Development and Validation of a New Zero-Dimensional Semi-Physical NOx Emission Model for a D.I. Diesel Engine Using Simulated Combustion Process 2015-01-1746

Reducing NO_x tailpipe emissions is one of the major challenges when developing automotive Diesel engines which must simultaneously face stricter emission norms and reduce their fuel consumption/CO2 emission. In fact, the engine control system has to manage at the same time the multiple advanced combustion technologies such as high EGR rates, new injection strategies, complex after-treatment devices and sophisticated turbocharging systems implemented in recent diesel engines. In order to limit both the cost and duration of engine control system development, a virtual engine simulator has been developed in the last few years. The platform of this simulator is based on a 0D/1D approach, chosen for its low computational time. The existing simulation tools lead to satisfactory results concerning the combustion phase as well as the air supply system. In this context, the current paper describes the development of a new NO_x emission model which is coupled with the combustion model.

The proposed zero-dimensional, semi-physical, NO_x prediction model is based mainly on a high frequency combustion model (Barba's approach) coupled with a thermodynamic calculation of the temperature (adiabatic flame temperature) in the burned gas products from a stoichiometric mixture. Furthermore, a new empirical correlation is created to relate the corresponding physical components such as the highest local temperature and the available oxygen concentration, describing a uniform progression of NO_x emissions. The model was built and validated over a large range of operating points on a 1.6 liter Euro 5 diesel engine. The simulation results show that the model can predict the multiple effects of exhaust gas recirculation and of the various injection parameters for both single and multi-injection cases.

The model provides the required trade-off between predictability (pure physical models) and simplicity (empirical models) which are mostly incompatible. A major advantage of the final NO_x model is that it does not require any calibration process once the combustion model is identified.