Microsoft Word - 42landucci.docx
CHEMICAL ENGINEERING TRANSACTIONS
VOL. 82, 2020
A publication of
The Italian Association
of Chemical Engineering
Online at www.cetjournal.it
Guest Editors: Bruno Fabiano, Valerio Cozzani, Genserik Reniers
Copyright © 2020, AIDIC Servizi S.r.l.
ISBN 978-88-95608-80-8; ISSN 2283-9216
Experimental Determination of Aluminum Burning Velocity
During Flame Propagation in a Tube
Clement Chanut*, Frederic Heymes, Pierre Lauret, Christian Lopez
Laboratoire de Génie de l’Environnement Industriel, IMT Mines Ales, Ales, France
clement.chanut@mines-ales.fr
Modeling the consequences of dust explosions is a challenging research topic. A key parameter of these
models is the burning velocity, which represents the consumption rate of the reactants by the flame front.
Especially, a relation between burning velocity and RMS (root-mean square) air velocity fluctuations has to be
implemented in such models; RMS velocity fluctuations representing the turbulence of the fresh air flow in
front of the flame front.This paper focuses on the experimental determination of this relation in the case of dust
flames propagating in a tube. The most commonly used method is the “open-tube method”, which assumes a
constant thermal expansion coefficient and the estimation of a 3D flame surface. These assumptions are
discussed, and a new method is proposed, based on the measurement of the fresh flow velocity in front of the
flame front. The corresponding optical setup implemented for analyzing aluminum burning velocity is then
exposed. Analysis method for obtaining burning velocity and RMS velocity fluctuations is detailed. Finally, first
results obtained are presented and commented.
1. Introduction
Dust explosion is a major concern in industries dealing with powders and dusts. The mechanisms of flame
propagation during such explosion are not well understood, especially for metallic dusts. Most studies aim to
characterize the sensitivity and severity parameters, using standardized setups like the Hartmann Tube or the
20L-sphere (Dufaud et al., 2012). Even if these parameters are useful for risk analysis in the industry, they are
not sufficient to model accurately the consequences of accidents in real conditions. To model accurately the
consequences, numerical simulations are needed. Most of the current models used for dust explosions are
based on models designed for gas explosions. One key parameter of these models is the burning velocity
(Su), which represents the consumption rate of the reactants. This burning velocity depends on the
characteristics of the turbulence of the flow. For this purpose, a key point for numerical simulations is the
determination of the relationship between the burning velocity and the RMS velocity fluctuations (VRMS) (Russo
and Di Benedetto, 2007).
This relationship can be obtained from the analysis of stabilized flames (Goroshin et al., 1996b; Julien et al.,
2017). However, for simulating real accidents, mechanisms involved during flame propagation have to be
understood. For this purpose, this relationship has also to be determined from propagating flames. Burning
velocity can be determined from experiments realized inside the 20-L standardized explosion sphere; for
example, from the evolution of the pressure during the flame propagation (Dahoe and de Goey, 2003).
However, because of the lack of optical accesses, flame propagation mechanisms cannot be analyzed from
these experiments. Tubes with optical accesses are used to visualize and analyze the flame propagation
processes. Especially from the visualization of the propagating flame, the turbulent burning velocity can be
deduced (Di Benedetto et al., 2011). The mostly used method is the “open-tube method”.
This study proposes an alternative method for estimating this turbulent burning velocity, but also for analyzing
the influence of RMS velocity fluctuations on this turbulent burning velocity. Indeed, in the first part of this
paper, the “open-tube method” is presented. Some limitations of this method are highlighted, especially for the
case of aluminum dusts. Then another method, the “direct method”, is proposed. Finally, first results obtained
are presented and analyzed.
DOI: 10.3303/CET2082035
Paper Received: 19 December 2019; Revised: 13 May 2020; Accepted: 18 August 2020
Please cite this article as: Chanut C., Heymes F., Lauret P., Lopez C., 2020, Experimental Determination of Aluminum Burning Velocity During
Flame Propagation in a Tube, Chemical Engineering Transactions, 82, 205-210 DOI:10.3303/CET2082035
205
2. Limitations of the open-tube method
2.1 Presentation of the open-tube method
The “open-tube method” has already been presented more in details in a previous paper (Chanut et al., 2018).
This method is used to determine the burning velocity from the visualization of the propagation of a flame in a
tube. Indeed, with the images obtained by direct visualization, the evolution of the position of the flame front is
obtained. The deduced propagation velocity (Vp) represents the flame speed in the laboratory referential; this
velocity depends among other on the geometry of the apparatus. This propagation velocity is different from the
turbulent burning velocity (Su) which represents the consumption rate of the reactants by the flame front. With
the “open-tube method”, this turbulent burning velocity is deduced from the propagation velocity following two
main steps. First, as the flame propagates from the bottom closed end of a tube to the open top end, thermal
expansion of the burned gases has to be taken into account. Indeed, these burned gases push the flame front
to the top; thus, the propagation velocity is higher than the burning velocity. A first burning velocity (Su1) is
deduced from the propagation velocity (Vp), where is the thermal expansion coefficient: = (1)
= ≈ (2)
Where ρu and ρb are the densities of the unburned and burned gases respectively, Tu and Tb are respectively
the temperature of the unburned gases (ambient temperature) and of the flame temperature (supposed
adiabatic). It has to be noticed that solid aluminum particles combustion will not participate to volume
expansion by solid to vapor change since reaction product is aluminum oxide (alumina) which will be
recovered in condensate state (flame temperature: 3500K (Bucher et al. 1996)). Oxygen consumption reduces
volume expansion but is neglected in Eq (1).
The second correction takes into account the shape of the flame front. Indeed, burning velocity is defined for
planar flames. However, planar flames are unstable and will not occur in experiments. Thus, the actual flame
shape will increase the burning surface, and the volume expansion. The final burning velocity (Su) is deduced
from the previous one (Su1) with the following relation:
= . ′ (3)
Where A′ is the projected flame area on a horizontal plane (corresponding to a virtual planar flame) and Af is
the actual flame surface.
2.2 Limitations of this method
With this method, burning velocity can be estimated from the visualization of the flame front propagation; from
these images, the flame position and the flame surface have to be extracted. An important point for this
analysis is the quality of these flame images. Indeed, as flames are very luminous, especially in the case of
metallic dusts, lots of images exposed in the literature are saturated. However, analysis with saturated images
can lead for example to over-estimating the burning velocity with the previous method, as previously
discussed (Chanut et al., 2018). In addition to some biases coming from the images quality, some inherent
limitations of these methods also exist, as precised above.
2.2.1 Limitations for estimating the turbulent burning velocity
The first step of the method consists on taking into account the thermal expansion of the burned gases. For
this purpose, burned gases temperature is supposed to be equal to the adiabatic flame temperature. However,
because of heat losses at the tube walls, the real burned gases temperature is less important than the
adiabatic flame temperature and will cool with time. For this purpose, the turbulent burning velocity is here
under-estimated. This flame temperature is used to calculate the thermal expansion coefficient. This latter is
considered as constant during all the flame propagation, but in the case of aluminum dust flames, a pulsating
behavior, in terms of light intensity, is observed during the flame propagation (Julien et al., 2015). Thus, this
coefficient should vary with these variations of light intensity.
The second step of the method consists on taking into account the flame surface area. However, especially for
turbulent flames, this estimation of the 3D flame surface area from 2D images of the flame front is tricky. In
most of the cases, the flame front is approximated by a parabolic shape in the plane orthogonal to the images
obtained (Khalili, 2012). With this surface approximation, the flame surface area is in general under-estimated
and thus the burning velocity is over-estimated. In a previous study, images were obtained in two orthogonal
206
planes. From these two points of view, results of burning velocity were compared. A difference of around 10%
was observed (Chanut, 2018). In any case, the error made from this approximation with parabolas is difficult
to quantify, especially in the case of turbulent flames.
2.2.2 Limitations for evaluating the influence of RMS velocity fluctuations
As mentioned in introduction, the turbulence level of the flow is expected to influence this turbulent burning
velocity. Thus, for each measurement of the turbulent burning velocity, the RMS velocity fluctuations of the
flow should be determined and specified. With this “open-tube method”, with only images of flame
propagation, the RMS velocity fluctuations of the flow cannot be determined. Several authors focused on the
influence of dispersion-induced turbulence on the burning velocity (Dufaud et al., 2012; Proust, 2017). This
dispersion-induced turbulence, also named initial turbulence, represents the level of turbulence due to the
dispersion of the dust in the tube at the moment of ignition. However, the important data is the turbulence level
just ahead the flame front while the flame is propagating in the tube. Indeed, the propagation of the flame in
the tube will increase this turbulence level; thus the burning velocity should increase (Eckhoff, 1992). This
increase of turbulence by the flame front propagation is in general called the explosion-induced turbulence.
3. Principle of an alternative method: the direct method
An alternative method is proposed to determine the turbulent burning velocity, but also to estimate the
turbulence level just ahead the flame front. This alternative method will be called the “direct method”. The
principle of this method has been previously proposed by Proust (2006). With this method, the determination
of the burning velocity (Su) is based on the measurement of two other velocities: the flame propagation
velocity (Vp, as previously introduced) and the flow velocity ahead the flame front (U). The following relation is
used to deduce the burning velocity, with the normal vector to the flame front: = . . (5)
This method is here adapted and applied with an alternative optical technique, as described hereafter.
Furthermore, a deeper analysis of the results obtained is proposed to determine the turbulence level of the
flow ahead the flame front.
4. Application of the direct method to aluminum flames
4.1 Presentation of the aluminum flame propagation tests
The experimental setup used for the study of aluminum dust flame propagation has been already described in
(Chanut et al., 2018), and is presented on Figure 1. The visualization part of the prototype is a vertical tube of
700 mm height with a 150 mm square-cross section. Aluminum dust is injected in the prototype by discharged
of two compressed air vessels through four injection tubes located in the corners of the prototype.
Figure 1: Sketch of the experimental setup
The dust cloud is then ignited by an electrical spark between two tungsten electrodes located at the bottom of
the prototype. For the experiment presented hereafter, the energy of the spark is around 15 J. From this
ignition spark, the flame propagates upward inside this tube; then the flame is conducted through an
evacuation tube. Aluminum dust with mean diameter of 5 microns is studied in the following example.
207
4.2 Presentation of the optical setup
For estimating the turbulent burning velocity with this “direct method”, two velocities have to be measured: the
flame propagation velocity and the flow velocity ahead of the flame front. A first high-speed camera is used to
visualize the flame front propagation, and thus to deduce the flame propagation velocity. A second optical
system is used to measure the flow velocity ahead the flame front, based on particle image velocimetry (PIV).
PIV is a non-intrusive optical technique used to visualize the movement of fluids. With this technique, 2D maps
of velocity on a plane of the flow are obtained. The optical setup consists on a laser generating a laser sheet
corresponding to the plane of measurement. A camera located perpendicular to this sheet records the
movement of the particles in this plane. Each image recorded by the camera is divided in different
interrogation areas (IA). Movement of particles inside each interrogation areas between two successive
images is deduced by a correlation algorithm. For PIV measurement, the flow has to be seeded with particles.
In this study, the seeding particles are the particles of the combustible dust.
However, some difficulties appear while implementing this method experimentally, because of intrinsic
characteristics of aluminum flames. Indeed, aluminum flame is a fast and very luminous phenomenon.
Moreover, a dense cloud is needed to observe flame propagation; so the laser beam is attenuated while
passing through the cloud.
For the PIV setup, the camera has to record the laser light dispersed by the cloud; not the light emitted by the
flame. For this purpose, a pulsed laser is used; to have the maximum of energy during the minimum of time.
With this pulsed laser, by decreasing the camera exposure time, the flame light recorded by the camera is
limited. A band-pass filter, centered on the wavelength of the laser is also located in front of this camera to
limit the flame light recorded. To increase the laser power, which is attenuated by the dense dust cloud, the
two pulses (from the two cavities of the pulsed laser) are emitted at the same moment.
For the results presented hereafter, the laser used is a Litron pulsed laser (30 mJ at 1 kHz, λ= 527 nm). To
have sufficient energy on the laser sheet, the frequency is fixed to 2 kHz. The corresponding “PIV camera” is a
Phantom V711 high speed camera, with a resolution of 1280x800 pixels at the frequency of 2 000 fps. This
camera is equipped with a Nikkor lens with a focal of 105 mm and an aperture of f/2.8. The “flame camera” is
a Phantom V2512 camera with the same resolution of 1280x800 pixels at this frequency of 2000 fps. This
camera is equipped with a 70-300 mm Nikkor lens, with an aperture of f/22. The exposure time of both
cameras is fixed at 1µs. For analysis purpose, these two cameras record the same field of view (height: 12
cm; width: 7,5 cm). For these first results and analysis, only the upper centered part of this field of view is
analyzed (height: 4 cm; width: 2,5 cm).
4.3 Analysis method
From the images obtained with both the “PIV camera” and the “flame camera”, the following steps allow the
determination of the turbulent burning velocity (Su) and the RMS velocity fluctuations (VRMS) in front of the
flame front. These steps are schematized on Figure 3.
1) From the images obtained with the “flame camera”, the flame front is detected.
2) From the images obtained with the “PIV camera”, the velocity field is determined. For this step, in a first
time, the images are pre-processed to homogenize the light intensity to improve the results from the PIV
analysis. Then the PIV analysis is performed with the DynamicStudio software (Dantec Dynamics). For this
analysis the size of the interrogation area (IA) is 8x32 pixels, resulting in 34 IA along the horizontal axis.
3) The mean flow velocity (Vmean) is then extracted from each vector field. This mean velocity could be
related to the global thermal expansion of the burned gases resulting in a global movement of the flow inside
the prototype to the open top end. This mean flow is calculated from the IA of the upper horizontal line. From
these IA, the mean velocity corresponds to the mean of the vertical component of the velocities.
4) This mean velocity is then subtracted to the velocity field; velocity field is now relative to the mean flow.
5) The velocity vectors just in front of the detected flame front are extracted. For each velocity vector
extracted, the corresponding normal to the flame front is deduced from the images obtained by the “flame
camera”.
6) As the turbulent burning velocity is defined in accordance to the flame front, as previously noticed in part
3, the previous velocity vector is projected relative to this normal vector; the resulting vector is now (u’,v’).
7) The mean of the v’ (velocity component normal to the flame front) is defined as the burning velocity.
8) In this step, the RMS velocity fluctuations in front of the flame front are estimated. This turbulence is
defined in a zone in front of the previous vectors representing the top of the flame front. The height of this
zone is defined to be 4 IA. Each velocity vector (u,v) is used for the calculation of the RMS velocity
fluctuations; these velocity vectors already result from the subtraction of the global mean flow velocity (step 4).
For defining turbulence, mean velocity ( and ̅) is defined; the mean is here defined as the spatial mean over
all the previous IA. Then the RMS velocity fluctuations (VRMS) are obtained from the following formula:
208
= 1 ² 1 ̅ ² (6)
Figure 2: Description of the analysis method
4.4 First results obtained
In this part, first results obtained during an aluminum dust flame propagation test are presented. For this
experiment, the aluminum dust concentration is 312 g/m3. Figure 3 presents the results obtained for the
evolution of the turbulent burning velocity in the field of view of the cameras. On this graph, the corresponding
evolution of the RMS velocity fluctuations is also plotted. As shown on this figure, the turbulent burning
velocity and the RMS velocity fluctuations are fairly constant. The calculation of the mean value of both data
gives a turbulent burning velocity of 37 cm/s for RMS velocity fluctuations of 11 cm/s.
Figure 3: Results of burning velocity and RMS velocity fluctuations
Table 1: Comparison of burning velocity data with the literature
Authors Apparatus used Burning velocity (cm/s)
(Goroshin et al., 1996b) Bunsen burner 20
(Julien et al., 2017) Counterflow burner 30-40
(Julien et al., 2015) Unconfined propagation 20
(Goroshin et al., 1996a) Open tube 30-40
Present study Open tube 37
This value is coherent with values of the literature, as exposed in Table 1 (adapted of the analysis of Julien et
al. (2017)). Indeed, this value is close to the laminar burning velocity of aluminum as the RMS velocity
fluctuations are low. Nevertheless, the main advantage of this study is the determination of the corresponding
RMS velocity fluctuations. Thus, with this test and this analysis, a first point of the relation between the
turbulent burning velocity and the RMS velocity fluctuations is obtained. Other tests, with different RMS
velocity fluctuations, are required to obtain more points and thus the final relation.
209
5. Conclusions
This study focused on the experimental determination of the burning velocity for dust flame propagating in
tubes. First, the commonly used “open-tube method” was presented. Some limitations of this method were
highlighted; mainly the estimation of the thermal expansion coefficient and the estimation of the 3D flame
surface area from 2D images. An alternative method, the “direct method”, was then presented. This latter is
based on the measurement of the fresh flow velocity just ahead the flame front. Difficulties of such
measurement in the case of aluminum propagating flames were exposed; indeed, aluminum propagating
flames are very luminous and fast phenomena. The corresponding analysis method was detailed. With this
method, turbulent burning velocity and RMS velocity fluctuations are evaluated.
First results obtained were analyzed and commented. From this first experiment, the mean turbulent burning
velocity is 37 cm/s for the corresponding RMS velocity fluctuations of 11 cm/s. This first result is in accordance
with the literature. More tests are needed to obtain the relation between turbulent burning velocity and RMS
velocity fluctuations. Additional tests will be realized in a longer prototype: with different measurement zones
along this new prototype, different RMS velocity fluctuations will be obtained and thus the corresponding
burning velocities will be deduced. Indeed, turbulence of the fresh flow increases when the flame propagates
inside the prototype because of explosion-induced turbulence. Moreover, in order to deeply study the
influence of RMS velocity fluctuations on burning velocities, obstacles will be located inside this longer tube.
Acknowledgments
The authors are grateful to IRSN (Institut de Radioprotection et Surete Nucleaire) for scientific and financial
support of this project, and to P. Slangen, L. Aprin and J.J.Lasserre for technical help to perform the tests
References
Bucher P., Yetter R.A., Dryer F.L., Parr T.P., Hanson-Parr D.M., Viceni E.P., 1996, Flames structure
measurement of single, isolated aluminium particles burning in air, Symposium on Combustion, 26, 1899-
1908.
Chanut C., 2018, Etude experimentale de la propagation du front de flamme et de la vitesse de combustion
d’une explosion de poussieres d’aluminium, PhD thesis, IMT Mines Ales, Ales, France
Chanut C., Heymes F., Lauret P., Slangen P., 2018, Visualization of aluminum dust flame propagation in a
square-section tube, Chemical Engineering Transactions, 67, 7–12.
Dahoe A.E., de Goey L.P.H., 2003, On the determination of the laminar burning velocity from closed vessel
gas explosions, Journal of Loss Prevention in the Process Industries, 16, 457–478.
Di Benedetto A., Garcia-Agreda A., Dufaud O., Khalili I., Sanchirico R., Cuervo N., Perrin L., Russo P., 2011,
Flame propagation of dust and gas-air mixtures in a tube, MCS 7 Seventh Mediterranean Combustion
Symposium, 11–13.
Dufaud O., Khalili I., Cuervo-Rodriguez N., Olcese R., Dufour A., Perrin L., Laurent A., 2012, Highlighting
the Importance of the Pyrolysis Step on Dusts Explosions, Chemical Engineering Transactions, 26, 369–374.
Eckhoff R.K., 1992, Influence of initial and explosion-induced turbulence on dust explosions in closed and
vented vessels Research at CM1, Powder Technology, 71, 181-187.
Goroshin S., Bidabadi M., Lee J.H.S., 1996a, Quenching distance of laminar flame in aluminum dust clouds,
Combustion and Flame, 105, 147–160.
Goroshin S., Fomenko I., Lee J.H.S, 1996b, Burning velocities in fuel-air rich aluminum dust clouds, Twenty-
Sixth Symposium on Combustion, The Combustion Institute, 1961-1967.
Julien P., Vickery J., Goroshin S., Frost D.L., Bergthorson J.M., 2015, Freely-propagating flames in aluminum
dust clouds, Combustion and Flame, 162, 4241–4253.
Julien P., Whiteley S., Soo M., Goroshin S., Frost D.L., Bergthorson J.M., 2017, Flame speed measurements
in aluminum suspensions using a counterflow burner, Proceedings of the Combustion Institute, 36, 2291–
2298.
Khalili I., 2012, Sensibilité, sévérité et spécificités des explosions de mélanges hybrides
gaz/vapeurs/poussières, PhD Thesis, Universite de Lorraine, Nancy, France.
Proust C., 2006, Flame propagation and combustion in some dust-air mixtures, Journal of Loss Prevention in
the Process Industries, 19, 89–100,.
Proust C., 2017, Turbulent flame propagation in large dust clouds, Journal of Loss Prevention in the Process
Industries, 49, 859-869.
Russo P., Di Benedetto A., 2007, The effect of turbulence on the theoretical evaluation of dust explosions
severity, Chemical Engineering Transactions, 11, 83–988.
210