Char fragmentation and fly ash formation during pulverized-coal combustion

Char fragmentation and fly ash formation during pulverized-coal combustion

COMBUSTION AND FLAME 90:174-184 (1992) 174 Char Fragmentation and Fly Ash Formation During Pulverized-Coal Combustion LARRY L. BAXTER Combustion Res...

839KB Sizes 20 Downloads 144 Views

COMBUSTION AND FLAME 90:174-184 (1992)

174

Char Fragmentation and Fly Ash Formation During Pulverized-Coal Combustion LARRY L. BAXTER Combustion Research Facility, Sandia National Laboratories, Livermore, CA 94551 0969 Experimental measurements of char and fly ash size distributions are reported in the size range from approximately 0.5 to 100 p~m for three coals, ranging in rank from high-volatile bituminous coal to lignite. These measurements are coupled with a theoretical model of fly ash formation to determine the extent of char fragmentation as a function of initial char particle size. These data reveal several mechanistic aspects of fragmentation. The extent of fragmentation is strongly dependent or size and coal rank. Bituminous coals may form over 100 fragments per char particle at large initial char particle sizes (above 80/.~m) and less than 10 at small initial char particle sizes (less than 20 p.m). Lignites fragment less extensively, with the number of fragments per original char particle being less than 5 at all particle sizes. These results partially resolve some apparent discrepancies in published studies of fragmentation.

INTRODUCTION Char fragmentation and fly ash formation during pulverized-coal combustion affect important aspects of boiler operation, including radiative heat transfer, fouling and slagging propensities, and particulate removal efficiency. Fly ash that escapes particle cleanup systems constitutes a significant environmental and health hazard and is the focus of substantial legislation and regulation [1]. These practical, environmental, health, and legal considerations have motivated a number of theoretical and experimental studies of fly ash formation. Reported mechanisms of fly ash formation include (1) vaporization and recondensation of volatile components of coal ash [2-6], (2) fragmentation of ash due to inorganic reaction [7, 8], (3) convective transport of ash during rapid organic reaction [9-11], (4) structural disintegration of chars [5], (5) shedding of ash material from the surface of chars during combustion [2, 4, 11, 12], and (6) coalescence of ash within a char particle [4, 7, 8]. The term fragmentation, as used in this article, encompasses all mechanisms that produce more than a single ash particle from a char particle and could involve any or all of the above processes. The number of fragments formed per char particle can exceed the number of fly ash particles formed per char particle, since some fragments formed by mechanism (4) above may contain 0010-2180/92/$5.00

no mineral matter. This article addresses char fragmentation as indicated by fly ash formation, which may underpredict the absolute extent of char fragmentation. Only those char fragments that produce residual ash particles are addressed by the methods of this article. Several mechanistic details of fragmentation are available in theoretical papers. Statistical studies ranging from number balances [13, 14] to percolation models [15, 16] describe the transient development of char particle size distributions, often for a series of assumed scenarios of fragmentation. These models vary widely in their level of detail, although none describes all aspects of fragmentation and parameters used in the models do not always lend themselves to experimental measurement. Although progress in this area continues, there exist no complete descriptions of fragmentation in the literature. Experimental techniques for studying fragmentation include both solid-sampling and laser-based, in situ techniques. Typical solidsampling studies involve size classification of fly ash by a cascade of impactors. Feed material and equipment used in these studies range from mechanically classified samples of coal entrained in laminar flow facilities, [2, 17] to utility-grind coal [18]. Results from various experiments appear discrepant when examined superficially. For example, the reported number of fly ash particles generated per char Copyright © 1992 by The Combustion Institute Published by Elsevier Science Publishing Co., Inc.

C H A R F R A G M E N T A T I O N AND FLY ASH F O R M A T I O N particle varies from near one [18, 19] to several tens of thousands [4, 20]. As will be illustrated, these discrepancies may be more superficial than actual. None of these investigations used the same analysis technique to characterize both the feed material and the fly ash under combustion conditions representative of commercial boilers. Specific objectives of this investigation are to (1) devise a fragmentation model that is sufficiently complete to analyze data collected from systems using utility grind coal as a feedstock, (2) measure experimentally the initial and final particle size distributions using the same in situ diagnostic and under conditions comparable to commercial scale operation, and (3) combine the theory and experiment to determine the extent of char fragmentation. THEORETICAL APPROACH The fragmentation model described below relates only the initial char (not coal) feedstock and final fly ash size distributions. The model is written in terms that are convenient for laboratory analyses. That is, the char particles have physical properties that can vary as a function of size but, at a given size, are described by a single-valued function. Particleto-particle variations in physical properties at a given size are neglected. Also, transient developments in particle size distributions, as discussed by many of the previously cited authors, are neglected. Size distributions of the initial char and final fly ash can be related, in general, through a particle number balance

,,lna(X) d x

jde=0 dc=~j~Og(xly) d x n c ( y ) dy. do1 f

=

t

"

(1) The functions n', and n'C represent particle concentration density functions for fly ash and char, respectively, with typical units of particles/(m 3 /xm). Dummy variables x and y are used in the integrands to represent fly ash (d~) and char (dc) particle diameters, respectively. Integration of n' with respect to particle diameter represents the cumulative concentration of particles with sizes between the limits of integration. Thus, the integral on the left rep-

175

resents the total concentration of fly ash particles with diameters larger than da] and smaller than da. The function g(xly) is a conditional fly ash particle size distribution (psd); It indicates the ultimate size distribution of fly ash particles formed from a char particle with initial size y. This represents the only portion of the above equation that is not determined experimentally in this investigation. Several published theoretical and experimental results suggest general forms of the function. One example [21] provides data from which g(xly) can be determined for one char particle size. Coal particles were mechanically sieved to a 140 tzm nominal diameter and combusted at a particle temperature of about 2000 K in a 20% O 2 environment. The number of fly ash particles generated was deduced as a function of fly ash particle size. From these data, a cumulative size distribution can be calculated. Differentiating this cumulative distribution with respect to particle size yields g(xly), which can be sufficiently described as

a(y)x b, g ( x l y ) = tO,

for x](y) < x < Xz(y), otherwise,

(2) where x] and x 2 represent the diameters of the smallest and largest fly ash particle generated, respectively. This conditional psd describes a power-law size distribution between the limits x I and x 2 for fly ash particles generated from each char particle. Note that this functional form can only describe the results from an ensemble of initial char particles with size y. The functional form of g for a single char particle is series of Dirac delta functions. Values for a and b of 3260 and -3.21, respectively, fit the data of Quann and Sarofim [20] with a coefficient of determination [22] (r 2) greater than 0.99 (in logarithmic coordinates with x expressed in microns). The original data of Quann and Sarofim have been recast in number density form, and this correlation is compared with these recast data in Fig. 1. The authors state that the smallest particles (those smaller than about 0.6 p~m) are generated by vaporization-recondensation

176

L. BAXTER

10 lo

lO e

i

........ i

,

->" 10°

.

.

.

.

.

.

.

.

,

.

.

.

.

.

.

.

.

,

.

.

.

.

.

.



Experimental Values



Correlation

in the char particle (Pa and wo, respectively) represent values of ash, as opposed to the mineral matter from which the ash is formed. It follows from Eq. 3 that

.

10 0 10 s "~ 10 4 :~ 10 s .~

a(y)

10 2

_~ 10' 100 10-+ .01

............................... .1

10

1

100

Fly Ash Particle Diameter (tam)

Fig. 1. Experimental data derived from Quann and Sarofim [20] indicating the distribution of fly ash particle generated from a single char particle. The correlation is based on Eq. 2. Parameters are discussed in the text.

mechanisms, whereas the larger particles are believed to be generated by surface drop shedding or disintegration of char structures. If the data at sizes smaller than 0.6 /zm are eliminated, values for a and b of 2370 and -2.95, respectively, fit the remaining data, also with a coefficient of determination greater than 0.99. Therefore, the functional form for g(xly) given by Eq. 2 represents at least this set of data regardless of the mechanism of fragmentation and the value of b used in the conditional psd can be assumed to be approximately - 3 . Results of a percolation model of char fragmentation [14] provide further theoretical motivation for this functional form of this conditional pdf. This formulation has been used in other theoretical studies of the transient evolution of char particle size distributions [15]. A mass balance on the fly ash determines the function a(y), as follows: ~Tr

ma(Y) = fo -ff pax3g(xly) dx 7r = -~Paa(Y) Ax(y)

wa(Y)pcY 3 p~ A x ( y )

(4)

For convenience in determining the upper and lower bounds of g, a parametric function f ( y ) is designated the fragmentation factor and is defined as follows: ( 09a Oc/Pa)l/3y

f(y) -

X2

(5)

The numerator represents the size of the fly ash particle that obtains when all of the ash in a char particle coalesces, that is, no fragmentation occurs. The function f equals the ratio of this hypothetical size to the size of the largest fragment actually generated from a char particle (x2); f increases with an increase in the extent of fragmentation. In the limit of no fragmentation, f is unity, independent of either initial char particle size (y) or char size dependencies in either 19 or oJ. (This statement may appear to contradict the explicit dependence of f on y indicated in Eq. 5. However, in the case of no fragmentation, this explicit dependence cancels with the implied dependence of x 2 and of the physical properties on y.) In general, f is expected to be greater than or equal to 1. The parameter f indicates the extent of fragmentation in that it increases as the extent of fragmentation increases. If char were to produce a monodispersion of fly ash particles as a consequence of fragmentation, the total number of fragments produced would be equal t o f 3 . A second function s(y), designated the fragment size ratio, is defined as

77"

= ma(y)m e = 09a(y)~pey 3,

(3)

where p represents density, w represents mass fraction, and Ax represents x 2 - x 1. Note that the density of the initial char particle (Pc) represents the density of the total particle, not just the organic portion. Also, the density of the final ash particle and the ash mass fraction

s(y) =-

xl(Y) x2(y)

(6)

and represents the mode of fragmentation in the sense that s decreases as the range of fragment sizes produced from a char particle increases. If there is no fragmentation, both f and s are unity. As s decreases from unity

C H A R F R A G M E N T A T I O N AND FLY ASH F O R M A T I O N toward zero, the range of sizes of fragments produced from a single char particle increases. Equation 2 can be written in terms of the parameters f and s as

g(xly) I t°aPcya.x_3= f3x23 = paA-"""~ AX x-3, for x l ( y ) < x <_x2(Y) 0,

(7)

(9)

(10)

where p is independent of y and typically has a value of order 0.5 or less. This allows the integral of the conditional psd to be expressed analytically in terms of y only. If all ash particles are included in the integration (d~l = 0), the relationship between char and fly ash concentration distributions can be written as

na(x) dx

( f3

= fod"/'ps)n'c(y)'2t 1-_sas2 )dy

=-2-1 s2- s3) +min 2(i-s)

It x2 ; (1;10/ x2 min(x2,d,l)

-

1-

s

+[d"/(PS)n'~(Y) f3 [1 - (py):] dy. "G/p 2(1 - s )

'

max(xl,da)

;]0) ,

,

(8) where s, f, and x 2 depend on y. The second term on the right side is negative for all y where d.1 > xl and the third term is negative for d. < x 2. Otherwise, both terms are zero. If all particle sizes are included in the integration (d.~ = 0 and d. larger than the largest char or fly ash particle in the stream), the only nonzero term on the right side of the equation is the first term, which represents the ratio of the total numbers of fly ash to char particles. In general, both f and s depend on initial char particle size y. However, the qualitative relationship between initial char and fly ash particle size distributions is conveniently illustrated by assuming that f, s, and the ratio are independent of initial char particle size. This means the char particles are homogeneous, or size-invariant, with respect to both their physical properties and their fragmentation properties. Under these assump-

(% Pc)/P,,

7 =PY

foda P

f3[1__$2 t

2(l-s)

( W a P c ) 1/3 Y

x2 = ~ p.

x 1 = sx 2 = psy,

otherwise

fa~ig( xly ) dx

+rain

tions,

and

yielding

x

177

(11) A linear transformation of this equation relates directly to the experimentally measured cumulative particle size distributions reported later in this article. This transformed expression is

Ka(d) =-fd n'a(x) dx d

- fo n ' a ( x ) d x -

fo na(x)dr,

(12)

where the integral from 0 to ~ is a constant (independent of d~). A similar expression defines the function K¢ for the char particles.

K (d)

--

fd ,,'c(y) dr

r = £ w~n:(y)dy- £ dnc(y ) dy.

(13)

These cumulative distributions are used in the discussion of the experiments that follows. Dependence of the fly ash psd on the extent of fragmentation ( f ) under these assumptions is straight forward; the ratio n ' a / f 3 is inde-

178

L. BAXTER

pendent of f when plotted as a function of the dimensionless size pd,. Dependence on mode of fragmentation (s) is more complex. Predicted cumulative fly ash psds are illustrated for two values of s with all other parameters constant in Fig. 2. Initial char psd and the predicted fly ash psd assuming no fragmentation ( f and s = 1) are also presented. These results assume an initial char psd similar to those reported below. Note that the cumulative fly ash psd can only exceed the char psd if the char fragments and that relatively small changes in the slope on this log-log plot represent large changes in the actual numbers of particles at a given size. Values of s close to unity produce fly ash particle concentrations that slightly exceed char particle concentrations at most diameters, and the features of the char psd are reflected in the fly ash psd, albeit displaced to the left due to combustion. As s decreases from a value of unity, the features of the char psd are smoothed in the fly ash psd and the slope of the fly ash psd becomes increasingly steeper. Values of s inferred from published literature are sometimes 0.001 or lower, but the trends remain the same. An experimental approach to determining s and f and an indication of their variation with size is presented in the following sections. Other limiting forms of Eq. 1, other than Eq. 11, are useful to consider. If the initial char particles are monodisperse (n'C is a Dirac

' ' '-'-.'.1

~

.......

rr

........

I

105

g 104

~

10 3

~- 10 2

~agmentation

~ 101 .....

"... " ~ , "...:"....~

. . . . . .

S=0.8

.......

s-. o. . ~. " -.. :.. ... ' , ~ ~.~,.

I

1

,

. . . . .

,,I

,

10

,

,,

,,-,',,1"". 100

Diameter(~m) Fig. 2. Predicted cumulative fly ash size distribution based on an initial char distribution similar to those of the fuels used in this investigation. The value of f is 2.5 for both values of s. The case of no fragmentation corresponds to both f and s = 1.

delta function), the integral on the left represents the total number of fly ash particles produced per char particle. Assuming Eq. 2 as the appropriate description for g, the total number of fly ash particles generated per char particle becomes (f3/2)[(1 + s)/s2]. This illustrates the dependence of the number of fly ash particles on the parameters f and s. For a given value of f, the number of fly ash particles increases dramatically as s decreases. (Recall that s is always less than or equal to 1.) Similarly, as f increases for a given value of s, the total number of fly ash particles increases as f3.

The valid range of the functions f and s illustrate that this theoretical approach is applicable only to ensembles of particles, not to individual particles. For example, when the ash in an individual particle coalesces to form a single fly ash particle ( f = 1), there is only one size fly ash particle produced (s = 1). However, only a fraction of the particles in an ensemble of nominally identical char particles may behave this way. Therefore, f can assume a value of unity for the ensemble without limiting the range of values s can assume. In the application of this theory in this article, the only theoretical limitations on these two parameters are that f > 1 and s < 1. The functions are independent of each other. EXPERIMENTAL APPROACH

The Multifuel Combustor (MFC) and a Particle Counter Sizer Velocimeter (PCSV) are the primary experimental facility and diagnostic, respectively, used in this examination. The MFC, schematically illustrated in Fig. 3, provides careful control of gas temperature and composition in a long-residence-time facility. The MFC was operated at a firing rate of 10 M W / m 2 with overall oxygen concentrations varying from 4 to 2 mol % during char oxidation, conditions similar to commercial-scale operation. Particle residence time is varied by moving the location of the coal injection lance to various ports along the length of the combustor. The PCSV performs in situ measurements of particle frequency (count), size, and speed at the base of the combustor, as indicated. These

C H A R F R A G M E N T A T I O N AND FLY ASH F O R M A T I O N

179

Natural Gas/ Gas

Coal Injection I

~

i

Burner ghtener

Laden

Flame

Stream Gas Flow

Heated/ Insulated

~ ~ ,

;% t:

::

1"

" ~ : i : i : Insulatingi V

pT~eo rm°c°uple

I" 15 cm"1 ~ " RefractoryLining

Layers . J r

_~e,, ;:% .°,%

PCSV f Laser Beam

~ ~ j pT. * , e. s t

section

• =.e.,

',:4","

to Exhaust

Fig. 3. Schematic diagram of the Multifuel Combustor (MFC) showing fuel ports, cross-section of a heated section, and overall construction. measurements are based on individual particles passing through an approximately ellipsoidal diagnostic volume with a major axis of about 1 mm and minor axes of approximately 300 p.m. The diagnostic is discussed more fully in the literature [21]. We determined the accuracy of the instrument by passing monosized standards through the diagnostic volume and comparing the measured particle sizes with the known actual sizes. Standards used included latex spheres entrained in air and transparent discs etched in an opaque reticule. The standards varied from 0.6 to 60 ~ m and indicate the PCSV reports particle sizes within one size bin or 0.1 p~m of the correct size, whichever is larger. We determined the precision of the data by repeating measurements of coal and fly ash in the MFC under nominally identical conditions

and computing coefficients of variation from the results. At least 5 and as many as 12 replicate experiments were performed for each coal at each sampling height. The coefficient of variation thus determined includes experimental uncertainty as well as actual fluctuations due, for example, to turbulent fluctuations and variations in coal feed rates. The impact of such fluctuations on the data can be minimized by computing the total mass flowrate from the measured particle concentrations and normalizing all the results to this value. Confidence intervals presented below are based on the conservative estimates of the coefficient of variation: those n o t accounting for the fluctuations. The upper and lower detection limits of the PCSV are determined in these experiments by the points at which the agreement among the

180

L. B A X T E R

repeated tests becomes poor, as indicated by the statistical confidence intervals. The lower limit is 0 . 5 - 0 . 7 / z m and is controlled primarily by signal-to-noise levels in the detector. The upper limit is controlled by sampling statistics in this experiment. Approximately 160,000 individual particles were sampled per measurement, and each measurement was repeated 5-12 times. Under these conditions, the upper detection limit is about 100/zm. Two bituminous coals (Illinois # 6 and Kentucky # 9 ) and a lignite (Beulah) with ash fractions of 0.10, 0.14 and 0.14, respectively, were examined in this investigation. The Kentucky # 9 and Illinois # 6 coals are similar, except for overall ash content. Beulah lignite contains about the same fraction of ash as the Kentucky # 9 coal but differs from both bituminous coals in most other respects. Utility-grind samples of all three coals were used. The nominal conditions under which these experiments were conducted are summarized in Table 1. The variations from run to run for a given coal type were approximately 7 relative percent from the values indicated. The variations from coal to coal were approximately 10 relative percent. Under these experimental conditions, the particles form a dilute phase in the gas, with the total particle volume fraction far less than 1% of the overall gas volume. Consequentially, the probability of particlewith-particle collisions producing agglomerated fly ash is vanishingly small.

exceeded 98% in all cases). Figure 4 illustrates the resulting cumulative distributions (see Eqs. 12 and 13) for both fly ash and char for each coal. Effects of different gas densities due to slightly differing gas temperatures are removed by normalizing all data to a standard gas temperature of 298 K. The fly ash size distributions of the bituminous coals are remarkably similar in both slope and functional features to those of the char. Fly ash concentrations exceed the char concentrations at all but the largest measured sizes, ,,,,, is

,

'1

,

,,,,

. . . . .

'1

''1

I

10 5 10 4 103 10 2 Char.___Flyash 101 ..... .....

0

I I

10 5 - ~ , .

',

l

:

I :l:l

I

I

. . . . . . . . . . . . . .

I I

Kentucky #9

~10 4 0 C 0

O 103

_¢ o

10 2

. . . . Flyash

¢1) >

101

RESULTS Char and fly ash particle size measurements were obtained for the three coals at the end of coal devolatilization and at the completion of heterogeneous combustion (overall burnout

E

0

. . . . .

10 s

103

Typical Local Gas and Particle Properties in the MFC at the Initial Char and the Fly Ash Measurement Locations

10 2

Property

101

Gas temperature (°C) Oxygen mole fraction Particle residence time z (s) Particle burnout (dad Particle temperature (°C)

Fly Ash

1500 0.04 0.2 0.65 1650

1000 0.03 2.1 0.99 + 1025

- ~

. . . . . . .

I

Beulah Ugnite

10 4

TABLE 1

Initial Char

I

. . . . Flyash

1

10 Particle Diameter (~m)

"'-

:~

100

Fig. 4. Experimentally measured cumulative char and fly ash size distributions for Illinois #6 and Kentucky #9 coals and Beulah lignite.

CHAR FRAGMENTATION AND FLY ASH FORMATION indicating that increases in particle number due to fragmentation offset decreases in particle diameter due to combustion. On the other hand, lignite fly ash particle concentrations are uniformly lower than char concentrations, indicating less significant fragmentation over the size range 0.6-100 /zm. The slope and functional features of the two distributions also differ significantly. Similarity of the initial char and fly ash concentration distributions for the bituminous coals suggests a narrow range of fly ash particle sizes generated per char particle (a value of s near unity), as illustrated in Fig. 3. The number and size of these fragments will be indicated by f and may vary with initial char particle size. Assuming a value of s of unity for all particle sizes, the size variation of f can be computed directly from these data using Eq. 1. Measured ash mass fractions in the coal were available at five particle sizes for all but the Kentucky #9 coal. These were used to approximate the size dependence of ¢.oa with size. Other physical parameters of importance inelude the density of the char and ash, which are assumed to be 0.7 and 2.5, respectively, independent of particle size. Figure 5 illustrates the results of such calculations for the three coals with respect to fly ash particle size. The extent of fragmentation, as indicated by f, is seen to vary with particle size for the high-rank coals. The variations are large compared with the confidence intervals, indicating that they are statistically significant. More fly ash particles with sizes of approximately 2 and 15 /zm are produced as a result of fragmentation than of other sizes for highrank coals. The size variation of f is minimal for the lignite. A value of f of unity reasonably approximates most of the lignite data, indicating minimal extent of fragmentation. These results can be used to estimate the number of fly ash particles produced per char particle and create a simpler model of fragmentation than is indicated in Eq. 1. Such a model may be useful for engineering calculations of fragmentation. One such model is postulated in which all fragments are assumed to have the same size (s = 1), and only the number of fly ash particles produced per original char particle need be specified. This quantity is

6

I

. . . .

181 ,

,

,

,

, , ,

lllinois #6

5

-Based on : Mean 4 •[ _ _ _ C o n f i d e n c e • Limits

95o,+>

.

.

.

.

.

.

.

I.

.

••

,' /¢'~, i'

: - -

3

s~'i~l

]

,

]

/

,' /

,'~'"

/

/ /

2

,~ ;

~.~•• ",~••

-\,,,

•1

1 0 v'

~" ~. _~

, , t l l , , , i

I

, ,

, ,

, +

, ,

, , , , , , , ,

i

Kentucky #9 Based on : Mean Confidence . . . . Limits (95%)

3

'

i

i

i i

'

'

'

' ' ' l

I I

I I

I I

I l

\

5

4

i

'

T





!

~ 2 $ 1 0 5

....

I

....

I

I

,

, ,,,,,I

. . . . . . .

I

I I

I I

I I

Beulah Lignite Based on : Mean Confidence . . . . Limits (95%)

4 3

2

I , " ' , ' , " . . . . . I-I 10 100 Flyash Particle Diameter (l~m) Fig. 5. Fly ash size dependence of the parameter f assuming s = 1 for Illinois # 6 and Kentucky # 9 coals and Beulah lignite. 0

.....

I

. . . . . . . .

1

shown in Fig. 6 as a function of initial char particle size. Variation in the number of fly ash particles produced per char particle with char particle size is large for the bituminous coals and less significant for the lignite. Extensive fragmentation is limited to the largest bituminous char particles. The number of particles generated per char particle is also slightly sensitive to ash loading. More fly ash particles are produced per char particle for the lower-ash Illinois # 6 than for the higher-ash Kentucky #9. Although these differences are apparent at

182

L. BAXTER

160 ,

. . . . . . . .

i

140 _~_

illinois#6

120 100 i 80 -

Based on : Mean Confidence ' " " - Limits

. . . . . .

~

f ~" 4 ,' //~ , , " /--~ ,-" / ] • I --I

...

:

}

60 40

~..

us BS



20 O

"~

0

o. 1 4 o

- K e n t u c k y #9

120

: : .-

~ 100

=_

,~

Based on : ~ Mean , Confidence - " " Limits

:j ,[

(95%)

,'/

-

--

###

••

40 20

0

i

i

i

i

i

i

ii

I

i

i

i

[

i

i

i

i

F

I

I

- Beulah Lignite

8

Based on : Mean . . . . Confidence Limits (95%)

6 4 2 0

I

1

I

I~T'~'1"l't~"

....

,

10 Initial Char Particle Diameter (~Lm)

I

100

Fig. 6. Number of fly ash particles generated per char particle as a function of initial char particle size for Illinois # 6 and Kentucky # 9 coals and Beulah lignite. Note the change of scale for the lignite.

all particle sizes, the confidence intervals overlap for the largest particle sizes, indicating that the difference is only statistically significant at lower particles sizes. DISCUSSION Variations in both the extent ( f ) and mode (s) of fragmentation with particle size cannot be determined uniquely from these data. However, these data do provide qualitative infor-

mation on the possible variations of s with size. The similarity in the fly ash and char psds from the bituminous coals indicates that s can only depart significantly from unity at the largest particle sizes; otherwise the slope of the fly ash distribution would be steeper. The data illustrated in Fig. 6 represent a lower bound for the number of fly ash particles produced per char particle at large initial char particle diameters and an upper bound at small initial char particle diameters. The number of fly ash particles produced per char particle at the largest initial char particle sizes increases by a factor of approximately (f3/2)[(1 + S)/S 2] as S decreases from unity. A decrease in s at these larger sizes must be associated with a decrease in f at the smaller sizes to satisfy the number balance. Therefore, if s is significantly less than unity, the result would be to increase further still the number of fly ash particles produced per char particle at large char particle sizes and to decrease the number at small particles sizes, thus amplifying the trend seen in Fig. 6 for the bituminous coals. By contrast, the lignite fly ash cumulative psd has a slightly steeper slope than the char, indicating a value of s of 0.1-0.5. The extent of fragmentation indicated by f is slight. Indeed, the value of f is slightly less than unity over some of the particle size range considered. These low values of f observed probably result from ash vaporization. Under these conditions, the ASTM ash fractions in the char used in analyzing the data are higher than the actual ash mass fraction in the particles, biasing the calculated values of f to smaller values. Analysis of the lignite fly ash indicates that up to 20% of the total ash may have vaporized. Cenosphere and plerosphere formation by fly ash would also be indicated by values of f less than unity, but formation of these fly ash structures typically is limited to a small fraction of the total number of particles, which would not significantly affect the average value for f. The measured differences in both the extent and mode of fragmentation with rank can be attributed to differences in char morphology. Bituminous coals form large voids and often char cenospheres during devolatilization (Fig.

CHAR FRAGMENTATION AND FLY ASH FORMATION

,res

pere

(a) =ores

(b) Fig. 7. Illustration of differences in char morphology for chars produced from (a) high-volatile bituminous coals and (b) lignites.

7a). As the organic portion of these cenospheres is removed by heterogeneous reaction, they become increasingly friable and ultimately collapse into a number of fragments. Lignite char morphology differs markedly from that of the bituminous coals. Lignite char resembles a partially delaminated structure (Fig. 7b), with approximately planar and parallel voids running through the organic matrix. Under conditions similar to boiler operations (2%-4% oxygen), heterogeneous combustion tends to be strongly influenced by oxygen transport to the particle surface, and chars tend to burn out with nearly constant density. This mode of burning does not increase the friability of a lignite char. However, unlike the high-rank coals, lignites produce significant numbers of ash droplets on the char surface during oxidation, as has been reported by others and confirmed by SEM micrographs of partially oxidized samples of these coals. The observed slight fragmentation of lignites can be attributed to shedding of these droplets, a

183

fundamentally different mechanism than the collapse of char structure involved with highrank coal fragmentation. These data help resolve some of the discrepant reports of the extent of fragmentation in the literature. Those who have reported extensive fragmentation have almost exclusively worked with large particles. Our data indicate that such large particles will potentially form large numbers of fragments. Those who report insignificant fragmentation have typically based their findings on overall number balances using utility-grind coal. Since the large particles make up a very small fraction of the number of particles in these coals, they do not substantially affect the number balance even when they fragment extensively. For example, the data presented here indicate that fragmentation of bituminous coal produces approximately two to five fly ash particles per char particle based on a total number balance, even though the largest char particles form hundreds of fragments. Based on these analyses, the apparent discrepancies in past studies may be at least partially due to this size dependence.

CONCLUSIONS A general theoretical model relating fly ash size distributions to initial char size distributions has been developed and illustrated. This model demonstrates that, on the basis of experimentally measured parameters, the size distribution of fly ash generated from broad initial char distributions is sensitive to both the extent and mechanism of fragmentation. Specific relationships are illustrated for the case when the extent and mechanism of char fragmentation is independent of char particle size. Experimental data are presented for two bituminous coals and a lignite and indicate a strong rank dependence of both the mechanism and extent of char fragmentation. Bituminous coals fragment more extensively than the lignite, with the extent of fragmentation exhibiting a strong particle size dependence and a weaker ash loading dependence. The number of fragments produced per original bituminous char particle varies from 1 to 10 for initial char

184 particle sizes less than 20 /.Lm to over 100, or possibly several hundred, for initial char particle sizes greater than 80 p~m. This measured size dependence may partially explain the apparent discrepancies in previous fragmentation studies. Lignite fragmentation is far less extensive and does not vary as dramatically with initial char particle size. The mechanism of fragmentation also differs, producing, on average, a broader distribution of fragment sizes from the largest char particles. Differences in the observed extents of fragmentation as a function of coal rank are attributed to measured differences in char structure.

The author gratefully acknowledges the assistance of Ephraim Arquitola in operating the MFC and in collecting the particle concentration versus size data. This work was sponsored by the U.S. Department of Energy through the Pittsburgh Energy Technology Center's Direct Utilization Advanced Research and Technology Development Program.

L. BAXTER

5. 6.

7. 8. 9. 10. 11.

12. 13. 14. 15. 16.

17.

REFERENCES 1. Wark, K., and Warner, C. F., Air Pollution, 2nd ed., Harper & Row, New York, chap. 5, 1981. 2. Quann, R. J., Neville, M., Janghorbani, M., Mims, C. A., and Sarofim, A. F., Era,iron. Sci. Technol. 16:776-781 (1982). 3. Neville, M., Quann, R. J., Haynes, B. S., and Sarofim, A. F., Eighteenth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1981, pp. 1267-1274. 4. Quann, R. J., and Sarofim, A. F., Nineteenth Sympo-

18. 19. 20. 21. 22.

sium (International) on Combustion, The Combustion Institute, Pittsburgh, 1982, pp. 1429-1440. Flagan, R. C., and Sarofim, A. F., Prog. Ener. Combust. Sci. 10:159-175 (1984). Helble, J., Neville, M., and Sarofim, A. F., Twenty-First Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1988, pp. 411-417. Baxter, L. L., and Mitchell, R. E., Combust. Flame 88:1-14 (1992). Raask, E., Mineral Impurities in Coal Combustion, Hemisphere, New York, 1985, chap. 7. Baxter, L. L., and Hardesty, D. R., SAND89-8201, Sandia National Laboratories, January 1989. Allen, R. M., and VanderSande, J. B., Fuel 63:24-29 (1984). Allen, R. M., and Mitchell, R. E., Proceedings of the 1985 International Conference on Coal Science, The International Energy Agency, 1985, pp. 401-404. Helble, J. J., and Sarofim, A. F., Combust. Flame 76:183-196 (1989). Dunn-Rankin, D., Combust. Sci. Technol. 58:297-314 (1988). Dunn-Rankin, D., and Kerstein, A. R., Combust. Flame 74:207-218 (1988). Kerstein, A. R., and Edwards, B. F., Chem. Eng. Sci. 42:1629-1634 (1987). Kang, S-G, Helble, J., Sarofim, A. F., and BeEr, J., Twenty-Second Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1989, pp. 231-238. Wibberly, L. J., and Wall, T. F., Combust. Sci. Technol. 48:177-190 (1986). Holve, D. J., Combust. Sci. Technol. 44:269-288 (1986). Sarofim, A. F., Howard, J. B., and Padia, A. S., Combust. Sci. Technol. 16:187-204 (1977). Quann, R. J., and Sarofim, A. F., Fuel 65:40-46 (1986). Holve, D. J., and Self, S. A., Appl. Opt. 18:1632-1645 (1979). Canavos, G. C., Applied Probability and Statistical Methods, Little, Brown, Boston, 1984.

Received 27 March 1990; revised 16 March 1992