Abstract
A method for LES of ternary mixing in turbulent flows is presented and partly validated against DNS data. The subgrid-scale mixing state is characterized by joint presumed discrete distributions, i.e. discrete ensembles which approximate multivariate FDFs. A closure for the transport equations of the first and second order moments of mixture fraction fluctuations, including the covariance, is formulated in the LES context. Biased mixing models are employed to generate particle ensembles with prescribed first and second order moments. Filtered reaction rates, which are computed as an average over the distributions, can in this way be parameterized by the first and second order moments of the mixture fractions. The mixture state taken from a DNS, which has been filtered onto a coarser grid as suitable for LES, is then compared against particle ensembles generated with the mixing model, where good agreement has been found.
