Quantification of prediction uncertainties is requisite for the development of truly predictive combustion models. Parameter uncertainties, such as those associated with reaction rate coefficients, give rise to combustion quantities of interest, such as ignition delay, that are uncertain. In this work, we assessed the applicability of polynomial chaos expansions based on least angle regression (LAR PCE) [1] for the uncertainty propagation (UP) and global sensitivity analysis (SA) of autoignition kinetics. The UP applicability assessment was done on atmospheric, stoichiometric constant volume 0-D autoignition of methane over initial temperatures of 1000-1800 K, and the global SA applicability assessment was done on the same pressure and equivalence ratio conditions but only at an initial temperature of 1000 K. The chemical kinetic mechanism GRI-Mech 3.0 [2] was used for all simulations and direct Monte Carlo results served as references for comparison. We demonstrated the effectiveness of LAR PCE due to its ability to construct accurate sparse PCEs. Further, we carried out UP on a comprehensive range of operating conditions. Given our findings, we anticipate that LAR PCE will be used in the future for the UP and global SA of very large combustion kinetic models involving complex fuels.
展开▼