LES prediction and analysis of knocking combustion in a spark ignition engine

Abstract

Highly boosted spark ignition (SI) engines are more and more attractive for car manufacturers in terms of efficiency and CO2 emissions. However, thermodynamic conditions encountered in these engines promote the occurrence of abnormal combustions like knock or super-knock, which are experimentally difficult to analyse due to the risks of engine damages. The Reynolds averaged Navier–Stokes (RANS) method mainly used in industry for piston engines is not the most appropriate as knock does not always affect the mean cycle captured by RANS. Using an accurate LES compressible code and improved versions of ECFM-LES (Extended Coherent Flame Model) and TKI (Tabulated Kinetics of Ignition) models allowing a full uncoupling of flame propagation and auto-ignition reaction rates, this work demonstrates for the first time that LES is able to describe quantitatively knocking combustion in a realistic SI engine configuration. Contrary to previous studies (Fontanesi et al., 2013 [5]) (Lecocq et al., 2011 [4]), a quantified knock analysis is conducted based on a specific post-processing of both numerical and experimental data. LES is able to predict the in-cylinder pressure variability, the knock occurrence frequency and the mean knock onset crank angle for several spark timings. A 3D analysis also demonstrates that knock occurs at random locations, mainly at the exhaust valves side. Knock intensity is found proportional to the fresh gases mass burned by auto-ignition at low knock intensities, while an exponential increase at the highest intensities suggests the influence of additional factors like the knock location in the cylinder or complex behaviour of knocking combustion.