Xue-Song Bai

Detailed soot modeling in turbulent jet diffusion flames

An approach for modeling the interaction between the formation and oxidation of soot, the radiative heat loss, and the flowfield in turbulent jet diffusion flames is presented. These interactions are modeled by the flamelet library approach in the framework of prescribed probability density functions (PDFs). The formation and oxidation of soot is calculated from a detailed chemical soot model. The laminar flamelet concept is applied to model the rates of soot particle inception, soot volume dependent surface growth, and oxidation, as well as species and temperature fields, Radiative heat transfer from the soot particles and the gas-phase species, CO2 and H2O, decreases the peak flame temperature, which in turn influences the flamelet structures. Experiments from Young et al. on a turbulent ethylene diffusion flame are used to validate the modeling approach. The calculated fields of mean mixture fraction, temperature, and soot volume fraction are found to be in agreement with the experimental data. The spatially resolved rates of soot inception, surface growth, and oxidation are presented. The maximum rates of the surface independent production occur in the fuel-rich region at a radial position of about 10 mm. In contrast, the maximum rates of surface growth and oxidation are found on the centerline with the oxidation occurring at a higher location in the flame. The total rate has its maximum on the centerline, whereas soot formation and destruction balances on the slightly rich side near the stoichiometric contour. This shows the strong interaction of soot formation and oxidation in the flame. A sensitivity analysis of the calculated soot volume fraction on different model parameters is presented. The rate coefficient of the heterogeneous surface growth reaction is the most sensitive parameter. A 20% increase of this rate leads to a 65% increase of the calculated maximum soot volume fraction.

Modelling of turbulent reacting flows past a bluff body: assessment of accuracy and efficiency

Abstract The accuracy and efficiency of numerical calculations of turbulent reacting flows past bluff bodies have been studied. The Reynolds-averaged Navier-Stokes equations, together with a two-equation k-ϵ model and a premixed flame model, are solved using finite differences on a system of global and locally refined grids. To improve the calculation efficiency, a multgrid (MG) method is employed. The influence of the grid resolution on the solution accuracy and the effects of the model parameters, boundary conditions and different reaction mechanisms are studied numerically. Comparisons are made between predictions and measurements of propane-air reacting flow past a bluff body placed in a wind tunnel. The results show that the computational time can be reduced considerably using an MG method, and the numerical accuracy can be maintained economically by using a local grid refinement technique. With proper model parameters, the solution of the simplified mathematical models is in good agreement with the experimental data.

A multi-grid method for calculation of turbulent and combustion

The application of the Multi-Grid (MG) method to the calculation of turbulent reacting flows is considered. Turbulence is handled by using the k - e model. The eddy-dissipation concept based on a reduced global chemical reaction scheme is used for modeling the chemical reactions. For low Reynolds number laminar flows the MG efficiency is best, with the convergence rate in the order of 0.8. For uniformly spaced grids the convergence rate can be better, and for highly skewed grid, slower. The introduction of turbulence and combustion generally slows down the converging process. However, the MG method still demonstrates considerable acceleration over the single grid solver.

On Modeling Aspects of Swirling Stabilized Diffusion Flames

Modeling of swirling stabilized diffusion flames in a cylinder combustor is considered. Two-equation turbulence models are inherently not well suited to handle swirling turbulent flows. We propose an LES based solver for the computation of the reacting flow fields. LES data is used to study the influence of inlet swirling motion on the combustion and the turbulent mixing process. PDF is extracted from the LES data. The shape of PDF profiles is similar to a ft—distribution. PDF distributions at different locations indicate that mixing of fuel and air is improved by increase of inlet swirl angle. Spectral analysis of different terms of the turbulent kinetic energy balance equation shows that the contribution of the heat release to turbulent kinetic energy balance decreases with the frequencies. We also used the LES data to assess the appropriateness of conventional eddy viscosity based gradient-diffusion models for accounting for the density-turbulence and temperature-turbulence correlations.

Large Eddy Simulation of Turbulent Combustion Using Cartesian Grid and Boundary Corrections

A high order Cartesian grid method is developed and applied to large eddy simulations (LES) of isothermal and reactive turbulent flows. Cartesian grid methods are potentially the most suitable numerical methods for LES because of its economical storage requirement, less computational effort per computational cell, fast numerical convergence and feasibility for constructing higher order finite difference schemes. An uniform Cartesian grid with high order interpolations is employed to handle the curved walls. The small cell limitation on the time steps is remedied by using an implicit scheme based on an efficient Multi-Grid (MG) acceleration method and local grid refinement technique. Application of the numerical method to LES of isothermal and reactive flows shows that the low order piecewise constant wall approximation affects the energy flux in the cascade near the walls. This affects the overall turbulence level and even affects the mean of the whole flow field. Numerical experiments show that 3rd order Lagrangian interpolation near the walls results in a stable LES, while higher order interpolations may trigger numerical instability. (Less)