Flame growth and wrinkling in a turbulent flow

High-speed planar laser-induced fluorescence (PLIF) and 3-D large eddy simulations (LES) are used to study turbulent flame kernel growth, wrinkling and the formation of separated flame pockets in methane/air mixtures. Turbulence was effected by a set of rotary fans situated in a cylindrical enclosure. Flame wrinkling was followed on sequential 2-D OH images captured at kHz repetition rates. Under stoichiometric conditions and low turbulence levels the flame kernel remains singly connected and close to spherical in shape. By increasing turbulence or reducing the stoichiometry of the mixture the formation of separated pockets could be observed and studied. The mechanisms behind these phenomena are investigated qualitatively by LES of a level-set G-equation describing the flame surface propagation in turbulent flows. PACS: 47.27.-i; 42.62.-b Although flame-turbulence interactions are of fundamental interest to combustion science [1] the phenomenon still poses a severe challenge to theorists and experimentalists alike because of the time-dependent and three-dimensional nature of the problem [2]. Turbulent eddies can wrinkle the flame and even quench it; on the other hand, eddies improve heat transfer and the mixing of species which increase the efficiency of technical combustion devices. Much of the present understanding on this topic stems from the study of simple and isolated flame vortex interactions (FVI) in laminar flames. Mostly two-dimensional vortices were studied exhibiting rotational motion. Because of the simple geometry of this problem and the controlability of the vortex, this situation is ideally suited for investigation and has indeed provided fundamental and important new insights. For example, both experimental [3–7] and theoretical [8, 9] FVI studies have confirmed that the domain of the laminar flamelet regime is much larger than suggested by the Klimov–Williams criterion on the Borghi This paper is dedicated to Professor Dr. Jürgen Wolfrum on the occasion of his 60th birthday. diagram. Ashurst and McMurtry [10] confirmed that the vortex strength may be increased by the flame from FVI studies. Microelectrostatic probe measurements of FVI [11] have shown that the scale of flame wrinkling is much larger than the Kolmogorov scale, consistent with the earlier hypothesis that it is mostly large eddies that wrinkle the flame [12]. For practical turbulent combustion situations these FVI studies are far too idealized systems since here the problem is always three-dimensional in nature and characterized by a large number of lengthand time-scales. Vortex motion is random and includes translational as well as rotational components. The objective of the present work was to study flameeddy interactions in a more realistic three-dimensional flame situation, using both advanced laser diagnostic imaging techniques and three-dimensional large eddy simulation (LES). The object of study was the growth of a premixed flame kernel subjected to controlled degrees of turbulence. Methane air mixtures were subjected to turbulence in a specially designed ignition cell [13] in which four high-speed fans were incorporated. The growth of the flame kernel in this system can be characterized by three phases: an initially laminar kernel moving in a turbulent flow field; subsequent growth with increasing degrees of flame wrinkling; and finally the development to a fully turbulent flame structure. In a related paper, flame initiation by the spark discharge in the cell was investigated both experimentally and theoretically [14]. In [14] high-speed spectroscopic imaging data was compared to a full direct numerical simulation (DNS) in two dimensions including detailed chemistry. At this early stage the chemistry is very complex (because of the high temperature peak and gradients during the discharge) although the flame remains small and its topology rather simple (laminarlike). At a later stage the situation becomes too complex and three-dimensional in nature to allow such an approach. Here, we present the results using a three-dimensional large eddy simulation (LES) into which chemistry is incorporated using the flamelet-G concept [12, 15]. The LES framework allows us to address not only the turbulent regime, but also the earlier transitional phase. We start the calculation es-