OCR Text |
Show 3 C.o1l11)ustion Engineering Furnace The S("('ond furnace modeled is the Large Scale Furnace Aerodynamics Test Facility (LSFATF) operated by Combustion Engineering in \\ indsor, Connecticut. The facility is a 0.46 scale model of C-E's Boiler Simulation Facility used in combustion tests. The furnace is corner-fired and contains a Dose as well as three rectangular flow obstructions hanging from the furnace top to model the superheater and rebeater platens and panels. Details of the geometry can be seen in Figures 2 and 3. Ambient air was blown into this furnace at an equal velocity in both the fuel and air ports of all four corner burners. Additional information concerning the furnace and operating conditions can be found elsewhere [9]. \\~ben coarse grid structures are elnployed in complex flow simulations: important features of the flow field can be lost. ~1an) of the recently developed three-dimension al gas dynamics and combustion models have only been demonstrated with coarse grids [2,3,4] due to computational restraints. In complex flows, such as encountered in the CoDsol furnace as seen in Figure 1 C, a fine grid structure is essential to correctly predict velocities. Corner-fired furnaces t) pically have less complex flow fields because the burners interact to form one large vortex. This is contrast.ed to wall-fired furnaces where the interaction between closely spaced s\\ irled burners can create numerous vortices. In order to study to effect of grid resolution on flo\\ field predictions, the LSFATF facility was modeled with grid structures consisting of 12 ,600 and 200,000 computational nodes. Velocity predictions for the two cases can be found in Figures 2 and 3. The coarse grid case predicted larger \iertical components of velocity throughout most of the furnace. The 200 ,000 node case revealed a larger downward recirculation zone at the top of the furnace , as seen in Figure 2B. Figure 3A shows that the coarse grid simulation failed to predict much of the swirling motion in the pendant/panel region. Both the fine grid case and the experimental data show several vortices in this area . Experimental velocity data was obtained for the LSFATF facility \ ia a calibrated five-hole pitot tube coupled to a computer controll ed traversing data acquisition system. Approximately 100 data points were obt.ained on each of seven different test planes. The filled arrowheads in Figures 2 and 3 signify experimental velocity vectors. In Figure 2A, the predictions generally appear to have larger con1ponents of yertical velocity than the dat.a. In the fine grid simulation, Figure 2B ~ the agreement i significantly better. The most significant improvement with the fine grid was in the region around the pendants/panels. The experin1ental velocities support the predictions of swirling eddies found only in the 200,000 node sin1ulation. This demonstrates that grid resolution is necessary for accurately modeling even the sin1pler three-dimensional flows found in furnoces. 5 |