Comprehending flame development and misfire at advanced engine conditions : detailed experimental characterizations and machine learning-assisted kinetic analyses

Abstract

Through comprehensive experimental and modeling efforts, this work unravels the underlying mechanisms governing flame development and misfire at advanced engine conditions that are representative of low-load and lean blow-out operations. Toward this, preliminary heat release, autoignition, and flame developing patterns are characterized, via a case study of n-heptane, at ultra-lean conditions in a well-controlled optical engine under various combustion modes including homogeneous charge compression ignition (HCCI), partially premixed combustion (PPC), and reactivity-controlled compression ignition (RCCI). Changes in preliminary heat release and flame developing patterns at three overall equivalence ratios (0.12, 0.18, and 0.24) are first characterized under the PPC mode. Flame development characteristics including flame areas and number of initial flame kernels at close-to-misfire conditions are further extracted and compared across the HCCI, RCCI, and three PPC modes, with two distinctive and one transition regimes identified. Further analyses indicate that sustainable flame development and misfire are largely controlled by the spatial distribution of local equivalence ratio (phi) and local temperature in the mixture, which dictate the initial flame kernel generation and the subsequent flame propagation through localized preliminary heat release and autoignition. Chemical kinetic modeling is also undertaken, using a recently updated gasoline chemistry model, in conjunction with a backpropagation neural network, where the predicted ignition delay map well captures the different regions of flame development. Further kinetic analysis and heat rate of production per reaction analysis corroborate the CH2O planar laser-induced fluorescence experiments and highlight the important chemical kinetics that govern the initial flame development patterns.