Particulate Flows in Aftertreatment Systems

Abstract: Emissions from internal combustion engines contain many components that have a detrimental effect on the environment and on human health, such as nitrogen oxides (NOx) and particulate matter (PM). In addition, the final product of any combustion of fossil fuel - carbon dioxide (CO2) - contributes to global warming. In order to reduce the emissions of CO2, more efficient engines are needed, and these typically necessitate the development of new exhaust gas aftertreatment systems. Lean-burn engines (e.g. the diesel engine) are more efficient than conventional petrol engines, but emit more PM and require addition of a reducing agent to reduce NOx. Usually, a urea-water solution or a hydrocarbon is sprayed into the system. In addition, the PM content of the exhaust must be reduced in terms of both mass and number. In the current work, detailed mathematical models are used to investigate the motion and deposition of PM and droplets in generic exhaust gas aftertreatment systems. It is shown that PM from internal combustion engines can be divided into three groups depending on their size, and that these groups are transported differently in the aftertreatment system. This is reflected in the extent and location of particle deposition, and can be taken advantage of in emission control engineering. Several particle transport models of differing complexity are presented and used to study the PM trapping characteristics of a number of filter designs. Also the influence of turbulence on the transport of particulate matter and droplets in aftertreatment systems is studied by means of numerical simulations. Finally, a model for simulations of gas-solids systems involving particles of size significant to that of the bounding geometry but also to the mean free path of the gas is presented. This very challenging flow situation is encountered inside the pores of a porous wall in a typical diesel particulate filter. It is shown that the new model can provide more accurate results than the previously available methods of similar computational cost.