Statistically Significant Results for the Propagation of a Turbulent Flame Kernel Using Direct Numerical Simulation

Two major issues are always associated with DNS of turbulent combustion. First, the Reynolds number is limited to small values, considerably below those of practical applications. Second, it is difficult to obtain statistically significant results, since this condition would lead to even higher requirements in computing time. Both issues are considered in the present work for the propagation of an initially spherical, premixed methane flame kernel in a turbulent environment, using three-dimensional Direct Numerical Simulations. Chemical processes are described in a realistic manner using tabulation (FPI method), relying on two coordinates to enter a look-up table constructed with 29 species and 141 reactions. Compared to previous results, a considerably higher Reynolds number has been considered by accessing parallel supercomputers. Furthermore, statistically significant results are obtained by repeating several times the DNS simulations, allowing for ensemble averaging. More particularly, the evolution of flame surface area, flame thickness, flame front curvature and flame shape factor are considered in a quantitative manner.