Вы находитесь на странице: 1из 5

CFD simulations of dust dispersion in the 20 L vessel: Effect of nominal

dust concentration
V. Di Sarli
a
, P. Russo
b
, R. Sanchirico
a
, A. Di Benedetto
c,
*
a
Istituto di Ricerche sulla Combustione, CNR, Piazzale V. Tecchio 80, 80125 Napoli, Italy
b
Dipartimento di Ingegneria Chimica Materiali Ambiente, Universit di Roma La Sapienza, Via Eudossiana 18, 00184 Roma, Italy
c
Dipartimento di Ingegneria Chimica, dei Materiali e della Produzione Industriale, Universit degli Studi di Napoli Federico II, Piazzale V. Tecchio 80, 80125
Napoli, Italy
a r t i c l e i n f o
Article history:
Received 17 September 2013
Received in revised form
25 October 2013
Accepted 28 October 2013
Keywords:
Computational uid dynamics
20 L bomb
Turbulent ow eld
Dust dispersion
Dust sedimentation
Dust nominal concentration
a b s t r a c t
Measurements of ammability and explosion parameters for dust/air mixtures require uniform disper-
sion of the dust cloud inside the test vessel. In a previous work, we showed that, in the standard 20 L
sphere, the dust injection system does not allowgeneration of a uniform cloud, but rather high gradients
of dust concentration are established. In this work, we used a previously validated three-dimensional
CFD model to simulate the dust dispersion inside the 20 L sphere at different dust nominal concentra-
tions (and xed dust diameter). Results of numerical simulations have shown that, as the dust nominal
concentration is increased, sedimentation prevails and, thus, when ignition is provided, the dust is
mainly concentrated at the vessel walls.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
In order to characterize the ammability and explosion features
of dust/air and/or dustegas/air mixtures, the evaluation of explo-
sion and ammability parameters is required. Most of these pa-
rameters (minimum explosible concentration, MEC; limiting
oxygen concentration, LOC; maximum explosion pressure, P
max
;
deagration index, K
st
) are evaluated in 20 L and/or 1 m
3
vessels,
according to a standard procedure (ASTM E1226; ASTM E1515; ISO
6184/1; NFPA 68; VDI 3673). Experimental tests may be signi-
cantly affected by level of pre-ignition turbulence and quality (i.e.,
uniformity) of dust dispersion. According to the standard proce-
dure, the ignition delay time is xed in order to ensure that the
turbulence level in the 20 L vessel is comparable to the turbulence
level achieved in the 1 m
3
vessel. Quality of dust dispersion is a
relevant issue and, in the standard procedure, it is recommended to
be sure about the uniform dispersion of the dust cloud inside the
sphere.
In previous papers (Di Benedetto, Russo, Sanchirico, & Di Sarli,
2013; Di Sarli, Russo, Sanchirico, & Di Benedetto, 2013), we
simulated, through a three-dimensional CFD model, the turbulent
oweld and the dust concentration distribution established in the
20 L bomb when feeding the dust according to the standard pro-
cedure. Results have shown the presence of multiple turbulent
vortex structures generating dead volumes for the dust inside the
sphere. The dust is mainly pushed toward the walls of the sphere,
accumulating at the boundaries of the vortices. As a consequence,
the dust concentration is not uniform, but higher than the nominal
value close to the vessel walls and lower at the center of the sphere.
It has also been found that the dust dispersion is affected by the
dust size: as the dust diameter is increased (from10 mmto 250 mm),
the dust concentration distribution becomes less uniform.
The role of non-uniformity of dust dispersion is particularly
relevant when dealing with the measurements of the minimum
explosive concentration (MEC). In this work, we simulated the
evolution of the turbulent ow eld and the dust concentration
distribution in the 20 L bomb for different dust nominal concen-
trations, including lowvalues such as those comparable to the MEC.
2. Model description
In this work, we used a previously developed and validated
model (Di Benedetto et al., 2013). Briey, the model consists of the
time-averaged NaviereStokes equations (Eulerian approach)
* Corresponding author. Tel.: 39 0817682265.
E-mail addresses: almerinda.dibenedetto@unina.it, dibenede@irc.cnr.it (A. Di
Benedetto).
Contents lists available at ScienceDirect
Journal of Loss Prevention in the Process Industries
j ournal homepage: www. el sevi er. com/ l ocat e/ j l p
0950-4230/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jlp.2013.10.015
Journal of Loss Prevention in the Process Industries 27 (2014) 8e12
written in polar coordinates. For turbulence, the standard ke3
model was used (Launder & Spalding, 1972). The ow of the solid
phase was solved with the Lagrangian approach using the Discrete
Phase Model (DPM).
The interaction between the uid phase and the solid particles
was assumed as two-way since, according to the classication of
Elghobashi (1994), the operating conditions (particle volume frac-
tion, particle density and particle concentration) suggest that the
uid ow affects the particle motion and vice versa, while the
particleeparticle collisions can be neglected.
In the momentum balance equation of the DPM, the gravita-
tional force was taken into account.
The uid ow equations were discretized using a nite-volume
formulation on a three-dimensional non-uniform unstructured
grid. The spatial discretization of the model equations used rst-
order schemes for convective terms and second order schemes for
diffusion terms. First-order time integration was used to discretize
temporal derivatives with a time step of 1$10
4
s.
Parallel computations were performed by means of the segre-
gated pressure-based solver of the code ANSYS Fluent (release 14)
(www.ansys.com).
The Discrete Phase Model is described by ordinary differential
equations. For particle tracking, we used an automated scheme
which provides a mechanism to switch in an automated fashion
between numerically stable lower order schemes and higher order
schemes, which are stable only in a limited range. In situations
where the particle is far from hydrodynamic equilibrium, an ac-
curate solution can be achieved very quickly with a higher order
scheme. When the particle reaches hydrodynamic equilibrium, the
higher order schemes become inefcient and the mechanism
switches to a stable lower order scheme. We choose the Euler
integration as lower order scheme and the semi-implicit trape-
zoidal integration as higher order scheme. The particle tracking
integration time step was taken equal to the uid ow time step
(1$10
4
s).
Fig. 1 shows the scheme of the standard 20 L sphere.
The spherical explosion chamber is made of stainless steel. It is
tested to withstand a static pressure of 30 bar. At the bottom of the
vessel, a rebound nozzle is placed in order to fed the dust. The dust
is initially loaded in a 0.6 L container (SC in Fig. 1) pressurized with
21 bar of compressed air.
Computations were run for a coal with density equal to 2046 kg/
m
3
. Simulation conditions are given in Table 1.
3. Results and discussion
In Fig. 2, the maps of the turbulent kinetic energy are shown as
obtained at t 60 ms (standard time of ignition) for different
nominal dust concentrations: C
n
100 g/m
3
; C
n
250 g/m
3
and
C
n
500 g/m
3
(central section of the sphere). Such values are
typical of the MEC for several dusts (C
n
100 g/m
3
) and of the
stoichiometric concentration (C
n
500 g/m
3
).
Fig. 1. Scheme of the standard 20 L sphere.
Table 1
Simulation conditions.
Dust container volume (L) 0.6
Sphere volume (L) 20
Initial pressure of container (bar) 21
Initial pressure of sphere (bar) 0.4
Dust density (kg/m
3
) 2046
Dust diameter (mm) 10
Dust concentration (g/m
3
) 100; 250; 500
Fig. 2. Maps of turbulent kinetic energy (m
2
/s
2
) for different nominal dust concentrations.
V. Di Sarli et al. / Journal of Loss Prevention in the Process Industries 27 (2014) 8e12 9
The turbulent kinetic energy decreases with increasing dust
concentration. Furthermore, the spatial distribution of the turbu-
lent kinetic energy is not symmetric, whatever the dust concen-
tration. Indeed, the presence of dust causes asymmetries of the
oweld owing to the entrainment effect of particle sedimentation
(Di Benedetto et al., 2013).
Fig. 3 shows the uid velocity vectors colored by the dimen-
sionless dust concentration, g C/C
n
, as computed for different
nominal dust concentrations (central section of the sphere;
t 60 ms). Two main vortices can be observed. Such vortices
generate dead volumes for the dust. Coherently, the dust concen-
tration is very low in the vortex cores, while it increases in
Fig. 3. Fluid velocity vectors colored by dimensionless dust concentration, g C/C
n
, as computed for different nominal dust concentrations.
Fig. 4. Time evolution of discrete phase particles colored by dimensionless dust concentration, g C/C
n
, for nominal dust concentration equal to 100 g/m
3
.
V. Di Sarli et al. / Journal of Loss Prevention in the Process Industries 27 (2014) 8e12 10
Fig. 5. Time evolution of discrete phase particles colored by dimensionless dust concentration, g C/C
n
, for nominal dust concentration equal to 250 g/m
3
.
Fig. 6. Time evolution of discrete phase particles colored by dimensionless dust concentration, g C/C
n
, for nominal dust concentration equal to 500 g/m
3
.
V. Di Sarli et al. / Journal of Loss Prevention in the Process Industries 27 (2014) 8e12 11
correspondence to the outer zones of the vortices (dashed lines,
Fig. 3 e100 g/m
3
case). It is worth noting that the dust particles are
partially entrained by the uid ow. Furthermore, at lower nominal
dust concentrations, a better dispersion is obtained.
In order to visualize the preferential paths of the dust, we
mapped the particle tracks in the sphere at different time instances
(20, 37, 60 and 150 ms). In Figs. 4e6, the time evolution of the
discrete phase particles colored by the dimensionless dust con-
centration, g C/C
n
, is shown as obtained for nominal dust con-
centration equal to 100 g/m
3
(Fig. 4), 250 g/m
3
(Fig. 5) and 500 g/m
3
(Fig. 6). The dust has preferential paths determined by the vortical
structures generated in the sphere after the dust injection from the
container.
In the 100 g/m
3
case (Fig. 4), the dust creates a three-
dimensional cross inside the vessel at all times. In particular, at
the ignition time (t 60 ms), the cloud is not uniform, suggesting
that the ame will start propagating in a stratied ammable
mixture.
As the dust nominal concentration is increased, the preferential
paths are still present, but a high degree of particle sedimentation is
found (Figs. 5 and 6).
When increasing the dust concentration, the uid owbecomes
slower and thus less able to sustain the dust particles. In these
conditions, sedimentation prevails with respect to the uid ow
motion. The dust particles are not positioned at the boundaries of
the vortices, but their distribution appears to be more disordered
and the dust is mainly concentrated at the vessel walls.
4. Conclusions
A previously validated three-dimensional CFD model was used
to simulate the dust dispersion inside the 20 L bomb at different
nominal dust concentrations. The velocity vector maps show that
multiple turbulent vortex structures are established within the
sphere. At low nominal concentration (100 g/m
3
), the dust mainly
accumulates at the boundary of the vortices, while at higher
nominal concentration (500 g/m
3
), the dust sedimentation prevails
giving rise to highly concentrated regions close to the vessel walls.
The results obtained suggest that the standard injection system
does not allowgeneration of a uniform cloud. Thus, further studies
are required to ensure the correct evaluation of the dust explosion
severity and the minimum explosive concentration (MEC).
Acknowledgments
The authors would like to thank Vincenzo Smiglio and Luigi
Muriello for the technical support provided during the computing
activity.
References
ASTM E1226. (2010). Standard test method for explosibility of dust clouds. West
Conshohocken, PA: ASTM International.
ASTM E1515. (2007). Standard test method for minimum explosible concentration of
combustible dusts. West Conshohocken, PA: ASTM International.
Di Benedetto, A., Russo, P., Sanchirico, R., & Di Sarli, V. (2013). CFD simulations of
turbulent uid ow and dust dispersion in the 20 liter explosion vessel. AIChE
Journal, 59, 2485e2496.
Di Sarli, V., Russo, P., Sanchirico, R., & Di Benedetto, A. (2013). CFD simulations of the
effect of dust diameter on the dispersion in the 20 l bomb. Chemical Engineering
Transactions, 31, 727e732.
Elghobashi, S. (1994). On predicting particle-laden turbulent ows. Applied Scientic
Research, 52, 309e329.
ISO 6184-1. (1985). Explosion protection systems e Part 1: Determination of explosion
indices of combustible dusts in air.
Launder, B. E., & Spalding, D. B. (1972). Lectures in mathematical models of turbulence.
London: Academic Press.
NFPA 68. (2007). Guide for venting of deagrations. Quincy, MA, USA: National Fire
Protection Association.
VDI 3673. (2002). Pressure release of dust explosions, part 1. 10772 Berlin, Germany:
BeuthVerlag GmbH.
V. Di Sarli et al. / Journal of Loss Prevention in the Process Industries 27 (2014) 8e12 12

Вам также может понравиться