Challenges in thedevelopmentof modern combustion technology
There are many interesting topics in numerical combustion, but two concern me the most.
This is, firstly, the gap between the underlying idea of most combustion models and their applicability to technically relevant applications and secondly, the challenges of Large-Eddy Simulations of near-wall turbulent, reactive flows.
Flame Index Approach
The first issue points on the fact that most combustion models are based on the ansatz of a perfectly premixed or non-premixed state. Depending on their level of simplicity or sophistication, they work fine as long as the application under consideration stays in the same combustion regime.
The order of error is usually known or can be well estimated, as there is a large pool of publications and evaluation data on model performances in geometrically simple cases with a single and distinctive premixed or non-premixed reaction zone.
However, many applications include either both regimes in the domain or can be considered as partially premixed from the beginning. Hence, models perform well when they treat their underlying regime but can remarkably fail, if they encounter the opposite flame-type.
Not to mention that it is difficult to grasp the quality of results, if the flame is partially premixed which is a state neither fully premixed or non-premixed.
Therefore, either modeling type can fail to handle this particular combustion regime.
My work aims to unite the advantages of a premixed and a non-premixed combustion model by the use of a Flame Index- a parameter that is meant to detect and identify reaction zones and their corresponding combustion regime. It was orignally introduced as a simple concept for post-processing purposes to check the sign of the scalar product of fuel and oxidiser gradients in order to identify premixed or non-premixed combustion.
In my work, however, I chose to use the Flame Index actively during the computations.
I implemented transport equations of certain properties to evaluate the combustion regime according to suggestions by E. Knudsen and H. Pitsch [2009,2012,2015].
The resulting Flame Index was then used as blending factor between two a priori tabulated chemistry manifolds.
One means a table of freely propagating premixed flames and the other of non-premixed counterflow flames.
Thus, in dependence on the accuracy of the Flame Index, the CFD-solver transports only few variables as access variables to the chemistry tables and picks the appropriate value or blends between two values according to the weighting of the Flame Index.
LES of Effusion Cooling
The second issue sets the focus on Large Eddy Simulations of effusion cooling flows. The full resolution of such near-wall flows is computationally expensive, but promises to capture essential aspects of turbulence, in return. A fine mesh allows to see the characteristics of the flow, such as the horse-shoe vortex at the edge of every effusion hole and the pair of counter-rotating vortices that forms a tube which penetrates first into the main flow in order to return and attach to the solid boundary. Although turbulent flows close to solid boundaries are anisotropic by nature, the interaction between main flow and effusion jet increases this effect and lifts this appearance higher, away from the wall. This stands in contrast to the generally isotropic turbulence far away from boundaries.
These aspects were numerically investigated with a case that has been analysed experimentally at the institute of Reaktive Strömungen und Messtechnik. The simulations were validated with the corresponding data of the cooling effciency. The turbulent invariant stress analysis was done by the evaluation of Lumley-triangles on different meshes.