A key challenge in combustion research is to model complex turbulence-chemistry
interactions during extinction and re-ignition of non-premixed
ames. These are
caused when the scalar dissipation, , is above (or below) certain limits such that
the diffusive temperature
ux from the
ame is much larger (or smaller) than chemical
heat-release. is characterised by its small-scale intermittency, i.e. large
uctuations
in are frequent and localised in space. This behaviour severely constrains
its resolvability and has implications for modelling extinction/re-ignition processes.
Scalar eld statistics from the direct numerical simulation (DNS) database of a
spatially-evolving, turbulent jet
ame with multi-step chemistry by Pantano (2004)
are studied using three different approaches, namely analysis of dissipation-spectra,
direct investigation of dissipation contours and spatial filtering of the instantaneous
dissipation signals. Out of these the spatial filtering method is found to be most
suited for capturing the intermittent dissipation length scales and an Reδ-1
scaling is
proposed for 'adequate' -resolution in turbulent jet
ame experiments/simulations.
Furthermore, a Multiple Mapping Conditioning (MMC) approach with two reference
variables is used to model extinction/re-ignition in inhomogeneous turbulent jet
ames. A new sub-model for the convective velocity term is employed that does not
need to presume Gaussian statistics and consistent closures for MMC drift and diffusion
coefficients are derived. Effect of temperature
uctuations on scalar diffusivity
is also accounted for. Joint scalar PDFs, conditional dissipation and conditional
species predictions from MMC, and also conditional species concentrations from
conventional singly- and doubly-conditioned moment closure (CMC1 and CMC2),
are assessed against the Pantano DNS database. CMC1 expectedly over-predicts extinction
and does not capture re-ignition, whereas extinction and re-ignition effects
on species concentrations (including atomic H-radical) are captured satisfactorily
using CMC2 and MMC. However, MMC scores over CMC2 because models for
the joint scalar PDF evolution and conditional dissipation are self-contained in the
former.