Uncertainty analysis of NO production during methane combustion

Abstract

Local and Monte Carlo uncertainty analyses of NO production during methane combustion were carried out, investigating the effect of uncertainties of kinetic parameters and enthalpies of formation. In Case I, the original Leeds methane oxidation mechanism with the NO x reaction block was used, but the enthalpies of formation of all species were updated. In Case II, the NCN-containing reactions of the prompt NO formation route were added and the rate parameters of several reactions were also updated. The NO production was examined at the conditions of the Bartok et al. experiments (PSR, T = 1565-1989 K, ϕ = 0.8-1.2, residence time 3 ms). The Monte Carlo analysis provided the approximate probability density function and the variance of the calculated NO concentration, and also its attainable minimum and maximum values. Both mechanisms provided similarly good to acceptable agreement with the experimental results for lean and stoichiometric mixtures, whereas only mechanism Case II could reproduce the experimental data for rich mixtures after a realistic tuning of the parameters. Local uncertainty analysis was used to assess the contribution of the uncertainty of each parameter to the uncertainty of the calculated NO concentration. Enthalpies of formation of NNH and HCCO, and rate parameters of 20 reaction steps cause most of the uncertainty of the calculated NO concentrations at all conditions. The relative importance of the four main NO formation routes was investigated via the inspection of the reaction rates, embedded in the Monte Carlo analysis. NO formation in rich mixtures was dominated by the prompt route, whereas in leaner mixtures the share of the NO formation routes depended very much on the values of rate parameters, when varied within the uncertainty limits of kinetic data evaluations.

Uncertainty of Arrhenius parameters

“…In several works dealing with the uncertainty of combustion systems [26][27][28], a normal distribution was assumed for parameters ln k, truncated at ±3σ . A normal distribution for ln k, but with ±2σ truncation, was used in another combustion uncertainty analysis study [30] and in the uncertainty analysis of atmospheric chemical systems [32,33].…”

A fundamental research on combustion chemical kinetic model’s precision property

“…In the recent years, detailed reaction mechanisms are widely used in various fields, such as combustion, atmospheric chemistry, and so on. Uncertainty analysis method is frequently used to study how variability or inaccuracies in the model's input ultimately lead to a quantifiable degree of uncertainty in the simulation results, for example, flame propagation characteristics [7,8] and pollutants emission [9] etc. However, so far, there are only few uncertainty analysis related researches in China [10,11].…”

Determination of rate parameters based on both direct and indirect measurements

“…However, these values have a relatively high uncertainty, which causes high uncertainty in the simulation results of combustion models [21,22,[32][33][34][35][36]. Repeated, direct measurements of important reactions, even made independently by several groups, did not decrease the uncertainty of the rate coefficients below a certain limit [4].…”

“…The usual assumptions are 3σ deviations [30][31][32][33][34][35][36] or 2σ deviations [21,22]; therefore, m f = 3 or 2, respectively. Note that assuming 3σ deviations instead of 2σ deviations means that smaller variance of ln k is assumed to the same k min and k max limits.…”