Modelling Turbocharged Spark-Ignition Engines: Towards Predictive Real Time Simulators 2009-01-0675

Due to increasingly stringent regulations, reduction of pollutant emissions and consumption are currently two major goals of the car industry. One way to reach these objectives is to enhance the management of the engine in order to optimize the whole combustion process. This requires the development of complex control strategies for the air and the fuel paths, and for the combustion process. In this context, engine 0D modelling emerges as a pertinent tool for investigating and validating such strategies. Indeed, it represents a useful complement to test bench campaigns, on the condition that these 0D models are accurate enough and manage to run quite fast, eventually in real time.

This paper presents the different steps of the design of a high frequency 0D simulator of a downsized turbocharged Port Fuel Injector (PFI) engine, compatible with real time constraints. The main part of this work lies in the combustion process modelling, the most critical point to reach real time performances. The methodology first consists in calibrating a 0D phenomenological combustion model to represent in a predictive and detailed way the different phenomena occurring in the cylinder. Then, this model is used to create an extended database of operating conditions which can not be provided by experiments for time and costs reasons. This numerical database also provides a detailed insight into the combustion processes by giving access to the evolution of many physical quantities in the cylinder during the engine cycle (pressure, temperature, turbulent flame surface, burned mass fraction…). These quantities are then involved in a learning process based on a neural network approach to build a fast combustion model which retrieves the rate of heat release evolution from in-cylinder conditions at each cycle. Finally, the fast model is embedded in a complete engine simulator running in real time.

Simulation results show good agreement with experimental data for a wide range of engine steady state operating points. The obtained simulator therefore appears to be a relevant tool for the development and test of engine control strategies.