Professional Documents
Culture Documents
Jaap 流化床流场仿真
Jaap 流化床流场仿真
A R T I C L E I N F O A B S T R A C T
Keywords: Pyrolysis of biomass and organic matter in fluidized bed reactors have become popular in recent years due to the
Pyrolysis reactor efficient conversion of biomass to bio-fuels. Among others, the presence of mesoscale structures produced by the
CFD transient nature of gas solid flow governs the performance of fluidized bed reactors. In the present study,
Gas-solid
monodispersed system with the particle size of 155 µm and bidispersed system with particle sizes of 128 and
Heterogeneity
1500 µm are simulated by the multi-fluid model with the kinetic theory of granular flow closure laws. The
objective is to investigate the effect of size distribution and solid volume fraction on the hydrodynamics of gas-
solid flows. The ANSYS Fluent® software is used for the simulations and the slip velocity, heterogeneity index as
well as mixing index are investigated. A two dimensional fully periodic domain is used to investigate the slip
velocity at five different solid concentrations in dilute range. The slip velocity, which represents the interaction
between solid particles and gas, calculated by simulation is higher than that calculated through Richardson-
Zaki’s correlation. The mixing index is found to be minimum for a solid volume concentration of 5%. The
heterogeneity index is significantly less than 1 for both cases with bidisperesed system having lower value as
compared to monodispersed system. The image analysis of solid fraction contours exhibits that mesoscale
structures are dense and non-uniform for solid fractions of 0.05 and are small and uniform of solid fractions of
0.025 and 0.075.
* Corresponding author.
E-mail address: atta@pieas.edu.pk (A. Ullah).
https://doi.org/10.1016/j.jaap.2022.105459
Received 15 October 2021; Received in revised form 30 January 2022; Accepted 31 January 2022
Available online 2 February 2022
0165-2370/© 2022 Elsevier B.V. All rights reserved.
S. Shahzad et al. Journal of Analytical and Applied Pyrolysis 162 (2022) 105459
CFB [18]. These interactions lead to the formation of various hetero Secondly, simulation of the bidisperse system (particle of two different
geneous structures over a wide range of scales both in space and time sizes) with a particle diameter of 128 and 1500 µm [28] with the mesh
[19]. Among these, mesoscale structures are of critical importance size of Δx = 0.33 cm is carried out. The relative slip velocity of this
owing to its impact on the static and dynamic properties of the sus system at five different solid concentrations from 1% to 10% are then
pended system. Despite their importance, these structures are difficult to compared with the monodisperse system. Another important parameter
characterize, as they constantly and randomly evolve both in space and known as degree of mixedness is studied and results are presented based
time [20]. These parcels of particles clusters perform a vital role in the on Lacey mixing index formulation [30]. Lastly, an attempt is made to
distribution of particles both in axial and radial directions and affect the characterize the time-averaged mesoscale clusters distributions based
overall hydrodynamic of the complex gas-solid system [21,22]. The upon the statistical analysis of snapshots of solid fraction contours.
mesoscale clusters may range up to 75% of the particles in the system
depending upon the solid concentration, gas velocities and particle size 2. CFD modeling
distribution [23]. A theoretical model is proposed by Subbarao [24] for
predicting the equivalent diameter of the particle clusters. There are mainly three most widely used modeling techniques which
The CFD approaches to study mesoscale structure consisting of both are numerically investigate the complex hydrodynamics of gas-solid
Eulerian Two-Fluid Model (TFM) [18,25] and Lagrangian Discrete flows, which are categoried as: (1) Eulerian-Eulerian modeling, (2)
Element Method (DEM) [25] have been reported in the literature. The Eulerian-Langrangian modeling, (3) Hybrid Eulerian-Langrangian
ability to handle the small number of particles limits the application of modeling [25]. The last two listed models are based on lagrangian
DEM approach, whereas TFM approach is the main technique to simu technique, which is computationally intensive and hence can not be
late large industrial processes [26]. EMMS mixture model approach with used for industrial applications, where one has to deal with a large
particle size distribution has been applied to analyze the hydrodynamics number of particles. In Eulerian-Eulerian two-fluid models, both the
of an industrial-scale riser [27]. Recently, Eulerian CFD analysis was solid phase and fluid phase are taken as interpenetrating continua and
carried out to study the size of clusters in circulating fluidized bed re equations for mass, momentum and energy conservation are solved for
actors and distribution in time [4]. individual phases separately [31]. Closure-related problems arise for
Multiple CFD models are available in the literature.Among these both solid and gas phases when averaged Navier Stokes equations are
models, Eulerian multiphase model based on the kinetic theory of employed. Here solid phase is taken as granular flow and the kinetic
granular flow is found to be less intensive computationally with the theory of granular flow (KTGF) is used to provide closure laws for
reasonable accuracy [13]. In order to capture the mesoscale structures solid-phase momentum equation. This two-fluid model is capable to
mesh resolution should be fine, typically of the order of 10–100 particle resolve heterogeneous structures such as streamers and clusters of par
diameter [28]. The continuum equations combined by the constitutive ticles [32]. But the efficacy of this model depends upon the mesh reso
equations for particle phase stress as derived from the Kinetic theory of lution. the Eulerian-Eulerian approach becomes computationally costly
Granular Flows (KTGF) can be applied to capture the mesoscale struc due to the requirement of small time scale and fine mesh resolution. The
tures. A coarse grid can not capture the mesoscale structure, as these reason behind these flow structures is the inertial instability, which is
structures are small in size. induced by the random motion of solid particles and their interparticle
In practice, gas-solid fluidized beds normally consist of solid parti and hydrodynamic interactions [21]. These interactions induce the
cles, having a broad size range as well as different densities [29], mesoscale structures, in which particles move as a lump (cluster), sup
however, we have restricted our study to gas-solid fluidized suspension pressing the individual motion of particles. The velocities of these
consisting of particles that only differ in size. clusters decrease and hence for the same drag force the relative velocity
This study is focused to analyze the effect of mesoscale structures on between gas and solid increases [33].
relative slip velocity of equivalent size particles in monodisperse and
bidispersed system and how much these structures cause the system to
2.1. Kinetic theory for granular flow
deviate from homogeneity. For this purpose, a two dimensional simu
lation of the monodisperse system (particle of same size) with particle
According to KTGF approach, the irregular motion of solid particles
diameter of 155 µm [19],using mesh size Δx = 0.39 cm is presented.
in a gaseous medium is considered analogous to the thermal motion of
2
S. Shahzad et al. Journal of Analytical and Applied Pyrolysis 162 (2022) 105459
Table 1
Governing equations and models for closure used in simulation.
Mass conservation equation: (m = g, s) ∂(ρm εm ) (1)
+ ∇⋅(ρm εm um ) = 0
∂t
Momentum conservation equation (gas phase):
( )
∂ ρg εg ug ( ) ( ) (2)
+ ∇⋅ ρg εg ug ug = − εg ∇p + ρg εg g + ∇⋅τg + β ug − us
∂t
Momentum conservation equation (solid phase):
∂(ρs εs us ) (3)
+ ∇⋅(ρs εs us us ) = − εs ∇p + ρs εs g − ∇Ps + ∇τs + β(ug − us )
∂t
Gas phase stress tensor: τg = 2μg Sg + εg λg ∇ug (4)
Solid phase tensor: τs = [− Ps + εs λs ∇us ]I + 2μs Ss (5)
Strain rate tensor: 1 [ ] 1 (6)
Sm = εm ∇um +(∇um )T − εm ∇um I
2 3
Particle phase pressure: Ps = εs ρp Θs + 2(1 +e)ε2s go ρp Θs (7)
Shear viscosity for solids:
√̅̅̅̅̅̅̅̅[ ]2
4 (Θ )1/2 10ρp dp Θs π 4 Ps sinϕ
μs = μs,col +μs,kin +μs,fr = ε2s ρp dp go (1 +e) s
+ 1 + εs go (1 + e) + √̅̅̅̅̅̅̅ (8)
5 π 96(1 + e)go 5 2 I2D( )
Granular bulk viscosity: 4 Θs 1/2 (9)
λs = εs dp ρp go (1 +e)
3 π
Granular energy equation:
[ ( ) ]
3 ∂ ρ p εs Θ s ( )
+ ∇. εs ρp us Θs = τs : (∇.us ) +∇.q − γ − 3βΘs (10)
2 ∂t
Granular energy diffusive flux: q = − κs ∇Θs (11)
6 (12)
2κk [1 + go (1 + e)εs ]2
κs = 5 + κc
(1 + e)go
√̅̅̅̅̅̅̅̅
75dp ρp πΘs (13)
κk =
384
(Θ )1/2
s (14)
κc = 2ε2s go dp ρp (1 +e)
π
Collisional energy dissipation: 12 (15)
γ = (1 − e2 )ε2s ρp go √̅̅̅ Θ3/2
s
[ d]p π
Radial distribution function: ( )1/3 − 1 (16)
εs
go = 1 −
εs,max
3
S. Shahzad et al. Journal of Analytical and Applied Pyrolysis 162 (2022) 105459
conditions is selected as the simulation domain (Fig. 1). Domain size particle Reynolds number (Re). It is stated earlier that mesoscale
with the dimension of 20 cm × 40 cm is chosen to avoid fluctuations in structures are intrinsically unstable and they manifest fluctuations.
the settling velocity and the swirls to avoid the finite-range correlations Fig. 2 is the time response of slip velocity showing that initially slip
that depend on volume fraction ϕ and particle radius. This finite-range velocity signal is a straight line, fluctuating n as time passes. The uni
correlation is known as correlation length which is found to 20 times form signal of slip velocity means that particles are homogeneously
the interparticle distance (ζ = 20a ϕ − 1/3) [36]. Cell size is chosen such distributed and are separated as fluidization starts. During fluidization
that Δx/dp is less than 26. The number of cells for monodisperse system these solid particles start interacting with each other due to random
and bidisperse system is calculated as 50 and 60 in x-direction and 100 motion, inducing particle aggregation and breakage, which is shown by
and 120 for y-direction respectively. the fluctuating velocity signal.
Table 2
Solver settings for mono and bidispersed system.
Description Case Setting
Dimension of System 20 cm × 40 cm
Mesh resolution (monodisperse) < 26
Mesh resolution (bidisperse) < 2.2
Boundary conditions Periodic
Operating pressure 1.01 × 105 Pa
Gravitational Acceleration 9.81 ms− 2
Solid particle density 2500 kg/m3
Fluid density 1.3 kg/m3
Fluid viscosity 1.78 × 10− 5
Momentum discretization Second order upwind
Volume fraction discretization QUICK
Temporal discretization First order implicit
Time step 1 × 10− 5 s
Drag law Wen and Yu [35]
Granular viscosity Gidaspow [39]
Granular bulk viscosity Lun et al. [34]
Frictional viscosity Schaeffer [40]
Granular temperature Algebraic
Internal friction angle 30◦
Maximum solid packing limit 0.63
Maximum iteration per time step 100
Convergence criteria 1 × 10− 3
Fig. 2. Time response of slip velocity for monodisperse system for ф = 0.025.
4
S. Shahzad et al. Journal of Analytical and Applied Pyrolysis 162 (2022) 105459
distinction is in the location of the peak of the curve for slip velocity, for heterogeneous structures is less as compared to that of the homogeneous
the monodisperse system it is at 5% of solid volume fraction and for the Wen-Yu drag model [26]. It is significant to point out that the curve for
bidisperse system, the peak point occurs at 7.5% of solid volume frac bidispersed system is even lower than that of the monodispersed system
tion. The work of Gidaspow’s group has demonstrated that the classic which shows that size distribution brings larger clusters in the system.
MFM has the capability of capturing heterogeneous structures like The continuous creation and breakage of clusters in the system is the
clusters in a fluidized bed [32]. signature of inducing heterogeneity in the flow. Therefore, the hetero
geneous drag coefficient is decreased [21]. As the solid concentration
4.2. Heterogeneity index remains unchanged in the simulation domain, higher slip velocity im
plies a lower drag coefficient. It indicates that the presence of mesoscale
A heterogeneity index [26] corresponds to the hydrodynamic clusters will reduce the drag significantly [21,48]. The plot of hetero
discrepancy between homogeneous and heterogeneous fluidization. It is geneity index is in good agreement with that obtained by using fine grid
the ratio of heterogeneous drag coefficient to homogeneous Wen-Yu periodic simulation for voidage range (0.4–1), as presented in the study
drag coefficient [35], defined as of Ullah et al. [26].
β
HD = , (21) 4.3. Mixing index
βWY
HD takes the value close to unity at extreme ends of void fraction M= , (24)
S2R − S20
range, and solid and gas phases compromise to form either particle
dominant or gas dominant flow. This trend in the HD curve represents Where S2 , S20 and S2R represents the variance of number fraction of a class
the homogeneity towards both the dilute and dense extremes [26]. The of particle, the maximum value of mixture variance and possible mini
multifluid model with homogenous drag model with fully periodic mum value of variance, respectively. These parameters are defined as;
boundary conditions is used to achieve structure-dependent drag and
the effect of heterogeneous structure on drag coefficient is observed 1 ∑ N
S2 = (xi − xm )2 , (25)
[47]. Fig. 4 is the plot of HD as a function of voidage, showing that the N − 1 i=1
drag coefficient decrease owing to flow heterogeneity. It can be
observed that for both mono and bidispersed systems, the value of HD is S20 = xm (1 − xm ), (26)
less than unity which means that the drag coefficient characterizing
5
S. Shahzad et al. Journal of Analytical and Applied Pyrolysis 162 (2022) 105459
xm (1 − xm ) 7.5% and 10%. It is apparent that at low solid volume fractions in the
S2R = . (27) monodisperse case, the mesoscale structures are few and virtually
n
nonexistent in the bidisperse case. At higher volume fractions of solids,
Lacey mixing index is plotted against the solid volume fraction in the trend shows that bidispersed particles produce fewer mesoscale
Fig. 5. It is worthy to notice here that mixing index values first decrease structures than the monodisperse.
as the solid concentration is increased and then increase. In an ideal The contour plots of solid volume fractions are processed to quantify
case, particles in the system should be completely mixed in a homoge the distribution of mesoscale structures in the domain. Gray scale con
nous manner or unsegregated fashion in the final particle mixture [51]. tours are plotted for solid volume fractions of 2.5%, 5% and 7.5%. The
There are some factors upon which the mixing process depends dimensions of images, and minimum and maximum contour scale limits
which include time, temperature, particles’ shape and size, shear rate are kept constant for all images. It is observed that the maximum
and particle loading [52]. The flow rate of gas is the main parameter to average concentration of particles in the domain has the value of
have an optimal value of mixing index for mixing by fluidized bed, ϕ = 0.62. This value of solid volume fraction is now fixed as the
however with an increase in the flow rate, power consumption increases maximum limit for all contours plots in post-processing of ANSYS®. The
[49]. In the present study, the gas flow rate is fixed and only the particle heterogeneous structure in gas-solid flow is now defined to be mesoscale
loading effect is studied. At minimal particle concentration, the value of structure for which pixel intensity is greater than the mean pixel in
the mixing index is high, indicating good mixing in the fluid. With the tensity value. The mean gray values (pixel intensity) for various solid
increase in particle loading, heterogeneous structures start developing, volume fractions are given in Table 3. In order to obtain the gray value,
inducing particle aggregation, decreasing the mixing in the system. With image processing is used with the help of imageJ® tool.
further increase in solid concentration, the clusters of particle start The pixel intensity represents the value of solid fraction at that
interacting with one another, leading to homogeneity [21], thus location in the image. The pixel intensities at various heights of the
increasing the mixing index. domain is obtained along the radial distances. The profile is drawn in
which y-axis represents the vertically averaged intensities of pixels and
4.4. Mesoscale structures characterization the x-axis represents the distance of domain in pixels. Fig. 8 represents
the intensity of pixels as a function of the distance measured in pixels. It
Mesoscale structures are particle-rich dense phase, having size range can be observed that the high intensity is found for ϕ = 0.05. It can also
of 10 – 100 particle diameters. As fluidization starts, the distribution of be observed that there are a few small heterogeneous structures that
solid particles becomes random and irregular, and the development of occur on the left side of the domain and one large structure appear on
mesoscale structures begins in the particle dense areas. Fig. 6 shows the the right side of the domain. For ϕ = 0.025, small structures are
evolution of mesoscale structures over time for monodispersed system observed with low gray values as at this volume fraction, indicating that
(Fig. 6a) and bidispersed system (Fig. 6b). It can be observed that the the formation of mesoscale structures is in progress. The formation of
development of mesoscale structures is intrinsically unstable and mesoscale structures is maximum at ϕ = 0.05 and with further increase
entirely fluctuating in nature, that is, clusters keep on building, growing in volume fraction these structures start interacting with each other,
and collapsing however an average value of heterogeneity can be which induces the breakage process. It is evident that at ϕ = 0.075 small
calculated. The development of these structures is the result of inertial but a number of gray value peaks still exist. It is evident from the figure
and local instabilities which come into the system from inelastic colli that the mesoscales structures are present in the system and intensities of
sion and dissipation of fluctuating movement of solid particles by pri pixels show how the structures form and break with the increase in solid
mary phase [13,21,27]. For bidispersed systems, mesoscale structures concentrations.
can also be observed, however, the intensity of such structures is rela
tively less as compared to the monodispersed system of the same average 4.5. Computational time
volume fraction. It is also observed that particles with large diameter
and higher mass fraction contribute more towards the formation of In an ideal case, three-dimensional simulations should be carried out;
mesoscale structures. however, three-dimensional simulation for each case is computationally
Fig. 7 displays the solid volume fraction distribution in the simula expensive than an equivalent two-dimensional simulation [21]. As
tion domain for the five solid volume fractions of 1.0%, 2.5%, 5.0%, several simulations have been performed to illustrate the main ideas
discussed in this study, three-dimensional simulations are simply
beyond our existing computing resources. For the current simulation, it
takes approximately 360 and 500 h of 5 s real time simulations of
monodisperse and bidisperse systems, respectively.
Monodispersed system deals with two phases, so conservation of
mass and momentum equations are solved separately for each phase
with the closure equations for the solid phase. In bidispersed system,
three sets of momentum and continuity equations along with the closure
equations for two solid phases are solved. As in the bidispersed system,
the large number of equations needs to be solved so computational time
is large. It is also observed that there is a quantitative difference between
the slip velocities for the mono and bidispersed system. Bidispersed
system is an acceptable representation of the polydisperse flow and it
should be favored against the mono-dispersed system in order to get
more realistic results. It is recommended to use a bidisperse system given
that enough computational resources are available.
5. Conclusions
6
S. Shahzad et al. Journal of Analytical and Applied Pyrolysis 162 (2022) 105459
Fig. 6. Evolution of mesoscale structures over time at ϕ = 0.05; a) monodispersed system, b) bidispersed system.
Fig. 7. Contours of solid volume fractions at five solid volume fractions; a) monodispersed system, b) bidispersed system.
10%. The observed increase in slip velocity with solids fraction is the
Table 3
due formation of mesoscale structures. At solid concentration around
Average gray value for different solid volume fractions.
0.075, the formation of mesoscale structures and the slip velocity attains
Solid Volume Fraction Mean Gray Value a maximum. However, values of slip velocity for the bidisperse case are
0.025 19.51 greater than for monodisperse case due to larger mesoscale structures
0.050 28.03 for bidispersed case. For the case of bidispersed system, more hetero
0.075 37.72 geneity is induced in the system and the HD value is lower as compared
to that of monodispersed system. Mixing index value of less than one for
bidispersed systems by varying the solid volume fraction from 1% to the bidispersed case indicates that simulated systems are not fully mixed
systems. The meso-scale structures (particle clusters) are a major cause
7
S. Shahzad et al. Journal of Analytical and Applied Pyrolysis 162 (2022) 105459
8
S. Shahzad et al. Journal of Analytical and Applied Pyrolysis 162 (2022) 105459
[36] P.N. Segrè, E. Herbolzheimer, P.M. Chaikin, Long-range correlations in [45] J.F. Richardson, W.N. Zaki, Sedimentation and fluidisation, Trans. Inst. Chem. Eng.
sedimentation, Phys. Rev. Lett. 79 (1997) 2574–2577, https://doi.org/10.1103/ 32 (1954) 35–53.
PhysRevLett.79.2574. [46] A. Ullah, W. Wang, J. Li, Evaluation of drag models for cocurrent and
[37] B. Lu, W. Wang, J. Li, Searching for a mesh-independent sub-grid model for CFD countercurrent gas-solid flows, Chem. Eng. Sci. 92 (2013) 89–104, https://doi.org/
simulation of gas-solid riser flows, Chem. Eng. Sci. 64 (2009) 3437–3447, https:// 10.1016/j.ces.2013.01.019.
doi.org/10.1016/j.ces.2009.04.024. [47] Y. Igci, S. Pannala, S. Benyahia, S. Sundaresan, Validation studies on filtered model
[38] L. Cammarata, P. Lettieri, G.D.M. Micale, D. Colman, 2D and 3D CFD simulations equations for gas-particle flows in risers, Ind. Eng. Chem. Res. 51 (2012)
of bubbling fluidized beds using eulerian-eulerian models, Int. J. Chem. React. Eng. 2094–2103, https://doi.org/10.1021/ie2007278.
1 (2002), https://doi.org/10.2202/1542-6580.1083. [48] W. Wang, J. Li, Simulation of gas-solid two-phase flow by a multi-scale CFD
[39] D. Gidaspow, Multiphase Flow and Fluidization, Elsevier, 1994, https://doi.org/ approach-of the EMMS model to the sub-grid level, Chem. Eng. Sci. 62 (2007)
10.1016/C2009-0-21244-X. 208–231, https://doi.org/10.1016/j.ces.2006.08.017.
[40] D.G. Schaeffer, Instability in the evolution equations describing incompressible [49] A.A. Alhwaige, S.M. Tasirin, A.M. Sowedan, W.R.W. Daud, Study The Homogeneity
granular flow, J. Differ. Equ. 66 (1987) 19–50, https://doi.org/10.1016/0022- of Mixing a Binary Polyethylene Granular Mixture in Fluidised Bed Mixer, in: Proc.
0396(87)90038-6. World Congr. Eng. Comput. Sci., International Association of Engineers, San
[41] D. Geldart, Types of gas fluidization, Powder Technol. 7 (1973) 285. Francisco, 2008, pp. 118–122. http://www.iaeng.org/publication/WCECS2008/
[42] A.R. Khan, J.R. Richardson, Fluid-particle interactions and flow characteristics of WCECS2008_pp118-122.pdf.
fluidized beds and settling suspensions of spherical particles, Chem. Eng. Commun. [50] S.H. Chou, Y. Lou Song, S.S. Hsiau, A study of the mixing index in solid particles,
78 (1989) 111–130, https://doi.org/10.1080/00986448908940189. KONA Powder Part. J. 2017 (2017) 275–281, https://doi.org/10.14356/
[43] F. Berruti, T.S. Pugsley, L. Godfroy, J. Chaouki, G.S. Patience, Hydrodynamics of kona.2017018.
circulating fluidized bed risers: a review, Can. J. Chem. Eng. 73 (1995) 579–602, [51] D.K. Rollins, D.L. Faust, D.L. Jabas, A superior approach to indices in determining
https://doi.org/10.1002/cjce.5450730502. mixture segregation, Powder Technol. 84 (1995) 277–282, https://doi.org/
[44] J.H. Pärssinen, J.X. Zhu, Axial and radial solids distribution in a long and high-flux 10.1016/0032-5910(95)02992-B.
CFB riser, AIChE J. 47 (2001) 2197–2205, https://doi.org/10.1002/ [52] R. Supati, N.H. Loh, K.A. Khor, S.B. Tor, Mixing and characterization of feedstock
aic.690471007. for powder injection molding, Mater. Lett. 46 (2000) 109–114, https://doi.org/
10.1016/S0167-577X(00)00151-8.