Optimization of methane simplified chemical kinetic mechanism based on uncertainty quantitation analysis by sparse polynomial chaos expansions

Abstract

A robust and compact simplified chemical kinetics mechanism for methane oxidation is newly developed based on the Directed Relation Graph with Error Propagation method (DRGEP) and the Full Species Sensitivity Analysis method (FSSA). It is constructed by adaptive self-tuning of reaction rates with a non-dominated sorting genetic algorithm II (NSGA-II) approach to satisfy ignition behaviors under varied engine conditions. Its high fidelity is achieved by uncertainty quantitative (UQ) analysis to retain the uncertainties caused by updated reaction rates. In UQ analysis, the sparse polynomial chaos expansions (SPCE) model based on the Least Angle regression (LAR) method is applied to the adaptive selection of significant polynomial bases, and the accuracy of the SPCE model is tested using a statistical methodology (e.g. leave-one-out cross-validation). Consequently, compared with the full PCE approximations, the UQ analysis scheme reduces the computational cost by more than one order of magnitude and effectively alleviates the “dimension disaster” dilemma of the classical PCE scheme. Therefore, the application of the adaptive SPCE scheme allows more reactions to be involved in quantifying the uncertainty propagation. Finally, the aforementioned new simplified methane mechanism is widely verified under engine-relevant conditions, and good auto-ignition performance is highlighted, especially in fuel-lean combustion.