Wt. Ashurst et Ig. Shepherd, FLAME FRONT CURVATURE DISTRIBUTIONS IN A TURBULENT PREMIXED FLAME ZONE, Combustion science and technology, 124(1-6), 1997, pp. 115-144
Distributions of flame front curvature obtained by laser sheet tomogra
phy agree with those derived from numerical simulations of passive fla
me propagation within three-dimensional Navier-Stokes turbulence. The
experimental configuration is that of grid turbulence impinging upon a
plate which stabilizes a premixed methane/air flame, planar images of
the flame allow construction of flame curvature as a function of flam
e location within the spatial zone that contains products and reactant
s. In the simulations the flame burning velocity is twice the turbulen
ce intensity and the Reynolds number based on the computed Taylor leng
th scale is approximately 55. The computed flame geometry and flame st
rain rate are obtained as a function of location based on the mean pro
gress variable (defined by the passive surface displacement or by the
scalar fluctuations defined over transverse planes). The shape of the
mean progress variable profile compares well with experiment and with
two reaction-diffusion models of propagation (KPP and an independent G
aussian model). From the simulations planar slices are created in orde
r to provide curvature information which is directly comparable to the
experimental data. Distributions of curvature, based on planar inform
ation, exhibit a change with location in the turbulent flame zone: an
overall positive curvature (convex to the reactants) at the front to a
negative value at the rear. however, this behavior is composed of pos
itive curvature (which by itself has an average value with no spatial
variation) and negative curvature (which increases in magnitude with d
istance from the front). A single length scale allows a good match bet
ween experimental and computed curvature throughout;he flame zone. The
passive flame simulations show the most probable flame shape to be cy
lindrical, and this feature, allows the planar information to be scale
d in order to match the curvature distributions based on three dimensi
onal information. The scaling factor is obtained by observing a cylind
er with planar slices at all possible angles.