Ind. Eng. Chem.
xv,fraction conversion of volatile N to NO xd,fractional conversion of volatile N to NO in additive case yo weight fraction nitrogen in char yt, weight fraction nitrogen in original fuel yv, weight fraction nitrogen in volatiles y;, weight fraction nitrogen in volatiles in additive case &, empirical char nitrogen conversion coefficient p,, empirical volatile nitrogen conversion coefficient $, dimensionless emission coefficient Literature Cited
Process Des. Dev., Vol. 18, No. 1, 1979
67
Merryman, E. L., Levy, A., "Nitrogen Oxide Formation in Flames: The Roles of NO, and Fuel Nitrogen", Fifteenth Symposium (International)on Combustion, The Combustion Institute, Pittsburgh, Pa., 1975. Pershing, D. W., Ph.D. Thesis, University of Arizona, 1976. Pershing, D. W., Wendt, J. 0. L., "Pulverized Coal Combustion: The Influence of Flame Temperature and Coal Composition on Thermal and Fuel NO, , Sixteenth Symposium (International)on Combustion, The Combustion Institute, Pittsburgh, Pa., 1977. Pohi, J. H., Saroflm, A. F., "Devolatization and Oxidation of Coal Nitrogen", Sixteenth Symposium (International)on Combustion, The Combustion Institute, Pittsburgh, Pa., 1977. Pereira, F. J., Be&, J. M., Gibbs, B., Hedley, A . B., "NO, Emissions from Fluidized-Bed Coal Combustors", Fifteenth Symposium (International) on Combustion", The Combustion Institute, Pittsburgh, Pa.. 1975. Sarofim, A. F., Williams, G. C., Modell, M., Slater, S.M., AIChE Symp. Ser. No. 748, 71 (1975). Sawyer, R. G., "Fuel Nkogen Studies", paper presented at the EPA Fundamental Combustion Research Contractors Meeting, Menlo Park, Calif., 1975. Sternling, C. V., Wendt, J. 0. L., "Kinetic Mechanisms Governing the Fate of Chemically Bound Sulfur and Nitrogen in Combustion", NTIS Report No. PB-230895, US. Department of Commerce, Springfield, Va., 1972. Turner, D. W., Andrews, R . L., Siegmund, C. W., AIChESymp. Ser. No. 726, 68 (1972). Wendt, J. 0. L., Pershing, D. W., Combust. Sci. Tech., 16, 111 (1977). Wendt, J. 0. L., Schulze, 0. E.,AIChE J., 22, 102 (1976).
Axworthy, A. E., "Flat Flame Burner Studies with HCN, NH3and NO Addition", paper presented at the EPA Fundamental Combustion Research Contractors Meeting, Menlo Park, Calif., 1975. Badzloch, S., Hawksley, P. G. W., Ind. Eng. Chem. ProcessDes. Dev., 9, 521 (1970). Bbir, D. W., Exxon Research and Engineering, Linden, N.J., private communication. July 1976. Blair, D. W., Wendt, J. 0. L., Bartok, W., "Evolution of Nitrogen and other Species during Controlled Pyrolysis of Coal", Sixteenth Symposium (International)on Combustion, The Combustion Institute, Pittsburgh, Pa., 1977. DeSoete, G. G., "Overall Reaction Rates of NO and N, Formation from Fuel Nitrogen", Fifteenth Symposium (International)on Combustion, The Combustion Institute, Pittsburgh, Pa., 1975. Fenimore, C. P.,Combust. Flame, 19, 289 (1972). Martin, G. B.,E. E. Berkau, AIChE Symp. Ser. No. 126,68 (1972).
Receiued f o r reuierc! April 18, 1977 Accepted August 11, 1978
An Analytical Study of Factors Affecting Gas Flow in Sintering Eric Rose" and Ian R. Dash Department of Control Engineering, University o f Sheffield, Sheffield S 73JD, England
A model of a sinter test pot based mainly on fundamental relationships is extended to include equations which facilitate the calculation of gas flow through the bed as the process progresses from ignition to burnthrough. The model is validated by comparing the results of simulation investigations with flow data from pilot plant trials. The dependency of flow on the viscosity and density of the gas, voidage of the mix, and effective particle diameter is illustrated. An understanding of the process is enhanced by a chart, obtained from the simulation studies, showing pressure contours and the progression of the heat wave as sinter is formed.
Introduction In practice, the investigation of the sintering characteristics of iron ore mixes and the effects of changing certain plant variables may be conveniently carried out by using a sinter test pot, the contents of which effectively represent a section of a sinter bed as formed in a Dwight-Lloyd type of sinter strand process. The raw material, usually in the form of moistened pellets of crushed ore and coke, is ignited from above and allowed to burn through under the influence of suction from an exhaust fan. The progress of the heat wave is monitored by thermocouples and when the process is complete, the sinter is cooled and subjected to various tests to assess its quality, particularly its strength. The test-pot process may therefore be regarded as an input-output process which yields vast quantities of data. Although the data are undoubtedly useful in providing guidance on the conditions required for the obtaining of sinter of acceptable quality from particular mixes, they provide a very limited picture of the detailed mechanism of the process. A thorough analytical study of the sintering process is currently being undertaken by the authors in the University of Sheffield and a detailed mathematical treatment of the heating, ignition, fusion, and cooling phases has been reported elsewhere (Rose and Dash, 1976). The results of this work show that heat wave propagation is particularly 0019-7882/79/1118-0067$01.00/0
dependent on the flow of gas through the bed. The aim of the present paper is to discuss how Szekely and Carr's equation, relating gas flow through a packed bed to the suction across the bed may be used in the simulation model to predict the gas flow history from ignition to burnthrough. The simulation results will be shown to agree well with flow data obtained from test-pot investigations and the way in which various factors such as gas density, gas viscosity, and the agglomeration of the mix affect gas flow will be fully discussed. Model of the Sinter Test-Pot Energy and mass balance equations representing the sintering process have been derived and the results of simulation studies have been reported elsewhere (Dash and Rose, 1976; Dash et al., 1974). However, in order to investigate the gas-flow problem, which is the main purpose here, it is necessary to first set the scene by discussing the essentials of the previous work. The method used in simulating the sintering process is to assume that the bed of material being processed can be divided into a large number ( n )of horizontal zones of small but finite thickness. Because of the large difference in the values of the density of the gas and the solid, it is possible to solve for gas temperature in space and solid temperature in time. Phases of the process included in the simulation model are heating, combustion of the coke particles, fusion, C
1978 American Chemical Society
68
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 1, 1979
I
Air flaw (initially ignition gas1
I
i
I I
area
Ill
'
,
I
_l___i
--lo
fan
Figure 1. Test pot showing process phases T-
7 . 7 -
{
1000-
0 182m
000-
,
-&
.U
f
h00-
f
400-
oGk%---l:O
I
4-
1bG
U A J 300 360 420
240
Tlme
I
Figure 2. Heat-wave leading-edge profiles.
limestone reduction, drying, and cooling (Figure 1). The input variables are diameter of raw mix particles, voidage, coke size, water content, limestone content, and ignition time and temperature. A further, and most important, input variable is a history of the exit gas velocity deduced from practical test-pot data. The governing equations and assumptions made in formulating the model are given in the Appendix. The heat wave leading-edge profiles obtained from the simulation and from test-pot data are compared in Figure 2. Each pair of curves represents the variation in temperature with time a t a particular level in the bed. The theoretical and practical results are in good general agreement. The results shown are for a hematite mix with a 4.5% coke content. The form in which the profiles are presented in Figure 2 tends to emphasize quite small discrepancies in the position at which the temperature is measured. A discrepancy of 0.01 m between the nominal depth and the true depth of the sensing element inside the material could bring the two sets of curves much closer together (see Figure 2). Bearing in mind the finite diameter of the thermocouple probes (0.006 m) and the forces on the probes due to the reacting material, a 1 cm discrepancy is quite feasible. By comparison, the effect on the curves of the thermocouple time constants (about 4 s) is negligible. Having established a model largely based on fundamental energy and mass balance equations which behaves in a manner corresponding to that of a practical test pot, the model has then been used to investigate the effect on burnthrough time and (indirectly) sinter quality of varying
T -e
to
bu-nthraugn
I
Figure 3. Effects of variations in input variables on burnthrough time.
the input quantities. Sinter quality may be assessed from the degree of fusion occurring in the process. Burnthrough time is an important variable in determining the production rate obtainable from a moving strand process. Important variables affecting burnthrough time are ore density, the initial mean diameter of the coke particles, voidage, and the velocity of the exit gas. The effects of varying each of these quantities are shown in Figure 3 for a sinter test pot with a typical bed depth of 0.25 m. It is normal practice to sinter a blend of iron ores. Provided that the ores are of high quality and properly blended, the input ore mixture has a nearly constant density. The initial size of the coke particles affects not only the burnthrough time but also the quality of sinter produced; for this reason it is usual to use coke sized between 3 and 5 mm. The value of voidage for the bed is dependent on the applied suction and a size analysis of the pellets forming the mix. In practice the voidage fluctuates by only 1 or 2% unless the bed collapses or segregation occiirs. Thus, the only variable affecting burnthrough which may be changed is the gas velocity through the bed and it is sorriewhat unsatisfactory to simply take test-pot data for this variable as input to the model. If, as an overall aim, methods of achieving high strand productivity are to be investigated, it is important that the factors affecting the gas velocity from ignition to burnthrough should be thoroughly understood and that this part of the process should be accurately modeled. The Calculation of Gas Velocity Using Szekely and Carr's Equation Of the many equations (Carman, 1937; Ergun, 1952) relating pressure drop across a packed bed to the flow of gas through the bed, Szekely and Carr's equation (Szekely and Carr, 1968) for nonisothermal flow appears to be the most appropriate
In 2 ui
where
+ 5 (AGp i- BG2) dz = 1
1
U
(1)
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 1, 1979 69
A =
150(1 - t)‘
;B=
1.75(1 - t )
(@dp)2t3
;G =
(@dp)c2
up, 99t
i signifies integration from the upstream end of the bed. Provided that the properties of the solid (effective particle diameter and voidage) and the properties of the gas from inlet to exit (density, viscosity, and temperature) are known, it is possible to use eq 1 to find the gas flow history from ignition to burnthrough. For isothermal flow of an incompressible fluid, Szekely and Carr’s equation reverts to a simpler form usually known as Ergun’s equation
- ApU h
+ Rp,l/PZ
dl
d2
dll
1 t2
In
“O U,
+ G 1F v A p k + G 21F CBAZ= -51 d,
(4)
also n
n
12 = 1
Simulation I
05
100
200
3W
I
1
I
100
500
6W
h;
I
Figure 4. Exit gas velocity showing comparison of simulation results with test-pot data. 110
-
100
p--x
/%msity
o t g a r constant
1
(3)
where f n is the fraction of particles between two sieve sizes with a mean size dn. When suction is applied across a sinter bed, the bed compacts causing changes in the voidage and a lowering of bed permeability. This is why, in practice, it proves uneconomic to sinter a t suctions of over 1.52 m of water (60 in.). The shape factor @ of the input mix may be calculated from Ergun’s equation by measuring the gas flow a t a low suction and assuming that the voidage remains unchanged. Provided that the input mix is satisfactorily balled the values of a and d, do not change with increasing suction; therefore a second measurement of the gas flow a t sintering suction gives the compacted voidage of the raw mix. The measurements of @, d,, and t for different mixes of the same ore (containing different amounts of return-fines and coke) have shown that the values of @ and t are nearly constant, indicating that changes in the initial bed permeability are probably due to changes in the size of the input mix. Due to the differences in the gas temperature and the gas constituents, u and p are different at different bed depths. The integrals have to be calculated numerically. Because the simulation is dependent on dividing the bed into n layers, Szekely and Carr’s equation is approximated by
5
Test-pot
Tlme from lgn
