Influence of thermal radiation on soot production in Laminar axisymmetric diffusion flames

Abstract

The aim of this paper is to study the effect of radiative heat transfer on soot production in laminar axisymmetric diffusion flames. Twenty-four C1–C3 hydrocarbon–air flames, consisting of normal (NDF) and inverse (IDF) diffusion flames at both normal gravity (1g) and microgravity (0g), and covering a wide range of conditions affecting radiative heat transfer, were simulated. The numerical model is based on the Steady Laminar Flamelet (SLF) model, a semi-empirical two-equation acetylene/benzene based soot model and the Statistical Narrow Band Correlated K (SNBCK) model coupled to the Finite Volume Method (FVM) to compute thermal radiation. Predictions relative to velocity, temperature, soot volume fraction and radiative losses are on the whole in good agreement with the available experimental data. Model results show that, for all the flames considered, thermal radiation is a crucial process with a view to providing accurate predictions for temperatures and soot concentrations. It becomes increasingly significant from IDFs to NDFs and its influence is much greater as gravity is reduced. The radiative contribution of gas prevails in the weakly-sooting IDFs and in the methane and ethane NDFs, whereas soot radiation dominates in the other flames. However, both contributions are significant in all cases, with the exception of the 1g IDFs investigated where soot radiation can be ignored. The optically-thin approximation (OTA) was also tested and found to be applicable as long as the optical thickness, based on flame radius and Planck mean absorption coefficient, is less than 0.05. The OTA is reasonable for the IDFs and for most of the 1g NDFs, but it fails to predict the radiative heat transfer for the 0g NDFs. The accuracy of radiative-property models was then assessed in the latter cases. Simulations show that the gray approximation can be applied to soot but not to combustion gases. Both the non-gray and gray soot versions of the Full Spectrum Correlated k (FSCK) model can be then substituted for the SNBCK with a reduction in CPU time by a factor of about 20 in the latter case.