OCR Text |
Show the coupling bet ween the solid and gas phases. Several ways of improving this difficult aspect using Multigrid methods have been identified, including using muhigrid methods for Eulerian particle calculations. Furt.hermore, in three dimensions, multigrid provides a greater proportional savings because a coarse grid has only 1/8 as many nodes a.s the next finer grid, compared to 1/4 for two dimensions. Local grid refinement also wil1 be an important tool for these geometries, because of the large disparity in scales found in industrial furnaces. 6 Conclusions \Ve have demonstrated the capabilities of current comprehensive coal combustion models and compared predictions with measurements taken from furnaces operating at practical, industrial conditions. The evaluation has emphasized the significance of fluid turbulence and energy losses on predictive capability. It has also emphasized the need for fine computational meshes and numerical methods able to deliver fine mesh calculations in reasonable computer run times. The simulation of the Con sol furnace illustrated the importance of turbulent transport in the closure model. Although the predictive capability of the commonly used k-f model for turbulent transport exceeded the capability of simpler closure schemes, it failed to predict all flow field details. The predictions from the Combustion Engineering facility illustrated the role of grid resolution in providing accurate flow field predictions. The fine grid st.ructure simulations revealed additional flow structure supported by the experimental dat.a which was not found in the coarse grid case. The coupling of the physico-cherrucal processes and heat transfer in coal fired furnaces produces effects that are neither in1mediately obvious nor predictable from beat transfer considerations alone. Although the observations on beat transfer made in this paper are specific to the BYU furnace operating under one set of conditions, some broad generalizations seem to be applicable. Heat loss from the gas pbase is significant and predominately by convective/conductive exchange with particles that are radiating to cooler heat transfer surfaces. The turbulent fluctuations occurring in pc furnaces result in flame structures (ie. temperatures and compositions) that are significantly different from non-fluctuating flames, and any computational method that attempts to simulate t.urbulent flames must explicitly account for fluctuations in stoichiometry and energy. The increased con1putational burden of n10re sophist icated three-dimensional turbulence n10dels, of performing fine grid structure calculations, of including the highly coupling processes of all of the various forms of energy exchange within the coal-fired furnace, and of treating the effect of local turbulence fluctuat.ions on all properties 18 |