Stochastic aspects of turbulent combustion processes

Abstract

Methods of using stochastic simulations to treat nonlinear interactions in turbulent combustion processes are described -- emphasizing the use of statistical time-series techniques to analyze the turbulence--radiation interactions of nonpremixed flames. Three aspects of the problem are considered, as follows: the statistics of scalar properties in turbulent flames, the formulation of algorithms to stimulate flame radiation based on flame statistics, and evaluation of the methodology using recent measurements for nonluminous flames. It is shown that the process becomes tractable through the laminar flamelet approximation whereby all scalar properties are taken to be solely functions of a conserved scalar like the mixture fraction. Thus, the simulations are designed to generate realizations of mixtures fractions along radiations path with the radiation properties of each realization found using a narrow-bond radiation model. An autoregressive process that reproduces probability density functions and spatial and temporal correlations of mixture fraction was found to yield reasonably good predictions of the statistical properties of spectral radiation intensities measured for turbulent carbon monoxide and hydrogen jet flames burning in still air. Although the approach appears to be promising, additional development is needed in order to treat some of the unique statistical features of turbulence that are not encountered during conventional use of statistical time-series techniques.