Tightening European standards on fuel consumption and pollutant emissions reduction lead to a sophistication of engine concepts and associated control. Since few years, downsizing (reduction of the engine displacement) appears as a major way to achieve those requirements for spark ignition engines. Efficient performance and drivability can be then achieved with a direct injection downsized engine with turbocharging and Variable Camshaft Timing (VCT). One of the major issues of the torque-oriented control is in-cylinder mass observation and control. To have an efficient torque response, the in-cylinder trapped mass, adjusted by the throttle and the waste gate, must be controlled with accuracy according to performance and drivability requirements. Depending on admission and exhaust pressures, the twin VCT will allow to control in-cylinder burned gases rate to reduce fuel consumption and pollutant emissions, and air scavenging to improve transient speed response. Another major issue is in-cylinder trapped mass and air scavenging prediction for AFR control. In this paper, we propose a model-based approach to achieve those engine control issues. The first challenge is to design accurate observers for non-measurable variables (in-cylinder burned gases and trapped air mass). The method is based on a complex high frequency 0D engine model, which has been validated on a large range of engine operating points and transient operations based on test bed results. Then, this model permits to design and learn open-loop nonlinear static observers of in-cylinder masses (based on neural network). The static and dynamic behavior of high frequency 0D engine model allows to achieve design of dynamic and closed-loop in-cylinder mass observation and prediction, multivariable and non-linear control of air path according to in-cylinder mass trajectory (trapped air mass & recirculated gases rate). Then, the complete engine control can be developed and validated on simulation and on a real time Software-In-the-Loop platform based on high frequency 0D engine model, before a complete validation and calibration on test bed. Finally, the complete torque-oriented engine control has been integrated on vehicle. From 0D engine model, a complete vehicle model has been set on real-time platform in order to validate engine control integration and design vehicle layout. The major issue is then supervision of engine control set points (torque, AFR, efficiency) according to engine states (start, idle, driver request, cut-off) and warm-up strategies.
[1]
Niels Kjølstad Poulsen,et al.
Neural Networks for Modelling and Control of Dynamic Systems: A Practitioner’s Handbook
,
2000
.
[2]
G. Corde,et al.
1D Simulation of Turbocharged Gasoline Direct Injection Engine for Transient Strategy Optimization
,
2005
.
[3]
John B. Heywood,et al.
Internal combustion engine fundamentals
,
1988
.
[4]
Bertrand Lecointe,et al.
Downsizing a Gasoline Engine Using Turbocharging with Direct Injection
,
2003
.
[5]
Gérard Bloch,et al.
Neural Control of Fast Nonlinear Systems— Application to a Turbocharged SI Engine With VCT
,
2007,
IEEE Transactions on Neural Networks.
[6]
Lennart Ljung,et al.
Nonlinear black-box modeling in system identification: a unified overview
,
1995,
Autom..
[7]
O. Colin,et al.
3d Modeling of Mixing, Ignition and Combustion Phenomena in Highly Stratified Gasoline Engines
,
2003
.
[8]
F. Le Berr,et al.
Modelling of a Turbocharged SI Engine with Variable Camshaft Timing for Engine Control Purposes
,
2006
.