Abstract
In the design of high-hydrogen content gas turbines for power generation, ashback of the turbulent ame by propagation through the low velocity boundary layers in the premix- ing region is an operationally dangerous event. Predictive models that could capture the onset of ashback would be indispensable in gas turbine design. For this purpose, modeling of the ashback process using the large eddy simulation (LES) approach is considered here. In particular, the goal is to understand the modeling requirements for predicting ashback in confined goemetries. The ow configuration considered is a turbulent channel ow, for which high-fidelity direct numerical simulation (DNS) data already exists. A suite of LES calculations with different model formulations and filterwidths is considered. It is shown that LES predicts certain statistical properties of the ame front reasonably well, but fails to capture the propagation velocity accurately. It is found that the ashback process is invariant to changes in the initial conditions and additional near-wall grid refinement but the LES filterwidth as well as the subfilter models are found to be important even when the turbulence is almost fully resolved. From the computations, it is shown that for an LES model to predict ashback, suffcient resolution of the near-wall region, proper represen- tation of the centerline acceleration caused by ame blockage, and appropriate modeling of the propagation of a wrinkled ame front near the center of the channel are considered the critical requirements.