Page 1
Timo Dymala, Ernst-Ulrich Hartge, Stefan Heinrich
Institute of Solids Process Engineering and Particle Technology,
Hamburg University of Technology
Fluidization XVI – May 26 - 31, 2019
The MP-PIC method for CFD-simulation of
fluidized beds – Comparison of two
different implementations
Page 2
www.tuhh.de/spewww.tuhh.de/spe
Motivation
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)2
Computational speed vs. accuracy
− Large-scale systems (e.g. combustors) involving up to 1018 particles are time-intensive to model
➢ Tradeoff between computational speed and accuracy for the simulation of fluidized beds
Solid
sre
pre
sen
tati
on
Generalized idea for particle interactions
CGPM(Coarse GrainedParticle Method)Masaaki et al., 2000Sakai and Koshizuka,
2009
CFD-DEM(Computation Fluid Dynamic-DiscreteElement Method)
Tsuhi et al., 1993
CGHS(Coarse Grained
Hard Sphere)Lu et al., 2017
ED/TD HS(Event Driven/ Time Driven Hard Sphere)Hoomans et al., 1996Ouyang and Li, 1999
MP-PIC(Multiphase –Particle in Cell)
Andrews and O‘Rourke, 1996
Collision resolvedMomentum conservation
Solid Stress Gradient
Pa
rtic
leP
arc
el
Reducedaccuracy
Increasedcomputationalspeed
[1] Lu et al. (2017): Ind. Eng. Chem. Res., 56 (27)
Page 3
www.tuhh.de/spewww.tuhh.de/spe
Methodology
3
Software implementation
Barracuda VR® 17.3
− Proprietary software by CPFD Software LLC
− Solver for compressible fluids
− Structured grid
OpenFOAM® v6
− Open source software by The OpenFOAM Foundation Ltd
− Solver for incompressible fluids
− Unstructured grid
Different software – Identical results?
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)
Page 4
www.tuhh.de/spewww.tuhh.de/spe
Methodology
4
Models
Particle stress
− Explicit inter-particle stress model
according to Harris and Crighton [2]
𝜏 Inter-particle stress
P𝑠 Modeling constant
𝛼𝑝 Particle volume fraction
𝛼𝑐𝑝Particle volume fraction at
close-pack
𝛽 Modeling constantTurbulence model
− Large Eddie Simulation (LES) according to Smagorinsky [3]
➢ Resolving large length scales, but model smallest
length scales to reduce the computational costs
[2] Harris and Crighton (1994): Journal of Fluid Mechanics, 266.
[3] Smagorinsky (1963): Monthly Weather Review, 91.
[4] Sagaut (2006): Large Eddy Simulation for Incompressible Flows, Springer.
[5] Gidaspow (1994): Multiphase Flow and Fluidization, Academic Press.
[6] Wen and Yu (1966): Chemical Engineering Progress Symposium Series, 62.
[7] Ergun (1952): Chemical Engineering Progress, 48.
Drag model according to Gidaspow [5]
− Combination of approaches by Wen and Yu [6] and Ergun [7]
− Homogeneous drag model commonly used in literature
𝜏 =𝑃𝑠 ⋅ 𝛼𝑝
𝛽
max 𝛼𝑐𝑝 − 𝛼𝑝, 𝛽 ⋅ (1 − 𝛼𝑝)
Schematic representation of the LES [4].
Grid scaleEddies are solved
Sub-grid scaleEddies are modeled
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)
Page 5
www.tuhh.de/spewww.tuhh.de/spe
Simulation Setup
5
Lab-scale reactor
Geometry
− Height: 1.4 m, diameter: 0.1 m
Material: Quartz sand (Geldart B)
− Initial bed height: hbed = 0.1 m
➢ Bed mass: mbed = 1.1 kg
− Solid density: 𝜌𝑠 = 2600 kg/m3
− Close pack fraction: αcp = 0.54
− Sauter diameter: d32 = 220 µm
Process conditions
− Isothermal flow (300 K)
− Velocity range from bubbling to
turbulent fluidization
➢ ug = 0.21 – 1.33 m/sGeometry of the lab-scale reactor.
0
20
40
60
80
100
0 100 200 300 400 500 600
Siz
e d
istr
ibu
tio
nQ
3[%
]
Particle diameter dP [μm]
Measured particle size distribution of F34 quartz sand.
0 m
0.5 m
1.0 m
1.5 m
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)
Page 6
www.tuhh.de/spewww.tuhh.de/spe
Results
6
Lab-scale reactor
Key Parameter: Solids concentration
averaged over the cross-section and
over time using the pressure drop
𝜀𝑠 =𝑑𝑝/𝑑𝑧
𝜌𝑠 ∙ 𝑔
− Similar results for low velocities
− Diverging results with increasing
velocities
➢ Error of pressure drop increases
with increasing velocity
Velocity
[m/s]
Calculated
mass [kg]
Relative
error
0.21 1.045 4.98 %
0.55 1.031 6.29 %
0.81 1.009 8.24 %
1.33 0.925 15.94 %
ug = 0.21 m/s
ug = 0.55 m/s
0
0.1
0.2
0.3
0.4
0.5
0.6
0 0.2 0.4 0.6
Ave
rage
so
lids
volu
me
fr
acti
on
[-]
Height [m]
BarracudaOpenFOAM
0
0.1
0.2
0.3
0.4
0.5
0.6
0 0.2 0.4 0.6
Ave
rage
so
lids
volu
me
fr
acti
on
[-]
Height [m]
BarracudaOpenFOAM
ug = 0.81 m/s
0
0.1
0.2
0.3
0.4
0.5
0.6
0 0.2 0.4 0.6
Ave
rage
so
lids
volu
me
fr
acti
on
[-]
Height [m]
BarracudaOpenFOAM
0
0.1
0.2
0.3
0.4
0.5
0.6
0 0.2 0.4 0.6
Ave
rage
so
lids
volu
me
fr
acti
on
[-]
Height [m]
BarracudaOpenFOAM
ug = 1.33 m/s
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)
Table 1: Relative error of calculated
mass (OpenFOAM).
Page 7
www.tuhh.de/spewww.tuhh.de/spe
Results
7
Lab-scale reactor
Comparison of video recordings: Recording frequency 50 Hz slowed down to 10 fps
ug = 0.55 m/s ug = 1.33 m/s
Barracuda ExperimentOpenFOAM ExperimentBarracudaOpenFOAMV
elo
city m
agnitude [m
/s]
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)
Page 8
www.tuhh.de/spewww.tuhh.de/spe
Simulation Setup
8
Pilot-scale reactor
Geometry [8,9,10]
− Height: 8.35 m, cross-section: 1 m x 0.3 m
Material: Quartz sand (Geldart B particles) [8]
− Initial bed mass: mbed = 300 kg
− Solid density: 𝜌𝑠 = 2600 kg/m3
− Sauter diameter: d32 = 150 µm
Process conditions
− Fluid velocity: ug = 3 m/s (isothermal flow at 300 K)
− External circulation rate Gs is adjusted to result in a constant
solid hold-up around 300 kg
➢ Experimental circulation rate: Gs = 20 kg/(m2∙s) [8]
Geometry of the pilot-scale reactor.
0 m
1 m
2 m
3 m
4 m
5 m
6 m
8 m
7 m
[8] Schlichthärle (2000): Doctoral thesis, TUHH.
[9] Hartge et al. (2009): Particuology, 7.
[10] Chen et al. (2013): Powder Technology, 235.
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)
Page 9
www.tuhh.de/spewww.tuhh.de/spe
Results
9
Pilot-scale reactor
− Simulated solids fractions too low in bottom zone
− Barracuda predicts a more homogeneous axial
distribution than OpenFOAM
− External circulation rates are adjusted to obtain a
constant solid hold-up of 300 kg
➢ Gs,exp = 20 kg/(m2∙s) [8]
➢ Gs,OF = 17 kg/(m2∙s)
➢ Gs,BC = 46 kg/(m2∙s)
DataSolid hold-
up [kg]
Pressure
drop [mbar]
Experimental [8] 300 99
Barracuda 300 87
OpenFOAM 300 73
➢ Lower pressure drop with OpenFOAM but more
homogeneous axial distribution with Barracuda
➢ Better agreement of external circulation rate with
OpenFOAM
[8] Schlichthärle (2000): Doctoral thesis, TUHH.
0.0
0.1
0.2
0.3
0.4
0 2 4 6 8
Ave
rage
so
lids
volu
me
fra
ctio
n [
-]
Height [m]
Experiment [8]
OpenFoam
Barracuda
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)
Page 10
www.tuhh.de/spewww.tuhh.de/spe
Summary
10
The MP-PIC method for CFD-simulation of fluidized beds
− Both implementations can predict fluidization behavior at lower
fluidization velocities with reasonable agreement
− Both implementations under-predict the resulting pressure drop
➢ Relative error increases with increasing velocity
− Despite the same conditions significant differences between
OpenFOAM and Barracuda
➢ Lower fluidization intensity and pressure drop with OpenFOAM
compared to Barracuda
➢ More realistic segregation of particles and external circulation rate
with OpenFOAM
Next step:
➢ Implementation of EMMS based drag model for a better
agreement with the experimental data
Instantaneous velocity magnitude of the particles (left)
and the fluid (right) in the pilot-scale riser (OpenFOAM).
Fluidization XVI – The MP-PIC method for CFD-simulation of fluidized beds – Comparison of two different implementations (Timo Dymala)
Page 11
Thank you
for your attention!
We gratefully acknowledge for the financial support the German Research
Foundation (DFG) (Germany) within the scope of the joint project “Multi-scale
analysis and optimization of chemical looping gasification of biomass” with the
Southeast University, Nanjing, China. Project number HE 4526/21-1.