Professional Documents
Culture Documents
Dixon 2005
Dixon 2005
Dixon 2005
www.fuelfirst.com
Abstract
Major advances have been made in the development of bagasse combustion technologies and understanding the many processes involved
in bagasse combustion and steam generation. CFD modeling has come to form an integral and critical part of this progression. The experience
with CFD in the sugar industry through Sugar Research Institute has encompassed the full range of applications from fundamental code
development, through the generation and commercialization of new ideas and technologies, to the resolution of practical plant problems. The
paper summarises the numerous applications where SRI has achieved successful results utilizing CFD. It is demonstrated that the full benefits
of CFD in the delivery of commercial outcomes, for new technologies and the solution of operating plant problems, are achieved through the
close interaction between the code development and validation via full scale plant simulation. This two-way interaction enhances the code
fundamentals by focusing on practical issues and similarly increases the confidence in the capabilities and accuracy of the CFD predictions,
to an extent that justifies firm engineering decisions on commercial plant based solely on the characteristics predicted by the code.
q 2004 Elsevier Ltd. All rights reserved.
have good agreement with qualitative trends and acceptable simulation. Information about particle impacts with indi-
agreement with quantitative data. vidual tubes (particle velocity, particle concentration and
impact angle) can be predicted and erosion rates estimated
using the empirical tube wear correlation.
2. Basis of CFD model Enhancement of the code is in progress for Conditional
Moment Closure (CMC) modeling of CO and NOx
SRI’s workhorse for CFD modeling in the combustion generation and dispersion/reaction specifically for bagasse
and boiler fields is the FURNACE code originally combustion.
developed for pulverized coal fired boilers at the University
of Sydney [1]. Use is also made of the CFX code for specific
applications where an unstructured grid configuration is
3. Model applications
required for adequate problem representation.
FURNACE is a three dimensional, structured grid CFD
In this section a range of CFD modeling applications are
code that can predict the standard gas and particle flow
presented. Depending on the particular design of the boiler
patterns, combustion and radiation in utility and industrial
under investigation, the modeling grid is assembled with
boilers firing a range of fuels. The code also has time
appropriate porous regions relevant to the problem. Fig. 1
stepping capability. The code has been modified over a
shows a typical grid configuration for a bagasse fired boiler
number of years to model the flow and combustion
incorporating a single stage superheater, asymmetrical
processes in bagasse fired boilers [2–4]. Details of a more
multiple pass convection tube bank and large gas transfer
recent version of the FURNACE code are described
duct to the vertical airheater/economiser passage.
elsewhere [5,6]. Some comments are pertinent. The
standard k-3 turbulence model is used for all simulations.
Lagrangian particle tracking is used to predict the motion of 3.1. Tube erosion
particles (bagasse and ash) through the boiler. Turbulent
dispersion of the particles is modelled using the technique of Bagasse fired boilers have convection bank tube
Gosman and Ioannides [14]. The code has been modified to configurations that can be susceptible to excessive tube
take into account the shape and aerodynamic behaviour of erosion. The tendency of these designs to erode is
bagasse particles. Bagasse combustion is modelled as a accentuated by the gas flow dynamics that differentially
four-stage process: particle heat-up, particle drying, devo- concentrate the ash particle streams and alter the angle of
latilisation and char burnout. impact onto tubes into directions that further increase
Several novel features of the code have been developed erosion rates. FURNACE simulations of the more intract-
in recent years. These include the dynamic simulation of the able problems typically use a dual particle size distribution
pile burning characteristics of bagasse that has deposited out consisting of representative ‘large’ size/variable density
of suspension on the boiler grate [5,7], gas turbine exhaust particles for bagasse and ‘small’ dense particles for the ash
injection in combined cycle configuration (unpublished fraction. It has been found in practice that specific tube
work), the simulation of tube erosion processes in individual erosion areas can be differentiated according to the different
tubes of heat transfer tube bundles including the incorpor- ash particle size fractions that occur in the gas circuit.
ation of porous regions [4,8], and the accurate simulation of
global convective heat transfer within tube bundles,
incorporating empirical correlations [9], for the prediction
of heat transfer coefficients within porous regions that are
used to represent the flow through tube bundles [6]. This
takes into account the pressure and temperature drop and
flow area restriction caused by the presence of the tube
banks. For tube erosion work, sub-grid modeling capability
has been added to the FURNACE code. This overcomes the
computer processing and storage limitations that prevent
detailed modeling of the flow patterns around individual
tubes. To use this approach, a FURNACE simulation of the
whole boiler is performed but with the inlet conditions
(particle velocity, particle size, etc.) to a user defined local
region within the boiler recorded by the model. These inlet
conditions are then used in a subsequent FURNACE
simulation of the flow patterns in this local region. As the Fig. 1. Side elevation view of the flow grid used to model the Union Saint
physical size of the local region is small, a much finer grid Aubin boiler (Mauritius) showing the positions of the superheater and
resolution (with improved accuracy) can be used for this convection bank porous regions.
T.F. Dixon et al. / Fuel 84 (2005) 1303–1311 1305
Fig. 2. Flow simulations for the as-constructed design. (a) Gas velocity; (b) particle trajectory.
Figs. 2 and 3 show a typical CFD erosion application in a 3.2. Convection bank heat transfer
tube bank for an Australian boiler. Gas velocity contours (a)
and trajectories for several particle fractions (b) are shown A difficulty that has emerged with tube erosion
for the as-constructed design (Fig. 2) and the modified simulations has been the conflict between reduced erosion
design (Fig. 3). and reduced convective heat transfer. It is rare that
Both the geometry of the tube rows in the bank and the improvements in both processes can be achieved simul-
configuration of flow deflection baffles have been manipu- taneously. In most cases the heat transfer of a tube bank will
lated to arrive at a reasonable resolution of the problem. It be decreased (less degree of crossflow) in order to achieve
has been found that satisfactory improvements in boiler improved flow conditions that reduce tube erosion. It has
tube erosion performance can be deduced by visual been necessary to include in FURNACE the simulation of
assessment alone of the predicted flow and trajectory global convective heat transfer across tube bundles. This
patterns. Particular attention is given to particle concen- approach has been taken to reduce computational effort
tration effects, altered angles of impact and gas velocity rather than to attempt the simulation of heat transfer to each
distributions. Only in difficult flow conditions is use made individual tube in the bundle. Standard correlations for
of the FURNACE capability to simulate gas and particle relevant correction factors for heat transfer to tube bundles
flows and calculate erosion rates at specific locations on have been incorporated [9]. These are applied in a
individual tubes [8]. For several boilers, second stage baffle framework that utilizes the predictive capability of FUR-
adjustments have been undertaken to refine flow patterns NACE to specify the point values of gas velocity and
based on observed changes in erosion patterns. These orientation throughout the porous region that defines the
adjustments usually involve only relatively small changes to tube bundle.
a baffle geometry or size to achieve the required erosion The global convective heat transfer in the tube bundle is
reduction. A summary of several tube erosion investigations matched to measured data by the use of a matching
can be found in [11]. coefficient or scaling factor that is applied throughout
Fig. 3. Flow simulations for the modified design. (a) Gas velocity; (b) particle trajectory.
1306 T.F. Dixon et al. / Fuel 84 (2005) 1303–1311
Fig. 4. Gas flow distribution patterns at an airheater inlet before (a) and after (b) modifications.
recirculation within the furnace is deliberately enhanced. the tip of the furnace exit baffle, centrifuging the larger
This is intended to achieve three objectives; (i) stabilize and partially burnt fuel particles into the recirculation flow
anchor the main flow column in the rear 25% of the furnace adjacent to the front wall, and (iii) establish the large, high
adjacent to the rear wall; (ii) accelerate the gas flow around strength recirculation flow adjacent to the front wall.
Fig. 5. Particle concentration distribution (kg/m3)—segregation and removal of ash particles from the gas stream. (a) Original design; (b) particle extraction.
Fig. 6. Advanced secondary air injection for furnace uprating. (a) Conventional configuration; (b) new configuration.
1308 T.F. Dixon et al. / Fuel 84 (2005) 1303–1311
This flow serves a dual purpose of returning the separated validation against operating plant data (where available) is
unburnt particles to the main flame zone and supplying always undertaken by the authors as a rigorous procedure to
higher temperature gases to the entrainment flow of the ensure realism and commercial accuracy of the model
bagasse spreaders to increase the particle drying rate and predictions. This is especially the case for simulations
improve ignition stability. involving previously untried plant configurations. The
reaction between CO and O2 to form CO2 was approximated
3.5. Secondary air distribution for CO minimisation by a modified version of the eddy break up model [10]. This
modified version takes into account the kinetic limitations
Carbon monoxide (and NOx) emissions are emerging as of the CO oxidation reaction at low temperatures and the
pollutants of concern for boiler design and operations in the mixing limitations of the reaction at higher temperatures.
Australian sugar industry. Recent investigations with The modified eddy break up model helps represent the
FURNACE have been directed to quantifying the gener- behaviour of ‘parcels’ of CO that are much smaller than the
ation and removal characteristics of carbon monoxide and computational cell size used in three dimensional models of
how these can be manipulated by the use of secondary air or full scale boilers.
overfire air injection. The code modifications associated Investigations have focused on examination of the
with this work are a companion to the CMC development number, location, size, orientation, injection velocities and
mentioned earlier. The CO modeling has been focused on mass flows of the overfire air system on a boiler design. The
the mean residual levels of CO that are ejected from a assessment of the success of a particular overfire air
furnace and the distribution of CO in the upper furnace configuration is based on the visual evaluation of the flow
region in the area where urea injection for NOx control is and mixing patterns of the overfire air jets with the bulk
typically implemented. furnace gases, the manipulation of the main flame column
Incorporation of the CO dynamics in the FURNACE up through the furnace, the variations in the CO contours
code has been implemented by assigning a fraction of the within the furnace and the mean concentration of CO
off-gas produced during both devolatilization and char present in the boiler discharge flue gas. Fig. 7 shows a
burnout to be CO. The ‘CO fraction’ factor has been typical result for a CO investigation. The ‘CO fraction’
selected by comparing the predicted value of the mean CO factor used in these simulations was 0.1. The flow velocity
concentration at the boiler convection bank exit plane with contours for a vertical cross-section through the upper
operating plant measurements. Typical CO fraction values furnace are shown as well as the companion CO
of 0.1 have been used in the simulations. Ideally the concentration contours for the whole furnace. The four
segment of the code incorporating the ‘CO factor’ should be rows of overfire air nozzles of varying size and injection
developed separately based on well controlled laboratory velocity are evident (two rows each on the front and rear
scale experiments. However, the authors are not aware of furnace walls). The CO contours show high levels of CO in
any published data on mean CO concentrations in the off- the devolatilization and ignition region in the lower section
gas from bagasse devolatilization and char burnout of the furnace (O12,000 ppm) consistent with in-furnace
experiments of any type on which to base this validation. flame zone measurements. The CO is oxidized rapidly
In the absence of suitable laboratory scale data, full scale within the main flame zone and a residual column of low
plant measurements are the only means available to concentration CO flows up the centre of the furnace where
calibrate the model. It is emphasised that global code the flame column is positioned. The overfire air jets act in
Fig. 7. Predicted contours for a particular overfire air configuration on a bagasse fired boiler. (a) Upper furnace velocity (m/s); (b) furnace carbon monoxide
(ppm).
T.F. Dixon et al. / Fuel 84 (2005) 1303–1311 1309
Fig. 9. Schematic of advanced combustion technologies for bagasse firing. (a) SRI swirl burner; (b) advanced swirl spreader.
1310 T.F. Dixon et al. / Fuel 84 (2005) 1303–1311
Fig. 10. Predicted furnace temperature characteristics for two settings of the swirl spreaders. (a) Zero swirl level; (b) swirl numberw1.0.
provide an interesting challenge when the CFD modelling is the device itself but also the interactive effects that derive
undertaken. from spreader integration with the host furnace. The
modeling also forms an integral stage of the engineering
3.7. Advanced swirl spreader technology design of the advanced boiler configuration of which the
swirl spreaders are a significant component.
Development is progressing at SRI for an advanced swirl
spreader device for enhanced bagasse combustion. The
swirl spreader (Fig. 9(b)) is designed to replace the current 4. Conclusions
technology linear pneumatic fuel spreader. The enhanced
mixing and lateral spreading of the fuel generated by the Major advances have been made in the development of
swirl action provides opportunities for boiler design bagasse combustion technologies and understanding of the
refinements that could result in significant capital cost many processes involved in bagasse combustion and steam
reductions for boiler plant. Two of the significant boiler generation over 20 years. CFD modeling has come to form an
design factors associated with the swirl spreaders (compared integral and critical part of this progression. The experience
to conventional technology) are a 50% reduction in the with CFD in the sugar industry through SRI has encompassed
number of spreader units required for a given firing the full range of applications from fundamental code
capacity, and increased drying and burning rates leading development, through the generation and commercialization
to increased bagasse firing rates within a given furnace of new ideas and technologies, to the resolution of practical
volume. The swirl spreader technology has been developed plant problems. The paper has summarized many of these
to the stage of large scale application on an operating boiler. applications where SRI has achieved successful results. It has
CFD has been used to optimize the internal aero- been illustrative for SRI that the full benefits of CFD in the
dynamics of the spreader design and to resolve erosion delivery of commercial outcomes, be it in new technologies
problems due to unwanted particle recirculation in a critical or the solution of operating plant problems, have emerged
area. It is also being used to examine the integration of the through the close interaction between the code development
swirl spreader technology with the boiler furnace. A process and the validation via full scale plant simulation. It is
of simulation matching of the observed flame characteristics fundamental to the applicability of CFD that regular and
of the swirl spreaders has been completed for measurements routine validation with full scale plant data is undertaken.
on the test boiler at Proserpine mill. Exploration of the The two-way interaction between development and appli-
design consequences of swirl spreader integration in a cation enhances the code fundamentals by focusing on
furnace is in progress. Fig. 10 shows the modeling of four practical issues and the need to resolve these effectively. In
swirl spreaders in a large furnace (330,000 kg/h steam some cases the physical size of the problem areas that are
output), demonstrating the sensitivity of the flame pattern to being investigated, within a commercial plant environment,
changes in the swirl intensity. The swirl intensity for the two can be very small. Similarly the confidence in the capabilities
predictions is (a) zero and (b) swirl numberw1.0. Increasing and ‘truth’ of the CFD predictions at the commercial scale is
the swirl level draws the flame column progressively closer enhanced, to the extent that firm engineering decisions are
to the front wall of the furnace. made based solely on the characteristics predicted by the
CFD is a critical element in the full development code. SRI now regularly applies CFD modeling to a
sequence for the swirl spreader technology. It is being wide variety of flow and combustion related areas as
applied to quantify the full operating envelope of the routine a routine investigation by and design tool.
spreaders that includes not only parameters specific to This notwithstanding, it has been surprising to SRI that
T.F. Dixon et al. / Fuel 84 (2005) 1303–1311 1311
there remains a continuing suspicion of the capabilities and [4] Mann AP, Pennisi SN, Dixon TF, Novozhilov V, Kirkpatrick M,
value of CFD modeling to plant design and process Kent JH, et al. Modeling of boiler tube erosion. Proc Aust Soc Sugar
Cane Technol 2001;23:369–75.
understanding in the sugar industry by both plant operators
[5] Woodfield PL. Combustion instability in bagasse-fired furnaces. PhD
and equipment vendors. Thesis. Australia: University of Sydney; 2001.
[6] Mann AP, Plaza F, Dixon TF. Improving the prediction of
convection bank heat transfer. Proc Aust Soc Sugar Cane Technol
Acknowledgements 2003;25.
[7] Woodfield PL, Kent JH, Dixon TF. Computational modeling of
combustion instability in bagasse fired furnaces. Mediterranean
Funding for the advanced swirl spreader and secondary combustion symposium, Antalya, Turkey, 813 June 1999.
air technologies was provided in part by the Australian [8] Novozhilov V, Kirkpatrick M, Kent JH, Sazonov V, Zhang L,
Greenhouse Office (through RECP) and the Sugar Research Mann A, et al. Computational fluid dynamics modeling of tube
and Development Corporation. Thermal Energy Systems erosion rates in bagasse fired boilers. J Inst Energy 2001;74:57.
are acknowledged for permission to publish the CO [9] Zhukauskas AA, Ulinskas R. Heat transfer in tube banks in cross-flow.
New York: Springer; 1988.
modeling predictions.
[10] Magnussen BF, Hjertager BW. On mathematical modelling of
turbulent combustion with special emphasis on soot formation and
combustion. 16th Symp (Int) Combust 1976;16:719–29.
References [11] Plaza F, Dixon TF, Dickenson NL, Fitzmaurice AL, Owens M.
Performance of baffled boilers with redesigned convection banks.
[1] Boyd RK, Kent JH. Three-dimensional furnace computer modelling, Proc Aust Soc Sugar Cane Technol 1999;21:432–7.
21st Symposium (International) on Combustion. Combust Inst 1986; [12] Dixon TF, Martel H. Commercial applications of the SRI swirl burner
21:265–74. combustion system. Proc Aust Soc Sugar Cane Technol 1997;19:
[2] Luo MC. Combustion of bagasse in a sugar mill boiler. PhD Thesis. 456–65.
Australia: University of Quensland; 1993. [13] Dixon TF, Plaza F, Mann AP. Prevention of airheater corrosion. Proc
[3] Woodfield PL, Kent JH, Dixon TF. Computational modeling of a Aust Soc Sugar Cane Technol 2000;22:491–7.
bagasse-fired furnace—effects of fuel moisture. Proc Aust Soc Sugar [14] Gosman AD, Ioamides J. Aspects of computer simulation of liquid-
Cane Technol 1998;20:458–64. fueled combusters. AIAA, 1981, Paper No. 81-0234.