Evaluation of combustion mechanisms using global uncertainty and sensitivity analyses: A case study for low-temperature dimethyl ether oxidation

Abstract

A global uncertainty analysis is performed for three current mechanisms describing the low-temperature oxidation of dimethyl ether (Aramco Mech 1.3, Metcalfe et al., Int J Chem Kinet 2013, 45, 638–675; Zheng et al., Proc Combust Inst 2005, 30, 1101–1109; Liu et al., Combust Flame 2013, 160, 2654–2668) with application to simulations of species concentrations (CH2O, H2O2, CH3OCHO) corresponding to existing data from an atmospheric pressure flow reactor and high-pressure ignition delays. When incorporating uncertainties in reaction rates within a global sampling approach, the distributions of predicted targets can span several orders of magnitude. The experimental profiles, however, fall within the predictive uncertainty limits. A variance-based sensitivity analysis is then undertaken using high dimensional model representations. The main contributions to predictive uncertainties come from the CH3OCH2 + O2 system, with isomerization, propagation, chain-branching, secondary OH formation, and peroxy–peroxy reactions all playing a role. The response surface describing the relationship between sampled reaction rates and predicted outputs is complex in all cases. Higher order interactions between parameters contribute significantly to output variance, and no single reaction channel dominates for any of the conditions studied. Sensitivity scatter plots illustrate that many different parameter combinations could lead to good agreement with specific sets of experimental data. The Aramco scheme is then updated based on data from a recent study by Eskola et al. (J Phys Chem A, in press), which presents quite different temperature and pressure dependencies for the rates of CH3OCH2O2 [RIGHTWARDS ARROW] CH2OCH2O2H and CH2OCH2O2H [RIGHTWARDS ARROW] OH+2CH2O compared with currently used values and includes well skipping channels. The updates from Eskola worsen the agreement with experiments when used in isolation. However, if the rate of the CH2OCH2O2H + O2 channel is subsequently reduced, very good agreement can be achieved. Owing to the complex nature of the response surface, the tuning of this channel remains speculative. Further detailed studies of the temperature and pressure dependence of the CH3OCH2O2 + O2, CH2OCH2O2H + O2 system are recommended to reduce uncertainties within current dimethyl ether mechanisms for low-temperature conditions.