Development of correlated and computational methods for predicting premixed turbulent flame speed

Abstract

Turbulent flame speed is an important property within turbulent flames and an invaluable parameter for gas turbine combustor design. A major issue however is the difficulty in accurately predicting turbulent flame speed due to the complexity of reactive turbulent flow. The aim of this research is to: (1) carry out an extensive study on literature analytical turbulent flame speed correlation, (2) develop a new empirical correlation, and (3) develop of a one-dimensional (1D) freely propagating turbulent flame speed model implemented in Cantera.

For the correlation study 15 state-of-the-art premixed turbulent flame speed correlations from the literature are examined in three studies to provide insight into each expression. A newly-developed empirical correlation is also introduced and assessed alongside the literature correlations. The assessment is carried out by comparing each expression against experimental data and determining accuracy using a mean absolute percentage error (MAPE). The combined findings of the three studies shows that a minimum of two correlations and two sets of adjustable parameters are required to accurately account for the entire range of data in the study and shows that there is currently no general correlation.

A predictive model is therefore developed based on the existing Cantera solver. To solve for turbulent flame speed, significant additions are made to Cantera to include: (1) the gradient transport model for the temperature and species transport equations to account for increase transport, (2) a Taylor series expansion of the reaction rate around the mean temperature to model the effects of temperature fluctuation on the Arrhenius rate, and (3) the eddy dissipation concept model with finite rate chemistry to define the influence of the turbulent-chemistry interactions. Validation shows a very small difference between the MAPE of the most accurate correlation of Muppala, (38.1%) and the predictive model (41.3 %). For gas turbine relevant conditions, it is found that the solver improved on existing empirical approaches, providing a smaller error. A series of trend studies showed that the model predicted the trends well over a wide range of conditions.