Semi-physical models to assess the influence of CI engine calibration parameters on NOx and soot emissions

Abstract

The progressive reduction of authorized emission levels in automotive Diesel engine standards has motivated the development of numerous technologies (exhaust gas recirculation (EGR), high pressure injection systems, sophisticated boosting systems, after-treatment devices, etc.) which, in turn, drastically increases the complexity of engine calibration. In this context the development of reliable simulation tools can help reduce the cost and time required for calibration. After a short introduction analysing the main currently existing models for evaluating engine emissions, this paper presents a novel 0D semi-physical model to assess engine-out NOx and soot emissions. The combustion process is modelled via Barba’s approach, while a thermodynamic two-zone calculation is used to evaluate adiabatic flame temperature. Emissions are modelled with semi-physical sub-models. This rather original approach does not evaluate emission on a crank-angle basis but only at exhaust valve opening (EVO), thus saving calculation time. The main physical parameters influencing pollutant formation are evaluated by the high-frequency 0D model and used as inputs for pollutant sub-models. NOx evaluation relies on a cartography linking NOx to O2 concentration and maximum values of in-cylinder bulk temperature and adiabatic flame temperature. Soot evaluation relies on a global equation, linking soot concentration to the main factors influencing formation and oxidation processes, in particular O2 concentration, in-cylinder pressure, temperatures and durations of some specific phases of the heat release rate (HRR), as well as turbulence intensity. The calibration of the models is thus quite easy and is described in the paper. The results of the models are then compared with measurements (different from those used for model calibration). NOx predictions are within ±20% of measured values for 95% of the tested operating points, with a R2 of 0.99, while for soot prediction a R2 coefficient of 0.93 is obtained and 96% of the tested points are within ±0.005mg/cycle. Moreover, engine parameters sweeps (at constant engine speed and load) involving EGR rate, boost pressure, injection pressure and timing are performed for five operating points. The agreement with experiments is good on both qualitative and quantitative points of view, as long as a conventional combustion mode is achieved. Although simple and fast, these models are not only able to interpolate between the training points but also to extrapolate with a reasonable accuracy when the engine calibration parameters are changed. This latter property is rarely demonstrated in existing models.