Numerical studies of turbulence effects on developing flames in lean methane/air homogeneous mixtures

Harinath Reddy, Purdue University

Abstract

Igniting the fuel reliably in lean-burn internal combustion engines is a challenge because the ignition kernel is susceptible to extinction. This work is a numerical study of the various factors affecting flame development from an ignition kernel in a natural gas-fueled engine. Methane is used as a surrogate for natural gas. Studies of kernel development in an initially quiescent mixture show that the initial kernel diameter has to be larger than a value determined by the thermal diffusion length scale in order to sustain a flame. The minimum size is smaller in spark-ignited engines where the kernel temperature is determined by the energy deposited during spark-ignition relative to hot-gas jet ignited engines where kernel temperature is determined by the adiabatic flame temperature of the mixture. Studies of the interaction of ignition kernels with vortices, employing length and velocity scales representative of engine conditions, reveal that there are two distinct kernel development regimes. A kernel breakthrough regime, characterized by local extinction and formation of secondary kernels, is observed with relatively smaller and faster vortices. A kernel wrinkling regime, where the kernel is stretched, deformed, and advected, occurs with larger and slower vortices. The rate of increase in surface area is greater in the wrinkling regime. Conditions which promote kernel development in the wrinkling regime will result in higher heat release rates and greater thermal efficiency in engines. Turbulence-kernel interaction studies reveal that the ratio λ of the integral length scale to the kernel diameter influences the outcome. For λ greater than 1, the kernel is initially convected by the turbulent flow. When λ is less than 1, if the ratio of turbulence intensity to laminar flame speed (u'/sL) is greater than 5, kernel breakup is observed. For lower values, kernel wrinkling is observed. Evaluation of the DNS data show that it takes approximately four eddy turn-over times for the kernel to become fully developed. The accuracy of closure models for different terms in the flame surface density (FSD) combustion sub-model for large-eddy simulations is evaluated employing the DNS data. It is observed that the sub-grid convection term has the largest modeling errors for smaller filter widths in the thin reaction zone regime whereas the strain rate term has the largest error for larger filter widths and/or kernel wrinkling regime. In addition, modeling errors in curvature term are large during the early evolution of the ignition kernel.

Degree

Ph.D.

Advisors

Abraham, Purdue University.

Subject Area

Mechanical engineering

COinS