You are on page 1of 14

Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.

1007/s12206-020-0141-4

Journal of Mechanical Science and Technology 34 (2) 2020


Original Article
DOI 10.1007/s12206-020-0141-4
Detailed assessment of mesh sensitivity
for CFD simulation of coal combustion
Keywords:
· CFD
· Coal
in a tangential-firing boiler
· Combustion
· Boiler Hyunbin Jo1, Kiseop Kang1, Jongkeun Park1, Changkook Ryu1, Hyunsoo Ahn2
· Mesh sensitivity and Younggun Go2
· Unburned carbon
1
School of Mechanical Engineering, Sungkyunkwan University, Suwon 16419, Korea, 2Doosan Heavy
Industries and Construction, Yongin 16858, Korea
Correspondence to:
Changkook Ryu
E-mail: cryu@skku.edu
Abstract Computational fluid dynamics (CFD) has become an essential tool for
optimizing the design and diagnosing the operation of a boiler. However, the validity of the
Citation: results depends on the degree of numerical diffusion as well as the reliability of the submodels.
Jo, H., Kang, K., Park, J., Ryu, C., Ahn, H.,
Go, Y. (2020). Detailed assessment of This study aims to assess mesh sensitivity in the reacting two-phase flow of pulverized coal in a
mesh sensitivity for CFD simulation of common tangential-firing boiler. Three mesh versions were constructed for the boiler with the
coal combustion in a tangential-firing number of cells ranging between 1.2 million and 5.4 million, corresponding to 0.0114 -
boiler. Journal of Mechanical Science and
Technology 34 (2) (2020) 917~930. 0.0022 m3 per cell in the burner zone. The velocity distribution was found to be highly sensitive
http://doi.org/10.1007/s12206-020-0141-4 compared to temperature, heat flux, and NO concentration. By contrast, the use of key
performance parameters such as total wall heat absorption, exit NOx concentration, and
Received July 16th, 2019 carbon conversion, was not appropriate criteria for the mesh sensitivity test. These parameters
Revised November 16th, 2019 were determined by integration over the entire surface or volume, which made them sensitive
Accepted December 20th, 2019 to the overall reaction stoichiometry instead of the mesh fineness. It suggests that the use of a
coarse mesh could be acceptable in evaluating the key performance parameters influenced by
† Recommended by Editor major operation variables, such as air distribution and fuel properties. However, sufficient mesh
Yong Tae Kang fineness is necessary for studies requiring accurate prediction of detailed flow patterns such as
the evaluation of burner tilting/yawing or ash deposition on the wall.

1. Introduction
Combined with efficient steam cycles and advanced combustion technologies with reduced
pollutant emission, coal-fired power generation has served as the most important source of
electricity worldwide for many years. However, it faces new challenges due to concerns involv-
ing greenhouse gas emissions as well as the increased use of renewable energy sources. This
means that coal-fired power plants must establish a new role for itself in the power industry by
improving their operational flexibility while further increasing energy efficiency and minimizing
environmental impact [1].
Most large-scale coal-fired boilers adopt one of the following two furnace configurations: Tan-
gential-firing (TF) and opposed wall-firing (WF). In a TF furnace, a series of coal burners and air
inlets are installed on the corners of a rectangular cross-section, which creates a large fireball
across the burner zone. In Korea, twenty plants adopting an almost identical TF furnace design
with a 500 MWe capacity are currently in operation, which are referred to as the ‘standard coal-
fired furnace’. In an opposed WF furnace, multiple layers of swirl burners are installed on the
front and rear walls, and these create individual jet flames colliding at the center of the furnace.
CFD is applied to the design optimization and operation diagnostics of coal-fired boilers
aimed to achieve high combustion efficiency, NOx reduction, alleviation of slagging/fouling, etc
[2]. It requires considering the detailed behavior and reactions of pulverized coal particles, typi-
© The Korean Society of Mechanical
Engineers and Springer-Verlag GmbH
cally having an average particle size of 50-55 μm. The calculation for the discrete particle
Germany, part of Springer Nature 2020 phase is performed using the Lagrangian scheme, tracking individual particles coupled with the

917
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

Table 1. Summary of CFD studies on a TF coal boiler in the literature.

Capacity No. of cells Mesh


Code Objectives and parameters Validation Ref.
(MW) (million) sensitivity
Wood chip burn-out and NOx in co-combustion by injection
150 2.04 CFX Exit values - [3]
level and burner yawing
NOx, heat absorption, and burn-out by burner yawing and the
200 5.00 Fluent Exit values - [4]
use of overfire air
300 0.95 Fluent Heat absorption and ash deposit by swirl-burner retrofit - - [5]
NOx and wall heat flux in straw co-combustion by cofiring ratio Measured temperature and
300 3.30 Fluent Tested [6]
up to 12 % exit values
500 0.96 Fluent NOx reduction and combustion with and without overfire air Exit values Tested [7]
Measured temperature and
500 3.00 Fluent Ash deposition by fuel blends and burner tilting (±30°) - [8]
velocity
Temperature and flow distribution in the crossover pass com-
500 7.00 Fluent Measured velocity - [9]
pared to measured data
Temperature and flow distribution, burn-out and NOx by differ-
500 4.13 Fluent Exit values Tested [10]
ent heat exchanger arrangement
NOx reduction and temperature distribution by the use of over- Measured NOx and exit
600 0.52 CFX Tested [11]
fire air values
Heat absorption, flow and temperature distribution by different
600 1.18 Fluent Measured NOx Tested [12]
overfire air ratios
Wall heat absorption, flow and reaction by the burner installed
600 0.74 Fluent Exit values and FEGT Tested [13]
on the corners and walls
Flow characteristics and heat absorption by SOFA yawing
600 1.18 - Measured velocity - [14]
(±20°)
Heat absorption and NOx by SOFA ratios up to 40 % and
600 1.64 Fluent Exit values Tested [15]
burner yawing
Measured temperature and
700 2.50 Fluent Burn-out, heat absorption and NOx by burner tilting up to 22° Tested [16]
exit values
800 3.40 CFX Heat absorption by coupled simulation with steam side Exit values Tested [17]
Exit temperature of heat
1000 3.20 Fluent Heat absorption by coupled simulation with steam side - [18]
exchanger and flue gas

gas phase in the Eulerian scheme. The interactions between mizer) and, in some cases [6, 15], through the in-furnace
the two phases involving momentum, heat and mass transfers, measurement of temperature or velocity. The number of cells
and heterogeneous reactions require repeated iterations to used to construct the mesh ranges between 0.52 and 5 million.
converge. The reliability of the solution depends on the mesh Most studies carried out mesh sensitivity analysis, but the wide
sensitivity, as well as the use of appropriate submodels and variations in the number of cells conflict with each other. Fur-
input conditions. thermore, the detailed methodologies used to assess the mesh
The mesh sensitivity is associated with the influence of nu- sensitivity differed between studies and were often not clear.
merical diffusion on the solution. Discretization of the governing For example, the temperature and concentration of O2 at the
equations with a choice of a numerical scheme for the mesh furnace exit were used as the criteria in Ref. [6], but the use of
resolution and shape induces the numerical diffusion. The nu- exit values may underestimate mesh sensitivity. The tempera-
merical diffusion needs to be minimized by refining the mesh ture profile along the furnace height was used in Refs. [11, 14]
combined with the alignment to the flow pattern and use of without any quantitative evaluation of the errors between the
higher-order discretization schemes. The mesh sensitivity test meshes.
to evaluate its influence on the key results is performed, typi- This study aims to improve the engineering practice of CFD
cally by increasing the number of cells. applications in a large-scale TF coal boiler to improve the de-
Table 1 lists some of the recent CFD studies on a TF furnace sign and diagnose operation, focusing on mesh sensitivity.
[3-18]. For a boiler capacity range of 150 - 800 MW, CFD was Three versions of mesh with up to 6 million cells were con-
performed using commercial codes for various purposes, in- structed for the standard 500 MW boiler in Korea. For each
cluding biomass co-combustion, NOx reduction, and the opti- mesh, CFD simulations were performed for three cases vary-
mization of operating parameters. The results were mostly ing the air distribution and burner tilting, with the use of ad-
validated using the values at the boiler exit (typically econo- vanced submodels for coal devolatilization, char conversion,

918
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

Table 2. Fuel properties of performance coal.

Analysis Coal
Total moisture 17.00
Inherent moisture 10.00
Proximate analysis
Volatile matter 31.36
(wt.%)
Fixed carbon 43.89
Ash 7.75
C 76.95
H 5.34
Ultimate analysis
(wt.%daf) O 16.01
N 1.19
S 0.51
HHV (kcal/kg) 5600

into its retrofit to improve its performance for low-rank coals


with poorer fuel quality than the original design coal. The instal-
Fig. 1. Schematic of 500 MW coal-fired boiler with proposed retrofit includ-
ing installation of separated overfire air (SOFA).
lation of separated OFA (SOFA) was also included to achieve
lower NOx emissions.
Table 2 presents the properties of the new performance coal
and heat transfer through user-defined functions. Based on the considered for the boiler retrofit; it was sub-bituminous with a
differences in the results, the mesh sensitivity was assessed in higher heating value (HHV) of 5600 kcal/kg, significantly lower
two ways: i) Local profiles of velocity, temperature, heat flux, than that of the original design coal (6300 kcal/kg).
and NO concentration, and ii) quantified values of key perform- In the operating condition for this study, 55.58 kg/s of coal
ance parameters. Then, an appropriate method for mesh sen- (equivalent to a thermal input of 1303 MWth) was pulverized
sitivity and sufficient degree of mesh fineness depending on and then injected through burners A-E transported by primary
the purpose of CFD simulation are discussed. air at 77 °C. Burner F was standby. During the pulverization,
the coal was dried to a residual moisture content of 8.6 %, and
therefore the moisture evaporated was mixed with the primary
2. Target boiler and operation conditions air. The secondary air (SA) with a total flow rate of 355.1 kg/s
Fig. 1 shows a schematic representation of the 500 MW at 326 °C was divided into the burner SA, CCOFA, and SOFA.
coal-fired boiler selected for this study, which consists of a The overall excess air ratio was 12.94 %.
burner zone with a TF configuration and a convective heat The detailed operating parameters varied into three cases for
exchanger section in the upper part. Its original design is often a detailed evaluation of mesh sensitivity. In case 1 (reference
referred to as the standard boiler in Korea, where 20 identical case), the flow ratios of CCOFA and SOFA were 5 % and 15 %
units are currently in operation. The burner zone had six levels of the total air, respectively, which resulted in an air-fuel
of coal burners (A to F levels) divided into three sections with equivalence ratio in the burner zone (λBZ) of 0.904. The burner
two burners each. Each section has an identical arrangement tilting of 15° upward was introduced in case 1T, which is per-
of coal and air supply ports, as illustrated in the inset of Fig. 1. formed during operation to control the heat absorption on the
In each burner, the flow of coal and primary air is split into con- wall. Deeper air staging was tested in case 2 by increasing the
centrated and weak ports, with more coal entering into the SOFA flow ratio to 20 % without tilting the burners.
former. Close-coupled over-fire air (CCOFA) is injected directly
above the burner zone and separated OFA (SOFA) 6.144 m 3. Numerical methods
above. The convective heat exchanger section incorporates a
3.1 Mesh constructed for the boiler
series of tube bundles for superheaters (SHs), reheaters (RHs),
and economizer (ECO) with different tube geometries and Three versions of the mesh were constructed for the boiler
steam conditions. The furnace itself is an evaporator built with with different numbers of cells, as illustrated in Fig. 2 and
membrane tube walls. The steam pressure and temperature summarized in Table 3. V1 has approximately 1.23 million cells.
leaving the final SH to the high-pressure turbine are 250.5 bar The mesh fineness for the burner zone was increased to 3.53
(gauge) and 596°C, respectively. The various characteristics of million cells in V2 and 5.47 million cells in V3, while maintaining
this boiler for combustion, heat transfer, NOx emission, and the identical mesh shape. The average cell volume in the
slagging/fouling have been previously reported in the Refs. [7, burner zone was 0.0114 m3 for V1, 0.0034 m3 for V2, and
8, 19]. Note that the convective section differs from that of the 0.0022 m3 for V3. The mesh employed hexahedral cells for
original boiler because this study was a part of the research most of the furnace volume, which was favorable for conver-

919
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

Table 3. Number of cells and quality in the three versions of mesh con- stability. The higher-order schemes caused instability in the
structed for the boiler. solution associated with radiation by particles especially in low-
Mesh V1 V2 V3 velocity regions.
Table 4 summarizes the detailed submodels that were
Total number of cells 1229410 3529358 5469908
adopted. In the discrete phase method (DPM), coal particles
Number of cells in the burner
zone
965121 3214169 5101373 were represented by 10 different diameters ranging between
5.9 μm to 204 μm, with an average of 50 μm. In total, 49560
Average cell volume in the
0.0114 0.0034 0.0022 particles were tracked, with considerations of the turbulent
burner-zone [m3/cell]
dispersion, coal devolatilization, and char conversion along the
Cell equiangle skewness 0.1281 0.1234 0.1268
path. The coal devolatilization was modeled as a single step
Orthogonality 0.9321 0.9365 0.9330
+
reaction with product yields and rate constants acquired using
Burner zone y 931 299 222
FLASHCHAIN of PCCoalLab [20]. Based on the semi-
empirical coal network model, FLASHCHAIN predicts the dy-
namics of coal devolatilization, providing the product yields and
kinetics. As the coal devolatilization finished within a short resi-
dence time at heating rates well over 103 K/s, the ultimate
yields of primary pyrolysis were taken for tar, CO, CO2, H2O,
CH4, C2H4, C2H6, C3H6, HCN, and H2S, and implemented into
the CFD code using a user-defined function (UDF). The C2+
hydrocarbons were simplified to an imaginary species of CxHy.
The yield and composition of tar were also slightly modified to
include the H and O fractions of char, which allowed the model-
ing of char as a pure carbon solid. The formation enthalpy of
tar was calculated according to the energy balance between
the input coal and devolatilization products. The reaction kinet-
ics was calculated using a single Arrhenius rate with rate con-
stants presented in Table 4.
The char conversions by O2, CO2, and H2O were solved us-
ing the unreacted core shrinking model (UCSM) [21], appropri-
ate for the high-temperature reaction in the industrial-scale
furnace. This model included three competing rates of hetero-
geneous reactions: The surface reaction on the char core, gas
diffusion onto the particle surface, and gas diffusion through
the ash layer. It was implemented in the CFD code using a
Fig. 2. Mesh constructed for the boiler with three versions of different fine- UDF.
ness in the burner zone. The gaseous reactions were modeled using the global reac-
tion schemes described by Jones and Lindstedt [22] for hydro-
gence. Tetrahedral cells were used for the top and bottom carbon and Smith and Smoot [23] for tar. The reaction rate was
ends of the burner zone leading to the heat exchanger section calculated using the kinetic-rate/diffusion-rate model [24] to
and the bottom hopper, respectively. Each heat exchanger in consider turbulence-chemistry interaction. Turbulence was
the upper furnace was simulated as a zone with local convec- solved by the realizable k-ε model [25] which considers the
tion and radiation heat transfer and directional flow resistances swirling flow better than the standard k-ε model and provides
depending on tube geometry and alignment as opposed to better convergence than the Reynolds stress model. Although
modeling the complex tube geometry. This modeling method advanced turbulence models such as the large eddy simulation
was implemented using user subroutines, which is explained are available [26], applying such models for engineering prac-
later. Therefore, the mesh in the heat exchanger sections in tice for the large-scale coal boiler is rare because of excessive
the upper furnace consisted of simple orthogonal hexahedral computation costs [27]. The models based on the Reynolds-
cells, and this was identical across the three mesh versions. averaged Navier-Stokes approach are common in engineering
applications listed in Table 1. Radiation was solved using the
discrete ordinate method with the weighted sum of the gray
3.2 CFD submodels
gases model for gas absorption [28].
CFD simulations were performed using ANSYS Fluent (ver- Regarding the heat transfer, the entire furnace wall (evapora-
sion 17.2). In the solution method, the PRESTO! scheme was tor) was assumed to have a fixed temperature of 652.15 K with
used for pressure-velocity coupling and the QUICK scheme for an overall heat transfer coefficient of 250 W/m2K and an inner
spatial discretization which provided the best convergence and wall emissivity of 0.7; the temperature corresponded to the

920
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

Table 4. Summary of submodels adopted for boiler simulation.

Category Submodels
Coal combustion -Devolatilization: FLASHCHAIN [20]
Dry coal→58.56 wt.%daf volatiles + 32.91 wt.%daf C(s)(Char)
Composition of Volatiles (wt.%daf): Tar 32.57, CO 3.31, CO2 3.89, H2O 14.35, H2 0.71, CH4 1.99, CxHy 1.73
-Devolatilization rate:
dV æ E ö 3
= A exp ç - ÷ (V0 - V ) ; E = 5.49 kcal/mol, A = 7.89×10 s-1
dt è RT ø
-Char combustion: Unreacted core shrinking model [21]
1
Rchar ,i =
1 1 1 æ1 ö i
( P - Pi * ) éëgcm-2 s-1 ùû
+ + ç - 1 ÷
kdiff ,i ks ,iY 2 kdash ,i è Y ø
kdash = kdiff ε 2.5 , Y = d char / d p
(R1) C(s) + 0.5 O2 → CO
0.75
ks = 8710 exp(-17967 / Ts ), kdiff = 1.383 ´ 10-3 (T / 1800 ) / ( Pt d p )
*
Pi - Pi = PO2
(R2) C(s) + H2O → CO+H2
0.75
ks = 247 exp(-21060 / Ts ), kdiff = 1 ´ 10-3 (T / 2000 ) / ( Pd
t p)

Pi - Pi = PH 2O - ( PH 2 × PCO ) / K eq , K eq = exp[17.644 - 30260 / (1.8Ts )]


*

(R3) C(s) + CO2 → 2 CO


0.75
ks = 247 exp(-21060 / Ts ), kdiff = 7.45 ´ 10-4 (T / 2000 ) / ( Pd
t p)

*
Pi - Pi = PCO2

Species, gas -Species: Tar, CO, CO2, H2, CH4, CxHy, H2, SO2, O2, N2
reaction -Reaction mechanism [22, 23]
(R4) CxHyOz (tar) + a .O2 → x CO + 0.5y H2
(R5) CnHm + 0.5n .O2 → n CO + 0.5m H2
(R6) CnHm + 0.5n H.2O → n CO + 0.5(m+n) H2
(R7) CH4 + 0.5 .O2 → CO + 2 H2
(R8) CH4 + 0.5 H.2O → CO + 2.5 H2
(R9) CO + H2O → CO2 + H2
(R10) H2 + 0.5 O2 → H2O
-Reaction rate: kinetic-rate/eddy-dissipation rate model [24]
Discrete phase -Lagrangian scheme with stochastic tracking for turbulence
-Number of particles: 49560
-Particle size: 5.88-204μm
Turbulence -Realizable k-ε model [25]
Radiation -Discrete ordinate method
-Gas absorption: Weighted sum of gray-gases model [28]
Tube bundles -Flow resistance by Jakob’s correlation [30]
-Volumetric heat source by convection (Zukauskas correlation [31]) and radiation
Wall condition -Overall heat transfer coefficient: 250 W/m2K
-Water/steam temperature: 352 °C
-Emissivity: 0.7
NOx -Thermal NOx: Extended Zeldovich mechanism [32]
-Fuel NOx: De Soete model [33] with volatile-N released as HCN and NH3 (5:1), and char-N as NO

average water/steam temperature between the inlet and outlet. vection, and radiation considering the detailed geometry infor-
The heat transfer coefficient was tuned to meet the design mation of tubes. The details of this method have been pre-
value of the wall heat absorption (562 MWth), but was consis- sented elsewhere [29]. Briefly, the inertial resistance in the
tent with the measured values (3.51-4.37 m2K/kW of thermal lateral and transverse directions was calculated using the Ja-
resistance equivalent to 229-285 W/m2K depending on coal kob’s correlation [30] as follows
types) in the boiler [19]. 0.14

The SHs, RHs, and ECO tube bundles were simplified as 2 f ¢Gmax 2 N L æ μs ö
Δp = ç ÷ (1)
porous zones having the source terms for flow resistance, con- ρ è μø

921
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

é ù -0.15 Table 5. Number of iterations and relative computation time for each of the
0.08 ( S L / D )
f ¢ = ê0.044 + 0.43+1.13 D / S L ú
Re (2) three meshes.
êë ( ST / D - 1) úû
Number of gas phase iterations DPM Relative
Method Cold flow Hot flow calcula- computation
where D denotes the tube diameter and S L and ST repre- Total tion time
simulation simulation
sent the spacings in the lateral and transverse directions, re-
V1 10000 10000 20000 333 1
spectively. The inertial resistance in each cell for the local ve-
locity and gas properties was updated using a UDF. Similarly, V2 10000 20000 30000 666 6.468
the heat transfer rate in each cell was calculated as the sum of V3 10000 22500 32500 750 9.444
convection ( q&conv ) and radiation ( q&rad ) as follows

k gas Nu
q&conv = f tuning As
D
(T
gas - Ttube ) éë W / m3 ùû (3)

q&rad = f tuning εσAs (Tgas 4 - Ttube 4 ) éë W / m3 ùû (4)

The Nusselt number ( Nu ) for convection was determined


using the Zukauskas correlation [31]:

Nu = 0.40 Re D 0.6 Pr 0.36 (5)

In the above equations, As denotes the specific tube surface


area per unit volume. The heat transfer rates were also imple-
mented using a UDF. Due to the simplified calculation and the
thermal resistance by slagging/fouling, a correction factor
( f tuning ) was introduced in Eqs. (3) and (4) to match the pre-
dicted heat transfer rates to the design values. Fig. 3. Comparison of heat absorption distribution between design values
and CFD results.
Fig. 3 compares the distributions of heat absorption in the
heat exchangers between the design values and the CFD re-
sults for the mesh V2. The heat transfer rates predicted by variation corresponded to 0.33 % variation in the temperature
CFD were very close to the design values in all of the heat profile and 0.12 % variation in the wall heat absorption for V2.
exchangers, which implies that the predicted temperature dis- Similar variations were monitored for other meshes.
tribution along the furnace height was reasonable. The computation times to reach the convergence for V2 and
V3 were 6.47 and 9.44 times that of V1, as also listed in Table 5.
3.3 Coupled iteration with the DPM
3.4 Post-processing for NO emission and ash
In the solution method, the DPM calculations coupled with slagging
the gas phase (hot flow simulation) were conducted after the
initial air flow pattern was established with 10000 iterations of NO emissions by thermal and fuel NOx mechanisms were
the gas phase alone (cold flow simulation). In the hot flow predicted by post-processing the CFD results. The prompt NO
simulation, each DPM calculation was performed in every 30 mechanism was ignored as its contribution was negligible. The
gas phase iterations, and the source terms from the particles thermal NOx reactions were based on the extended Zeldovich
were updated with an under-relaxation factor of 0.15. This mechanism with rate constants taken from Hanson and Sali-
meant that at least 83 times of DPM calculation and 83×30 mian [32]. The local concentrations of O and OH radicals were
iterations for the gas phase were required for the updated determined by the partial equilibrium assumption. The fuel NOx
source term to reach 99.9999 % of the total, even without any mechanism considered the release of fuel-N during the devola-
interactions with the gas phase. As summarized in Table 5, the tilization and char conversion [33]. The partitioning of fuel-N
actual solution required a substantially larger number of itera- into volatile-N and char-N was acquired using FLASHCHAIN.
tions (10000 iterations for the mesh V1, 20000 for V2, and In the volatile-N, the N intermediates were assumed to be HCN
22500 for V3 after the cold flow simulation) for the conver- 5: NH3 1. Char-N was directly converted into NO. The reduction
gence with a criterion of less than 3 % variation over 500 itera- of NO to N2 by active carbon on the char surface was also
tions in the velocity profile along the centerline. This involved considered, with a BET surface area of 100 m2/g.
repeated DPM tracking by 333 to 750 times depending on the Furthermore, the ash slagging caused by the deposition of
mesh version, which decreased the dependency on the limited sticky ash particles onto the wall was evaluated for the three
number of particles (49560) at the same time. Also, the velocity mesh versions. Ash slagging is associated with boiler efficiency

922
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

[m/s]

(a) V1 (b) V2 (c) V3

Fig. 4. Comparison of velocity contours on selected cross-sections for case 1.

and operational issues. It reduces the wall heat transfer and, if μash = 4.0378 ´ 10-11T exp ( 36.9753 ´ 103 / T ) (11)
severe, damages the boiler. The slagging tendency was pre-
dicted by developing a UDF that integrated the rate of ash ηc was introduced to ηcapture based on the fact that the char
particles impacting on the wall cells by post-processing of the particle requires sufficient ash buildup on its surface to be
CFD results for trajectories of approximately 2.2 million parti- sticky. The critical carbon conversion ( x crt ) as the threshold
cles. value was 0.9 [36].

m& slag = åm& impact ηcapture [ kg / s ] (6) 4. Results and discussion


π
m& impact = w ρd p 3 [ kg / s ] (7) 4.1 Comparison of results for case 1
6
Fig. 4 compares the velocity magnitude on the center and
The particle mass ( m& impact ) included a weighting factor ( w ) horizontal planes between the three meshes in case 1. On each
to consider the actual mass flow rate represented by each horizontal cross-section, the jet flow from each corner can be
particle tracked. It was multiplied by capture efficiency ( ηcapture ) identified clearly. This developed a swirling flame (fireball) over
[34] which was determined by the sticking probability ( ηv ) and 20 m/s across the burner zone. The fireball had three crescent-
a threshold of carbon conversion ( ηc ) for each particle. shaped high-velocity regions on the vertical cross-section, which
coincides with the three burner sections (A/B, C/D, and E/F
ηcapture = ηc ηv (8) burners). This was because the fireball was forced towards the
ì 1 if μash < μcrt center by the air supply in the three burner sections and was
where ηv = í (9) then expanded to the wall by the centrifugal force in the gap
î crt ash if μash > μcrt
μ / μ
between the burner sections. Above the burner zone, the swirl-
ì 0 if x < xcrt ing flow expanded to the wall because the momentum keeping
and ηc = í (10)
î 1 if x > xcrt the fireball at the center no longer existed, resulting in an
enlarged low-velocity region at the center. The meshes V2 and
ηv was based on the ash viscosity of ash with critical vis- V3 exhibited similar patterns when these high- and low-velocity
cosity ( μcrt ) varying between 104 and 108 Pa·s [34]. In this regions were compared. By contrast, differences were notice-
study, 105 Pa·s was selected for comparison between the able in V1 in various locations, as indicated by dotted circles in
meshes. If the particle temperature was high enough to have Fig. 4(a). The results are evaluated quantitatively later.
the viscosity below the critical viscosity, ηv becomes 1; oth- Fig. 5 compares the temperature distribution for the refer-
erwise, it takes the value of the viscosity ratio. The ash viscos- ence case. The fireball had high temperatures as a result of the
ity was determined using the Urbain model [35] for the compo- mixing and reactions between coal and air. The temperature
sition of design coal (CaO 2.69, MgO 0.90, SiO2 57.29, Al2O3 gradually increased along the burner height as more coal and
29.34, Fe2O3 4.39, Na2O 0.30, K2O 0.70, TiO2 1.30, and mis- air were supplied, and ultimately peaked in the uppermost
cellaneous 3.09 wt.%), as follows. burner level. It then decreased with the injection of SOFA. As

923
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

[℃]

(a) V1 (b) V2 (c) V3

Fig. 5. Comparison of temperature contour for case 1.

soon as the gas entered the heat exchanger zone, the tem- velocity magnitude and 2.2 % for temperature, whereas it was
perature dropped rapidly. Part of the flow from the A/B burner lowered to 6.4 % and 0.6 %, respectively, for V2. In particular,
section entered the bottom hopper and was cooled down the velocity profile across the fireball in V1 was significantly
through heat transfer to the wall. The cold flow then entered the attenuated compared to that of V2 and V3.
center of the fireball, creating a convex shape in the low- The results were further compared qualitatively for heat trans-
temperature region in the lower burner levels. By comparing fer, char conversion, and NOx reaction, which are associated
the temperature patterns between the meshes, it was observed with the performance of the boiler design and operation. Fig. 7
that the convex shape of the low-temperature region at the compares the total heat flux by radiation and convection on the
center of the fireball was narrower in V1. In addition, the high- wall among the three meshes for the reference case. The heat
temperature region in the E/F burner section was enlarged. flux increased when the hot flow including particles (char and
Some differences were noticeable on the horizontal cross- ash) approached the wall. Compared to the complex pattern of
sections between V2 and V3, but these were not significant to V3, the region of high heat flux, above 200 kW/m2, between the
those between V1 and V3. burner C level and SOFA was flattened significantly in V1. This
For quantitative analysis, the temperature and velocity mag- was associated with the numerical diffusion, which did not cap-
nitude profiles along the vertical centerline across the furnace ture the secondary flow patterns and sharp temperature varia-
and the horizontal centerline on the burner D level are com- tions near the wall sufficiently. By contrast, V2 exhibited very
pared in Fig. 6. The temperature profiles were very similar be- similar patterns to those of V3.
tween the meshes. However, the velocity magnitude profile of Fig. 8 compares the reaction rate of char by O2 between the
V1 deviated largely from that of V3, whereas V2 had a very three meshes for the reference case. The char reactions oc-
similar pattern to that of V3. For the data sets, the relative er- curred rapidly in the high-velocity regions of the fireball where
rors in the results of V1 and V2 were determined with respect O2 was available. On the other hand, the reaction was not sig-
to those of V3, as follows. nificant at the center of the fireball because the char particle
concentration was low and O2 was depleted. Above the burner
zone, the char reaction continued in the fireball up to the first
Relative error ( % ) =
ò φ - φ ds
i 3
(12) heat exchanger section (primary SH). The high concentration
òφ ds
3
of char particles near the wall without efficient mixing led to the
release of unburned carbon in fly ash. Although the reaction in
φi denotes the temperature or velocity magnitude for a the mesh V1 was more spread out at the center of the fireball,
mesh version i, and s denotes the coordinate along the line. the char reaction patterns were overall similar between the
The relative errors for V1 were 17.6 % for velocity magnitude mesh versions.
and 1.1 % for temperature on the vertical centerline. By con- Fig. 9 compares the NO reaction rate between the three
trast, the values for V2 were decreased to 4.8 % and 0.7 %, meshes for the reference case. Here, a negative value indicates
respectively. Similar trends were observed in the profiles on the the reduction of NO by the fuel and thermal NO reaction
horizontal centerline. The relative error of V1 was 28.9 % for mechanisms. NO was rapidly generated near the burner jets

924
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

(a) (b)

(c) (d)

Fig. 6. Profiles of velocity magnitude and temperature on the vertical centerline (a) and (b); on the horizontal centerline of the burner D level (c) and (d) for
case 1.

[kW/m2]
formation above 1x10-6 kmol/m3s coincided with those of rapid
char oxidation shown in Fig. 8. Outside these regions, NO re-
duction reactions were active due to the depletion of O2. In
particular, the reduction reactions were dominant in the wide
volume between the top of the burner zone and SOFA. This
was the main purpose of installing the SOFA in the boiler retro-
fit. Overall, the reaction patterns were similar among meshes
because of the identical reaction stoichiometry, as determined
by the distribution of combustion air.
If the local profiles of wall heat flux or NO concentration were
analyzed, the magnitude of numerical diffusion increased to a
similar degree with those of temperature and velocity magni-
tude. For example, the relative error in the profile of wall heat
flux shown in Fig. 10(a) was 5.6 % for the mesh V1 and 2.8 %
for V2. These values were comparable to those for tempera-
ture. The NO concentration as shown in Fig. 10(b) was more
sensitive to the mesh with relative errors of 16.2 % and 7.6 %
(a) V1 (b) V2 (c) V3
for V1 and V2, respectively. In particular, the profile was flat-
Fig. 7. Comparison of wall heat flux for case 1. tened significantly in the NO profile for V1, which can be mis-
leading if employed for detailed analysis to adjust the burner
design and operation for NOx reduction.
by the oxidation of HCN and NH3 from coal devolatilization and In addition to comparing profiles, the key parameters associ-
release of NO from the char-N conversion. The regions of NO ated with boiler performance were analyzed: Carbon conversion,

925
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

Table 6. Comparison of key performance parameters for cases 1, 2 and 1T.

Case Case 1 Case 2 Case 1T


Mesh V1 V2 V3 V1 V2 V3 V1 V2 V3
Char + O2 66.88 63.98 65.19 65.02 62.31 64.59 70.55 66.15 66.61
Char + CO2 21.48 23.21 22.61 23.73 26.05 24.21 19.10 21.50 21.26
Carbon conversion [%]
Char + H2O 11.42 12.45 11.96 10.97 11.22 10.83 10.35 12.36 12.13
Total 99.78 99.64 99.76 99.72 99.59 99.64 99.86 99.81 99.78
Exit O2 [mol.%] 2.485 2.494 2.476 2.495 2.516 2.515 2.474 2.477 2.483
Exit NO [ppm, 6%O2] 109.4 107.4 109.4 98.8 97.7 98.3 126.4 112.3 109.0
Total 592.4 598.6 599.5 577.1 580.7 583.7 577.47 579.70 574.91
Furnace wall heat
Radiation 555.7 560.5 561.9 541.0 543.7 546.9 540.53 542.24 538.25
absorption [MWth]
Convection 36.7 38.1 37.6 36.1 37.0 36.8 36.94 37.46 36.66

[kgmol/m3·s] exit NO and O2 concentrations, and wall heat absorption.


(log-scale)
These values have been used for mesh sensitivity analysis in
several studies listed in Table 1. Table 6 compares the pa-
rameters for the three mesh versions. Surprisingly, the values
between the meshes were found to be very similar. In particu-
lar, the results of V1 deviated by less than 2 % from those of
V3, despite the large differences in the flow and temperature
patterns in the burner zone. This is further discussed with the
results for cases 2 and 1T.
Finally, the ash deposition rate on the front wall was calcu-
lated by post-processing of the CFD results. As shown in Fig.
11, it was influenced by the swirling fireball and the secondary
flow pattern circulating near the wall. The deposition patterns
were overall similar, but the deposition rate showed a strong
dependence on the mesh fineness. In terms of the total deposi-
tion rate, the mesh V1 was 32.5 % larger than that of V3
whereas V2 was 7.3 % larger. This was comparable to the
(a) V1 (b) V2 (c) V3 relative errors in the velocity profile mentioned previously. The
Fig. 8. Comparison of char oxidation rate for case 1. results suggest that V1 would not be acceptable for the quanti-
tative evaluation of ash slagging. Even V3 may not be fine
[kmol/m3·s]
enough to be free from the mesh sensitivity issue for the ash
deposition rate. The strong mesh dependence in ash slagging
is associated with the flow pattern in the near-wall region. In a
fine mesh, detailed velocity distribution can be acquired in
small near-wall cells for particle trajectories to become evasive
from wall impaction. In contrast, a single velocity vector in the
large near-wall cell of a coarse mesh can lead to over-
estimation of ash deposition. Therefore, a fine mesh should be
used because of the strong mesh dependence if the purpose
of the CFD study is on ash slagging and related behaviors.

4.2 Comparison of results for case 2


In order to confirm that the same level of differences in the
key performance parameters between the three meshes is
achieved for different operating conditions, the proportion of
SOFA was increased from 15 to 20 % in case 2. Correspond-
ingly, the amount of secondary air in the burner zone was de-
(a) V1 (b) V2 (c) V3 creased. Table 6 includes the key performance parameters
Fig. 9. Comparison of NO reaction rate for case 1. between the meshes for case 2. Due to the reduced λBZ (case

926
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

(a) V1 (b) V2 (c) V3

Fig. 12. Comparison of z-velocity contour on the burner B level in case 1T.

(a)

(a)

(b)

Fig. 10. Profiles of (a) wall heat flux on the side wall; (b) NO concentration
on the centerline of burner D level in case 1.

(b)

Fig. 13. Comparison of (a) carbon concentration; (b) velocity magnitude


profile on the centerline of burner A level for case 1T.

versions exhibited the same trend. The heat absorption on the


furnace wall tended to become lower in the meshes V1 and V2
with deviations of 1.13 % and 0.51 % from that of the mesh V3,
(a) V1 (b) V2 (c) V3
respectively. The deviations in NOx emission from that in V3
Fig. 11. Comparison of ash deposition rate on the front wall. were within 0.6 ppm.
The above results suggest that the key performance pa-
1: 0.904, case 2: 0.847), a decrease in gas temperature and rameters have little dependence on the mesh fineness for
depletion of oxygen in a wider region was observed. As a result, change in the air distribution. This was because these parame-
the heat absorption in the furnace wall was 15-18 MWth lower ters were determined by integration over the entire furnace
than that in case 1. Further, NOx emission decreased by ap- surface or volume. This made them relatively insensitive to
proximately 10 ppm with a smaller contribution of thermal NO local variations in temperature, velocity, or species concentra-
mechanism. Similar to the results in case 1, the three mesh tion and, therefore, the influence of numerical diffusion was

927
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

diminished. Instead, the parameters were more influenced by and shape/size of air supply ports.
overall input variables such as burner zone stoichiometry (λBZ)
and fuel properties that shifted the entire reaction and flow
characteristics in a certain direction. On the other hand, a mesh
5. Conclusions
comparable to V1 can be acceptable if the purpose is to inves- This study investigated the issues associated with mesh
tigate the influence of those major variables on the key per- sensitivity to improve the current engineering CFD practice
formance parameters. applied to the design and operation diagnostics of a coal-fired
boiler. Differences in local variables, ash slagging on the wall,
4.3 Comparison of results for case 1T and key performance parameters were evaluated for three
versions of the mesh with average cell volumes ranging be-
The burner tilting by 15° upward in case 1T is a minor tween 0.0114 and 0.0022 m3/cell which corresponded to ap-
change in the operation parameter from case 1. Table 6 also proximately 0.97 and 3.2 million cells in the burner zone for the
compares the key performance parameters acquired for this 500 MW TF boiler selected in this study. The mesh sensitivity
case. Because the fireball and corresponding high heat flux could be evaluated better by velocity magnitude profile which
regions on the wall were moved upward by the burner tilting, was the most sensitive to the mesh fineness compared to NO
the wall heat absorption was decreased by approximately 25 concentration, temperature, and wall heat flux. The mesh ver-
MWth to 574.9 MWth from case 1 for V3, which in turn increased sion with 0.0114 m3/cell led to a larger fireball formed in the
the proportion of heat absorption in the tube bundles. V1 and burner zone and flattened the heat flux on the wall because of
V2 had similar values in the wall heat absorption and carbon the increased numerical diffusion. In contrast, the overall per-
conversion with V3. However, a significant difference was ob- formance parameters, such as the total wall heat absorption,
served in NO emission for V1 which was 17.4 ppm higher than NO and O2 concentrations, and carbon conversion, were not
that of V3. The difference was large enough to mislead the sensitive to mesh size. This suggests that the use of a coarse
effect of burner tilting. In contrast, V2 was 3.3 ppm higher, mesh is acceptable for evaluating the influence of major design
which is an acceptable deviation. and operating variables such as the burner zone stoichiometry
Elucidating the reason for the larger NO emission in V1 for and fuel properties on those parameters. Therefore, the find-
this case required a detailed analysis of the results which was ings of CFD studies with relatively coarse meshes are justified
caused by the subtle change in the operating parameter. It was in relation to the reported literature. However, a mesh finer than
associated with different secondary flow patterns that influ- 0.0034 m3/cell is required to accurately capture the secondary
enced particle trajectories, char conversion, and fuel NOx reac- flow pattern and associated particle behaviors if minor de-
tions. Fig. 12 compares the z-velocity contour on the burner B sign/operation changes are to be considered such as burner
level. The secondary flows near the wall in V1 were stronger tilting and yawing. Finally, the finest mesh should be employed
and headed upward (positive z-velocity) whereas those in V2 for analysis of ash slagging because of its strong mesh de-
and V3 were weaker and mostly downward. Because the char pendence in the near-wall region.
particles were concentrated in near-wall region, those from the
lowermost burners in V1 were mostly entrained with the up-
ward swirling flow. In contrast, the particles in V2 and V3
Acknowledgments
flowed downward to enter the bottom cone region or upward This work was supported by the Korea Institute of Energy
along the near-wall region. Those in the bottom cone region Technology Evaluation and Planning (KETEP) affiliated with
then flowed upward through the center of the fireball. As a the Ministry of Trade, Industry and Energy of the Korean Gov-
result, char particles were more concentrated in the near-wall ernment (Contract No. 20181110200190). The authors would
region on the burner A level in V2 and V3 as shown in Fig. also like to thank Doosan Heavy Industries and Construction
13(a). According to the fuel NO mechanism, the char particles for their help and support.
produced NO from char-N and at the same time reduced NO to
N2. The resultant NO concentration became lower compared to
References
that in V1. Such a difference would have been difficult to notice
while using the velocity magnitude profile on the centerline as [1] M. A. Gonzalez-Salazar, T. Kirsten and L. Prchlik, Review of
shown in Fig. 13(b). the operational flexibility and emissions of gas-and coal-fired
To summarize, the mesh sensitivity was more influential for power plants in a future with growing renewables, Renewable
the minor operational change and the large relative error in the and Sustainable Energy Reviews, 82 (2018) 1497-1513.
velocity profile of V1 led to significant differences in NO emis- [2] V. V. Ranade and D. F. Gupta, Computational Modeling of
sion and particle behaviors. Therefore, a mesh comparable to Pulverized Coal Fired Boilers, CRC Press (2014).
V2 or finer should be used to achieve improved accuracy for [3] H. Gao, A. Runstedtler, A. Majeski, P. Boisvert and D. Camp-
simulations requiring detailed analyses of flow patterns and bell, Optimizing a woodchip and coal co-firing retrofit for a
local flow/reaction phenomena by minor changes in design or power utility boiler using CFD, Biomass and Bioenergy, 88
operation variables such as burner angle for tilting or yawing (2016) 35-42.

928
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

[4] X. Zhang, J. Zhou, S. Sun, R. Sun and M. Qin, Numerical in- [18] T. Chen and W. Z. Wang, Modeling of combustion and hy-
vestigation of low NOx combustion strategies in tangentially- drodynamics for a coal-fired supercritical boiler with double-
fired coal boilers, Fuel, 142 (2015) 215-221. reheat cycle, International Journal of Numerical Methods for
[5] N. Modlinski, Computational modeling of a utility boiler tangen- Heat & Fluid Flow (2019).
tially-fired furnace retrofitted with swirl burners, Fuel Process- [19] H. Y. Park, J. E. Lee, H. H. Kim, S. Park, S. H. Baek, I. Ye
ing Technology, 91 (11) (2010) 1601-1608. and C. Ryu, Thermal resistance by slagging and its relation-
[6] S. R. Gubba, D. B. Ingham, K. J. Larsen, L. Ma, M. Pourka- ship with ash properties for six coal blends in a commercial
shanian, H. Z. Tan, A. Williams and H. Zhou, Numerical model- coal-fired boiler, Fuel, 235 (2019) 1377-1386.
ling of the co-firing of pulverised coal and straw in a 300 MWe [20] S. Niksa, PC Coal Lab Version 4.1: User Guide and Tutorial,
tangentially fired boiler, Fuel Processing Technology, 104 Niksa Energy Associates LLC, Belmont, CA (1997).
(2012) 181-188. [21] C. Y. Wen and T. Z. Chaung, Entrainment coal gasification
[7] C. R. Choi and C. N. Kim, Numerical investigation on the flow, modeling, Industrial & Engineering Chemistry Process Design
combustion and NOx emission characteristics in a 500 MWe and Development, 18 (4) (1979) 684-695.
tangentially fired pulverized-coal boiler, Fuel, 88 (9) (2009) [22] W. P. Jones and R. P. Lindstedt, Global reaction schemes for
1720-1731. hydrocarbon combustion, Combustion and Flame, 73 (3)
[8] K. Jang, K. Han, G. Lee, S. H. Baek, H. Y. Park and K. Y. Huh, (1988) 233-249.
Prediction of the ash deposition characteristics of blended [23] L. D. Smoot and P. J. Smith, Coal Combustion and Gasifica-
coals in a 500 MWe tangentially fired boiler, Energy & Fuels, tion, Springer Science & Business Media (2013).
32 (7) (2018) 7827-7840. [24] B. F. Magnussen and B. H. Hjertager, On mathematical mod-
[9] B. Akkinepally, J. Shim and K. Yoo, Numerical and experimen- eling of turbulent combustion with special emphasis on soot
tal study on biased tube temperature problem in tangential fir- formation and combustion, Proceedings of the Combustion In-
ing boiler, Applied Thermal Engineering, 126 (2017) 92-99. stitute, 16 (1977) 719-729.
[10] P. Tan, Q. Fang, S. Zhao, C. Yin, C. Zhang, H. Zhao and G. [25] T. H. Shih, W. W. Liou, A. Shabbir, Z. Yang and J. Zhu, A
Chen, Causes and mitigation of gas temperature deviation in new k-ϵ eddy viscosity model for high Reynolds number turbu-
tangentially fired tower-type boilers, Applied Thermal Engineer- lent flows, Computers & Fluids, 24 (3) (1995) 227-238.
ing, 139 (2018) 135-143. [26] C. D. Argyropoulos and N. C. Markatos, Recent advances on
[11] L. I. Díez, C. Cortés and J. Pallarés, Numerical investigation the numerical modelling of turbulent flows, Applied Mathemati-
of NOx emissions from a tangentially-fired utility boiler under cal Modelling, 39 (2015) 693-732.
conventional and overfire air operation, Fuel, 87 (7) (2008) [27] W. P. Adamczyk, B. Isaac, J. Parra-Alvarez, S. T. Smith, D.
1259-1269. Harris, J. N. Thornock, M. Zhou, P. J. Smith and R. Zmuda,
[12] Y. Liu, W. Fan and Y. Li, Numerical investigation of air-staged Application of LES-CFD for predicting pulverized-coal working
combustion emphasizing char gasification and gas tempera- conditions after installation of NOx control system, Energy, 160
ture deviation in a large-scale, tangentially fired pulverized-coal (2018) 693-709.
boiler, Applied Energy, 177 (2016) 323-334. [28] T. F. Smith, Z. F. Shen and J. N. Friedman, Evaluation of
[13] S. Chen, B. He, D. He, Y. Cao, G. Ding, X. Liu, Z. Duan, X. coefficients for the weighted sum of gray gases model, Journal
Zhang, J. Song and X. Lia, Numerical investigations on differ- of Heat Transfer, 104 (4) (1982) 602-608.
ent tangential arrangements of burners for a 600 MW utility [29] J. K. Park, S. Park, M. Kim, C. Ryu, S. H. Baek, Y. J. Kim and
boiler, Energy, 122 (2017) 287-300. H. Y. Park, CFD analysis of combustion characteristics for fuel
[14] Y. C. Liu, W. D. Fan and M. Z. Wu, Experimental and numeri- switching to bioliquid in oil-fired power plant, Fuel, 159 (2015)
cal studies on the gas velocity deviation in a 600 MWe tangen- 324-333.
tially fired boiler, Applied Thermal Engineering, 110 (2017) [30] M. Jakob, Heat transfer and flow resistance in cross flow of
553-563. gases over tube bank, Transaction ASME, 60 (1938) 384.
[15] Q. Zha, D. Li, C. A. Wang and D. Che, Numerical evaluation [31] R. K. Hanson and S. Salimian, Survey of rate constants in the
of heat transfer and NOx emissions under deep-air-staging N/H/O system, Combustion Chemistry, Springer, New York,
conditions within a 600 MWe tangentially fired pulverized-coal NY (1984) 361-421.
boiler, Applied Thermal Engineering, 116 (2017) 170-181. [32] G. G. De Soete, Overall reaction rates of NO and N2 forma-
[16] P. Tan, D. Tian, Q. Fang, L. Ma, C. Zhang, G. Chen, L. Zhong tion from fuel nitrogen, Proceedings of the Combustion Institute,
and H. Zhang, Effects of burner tilt angle on the combustion 15 (1975) 1093-1102.
and NOx emission characteristics of a 700 MWe deep-air- [33] A. Žukauskas, Heat transfer from tubes in crossflow, Ad-
staged tangentially pulverized-coal-fired boiler, Fuel, 196 vances in Heat Transfer, 8 (1972) 93-160.
(2017) 314-324. [34] L. Y. Huang, J. S. Norman, M. Pourkashanian and A. Wil-
[17] H. Y. Park, M. Faulkner, M. D. Turrell, P. J. Stopford and D. S. liams, Prediction of ash deposition on superheater tubes from
Kang, Coupled fluid dynamics and whole plant simulation of pulverized coal combustion, Fuel, 7 (1996) 271-279.
coal combustion in a tangentially-fired boiler, Fuel, 89 (8) [35] G. Urbain, F. Cambier, M. Deletter and M. R. Anseau, Viscos-
(2010) 2001-2010. ity of silicate melts, Transactions and Journal of the British Ce-

929
Journal of Mechanical Science and Technology 34 (2) 2020 DOI 10.1007/s12206-020-0141-4

ramic Society, 80 (1981) 139-141. Hyunbin Jo is a Ph.D. student at the


[36] S. Li, Y. Wu and K. J. Whitty, Ash deposition behavior during School of Mechanical Engineering,
char-slag transition under simulated gasification conditions, Sungkyunkwan University, who received
Energy & Fuels, 24 (2010) 1868-1876. his B.S from the same school. His re-
search subject is computational model-
ing of coal and biomass combustion in
Changkook Ryu is Professor at the industrial-scale boilers for NOx reduction
School of Mechanical Engineering, and improved performance.
Sungkyunkwan University. He received
his B.S., M.S. and Ph.D. in Mechanical
Engineering from Korea Advanced Insti-
tute of Science and Technology (KAIST).
His research interests include combus-
tion, gasification, and pyrolysis of solid
fuels, and various applications to industrial-scale plants.

930

You might also like