Large eddy simulation of combustion using linear-eddy subgrid modeling

Doktorsavhandling, 2019

As emissions regulations are getting stricter and efficiency requirements of internal combustion engines (ICE) are increasing, different concepts to improve combustion are being investigated. For example lean stratified premixed combustion, homogeneous charge compression ignition (HCCI), use of more exhaust gas recirculation (EGR) to reduce NOx etc. In all these concepts, combustion happens at lower temperatures, higher pressures, and higher level of air dilution than today's typical spark ignition or diesel engines.
Many combustion models in computational fluid dynamics (CFD) today describe either premixed or non-premixed mode of combustion, assuming fast chemistry regimes only. There is a great need for new combustion models that are mode (premixed/non-premixed) and regime (fast/non fast chemistry) independent. The linear-eddy model (LEM) of Kerstein used as a subgrid combustion model for large eddy simulation (LES) called LES-LEM is regarded as a truly mode and regime independent combustion model as it models all the physical processes, i.e. large and small scale turbulent advection, molecular diffusion and chemical reactions at their respective length and time scales.
In this dissertation, a new LEM closure for LES-LEM using the reaction-rate approach is proposed in which the LEM provides closure for the chemical source terms in the conservation equations of the sensible enthalpy and species mass. The new LEM closure is tested on a bluff-body premixed flame problem and simulation results are compared with experiments.
Furthermore, a new splicing approach for modeling large-scale advection in LES-LEM is presented. The approach links the subgrid LEM implementation
to a concept of control-volume crossing rate. A dedicated investigation of splicing is done by simulating passive scalar mixing without the complexity of
chemically reacting flow physics.
Lastly, an improved modeling technique called super-grid LES-LEM is proposed to computationally speed-up LES-LEM.