Flame propagation and quenching in iron dust clouds

Flame propagation and quenching in iron dust clouds

Available online at www.sciencedirect.com Proceedings of the Proceedings of the Combustion Institute 32 (2009) 1905–1912 Combustion Institute www.e...

709KB Sizes 0 Downloads 7 Views

Available online at www.sciencedirect.com

Proceedings of the

Proceedings of the Combustion Institute 32 (2009) 1905–1912

Combustion Institute www.elsevier.com/locate/proci

Flame propagation and quenching in iron dust clouds Francßois-David Tang *, Samuel Goroshin, Andrew Higgins, John Lee Mechanical Engineering Department, McGill University, 817 Sherbrooke Street West, Montreal, QC, Canada H3A 2K6

Abstract Laminar flames propagating in fuel-rich suspensions of iron dust in air were studied in a reduced-gravity environment provided by a parabolic flight aircraft. Experiments were performed with four different dusts having average particle sizes in the range 3–27 lm. Uniform dust suspensions were created inside glass tubes (ID = 48 mm, L = 70 cm) and then ignited at the open end via an electrically heated wire. Quenching distances were determined as the flames propagated through assemblies of equally spaced steel plates installed in the tubes. Flame propagation speeds in the open tubes and within the quenching plates were determined from video recordings, and emission spectra recorded by a spectrometer were used to determine flame temperature. Flame quenching distance was found to increase linearly with particle size from less than 2 mm quenching distance for the 3 lm-sized dust to 10 mm quenching distance for the 27 lm-sized dust. The flame speeds in the open tubes were found to be inversely proportional to the dust particle size, and pffiffiffi the minimum speeds observed near quenching within the plate assemblies were found to be a factor of e smaller than the flame speeds in the open tube. The experimental results were in good agreement with the predictions of a simple one-dimensional dust flame model with conductive heat loss that assumes the diffusive regime of particle combustion. Ó 2009 The Combustion Institute. Published by Elsevier Inc. All rights reserved. Keywords: Dust; Iron; Flame; Reduced gravity

1. Introduction Combustion of solid fuels suspended in gases is encountered in a number of applications, such as burning of fossil fuels (e.g. coal), safe handling of flammable dust in industrial settings, and solid-fueled propulsion. Despite their practical importance, theoretical understanding of dust flames is in an underdeveloped state compared to the well-established theory of homogeneous gas flames. A comprehensive theory of dust com-

*

Corresponding author. Fax: +1 514 398 7365. E-mail address: [email protected] (F.-D. Tang).

bustion is complicated by the enormous diversity of solid fuel properties, as well as the experimental difficulty of creating well-characterized particulate suspensions in the laboratory. The experimental difficulty in generating a uniform, laminar dispersion of dust accounts for the scarcity of fundamental data on dust flame propagation (burning velocity, quenching distance, flame thickness, etc.). Performing experiments in a reduced gravity environment can largely eliminate particle sedimentation. The low gravity environment also diminishes buoyancy-induced convective flows, permitting very low flame speeds to be observed in suspensions of large particles and slow-burning, nonvolatile fuels. This clear rationale for reduced gravity-based studies of dust flames motivated

1540-7489/$ - see front matter Ó 2009 The Combustion Institute. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.proci.2008.05.084

1906

F.-D. Tang et al. / Proceedings of the Combustion Institute 32 (2009) 1905–1912

drop tower experimentation as far back as the 1970s with the work of Kumagai et al. [1] and Ballal [2]. More recently, Okuyama et al. [3] have obtained flame propagation speeds in PMMA dust clouds utilizing a 10 s drop tower facility, and Dreizen and Hoffmann [4] have studied the structure of dust flames in magnesium and aluminum alloy suspensions in a 2.2 s drop tower. The authors of the current paper have obtained quenching distance measurements for two different sizes of aluminum dust in parabolic flight experiments [5]. The objective of the present paper is to obtain flame speed and quenching distance measurements of iron dust suspensions in air for a wide range of particle size. These results are compared to an analytical model for fuel-rich heterogeneous flame propagation that has previously been developed by Goroshin et al. [6]. While the model has demonstrated good correlation with results obtained using suspensions of small particle sized aluminum dust in air [6], the model predictions have not yet been tested with other fuels or with a wider range of particle size due to a lack of data in the literature. The use of a nonvolatile metal, such as iron, that burns entirely in the condensed phase is of particular interest to validate models that attempt to capture the heterogeneous characteristics of dust combustion.

above their flame temperature, such as carbon and boron, may still partially occur in the gas phase due to the formation of gaseous suboxides (BO, BO2, CO, etc.). Thermodynamic calculations show that a few metals (iron, zirconium, and hafnium) are good candidates to burn as a pure heterogeneous flame uncontaminated by gas phase combustion, since they form only condensed phase oxides. For safety reasons, iron was selected for this study. Equilibrium thermodynamic calculations predict a stoichiometric adiabatic flame temperature in air of 2230 K (well below the boiling point of iron at 3130 K), with the products being condensed phase (liquid) Fe3O4 (83%) and FeO (17%). The iron dusts used (Alpha sar, Atlantic Equipment Engineers, Inc.), were characterized using the Mie scattering technique with a Malvern Mastersizer 2000E system; the arithmetic mean particle size of the powders used are given in Table 1. Scanning electron microscope photographs of the iron dusts used are shown in Fig. 1. 2.2. Apparatus and diagnostics The reduced gravity experiments with constant pressure flames propagating in iron dust– air suspensions were performed in four Pyrex glass tubes with an internal diameter of 48 mm and 70 cm in length (Fig. 2). The quenching dis-

2. Experimental details 2.1. Iron dust characterization The selection of iron dust for the current experiments was motivated by the desire to realize a ‘‘pure” heterogeneous flame, as reflected in the analytical model. This requirement excluded organic dusts such as PMMA, since their volatility results in their burning being not that dissimilar from liquid spray combustion, in which droplet evaporation and combustion of the gaseous fuel vapors are the controlling mechanisms. Similarly, moderately volatile metals such as magnesium and aluminum were also excluded since their flame temperature exceeds the boiling point of the metal, a feature that may result in the formation of a metal vapor–air flame. Even the combustion of refractory elements with boiling points

Fig. 1. SEM photographs of the iron dusts used. Note different scale for image (d).

Table 1 Characteristics of iron powders Dust

A

B

C

D

Supplier Purity (%) Specific surface area (m2/g) Diameter d10 (lm)

Alfa Aesar 99.0 0.175 3.3

Alfa Aesar 99.5 0.078 7.0

Alfa Aesar 99.5 0.055 9.9

AEE 99.9 0.013 26.8

F.-D. Tang et al. / Proceedings of the Combustion Institute 32 (2009) 1905–1912

Acrylic tube

Pyrex tube Air

1907

Steel balls bed Ceramic filter To second stage filter and vacuum

Flame Air

Ignition coil

Dust dispersion Combustion tube

First stage filters Electromechanical actuator

Quenching plate assembly

Fig. 2. The experimental apparatus for studying flame propagation in dust suspensions in the reduced gravity environment, showing four combustion tubes. The upper insets show the coupling between the combustion tubes and the exhaust system and the lower inset shows the quenching plate assemblies located inside the tubes.

tance measurements were made by observing flames propagating through two assemblies of equally spaced steel plates positioned at 15 and 45 cm from the ignition end of the tube. The individual quenching plates had a thickness of 0.6 mm, and each assembly had a fixed spacing between plates, varying between Dq = 2 and 20 mm for the different assemblies (see Fig. 2 inset). Each tube was equipped with a dispersion system that used an impinging jet of high-speed air to disperse the iron dust fed by a piston–syringe mechanism; the details of the dispersion technique are given in [6]. The suspension was ignited by a 100 lm tungsten wire at the open end of the tube, with the flame propagating toward the closed end (dispersion system) of the tube. The combustion products were mixed with large amount of ambient air and were drawn into a multi-stage venting and filtering system. A wide annular opening (see Fig. 2 inset) provided unobstructed access for ambient air, ensuring constant pressure conditions were maintained during the experiment. Results of an experiment were recorded with a video camera (30 frames/s) viewing the entire tube length and a high-speed video camera (250 frames/s) focusing on the first assembly of quenching plates. High-resolution still images were obtained with a digital SLR camera. A fiber optic spectrometer (OceanOptics USB-4000) coupled to the optics of a conventional SLR camera (details in [7]) was used to acquire emission spec-

Fig. 3. The dust combustion apparatus and optical diagnostic rack mounted inside the Falcon-20 aircraft.

tra in the 400–1000 nm range at a rate of 20 spectra/s. The spectrometer field of view was 2 mm and focused at the entrance to the first quenching plate assembly. The diagnostic rack was mounted facing the combustion tubes onboard the Falcon20 research aircraft (Fig. 3). 2.3. Procedure Experiments were performed in an individual tube during the 20–24 s of reduced gravity provided by the parabolic trajectory of the aircraft. Each tube was used only once per flight and all four tubes were replaced between flights. The reduced gravity environment was an average of

1908

F.-D. Tang et al. / Proceedings of the Combustion Institute 32 (2009) 1905–1912

3. Results

0.1

3.1. Observation of flame propagation g

0.05

0

Gravity acceleration, g

2 -0.05 30

31

32

33

34

35

Time, seconds

Dispersion

1

Combustion

0 20

25

30

35

40

45

50

55

60

Time, seconds

Fig. 4. Typical accelerometer measurement of the gravity environment during a parabolic flight maneuver. Inset illustrates the level of g-jitters encountered.

0.02 g, with g-jitter spikes reaching 0.07 g during some parabolas, as shown in Fig. 4. The more severe g-jitters likely affected some experimental results (see Section 3). The automated experimental sequence was initiated upon the onset of reduced gravity. The sequence consisted of dust dispersion into the tube for 10 s, sufficient to fill the tube completely, followed by at 1-s delay after the termination of the dispersion flow to ensure a quiescent suspension. The ignition wire was then energized for 5 s. Flame propagation through the tube and quenching plate assemblies occurred during the remaining 10 s of reduced gravity. The dust concentration was calculated from the feeding rate of the piston, the dust packing density, and flow rate of dispersion air. The concentration calculations were verified experimentally in ground-based trials using the finest dust (A) in which the dispersed dust was collected without combustion using a fine material filter. The dust concentration in the present experiments varied from 900 to 1200 g/m3, corresponding to a fuel-rich iron–air suspension with equivalence ratio / = 1.43–1.90. The fuel-rich suspensions were chosen for the present study as prior experimental and theoretical results have established that flame propagation and quenching in dust clouds are insensitive to variations in fuel concentrations in fuel-rich suspensions [6,8]. The ability to generate uniform suspensions with the finest dust (A) in normal gravity permitted additional tests to be completed with this dust on the ground.

The propagating flames exhibited a parabolic-shaped front, as shown in Fig. 5, similar to that observed in gas flames propagating in tubes (see Supplementary material, Video V1). Tests with the finest particle size dust (A) both on the ground and in reduced gravity exhibited a folded structure consisting of cells of approximately 1 cm width, reminiscent of earlier observations of acoustically excited dust flames in magnesium and aluminum suspensions [9]. Analysis of the high-speed videos, however, did not reveal any measurable oscillations in flame position or speed that would indicate acoustically coupled combustion. For the larger sized particles, the combustion front became highly discrete, with individual burning particles clearly visible (Fig. 5c). Upon encountering the quenching plate assemblies, the flame continued as individual flamelets (see Supplementary material, Video V2) in the case of successful propagation (no quenching). Upon emerging from the quenching plate assembly, the front had a highly corrugated structure, but quickly reestablished its usual parabolic shape within 1–2 tube diameters (see Fig. 6). 3.2. Flame propagation speed The measured values of flame propagation speed in the open tube (outside the quenching plate assemblies) are shown in Fig. 7, as determined by analysis of the video recordings. The error bars represent the standard deviation from 5 to 11 repeated measurements with each particle size used. Flame speed measurements with the finest particle size used (A) were found to be practically the same in reduced gravity and groundbased experiments. The propagation speed of the individual flamelets inside the quenching plate assemblies are shown in Fig. 8, normalized by the average flame speed for the same particle size in the open tube (outside the quenching plates). These measure-

Fig. 5. Flame fronts observed in the smallest sized particle (a), intermediate particle (b), and larger particle (c) iron dust suspensions.

F.-D. Tang et al. / Proceedings of the Combustion Institute 32 (2009) 1905–1912

1909

/V

0

0.8

Minimum speed, V

min

Experiment Theory

0.6 1.2

0.4 0

10

20

30

Particle diameter, μm

Fig. 6. Dust flame in iron–air suspension emerging from a quenching plate assembly (no quench result) using iron B.

80

theory experiment

70

Normalized flame speed, V/ V

0

1

0.8 Dust A, exp. Dust A, theor. Dust B, exp. Dust B, theor.

0.6

0.4

60

0

2

4

6

8

10

12

14

Flame speed, cm/s

Distance between plates, mm 50

Fig. 8. Flame speed inside the quenching plate assemblies, normalized by open tube results for iron A and B. Inset graph shows the minimum flame speed observed in the quenching plate assemblies for all four iron dusts used. Curves are the predictions of theoretical model.

40

30

20

10

14

0 5

10

15

20

25

30

35

12

40

Particle diameter, μm

Fig. 7. Flame speed as a function of particle size. Experimental results are shown with standard deviation error bars. The solid line is the prediction of the theoretical model.

ments were made from analysis of the high-speed video of the fastest flamelets in the quenching plate assemblies (i.e. the first flamelet to emerge from the opposite side of the assembly). The inset in Fig. 8 shows the smallest observed flame speed inside plates for each of the four particle sizes used. 3.3. Quenching distance measurements The ‘‘quench/passed” results of the quenching plate tests are summarized in Fig. 9. The flame quenching results were always unambiguous, with quenching occurring promptly upon the flame encountering the quenching plates. Flame quenching with the finest particle size (A, d10 = 3.3 mm) was not observed even with the smallest quenching plate assembly used (Dq = 2 mm). The large scatter of results obtained with the largest particle

Distance between plates, mm

0

passed quenched theory

large g-jitters

10

8

6

4

2

A

0 0

B 5

C 10

D 15

20

25

30

Particle diameter, μm

Fig. 9. Results of quenching distance experiments, showing ‘‘quenched” and ‘‘passed” results as a function of the particle size. Solid line is the prediction of theoretical model.

size used (D) in Fig. 9 is suspected of being a consequence of greater than average g-jitter (0.07 g) observed on the particular parabolas for these tests, which had very slow flame speeds and thus were more sensitive to perturbation.

1910

F.-D. Tang et al. / Proceedings of the Combustion Institute 32 (2009) 1905–1912

y = 49.488 - 7907x R= 0.99845

−5

Ln (I x λ )

40

T=1819 K

37.5

0.0012

0.0014

0.0016

-1

Relative intensity

1/λ, nm

500

550

600

650

700

750

800

850

λ nm

Fig. 10. Flame emission spectra (for iron C) and fit to Planck’s law for a grey-body emitter (inset).

3.4. Combustion spectra and temperature A typical example of the iron flame spectra registered by spectrometer is shown in Fig. 10. The total absence of any atomic lines or molecular bands on the background of continuous spectral component observed for all powders confirms the pure heterogeneous character of iron particle combustion predicted by thermodynamics. The temperature of emitters (i.e. particles) was found by linear fitting of the spectral intensity to Planck’s law. It shows that light emitters are practically ‘‘grey”, i.e. their emissivity does not depend on wavelength in 500–850 nm range (see cutaway in Fig. 10). 4. Comparisons to theoretical model and discussion In contrast to gaseous flames that are modeled as a continuum, heterogeneous dust flames consist of individual burning particles that might be at much higher temperature than the surrounding gas. However, if the nondimensional parameter v = sca/l2 (representing the ratio of particle combustion time to inter-particle heat diffusion time, where l is the inter-particle spacing and a the thermal diffusivity) is large compared to unity, then a dust flame can still be appropriately modeled as a continuum in which the heat release function is spatially uniform [10]. Estimations for relatively slow burning iron show that this parameter is larger than 5 for fuel-rich mixtures, meaning the ‘‘classical” continuum approach to modeling the dust flame structure can be used. Within this

framework, a model that has been previously developed by Goroshin et al. [6] is briefly outlined here. The model assumes mechanical equilibration between the particles and the gas (no slip), and the particles are assumed to burn in the diffusive regime after ignition, with the burning rate proportional to the local oxygen concentration. Subsequent quenching of particles due to local oxygen depletion is neglected. The calculated temperature of the gas at the moment of ignition takes into account the dependence of particle ignition temperature on local oxygen concentration as well as the thermal inertia of the particles in the preheat zone. Heat losses to the walls are assumed to be due to molecular heat diffusivity only and are written in volumetric form, within the context of the quasi-one-dimensional model. With these approximations, the governing heat and oxygen diffusion equations describing the preheat and combustion zones become linear and can be solved in closed form. The flame speed is then determined by matching the solution for thermal and oxygen flux at the boundary between the preheat and combustion zones. The quenching distance emerges as a turning (i.e. bifurcation) point in the solution for flame speed. The only inputs required to the model (other than the thermodynamic and transport properties of the gas and particulate fuel) are the ignition temperature and burning time of individual particle. For the present study, the ignition temperature was obtained from the literature [11] and the burning time was calculated from the mass balance equation describing the fuel consumption resulting from the diffusion of oxygen to the particle surface and the rate of mass loss of the particle. Comparisons for the predictions of this model to the experimental results obtained are discussed below. 4.1. Flame speed The flame speed predicted by the model is the laminar burning velocity of a planar front. In order to compare this prediction to the experimental results with a curved flame propagating in a tube, the analytic results for flame speed were multiplied by a factor of 2, following the relation established between burning rate and flame speed for similarly shaped flames in gases [12]. The experimental results shown in Fig. 7 exhibit good correlation with the V  1/dp law predicted by this model as well as other theoretical descriptions of flames in solid suspensions without heat loss [13]. The reduction of flame speeds predicted by the model inside quenching plates also demonstrates good correlation with experimental data for different particle sizes (Fig. 8). The predicted minimal values of flame speed at quenching is compared to the experimentally observed minimum flame speeds inside the quenching plates in

F.-D. Tang et al. / Proceedings of the Combustion Institute 32 (2009) 1905–1912

1911

the inset to Fig. 8. It is interesting to note that the flame speed at quenching does not depend on particle size and is a constant ratio of the open-tube pffiffiffi flame speed, inversely proportional to e. This result is an inherent property of bifurcations that is independent of the particular combustion mechanism [14].

3.4. The smallest iron particle size used (A) exhibited a flame temperature about 100 K lower, which could be attributed to the larger surface area resulting in greater radiative losses and earlier quenching of smaller particles leading to incomplete combustion.

4.2. Quenching distance

5. Conclusions

In the analytic model of the dust flame described above, heat losses to the walls appear as the ratio of the characteristic combustion time sc to the heat diffusion time across the channel, D2/a (where D is the channel width and a is the average thermal diffusivity of the gas in the channel). Thus, it follows that pffiffiffiffithe quenching distance Dq is proportional to sc . For particles which burn in the diffusive regime (sc  d 2p , where dp is the particle diameter), the model therefore predicts a linear dependence of quenching distance on particle size. This linear prediction is shown as a solid line in Fig. 9, in satisfactory agreement with the data. In comparison to gas flames such as stoichiometric methane–air or propane–air, which exhibit quenching distances similar or even larger (in comparison to the finest dust used) to the iron–air flames, it is interesting to note that the estimated burning time of iron particles is more than an order of magnitude greater than the characteristic combustion time in a gas flame. The mechanism of particle combustion in a heterogeneous dust flame provides an explanation for the remarkable ability of the thicker dust flame to propagate through a comparatively smaller quenching distance. This is because the dust flame can tolerate a much larger decrease in bulk gas temperature before quenching since all that is necessary to sustain propagation is elevating the particles to their ignition temperature (for iron dust, approximately 850 K). Once ignited, the particles will burn in the diffusive regime as ‘‘micro-reaction” centers, independent of the surrounding gas temperature. A gas flame governed by Arrhenius kinetics, in contrast, cannot continue to propagate if the flame temperature decreases by one characteristic interval RT 2af =EA [15], which for a gas flame is of the order of a few hundred degrees, making it much more susceptible to quenching.

As the experimental results of this work have demonstrated, the reduced gravity environment has proven to be an invaluable tool in dust combustion research, permitting measurements of the fundamental flame parameters over a wide range of particle sizes for the first time. The comparison of the data obtained with a previously developed theoretical model [6] reinforce the authors’ conclusion that the dominant mechanism in differentiating dust flames from homogeneous combustion is that the ignition of individual particles results in the appearance of micro-diffusion flames within the global reaction zone (i.e. ‘‘flames within the flame”). This feature, in the authors’ opinion, accounts for the majority of characteristics that distinguish dust flames from gas-phase combustion.

4.3. Flame temperature

References

The flame temperature measured by emission spectroscopy can be compared to the adiabatic flame temperature calculated by the THERMO thermodynamic equilibrium code [16]. Over the fuel equivalence ratio range of / = 1.4–1.9, the adiabatic flame temperature is predicted to be Taf = 1810–1990 K, which is within the range of the experimentally measured flame temperature for the (B) and (C) iron dusts reported in Section

Acknowledgments This work was supported under Canadian Space Agency Contract 9F007-052073/001/ST, with Dr. Marcus Dejmek serving as scientific authority. The authors thank the personnel of the NRC Flight Research Laboratory for their outstanding support of the reduced gravity experiments reported in this paper. The authors also thank Dr. Jeffrey Bergthorson, Malcolm Cairns and Andrew Barkman for assistance in performing the parabolic flight experiments. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.proci.2008.05.084.

[1] S. Kumagai, T. Sakai, S. Okajima, Proc. Combust. Inst. 13 (1971) 779–785. [2] D.R. Ballal, Proc. R. Soc. Lond. A 385 (1788) (1983) 21–51. [3] Y. Okuyama, Y. Ohtomo, K. Maruta, H. Kobayashi, T. Niioka, Proc. Combust. Inst. 26 (1996) 1369–1375. [4] E.L. Dreizen, V.K. Hoffmann, Combust. Flame 122 (1–2) (2000) 20–29.

1912

F.-D. Tang et al. / Proceedings of the Combustion Institute 32 (2009) 1905–1912

[5] S. Goroshin, J. Mamen, J. Lee, K. Sacksteder, Proceedings of the 20th ICDERS, McGill University, Montreal, Canada, 2005. [6] S. Goroshin, M. Bidabadi, J.H.S. Lee, Combust. Flame 105 (1) (1995) 147–160. [7] S. Goroshin, J. Mamen, A. Higgins, T. Bazyn, N. Glumac, H. Krier, Proc. Combust. Inst. 31 (2007) 2011–2019. [8] S. Goroshin, I. Fomenko, J.H.S. Lee, Proc. Combust. Inst. 26 (1996) 1961–1967. [9] S. Goroshin, V. Shevchuk, N. Ageyev, Combust. Explo. Shock Waves 17 (6) (1981) 595–600. [10] S. Goroshin, J.H.S. Lee, Y. Shoshin, Proc. Combust. Inst. 27 (1998) 743–749.

[11] L. Breiter, V.M. Mal’tsev, E.I. Popov, Combust. Explo. Shock Waves 13 (4) (1977) 475–485. [12] B. Lewis, G. von Elbe, Combustion, Flames, and Explosions of Gases, Academic Press, New York, 1961. [13] H.M. Cassel, Some Fundamental Aspects of Dust Flames, Report No. 6551, US Bureau of Mines, 1964. [14] V.I. Arnold, Catastrophe Theory, Springer, Heidelberg, 1992. [15] Y.B. Zeldovich, G.I. Barenblatt, V.B. Librovich, G.M. Makhviladze, The Mathematical Theory of Combustion and Explosion, Plenum Press, New York, 1985. [16] A.G. Merzhanov, Int. J. SHS 2 (2) (1993) 133–156.