3 D Simulation of Multiple Injections in DI Diesel Engine

Paper in proceeding, 2004

In numerical simulation of spray combustion in DI Diesel engines, a correctly modeled spray behavior is essential, since it forms a “fundamental” basis for the combustion process. Droplet penetration and vapor distribution determines combustion characteristics such as ignition delay and position, flame lift-off, heat release rate, emissions formation, etc. The process of spray combustion development is modeled in several distinctive stages and much research has been devoted to this area. Stages such as spray primary- and secondary -break-up, droplet collision and evaporation can all be modeled in various ways with not always consistent results. In this paper, among these stages, a special attention is paid on the development of a new collision model and it's effect on spray combustion in both the 2-D constant volume vessel and in DI Diesel engine combustion chamber represented on a 3-D sector grid. The calculations were performed using the KIVA3V, release 2, CFD code supplemented by the Partially Stirred Reactor (PaSR) model for turbulence-chemistry interaction coupled with the detailed chemical mechanism (68 species, 280 reactions) of diesel oil surrogate represented by a mixture of n-heptane and toluene. The modified collision model (Nordin, 2001) includes elements of a classic collision theory to calculate a probability of droplet collision/coalescence with a highly reduced grid dependence compared with the original O’Rourke model. The model has a considerable impact on the spray development, and it has been validated using constant volume experimental data. When simulating the Volvo NED5 DI Diesel engine in a split injection mode, it was found that the collisions play a crucial role in predicting the rate of heat release during the pilot injection. More specifically, if a collision probability is over-predicted, it causes a droplet cluster formation and too fuel lean conditions resulting in a decrease in combustion intensity. Since the O'Rourke collision probability is proportional to the number of droplets in the cell, the droplet displacement volume based on relative velocity of colliding droplets and inverse proportional to the cell volume, it causes both high grid dependence and does not discriminate droplet trajectories which are not intersecting each other. In a new grid independent formulation, two vital requirements have to be met for collision between two droplets to occur: first, they have to travel towards each other being in a so-called “collision” cone and second, all these colliding droplets are constrained to a certain exponential time-decaying probability. The former weak rate of heat release for the pilot injection (Gustavsson, Golovitchev, 2003) related to the O´Rourke collision model has been replaced by better predictions of cylinder pressure and heat release rate resulting in a closer agreement to experimental data from the single cylinder engine.