A multi-resolution analysis (MRA) technique in which coupled sequences of LES simulations are solved on successively coarsened meshes is used to analyze the relative performance of several algebraic partially-stirred reactor (PaSR) subgrid models for turbulent combustion. The test case is a methane-hydrogen bluff-body flame, and the ‘truth solution’ is obtained from a fine-mesh LES that resolves the flow into the dissipative scales. The velocity field from this solution drives the solutions obtained on the underlying coarser meshes, enabling a tight correlation of eddy structures. Conditional distributions of chemical production rate norms and local heat release over a subgrid Reynolds-number / Damköhler sample space are extracted from each coarse-mesh level. The results show that none of the tested PaSR methods accurately predicts the combustion characteristics evidenced from the finest mesh; results improve when coupling with an ‘outer environment’ is included. Optimal forms for the characteristic PaSR time scale and fine-scale volume fraction are regressed from the data and can be modeled effectively using resolved scale information. The use of these forms provides improved results and yields a level of mesh-independence that is not found in the original algebraic PaSR models.
Read full abstract