Energy & Fuels 1993, 7, 774-781
774
Simulation of Ash Deposit Growth in a Pulverized Coal-Fired Pilot Scale Reactor Galen H. Richards, Peter N. Slater, and John N. Harb' Department of Chemical Engineering, Advanced Combustion Engineering Research Center, Brigham Young University, Provo, Utah 84602 Received April 8, 1993. Revised Manuscript Received August 31, 199P
A model has been developed to relate the deposition behavior of ash under slagging conditions to boiler operating conditions and coal composition data. This model has been incorporated into a comprehensive combustion code and used to investigate the effects of ash deposition rate, thermal conditions, and ash chemistry on slag growth in a pilot-scale combustor. Results for simulated deposits from a coal blend fired at 3.7 MBtu/h showed a relatively high liquid fraction corresponding to denser and presumably stronger deposits. The same coal blend fired at a lower rate produced deposits which were less dense because of the lower temperatures and heat flux levels in the combustor, as well as the lower ash deposition rates. Deposition from a cleaned version of the same blend was also simulated at 3.7 MBtu/h and showed less potential for liquid-phase formation than the uncleaned blend. These results are in qualitative agreement with experimental results and illustrate the importance of operating conditions on deposit formation.
Introduction Ash deposition on heat-transfer surfaces is one of the major factors limiting the efficient use of pulverized coal to generate electricity. Inorganic material from the coal deposits on heat transfer surfaces and leads to reduced heat transfer and corrosionof boiler tubes which may result in reduced generating capacity and unscheduled outages. The behavior of the inorganic material upon combustion depends upon the amount, the composition, and the mode of occurrence of the inorganic matter in the individual coal particles, as well as the boiler design and operating conditions. Since all coals have varying amounts and types of inorganic constituents, the deposition behavior of different coals will be different. Typically, pulverized coal-fired boilers are designed to burn a particular class of coals. The size of the combustion chamber of a boiler reflects the behavior of the mineral matter in the coal upon combustion. The combustion chamber of a boiler burning a low-rank western coal can be twice as large by volume as the chamber of a boiler of the same operating capacity burning a high-rank eastern coal,' even though chars generated from low-rank coals are generally more reactive. Utilities which are currently burning high-sulfur eastern coals are changing coals, cleaning their coals, or blending their coals with low-sulfur western coals in order to meet emissions standards. It is difficult to predict what types of ash deposition problems may occur when burning a new or modified fuel. In fact, it is sometimes necessary to derate the boiler in order to manage ash deposits in day-to-day operation and prevent catastrophic deposition from occurring. A method for predicting the behavior of the coal inorganic matter as a function of operating conditions would be extremely valuable. In particular, we would like to be able to (1) predict the potential for catastrophic failure leading to unplanned outages, and (2) use the predictive model to e Abstract
published in Advance ACS Abstracts, October 15, 1993. (1) Borio, R.W.; Levaaseur, A. A. In Mineral Matter and Ash in Coal, Vorres, K. S.;Ed.; American Chemical Society, Washington, DC, 1986.
0887-0624/93/2507-0774$04.00/0
help develop a viable, if not optimum, strategy for ash management. This paper describes our initial efforts to incorporate a deposition model into a comprehensive computer code developed to simulate turbulent combustion. The combustion code is used to predict the flow field, temperature field, radiation field, and particle transport in an axisymmetric combustor for different coals at different firing rates. The deposition model is then used to examine the effect of different fuels and operating conditions on deposit growth, thermal properties, porosity, and composition. Calculations are applicable to slagging deposits formed in the radiant section of a boiler.
Mathematical Model Combustion Code. This study used a modified version of PCGC-2, a comprehensive combustion code developed at Brigham Young University by Fletcher, Smith, and Smoot2s3 to simulate combustion and gasification of pulverized coal. The typical geometry modeled by this code is two-dimensional and axisymmetric. In PCGC-2, the gas-phase flow field is calculated with use of steadystate finite difference techniques4s5and the k-t turbulence model6 for closure. Gas-phase turbulence is adjusted for the presence of particles, and the effect of turbulence on particle motion is modeled semiempiri~ally.~.~ (2) Fletcher, T.H. Ph.D. Dissertation, Chemical Engineering Department, Brigham Young University, Provo, UT 84602,1983. (3) Smith, P. J.; Fletcher,T. H.; Smoot, L. D. Eighteenth Symposium (International) on Combustion;The Combustion Institute: Pittsburgh, 1980; p 1285.
(4)Gceman,A.D.;Pun,W.M.;Ruchal,A.K.;Spalding,D.B.;Wolfstein,
R. Heat and Mass Transfer in Recirculating Flows; Academic Press: London, 1969. (5) Patankar, S. V. Numerical Heat Transfer and Fluid Flow; Hemisphere Publishing Corp.: New York, 1980. (6) Launder, B. F.; Spalding, D. B. Mathematical Models of Turbulence; Academic Press, London, 1972. (7) Fletcher, T.H.;M.S. Thesis, Chemical Engineering Department, Brigham Young University, Provo, UT 84602, 1980. (8) Melville, E. K.; Bray, N. C. Int. J. Heat Mass Transfer 1979,22, 647.
0 1993 American Chemical Society
Ash Deposit Growth in a Coal-Fired Reactor
Gas-phase chemistry is treated assuming that reactions are micromixing-limited, with transport equations governing the mixture fractions of inlet gas and coal "offgas", as well as their mean square fluctuation^.^ The coal "off-gas" includes volatiles liberated during the rapid devolatilization process, as well as the subsequent gaseous products of char oxidation. Instantaneous gas properties, such as species concentrations, are calculated from local elemental composition as a function of the mixture fractions. Time-mean gas properties are calculated by convoluting the instantaneous properties over a clippedGaussian probability density function (PDF) based on the turbulence statistics of the mixture fractions. Particle mechanics are solved along Lagrangian trajectories using the PSI-CELL technique.1° Devolatilization is modeled with a simple two-step kinetic scheme,l1J2with kinetic coefficients and yields dependent on the particular coal type. Char oxidation rates are based on the external surface area, with coal-dependent rate coefficienh. Typically, a limited number of particle trajectories (e.g., 150) are sufficient to approximate the combustion behavior. The combustion code also includes a description of energy transport for both the particle and gas phases, as well as coupling between the two phases. The radiation field is solved with a discrete ordinates method (5'4 quadrature) which includes anisotropic multiple scattering.'3J4 Particle scattering and absorption are coupled to the Lagrangian particle trajectories. Absorption and scattering efficiencies change with particle burnout and reflect the transition from coal to ash. Particle Impaction Rates. Although a limited number of particle trajectories were adequate for the combustion simulations, more information was required to approximate impaction rates at the wall. Consequently, a stochastic separated flow (SSF) modeP was added as a postprocessor to calculate particle impaction rates. In this model, the influence of turbulent velocity fluctuations in the gas phase on the particle trajectories is accounted for through random particle-eddy interactions. Velocity fluctuations are assumed to be isotropic and are obtained by sampling a Gaussian PDF with a standard deviation of (2k/3).0.6 The particle motion within an eddy is modeled deterministically by solving the instantaneous particle momentum equation. Velocity fluctuations are assumed to remain constant during a particle-eddy interaction. SSF calculations were performed for reacting particles similar to those used in the combustion simulation. The carbon burnout was tracked along the particle trajectory. It was assumed that inorganic matter in each coal particle agglomerated to form a single ash particle.lsJ7 Note that (9)Spalding, D. B. Chem. Eng. Sci. 1971,26,95. (10)Crowe, C. T.;Sharma, M. P.; Stock, D. E. J. Fluids Eng. Trans. ASME 1977,325. (11)Kobayashi,H.; Howard, J.B.;Sarofii, A. F.SixteenthSymposium (International)on Combustion:The Combustion Institute: Pittsburgh, - . i978; p 411. . (12)Ubhayakar, S. K.; Stickler, D. B.; von Rosenberg, C. W.; Gannon, R. E. Sixteenth Symposium (International) on Combustion; The Combustion Institute: Pittsburgh, 1976;p 427. (13)Fiveland, W. A. ASME J. Heat Transfer 1984,106,699. (14)Jamaluddin, A. S.;Smith, P. J. Combust. Sci. Technol. 1988,59, 321. (15)Shuen, J-S.; Solomon, A. S. P.; Qhang, Q-F.; Faeth, G. M. AZAA J. 1985,23,396-404. (16)Wilemski, G.;Srinivasachar, S.; Sarofii, A. F. In Inorganic Transformations and Ash Deposition During Combustion; Benson, S. A,, Ed.; Engineering Foundation Prese, ASME: New York, 1992;pp 545564. (17)Baxter, L. L.;Hardesty, D. R. In Coal Combustion Science: Quarterly Progress Report; Sandia Report, SAND91-8233,1991.
Energy & Fuels, Vol. 7, No. 6, 1993 775
this assumption does not account for submicron particles which may form from a portion of the organically bound inorganics in low-rank coals. Because of the stochastic nature of the SSF calculations, it was necessary to simulate a large number of particles in order to obtain an acceptable representation of the particle impaction rates. Typical SSF simulations included 15 000 particle trajectories. Impaction rates were determined as a function of particle size and location. The temperature and velocity of impacting particles were also available from the SSF simulations. The compositionof ash particles was assumed equal to the composition of fly ash from a pilot-scale reactor measured by SEMPC.18 Particle Sticking. Amodel was also needed to predict which of the impacting particles would stick based on particle temperature and composition. The capture efficiency (ocap)or fraction of the impacting particles which adhere to the surface was approximated by the following expression:19 N
where pi(Tp,) is the sticking probability of particles of composition i, TPis the particle temperature on impaction, p8(T8)is the sticking probability of the deposit surface, and T,is the temperature of the deposit surface. Factors which are expected to influence sticking probability include particle. viscosity, surface tension, velocity, and angle of impact, as well as the chemical and physical state of the surface upon which deposition is occurring. Clearly, a complex relationship would be needed to incorporate the influence of all of these factors. However, as a first approximation, particle viscosity has been found to provide a reasonable measure of the particle properties which affect sticking behavior.19 The sticking probability of a particle of composition i was therefore defined as c1 'ret
PiCTpJ 1 P 5 (2) The viscosity is a function of both temperature and composition which enables it to account for changes in sticking behavior owing to variations in both the composition and the temperature of the incoming particles. The modified Urbain method was used to approximate particle viscosities in the present study.20 Deposit Properties. In order to model deposit growth and heat transfer, it was necessary to describe the thermal and physical properties of the deposit. In the present work, a correlation developed by Sugawara and Yoshizawa21was used to describe the thermal conductivity of the deposit as a function of the deposit porosity and the thermal conductivities of the gas and solid phases as follows:
-(
F = 2" 2"-1
1-
-) 1 +
(1 4)"
(18)Thomock, D. E.; Borion, R. W. 'Developing a Coal Quality Expert: Combustionand Fireside Performance Characterization Factors". Report prepared for CQ, Inc. and US. Department of Energy, ABBCombustion Engineering, 1992. (19)Walsh, P. M.; Sayre, A. N.; Loehden, D. 0.;Monroe, L. S.; Beer, J. M.; Sarofim, A. F. h o g . Energy Combust. Sci. 1990,16,327. (20) Kalmanovitch, D. P.; Frank, M. In Mineral Matter and Ash Deposition from Coal; Bryers, R. W., and K. S., Vorres, Eds.; Engineering Foundation Conference: Santa Barbara, CA, 1990; pp 89-101. (21)Sugarawa, A.; Yoshiwaza, Y. Aust. J.Phys. 1961,14,469.
Richards et al.
776 Energy & Fuels, Vol. 7, No. 6, 1993
deposited particles (except oxides) at deposit temperature
Test viscosity of individual particles kr?
Determine equilibrium composition of particles where p e k r from equilibrium table
v Determine total amount of liquid and solid
Calculate new deposit porosity I 1 Figure 1. Flowchart illustratingthe procedure used to estimate the porosity of the deposit.
k
(1-F)kO + Fk,
erties have recently been coded and will be described in future publications. A key property in the deposit growth simulations was the local porosity of the deposit. The porosity affects the thickness, thermal conductivity, and strength of the deposit. The initial deposit was assumed to have a porosity of 0.6, characteristic of powdery or perhaps slightly sintered dep0sits,~5 and consistent with values measured by Anderson et al.26 The local porosity of the deposit decreased as the deposit grew by the formation of liquid phases which filled the pores. The amount of liquid present in the deposit was approximated by the procedure illustrated in Figure 1. First, the viscosity of each of the particles which deposited, except for those identified as silica or other pure oxides, was calculated at the deposit surface temperature for each of the compositions from the SEMPC analysis. Note that the amount of oxide species other than silica was very small. Particles with a viscosity below a critical viscosity (per = lo5 P) were assumed to approach their equilibrium composition. A lower value of the critical viscosity would delay the approach of a particle to equilibrium until higher temperatures. However,calculations performed with pa values of 103 and lo4 P were not significantly different from the results presented in this paper for a wcr of los P. The equilibrium species and phase composition of particles a t equilibrium was taken from a table of equilibrium data calculated for the temperature and composition ranges of interest with a computer code described elsewhere.27Use of the equilibrium table greatly decreased the CPU time required for the deposit growth simulations. The total amount of liquid in the deposit was obtained by summing the amount of liquid from each of the particles at equilibrium. The amount of solid was the sum of (1) the mass of the solid fraction of the equilibrium particles, (2) the mass of particles whose viscosity was greater than per, and (3) the mass of the silica and other oxide particles. The new deposit porosity was estimated by assuming that the solid phase formed a matrix with a porosity of 0.6 and that the liquid phase filled the pores as follows:
(3)
where k, is the thermal conductivity of the gas phase, k, is the thermal conductivity of the solid, n is an empirical parameter (6.5), F is the fraction of the conductivity attributable to the gas, and 4 is the porosity of the deposit. The thermal conductivity of the gas phase was approximated as that of nitrogen at the local deposit temperature. The thermal conductivity of the solid phase was assumed to be constant at 4 W/(m K), a value representative of silica-containing materials at high temperature^.^^^^^ Determination of the deposit porosity was somewhat more complex and will be described below. Once the porosity was determined at a particular location in the deposit, eq 3 was used to determine the local thermal conductivity. An approximate correlation developed by D a ~ i e was s~~ used to calculate the emissivity of the deposit. More sophisticated algorithms to describe the radiative prop-
where 4 is the deposit porosity, $0 is the initial porosity of the deposit, V Iis the volume of the liquid, and V, is the volume of the solid. Note that, according to eq 4, zero porosity is reached before the deposit is completely liquid. As the volume of liquid approaches that of the solid, it becomes less accurate to approximate the total amount of liquid from particle-by-particle calculations. However, these calculations still provide an indication as to whether or not very low porosities are likely to be reached in a deposit. The use of equilibrium to approximate the phase and species composition of the deposit is clearly a simplification. Equilibrium does not take into account the time dependency of crystallization and liquid phase formation. Also, the equilibrium approximation may lead to sudden changes in the phase composition of the deposit as a critical
(22) Bever, M. B. Concise Encyclopedia of Advanced Ceramic Materials; Pergamon Press, New York, 1991; p 418. (23) Kingery, W. D.; Bowen, H. K.; Uhlmann, D. R. Introduction to Ceramics, 2nd ed.; John Wiley & Sons: New York, 1976. (24) Davies, P. R. M.S. Thesis, Chemical Engineering Department, Brigham Young University, Provo, UT, 84602, 1988.
(25) Wesse1,R.A.;Wagoner,C.L.PaperNo.86-JPGC-FACT-7,ASME, 1973. (26)Anderson, D. W.; Viskanta, R.; Incropera, F. P. J. Eng. Gas Turbines Power 1987,109,215-221. (27) Harb, J. N.; Munson, C. L.; Richards, G. H. Energy Fueb 1993. 7, 208-214.
Energy & Fuels, Vol. 7, No. 6, 1993 777
Ash Deposit Growth in a Coal-Fired Reactor
Table I. Experimental Runs from CQE Tests in a Pilot-Scale Combustor fuel firing rate, MBtuIh excess air, % 70% WY" 3.1 20 30% OKb 70% WY 30% OK 70% WY 30% OK cleaned
capture rate
a
Increment time
Yes
u Figure 2. Flowchart illustrating the procedure used to simulate deposit growth.
condition (e.g., temperature) is reached. These changes would be expected to occur more gradually in an actual system. However, the time required to reach equilibrium decreases as the surface temperature increases, making the equilibrium a better assumption for the outer layers of the deposit where liquid is more likely to be present. Simulation of Deposit Buildup. Figure 2 a flowchart which illustrates the procedure for simulation of deposit growth. The first step in the procedure was to calculate the properties of the deposit at the beginning of the current time step. These properties were assumed constant for each time step but were allowed to vary from time step to time step. Second, the particle deposition rate (kg/(m2 s)) was calculated as the product of the particle impaction rate and the particle capture efficiency. The particle deposition rate was then multiplied by the number of seconds in the current time step to yield the mass of the deposit per wall surface area (kgJm2). This mass was used to calculate the thickness of the additional deposit formed during the current time step by the following equation: li = m i / b , ( 1 - 491
20
3.6
20
Wyoming coal. b Oklahoma coal.
required to establish a steady-state temperature profile through the deposit. The first step in the heat transfer calculations was to determine the thermal resistance of the deposit at the current time step. In this discrete model, the thermal resistance included a contribution from the deposit formed during each time step as described by the following equation:
Iterate to determine surface temperature and heat flux
L
3.0
(5)
where Zi is the thickness, mi is the mass per area, pp is the solid particle density, & is the deposit porosity, and i refers to the current time step. Note that the effects of erosion and periodic shedding of deposits were not considered in the present study. After the deposit thickness had been determined, it was then possible to quantify heat transfer to and through the deposit. Heat was transferred to the deposit by radiation and convection. This heat was then transferred through the deposit by conduction to the cool wall. Conduction through the deposit was approximated as quasi-steady state since the characteristic time associated with buildup of the deposit is much greater than the time
where l j is the thickness of the deposit formed during time step i, ki is the thermal conductivity of the deposit formed during the same time step, and N is the current time step. Equation 6 is equivalent to summing a set of resistances in series to obtain a total resistance. Note that the resistance in eq 6 is related to the effective thermal conductivity as follows: where k,ff is the effective thermal conductivity and L is the total thickness of the deposit. Once the thermal resistance of the deposit had been calculated, it was then possible to calculate the temperature at the outside surface of the deposit. If the flux (qtot) at the outside surface of the deposit were known, the surface temperature could be calculated by the following equation: The flux, however, contains both a convective part and a radiative part, both of which are a function of the surface temperature and are therefore unknown:
Consequently, an iterative procedure was required and a predictor corrector technique was used to solve for the surface temperature and flux. Once the temperature and flux had been determined, the deposit properties were recalculated and the procedure shown in Figure 2 was repeated until the total time elapsed equaled the total desired deposition time.
Experimental Data Experimental data for model validation were supplied by ABB Combustion Engineering from work performed for the Coal Quality Expert (CQE) Program funded by the Department of Energy and EPRL1* Three different runs were chosen to illustrate the effects of firing rate and coal cleaning on the deposition process (see Table I).These runs were performed with a 70130 blend of a Wyoming and Oklahoma coal, as well as with a blend of the Wyoming coal and cleaned Oklahoma coal. Standard data on both the blend and the cleaned blend are given in Table 11. Note that the cleaned coal has a lower ASTM ash content
Richards et al.
778 Energy & Fuels, Vol. 7,No. 6,1993 Table 11. ASTM Analyses of Coals analysis
HHV,Btu/lb proximate, w t % moisture volatile matter fixed carbon ash ultimate, wt % moisture hydrogen carbon sulfur nitrogen oxygen ash ash composition, wt % Si02 A1203 Fen03 CaO MgO Nan0
KzO Ti02 P206 SOa
IO % WY/30% OK
10% WY/30% OK CLN
11332
11484
8.5 40.2 43.2 8.1
8.0 41.4 44.0 6.6
8.5 4.4 64.5 0.6 1.3 12.6 8.1
8.0 4.1 65.5 0.6 1.2 13.4 6.6
37.1 15.6 5.8 16.0 2.8 0.1 1.6 1.0 0.5 11.8
35.4 16.3 6.1 16.5 3.5 0.1 1.1 1.2
minerals gehlenite anthorite albite pyroxene calcium silicate spurrite calcium aluminate quartz iron oxide calcium oxide ankerite anhydrite pure kaolinite kaolinite derived illite (amorphous) montmorillonite unclassified
70% WY/30% OK 14.4 7.6 0.0 0.0 0.4
0.0 0.8 4.4 0.4 0.4 0.4 0.8 1.6 12.4 2.0 7.6 46.8
I
L5-PANEL5
1v Itl"\
L4-PANEL4
u - PANEL 3
-
L1 PANEL 1
BURNER
PRIMARY AIR
AN9 FUEL
SO?rOM ASH DISCHARGE
Figure 3. Fireside Performance Test Facility (FPTF) at ABB
0.8
Combustion Engineering.lB
15.1
Table 111. Radiant Section In-Flame Solids (wt SEMPC
w e
W),
70% WY/30% OK CLN 11.6 9.6 0.0 0.4 0.0 0.4 1.6 5.2 0.0 0.4 0.0 0.8 1.2 9.6 0.8 4.4 53.6
and is enriched slightly in iron over the original blend. The phase composition of the in-flame solids was also measured (by SEMPC) as part of the CQE study as shown in Table 111. The experimental data described above were taken in the Fireside Performance Test Facility (FPTF) at ABB Combustion Engineering (Figure 3). This combustor is up-fired with a single central burner. Its construction is nearly axisymmetric, except for two asymmetric legs at the bottom of the combustor, and the outlet to the convectivepass at the top of the reactor. A 3-D code would be required to accurately model these features. An approximate axisymmetric geometry was used in the present study; no attempt was made to model the nonaxisymmetric features of the reactor. It is therefore reasonable to expect differences between the simulated results and the experimental observations. However, the potential effect of input parameters which were not accurately known (e.g., swirl number) did not seem to justify additional sophistication in the geometry at this point. Deposit growth calculations were performed on the first slag panel in the wall of the combustor (see Figure 3).
I 1 1 Combustion Simulation
Particle Impaction
with Time Figure 4. Calculation sequence.
Calculations The overall sequence of calculations is shown in Figure 4. It was assumed that the combustion conditions did not
change significantly during buildup of the deposit. This assumption is obviously invalid for boilers where the principal objective is to transfer heat in order to raise steam. In contrast, deposition in pilot and laboratory scale combustors is typically studied on a small cooled panel, which tends to minimize the effect of the deposition on the operation of the combustor. All calculations were performed on an H P 750 workstation. Simulations were performed for the conditions shown in Table I, except that the clean blend was simulated at 3.7 MBtu/h instead of 3.6 MBtuJh. Combustion simulations were performed with a modified version of PCGC-2. A typical simulation required approximately 45 min to complete. The number of particle iterations required to reach convergence was approximately 15, with a maximum of 250 gas-phase iterations per particle iteration. Closure of an overall energy balance was of particular concern since simulations were compared at different operating conditions. The discrepancy between the energy entering the system (with coal, primary gas, and secondary gas) and that leaving (by radiation, convection, and with exiting gas) was typically less than 0.6%. The stochastic separated flow (SSF) model was run as a postprocessor to each combustion simulation in order to
Ash Deposit Growth in a Coal-Fired Reactor
Energy & Fuels, Vol. 7, No. 6,1993 779
-
7.5
1 -.-.-.-
3.0MBTURlR 3.7 MBNIHR CLEAN
I
Figure 5. Flow streamlines for the Wyoming/Oklahoma blend fired at 3.7 MBtu/h with 20% excess air.
1- "
~~~~~
0
200
T
I
I
400
600
800
Time (min) Figure 6. Temperature contours for the Wyoming/Oklahoma blend fired at 3.7 MBtu/h with 20% excess air.
........ A
Figure 8. Deposit thickness aa a function of time.
cplcumed
Exp.rlmental Calculated Exp.rlmental
3.7 MBtu/hr
@ a 500
0
1
2
3
600
0
4
200
approximate particle deposition rates at the wall. Each SSF run consisted of 15 000 particles and required approximately 5 h to complete. Even though only a limited number of SSF computations were performed, use of this model to estimate particle impaction rates represented the most computationally intensive portion of the calculation sequence. The deposition model was used to estimate deposit growth a t a specific location (node) in the axisymmetric combustor. Each simulation included a large number of particle compositions (250) and required approximately l / 2 h tocomplete. This time could be reduced considerably by considering a smaller number of compositions. An additional 4 h were required to generate the equilibrium tables used for the deposition calculations.
Results and Discussion Combustion Simulations. Results from combustion simulations are given in Figures 5,6, and 7. Figure 5 shows the flow streamlines for the coal blend fired at 3.7 MBtu/h with 20 % excess air. A main recirculation zone is evident in the center of the combustor due to the swirling flow. Flow recirculation is also evident in the corner of the reactor, although the shape of this recirculation zone would undoubtedly be influenced by the reactor legs as discussed above. Temperature contours for the same firing conditions are given in Figure 6. Temperatures are similar to those
IO
Time (min)
Axial Distance (m)
Figure 7. Comparison of measured and predictedheat flux along the reactor wall.
400
Figure 9. Temperatureat the surfaceof the depositas a function of time.
0.6
0.4-
0
200
400
600
800
Time (min) Figure 10. Porosity of the outside layer of the deposit aa a function of time.
expected in a full-scale unit. Figure 7 provides a comparison of the calculated heat flux along the reactor wall with values measured by Combustion Engineering a t both the high and low firing rates. These results indicate that, in spite of the approximations, the combustion simulations provide a reasonable estimate of the conditions inside the combustor under different firing conditions. Deposit Growth. Predictions of deposit thickness as a function of time are given in Figure 8. The blend fired at the 3.7 MBtu/h firing rate resulted in the thickest deposit '
Richards et al.
780 Energy & Fuels, Vol. 7,No. 6, 1993 100
80
-
h
sf:
v
-260 -m U Y
I aJ F
-40
-
C
-
0-
20
-
0 3.0 MBtu/hr
3.7 MBtu/hr
3.6 MBtuhr Cleaned
Figure 11. Experimentallymeasured heat flux recovery for deposits in the Fireside Performance Test Facility at ABB Combustion Engineering.18
due to the relatively high amount of ash in the system. Both the blend at the 3.0 MBtu/h firing rate and the cleaned blend showed lower rates of particle impaction and deposit buildup. The difference between the results for the clean blend and the 3.0 MBtu/h firing rate are greater than expected based on ash content alone, and reflect differences in the particle impaction rate due to differences in the flow field, particle size distribution, etc. Figure 9 shows the calculated surface temperature of the deposit as a function of time. The surface temperature at the 3.7 MBtu/h firing rate increases more quickly than the others owing to its higher deposition rate. The 3.0 MBtu/h firing rate flattens out at a lower temperature than the high firing rate because of lower temperatures and lower heat fluxes in the combustor (see Figure 7). The crossover of the surface temperature in the clean case relative to 3.0 MBtu/h firing rate is expected since both the heat flux levels and gas temperatures are higher for the clean case, even though the rate of deposition is substantially lower. It is significant to note that most of the temperature change occurs in the first two hours of deposition, consistent with experimental heat flux data from the Coal Quality Expert Study18. Hence, almost all of the temperature change results from a deposit 1or 2 mm thick. The change in deposit porosity as a function of time is illustrated in Figure 10. The change in the porosity occurs first for the 3.7 MBtu/h firing rate, followed by the 3.0 MBtu/h firing rate and the clean case. The most distinctive feature of this figure is the sharp drop in the porosity at the 3.7 MBtu/h firing rate at just under 600 min of deposition. The sharpness of the drop-off reflects the discrete nature of the calculations where changes occur in discrete time steps and not continuously. However, the fact that the drop-off occurred is significant and indicates the formation of substantial amounts of liquid in the deposit. This decrease in porosity may be associated with a stronger deposit for the 3.7 MBtu/h firing rate than for either of the other two cases. This conclusion is supported by experimental data from the CQE program (Figure 11)which show essentially zero heat recovery at the end of 12 h for the 3.7 MBtu/h firing rate, and substantial heat recovery for each of the other two cases. Note that the experimental data for the clean blend were
z 4 E
Local Thermal Conductivity
-...-....-Effective Thermal Conductivity
0
Time (min) Figure 12. Local and effective thermal conductivities for the 3.7 MBtu firing rate case.
actually taken a t 3.6 rather than 3.7 MBtu/h, which may have had some effect on the results, although less deposition would be expected a t the lower firing rate. Figure 12 shows both the local and effective thermal conductivityas a function of time for the 3.7 MBtu/h firing rate. Again, the distinguishing feature of this figure is the sharp change which occurs at just under 600 min of deposition. Although the sharpness of the change may be overestimated, the magnitude of the change is consistent with the results of Anderson et al.26who also observed a factor of 10 change in the local thermal conductivity. In contrast, the effective conductivity increases slowly and steadily throughout the simulation, reaching a final value of 0.45 W/(m K). This value is similar in magnitude to the effective conducitity of 0.5 W/(m K) given by Walsh et al.19 Deposition rates, thermal conditions (i.e., heat flux and temperature), and deposit chemistry are key factorsleading to differences in deposit growth and properties. To examine the relative importance of these variables, the results, discussed above were replotted in two different ways as described in the paragraphs that follow. A plot of porosity vs accumulated mass (deposit mass per wall surface area in kg/m2)is given in Figure 13. This type of plot should factor out differences due to deposition rate, while differences due to thermal conditions and
Energy & Fuels, Vol. 7, No. 6, 1993 781
Ash Deposit Growth in a Coal-Fired Reactor
Finally, the porosity at the surface of the deposit was I plotted against the deposit surface temperature (Figure
0.4
14) in order to isolate the influence of deposit chemistry. In the model, the local chemistry of the ash should be the dominant factor affecting the deposit porosity at a given temperature. Figure 14 shows that the porosities of all three ashes were similar when ploted in this fashion, indicating that the ash chemistry did not play a dominant role in the present simulations.
4
Conclusions
Figure 14. Local porosity as a function of deposit surface temperature.
A model has been developed to simulate deposit growth under slagging conditions. This model was coupled with a comprehensive combustion code to examine the effects of deposition rate, thermal conditions, and deposit chemistry on the thermal and physical properties of deposits. The model was applied to an experimental system studied at ABB Combustion Engineering under the Coal Quality Expert program funded by the Department of Energy and EPRI. Qualitative agreement was observed between predicted and observed deposition behavior. Calculations of both the local and effective thermal conductivity agreed with expected trends. The relative contributions of deposition rate, thermal conditions, and deposit chemistry to deposition were also evaluated. It was found that both the heat flux and the deposition rate had a significant influence on the thermal and physical properties of deposita formed from the same coal blend fired a t different rates. Deposits formed from the clean blend appeared to be most strongly influenced by a particle deposition rate which was significantly lower than that of the other coal blend examined. The chemistry of the three deposita was similar, although the clean blend chemistry was slightly worse. These calculations illustrate the importance of operating conditions on deposit formation, and the use of a mathematical model to predict deposit behavior.
chemistry should remain. Figure 13 shows that the clean case and the 3.7 MBtu/h firing rate are similar in behavior. Differences in thermal conditions were nbt expected to be large as these simulations were performed at similar firing rates. Apparently, the chemistry of the two deposits is also very similar. Note that the curve for the clean blend ends much sooner than that for the 3.7 MBtu/h firing rate, indicative of the smaller mass of the “clean” deposit. The difference between the 3.0 MBtu/h firing rate and the 3.7 MBtu/h firing rate is due primarily to thermal conditions since the same chemistry was used for these two ashes.
Acknowledgment. The authors thank David Thornock of Combustion Engineering and David Kalmanovitch of Riley Stoker for their helpful insight and discussion. Use of experimental data from ABB Combustion Engineering taken under the Coal Quality Expert program funded by the U.S.Department of Energy and EPRI is gratefully acknowledged. This work was sponsored by the Advanced Combustion Engineering Research Center. Funds for this Center are received from the National Science Foundation, the State of Utah, 29 industrial participants, and the U.S.Department of Energy.
Accumulated Mass (kg/m2) Figure 13. Local porosity as a function of accumulated deposit mass.
* *z L
9
0.4
P
I I
I I 500
I
I
I
I
750
1000
1250
1500
i I I
1750
Surface Temperature (K)
