You are on page 1of 13

Applied Mathematical Modelling 30 (2006) 854–866

www.elsevier.com/locate/apm

CFD application for coal/air balancing in power plants

Sowjanya Vijiapurapu a, Jie Cui a,*
, Sastry Munukutla b

a
Mechanical Engineering Department, Tennessee Technological University, P.O. Box 5014, Cookeville, TN 38505, USA
b
Mechanical Engineering Department, Center for Energy Systems Research, Tennessee Technological University,
P.O. Box 5014, Cookeville, TN 38505, USA

Received 1 May 2004; received in revised form 1 May 2005; accepted 27 June 2005
Available online 18 August 2005

Abstract

Unbalanced coal/air ﬂow in the pipe systems of coal-ﬁred power plants will lead to non-uniform com-
bustion in the furnace, and hence a overall lower eﬃciency of the boiler. A common solution to this prob-
lem is to put oriﬁces in the pipe systems to balance the ﬂow. It is well known that if the oriﬁces are sized to
balance clean air ﬂow to individual burners connected to a pulverizer, the coal/air ﬂow would still be unbal-
anced and vice versa. However, the current power industry practice throughout the world is to size oriﬁces
for balancing the clean air ﬂow and accept the resulting imbalance in coal/air ﬂow. Field tests are mostly
conducted to verify a balanced clean air ﬂow.
It is now proposed to size the oriﬁces for balancing the coal/air ﬂow and then calculate the unbalanced
clean air ﬂow distribution to be known as the ‘‘tailored clean air ﬂow’’. Commercially available Computa-
tional Fluid Dynamics (CFD) code CFX was used to simulate the complex ﬂows in the piping systems in a
power plant. The two-phase modelling technique was employed to estimate the pressure drop coeﬃcients
with both clean air and coal/air ﬂows in order to size the oriﬁces. The results indicate that the pressure drop
is strongly dependent on the piping system geometry. With this proposed method, ﬁeld tests can be con-
ducted to correspond with the tailored clean air ﬂow, and the coal/air ﬂow balancing would be achieved.

Keywords: CFD; Numerical modelling; Coal/Air balancing; Power plant

*
Corresponding author. Tel.: +1 931 372 3357; fax: +1 931 372 6340.

doi:10.1016/j.apm.2005.06.005
S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866 855

1. Introduction

There are a number of operational problems of coal-ﬁred boilers, which are suspected to be
caused by non-uniform combustion in the furnace. One source of non-uniform combustion is un-
even distribution of fuel inputs to the furnace. Several coal pipes connect the exit of the pulverizer
to the individual burners, all of which are located at the same elevation in the furnace for that
particular pulverizer [1]. Though the elevation change is identical for each system, the pipes have
diﬀerent horizontal, vertical, and inclined lengths, and bends. Thus, the resistance to ﬂow is dif-
ferent for each system, leading to uneven distribution.
The resistance of a system is diﬀerent for single-phase ﬂow (clean air) or two-phase ﬂow (coal/
air) [2]. The resistance is expressed as a dimensionless pressure drop coeﬃcient. Thus, the single-
phase and the corresponding two-phase pressure drop coeﬃcients for any given system are diﬀer-
ent. One signiﬁcant point is that the pressure drop coeﬃcient for a sharp edged oriﬁce is the same
for both single-phase and two-phase ﬂows for dilute suspension ﬂows encountered in power
plants. A typical volume fraction of coal to air is 1/1600 [3].
Since the resistance coeﬃcients are diﬀerent for clean air and coal/air ﬂows, the oriﬁce sizes
needed to balance the clean air ﬂow would be diﬀerent from the oriﬁce sizes needed to balance
the coal/air ﬂow. The electric power industry practice throughout the world so far has been to size
the oriﬁces based on clean air ﬂow pressure drop coeﬃcients. The clean air ﬂow would, therefore,
be balanced but the corresponding coal/air ﬂow would still be unbalanced. In one instance, it was
reported that after the clean air ﬂow was balanced to within 3%, the coal/air ﬂow was still unbal-
anced by 27% [4]. On the other hand, if the oriﬁce sizing is based on the coal/air ﬂow balancing,
the corresponding clean air ﬂow would be unbalanced. This would not present a problem since
coal/air balancing is what is needed for eﬃcient combustion.
For calculating the pressure drop coeﬃcients, standard handbook data can be used [5]. How-
ever, the data for two-phase ﬂow (coal/air) is found to be unreliable. So, the pressure drop coef-
ﬁcients for the coal/air as well as the clean air ﬂow were calculated using CFD in the present work
[6]. The multi-phase models were used to simulate the complex ﬂow ﬁeld of each phase and the
interaction between air and coal particles. It is believed that the numerical solution was able to
provide reasonable pressure drop coeﬃcients in the current study.
In most power plants the geometry for individual pipe systems is speciﬁed by the total horizon-
tal run, total vertical run, number of bends, and bend geometry. Since exact geometry of the sys-
tem is not speciﬁed, it is not possible to calculate the pressure drop of the entire system in one
pass. It was therefore decided to calculate the pressure drop in a system by breaking it into indi-
vidual components. The calculated pressure drops across various components were then added
together, to estimate the pressure drop of the entire system. In the current paper, this procedure
will be implemented in one of the coal-ﬁred electric utilities. A description of this method and
results from ﬁeld experience will form the rest of the contents of this paper.

2. Use of CFD in power industry applications

In the power generation industry, many processes involve multi-phase ﬂow, phase transforma-
tion, combustion, and complex chemical reactions [7]. This is particularly true for coal ﬁred power
856 S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866

plants. However, due to the complex nature of these phenomena, traditional research approaches
were not adequate in providing a thorough understanding of these processes. Recently, CFD has
been used in power industry to gain a qualitative as well as quantitative understanding of these
processes.
Pulverized coal combustion in a 2.5 MW burner was modelled using two commercial CFD
codes [8]. Despite discrepancies between the solutions of the two codes, the predicted velocity,
temperature, and species concentration were in overall agreement with the experimental data, sug-
gesting that the two codes were capable of predicting good ÔtrendÕ solutions. The Eulerian–Eule-
rian multi-phase model in conjunction with the k–e model was applied to predict erosion in slurry
pipeline tee-junctions [9]. The CFD model was used to assess several potential solutions to the ero-
sion problem, and the results demonstrated the eﬀectiveness with which CFD techniques can be
used in industrial applications. Similar multi-phase model was employed to develop a computa-
tional model of erosion in a ﬂuidized bed [10]. It was shown that the model predictions were in
good agreement with the experiment results. Stopford reviewed some of the recent applications
of CFD to the power generation and combustion industries [11]. Examples included coal-ﬁred
low-NOx burner design, furnace optimization, over-ﬁre air, gas reburn, and laminar ﬂames.
The results showed that CFD modelling was well established as a design tool and has been widely
used in the power generation industry to help engineers reduce emissions, increase thermal eﬃ-
ciency, select fuel, and extend plant lifetime.
In this paper, an attempt will be made to take this new approach—CFD—to a problem faced
by the power industry for a number of years: coal/air balancing. The multi-phase model in the
unstructured version of CFX (version 5.5.1) will be applied to determine pressure drop in the pip-
ing systems in a power plant. Based on the predicted pressure drop, oriﬁces in the piping systems
will be added or modiﬁed and balanced coal/air ﬂow will be achieved.

3. Principle of the method

Fig. 1 shows two systems with the same elevation change, but diﬀerent horizontal, vertical runs,
and bends. Both systems start at the same elevation at the pulverizer exit and discharge at the same
elevation in the boiler. System 1 has one vertical run, one horizontal run, and one bend. System 2
has 3 vertical runs, 3 horizontal runs, and 5 bends. The total pressure drop consists of frictional
loss over the horizontal and vertical pipe lengths, a loss due to bends, and a loss due to gravity.
In general if the total ﬂow rate is Q, with ﬂow rate in system 1 being Q1 and that in system 2
being Q2, and the pressure drop coeﬃcients being K1 and K2, respectively, we have
Q ¼ Q1 þ Q2 ;
K 1 Q21 ¼ K 2 Q22 ;
ð1Þ
Q Q
) Q2 ¼ qﬃﬃﬃﬃ and Q1 ¼ qﬃﬃﬃﬃ .
1 þ KK 21 1 þ KK 12

Let the pressure drop coeﬃcients for clean air and coal/air for systems 1 and 2 be K1A, K2A, and
K1C, K2C respectively. Then the ﬂow distribution with clean air is given as
S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866 857

Fig. 1. Illustration of two systems with the same elevations change, but diﬀerent horizontal and vertical runs, and
bends.

Q Q
Q2A ¼ qﬃﬃﬃﬃﬃﬃ and Q1A ¼ qﬃﬃﬃﬃﬃﬃ ð2Þ
K 2A K 1A
1þ K 1A
1þ K 2A

and that with coal/air is given as

Q Q
Q2C ¼ qﬃﬃﬃﬃﬃﬃ and Q1C ¼ qﬃﬃﬃﬃﬃﬃ . ð3Þ
K 2C K 1C
1þ K 1C
1þ K 2C

In most cases, it has been observed that the ratio of K2C/K1C is very nearly equal to K2A/K1A. In
other words, clean air and coal/air are unbalanced in a similar manner. The pressure drop coef-
ﬁcient KOR of the oriﬁce plate is calculated as the absolute diﬀerence in the pressure drop coeﬃ-
cients of the pipes. In order to achieve clean air balance, an oriﬁce yielding a pressure drop
coeﬃcient of KOrf,A = K2A  K1A should be introduced. In order to achieve coal/air balance,
an oriﬁce yielding a pressure drop coeﬃcient of KOrf,C = K2C  K1C should be introduced.
As a general rule the pressure drop coeﬃcients for the coal/air (two-phase) ﬂow K1C and K2C,
are always greater than the corresponding clean air (single-phase) pressure drop coeﬃcients K1A
and K2A (typically by a factor of 2–3) [12]. Hence, the oriﬁce diameters based on the clean air dis-
tribution are larger than those based on the coal/air distributions. As a result of the above, it has
been a normal trend in the power industry to under estimate the pressure drop requirement.
The diameter of the oriﬁce can be calculated from available empirical equations [13], one of the
commonly used equations is
"   0.375 !#2
F1 F0 F0
K OR ¼  1 þ 0.707  1 ; ð4Þ
F0 F1 F1
 2
where FF 01 ¼ AA01 ¼ DD01 , A0, A1 correspond to the area of the oriﬁce and pipe, respectively, and, D0,
D1 correspond to the diameters of the oriﬁce and pipe, respectively.
858 S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866

4. The systems modelled

The four systems modelled using the numerical method are given in Table 1. As shown in the
Table, there are 4 systems that carry coal/air mixture from the pulverizer to the furnace. The indi-
vidual lengths of the horizontal and vertical sections between the bends are not known. The only
data available is the total horizontal and vertical length of the systems, and the number of diﬀer-
ent bends in each system. Also, the bend angles for each system are known.
If K1, K2, K3, and K4 are the pressure drop coeﬃcients for systems 1, 2, 3, and 4, respectively,
ﬂow distributions V1, V2, V3, and V4 can be calculated by extending Eq. (3) to a 4-branch system.
rﬃﬃﬃﬃﬃﬃ
400 K1
V 1 ð%Þ ¼ qﬃﬃﬃﬃ qﬃﬃﬃﬃ qﬃﬃﬃﬃ ; V 2 ð%Þ ¼ V 1 ;
K1
1 þ K2 þ K3 þ K4K1 K1 K2
rﬃﬃﬃﬃﬃﬃ rﬃﬃﬃﬃﬃﬃ
K1 K1
V 3 ð%Þ ¼ V 1 ; V 4 ð%Þ ¼ V 1 . ð5Þ
K3 K4
Note that for a perfectly balanced system K1 = K2 = K3 = K4, and in that case V1 =
V2 = V3 = V4 = 100%. For an unbalanced system V1, V2, V3, and V4 could be diﬀerent from
100%, but however, would still add to 400%. It can be seen from Table 2 that in the existing sys-
tem the clean air and coal/air ﬂows are both unbalanced. It can also be observed that the two
unbalances are not similar due to the presence of oriﬁces. Hence, it is proposed to retroﬁt the
whole oriﬁcing system based on the coal/air balancing. Since the two-phase ﬂow in the existing
systems is quite complex and the data from standard handbook is unreliable, it was decided to

Table 1
Detail of four systems that exit from one pulverizer and feed the boiler at the same elevation
System Total vertical Total horizontal Bends in system Diameters of existing
length (m) length (m) oriﬁces in system (mm)
1 10.39 48.11 150, 160, 160, 135 442 at 1 m distance
442 at 56 m distance
2 10.39 36.83 120, 135 437 at 1 m distance
437 at 45 m distance
3 10.39 64.34 165, 90, 135 442 at 1 m distance
4 10.39 75.62 135, 90, 160, 160, 135 No oriﬁces
Note: Oriﬁces are inserted according to the maximum pressure drop in the systems. Calculated pressure drop in system 4
is the maximum. Therefore, oriﬁces were inserted in systems 1–3 to balance the pressure drop.

Table 2
Existing ﬂow distribution for clean air and coal/air ﬂows
System Clean air ﬂow distribution (%) Coal/air ﬂow distribution (%)
1 95 102
2 97 108
3 100 96
4 108 94
S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866 859

utilize the two-phase modelling technique to obtain the pressure data in each system. Then the
oriﬁce diameters are recalculated based on the CFD results and a well-balanced system is
obtained.

5. CFD application to calculate pressure drop

As discussed in Section 2, the exact geometries of the individual systems in a power plant are
not always available. This problem was overcome by breaking up the geometry of the system into
various components like the horizontal section, vertical section, and various bends. The pressure
drop across each component is calculated and then put together to give the pressure drop along
the whole geometry of the system. Examples of the geometries used for the modelling are shown in
Fig. 2.
The pressure drops for the horizontal and vertical lengths were calculated initially for a 60D
pipe, where D is the diameter of the pipe. The pressure drops across unit length were then calcu-
lated and applied to the existing lengths of the pipe. A length of 60D was chosen to ensure that the
ﬂow became fully developed. The pressure drop per unit length was calculated for the fully devel-
oped region. It was found that for the current conﬁguration and ﬂow condition (Reynolds number
9.15 · 105, surface roughness e = 0.0026 mm) the ﬂow became fully developed within 30D from
the inlet. Therefore, for the bends, the upstream length was assumed to be 40D and downstream
to be 20D and the pressure drop across the bend was calculated.
For calculation of the steady state ﬂow in the piping systems, continuity and momentum equa-
tions were solved along with the standard k–e turbulence model. Two-phase ﬂow calculations
were adopted to simulate the air ﬂow and coal particles. The equation of continuity for a mixed
ﬂuid is expressed by Eq. (6), where a is the phase, ra is the volume fraction of that phase, qa is the
density of the ﬂuid in the phase a, xj is the coordinate with the index j ranging from 1 to 3, and U ja
is the mean velocity in the phase a along the direction j. The continuity equation expressed for
each control volume is shown in Eq. (7).

Fig. 2. Schematics of the components modelled.

860 S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866

oðra qa Þ oðra qa U ja Þ
þ ¼ 0; ð6Þ
ot oxj
X X 1 oðra qa Þ oðra qa U j Þ
a
ra ¼ 1 or þ j
¼ 0; ð7Þ
a a
qa ot ox

where repeated indices imply summation from 1 to 3. The equation of motion is expressed by Eq.
(8)
oðra qa U ia Þ oðra qa U ia U ja Þ op oðra sji Þ
þ j
¼ ra i þ ra qa gi þ ; ð8Þ
ot ox ox oxj
where p is the pressure and sij is the stress tensor given by
 i 
oU oU j
sij ¼ leff þ i ; ð9Þ
oxj ox
where leﬀ is eﬀective viscosity, which is deﬁned as the sum of dynamic viscosity l and eddy vis-
cosity lt, leﬀ = l + lt. The eddy viscosity is provided by the k–e turbulence model.
k2
lt ¼ C l q ; ð10Þ
e
where Cl is a constant and is equal to 0.09, k is the turbulent kinetic energy, e is the dissipation
rate, and both are provided by the k–e turbulence model.
   

o o j l þ a ok a
ðra qa k a Þ þ j ra ðqa U a k a Þ  l þ ¼ ra ðP a  qa ea Þ þ T kab ; ð11Þ
ot ox rk oxj
   

o o l þ a oea ea
ðra qa ea Þ þ j ra ðqa U ja ea Þ  l þ ¼ ra ðC e1 P a  C e2 qa ea Þ þ T eab ; ð12Þ
ot ox re oxj ka
where rk = 1.0, re = 1.3, Ce1 = 1.44, and Ce2 = 1.92 are constants from the k–e model. T kab and T eab
are the terms that represent the inter-phase transfer for k and e respectively. P represents the shear
production due to turbulence for that particular phase, which is
 
oU j oU j oU k
P ¼ lt k þ j . ð13Þ
ox oxk ox
The continuity equation, momentum equation, and the turbulence model equation were solved
for each phase. The pipe diameter was 0.527 m. For clean air ﬂow, the air density was taken as
1.284 kg/m3, which is the value at room temperature. For coal/air ﬂow, the air density was
0.977 kg/m3, which is the value at 88 C, and the coal density was 1398 kg/m3. The coal ﬂow rate
was 2.36 kg/s and the air ﬂow rate was 5.00 kg/s.
The boundary conditions were assumed to be uniform velocity distribution at the inlet and zero
gage pressure (open to atmosphere) at the outlet. Grid reﬁnement near the walls was used for grid
generation (Fig. 3). This allows for much better resolution of the velocity ﬁeld near the wall where
it changes rapidly. Reynolds Averaged Navier–Stokes equations (RANS) were solved using ﬁnite
volume method on an unstructured mesh with the standard k–e model for turbulence.
S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866 861

Fig. 3. Cross-section and plan view of the unstructured grid.

A typical mesh size for the CFD simulation was around 300,000 mesh cells, each steady state
case took 11–16 wall clock hours on Dell Optiplex GX270 computers. These computers had 2
Intel P4 3.20 GHz processors with 2.00GB RAM. The steady state solutions used pseudo time
step, which was determined by the CFL requirement of the solver. Convergence was claimed when
a stable pressure drop was achieved, usually it took 7000 iterations.
The friction factor was calculated by using the pressure drop in the fully developed region of the
pipe. The friction factor for clean air was calculated by
DP
f ¼ ; ð14Þ
L 1
D 2
qU 2avg
where f is the friction factor, DP is the pressure drop along the pipe over a distance of L in the
fully developed region, q is the density of air, Uavg is the average velocity, and D is the diameter
of the pipe. For a fully developed ﬂow, the friction factor is a constant. From this calculation, the
friction factor was found to be f = 0.0164. The results were compared to those in the Moody chart
(Reynolds number 9.15 · 105, surface roughness e = 0.0026 mm, friction factor f = 0.017).
The boundary conditions for the coal/air ﬂow were imposed in the same manner as those for
clean air ﬂow. The boundary condition for the particles was no slip i.e., the particles stick to the
surface if they hit the surface. The volume fraction for coal at inlet was considered to be uniformly
distributed. The coal particles were all considered to be spherical and of the same uniform size
(diameter 1 lm). Table 3 gives the pressure drop per unit length for the horizontal and vertical
components and also the pressure drop across the bends for both clean air and coal/air ﬂows.
As can be seen, the pressure drop for the coal/air ﬂow is always greater than that of the clean
862 S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866

Table 3
Pressure drop across various components for clean air and coal/air ﬂows
Horizontal (Pa/m) Vertical (Pa/m) Bend angles
90 (Pa) 120 (Pa) 135 (Pa) 150 (Pa) 160 (Pa) 165 (Pa)
Clean air 11.02 10.82 48.68 32.06 23.93 13.79 9.18 5.71
Coal/air 27.52 16.6 104.33 51.76 29.63 15.64 11.47 8.38

air ﬂow. It can be said that this is mainly due to the resistance oﬀered by the interaction between
air and coal particles and among coal particles themselves, as well as blockage eﬀects.

6. Results

In this study, oriﬁce sizes were calculated for coal/air balancing. Based on the given data, the
calculations have been performed with the existing system conﬁguration. The clean air and coal/
air pressure drop coeﬃcients, and ﬂow distributions were ﬁrst calculated. Oriﬁce sizes for coal/air
balancing were then calculated. Note that in each system if one or more oriﬁces already exist, then
there was at least one oriﬁce at 1-m distance from the pulverizer exit. It was then tacitly assumed
that the new oriﬁce would replace that oriﬁce.
Table 1 gives the data from the power plant. This table is self-explanatory. The existing ﬂow
distributions are given in Table 2, which are based on the pressure drop calculations shown in
Table 3. The oriﬁce diameters for balancing coal/air ﬂow are shown in Table 4. The tailored clean
air ﬂow and the balanced coal/air ﬂow distribution after insertion of new oriﬁces is shown in
Table 5. As can be seen from Table 5, the pressure drop for each system is the same and

Table 4
Oriﬁce diameter calculation for coal/air balancing
System K Loss coeﬃcient Existing oriﬁce KOR Total K New KOR Oriﬁce diameter (mm)
1 5.84 0.95 6.79 2.32 401.78
2 4.73 1.07 5.80 3.31 389.08
3 7.78 0.00 7.78 1.32 427.85
4 9.10 0.00 9.10 0.00 –
Note: K is the pressure loss coeﬃcient of each system and KOR is the pressure loss coeﬃcient of the oriﬁce calculated
based on Eq. (4).

Table 5
Tailored clean air ﬂow and balanced coal/air ﬂow distribution after insertion of new oriﬁces
System Clean air Coal/air
Total K after retroﬁt Flow distribution (%) Total K after retroﬁt Flow distribution (%)
1 5.25 91 9.10 100
2 6.01 85 9.10 100
3 3.88 105 9.10 100
4 3.01 120 9.10 100
S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866 863

consequently the coal/air ﬂow is balanced. In each instance, an existing oriﬁce at 1 m distance is to
be replaced by the new oriﬁce.
The CFD simulations also provide detailed information of the two-phase ﬂow ﬁeld. Fig. 4
shows the close-up view of the air velocity magnitude contours for bends with various angles.
The gravity force is vertically downward (in the negative y-direction). All contours shown in Figs.
4 and 5 are on the center plane (xz-plane) that passes the axis of the bend. In the horizontal por-
tion of the bend, all ﬂows exhibit similar velocity distributions and there is no variation in the

Fig. 4. Air velocity magnitude contours for bends with various angles.
864 S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866

Fig. 5. Contours of volume fraction of coal for bends with various angles.

streamwise direction. This is because the ﬂow has gone through 40D upstream length before it
reaches the bend, and ﬂow has become fully developed. It is observed that the velocity is lower
near the bottom of the horizontal portion. This is due to the higher resistance to the air induced
by the coal particles when they deposit to the pipe bottom. For the ﬂow conditions considered in
the current study, the eﬀects of coal particle deposition is signiﬁcant. It can be clearly seen from
Fig. 4 that the characteristics of the air ﬂow downstream of the bend strongly depend on the bend
angle. For sharp turns (Fig. 4a–c), the ﬂow has to change direction quickly and the higher velocity
ﬂow is shifted towards the outer radius of the downstream bend. While near the inner radius of
S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866 865

the bend, a lower velocity region can be identiﬁed. As shown in Table 3, this sharp ﬂow direction
change results in signiﬁcant pressure loss, with the 90 bend giving the highest. For higher bend
angles (Fig. 4d), the ﬂow goes through the bend quite smoothly, and the pressure loss is insignif-
icant compared to the sharp turning cases.
Fig. 5 displays the contours of volume of ﬂuid (VOF) of coal for bends with various angles. In
the horizontal portion, all bends show similar coal VOF distribution, just like the velocity con-
tours in Fig. 4. Higher coal VOF is found at the lower part of the upstream pipe. This is again
attributed to the deposition of the coal particles. In the bend, maximum VOF of coal can be iden-
tiﬁed near the outer radius of the bend. This is explained by the ﬂow impingement in the bend, and
the attachment and accumulation of coal particles at this location. For sharp turning bends, the
overall VOF of coal becomes much less downstream of the bend, indicating signiﬁcant deposition
and accumulation of coal particles in the bend and upstream pipe (Fig. 5a–c). While for smooth
turning bend, the eﬀects of coal particle deposition and accumulation are not as prominent
(Fig. 5d). It should be pointed out that for the 90 bend, immediately downstream of the bend
and close to the inner radius, there is a localized region with a higher coal VOF than its local
ambient. This is the indication of ﬂow separation. The local recirculating ﬂow traps a signiﬁcant
amount of coal particles and a region of higher VOF of coal is formed. For 120 and 130 bends,
this localized region can still be seen, but the size is much smaller. While for the 160 bend, it is not
recognizable, suggesting negligible or non-existent ﬂow separation downstream of the bend.

7. Conclusions

There is currently no easy way of measuring coal/air ﬂow in a power plant. In order to balance
the coal/air ﬂow to the individual burners, it would, therefore, be necessary to rely on clean air
tests. The industry practice has so far been to balance the clean air ﬂow and accept the resulting
imbalance in the coal/air ﬂow. It was observed in one speciﬁc case that, when the clean air is bal-
anced with a maximum deviation of 3% from the average, the corresponding coal/air ﬂow devi-
ation was approximately 27%. This is unacceptable from the consideration of eﬃcient unit
operation.
A new method in which creating a tailored imbalance in clean air ﬂow corresponding to a bal-
anced coal/air ﬂow distribution has been successfully proposed in this paper. Commercially avail-
able software CFX was used to calculate pressure drops in systems. The two-phase ﬂow
phenomena was simulated using the multi-phase modelling technique. Using this, several geom-
etries involving any number of independent lines starting from a mill and discharging to a given
burner can be handled. The results show that the pressure drop in the systems strongly depends on
the system geometry. Oriﬁces are sized based on calculated coal/air pressure drops, and ﬁnally
imposing a tailored imbalance in clean air ﬂow distribution leading to a balanced coal/air distri-
bution. It should be pointed out that in order to implement this method in power plant, veriﬁca-
tion can still be done by measuring the tailored clean air distribution. The numerical results also
provide detailed information of the two-phase ﬂow ﬁeld in the piping systems in a power plant,
based on which physical insights were obtained and better understanding of the complex ﬂow phe-
nomena was achieved. This study demonstrated that CFD can be used as an eﬀective tool for
design and research for power industry applications.
866 S. Vijiapurapu et al. / Applied Mathematical Modelling 30 (2006) 854–866

References

[1] A.A. Vetter, R.S. Vetter, Balancing pulverized coal ﬂows in parallel piping, J. Eng. Gas Turb. Power 107 (July)
(1985) 679–684.
[2] R. Avancha, Coal/air balancing in pulverized coal ﬁred units, MasterÕs Thesis, Tennessee Technological University,
1995.
[3] M.P. Sharma, Numerical and experimental study of gas-particle ﬂows in oriﬁces and venturis: application to
ﬂowmeter design, Ph.D. Dissertation, Washington State University, 1977.
[4] G.R. Jones, Pulverized Coal Mill Fuel/Air Ratio Testing, paper presented at EPRI Heatrate Improvement
Conference, Knoxville, Tennessee, September 1989.
[5] ASHRAE Handbook Fundamentals-2001, I.P. Edition.
[6] AEA Technology, CFX-5.5.1 Flow Solver User Guide, AEA Technology Inc., 2003.
[7] C. Yin, L. Rosendahl, S.K. Kar, T.J. Condra, Use of numerical modeling in design for co-ﬁring biomass in wall-
ﬁred burners, Chem. Eng. Sci. 59 (2004) 3281–3292.
[8] C.N. Eastwick, S.J. Pichering, A. Aroussi, Comparisons of two commercial computational ﬂuid dynamics codes in
modeling pulverized coal combustion for a 2.5 MW burner, Appl. Math. Modell. 23 (1999) 437–446.
[9] G.J. Brown, Erosion prediction in slurry pipeline tee-junctions, Appl. Math. Modell. 26 (2002) 155–170.
[10] D. Achim, A.K. Easton, M.P. Schwarz, P.J. Witt, A. Zakhari, Tube erosion modelling in a ﬂuidised bed, Appl.
Math. Modell. 26 (2002) 191–201.
[11] P.J. Stopford, Recent applications of CFD modelling in the power generation and combustion industries, Appl.
Math. Modell. 26 (2002) 351–374.
[12] A.A. Vetter, Theoretical Evaluation of the Diﬀerences Between Single and Two-Phase Flow as Applied to Coal
Transport Piping, Humburg Mountain Research Laboratories Report, HMRL R—31:1, September 1981.
[13] I.E. Idelchik, Handbook of Hydraulic Resistance, third ed., CRC Press, Inc., Boca Raton, 1994, p. 221.