Date of Completion
Field of Study
Doctor of Philosophy
Three approaches toward efficient and predictive turbulent combustion modeling are investigated in this dissertation. The first approach focuses on the development of locally reduced chemistry and advanced solvers for efficient time integration of stiff chemical kinetic systems. In particular, a numerical technique using dynamic adaptive chemistry (DAC) with splitting schemes is developed and demonstrated in one-dimensional (1-D) premixed flames. A sparse stiff chemistry solver based on dynamic adaptive hybrid integration (AHI) and sparse matrix techniques (AHI-S), and an iterative uncoupled quasi-steady-state (IU-QSS) method for improved stability of explicit solvers, are further developed and shown to be more computationally efficient than other chemistry solvers in various flame configurations. In the second approach, a computational diagnostic tool, namely the chemical explosive mode analysis (CEMA), is extended to account for the interactions between chemical reactions and transport processes. Different local combustion modes, including the auto-ignition, diffusion-assisted ignition, and extinction modes, are demarcated by projecting the chemical and transport source terms to the chemical explosive mode. A criterion based on the local combustion modes is proposed to distinguish between two premixed flame propagation modes, that is the auto-ignition and diffusion-controlled deflagration waves, respectively. The new criterion is validated in 1-D premixed flames and 2-D homogeneous charge compression ignition (HCCI) systems. CEMA-based diagnostics are then employed to investigate the local structures of strongly turbulent premixed n-dodecane flames, and to understand the propagation modes and stabilization mechanisms of a turbulent lifted dimethyl ether (DME) jet flame, based on direct numerical simulation (DNS) data. The third approach is to construct a dynamic adaptive combustion modeling framework for turbulent flames that involve both premixed and non-premixed features. CEMA is adopted as a flame segmentation tool, and appropriate sub-models are assigned on-the-fly to different flame zones. The proposed modeling framework is tested in a turbulent lifted n-dodecane spray flame using large eddy simulations (LES). The new model is found to predict the ignition delay and lift-off length more accurately compared with the low-cost flamelet models, while the overall computational cost can be substantially reduced compared with the high-cost regime-independent models that incorporate finite rate chemistry.
Xu, Chao, "Advanced Chemistry Solver Development and Computational Diagnostics and Dynamic Adaptive Modeling of Turbulent Combustion" (2018). Doctoral Dissertations. 1836.