OCR Text |
Show atoms are available to initiate the thermal N O reactions. Also the H C N and N, formed by reaction between C H and N 2 through the prompt mechanism in the fuel-rich region, are oxidized by the available O and O H to form N O . The measurements taken over this interface confirm this sharp increase in the radial direction as shown in Figure 11. Especially at axial distances x = 2.6 to 4.9D, the measured and predicted N O concentrations are very close over this transitional region. The predicted N O level near the exit is, however, about 2 0 % less than that measured (see the profile at x = 12.8D in Figure 11). This discrepancy is also manifested in regions near the furnace walls at all axial distances because the N O at the back of the furnace is brought back to the front in the near wall region through the recirculation process. The inadequacies of using a reduced chemistry approach to model N O formation may have caused this error in the N O prediction in the near wall region and at the furnace exit. However, the model does perform very well in the regions where N O formation is most intense (i.e. where the fuel-rich zone merged with the tertiary air jets). Perhaps the same imperfect assumption concerning the entrance conditions for the secondary air inlets in the burner tile which may result in the aforementioned error in CO prediction, may also cause an inaccurate estimation of CH; radicals in the large fuel-rich zone of this flame. Since C H plays a key role in the formation and reduction of N O (see R4 and R8), an under-prediction of C H on axis (like the under-prediction of C O on axis shown in Figure 8) can result in under-estimating the N O formed at the tip of the flame, possibly accounting for the 2 0 % discrepancy observed above. CONCLUDING REMARKS A strategy using the k-e model for turbulence closure, a non-premixed flamelet model for combustion, a presumed pdf method for turbulence/chemistry interaction, and a reduced chemistry approach for N O x formation, was proposed to simulate the combustion and N O x characteristics in non-swirling diffusion natural gas flames. The strategy, deemed computationally acceptable for industrial-scale applications, was implemented in this study to simulate two pilot-scale natural gas flames generated by selected commercial burners and the following conclusions are drawn: 1. Comparisons between measurements and predictions indicate that the overall details of combustion, temperature and N O characteristics are fairly well simulated by the proposed model. For engineering purposes, the model does provide the correct trend in the variation of these quantities from the near burner region to the exit of the furnace. 2. The largest discrepancy between measurement and prediction of C O and temperature occurs in the fuel-rich core of the flames along the central axis of the furnace. The cause can be due to a combination of the limitations of the non-premixed flamelet model and the possible error in the assumed secondary air boundary condition in the Ultra Low N O x Burner test case. These factors may have affected the quality of the N O x prediction at the furnace exit. 3. Near the thin flame front region where N O formation is most intense, the proposed modelling strategy provides a very accurate estimate of N O concentration in both test cases. The NOx prediction at furnace exit is within 3 ppm of the measured value (135 ppm) in the High N O x Burner test case and 2 0 % less than that measured (112 ppm) in the Ultra L o w N O x Burner test case. Considering the complexity of N O x formation and measurement accuracy, this relatively simple N O modelling approach has performed satisfactorily well. Nevertheless, given the size of error that this model can incur, it should be utilized with great caution for applications where an accuracy of less than 10 ppm of N O x is required. 9 |