Optimal-Behavior-Based Dynamic Calibration of the Automotive Diesel Engine

This paper presents a novel automotive engine management system (EMS) calibration methodology for rapid dynamometer test-bed data collection and mapping I.C. engine controllers with significant dynamic and transient performance requirements. This paper applies the methodology to an industrial state of the art WAVE-RT1 model of a 1.5 L Turbo EU6.1 Diesel engine acting as a virtual dynamometer based engine. The approach directly yields a feedforward controller in a nonlinear polynomial structure that can be either directly implemented in the EMS or converted into a dynamic or static lookup table format. The methodology is based on multistage blackbox modeling and dynamic system optimization. The process can exploit the power of global constrained numerical optimization codes and use system identification techniques with dynamic design of experiments. The objective of the engine controller optimization is to improve the fuel economy while maintaining specified (legislated) limits on emissions and map smoothness for driveability. The key contribution is a novel approach to obtaining a feedforward dynamic calibration controller from the system identification of the computed optimal behavior. Model structure selection techniques are shown to be usefully employed to further improve the accuracy of the system identification and so enhance the control performance of the dynamic controllers. The results indicate that the dynamic calibration methodology leads to a considerably reduced requirement for testing time and an improvement of between 1.9% and 2.6% in fuel economy over extra urban driving cycle without violating the emission constraints compared with current steady-state model-based calibration methods.