Investigation on the Potential of Quantitatively Predicting CCV in DI-SI Engines by Using a One-Dimensional CFD Physical Modeling Approach: Focus on Charge Dilution and In-Cylinder Aerodynamics Intensity 2015-24-2401

Increasingly restrictive emission standards and CO₂ targets drive the need for innovative engine architectures that satisfy the design constraints in terms of performance, emissions and drivability. Downsizing is one major trend for Spark-Ignition (SI) engines. For downsized SI engines, the increased boost levels and compression ratios may lead to a higher propensity of abnormal combustions. Thus increased levels of Exhaust Gas Recirculation (EGR) are used in order to limit the appearance of knock and super-knock. The drawback of high EGR rates is the increased tendency for Cycle-to-Cycle Variations (CCV) it engenders. A possible way to reduce CCV could be the generation of an increased in-cylinder turbulence to accelerate the combustion process. To manage all these aspects, 1D simulators are increasingly used. Accordingly, adapted modeling approaches must be developed to deal with all the relevant physics impacting combustion and pollutant emissions formation. In this study, a CCV modeling approach for system simulation integrated into the 12 gas CFM1D combustion model has been used to reproduce CCV in a downsized Direct Injection (DI) - SI single cylinder engine. The CCV model was developed by integrating physical understanding gained from 3D CFD based on a Large-Eddy Simulation (LES) approach. The experimental database includes a complete engine map, single-parameter variations of dilution rate as well as standard and increased aerodynamics operating conditions. Once the CCV model was calibrated, it was applied to simulate CCV over the complete experimental database and showed a good reproduction of experimental observations. As a result of using a physics based CCV model, it was possible to acquire some understanding of the reasons for the observed combustion variability from the performed simulations. Finally, with the perspective of using the model for engine control applications, its potential to run in Real Time (RT) was evaluated.