| OCR Text |
Show 6 Steele et ale (1994) compared chemical reactor modeling to some of the NOx data plotted in Figure 1. The model consisted of a PSR (perfectly stirred reactor) followed by a PFR (Plug flow reactor). The C1-C2 hydrocarbon oxidation/NOx formation mechanism of Miller and Bowman (1989) was used in these calculations. The PFR component comprised no more than 20% of the total reactor. The PFR component had little effect on the NOx; that is, the experimental NOx could be matched by treating the whole reactor as a PSR. However, the PFR component was required to match the experimental temperature (for the short residence time, nearly adiabatic cases) and the CO concentration. That is, the PFR component permitted the gas to relax towards (but not to) equilibrium. (Also, the CO relaxation in the probe was modeled using a second PFR.) Close agreement of the modeled NOx to that measured was obtained for the 1. 7ms and 3.4ms cases (of Figure 1). The 1.7ms model assumed an adiabatic reactor. The 3.4ms model was assigned the measured temperature. For the 6.0/6.9ms cases, the model (based on the measured temperature) under-predicted the NOx by 20 to 30%. This is thought to be due to slight temperature non-uniformities and/or finite rate mixing effects within the reactor. In these long residence time experiments, the combustion loading is decreased from that of the other experiments, and the reactor is stirred by subsonic jets of about 25% pressure loss. The modeling also gave the percentage contributions of the three NOx formation routes. The percentages of NOx formed by the nitrous oxide, Zeldovich, and prompt mechanisms at atmospheric pressure varied from 65:25:10 at 1650K to 35:50:15 at 1850K. In Figure 2, the NOx data are re-expressed on a dry, 15 % 02 basis, and are plotted as NOx versus the measured combustion temperature. Each of the datasets is fit with an exponential function. It should be noted that for each of the datasets the residence time decreased slightly as the temperature was increased (because the air flow rate was held constant). For these cases the actual residence times (from lowest to highest temperature) are as follows: 1.7ms nominaV1.76 to 1.58ms actual; 1.9ms nominal/2.04 to 1.78ms actual; 3.4ms nominal/3.54 to 3.22ms actual; 3.5ms nominal/actual residence time held close to 3.5ms; 6.9ms nominaln.21 to 6.58ms; 6.0ms nominaV6.28 to 5.70ms actual. Because of the residence time variation, the present data plotted as NOx versus reciprocal temperature show a somewhat lower apparent activation energy (by about 2kcal/ gmol) than data plotted as NOx divided by residence time versus reciprocal temperature. In Figure 3, the data are plotted as NOx (15% 02 dry) divided by residence time (tau) versus the reciprocal of the measured temperature. The change from NOx (wet, actual 02) to NOx (dry, 15% 02) decreases the apparent activation energy of the NOx formation by about 3kcal/gmol. Two curve fits are shown in Figure 3: one for 600K inlet mixture temperature and (nominal) residence times of 1.7, 1.9, and 3.4ms, and one for 300K inlet temperature and the residence time of 3.5ms. As in Figure 1, NOx/tau shows dependency on the combustion temperature. However, in the 15% 02 format, NOx/tau shows dependency also on the inlet mixture temperature. The respective apparent activation energies for the two curve fits of NOx formation are 44 and 34kcal/gmol. Figure 3 also indicates a more complex dependency of the NOx on residence time than noted in Figure 1. That is, the long residence time data (i.e., the 6.0 and 6.9ms cases) do not scale simply as NOx/tau versus Iff. |
