Particuology: Hendrik Otto, Kristin Kerst, Christoph Roloff, Gábor Janiga, André Katterfeld

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Particuology 40 (2018) 34–43

Contents lists available at ScienceDirect

Particuology
journal homepage: www.elsevier.com/locate/partic

CFD–DEM simulation and experimental investigation of the flow


behavior of lunar regolith JSC-1A
Hendrik Otto a,1 , Kristin Kerst b,1 , Christoph Roloff b , Gábor Janiga b,∗ , André Katterfeld a
a
Chair of Conveying Technologies, University of Magdeburg “Otto von Guericke”, Magdeburg, Germany
b
Laboratory of Fluid Dynamics and Technical Flows, University of Magdeburg “Otto von Guericke”, Magdeburg, Germany

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

Article history: Rovers on Mars and the Moon analyze the local geology by collecting samples of the upper layer in
Received 27 January 2017 containers and ovens. After the analysis, the complete discharge of samples from the reservoir must
Received in revised form be ensured. Because of the low atmospheric pressure, reduced gravity, and different grain shapes of
10 November 2017
the bulk material, the discharge process is very different compared to that on Earth. In this study, the
Accepted 8 December 2017
behavior of lunar regolith JSC-1A in closed containers during discharge was investigated by analyzing
Available online 24 April 2018
the flow in an hourglass under the Earth’s atmosphere. Reproducible fluidization of the top particle layer
was observed during the outflow of the upper half of the hourglass. These particles were fluidized by the
Keywords:
CFD–DEM simulation
displacement flow initiated by falling particles in the completely closed container. This complex problem
Flow behavior was simulated by coupling computational fluid dynamics (CFD) with the discrete element method (DEM).
Lunar regolith JSC-1A A CFD–DEM simulation with 1 million particles was performed. Because billions of particles are present
Experiment in the actual system, the use of a coarse graining approach was required. In addition, high-speed camera
measurements were used to determine the velocities of individual particles to validate the simulation.
The fluidization effect was successfully simulated using the coupled method.
© 2018 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of
Sciences. Published by Elsevier B.V. All rights reserved.

Introduction sis methods, the sample is placed into a container or small oven,
where it is heated to investigate the released gases. After the heat-
In recent decades, the exploration of celestial bodies has rapidly ing process, it is critical to ensure that the contents of the sample
developed. However, missions to the Moon and Mars have proven container are completely discharged. If material remains in the con-
challenging for engineers and scientists because the material prop- tainer, further tests may not be feasible. In the worst case scenario,
erties are often completely different owing to the different physical a complete blockage of the container or device could occur, pre-
environments. Corrias, Licheri, Orrù, and Cao (2012) described the venting any further investigation and leading to an interruption of
optimization of the in situ self-propagating high-temperature pro- the mission. The discharge of the excavator bucket as well as the
cess for fabricating lunar construction materials, and He (2010) filling and discharge of the container are critical steps in ensuring
characterized lunar regolith simulants. However, further studies the safe functionality of the rover.
pertaining to bulk material characteristics on other celestial bod- Appropriate container design strongly depends on the bulk
ies are scarce in the scientific literature. Autonomous aerospace material behavior. In this study, a substitute material with the same
rover and robots such as the Mars Science Laboratory (“Curios- properties as the regolith on the Moon was employed. Astronauts
ity”) are used to explore the atmosphere, radiation, and geology of involved in the Apollo Program collected soil and regolith on the
extraterrestrial celestial bodies. All rovers are equipped with vari- surface of the Moon and brought it back to Earth. This material
ous analytical instruments. Small excavator booms are often used was characterized by a very rough appearance because the grains
to extract a sample of rocks and regolith, which make up the upper were not rounded by erosion as they would be on earth. Only
dust layer, for the analysis of geological samples. For some analy- impacts from asteroids, radiation, and temperature formed their
shape. Consequently, the material behavior differed from that of
sand found on Earth with the same grain size. NASA developed the
material “JSC-1A”, which has the same bulk density, friction, grain
∗ Corresponding author.
size distribution, and shape as lunar regolith (McKay, Carter, Boles,
E-mail address: janiga@ovgu.de (G. Janiga).
1
The first two authors contributed equally to this work. Allen, & Allton, 1994) for this reason. This material has been used for

https://doi.org/10.1016/j.partic.2017.12.003
1674-2001/© 2018 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of Sciences. Published by Elsevier B.V. All rights reserved.
H. Otto et al. / Particuology 40 (2018) 34–43 35

further investigations. On Earth and several larger celestial bodies, 2. the particles affect the fluid flow,
the atmosphere cannot be neglected when studying the dynamic 3. the particle disturbance in the fluid locally affects the motion of
behavior of fine powders and bulk material. Reiss, Hager, Hoehn, other particles (an effect known as particle swarm), and
Rott, and Walter (2014) investigated the bulk material discharge of 4. the particles can collide with each other and the walls.
lunar regolith in an hourglass during parabolic flights. Fluidization
was observed in the top particle layer and occurred under varying The advantages of CFD–DEM simulations over other multi-
atmospheric pressure and gravity. During the fluidization, the air- phase models, e.g., DPM simulations, are the direct consideration
flow decreased the inner pressure of the bulk material and reduced of particle–particle interactions and the flow modifications due to
the density of the particle phase, causing the particles to move away the swarm effect. In this section, the computational details of the
from each other. The inner friction of the bulk material was reduced CFD–DEM simulation are briefly described. Further details can be
because of the diluted particle bed, leading to an increased air flow found in, Chen, Zhong, Zhou, Jin, and Sun (2012), Zhu, Zhou, Yang,
rate. The fluidization was reproducible; however, the point in time and Yu (2007), and Saidi, Tabrizi, Grace, and Lim (2015).
at which particles were fluidized varied as a stochastic value in the
experiments. At low pressure, the fluidization effect was no longer CFD–DEM
observed. The experimental tests conducted by Reiss et al. (2014)
under varying gravitational conditions were extremely difficult and The continuity equation and the Navier–Stokes equations
expensive, and different gravitational effects can be easily consid- (momentum conservation) in modified form are the basis for calcu-
ered using simulations. Hence, several research groups, including lating the flow of an isothermal incompressible fluid with particles:
Kuang, LaMarche, Curtis, and Yu (2013) and Metzger, Smith, and
Lane (2011), are working to develop a simulation model of the ∂ (f εf )
+ ∇ · (f εf uf ) = 0, (1)
flow behavior of lunar regolith. The aim of this study was to adapt ∂t
the previously presented methods and use a coupled simulation
∂ (f εf uf )
to reproduce the observed effects of the bulk material flow in a + ∇ · (f εf uf uf ) = −εf ∇ p + Rf,p + ∇ · (εf f ) + f, (2)
simulation.
∂t
This study focuses on the described fluidization behavior that where the index f denotes the fluid, εf represents the volume frac-
occurs owing to particles falling in a closed container and thus tion occupied by the fluid, f represents the density of the fluid,
displacing the particles in the upper layer. The bulk material and uf represents the velocity of the fluid, and  f represents the stress
air flow were simulated by coupling the discrete element method tensor of the fluid. Rf,p represents the semi-implicit momentum
(DEM) and computational fluid dynamics (CFD). This problem has a exchange between the fluid and particulate phase, which is cal-
high numerical complexity because a dynamic flow boundary con- culated for each cell and assembled from the particle-based drag
dition (e.g., the inlet–outlet condition) is not defined from outside forces. Finally, f denotes the explicit momentum exchange term.
the system; the flow velocity is directed against the particle flow, In practice, if a large number of particles is simulated (as in the
and also driven by the displacement of the particle volume. The present case), a so-called non-resolved explicit approach is used (Di
particle and air flows are strongly coupled; therefore, the solver Felice, 1994). In this case, the particles are smaller than the com-
must adapt to a very stiff system. The flow rates that occur are putational grid used for the finite-volume method. Consequently,
only dependent on the movement of the falling particles. These the particles are not resolved in the CFD simulation. The compu-
constraints exceed the limits of most other conventional multi- tation only considers their interaction with the fluid phase (e.g.,
phase simulation methods, e.g., the discrete particle method (DPM). momentum exchange and the resulting porosity) (Goniva, Kloss,
CFD–DEM simulation is a promising method to simulate both the Deen, Kuipers, & Pirkers, 2012).
interaction between particles and their bidirectional effect on the The drag force in the examined multi-particle-fluid interaction
fluid phase. Recently, CFD–DEM simulations have been increasingly system is considered using the drag model presented by Di Felice
applied to engineering problems thanks to an increase in compu- (1994). The force balance for a single particle in a fluid environment
tational power (Brosh, Kalman, Levy, Peyron, & Ricard, 2014; Chen is
et al., 2015; Salikov et al., 2015). DEM is a meshless method in which dp
the particles are considered free rigid bodies. Contact between the Fd − V = Vp g − Fc , (3)
dz
particles is considered by force vectors acting on the center of
mass. CFD considers the fluid as a continuous phase by meshing the where V, p , g, Fd , and Fc are the volume of the particle, density
computational domain. The discretized fluid dynamic conservation of the particle, gravity, drag force, and contact forces, respectively.
equations are computed for each grid cell. The term − dp
dz
is the pressure gradient force, where p represents the
In the present work, a CFD–DEM simulation with one million fluid pressure.
particles was performed. In the real system, billions of particles In the absence of other particles, the drag force Fd0 on a single
are present; thus, coarse graining needed to be applied. In addi- particle is
tion, high-speed camera measurements were used to determine the f u2rel ␲d2
velocities of individual particles to validate the simulation method. Fd0 = Cd0 , (4)
2 4
where Cd0 , f , urel , and d are the drag coefficient, fluid density,
Model, calibration, and computational details relative velocity, and particle diameter, respectively. For spheri-
cal particles, established correlations for the drag coefficient are
Coupled CFD–DEM simulations constitute relatively new meth- available such as that proposed by Dallavalle (1948):
ods for the simulation of multiphase flow (Harris et al., 1996).
 4.8
2
Instead of pure DEM simulations (Müller & Tomas, 2014), the effect
Cd0 = 0.63 + , (5)
of the fluid phase is considered (Liu, Wen, Liu, Liu, & Shao, 2015). Re0.5
CFD–DEM computations are so-called four-way coupling methods
which is an empirical function dependent upon the Reynolds num-
(Crowe, 2005), meaning that
ber, Re.
According to Di Felice (1994), the effect of neighboring particles
1. the fluid (continuous phase) affects the particles, on the drag force experienced by a single particle may be considered
36 H. Otto et al. / Particuology 40 (2018) 34–43

Fig. 1. Process of agglutination after micrometeorite impacts that results in vaporization in vacuum and condensation on the surrounding surface particles according to
Taylor (1988).

Fig. 2. Comparison of standard river sand (left) and lunar regolith (right) under a microscope. The lunar regolith has obvious sharp edges, which may be interpreted as a high
resistance to rolling.

only as a function of the local volumetric particle concentration, or Di Felice (1994) evaluated the experimental data of Happel and
void fraction, ε: Epstein (1954), Rumpf and Gupte (1971), and Ergun (1952) and
observed the following correlation for g(ε):
Fd = Fd0 f (ε) . (6)
g (ε) = ε−ˇ , (12)
In dense particle systems in which particle–particle contacts affect
the general flow behavior, the action of the drag force between with
a particle and the fluid results in energy dissipation. For dense  
particle systems, numerous data for the piezometric head loss (1.5 − x)2
ˇ = 3.7 − 0.65 exp − , (13)
P are available. The left side of Eq. (3) represents the mutual 2
fluid–particle interaction force that acts on a particle and (in the
which has been successfully applied to the simulation of fluidized
opposite direction) on the fluid. Thus, the momentum equation
beds, e.g., in Zhou, Kuang, Chu, and Yu (2010).
for the fluid phase contained in a unit volume of suspension
(fluid + solid) may be written as
 dp
 1−ε dp
Calibration of DEM parameters
−εg − Fd − V − = 0. (7)
dz V dz Certain relevant parameters for the DEM simulations (e.g.,
The piezometric pressure gradient is defined as Coulomb and rolling friction) cannot be directly obtained by
measurements. To observe the correct material behavior, these
dP dp
= + g. (8) parameters can only be obtained by performing calibration simu-
dz dz
lations. As such, the relevant parameters are varied systematically
Transforming Eq. (3) using Eq. (8) results in until the behavior of the bulk material is identical to the simu-
␲d3
 ε  P lated behavior of the bulk material. Another important part of this
Fd = , (9) method is the application of coarse graining, which is used because
6 1−ε L
the real particle size is very small and hence the simulation of the
where real number of particles would not be feasible. There are several
dp coarse graining methods. The approach applied here scales up par-
P = −L (10)
dz ticles to simulate the mass and properties of many particles with
and L is the bed height (in this case, the height of the hourglass). one single particle, as described by Sakai, Takahashi, Pain, Latham,
Inserting Eq. (9) into Eq. (6) along with the expression for Fd0 (Eq. and Xiang (2012), and Di Maio and Di Renzo (2013). The parameters
(4)) results in pertaining to friction can also not be obtained by a direct measure-
ment because of the scaling as well as the heterogeneous grain
4 ε dP
f (ε) = . (11) shape of the particles in the bulk material. In the CFD–DEM sim-
3Cd0 1 − ε u2 L
rel ulation, it is necessary to consider the grain size by adjusting the
H. Otto et al. / Particuology 40 (2018) 34–43 37

Fig. 3. Particle size distribution function Q3 and particle size distribution density q3
of lunar regolith JSC-1A measured using a CAMSIZER.

porosity because the fluid dynamic resistance for agglomerations


or clusters of particles is higher than the drag coefficient of a single
particle. Other parameters such as the bulk density and the friction
between the wall and bulk material can be directly extracted from
experimental data.

Characterization and appearance of bulk material


The lunar regolith JSC-1A investigated here is a synthetic prod-
uct that approximates the properties of the upper fine-grained
layer on the Moon. Examples of regolith and soil were collected
by astronauts of the Apollo missions and brought back to Earth.
On Earth, different types of simulants were created to achieve the
same bulk chemical and physical properties as the original sam-
ple, reproducing the flow behavior, grain size, and shape. JSC-1A is
the second generation of JSC-1, which was mined from the Merriam
Crater (McKay et al., 1994). To understand the behavior of the Moon
regolith, it is important to be cognizant of the processes that act on
the Moon surface. The main formative processes on the Moon are
micrometeorite impacts (Taylor & Meek, 2005). Larger rocks break
into small pieces as the result of such impacts, and the thermal
energy of the impact causes smaller particles to melt together.
The melting process produces a large amount of glass, leading to
a fine glass layer (Hu & Taylor, 1977). The destruction leads to small
grain sizes, and agglutination causes the particles to melt together.
This process is exemplified in Fig. 1.
A regolith containing a large amount of glass can be described as
a rough surface with hard and sharp edges. Because of these edges,
the frictional resistance between particles is expected to be high. A
comparison of washed river sand and lunar regolith simulant JSC-
1A is presented in Fig. 2. The sharp edges of the glass particles of
JSC-1A are clearly visible in Fig. 2 (right), whereas all the sharp edges
and corners of the river sand in Fig. 2 (left) have been rounded by
washing. Fig. 5. Upper: Angle of repose test of the static experiment for the DEM calibra-
The particle size distribution Q3 and particle size distribution tion and lower: angle of repose test from the dynamic experiment for the DEM
density q3 of lunar regolith JSC-1A is depicted in Fig. 3. The A in calibration.

Fig. 4. Angle of repose experiment with lunar regolith.


38 H. Otto et al. / Particuology 40 (2018) 34–43

Fig. 6. Experimental setup for the discharge flow behavior (left) and resulting angle of repose (right) for the discharge of 5 g of lunar regolith.

JSC-1A represents a grain diameter of less than 1 mm for all the


particles. The measurement of Fig. 3 confirms that all the particles
were smaller than 1 mm.

Angle of repose
As one of the important tests for calibration of the simulation,
the angle of repose (AOR) is shown in Fig. 4. The bulk-filled cylinder
with a diameter of 100 mm and volume of approximately 1.5 dm3
was moved upwards with a defined velocity of 8 mm/s. Because of
the continuous discharge at the bottom of the cylinder, a character-
istic pile was formed, the angle of which could be determined by
optical measurements. For the lunar regolith JSC-1A, the average
AOR was 37◦ .
To achieve the same AOR in the simulation, the estimated angle
must be calibrated. In the upper part of Fig. 5, the AOR values
from simulations with a systematic variation of particle friction
and rolling friction are shown. The variation in particle friction
and rolling friction led to contour lines with different frictional val-
ues but the same AOR. The procedure is equivalent to the method
presented by Wensrich and Katterfeld (2012).
The target angle of 37◦ led to an infinite count of parameter com-
binations of particle friction and rolling friction. These parameters
can be identified by following the contour line of 37.1◦ . From the
grain shape analysis in Fig. 2, it is apparent that the rolling friction
must be higher than the Coulomb friction because of the very irreg-
ularly shaped particles. Based on this information, the parameter
set from the lower part of Fig. 5 was reduced to the lower half of
the diagram. To identify the correct parameter set, a second exper-
iment was needed. Therefore, the regolith discharge process from
a funnel was used to measure the dynamic AOR. The experiment is
shown in Fig. 6.
The funnel was placed 170 mm above the plate and filled with
5 g of bulk material from 240 mm above ground level. Within 3 s,
all of the JSC-1A passed through the funnel, which channeled the
flow. The average value of the resulting AOR was 34.5◦ . Thus, the
dynamic angle was lower than the static AOR. The experiment was
performed using a set of 100 parameter combinations, which led
to the diagram shown in the lower part of Fig. 5.
To find the best combination that fulfilled the requirements of
both experiments, the intersection of the upper and lower parts of
Fig. 5 was taken. Therefore, a range of 35.9◦ –37.9◦ for the static AOR
and a range of 34.1◦ –34.9◦ for the dynamic AOR were selected. Both
overlaid areas are shown on the upper side of Fig. 7. The intersection
of the areas is shown on the lower side of Fig. 7.
The calibration simulations resulted in a sliding friction for
the particles p of 0.3 and a rolling friction r of 0.7. This result
Fig. 7. Overlay of the ranges of 35.9◦ –37.9◦ for the static AOR and 34.1◦ –34.9◦ for
represents the highest possible particle rolling friction for the inter- the dynamic AOR (upper) and intersection areas of both areas (lower).
section area, which is forced by the irregular particle shape, as
H. Otto et al. / Particuology 40 (2018) 34–43 39

Fig. 8. Comparison of simulated and experimental AOR values.

described above. The resulting AOR is compared with the exper-


imental measurement in Fig. 8.

Wall friction
The wall friction value is one of the simulation parameters
that can be obtained by direct measurements using two separate
methods. The first measurement is performed by simply increas-
ing the inclination of a plane. This method determines the value of
the static friction because only the value of the first movement is
recorded. For the simulation of continuous bulk solid flow, the slid-
ing friction is more important. Its value is measured using a Jenike
shear cell (Jenike, 1970).

Test under reduced gravity


The calibration of the simulation parameters was compared
with the experimental investigations, which could be performed
with limited effort under the Earth’s conditions. Reiss et al. (2014)
presented experimental studies of an hourglass test under reduced
atmosphere and gravity at the Technical University of Munich. For
these investigations, a suitable geometry, approaching the form
of the sample reservoir bin or hopper, was filled with the lunar
regolith. The flow mechanisms were investigated both under low
pressure and reduced gravity as well as under the conditions here
on Earth. Discharging the bulk material from one side to the other
was achieved by flipping the geometry upside down, which intro-
duced a large amount of energy into the bulk material, leading to an
obvious increase in the flow rate. The geometry in this study had a
slider valve in the narrow channel for this reason. In the presented
study, the flow behavior of the lunar regolith was examined under
the Earth’s gravity and atmospheric pressure.

Computational details

The application of powerful parallel computers is critical for


CFD–DEM simulations that consider a large number of particles
Fig. 9. Computational grid for considered configuration.
because computational effort is very high for these simula-
tions (Almohammed, Alobaid, Breuer, & Epple, 2014; Wu, Ayenia,
Berrouk, & Nandakumara, 2014). In this work, a complete three- Group), 2016), was used for the CFD simulation. The solver uses the
dimensional model of the test configuration filled with 1 million PISO algorithm (Pressure Implicit with Splitting of Operator) for the
particles was simulated using an in-house Linux cluster with 128 velocity–pressure coupling. It is a freely available CFD solver used
processors (2100 GHz/processor). by the coupled software (CFDEMcoupling).

CFD simulation
Fluid dynamic weight factor
The three-dimensional geometry and applied block-structured
Because of the application of a coarse graining method, the par-
computational mesh of the test configuration are shown in Fig. 9.
ticle volume occupied in the CFD domain had to be adjusted using
The regions relevant for the bulk material flow behavior were
a weight factor fw . The weight factor affects the volume as follows:
resolved using a finer mesh.
The fluid was considered to be incompressible (air). The tran- 4 3
Vp = r fw . (14)
sient solver pisoFoam, a solver in OpenFOAM (OpenCFD Ltd. (ESI 3
40 H. Otto et al. / Particuology 40 (2018) 34–43

Table 1
CFD–DEM simulation parameters.

Particle sizes (␮m) 100, 150, 200


(Mass fraction) (1/3), (1/6), (1/2)
Particle density (kg/m3 ) 2997.68 (lunar regolith)
Number of particles 1,000,000
DEM time step (␮s) 1
Particle sliding friction, p 0.7
Particle rolling friction, r 0.3
Particle wall friction 0.465
Poisson number 0.3
Young’s modulus (Pa) 5,000,000
Gravity (m/s2 ) 9.81
Fluid density (kg/m3 ) 1.293 (air)
Kinematic viscosity (m2 /s) 14.25 · 10− 6 (air)
Velocity at wall (m/s) 0 (no slip)
Flow model laminar
Number of computational cells 33,936
Fig. 10. Information exchange between the CFD and DEM solvers.
CFD time step (␮s) 10
Force models Di Felice (1994), Archimedes
Coupling interval (␮s) 10
Total physical simulation time (s) 0.6 described by Goniva et al. (2012). Fig. 10 illustrates the information
Computing time (128 processors) ≈9 d 10 h 39 min exchange between the solvers.

Results and discussion


For the approach presented here, the weight factor was set to 5.
The effect of fluidization could only be observed with a factor of 5 First, the apparent flow velocities of the bulk material and fluid
or higher. With the factor used here, an intermittent fluidization of phase were compared. The solid phase, color-coded according to
particles is visible, which is similar to the experimental measure- the particle velocity, is shown in Fig. 11 (left), and the streamlines
ments. It is assumed that a factor much higher than 5 would lead of the air are shown in Fig. 11 (right). An animation showing the
to continuous particle fluidization. particle motion is available as Supplementary material. For both
the solid and gas phase, approximately the same velocities were
CFD–DEM simulation observed. The image of the streamlines in Fig. 11 (right) clearly
The ratio of the particle volume to the CFD grid cell volume demonstrates that the particle fluidization occurred due to inter-
should be between approximately 3 and 20 according to Goniva mittent air pockets, followed by pressure compensation.
et al. (2012), which is fulfilled with a ratio of 12 to 28 in the
regions that are relevant for bulk material flow behavior. CFD–DEM Comparison of CFD–DEM simulation with experimental results
coupling typically exchanges information every 10–100 DEM time
steps. A coupling interval of 10 was used here. The simulation To validate the results of the CFD–DEM simulation with exper-
parameters for the CFD–DEM simulation are summarized in Table 1. imental results, the particle flow behavior was recorded using a
The CFD–DEM simulations were performed using the coupled high-speed camera system. Direct measurement of the particle
open-source solver CFDEM-coupling (Goniva & Kloss, 2016). This velocities in the experiment was not possible via laser measure-
software contains a solver that was developed by combining the ment or particle image velocimetry (PIV) because the laser was
C++-based open-source software environments OpenFOAM (for blocked by the particles in the observed volume and the laser beam
CFD) (OpenCFD Ltd. (ESI Group), 2016) and LIGGGHTS (for DEM). could not penetrate into the dense particle stream. In addition,
The solver performed the fluid (CFD) and particle (DEM) computa- the inner movement of the bulk material with increasing depth
tions by applying two separate codes. The interaction was realized proved to be too high for PIV (Hagemeier, Roloff, Bück, & Tsotsas,
by exchanging relevant information with a predefined time step, as 2015). In addition, the low contrast within the bulk material caused

Fig. 11. Solid phase color-coded by particle velocity (left) and streamlines of the air (right).
H. Otto et al. / Particuology 40 (2018) 34–43 41

Fig. 12. Comparison between CFD–DEM simulation (top) and experimental measurements (bottom) at three different time steps.

problems for optical measurement, as single particles could not be


resolved and were difficult to differentiate. Single particle track-
ing constituted an applicable alternative. In this case, styrofoam
spheres were used as tracer particles.
A comparison of the CFD–DEM simulation and experimental
measurements for a fluidization example is shown in Fig. 12 for
three different time steps. The particle fluidization effect was visible
for both the simulation and experiment; however, there were fewer
particles in the simulation, as explained in Section “Calibration of
DEM parameters”.
Data analysis was performed via manual reconstruction of the
particle velocities. Three arbitrarily selected particles that were
transported upward with the flow are shown in Fig. 13. The mea-
sured velocities of the three selected particles ranged from 0.2 to
0.7 m/s. The average velocity plotted in Fig. 13 on the lower side
was 0.353 m/s. The velocity of the three arbitrary selected particles
that were transported upward during the entire simulation was
also determined for the CFD–DEM simulation shown in Fig. 13 on
the upper side.
The fluidization in the experiment began with an open space
within the bulk material and culminated in an explosive pressure
compensation by the air, which was responsible for the resulting
height of three arbitrarily selected particles from the outer surface
of the fluidized particles. The particles were accelerated upwards
from t = 0.31 s until t = 0.37 s. To compare the experiment with the
simulation, the average velocity in the time span from t = 0.31 s
Fig. 13. Variations of tracked particle velocities with time in the simulation (upper)
to t = 0.33 s was calculated. For the experiment, the average value
and experiment (lower).
was vexp = 0.41 m/s, and for the simulation the average value was
vsim = 0.33 m/s.
42 H. Otto et al. / Particuology 40 (2018) 34–43

Fig. 14. Fluidization with particles color-coded according to depth to visualize a single blister and CFD-DEM simulation at three different time steps. In the lower series of
images, the fluidization of JSC-1A with styrofoam particles to aid in visualization is shown to help visualize the fluidization effect in the considered configuration.

Because even manual tracking was very difficult because of vacuum or reduced gravity, is also required. It can be assumed that
movements within the axis of view, only the point of fluidization it is also possible to simulate bulk behavior for missions to Mars.
and main upward movement were tracked. The fluctuation of the Therefore, it is necessary to calibrate a new model for a Marsian
measured values can be explained by the manual tracking. One simulant to consider the CO2 atmosphere and difference in gravity.
tracer particle was only resolved by a few pixels during the high-
speed camera measurement. The movement of the particle was also Acknowledgements
only stepwise resolved.
Fig. 14 clearly shows the depth information; the particles are The financial support of the German Research Foundation (DFG)
color-coded according to their distance to the wall along the y-axis within the Research Program SPP 1679 “Dynamische Simula-
(i.e., the depth). The same visualization of fluidization of particles tion vernetzter Feststoffprozesse” is gratefully acknowledged. Ms.
is shown for a single experiment in the lower part of Fig. 14. The Kristin Kerst would also like to thank the German Federal State of
moment of fluidization and the number of air pockets could not “Saxony-Anhalt” for the financial support of her work. The sup-
be predicted by a single simulation. Neither the experiment nor port of Dr. Thomas Hagemeier for the laser measurements is also
the simulation could predict the exact beginning of the fluidization gratefully acknowledged. Furthermore, the support of Dominique
because random effects such as particle arrangement and turbu- Dauven from Universitätsfrauenklinik Magdeburg with the micro-
lence affect the particle discharge and fluidization. scopic examinations is gratefully acknowledged.

Conclusions and outlook Appendix A. Supplementary data

In conclusion, CFD–DEM simulations result in the same inter- Supplementary data associated with this article can be found, in
mittent fluidization effect as that observed in experiments in an the online version, at https://doi.org/10.1016/j.partic.2017.12.003.
hourglass, which is a great success of this work because a dynamic
flow boundary condition (e.g., the inlet–outlet condition) was not References
defined from outside the system. The falling particles displace the
air at the bottom of the hourglass geometry, leading to a rapid Almohammed, N., Alobaid, F., Breuer, M., & Epple, B. (2014). A comparative study on
acceleration of the upwards moving air flow. The rapidly chang- the influence of the gas flow rate on the hydrodynamics of a gas-solid spouted
fluidized bed using Euler-Euler and Euler-Lagrange/DEM models. Powder Tech-
ing dynamic conditions of the fluidization possess a high numerical nology, 264, 343–364.
complexity. The simulated and experimentally determined parti- Brosh, T., Kalman, H., Levy, A., Peyron, I., & Ricard, F. (2014). DEM-CFD simulation of
cle velocities were in a comparatively good agreement; however, particle comminution in jet-mill. Powder Technology, 257, 104–112.
Chen, J., Wang, Y., Li, X., He, R., Han, S., & Chen, Y. (2015). Erosion prediction of liquid-
the moment of fluidization and number of air pockets could not be
particle two-phase flow in pipeline elbows via CFD-DEM coupling method.
predicted by a single simulation. Powder Technology, 275, 182–187.
The outlook for work in this area involves improving the DEM Chen, X., Zhong, W., Zhou, X., Jin, B., & Sun, B. (2012). CFD-DEM simulation of par-
ticle transport and deposition in pulmonary airway. Powder Technology, 228,
model to simulate compacting effects, e.g., bridge building, using
309–318.
the same coupled simulation model. If this objective were achieved, Corrias, G., Licheri, R., Orrù, R., & Cao, G. (2012). Optimization of the self-propagating
a validated model for simulations under varying conditions could high-temperature process for the fabrication in situ of lunar construction mate-
be assumed for both the particle and fluid phase. Ensuring the rials. Chemical Engineering Journal, 193–194, 410–421.
Crowe, C. T. (2005). Multiphase flow handbook. Boca Raton: Taylor and Francis., ISBN
model calibrated on Earth may be used for simulations for which 9780849312809
conditions present on the Moon are considered, such as complete Dallavalle, J. M. (1948). Micrometrics. New York: Pitman Publ. Corp.
H. Otto et al. / Particuology 40 (2018) 34–43 43

Di Felice, R. (1994). The voidage function for fluid-particle interaction systems. Inter- Metzger, P. T., Smith, J., & Lane, J. E. (2011). Phenomenology of soil erosion due
national Journal of Multiphase Flow, 20(1), 153–159. to rocket exhaust on the Moon and the Mauna Kea lunar test site. Jour-
Di Maio, F. P., & Di Renzo, A. (2013). Verification of scaling criteria for bubbling nal of Geophysical Research: Planets, 116, E06005. http://dx.doi.org/10.1029/
fluidized beds by DEM-CFD simulation. Powder Technology, 248, 161–171. 2010JE003745
Ergun, S. (1952). Fluid flow through packed columns. Chemical Engineering and Pro- Müller, P., & Tomas, J. (2014). Simulation and calibration of granules using the dis-
cessing, 48, 89–94. crete element method. Particuology, 12, 40–43.
Goniva, C., & Kloss, C. (2016). CFDEM coupling. http://www.cfdem.com. Reiss, P., Hager, P., Hoehn, A., Rott, M., & Walter, U. (2014). Flowability of lunar
Goniva, C., Kloss, C., Deen, N. G., Kuipers, J. A. M., & Pirkers, S. (2012). Influence regolith simulants under reduced gravity and vacuum in hopper-based convey-
of rolling friction on single spout fluidized bed simulation. Particuology, 10, ing devices. Journal of Terramechanics, 55, 61–72.
582–591. Rumpf, H. C. H., & Gupte, A. R. (1971). Einflüsse der Porosität und Korngrößen-
Hagemeier, T., Roloff, C., Bück, A., & Tsotsas, E. (2015). Estimation of particle dynam- verteilung im Widerstandsgesetz der Porenströmung. Chemie Ingenieur Technik,
ics in 2d-fluidized beds using particle tracking velocimetry. Particuology, 22, 43(6), 367–375.
39–51. Saidi, M., Tabrizi, H. B., Grace, J. R., & Lim, C. J. (2015). Hydrodynamic investigation of
Happel, J., & Epstein, N. (1954). Cubical assemblages of uniform spheres. Industrial gas-solid flow in rectangular spout-fluid bed using CFD-DEM modeling. Powder
and Engineering Chemistry, 46, 1187–1194. Technology, 284, 355–364.
Harris, C. K., Roekaerts, D., Rosendal, F. J. J., Buitendijk, F. G. J., Daskopoulos, P., Vreene- Sakai, M., Takahashi, H., Pain, C. C., Latham, J.-P., & Xiang, J. (2012). Study on a
goor, A. J. N., et al. (1996). Computational fluid dynamics for chemical reactor large-scale discrete element model for fine particles in a fluidized bed. Advanced
engineering. Chemical Engineering Science, 51, 1569–1594. Powder Technology, 23, 673–681.
He, C. (2010). Geotechnical characterization of lunar regolith simulants. Doctoral dis- Salikov, V., Antonyuk, S., Heinrich, S., Sutkar, V. S., Deen, N. G., & Kuipers, J. (2015).
sertation. Cleveland, Ohio, USA: Case Western Reserve University. Characterization and CFD-DEM modelling of a prismatic spouted bed. Powder
Hu, H. N., & Taylor, L. A. (1977). Lack of chemical fractionation in major and minor Technology, 270, 622–636.
elements during agglutinate formation. Lunar and Planetary Science Conference Taylor, L. A. (1988). Generation of native Fe in lunar soil. In Engineering, Constructions,
Proceedings (Vol. 8), 3645–3656. Operations, and Business in Space. pp. 67–77.
Jenike, A. (1970). Storage and flow of solids. Doctoral dissertation. Salt Lake City, USA: Taylor, L. A., & Meek, T. T. (2005). Microwave sintering of lunar soil: Properties,
University of Utah. theory: and practice. Journal of Aerospace Engineering, 18, 188–196.
Kuang, S., LaMarche, C., Curtis, J., & Yu, A. B. (2013). Discrete particle simulation of Wensrich, C. M., & Katterfeld, A. (2012). Rolling friction as a technique for modeling
jet-induced cratering of a granular bed. Powder Technology, 239, 319–336. particle shape in DEM. Powder Technology, 217, 409–417.
Liu, M., Wen, Y., Liu, R., Liu, B., & Shao, Y. (2015). Investigation of fluidization behavior Wu, C. L., Ayenia, O., Berrouk, A. S., & Nandakumara, K. (2014). Parallel algorithms
of high density particle in spouted bed using CFD-DEM coupling method. Powder for CFD-DEM modeling of dense particulate flows. Chemical Engineering Science,
Technology, 280, 72–82. 118, 221–244.
OpenCFD Ltd (ESI Group) (2016). OpenFOAM. http://www.openfoam.com. Zhou, Z. Y., Kuang, S. B., Chu, K. W., & Yu, A. B. (2010). Discrete particle simulation of
McKay, D. S., Carter, J. L., Boles, W. W., Allen, C. C., & Allton, J. H. (1994). JSC-1: A new particle-fluid flow: Model formulations and their applicability. Journal of Fluid
lunar soil simulant. In Engineering, Construction, and Operations in Space IV. pp. Mechanics, 661, 482–510.
857–866. Zhu, H. P., Zhou, Z. Y., Yang, R. Y., & Yu, A. B. (2007). Discrete particle simulation
of particulate systems: Theoretical developments. Chemical Engineering Science,
62, 3378–3396.

You might also like