Large-Eddy Simulation of Turbulent Flames with Radiation Heat Transfer

Open Access
Gupta, Ankur
Graduate Program:
Mechanical Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
August 25, 2011
Committee Members:
  • Daniel Connell Haworth, Dissertation Advisor
  • Daniel Connell Haworth, Committee Chair
  • Michael F Modest, Committee Chair
  • Stephen R Turns, Committee Member
  • Philip John Morris, Committee Member
  • Filtered-density function
  • Photon Monte Carlo
  • Turbulence-radiation interaction
  • Large-eddy simulation
  • Combustion
  • Thermal radiation
  • Turbulence-chemistry interaction
  • Probability density function
Most practical combustion devices involve turbulent flow and operate at high temperatures. Reliable predictive models for these devices must not only represent each individual physical process (combustion, turbulence and radiation) with high accuracy, but also must capture the highly nonlinear interactions among these processes. In this work, a comprehensive computational tool is developed for numerical modeling of turbulent combustion systems with accurate representation of turbulence, chemistry, turbulence-chemistry interaction (TCI), radiation, and turbulence-radiation interaction (TRI). A hybrid finite-volume (FV)/particle-Monte-Carlo procedure is employed wherein a compressible FV Large-Eddy simulation (LES) formulation with a composition filtered-density function (FDF) method is used to model turbulence-chemistry interactions (TCI) and emission TRI. Nongray radiation and absorption TRI is modeled through a photon Monte Carlo (PMC) method where stochastic schemes are developed for treating thermal radiation in a turbulent flow field characterized by the notional particles of the Lagrangian-FDF method. LES/FDF/PMC computations are computationally highly expensive, and a novel "computational time-map"-based domain-decomposition technique is implemented in this study for effective parallelization of the computational code. A nonpremixed methane/air flame is simulated to demonstrate the accuracy of the code developed here. Since LES is inherently time-dependent, the PMC solution at each time step needs to be reasonably reflective of the instantaneous fields to preserve the transient nature of LES, which might require considering large number of photon bundles. Investigations are made in this work for a wide range of flames to estimate the statistical uncertainties in the PMC solution for various number of photon bundles for an instantaneous LES/FDF snapshot. The time-averaged solution is also compared for different bundle sizes. The effect of thermal radiation appears as a source term in the energy equation, which consists of filtered emission and filtered absorption terms in the LES context. In LES, since only large scales are explicitly resolved, the contribution of subfilter-scale (SFS) fluctuations to filtered emission and absorption terms (referred to as SFS emission TRI and SFS absorption TRI, respectively) need to be modeled. The importance of SFS TRI is assessed here for a wide range of flames. A state-of-the-art, advanced LES-based numerical tool for comprehensive modeling of turbulent reacting flows, encompassing all key processes in detail, has become available for the first time as a result of this work. An effective parallelization scheme is implemented in the code that scales well irrespective of the computational cost for chemistry calculations. Stochastic PMC schemes are devised that are consistent with the notional particle representation of the FDF method. It is estimated that approximately three-to-four photon bundles per grid-cell are sufficient to ensure accurate time-averaged solutions for a wide range of flames (ranging from small, optically thin to relatively-large, optically thick). SFS emission TRI is found to be more important than resolved emission TRI for all flames for a grid resolution that is representative of engineering meshes, whereas SFS absorption TRI is found to be negligible for all flames.