Computational assessment of methane-air reduced chemical kinetic mechanisms for soot production studies

Abstract

Accurate predictions of minor chemical species and radicals are crucial for determining the production of pollutants such as soot. Reduced chemical kinetic mechanisms compromise their accuracy in favor of a lower computational cost. When using these reduced mechanisms then, it is important to assess both the accuracy of the results obtained, and the amount of computational time saved. This work describes such an assessment of seven methane-air reduced chemical kinetic mechanisms to be used for carrying out soot formation studies. The reduced mechanisms evaluated involve different numbers of chemical species and reaction steps. The assessment is carried out using partially stirred reactors, with and without in situ adaptive tabulation-based chemistry acceleration techniques, by comparing the results with detailed chemical kinetics baseline computations. In terms of accuracy, for equivalence ratios featuring significant amounts of soot (above ~1.5), and considering only those mechanisms that are readily used with the soot model utilized, the reduced mechanisms results show that major species are in general predicted reasonably well (~0–10 % discrepancies). Larger discrepancies between detailed and reduced mechanisms results (~0.2–16 %) are observed, however, when predicting minor species. The in situ adaptive tabulation technique used in this work leads to further reductions in the accuracy of the minor species predicted, i.e., to further increases in discrepancies (~0.1–7 %). Regarding the computational cost, the results show that savings of up to 57 % can be obtained when using the reduced mechanisms analyzed. The use of chemistry acceleration techniques results in further cost reductions ranging from 1 to 43 %. Additionally, discrepancies in the predictions of soot volume fraction of the order of 4–11 % are observed when using reduced mechanisms. The results obtained in this work emphasize overall the need of carefully selecting the reduced mechanism that is more suitable for a given application.

Appendix

Regarding the recommended settings for the ISAT-based chemistry acceleration technique utilized in this work, there are several aspects to be considered, including (1) the ISAT error tolerance, (2) the number of binary trees used to tabulate the composition space accessed region, and (3) the amount of memory allocated for storing the tabulated data. Increasing the ISAT error tolerance implies not only obtaining more cost savings, but also introducing larger errors associated with the linear extrapolation that is part of the ISAT approximation. Acceptable values for this error tolerance were found to be of order of 10⁻⁶–10⁻³. When accuracy is a critical issue, the tightest tolerances are preferred, which in turn penalize the computational cost. In simulations involving transient periods, such as those carried out in this work, it is important to recreate the ISAT binary trees, as this allows discarding stored data that may be seldom used during the statistical steady state part of the computation. The number of binary trees to be utilized, i.e., the number of binary tree recreations, and the corresponding size, needs to be determined in a case-by-case basis. Accordingly, the simulations performed have shown that the best results, in terms of cost savings, are obtained using two binary trees only, with binary tree recreation occurring between 5 and 10 residence times. The results obtained have shown as well that small trees (~150 MB, in memory size) are preferred to larger ones. These figures are expected to change of course when dealing with different physical problems. Nevertheless, they should provide a general idea of what could be expected when employing ISAT-based chemistry acceleration techniques, as those utilized in this work.