Combustion in DI-SI engines using G-equation model

It is well established nowadays that turbocharged Direct Injection Spark Ignited Engine (DI-SI) appears to be the most relevant way to improve fuel efficiency of Spark Ignited engines with fuel economy and pollutant reductions control.

The increasing number of published industrial and research works over the past decade indicated the high interest of the technology in the automotive industry, undoubtedly it is also a sign that the development of such engine is complex and optimization would be quite difficult to achieve.

The combustion process is key phenomena, its modeling using CFD is a useful tool to understand the processes taking place and a way to support the design.

An ensemble of physics like spray and mixing processes, spark ignition and initial flame kernel growths, flame propagation including auto-ignition and pollutants formation need to be taken into account simultaneously and included in the CFD model. This task is quite challenging in order to achieve a reliable and predictive tool.

A hierarchical approach is needed to achieve reliable model and in the same time a tool with a reasonable computing time compatible with the engine design.

Therefore, we will establish the methodology based on URANS only, LES is still very expensive. We will look at carefully to time and length scale for each phenomenon, which will allow us to decouple the complexity of the chemistry from the complexity of the flow and the mixing processes.

For the reasons mentioned above, we have decided to model the turbulent flame front propagation using an enhanced G-equation (level set) approach combined with flamelet library for emissions (soot and NOx). The link between the two is established using a presumed probability density function extracted using mean and turbulent extracted from the flow. 

To compute knock the unburnt gas temperature is used to compute reaction rates in fresh gases thanks to a conditional approach. The chemical source terms are evaluated from detailed chemistry using a reduced mechanism established for a surrogate fuel. A stochastic approach which provides the probability of auto-ignition and distinct criteria to determine the mean knock onset as well as the number of knocking cycles. This approach is similar to the approach developed by Linse and Al.

The set of sub-models and methodology presented above has been implemented in STAR-CD. Applications and validations have been performed for typical automotive turbocharged DI-SI engines operating at part load and full load. The results of the global thermodynamic quantities, knock onset as well as the emission levels are compared to experimental data.

Siemens PLM Software
Marc Zellat
Session Time Slot(s): 
07/03/2017 - 9:20am-07/03/2017 - 9:45am
Room III
Presentation Image: 
Session Track: 

Session Tracks (STAR Global Conference 2017)