OCR Text |
Show ( X>1 ^±__1 3 100 200 200 FIGURE 17: TYPE-1 FLAME; COMPUTED LOCAL STOICHIOMETRY (TOP) AND CONTOURS OF TOTAL-NO CONCENTRATIONS IN PPM (BOTTOM) flame measured and predicted N O concentrations leads to an observation (sec Peters and Weber, (1991)) that the computed lolal-NO values arc undcrprcdictcd in fuel rich regions and over-predicted in oxygen rich regions. This is mainly due to Eq.(28) used to calculate the yield of N O which is in fact applicable to high temperatures of about 2450 K. It is possible to replace Eq. (28) wilh another expression applicable to temperatures around 1700 K but it should be remembered that a proper way of improving this model deficiency is to use a good kinetic submodel for M N x species reduction coupled into the turbulent combustion model. The NO post-processor has distinguished the high NOx flame from the low N O x one mainly because the flame model has correctly identified the large differencies in the size of the fuel rich region of the two flames (see Figures 16 and 17.) The N O x postprocessor is bound to fail if the fuel rich zone is wrongly predicted. CONCLUSIONS The mathematical modeling technique for simulating pulverized coal flames including the N O predictions has been scrutinized. The objective is lo analyze whether the mathematical modeling can provide engineers wilh the three-level technical information (see Introduction) required for designing large scale burners and flames. The following has been concluded : The first-order information containing the knowledge of the flame shape, length, and a global estimate of flame temperatures wilh accuracy around 200° C is generally obtainable from the modeling. Here ihc modeler can use as a starting point the proximate and ultimate coal analyses and estimate the dcvolatilizalion rale, high temperature volatile yield and char combustion rate Such computations can be performed using the k-e turbulence model, but coarse numerical grids and first-order numerical schemes should be avoided. Neither predictions of char burnout nor emissions of N O x or C O are reliable The second-order information containing detailed knowledge of temperature distribution, oxygen and unburned fuel can be obtained if both the inlet conditions lo the burner arc precisely known and the coal combusted has been characterized e.g. its pyrolysis rate, the high temperature volatile yield and char combustion rate are measured prior to the modeling. The combustion of volatilcs should proceed lo carbon monoxide with the mixing rale constant prescribed a value of 0.6 rather than 4. The computations should be performed using a second-order numerical scheme and a fine numerical grid. Reliable flow-field predictions of non-swirling jct-flamcs and typc-2 swirling flames (see Figure 1) of high- or medium-volatile coals can be obtained using the k-e model of turbulence. In the case of type-2 flames, when the inlet swirling flow is combustion accelerated in the vicinity of the fuel injector, the swirl induced effects are rapidly reduced and the k-e turbulence model works satisfactory. For type- 2 flames of low volatile or anthracite coals better predictions of the flow field inside the burner quarl can be obtained if a second-order turbulence model is applied. Obtaining the second-order information by computing lype-1, low N O x flames which arc typical for internal air staged burners, is very difficult It has been demonstrated that the quality of the near burner zone predictions of type-1 flames is substantially worse than for type-2 flames. This has been attributed lo difficulties in computing correctly the penetration depth of the fuel jet inside the swirl induced reverse flow. Both full penetration and partial penetration flames can be generated in such burners. Similarly, the computations may lead to more than one mathematical solutions (see text). At this stage of the modeling the third-order information is obtainable only in very special circumstances when the modeler "extrapolates gently" from rigorously verified predictions. Firstly a confidence in the predictions must be gained by comparing them with in-flame measurements of temperature and chemical species including N O . If a good agreement is obtained, then a subsequent computer run, wilh either slightly altered burner geometry or with different inlet or boundary conditions can provide the desired third-level information. ACKNOWLEDGEMENT The authors gratefully acknowledge the financial support of the IEA Coal Combustion Science Program, Annex 2, sponsored by organizations within Canada, Germany, the Netherlands and the United Kingdom. W e thank the Joint Committee of the IFRF for permission to publish this paper. REFERENCES Abbas. T.t Costen, P., and Lockwood, F.C., 1992, "The Influence of Near Burner Region Aerodynamics on the Formation and Emission of Nitrogen Oxides in a Pulverized Coal- Fired Furnace", Comb, and Flame, pp. 346-363. Adams. B.R., and Smith, P.J., 1992, "Thrcc-Dimcnsional Discrcte-Ordinates Modeling of Radiative Transfer in a Geometrically Complex Furnace", Combustion Science and Technology, 83, pp. 1-15. 14 1 1 - 11 |