Using large eddy simulations to study diesel DI-HCCI engine flow structure, mixing and combustion

Jhavar, R. Using Large Eddy Simulations to Study Diesel DI-HCCI Engine Flow Structure, Mixing and Combustion. University of Wisconsin-Madison, 2007.

The use of an ?engineering large eddy simulation (LES)? approach, as an advanced turbulence modeling technique for engines, was further substantiated. This technique was used to study diesel direct injection (DI) HCCI engine flow and its effects on mixture formation and combustion. In LES, the larger three dimensional unsteady fluid motions are directly represented, whereas the effects of smaller-scale motions are modeled. This makes LES more suited for investigating complex time dependent engine flows; as compared to the more traditional Reynolds averaged Navier-Stokes (RANS) turbulence models.

Four LES models were implemented into the KIVA-3V engine simulation CFD code: Smagorinsky model, one-equation Menon model, dynamic Smagorinsky model and dynamic structure model. DI-HCCI cases simulated on a CAT 3401 engine grid showed that LES provides additional details of in-cylinder large scale flow structures, including how they develop, evolve and impact mixing, as compared to RANS. The detailed kinetics solver CHEMKIN was incorporated with the LES-KIVA code to more accurately predict low temperature chemistry for early injection HCCI conditions. Simulations performed on two different engines showed that LES-CHEMKIN predicts flow structure as well as temperature and intermediate chemical species distribution more accurately than RANS-CHEMKIN when compared with experimental images. A grid sensitivity study also concluded that LES, unlike RANS, shows increased sensitivity to geometric details with increasing resolution.

The preceding studies helped validate the use of a computationally efficient ?engineering LES? approach to investigate engine flows. The next step was to use this approach to study engine processes that were difficult to analyze using traditional RANS based turbulence models. Multiple engine cycles were simulated to identify sources of cycle-to-cycle variability under early injection HCCI conditions. Unlike BANS models, LES was able to explain some of the causes of cyclic variability arising from intake flow and fuel vaporization. LES simulations were also used to evaluate variable valve actuation (VVA) strategies in controlling HCCI combustion. Emissions trends predicted by LES matched experimental measurements better than BANS. LES predicted engine flow showed increased sensitivity to the change in valve profile, indicating a greater influence of flow structure on emissions formation.