Automatic Chemistry Mechanism Reduction of Hydrocarbon Fuels for HCCI Engines Based on DRGEP and PCA Methods with Error Control

Abstract

The chemical kinetics of hydrocarbon fuels determines the combustion characteristics and pollutant emissions of homogeneous charge compression ignition (HCCI) engines. Including comprehensive chemical mechanisms in HCCI engine models provides accurate predictive results that can be used to improve engine designs. However, a large number of simulations are usually required to optimize an HCCI engine, and the use of comprehensive chemical mechanisms is prohibitive. Furthermore, an increased demand for surrogate fuels that better represent real fuels has resulted in further increases in the size of chemical mechanisms as the carbon number of surrogate fuel species and the number of fuel components considered increases. Consequently, reduced mechanisms of smaller sizes, which are able to represent their corresponding comprehensive mechanisms over a wide range of conditions are necessary. This paper presents an approach that fully automates the process of reducing comprehensive chemical mechanisms of fuels for HCCI engines. The approach is based on the directed relation graph with error propagation (DRGEP) and principal component analysis (PCA) methods. In the first stage, the DRGEP method is used to efficiently remove redundant species. This is followed by the use of the PCA method to further remove insignificant reactions and species. During the entire process, the performance of the reduced mechanism is monitored to ensure that the generated mechanism satisfies user-specified error tolerances. In the present study three comprehensive mechanisms that include n-heptane, iso-octane, and methyl decanoate (MD) were investigated. The proposed approach successfully reduced the comprehensive mechanisms of n-heptane (561 species and 2539 reactions), iso-octane (857 species and 3606 reactions), and MD (2878 species and 8555 reactions) to reduced mechanisms with sizes of 140 species and 491 reactions, 195 species and 647 reactions, and 435 species and 1098 reactions, respectively, while maintaining small errors compared to the full mechanisms.