플라즈마 흐름 반응기 측정을 이용한 산화우라늄 반응 메커니즘의 확률론적 최적화

where \({\varvec{k}}\) is a vector containing the reaction rate coefficients and \(n^{exp}_i\) and \(n^{calc}_i({\varvec{k}})\) are the measured and calculated species number densities at time point i, respectively. The optimization problem is solved by employing an iterative procedure that finds an optimal parameter set \({\varvec{k}}\) that minimizes the objective function \(\phi\). In the context of the current problem, an optimized \({\varvec{k}}\) value would represent a set of rate coefficients that closely match the uranium oxide formation rates observed in the laser ablation or PFR experiments. Typically, deterministic nonlinear least squares methods, such as the Gauss-Newton or Levenberg-Marquadt methods3.0.CO;2-R (1998)." href="#ref-CR21" id="ref-link-section-d84530096e1787"21,22,23, are employed for such optimization problems. Modern computational techniques, such as neural networks24, can also be used to this end./p>0.5\) for all solution curves can be used. If a solution is retained, then the statistical values calculated as part of Eq. (14) and the corresponding sets of modified rate parameters from Eq. (3) are stored. It is preferable to have a lax selection criterion at this stage in order to build up a sizable number of candidate mechanisms. The mechanisms passed to the genetic algorithm are then selected from these candidates based on the fitness metric used by the algorithm./p>0\)). This criterion was satisfied by 8.61% of generated mechanisms from 2.3 million samples. These 200,000 candidate mechanisms could be used as a starting population for the GA portion of the algorithm. However, to keep the run time of the GA reasonable, only a subset of the candidate mechanisms is used. This is done by first sorting the mechanisms according to two versions of the objective function \(\phi\) previously defined in Eq. (14). The two objective functions are: \(\phi _1\), which includes all terms in the objective function:/p>\sigma _k\) (where \(\sigma _k\) is the standard deviation of k)./p>3 cm) are covered. These reactions could be better constrained in the future by performing upstream measurements over a wider range of \(\hbox {O}_2\) conditions or with reduced \(\hbox {H}_2\)O concentrations (i.e. using a desolvating nebulizer). The current dataset is also limited in the range of temperatures and cooling rates covered, which also inhibits the location of a true global optimum. This is related to the temperature dependence of the Arrhenius rate expression, since similar reaction rates can be achieved using different combinations of coefficients if the reaction is active over a limited temperature range in the system. Based on the above observations, the lack of convergence toward a singular global optimum appears to stem more from the limited constraining data rather than from a shortcoming of the genetic algorithm. Even so, four reactions (R1, R4, R6, and R11 in Table 2) are consistently well constrained by the current optimization, demonstrating the reliability of the MCGA method and highlighting the dominant reaction channels for UO formation in the PFR./p>3 cm) behavior of U is poorly captured. While the R\(^2\) values appear to be adequate (outside of the 0 \(\hbox {O}_2\) dataset 1 case), visual inspection reveals that neither the 3–5 cm decrease nor the 5–8 cm saturation behavior is well matched for any dataset. This suggests that the saturation may be driven by a non-chemical effect that is not accounted for in the current 0D treatment of the PFR. As discussed previously, this behavior is likely caused by optical scattering of the strong upstream U emission line. Therefore, we will consider it an invalid constraint for the current optimization problem and will focus our subsequent analysis on the \(\phi _2\) optimized mechanism. As shown in the next section, the \(\phi _2\) result captures the 3–5 cm decrease in U intensity while disregarding the subsequent signal saturation due to the exclusion of the log U term./p> 7\) cm where temperature drops below 3000 K./p>3 cm). The UO sensitivity plots show that the upstream UO evolution is relatively insensitive to the reaction mechanism. Further downstream, UO is most sensitive to R6 and R13. The remaining optimized channels from Table 5, R9 and R10, appear to make finer adjustments to the downstream UO evolution. Overall, the upstream behavior of the mechanism is constrained mainly by U data, while both U and UO measurements play a role in constraining the downstream behavior./p>

Tadi, M. & Yetter, R. A. Evaluation of the rate constants in chemical reactions. Int. J. Chem. Kinet. 30, 151–159. 3.0.CO;2-R"https://doi.org/10.1002/(SICI)1097-4601(1998)30:2<151::AID-KIN7>3.0.CO;2-R (1998)./p>

3.0.CO;2-R" data-track-action="article reference" href="https://doi.org/10.1002%2F%28SICI%291097-4601%281998%2930%3A2%3C151%3A%3AAID-KIN7%3E3.0.CO%3B2-R" aria-label="Article reference 21" data-doi="10.1002/(SICI)1097-4601(1998)30:23.0.CO;2-R"Article CAS Google Scholar /p>