Publisher's Synopsis
Downsizing of modern gasoline engines with direct injection is a key concept for achieving future CO22 emission targets. However, high power densities and optimum efficiency are limited by an uncontrolled autoignition of the unburned air-fuel mixture, the so-called spark knock phenomena. By a combination of three-dimensional Computational Fluid Dynamics (3D-CFD) and experiments incorporating optical diagnostics, this work presents an integral approach for predicting combustion and autoignition in Spark Ignition (SI) engines. The turbulent premixed combustion and flame front propagation in 3D-CFD is modeled with the G-equation combustion model, i.e. a laminar flamelet approach, in combination with the level set method. Autoignition in the unburned gas zone is modeled with the Shell model based on reduced chemical reactions using optimized reaction rate coefficients for different octane numbers (ON) as well as engine relevant pressures, temperatures and EGR rates. The basic functionality and sensitivities of improved sub-models, e.g. laminar flame speed, are proven in simplified test cases followed by adequate engine test cases. It is shown that the G-equation combustion model performs well even on unstructured grids with polyhedral cells and coarse grid resolution. The validation of the knock model with respect to temporal and spatial knock onset is done with fiber optical spark plug measurements and statistical evaluation of individual knocking cycles with a frequency based pressure analysis. The results show a good correlation with the Shell autoignition relevant species in the simulation. The combined model approach with G-equation and Shell autoignition in an active formulation enables a realistic representation of thin flame fronts and hence the thermodynamic conditions prior to knocking by taking into account the ignition chemistry in unburned gas, temperature fluctuations and self-acceleration effects due to pre-reactions. By the modeling approach and simulation methodology presented in this work the overall predictive capability for the virtual development of future knockproof SI engines is improved.
