Hybrid Phenomenological and Mathematical-Based Modeling Approach for Diesel Emission Prediction 2020-01-0660

In order to reduce the negative health effects associated with engine pollutants, environmental problems caused by combustion engine emissions and satisfy the current strict emission standards, it is essential to better understand and simulate the emission formation process. Further development of emission model, improves the accuracy of the model-based optimization approach, which is used as a decisive tool for combustion system development and engine-out emission reduction.

The numerical approaches for emission simulation are closely coupled to the combustion model. Using a detailed emission model, considering the 3D mixture preparation simulation including, chemical reactions, demands high computational effort.

Phenomenological combustion models, used in 1D approaches for model-based system optimization can deliver heat release rate, while using a two-zone approach can estimate the NO_x emissions. Due to the lack in modeling of 3D mixture preparation phenomena, such models are unable to predict soot or HC emissions. However, employing physical-based air-path and combustion modeling, these models can predict the engine behavior outside of the training area.

Data driven mathematical models are very fast, with sufficient accuracy within the training area to simulate the engine-out emissions. However, they are not capable of extrapolation, if the input parameters and engine operating conditions deviate from the training conditions.

Combining the phenomenological combustion model with mathematical emission prediction model can provide multiple advantages. This so-called hybrid emission modeling approach includes a predictive air-path and combustion model to calculate the characteristic combustion parameters, like ignition delay, flame temperature, etc. for prediction of emissions by mathematical approaches. The mathematical model that uses such input parameters from the combustion process is first trained and then used for prediction of NO_x, soot, HC and CO emissions. Using phenomenological combustion model, the effects of changing the engine operating conditions outside the training area on characteristic parameters coming from the combustion model can be simulated. Using these characteristic parameters, the engine emissions can be well predicted by the mathematical model, using combustion key parameters from engine operating conditions.

In this paper, the novel hybrid emission model is presented along with its advantages and limitations, comparing to traditional physical and mathematical emission modeling using examples from both light and medium duty diesel engines. Accuracy, computational effort and application fields are shown and future development potentials are discussed.