Study of the ignition of single coal and char particles in a drop tube furnace by a probability method

Twenty-Fourth Symposium (International) on Combustion/The Combustion Institute, 1992/pp. 1127-1133

STUDY OF T H E I G N I T I O N OF SINGLE COAL AND CHAR PARTICLES I N A D R O P T U B E F U R N A C E BY A PROBABILITY M E T H O D R. BOUKARA, R. GADIOU, P. GILOT, L. DELFOSSE AND G. PRADO Laboratoire Energdtique--Combustion et Environnment C N R S - - S D I 6302 Ecole Nationale Supdrieure de Chimie de Mulhouse 3, rue Alfred Werner F68093 Mulhouse cedex, France

A drop tube technique coupled with optical particle detection and three color pyrometry has been applied to study the ignition of single particles of a petroleum coke, a high volatile bituminous coal (Freyming--France) and their chars. The objective of the study was to measure the critical ignition temperatures and the ignition delays. This was performed in both cases in a pure oxygen atmosphere using five particle sizes,: 100-125 p.m, 125-160 I.Lm, 160200 Ixm, 200-250 ixm and 250-315 p,m. Forced ignition delays were measured only for chars. Results show, that critical ignition temperatures decrease with increasing char particle diameter and with decreasing coal or petroleum coke particle diameter. The Semenov's steady state thermal ignition theory was used to predict critical ignition temperatures. A transient heat balance equation was developed to predict ignition delays. The experimental results are correctly accounted for, except in the case of the smallest particles, for which the model over-predicts the ignition delays.

Introduction The experimental study of the self ignition of solid particles has been most often, performed using facilities in which particles were introduced into the reaction zone under the form of a cloud. Without reviewing here all the studies that have been done in that way, examples of some recent experimental or theoretical studies are Juniper et al., 1 Krishna et al., ~ Gupta et al., 3 a recent exhaustive review on the subject was done by Essenhigh. 4 The most important problem encountered when using clouds is that one cannot easily avoid cooperative phenomena which may possibly mask a number of specific features relevant to each isolated particle. A number of authors have been aware of this problem and the current tendency has been to decrease the particle concentration in the cloud, so that each particle could be considered as not interacting with its neighbours. This was the case for example with Chen et al. 5 Experiments with a single particle are more difficult to perform for several reasons including problems related to feeding the reactor, holding the particle, detecting light threshold. Besides, except in the case of a falling particle, or in the case of the electrodynamical balance, 6 problems in heat transfer calculation arise due to the necessity of using a holder (grid, thermocouple, quartz tip etc.). Nevertheless a number of works, in which isolated or very diluted particles have been considered, can

be found among which Bandyopadhyay et al., 7 Ivanova et al., s Yang et al., 9 Karcz et al., 1~ Kordylewski et al., 11 Tognotti et al., 12 and Gomez et al.13 The most common target parameter of these experiments is to determine the ignition temperature, which is often equated to the temperature of the surrounding gas at ignition fig). The aim of the present paper is to report about a number of experiments done using a drop tube furnace with single char or coal particles, mainly looking at the influence of particle size on ignition delays and temperatures. Finally, Semenov's thermal explosion theory (TET) used in unstationary or stationary conditions will be used, to account for the obtained results. A number of recent applications tend to use pure oxygen or at least enriched oxygen atmospheres (for example blast furnaces). Among the main advantages bound to the use of pure oxygen are: smaller volume of fumes to clean, higher temperatures. The present study was performed mostly using pure oxygen.

Experimental The main feature of the experimental apparatus is a drop tube furnace (Fig. 1). The highest temperature that can be obtained is 1700 K. The reactor is made of a cylindrical alumina tube with an internal diameter of 45 mm and a length of 1400

1127

1128

COAL COMBUSTION 12,

1.6

1.2

9~

1

0.8

~ ~

0,6

0.4

0.2

0

0

I

I

t

I

I

50

1 O0

150

200

250

300

time ( ms )

FIG. 2. Recording of the light emission by the three color pyrometer at 633 nm. 14

, ..

F~c. 1. Scheme of the drop tube furnace 1 Secondary gas injection 2 Primary gas injection 3 Single particle injection 4 Water--cooled injection probe 5 Saphire window 6 Quartz window 7 Signal comparator 8 Interference filter (632,8 nm) and photo-cell 9 Diaphragme 10 Diaphragme 11 Lenses 12 Three wavelength pyrometer 13 Data acquisition and Pyrometer's curves treatment computer 14 Ash evacuation 15 HeNe laser (0,5 mW) mm. A rotating plate is used as a distributor. The single particle is injected in the hot gas stream through a water cooled probe. A primary gas stream of 40 1 (STP) per hour is preheated in the first 40 cm of the reactor. Previous calculations]4 made in our laboratory have shown this preheating to be sufficient to obtain a gas temperature equal to that of the furnace wall. A secondary gas stream of 0.5 1 (STP) per hour is maintained in the injection probe. A three color pyrometer is used to analyze the radiation from the coal particle falling along the reactor axis. The three wavelengths used in our experiments were: 515, 633

and 746 nm. Calibration of the apparatus is done using a tungsten ribbon lamp in the temperature range 1300-2000~ K. Two independent temperatures can be computed based on the three wavelengths emission curves. If the particle temperature lies inside the calibration range, measurements have shown that these two measured temperatures usually differ by less than 30 ~ K. Beyond 2000~ K, the difference may be more important. A 0.5 mW H e - - N e laser is used to detect the entrance of the cold particle in the hot gas stream. This is done through a photocell opposed to the laser, which measures the attenuation of the signal. A comparator sends a trigger signal to the data acquisition computer. This particle detector can be used for particles having a diameter greater than 80 txm. Figure 2 shows a typical curve for the pyrometer signal. Different characteristics of the particle combustion can be measured: t i : ignition delay t r : volatile flame duration th : heterogeneous combustion duration Th : heterogeneous combustion temperature Ti : ignition temperature In this paper, we will report only on ignition temperatures and delays. Ignition Temperatures of Single Coal or Char Particles: In this type of experiments, single particles of carbonaceous materials were injected in the hot pure oxygen stream of the furnace, varying oven temperature stepwise from 873 to 1373 K. Ignition was detected using the above mentioned three color pyrometer. The percentage of ignited particles (hereafter ignition ratio) was calculated from experiments involving approximately one hundred particles. Four

IGNITION OF COAL AND CHAR PARTICLES samples were used, which correspond to the following nomenclature: FI High volatile bituminous coal from Freyming (France)(ash = 3% df, elemental analysis (dat): C 82.9%, H 5.5%, O 9.3%, N 1.1%, S 0.7%) F2 Char of F1 obtained by flash pyrolysis under nitrogen at 1000~ C in the drop tube furnace P1 A petroleum coke P2 Char of P1 obtained in the same conditions as for F2.

1 O0

90

§

o o

~

so

el_ = .2

40

:~

+ _ o

6o

~

9 o

x

70

x

-

-

§

30 •

20

x

10

o

800

x

I

I

;

I

I

900

1000

11 O0

1200

1300

0

g;=~ t a r n n e r a t u r e ( K )

(a) 100 x

9O x

~

o

x

8O

~-e

o x +

70 x

~

60

~ o

50

o

+ +

g 40 ~ 3o Results

.

80

•

The volatile matter removed by the inert pyrolysis of F1 and P1 were 59 and 24 mass percent (dry basis), respectively. Four particle sizes were used, 100-125 p.m, 160-200 p.m, 200-250 p.m and 250315 p.m. In order to evaluate precisely ignition temperatures, we used the probit method 16 which allows to fit the ignition probabilities to Gaussian curves. The ignition temperatures are therefore obtained at the inflection point of the probability curves; above these temperatures, ignition delays were measured.

1129

• o

-

x +

x

20

o x

10 +

Ignition Temperatures of Single Coal Char Particles:

0 800

900

gas

Examination of curves obtained for chars F2 and P2 Fig 3a and 4a, which are volatile free, shows the strong influence of the particle diameter on the heterogeneous ignition temperature, as already pointed out in earlier studies. ~-7'1~The results obtained through the Probit analysis are shown on Fig. 5. The sigma values calculated for the F2 experiment were constant (mean value 60 K) while they were constantly decreasing from 147 K to 96 K with increasing diameter in the case of P2. The obtained values are higher than usual gas temperatures at ignition measured in other studies, zA~ Ignition experiments performed using non devolatilized samples (Fig 3b and 4b) show quite a different influence of particle diameter. The ignition probability here, decreases with increasing diameter, the sigma values were both decreasing on increasing diameter (from 100 to 53 K in the case of F1, and from 79 to 42 K in the case of PI). Another particular observation is that the ignition probability sharply increases in the 750-800~ C range for the two samples, though this is not so obvious with the petroleum coke sample. It seems therefore reasonable to relate this behaviour to the different volatiles matter contents. (24 p. cent in mass for petroleum coke versus 59 p. cent for Freyming coal).

I

J

I

I

1000

11 O0

1200

1300

temperature

( K

}

(b) FIc. 3(a). Ignition probabilities of Freyrning coal for different sizes: +100 to 125 t.tm; "125 to 160 Ixm; • to 250 p.m; -250 to 315 p.m. (b) Ignition probabilities of Freyming coal char for different sizes: +I00 to 125 p.m; "125 to 160 p.m; o 160 to 200 p.m; • to 250 Ixm; -250 to 315 p.m.

Ignition Delays: The obtained ignition delays in pure oxygen at 900~ C and 1000~ C using various sizes of Fe and P2 chars are shown in Fig 7 and 8. It is obvious that the ignition delay is almost independent of diameter.

Discussion and Interpretation

Ignition Temperatures of Single Coal or Char Particles: The critical conditions for ignition of the solid particles were computed from Semenov's steady state

1130

COAL COMBUSTION

100

1200-

x o

90 80

~

1150-

o

70

I

x

~

5o

g

40

s

..'

.,'

.X

..,, ,,-X"

0

"'...

..X ......"

""-I-'...

X ........ "......... ".. 0 "..... '..

""-....

. ...§

.~

1050-

20

~x

10

~

x

x

•

x~ I

0 80O

•

900

i

I

l

I

1000

11 O0

1200

1300

gas t e m p e r a t u r e

""'0

1000

( K )

(a)

95~00

1 O0 • 5

80

o i

70

o

>" 60

30

FIG. 5. Ignition temperatures calculated using the Probit method. +F2; o P2; xF~; *Pl.

x

+

x +

+

"g

o

o

l

L~

40

• x

+

x

=

1~0 1,0t 1~0 1~0 2~0 220~ 2~0 2~0 2,0 ,~0 mean particle diameter (Fm)

90

and the rate of heterogeneous reaction Ra must account for diffusional and kinetic limitations. This gives:

20

o

10 0

"'.. 0 "".. ".., ""..,.. 9. ",...j. " ''''"-..

1100-

.~ 3o

~-e

+..

~

800

§

~1 900

Ro = / 3

+

I

I

~

I

1000

1100

1200

1300

gas t e m p e r a t u r e

( K )

with

kK =

+

Xo2

(4)

ko exp (-E/RTp)

(5)

ko = Sh D/dp

and

(6)

(b) FIG. 4(a). Ignition probabilities for different sizes: +100 to 125 I~m; o 160 to 200 Ixm; • to 315 p.m. (b) Ignition probabilities char for different sizes: +100 160 I~m; o 160 to 200 Ixm; • to 315 I~m.

of Petroleum coke i~m; "125 to 160 250 I~m; - 2 5 0 to of Petroleum coke 125 txm; "125 to to 250 I~m; - 2 5 0

1200.

1150-

1100-

o.

"...... ;- .......... "'.

theory. According to this theory we m u s t have at ignition:

gl = g2

0 +

"-,, 1050-

""....,... 0 ""..

(]) 1000-

"'"......,...

o

and

Ogl - -

aT

950 100

Og2 =

- -

aT

(2)

gl is calculated from:

gl = q~At,Ra

(3)

1210 1~ 160 1/0 2100 210 240 I 2160 2/0 300 I mean particle diameter (Fm)

FIG. 6. C o m p a r i s o n b e t w e e n predicted and measured values of ignition temperature. +Fz measured; o P2 measured; --- Fz predicted using ko = 6.75 10 a m / s ; " " P2 predicted using ko = 9 103 m / s .

IGNITION OF COAL AND CHAR PARTICLES 400'

1131

pressed as: '\",

g2 = Ap [h(T,, - Tr + co" (Tv 4 - T,~4)]

%E E

300,

h is calculated from the Nusselt number:

"% %',,

200-

0

"..

0

:.D

0 "'''"-......

+

.............................. + +

and

0

Nu = (h do~A)

(9)

Nu = 2 (1 + 0.3 Rel/2pr I/3)

(10)

By numerically solving equations (1) and (2) in order to fit experimental and computed values, we determined ko and E for both F2 and P2 samples, this leads to the following values:

-

+

100-

ko=9.5103m/s

%

(8)

1201/01101801

20102~021026102~03~0

and

E = 128kJmole -1

For P2 samples, we obtained:

mean porticle diameter (tam)

ko=4.25.104m/s

FIG. 7. Comparison between predicted and measured ignition times for Freyming coal chars at various supercritical temperatures, o 900~ C measured; --- 900~ C predicted; +1000~ C measured; ..- 1000~ C predicted. was measured by calorimetry and the Sherwood number was calculated according to the formula: Sh = 2 (1 + 0.3 Rel/2Sc l/3)

(7)

The heat loss, g2, from the particle can be ex-

E = 9 4 k J m o l e -1

The computed critical values of Tr for ignition are given in fig 6 and compared with the experimental ones. Predictions appear to be very good for bituminous coal. For petroleum coke, the reactivity of char is intermediate between bituminous coal and anthracite. Ignition Delays Transient Model:

When gas temperature becomes much higher than the ignition temperature, the steady model is no longer valid. The heat balance for the particle is written: dT mp C,,-~tt = gl - g2

400.

(11)

Expressing gl and g2 this equation becomes (assuming spherical symmetry):

300-

% E E

and

,.

0

0

+

..... ...-

C

..... "'"

" d--T= r

+

Xo~

200-

0

C 0

+

.+... ..-" ...."

+

g

- h(Tp - Tg) - ,a(T~ - T4w) (12)

The specific heat of the char is calculated using Merrick's iv model as a function of particle temperature. The criterion for ignition is now obviously:

100-

~00 ~201I~0 1801I~02~02~02~02~02~03~0

O~TP = 0 Ot2

(13)

mean panicle diameter (tam)

FIG. 8. Comparison between predicted and measured ignition times for petroleum coke chars, o 900~ C measured; --- 900~ C predicted; + 1000~ C measured; "" 1000~ C predicted.

Equations (11) to (13) were solved, taking into account the drop velocity of the particle in the temperature gradient in the injection probe and allowing for the physical properties variations of the gas with temperature. Computed ignition delays are

1132

COAL COMBUSTION

plotted in fig 7 and 8, compared to the experimental values. The prediction appears reasonably good for particles of higher diameter showing particularly the independence on particle diameter. This reflects the competition between heating rate and drop velocity for particle of different sizes. The disagreement observed for lower diameters may be due to the t:act that, for these particles, the temperature of experiment is close to their critical ignition temperature. The model therefore has a marked tendency to predict very large ignition delays.

Ignition of Coals: Ignition of coals could not be modelled using such simple theories because one must take into account the possible hetero-homogeneous mechanism or merely the homogeneous mechanism involving volatiles. This requires more information on devolatilization rate, volatile matter ignition and combustion. Work on this subject is in progress in our laboratory.

Conclusion The use of the drop tube furnace technique coupled with simultaneous optical detection and uptical pyrometry has alh)wed the determination of ignition temperatures of single particles of char and their parent fuels using a statistical method. At higher temperatures, it was also possible tu measure their forced ignition delays. Heterogeneous ignition of char shows that ignition temperatures decreases when increasing the particle diameter. The reverse situation is encountered with parent fuels. It is suggested that this complete reverse behaviour is due to the importance of devulatilization rate. Semenov's thermal ignition theory, applied nnder steady conditions for critical ignition temperature calculations, allows their correct prediction. A transient model has been used to model ignition at temperatures higher than the critical one. This allows the correct prediction of the ignition delays for larger particle diameters but strongly overpredicts the ignition delays for smaller ones. This is ascribed to the fact that, for these particles, the explored temperature range is too close to their critical ignition temperature.

Nomenclature A Cv D d E

: Surface area : Specific heat capacity of particle at constant pressure : Diffusion coefficient of oxygen : Diameter : Activation energy

gl g2 h k

ko m Nu Pr R

Ra Re Sc Sh T t X

: : : : : : : : : : : : : : : :

tieat evolved by chemical reaction Heat loss Convective heat transfer coefficient Rate constant Pre-exponentional factor Mass Nusselt number Prandtl number Perfect gas constant Overall reaction rate Reynold number Schmidt number Sherwood number Absolute temperature Time, durations Molar fraction

Subscripts c D g i K P r

: : : : : : :

Convective Diffusion Gas Ignition Kinetics Particle Radiative

Greek symbols /3 h o" ~b p

: : : : : :

Stoichiometric coefficient Emissivity Heat conductivity Stefan constant Combustion enthalpy Density

REFERENCES 1. JUNIPER, I,. A. AND WALL, T. F.: Combust. Flame 39, 69 (1980). 2. KBISHNA, C. R. AND BERLAD, A. L.: Combust. Flame 37, 207 (1980). 3. GuerA, R. P., GURURAJAN,V. S., LUCAS, J. A. AND WALLS, T. F.: Combust. Flame 79, 333 (1990). 4. ESSENHIGH,R. ti., Mlsrt~, M. K. AND SHAW, D. W.: Combustion and Flame, 77, 3 (1989). 5. CtlEN, M. R., FAN, L. S. AND ESSENHIGH, R. I1.: Twentieth symposium (International) on Combustion 1513 (The Combustion Institute)(1984). 6. BAR-ZIv, E. AND SAROFIM, A. F.: Progress in Energy and Combustion Sci 17, 1, 1 (1991). 7. BANDYOPADHYAY, S. AND BHADURI, D.: Com-

bust. Flame, 18, 411 (1972). 8. IVXNOV^, I. P. AND BABH, V. L. : Teploenergetika--13, 54 (1966). 9. YANG, J. T. AND TSAI, G. T.: Fuel, 69, 696 (i~o).

IGNITION OF COAL AND CHAR PARTICLES 10. KARCZ, M., KORDYLEWSKI, W. AND RYBAK, W.: Fuel, 59, 799 (1980). 11. KORDYLEWSKI,W., KRUCZEK, a . AND RYBAK, W.: Combust. Sci. Tech., 26, 157 (1985). 12. TOGNOTrl, L., MALOTrI, A., PETARCA, L. AND ZANELLI, S.: Combust. Sci. Tech., 44, 15 (1985). 13. GOMEZ, C. O. ANO WASTOLA, F. J.: Fuel 64,

558 (1985). 14. GADIOU, R.: Ph.D. Thesis--Universit6 de Haute Alsace Mulhouse n ~ 159 (1990).

1133

15. CHICK, A.: Ph. D. Thesis--Universit6 de Haute Alsace Mulhouse--6.24 (1986). 16. OUTASSOURT, T., CHARON, O., PRADO, G. AND LAHAYE, J. : Joint Meeting of the French and Italian Section of the Combustion Institute, Naples. Preprints, Sect. 9.10 (1987). 17. NARTRELLA, N. G.: Experimental Statistics, United States Department of Commerce N.B.S. (Handbook 91). 18. MERmCK, D.: Fuel, 62, 540 (1983).

COMMENTS Prof Stephen Niksa, Stanford University, USA. Were you able to identify the ignition mode, either homogeneous or heterogeneous, from the signals from your photo detectors? If so, please describe any transitions between modes as furnace temperature is increased for the whole coal sample.

Author's Reply. With non-devolatilized samples, heterogeneous ignition could result in two different cases. If there is not any substantial heterogeneous combustion before the stabilization of a volatile matter flame around the particle, there will be no difference between photomultiplier signals recorded for homogeneous and heterogeneous mechanisms. If there is a sufficient amount of heterogeneous combustion to raise the particle temperature, there will be a light emission which could be detect by the pyrometer before the appearance of the volatile matter flame. This last case was not observed in our experiments, it may be possible with samples having lower volatile matter content than ours (10% or less).

Herman Krier, University of Illinois at UrbanaChampaign, USA. Both the steady-state and transient lumped parameter ignition models you described would require coal/char surface area as an input. Does the mean particle diameter properly give these surface areas? For particles less than 100 Ixm the answer to this can be more critical. I've noted that your predictions weren't "as good" at your 'smaller" size coals/chars.

Author's Reply. Particle surfaces used in our models were based on nitrogen adsorption measurements. The most probable explanation for the bad predictions of ignition times of small particles are that we may detect only the highest diameter in the 100-125 i~m size.

Adel F. Sarofim, Massachusetts Institute of Technology, USA. Do you have an explanation for the spread in the probability of ignition? Is it a consequence of the variability in the reactivity of individual char particles? Or that of the specific heat which was shown by your laboratory to influence ignition time? It would be interesting to match the variance as well as the mean by your model.

Author's Reply. We effectively tried to correlate the shape of the ignition probability curves with variations in particles size for the differents samples, but it was not sufficient to explain the observed sigma values. Variations in specific heat may be a consequence of the mineral matter content of the coal,' but we did not test it. The variance experimentally observed is in fact the sum of the different variances of the coal in diameter, surface, shape, composition. REFERENCE 1. OUTASSOURT, T., CHARON, O., PRADO, G. AND LAHHAYE, J.: Joint Meeting of the French and Italian Sections of the Combustion Institute, Amalfi, Italy (1987).

Prof T. F. Wall, University of Newcastle, Australia. You define ignition at a 50% probability of detection by the pyrometer. Was the spread in the probability of detection different for the coals and chars, and does it help you decide if ignition is homogeneous or heterogeneous?

Author's Reply. The sigma value observed in our experiments seems to be mainly a consequence of particle-to-particle variations for a given experiment, so that the influence of the ignition process is very difficult to detect. Numerical simulations are being performed in our laboratory to analyze the variation of sigma values.