Large eddy simulations of turbulence -chemistry -radiation interactions in diffusion flames

Abhilash J Chandy, Purdue University

Abstract

The efficiency and pollutant emission characteristics of practical combustion devices often depend critically on interactions between turbulent flow, finite-rate combustion chemistry, and thermal radiation from combustion products and soot. Due to the complex nonlinear coupling of these phenomena, modeling and/or simulation of practical combustors or even laboratory flames undergoing significant extinction and reignition or strong soot formation remain elusive. Methods based on the determination of the probability density function (PDF) of the joint thermochemical scalar variables are one of the most promising approaches for handling turbulence-chemistry-radiation interactions in flames. PDF methods have gained wide acceptance in the context of Reynolds-Averaged Navier-Stokes (RANS) approaches to predicting mean flowfields as evidenced by their availability in commercial CFD codes such as FLUENT™. Over the past 6 years, the development and application of the filtered mass density function (FMDF) approach in the context of large eddy simulations (LES) of turbulent flames has gained considerable ground. Some of the key issues remaining to be explored regarding the FMDF approach in LES are related to mixing model and chemical mechanism sensitivities of predicted flame statistics, especially for flames undergoing significant extinction and reignition, and application of the approach to more realistic flames, for example, those involving soot formation and luminous thermal radiation. In this study, we explore the issue of mixing model sensitivity, as well as the role of the presumed constant (independent of chemistry and species) mixing frequency, for several laboratory and idealized piloted turbulent diffusion flames at different Reynolds numbers and hence, different levels of local flame extinction/reignition. The laboratory flames are modeled after the Sandia Turbulent Nonpremixed Flames D, E, and F and are predicted using a RANS/PDF transport model in FLUENT. The idealized flames are simulated using an in-house LES/FMDF code modified to allow different mixing models. In addition, we extend the in-house LES/FMDF code to include luminous thermal radiation from a flamelet soot model, and conduct simulations of idealized strongly radiating turbulent flames. A new parallel radiation solver, employing the discrete ordinates method (DOM), is developed and tested as part of this effort. Our findings from both studies confirm that the level of local extinction/reignition predicted in the flame is sensitive to the choice of mixing model and the mixing frequency and suggest the use of a variable mixing frequency could improve the models. Also, our idealized strongly radiating flame studies demonstrate the utility of the LES/FMDF approach for such flames, highlight the importance of turbulence-radiation interactions, and pave the way for the inclusion of finite-rate soot transport and kinetics models and quantitative prediction of laboratory scale sooting flames in the future.

Degree

Ph.D.

Advisors

Frankel, Purdue University.

Subject Area

Mechanical engineering|Fluid dynamics|Gases

COinS