Format Type: PDF, Docs
Read Online: 919
The global community requires increasing supplies of cheap, clean, and transportable energy. Combustion devices currently provide for the bulk of these energy needs, but their operation also produces a variety of undesirable consequences such as instabilities, noise, and pollution. Accordingly, new energy strategies emphasize the importance of improving combustion device designs. The design process itself remains a particular challenge, however, and not even the newest designs can be said to optimally control for the products of combustion. The technique of large eddy simulation (LES) provides a framework for improving this device design process. But while a variety of LES combustion models have been proposed, no champion model yet exists that could be considered truly predictive and regime independent. This work seeks to improve the accuracy and the predictive capabilities of reactive LES. The flamelet approach is accepted here as a working baseline model because of its ability to represent asymptotic combustion physics. Several deficiencies exist in current flamelet implementations, however. First, premixed flamelet LES models are underdeveloped, and cannot yet accurately describe the full range of burning behavior seen in many devices. Second, because the flamelet approach is asymptotic in nature, it only accounts for a single combustion regime of a time. Flamelet methods therefore fail when applied to arbitrarily complex or partially premixed flows. In the first part of this work, a generalized flamelet transformation is derived. This transformation describes how chemical source terms are balanced in the asymptotic limit of either the non-premixed, the premixed, or the auto-ignition regime. The transformation is then used to formulate a method of determining which combustion regime exists locally in a simulation. The indicator is validated using fully resolved triple flame simulations, and is applied in a flamelet based LES of a low swirl burner. The simulations show that the indicator describes combustion regimes more accurately than the traditional flame index. Next, a level set equation is formalized for use in premixed combustion LES. Level set methods are employed in the proposed generalized model because premixed flame fronts tend to be significantly underresolved in LES. Typical transport equation approaches are consequently subject to significant numerical error. Level set methods have historically been problematic, however, when applied in the context of a filtering procedure. Once it has been derived, the appropriate front tracking equation is used to formulate a dynamic model for the turbulent burning velocity of a flame. This dynamic burning velocity model is validated in a direct numerical simulation of a propagating turbulent front, and applied in an LES of a premixed turbulent jet flame. Finally, premixed flamelet models are considered with respect to the issue of flame structure. A new variation of a coupled level set and progress variable approach is proposed in an effort to improve descriptions of turbulence and front interactions. This method is tested using a direct numerical simulation of a premixed turbulent flame propagating in the thin reaction zones regime. The new coupling approach is then applied in an LES of a spray fueled lean direct injection combustor. This simulation makes use of both the new coupled premixed approach, and a traditional non-premixed approach. These approaches are integrated using the proposed combustion regime indicator.