Fuel Processing Technology 143 (2016) 7–17

Contents lists available at ScienceDirect

Fuel Processing Technology

journal homepage: www.elsevier.com/locate/fuproc

Modeling of unburned carbon in fly ash and importance of

size parameters
Mehmet Bilen ⁎, Sait Kizgut
Faculty of Engineering, Zonguldak, Bülent Ecevit University, 67100 Zonguldak, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: Unburned carbon (UBC) percentage in coal ash can be considered as an economic and environmental constraint
Received 7 August 2015 since it is not only an important characteristic in terms of combustion efficiency but also it becomes more of an
Received in revised form 6 October 2015 issue of further utilization of ash. In this study, particle size distribution (PSD) effect on UBC in fly ash (FA) was
Accepted 27 October 2015
discussed and efficient size parameter of PSD in terms of UBC formation was determined. At the first stage of
Available online 19 November 2015
the study, regarding ash formation, percentage of any set of coal particles which are more likely to form fly ash
Keywords:
(FA) was theoretically determined. For this purpose, a formulation including the size parameters was proposed.
Unburned carbon Secondly, with the help of a 3-day long systematical control of power, the effect of size distribution on UBC was
Pulverized coal size obtained. The size distributions of these samples showed that an increase in D32 size parameter causes an in-
Size parameters of PSD crease in the amount of UBC as well. Relationships between the amount of UBC and mean values of size param-
Fly ash eters (D10, D50, D90, D32, D43) were all statistically meaningful for both units of power plant concerned but D32 size
parameter has a better fit among all the size parameters. In summary, D32 & UBC have a regression coefficient of
0.90 (R2 ~ 0.9), while other size parameters (D10, D50, D90, D43) & UBC have regression coefficients of 0.2 (R2 ~ 0.2)
for both units. Finally, a model including these size parameters was proposed to predict UBC in FA. The model
proposed was in good agreement with the measured UBC in FA.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction
Decreasing the amount of UBC have been gaining more importance
as new technologies about thermal power plants develop. UBC content
in FA usually is in the range of 2–12%, but in some cases it can be up to
20% [3] even 23% [17]. UBC is not only the measure of combustion effi-
ciency but also a measure of possible usage of FA in the cement industry.
Possible usage of FA in Turkey is crucial as 29% of the total electric ener-
gy demand is supplied by thermal power plants [1]. Therefore, Turkey
has a significant fly ash potential. As an example, Afsin Elbistan power
plant only consumes 18.0 × 106 metric tonnes of coal per year and gen-
erates about 3.24 × 106 metric tons of FA itself [21]. Considering the last
5-year average, it can be estimated that almost 20 million tons of ash
most of which is FA has been generated in Turkey and this amount is
expected to increase in the near future. According to Acar and Atalay's
[2] study, worldwide annual production of coal FA is estimated around
500 million tones, in total.
Minerals in the coal transforms into ash while coal in a pulverized
coal (PC) combustion undergoes two different conversion steps, i.e. py-
rolysis and char oxidation. Char oxidation is the rate limiting step and
determines the carbon conversion and the ash formation. Ash with
small amount of residual carbon (UBC) is formed after these conversion
⁎ Corresponding author.
E-mail address: mehmetubilen@yandex.com (M. Bilen).
steps [13]. Porous structure of coal has significant influence on the forma-
tion and the characteristics of ash [6,13,22,25]. Char fragmentation and
ash formation mechanism proposed by Wu et al., [22] is given in Fig. 1.
Referring to Fig. 1, fragmentation of highly porous char results in fine
ash particles while chars with low porosity result in more compact and
coarser ash particles. The effect of particle size of PC on ash formation
keeps its complexity. The expected and mostly encountered fact is
that the larger the particle size of coal the larger the ash size and more
likely to form BA, and if the coal size is finer formation of FA is favored.
This may not always be the case, since larger particles with higher po-
rosity and higher internal and external surface area may breakdown
and formation of FA is favored. This is also supported by Baxter's [5]
study which claims the fact that large char particles have much higher
tendency to fragmentation than small particles. Since the PSD of FA of
many bituminous coals is generally less than 75 μm [3], formation of
these fine particulates may be mostly due to breaking down of large
coal chars. In order to understand FA formation clearly not only PSD of
Pulverized Coal (PC) and FA but also size means, such as Sauter mean,
should be considered. Sauter mean was taken into account by Senneca's
study [17] in which Sauter mean was only observed for FA samples and
a change between 10 and 27 μm was reported.
The amount of UBC is high due to low oxygen and low temperature
combustion conditions to meet emission requirements of NOx [3] most
of the time, however the role of non-ideal PSD of PC on UBC should not
be underestimated. Many factors affect the combustion efficiency in

http://dx.doi.org/10.1016/j.fuproc.2015.10.039
0378-3820/© 2015 Elsevier B.V. All rights reserved.
8 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17
Fig. 1. Char fragmentation and ash formation mechanism proposed by [22].
pulverized fuel boilers. Most effective factors on UBC, i.e. combustion ef-
ficiency, are residence time, oxygen feed and the coal properties. Coal
properties were studied by Bilen et al. [7] who discussed the effects of
moisture content and PSD of coal on UBC in BA. According to Xue et al.
[23] coarse fraction of 110–480 μm is important for an efficient PC burn-
out in coal-fired power plant boilers. Jiménez and Ballester [12]
discussed the factors influencing the evolution of particle size during
combustion of PC, as well as their consequences for the interpretation
of burnout curves. They reported that at early stages of combustion par-
ticles (53–63 μm) break into fragments (20–40 μm). Ganguli and
Bandopadhyay [10] claimed that there is negligible correlation between
PSD of PC and efficiency. Although Ganguli and Bandopadhyay [10]
claimed no correlation, it seems that they failed to notice the strong re-
lationship between PSD and efficiency (UBC). Regarding the findings of
their study, while PSD 76 of PC decreases from 81% to 46%, UBC in FA de-
creases from 3.6% to 0.8%. The effect of PSD on UBC was also revealed by
Gao et al. [9] that particles larger than 140 μm have contribution of 70%
on total UBC although fractional mass ratio of these particles is about
Fig. 2. Schematic representation of sampling. (Representation of boiler is adapted from the study of Li et al. [14]).
20%. Particle size is also important in terms of heating rate since maxi-
mum heating rate decreases linearly with increasing particle size as re-
ported by [24] who claimed that the combustion time increases with
increasing particle sizes of biomass from 10 μm to 20 mm. Regarding
the combustion time, Atas et al. [4] emphasized the fact that finer coal
particles burn more rapidly, and fineness of coal directly reduces the
UBC in FA and BA.
Bilen et al. [7] proposed a model to predict UBC in BA which employs
D90 size and moisture content of feed coal and it was in good agreement
with the experimental findings. The achievement of aforesaid correla-
tion simply is a proof to the importance of PSD in terms of UBC either
in BA or in FA. In this study, the effect of PSD of PC on UBC in FA was
discussed. Size parameters were introduced in FA formation and related
formulation was proposed. PSD effect was also investigated with size
parameters and amount of UBC in FA was correlated with these size
parameters.
2. Materials and experimental procedures
2.1. Combustion units and sample collection
The samples were taken from a local power plant in Zonguldak,
Turkey. Schematic representation of these studied samples is given in
Fig. 2. Unit 1 and Unit 2, which the samples were provided from have
the same type of boilers operated at combustion zone temperature of
around 1200 °C. The stockyard coal sample was taken from a huge
pile feeding both units. This sample was taken once and proximate anal-
ysis was carried out just to have a general idea about the coal character-
istics and boiler design requirements. All other samples (PC samples)
were taken from the burner points shown in Fig. 2. A total of 801 PC
samples, 369 from Unit 1 and 432 from Unit 2, were collected for nine
shift periods. The number of PC samples for each shift was 48, PC sam-
ples were taken from mill exits twice at each shift. Since there are four
exits of each mill and the plant has six mills a total of 48 PC samples
were gathered for each shift for each unit. Representation of the com-
bustion environment in terms of size distribution of PC is provided by
such a large number of samples.
2.2. Characterization of pulverized coal samples
Sieve analysis of PC samples was done by using Malvern Mastersizer
S 2000 utilizing wet method and PSD of PC samples was obtained over a
range of 0.05 μm–878.67 μm. Refractive index of water and PC was set to
1.33 and 1.64, respectively. Obscuration of the Mastersizer experiments
 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17 9

was kept between 12% and 18%, mostly about 15%. Prepared coal sam-
ples of different shifts of 3 days of operation of power plant were ana-
lyzed and size distribution parameters (D10, D50, D90, D32, D43) for two
units of power plant were obtained. Size parameters and their descrip-
tions are as the following.
D10 (μm): sieve opening which 10% of particles passing through
D50 (μm): sieve opening which 50% of particles passing through
D90 (μm): sieve opening which 90% of particles passing through
D32 (μm): volume/surface mean (Sauter mean).
Sauter mean is defined as the diameter of a sphere that has the same
volume/surface ratio as a particle of interest.
Xn
Di 3 ν i
D32 ¼ X1n
1
Di 2 ν i
obtained from the ESP design. Since the FA from the plant is marketed
to cement producers, design capacities of each row could be double
checked with the marketed amount and it was found that the marketed
amount was compatible with the designed amount. The capacities of
the ESP rows are given in Table 1. By mixing eight samples of FA from
each ESP row, a total of nine combined FA sample for each unit was ob-
tained. Then, the FA samples were analyzed in terms of UBC content.
The FA samples were firstly dried at 105 °C to remove the incidental
moisture content. The samples do not need to be ground to pass
through a 200 (units) mesh since PSD of the most bituminous coal FA
is generally less than 75 μm [3]. About 1 g of FA sample was heated at
500 °C for 1 h and at 850 °C for 2 h. The decrease in mass of FA samples
showed the Loss-on-ignition (LOI) [18]. The ratio of the burned amount
of 1 g sample was accepted as the percentage of UBC. LOI is accepted as
unburned combustible carbon (UCC) content of ash samples.
ð1Þ
3. Fly ash formation and models

D43 (μm): the mean diameter over volume (DeBrouckere mean)

3.1. Fly ash formation
Xn
Di 4 ν i
D43 ¼ X1n ð2Þ The ash particles which neither melt nor agglomerate inside the PC
1
Di 3 ν i boiler are mostly FA with finer in size but heavier in bulk than BA.
Some coal particles are already coarse and take place in the formation
If we assign 3 spheres with diameter 1, 2, 3 units, the calculation of of BA, while some others are already fine and take place in the formation
Sauter and De-Brouckere means of these spheres is exemplified as in of FA. In terms of FA formation, surface area and mass of PC particle and
the following equations (Eqs. (3) and (4)). temperature inside the boiler are the significant factors. Surface of a coal
particle is directly proportional to the 2nd power of the dimension
1 þ 8 þ 27 while the volume of it is the function of the 3rd power. Likewise, it
D32 ¼ ¼ 2:57 μm ð3Þ
1þ2þ3 can be assumed that the 4th power could be related to the mass of the
particles. Particles in the boiler either fly or fall due to gravity. As the
1 þ 16 þ 81 simplest approach, particles more likely to fall are usually heavy with
D43 ¼ ¼ 2:72 μm ð4Þ
1 þ 8 þ 27 small surface areas. The surface area of a particle is important since
the drag force is dependent on the surface area. The particles in combus-
2.3. Characterization of FA samples tion zone cannot be expected to behave like the particles in normal con-
ditions, since reactions follow each other in milliseconds. However,
FA samples were taken from the plant to observe the effect of PSD of there must be a method to understand whether a set of coal particles
PC samples on UBC. In Fig. 3, the method of collection of FA samples is cause formation of mostly FA or BA. Therefore, we proposed the
presented. The FA samples were taken 3–4 h after the PC samples had DeBrouckere Mean over Sauter Mean (DBMSM) in this study, which is
been collected in order to ensure that the corresponded PC combusted the comparison of the DeBrouckere Mean and Sauter Mean of the parti-
to represent the FA. FA samples were collected twice for each shift cles. Formulation of DBMSM is as the following:
from each row of the ESP unit. Depending on the capacities of each
row, a total of 8 samples from each unit in each shift were combined D43
DBMSM ¼ ð5Þ
to a single FA sample. Each FA sample from each row has different D32
amounts of UBC content; that is why all the FA samples were mixed in Xn
the order of the capacities of the rows and UBC in FA was calculated as Di 4 ν i
X1n
weighted average of the ESP rows. Individual row capacities were Di 3 ν i
DBMSM ¼ X1n 3 ð6Þ
Di ν i
X1n 2
D νi
1 i

Since the percentage of the particles play role in FA formation, it

should be between 0 and 100. Ash Defining Factor (ADF) can be calcu-
lated as follows:

ADFðFAÞ ¼ ð100−DBMSMÞ ð7Þ

ADFðBAÞ ¼ DBMSM ð8Þ

Table 1
Capacities of the ESP rows (same for Unit 1 and Unit 2).

ESP rows Mass flow (t/h) %

Row 1 23.80 85.83

Row 2 3.35 12.09
Row 3 0.50 1.80
Row 4 0.076 0.28
Total 27.73 100.00
Fig. 3. The method of collection of the FA samples.
10 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17
Fig. 4. The representation of the fragmentation model by Holzleithner [11].
where ADF(FA) stands for ash defining factor for FA and ADF(BA) stands
for ash defining factor for BA.
At normal conditions about 80–90% of the ash is present as FA and
that is why ADF should favor the FA formation. Regarding the formula-
tion of ADF(FA) the value changes between 0 to 100, which would be 80
or 90 at normal conditions. ADF(FA) can be better approximated with
the real life FA production ratio. In this study, ADF was calculated by
shift based D43 ratio to D32.
3.2. Fragmentation model
The fragmentation model of coal particles in slowly moving fixed
beds was discussed by Holzleithner et al. [11]. Since fragmentation is a
governing factor for the chemical processes like drying, devolatilization
and gasification [11], it can be claimed that fragmentation is also a
governing factor in combustion. Fragmentation model is based on PC
combustion models as previously suggested by Mitchell [16] and Shah
et al. [19], which was also employed in Bilen et al.'s study [7]. The six
major factors which are supposed to be the driving forces for the frag-
mentation of coal particles inside the slowly moving fixed bed gasifier
explained by Holzleithner et al. [11] can still be used for the PC boilers.
These factors are:
• Solids pressure
• Shear stress
• Rate of heating
• Rate of drying
• Rate of devolatilization
• Rate of gasification
In addition to the major factors above the rate of combustion can
also be considered as a major factor for the fragmentation of PC parti-
cles. The representation of the fragmentation model by Holzleithner
[11] is given in Fig. 4.
The rate of change of the particle i due to fragmentation is generally
assumed to be proportional to the number of particles in each size class
at the time t on a certain position on a distinct stream line. The numer-
ical distribution of a particle size class i into smaller particle size classes j
due to the processes k is given by the elements of the fragmentation ma-
trix bk.i,j The term Sk,j is the fragmentation rate constant for each process
k and each size class i. The variable Kk contains the driving force of the
process k. For example, Kk for the process ‘heating rate’ is the rate of
temperature change of the particle in K/s. The variable Cj accounts for
the reduction of the particle size due to gasification. The fragmentation
matrices bki,j and the fragmentation rate constants Sk,j are determined by
suitable experiments [11]. By involving the rate of combustion in frag-
mentation model, the variable Kk which contains driving forces of the
process k would be 7, instead of 6, just for gasification. And this time
Kk for the process ‘rate of combustion’ is the rate of conversion coal to
ash, i. e. mass/time. Driving factors of gasification are all valid for the
modeling of PC fragmentation but ‘rate of combustion’ should be
noted as well.
This model would give the most accurate results if each process con-
stant is evaluated in a proper way. Obtaining the ash size after the pro-
cesses like heating, drying, devolatilization and combustion would let
one to obtain the amount of UBC. However, execution of this model
would be time consuming since 7 major driving factors involved
would make the problem even more complex.
3.3. Combustion model
As an alternative to fragmentation and population model, combus-
tion model employing size distribution of PC was proposed in terms of
UBC prediction. The model was obtained by manipulating a previously
suggested [8] grinding model. In terms of size distribution, production
of PC in the mills of a power plant can be considered to be similar to
the production of ash and combustion of PC. According to the study of
Liakos et al., [15] for PC combustion there are three controlling mecha-
nisms which are chemical reaction controlling mechanism, diffusion
controlling mechanism and combination of these two. These mecha-
nisms control the UBC fraction and combustion efficiency as well. The
mechanisms are affected by gas temperature, volume of CO2, diffusion
combustion rate, chemical reaction rate and particle size change.
Boiler performance and combustion efficiency is strongly related to
the particle size of feed PC since it directly designates the controlling
mechanisms. Grinding process is similar to combustion process where
initial energy is given to volume and breakage occurs. During steady
state combustion of PC particles, i size material fed and i size material
combusted and i size material is becoming ash. Thus, modification of
the grinding model from the study of Bwalya and Moys [8] can be
used as a combustion model and the steady state combustion process
for a specific size fraction (i) can be stated as:
“Rate of combusted i-size material in product = rate of addition of
i-size material in feed + rate of production of i-size material by com-
bustion of all larger size − rate of combustion of i-size material”.
According to the study of Bwalya and Moys [8], the steady state mill-
ing process for a specific size fraction i can be written as shown in the
Table 2
Proximate and petrographic analyses of stockyard coal sample.
Proximate analysis Petrographic Analysis
Constituent Maceral group composition
(wt.%) (vol%, mmf)
Ash VM FC S GCV V L I SI Rm
9.57 37.14 52.91 0.38 6640 78.90 0.81 9.74 10.55 0.569
VM: Volatile matter, FC: Fixed carbon, S: Total sulfur, CV: Gross Calorific value (Kcal/kg),
V: Vitrinites, L: Liptinites, I: Inertinites, SI: Semi-inertinites, Rm: Mean maximum vitrinite
reflectance.
 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17 11

Table 3 and breakage function is:

Float and Sink analysis of stockyard coal sample.
  X −1β
Density Product X −1 γ 
Bij ¼ ϕ i þ 1−ϕ j i
ð11Þ
(g/cm3) Xj Xj
Amount (%) Ash (%)

−1.60 96.08 6.90

where pi is the fraction of the product which is within size class i, F is the
+1.60–1.90 0.16 36.99
+1.90 3.76 78.47 feed rate, fi is the fraction of the feed that is within size class i, W is the
Total 100.00 9.64 weight of the material that is present in the mill, while wi represents the
fraction in the mill load in the size class i. Si is the breakage rate and bij
gives the fraction that breaks into size class i when breakage occurs in
Table 4 size class j.
Ground coal size distribution parameters for Unit 1 for the period of 9 shifts. The combustion model, manipulated version of the grinding
Unit no. Shift no. D10 (μm) D50 (μm) D90 (μm) D32 (μm) D43 (μm) model by Bwalya and Moys [8], is composed of selection function and
‘combustion function’ instead of breakage function and it is represented
Unit 1 1st shift 6.38 42.06 114.44 6.02 52.72
2nd shift 8.21 50.49 125.60 8.65 60.28 as in the Eqs. (12), (13) and (14).
3rd shift 9.16 56.79 141.30 9.26 67.56
Xi−1
4th shift 8.74 57.34 147.95 9.20 69.14 pi F ¼ f i F þ W cij S j w j −Si wi W ð12Þ
5th shift 8.60 56.19 141.33 9.27 67.21 iN1
6th shift 9.51 59.70 151.56 17.19 71.75
7th shift 10.99 63.88 155.97 10.15 75.24 where selection function:
8th shift 11.64 64.11 159.15 11.99 76.42
9th shift 15.99 81.17 191.60 15.72 94.32 axi
Si ¼   ð13Þ
x
1þ i
μ
following (Eqs. (9), (10) and (11)) in terms of selection and breakage
function; and breakage function:
Xi−1   X −1β
pi F ¼ f i F þ W bij S j w j −Si wi W ð9Þ X −1 γ 
iN1 C ij ¼ ϕ i þ 1−ϕ j i
ð14Þ
Xj Xj
where selection function is:
where pi is the fraction of the product which is within size class i, F is the
axi feed rate, fi is the fraction of the feed that is within size class i, W is the
Si ¼   ð10Þ
x weight of the material that is present in the boiler, while wi represents
1þ i
μ the fraction in the boiler load in the size class i. Si is the combustion

Fig. 5. Changes on size parameters for 9 shifts of Unit 1 (blue lines show the average for each shift and PC stands for pulverized coal).
12 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17

Table 5 Boiler Population Model for UBC in FA;

Ground coal size distribution parameters for Unit 2 for the period of 9 shifts.

Unit no. Shift no. D10 (μm) D50 (μm) D90 (μm) D32 (μm) D43 (μm) dS=dt ¼ kB S :
Unit 2 1st shift 9.09 56.19 137.88 9.01 66.42 ðwhich S is the average size t is the time and kB is constant for BPMFAÞ
2nd shift 7.69 50.13 125.57 10.92 59.72 ð15Þ
3rd shift 9.08 51.18 123.84 11.13 60.23
4th shift 11.48 61.00 144.42 9.49 70.95
5th shift 10.27 58.26 136.05 9.65 67.23 Analytical solution to this differential equation is;
6th shift 8.50 47.21 118.18 8.67 56.65
7th shift 10.44 57.16 134.99 14.16 66.35 SðtÞ ¼ S0  exp:ð−kB tÞ ð16Þ
8th shift 10.81 61.22 140.12 10.96 69.71
where kB constant is taken as 1:21639 and t as 3:33 s:
9th shift 13.52 64.20 147.53 13.17 73.76

It should be noted that ash formation constant is approximated as in

rate and cij gives the fraction that combusts into size class i when com-
bustion occurs in size class j.
Calculation of ash fraction within a presumed set of ash particles in
the same size fraction can be realized with the execution of combustion
model. Obtaining the ash size distribution might let one to predict UBC
with valid assumptions. This model can easily be used however the con-
stants in the selection and combustion functions should be experimen-
tally obtained before the estimation of ash size distribution and UBC.
3.4. Population model
The Boiler Population model was suggested by Bilen et al. [7] in
order to estimate ash size distribution and UBC content in BA. The
abovementioned method is successful and applicable which a modified
version that can be suggested for UBC in FA. The Boiler Population
model which was previously used for the prediction of UBC in BA is
also used for the prediction of UBC in FA and it is represented as in the
following Eq. (15).
the previous study. This constant, kB can further be approximated and
manipulated in the order of boiler conditions with the better under-
standing of ash formation and FA formation.
Conversion of ash size into UBC content can still be used for the pre-
diction of UBC in FA similar to the previous study about UBC in BA. Some
manipulations had been applied and the proposed formulation for UBC
in FA is as follows:
3
UBC content in FA ð%Þ ¼ ADFðFAÞ4  ðDxx size of ashÞ  ð6:6
 10−7  ðOPCÞ: ð17Þ
ADF(FA) is predicted on a shift based with the data of PC size param-
eters, i.e. D43 and D32.
OPC: Operational constant; varies with the restriction of emission or
other factors shift by shift.
In the order of the restriction of emissions or other constraints in
similar UBC content in FA, as Ahmaruzzaman [3] mentioned, increases
up to 20%. This ‘OPC’ constant varies with these conditions and depend-
ing on the high or low UBC, OPC can be approximated. Verification of

Fig. 6. Changes on the size parameters for 9 shifts of Unit 2 (blue lines show the average for each shift and PC stands for pulverized coal).
 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17 13

Table 6 petrographical studies showed that the coal from stockyard is one
Amount of UBC for 9 shifts of Unit 1 and Unit 2. type i.e. not a blend. The amount of heavy fraction in the coal (less
Shift no. UBC (%) than 4%), found by the float and sink analysis (Table 3), revealed that
Unit 1 Unit 2
the stockyard coal had been passed through a coal preparation plant
and ash making materials had been removed.
1st shift 1.53 1.44
2nd shift 3.26 2.50
3rd shift 2.57 3.08 4.2. PSD results for unit 1
4th shift 3.21 2.68
5th shift 3.24 1.79 The average values of the size parameters (D10, D50, D90, D32, D43) for
6th shift 5.20 1.50 each shift (about 48 samples) were obtained for both units. For Unit 1, in
7th shift 3.13 4.84
8th shift 4.11 3.27
some of the shifts 48 PC samples were collected. However, due to mill
9th shift 4.72 4.58 bypasses less number of samples collected for the 3rd shift in particular
as well as the 6th, 7th, 8th and 9th shifts. That is why a total of 369 coal
samples were collected for Unit 1. The values of the size parameters are
the approximated OPC can be justified with the validity of the predic- provided in Table 4 and the distribution of each size parameter with re-
tion of UBC in FA. spect to each PC sample is provided in Fig. 5.
Since multiple size parameters (i.e. D10, D50, D90, D32, D43) are eval- Referring to Fig. 5, a significant overdispersion was observed for the
uated, all of the results of size distribution can be substituted for Dxx D32 size parameter for Unit 1 and referring to Table 4 ground coal PSDs
in Eq. (17). This would lead an opportunity to clarify which size increased with the following shift except the 4th and 5th ones. The D10
parameter is more related to UBC. Having clarified the most effective size parameter of PSD changed from 6.38 μm to 15.99 μm which is al-
size parameter on UBC, prediction of UBC in FA can be realized easily most more than twice. The change in the D50 size parameter was be-
with that specific size distribution parameter. tween 42.06 μm and 81.17 μm. The same trend was observed for the
In this study FA samples were analyzed in terms of UBC while PC D90 size parameter and it increased from 114.44 μm to 191.6 μm. This in-
samples were analyzed in terms of size distribution parameters. crease is not such significant like the D10. Change in the D32 somehow
Correlation between these two was established and confidence of the showed the same linearity and the values of the D32 were very close
correlation was provided with the use of vast number of the samples. to the values of the D10. The lowest D32 value was 6.02 μm and the
Among the models (fragmentation, combustion, population), the popu- highest was 15.72 μm. The D43 size parameter revealed an increase
lation model due to its ease of applicability and success in obtaining from 52.72 μm to 94.32 μm which had almost a similar increase percent-
good correlation is suggested [7]. age like the D90 and D50.
Abnormal increases in each size parameter were reasoned by any
4. Results and discussion disturbance on grinding. Considering the results, there might have
been disturbances at the 3rd and the 9th shifts. However, the results
4.1. Experimental results of the size parameters show an increased level of disturbance for all
the 9 shifts for Unit 1. This situation might be attributed to the changing
Proximate and petrographic analyses were carried out on the stock- capacities of the mills. Along with the stable operation with no mill by-
yard coal and the results are given in Table 2. Considering vitrinite re- pass, a total of 432 PC sample would have been collected, however, only
flectance value of 0.57% and volatile matter of 37.1%, and referring to 369 samples could be collected.
Table 2, the coal was ranked as sub-bituminous A coal. With its low
ash (9.6%) and low sulfur (0.3) contents and a calorific value of 6640 4.3. PSD results for the unit 2
kcal/kg it can also be regarded as a suitable fuel for power generation
[20]. Inert and semi-inert components account only for 20%. Taking For Unit 2, a total of 48 PC samples were collected for each shift, and
into account the reactive character of some semi-inert constituents, mill bypass was not experienced. Thus, a total of 432 coal samples could
the amount of reactive constituents is about 85%. In this context, the be collected and analyzed for Unit 2. The values of the size parameters
petrographic composition is also an indication of its reactive behavior are provided in Table 5 and the distribution of the each size parameter
in combustion. Reflectance measurements performed in the context of with respect to each PC sample is provided in Fig. 6.

Fig. 7. Size parameter evaluations in terms of UBC content (%) for Unit 1 (PC stands for pulverized coal).
14 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17

Fig. 8. Size parameter evaluations in terms of UBC content (%) for Unit 2 (PC stands for pulverized coal).

Referring to Fig. 6, a significant overdispersion was observed for the
D32 as in the case for the Unit 1 and for the D43 size parameter for some
extent. Referring to Table 5, ground coal PSDs showed a random behav-
ior. The D10 size parameter of PSD changed from 7.69 μm to 13.52 μm;
which can be inferred as almost a 75% increase was realized. The lowest
D10 size parameters were for the 2nd shift and the highest ones were for
the 9th shift. The D50 size parameter increased from 47.21 to 64.2. For
the D50 size parameter, the 6th shift is the lowest one. Comparing Unit
1 and Unit 2, respectively closer range for the D90 size parameter was
evaluated. The D90 size parameter increased from 118.18 μm to
147.53 μm, the lowest at the 6th shift and the highest at the 9th shift.
For Unit 2 there is a comparable difference between the D10 and D32, un-
like Unit 1. The range for the D32 (8.67 μm–14.16 μm) somehow showed
almost the same increase in percentage from the lowest to the highest.
It is interesting that the result of the 6th shift was the lowest whereas it
was the highest for the 7th shift. Regarding the D43 size parameters, the
highest (73.76 μm) was not so high for Unit 2 unlike Unit 1 (94.32 μm)
and the lowest (56.65 μm) was not so low unlike Unit 1 (52.72 μm) as
well. The highest D43 size value was evaluated in the 9th shift and the
lowest one was in the 6th shift. Abnormal increases were only observed
between the 8th shift to the 9th for the D10 and between the 6th shift to
the 7th shift for the other size parameters.
Considering the results, it can be inferred that there might have been
disturbances at the 6th and 8th shifts. The results of the size parameters
show random distribution starting from the 1st shift to the 9th for Unit
2. This might be attributed to many factors such as coal properties and
properties of the grinding environment. Along with the stable operation
with no mill bypass, with no significant changes of coal properties and
with no wear of the mills or classifiers, this random size parameter
distribution might not have been experienced. The main reason of

Table 7
ADF(FA) values of Unit 1 and Unit 2.

1st shift 2nd shift 3rd shift 4th shift 5th shift 6th shift 7th shift 8th shift 9th shift

ADF(FA) 91 93 93 92 93 96 93 94 94
Unit 1
ADF(FA) 93 95 95 93 93 93 95 94 94
Unit 2

Table 8
Size parameters of pulverized coal samples (μm) and predicted ash size parameters (μm).

Unit no. Shift no. D10 coal D10 size of ash D50 coal D50 size of ash D90 coal D90 size of ash D32 coal D32 size of ash D43 coal D43 size of ash

Unit 1 1st shift 6.38 0.11 42.06 0.73 114.44 1.99 6.02 0.10 52.72 0.92
2nd shift 8.21 0.14 50.49 0.88 125.60 2.19 8.65 0.15 60.28 1.05
3rd shift 9.16 0.16 56.79 0.99 141.30 2.46 9.26 0.16 67.56 1.18
4th shift 8.74 0.15 57.34 1.00 147.95 2.58 9.20 0.16 69.14 1.20
5th shift 8.60 0.15 56.19 0.98 141.33 2.46 9.27 0.16 67.21 1.17
6th shift 9.51 0.17 59.70 1.04 151.56 2.64 17.19 0.30 71.75 1.25
7th shift 10.99 0.19 63.88 1.11 155.97 2.72 10.15 0.18 75.24 1.31
8th shift 11.64 0.20 64.11 1.12 159.15 2.77 11.99 0.21 76.42 1.33
9th shift 15.99 0.28 81.17 1.41 191.60 3.34 15.72 0.27 94.32 1.64
Unit 2 1st shift 9.09 0.16 56.19 0.98 137.88 2.40 9.01 0.16 66.42 1.16
2nd shift 7.69 0.13 50.13 0.87 125.57 2.19 10.92 0.19 59.72 1.04
3rd shift 9.08 0.16 51.18 0.89 123.84 2.16 11.13 0.19 60.23 1.05
4th shift 11.48 0.20 61.00 1.06 144.42 2.51 9.49 0.17 70.95 1.24
5th shift 10.27 0.18 58.26 1.01 136.05 2.37 9.65 0.17 67.23 1.17
6th shift 8.50 0.15 47.21 0.82 118.18 2.06 8.67 0.15 56.65 0.99
7th shift 10.44 0.18 57.16 1.00 134.99 2.35 14.16 0.25 66.35 1.16
8th shift 10.81 0.19 61.22 1.07 140.12 2.44 10.96 0.19 69.71 1.21
9th shift 13.52 0.24 64.20 1.12 147.53 2.57 13.17 0.23 73.76 1.28
 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17 15

Table 9
The measured and the predicted UBCs.

Unit Shift no. Measured UBC Predicted UBC (%) with Predicted UBC (%) with Predicted UBC (%) with Predicted UBC (%) with Predicted UBC (%) with
no. (%) D10 D50 D90 D32 D43

Unit 1 1st shift 1.53 0.63 0.90 1.81 0.79 1.77

2nd shift 3.26 1.44 1.68 2.59 2.53 2.86
3rd shift 2.57 1.98 2.36 3.63 3.07 3.97
4th shift 3.21 1.70 2.40 4.13 2.98 4.21
5th shift 3.24 1.64 2.29 3.64 3.08 3.91
6th shift 5.20 2.53 3.13 5.11 22.39 5.43
7th shift 3.13 3.40 3.34 4.86 4.02 5.45
8th shift 4.11 4.22 3.53 5.40 6.92 5.97
9th shift 4.72 11.12 7.27 9.57 15.85 11.41
Unit 2 1st shift 1.44 1.93 2.28 3.36 2.81 3.76
2nd shift 2.5 1.27 1.75 2.75 5.43 2.96
3rd shift 3.08 2.09 1.87 2.65 5.77 3.05
4th shift 2.68 3.86 2.90 3.85 3.27 4.56
5th shift 1.79 2.83 2.58 3.29 3.52 3.97
6th shift 1.50 1.63 1.40 2.19 2.60 2.42
7th shift 4.84 3.27 2.69 3.54 12.25 4.20
8th shift 3.27 3.38 3.07 3.68 5.29 4.54
9th shift 4.58 6.84 3.66 4.44 9.48 5.55

the variance in the size parameters might be the malfunction of the
classifiers of the mills of Unit 2. In addition to mill and classifier
malfunctioning, feed coal properties such as moisture content and
even the minor differences in coal compositions might lead to the vari-
ance in the size parameters from shift to shift. The PSD of feed coal prior
to entering the mill might also be one of the most governing reasons of
this variance and this was established in the previous study about UBC
in BA. It was claimed that the moisture content of the feed coal had
been found between 8.02% and 18.37% for the two combustion units,
and the resulting UBC in BA was reported between 0.49% and 0.71%
for Unit 1 and 0.26% to 0.69% for Unit 2 [7].
4.4. UBC analysis
The PC samples are injected into boiler and they are combusted
within the boiler conditions. Thus, the events occurred in the boiler

Fig. 9. Predicted UBC vs Measured UBC for Unit 1 (prediction of UBC with each size parameters D10, D50, D90, D32, and D43).
16 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17
can only be understood with the correct analysis of the ash produced.
For this purpose, the collected representative FA samples were analyzed
in terms of UBC content in the context of this study. The results obtained
for Unit 1 and Unit 2 are given in Table 6.
Regarding the results of UBC content in FA, it was observed that it only
changed between 1.44% and 5.2% for both units. These percentages are re-
spectively very low comparing to the claims of Ahmaruzzaman [3] (2% to
12%) and Senneca [17] (10–23%). The UBC content is the lowest (1.53%)
for the 1st shift and the highest for the 6th shift (5.2%) for Unit 1 while
it is the lowest for the 1st shift (1.44%) and the highest for the 7th shift
(4.84) for the Unit 2. The UBC content can be associated to the results of
the size parameters of PSD in order to understand which size parameter
is more effective on UBC. Figs. 7 and 8 show the comparisons between
size parameter results and UBC content for Units 1 and 2 respectively.
Regarding comparisons between the UBC content and the size pa-
rameters (D10. D50. D90. D32. D43) for Unit 1, the relationship between
the UBC and the size parameters (D10. D50. D90. D43) has a correlation co-
efficient of around 0.5 whereas it is 0.907 between the UBC and the D32.
For Unit 2 a very similar result was obtained. R2 of the relationship be-
tween the UBC and the size parameters (D10. D50. D90. D43) has the cor-
relation coefficient of around 0.2 whereas it is 0.92 for the relationship
between the UBC and the D32. This is an evidence of the PSD effect on
UBC and employment of D32 would enable the rough prediction of
UBC content in FA.
4.5. Modeling results and discussion
Employment of the size parameters in a model for the prediction of
UBC is the last step of this study. By the introduction of new parameters
Fig. 10. Predicted UBC vs Measured UBC for Unit 2 (prediction of UBC with each size parameters D10, D50, D90, D32, and D43).
like ADF(FA) and OPC and the size parameters other than D90 (D10. D50.
D32. D43), a new model was proposed (see Eq. (17)). Dxx size of ash was
evaluated as in the previous study [7] about BA. The ADF(FA) values for
each shift was calculated with performing the formulation (Eq. (7)) pro-
posed. The ADF(FA) values for both units are given in Table 7. The size
parameters of PC samples and the predicted size parameters of ash
(Dxx size of ash) are given in Table 8. In Table 9, the measured and the
predicted UBC contents are presented. The OPC constant were taken
10, 0.05, 0.005, 15, and 0.05 for D10, D50, D90, D32, and D43, respectively.
In Figs. 9 and 10 graphical representations of the measured and predict-
ed UBC values are provided for both units.
For Unit 1, the experimentally measured amounts of UBC and the
predicted amounts of UBC in FA by the model proposed yielded correla-
tion coefficients (R2) of 0.35, 0.48, 0.55, 0.77, and 0.51 for D10, D50, D90,
D32, and D43, respectively whereas for Unit 2, correlation coefficients
were 0.45, 0.38, 0.34, 0.88, and 0.38 for the same size parameters re-
spectively. Among these values, 0.77 found for Unit 1 and 0.88 found
for Unit 2 are the most reasonable correlation coefficients which corre-
spond to UBC vs D32.
5. Conclusions
High amount of UBC in FA or BA is strictly related to the size distribu-
tion of PC since no other factor could be associated to the shift by shift
variance of UBC under stable operations. The change in PSD of PC, i.e.
the change in size parameters, was correlated with UBC content with
the assumption of negligible effect of possible factors affecting UBC.
This assumption is valid because not only operational parameters
 M. Bilen, S. Kizgut / Fuel Processing Technology 143 (2016) 7–17
under stable period had limited change but also cordial relation was
achieved between the UBC and the size parameters.
Combustion, fragmentation and population models were introduced
and among them population model was suggested due to ease of appli-
cability. All of the models were explained in detail, and prediction of any
size parameter of the ash formed after combustion was the sole purpose
of each model. A modified population model was proposed to estimate
the UBC by having PC size distribution and amount of UBC in FA. A step-
wise prediction of ash size distribution for each size parameters and
UBC was realized in the context of the study. In particular, the model
proposed which employs the D32 size parameter was in good agreement
with the experimental results. Corresponding R2 values for the predict-
ed UBC percentages with respect to the measured ones were found as
0.78 for Unit 1 while R2 value is 0.88 for Unit 2. Some of the constants
(kB, ADFFA, OPC) used to model boiler environments were adapted
from a previous study performed by Shah et al. [19] and Bilen et al.
[7]. The conversion constant (6.6 × 10−7), which can be recalled as
UBC-Size conversion constant, is still applicable on the prediction of
UBC in FA. This constant (UBC-Size conversion) could further be applied
in future studies on UBC. Notably, understanding the effect of the size
parameters, especially the effect of D32, can give useful information for
the plant operators and researchers.
The D32 was found to be the most affecting factor on UBC that is why
D32 can be regarded as a replace parameter to the effect of coal proper-
ties. Ultimately, the validity of the hypothesis that unburned carbon
percentage in FA of a PC combusted is strongly related to D32 of that
coal was discussed in detail.
References
[1] MENR (Ministry of Energy and Natural Resources of Turkey), Sector Report, 2014
http://www.enerji.gov.tr (accessed October 2015).
[2] I. Acar, M.U. Atalay, Characterization of sintered class F fly ashes, Fuel 106 (2013)
195–203.
[3] M. Ahmaruzzaman, A review on the utilization of fly ash, Prog. Energy Combust. Sci.
36 (2010) 327–363.
[4] S. Atas, U. Tekir, M.A. Paksoy, A. Çelik, M. Çam, T. Sevgel, Numerical and experimen-
tal analysis of pulverized coal mill classifier performance in the Soma B Power Plant,
Fuel Processing Tehnology 126 (2014) 441–452.
[5] L.L. Baxter, Char fragmentation and fly ash formation during pulverized coal com-
bustion, Combustion and Flame 90 (1992) 174–184.
[6] E. Bar-Ziv, I.I. Kantorovich, Mutual effects of porosity and reactivity in char oxida-
tion, Prog. Energy Combust. Sci. 27 (2001) 667–697.
17
[7] M. Bilen, S. Kizgut, B. Akkaya, Prediction of unburned carbon in BA in terms of mois-
ture content and sieve analysis of coal, Fuel Process. Technol. 138 (2015) 236–242.
[8] M.M. Bwalya, M.H.A. Moys, model for pulverised fuel production in an air-swept
tube mill, Miner. Eng. 43–44 (2013) 154–158.
[9] H. Gao, A.J. Majeski, A. Runstedtler, A method to target and correct sources of un-
burned carbon in coal-fired utility boilers, Fuel 108 (2013) 484–489.
[10] R. Ganguli, S. Bandopadhyay, Relationship between particle size distribution of low-
rank pulverized coal and power plant performance, Journal of Combustion
(2012)http://dx.doi.org/10.1155/2012/786920.
[11] F. Holzleithner, R. Eisl, M. Haider, G. Aichinger, CFD simulation of process-driven
particle fragmentation in a coal bed gasifier, 27th Annual International Pittsburgh
Coal Conference 2010 PCC, vol. 2 2010, pp. 1139–1151 (ISBN: 9781617823213,
161782321X).
[12] S. Jiménez, J. Ballester, Study of the evolution of particle size distributions and its ef-
fects on the oxidation of pulverized coal, Combust. Flame 151 (2007) 482–494.
[13] S. Li, Char Slag Transition During Pulverized Coal GasificationPh.D. Dissertation The
University of Utah, 2010.
[14] J. Li, R. Jankowski, M. Kotecki, W. Yang, D.C.F.D. Szewczyk, Aproach for unburned
carbon reduction in pulverized coal boilers, Energy Fuel 26 (2012) 926–937.
[15] H.H. Liakos, K.N. Theologos, A.G. Boudouvis, N.C. Markatos, Pulverized coal char
combustion: the effect of particle size on burner performance, Appl. Therm. Eng.
18 (1998) 981–989.
[16] R. Mitchell, An intrinsic kinetics based, particle population balance model for char
oxidation during pulverized coal combustion, Proc. Of the Combustion Institute 28
(2000) 2261–2270.
[17] O. Senneca, Burning and physico-chemical characteristics of carbon in ash from a
coal fired power plant, Fuel 87 (2008) 1207–1216.
[18] P. Straka, J. Náhunková, M. Zaloudková, Analysis of unburned carbon in industrial
ashes from biomasscombustion by thermogravimetric method using Boudouard
reaction, Thermochim. Acta 575 (2014) 188–194.
[19] K.V. Shah, M.K. Cieplik, I. Betrand, W.L. Kamp, H.B. Vuthaluru, A kinetic-emprical
model for particle size distribution evlotion during pulverized fuel combustion,
Fuel 89 (2010) 2438–2447.
[20] A. Restrepo, R. Miyake, F. Kleveston, E. Bazzo, Exergetic and environmental analysis
of a pulverized coal power plant, Energy 45 (2012) 195–202.
[21] S. Ural, Comparison of fly ash properties from Afsin–Elbistan coal basin, Turkey
Journal of Hazardous Materials 119 (2005) 85–92.
[22] H. Wu, T. Wall, G. Liu, G. Bryant, Ash liberation from included minerals during com-
bustion of pulverized coal: the relationship with char structure and burnout, Energy
Fuel 13 (1999) 1197–1202.
[23] X. Xue, Y. Wang, Particle size distribution as a nonindependent variable affecting
pulverized-coal burnout in coal-fired power-plant boilers energy, Fuels 27 (2013)
4930–4934.
[24] Y.B. Yang, V.N. Sharifi, J. Swithenbank, L. Ma, L.I. Darvell, Combustion of a single par-
ticle of biomass, Energy Fuel 22 (2008) 306–316.
[25] X. Zhang, A. Dukhan, I.I. Kantorovich, E. Bar-Ziv, Structural changes of char particles
during chemically controlled oxidation, Proceedings of the 26th International
Symposium on Combustion, Napoli, Italy, July 28–August 2, 1996;The Combustion
Institute, Pittsburgh, U.S. 1996, pp. 3111–3118.