OCR Text |
Show In this paper we examine three sepa.rate pilot sC.(1le furnaces . Evaluation is significantly enhanc.ed when local data are available for the furnace, and thus the selection of these three furnaces . To decouple the various subprocesses evaluation is presented first for non-reacting flow conditions, then on a ful1 coal-fired system. The foUowing discussion emphasizes results of the evaluation. Additional information concerning the formulation and capabilities of the model can be found elsewhere [5,6,7]. 2 Consol Furnace The first data set was obtained from a pilot scale furnace operated by Consolidation Coal Corporation in Library, Pennsylvania. The furnace is approximately ~o ~cale of a wall-fired full utilit) boiler and is fed b) four swirled burners. Although the burners were designed for the firing of pulverized coal, experimental data were collected for air flowing in both the primary and secondary inlets. The burners are located on a single wall in a diamond configuration and are all swirled in the same direction. The outlet is located above the burners on the same wall. Further details about the furnace configuration can be seen in Figure 1, which depicts velocity vectors in a vertical plane passing through the outlet and the upper and lower burners. The furnace geometry is similar in design to large industrial boilers and contains an ash bin, furnace nose, and several clipped corners. This geometry was modeled with a Cartesian grid that contained 35 points in the depth direction , 45 points in the width direction, and 65 points in the height direction. Converged results for this 102,375 node case were obtained using three different turbulence submodel options. Each turbulence model produced significantly different velocity field s. Turbulence closure su bmodels are needed to predict time average properties in the turbulent two-phase coal-fired furnace. Although direct simulations of turbulence are being performed for simple flows, these alternatives are not currently available for industrial furnaces. In this paper we test three turbulence models based on the gradient diffusion approximation . The simpliest turbulence option is to assume a constant value for eddy diffusivity (Pt) throughout the flow field . The second turbulence option employs the Prandtl s mixing length model, given by Equation 1 v-. hich relates the eddy diff usiyity to the mean velocity gradient and a rrllxing length , 1m. The third turbulence submodel option is the the st.andard k-{ model. The k-{ turbulence model provides a means of modeling the transport of k, the turbulent kinetic energy, and {, the dissipation of turbulent kinetic energy. Transport equations are de\'ised for both k and { that include tern1S to model convection, diffusion, production , and dissipation . After calculating nodal values for k and {, the local eddy diffusivity is calculated from the Prandtl-Koln10gorov relationship, Equation 2. 2 |