–solid flow in
Bioengineering 1988;31(5):487–94. an internally circulating fluidized bed. Industrial &
[22] Tardieu E, Grasmick A, Geaugey V, Manem J. Hydrodynamic Engineering Chemistry Research 2003;42(1):214–21.
control of bioparticle deposition in a MBR applied to [34] He YR, Lu HL, Sun QQ, Yang LD, Zhao YH, Gidaspow D, et al.
wastewater treatment. Journal of Membrane Science 1998; Hydrodynamics of gas–solid flow around immersed tubes in
147(1):1–12. bubbling fluidized beds. Powder Technology 2004;145(2):
[23] Allen DG, Robinson CW. Hydrodynamics and mass-transfer 88–105.
in Aspergillus-niger fermentations in bubble column and [35] Ren TT, Mu Y, Ni BJ, Yu HQ. Hydrodynamics of upflow
loop bioreactors. Biotechnology and Bioengineering 1989; anaerobic sludge blanket reactors. Aiche Journal 2009;55(2):
34(6):731–40. 516–28.
[24] Jin B, Lant P. Flow regime, hydrodynamics, floc size [36] Diez L, Zima BE, Kowalczyk W, Delgado A. Investigation of
distribution and sludge properties in activated sludge bubble multiphase flow in sequencing batch reactor (SBR) by means
column, air-lift and aerated stirred reactors. Chemical of hybrid methods. Chemical Engineering Science 2007;62(6):
Engineering Science 2004;59(12):2379–88. 1803–13.
[25] Krishna R, van Baten JM, Urseanu MI. Scale effects on the [37] Elghobashi SE, Abouarab TW. A 2-equation turbulence
hydrodynamics of bubble columns operating in the model for 2-phase flows. Physics of Fluids 1983;26(4):
homogeneous flow regime. Chemical Engineering & 931–8.
Technology 2001;24(5):451–8. [38] Dou BL, Dupont V, Williams PT. Computational fluid
[26] Lee KS, Lin PJ, Chang JS. Temperature effects on biohydrogen dynamics simulation of gas–solid flow during steam
production in a granular sludge bed induced by activated reforming of glycerol in a fluidized bed reactor. Energy &
carbon carriers. International Journal of Hydrogen Energy Fuels 2008;22(6):4102–8.
2006;31(4):465–72. [39] Sun J, Battaglia F. Hydrodynamic modeling of particle
[27] Ferng YM, Su A. A three-dimensional full-cell CFD model rotation for segregation in bubbling gas-fluidized beds.
used to investigate the effects of different flow channel Chemical Engineering Science 2006;61(5):1470–9.
designs on PEMFC performance. International Journal of [40] Huilin L, Gidaspow D, Bouillard J, Wentie L. Hydrodynamic
Hydrogen Energy 2007;32(17):4466–76. simulation of gas–solid flow in a riser using kinetic theory of
[28] Hawkes G, O’Brien J, Stoots C, Hawkes B. 3D CFD model of granular flow. Chemical Engineering Journal 2003;95(1–3):
a multi-cell high-temperature electrolysis stack. 1–13.