Large Eddy Simulation (LES) coupled with the Linear Eddy Model (LEM) provides a robust method for studying turbulent combustion, but it is computationally expensive due to the need for highly resolved sub-grid LEM domains. These domains simulate sub-grid stirring through stochastic rearrangements of scalar fields, while large-scale transport is modelled using a Lagrangian ‘splicing’ scheme. To address the computational cost of LES-LEM, a super-grid (SG) framework for LEM closure was developed by the authors (Comb. Theor. Model. 28, 2024), which uses coarse-graining, on-the-fly chemistry tabulation and a presumed PDF approach to reconstruct thermochemical fields at LES resolution. This study applies SG-LEM to a challenging setup, Case 1 of the Volvo Validation Rig, which involves a bluff-body-stabilised turbulent premixed propane-air flame, as a stress test to identify limitations that were not revealed by the previous application, in particular that of the coarse-graining parameters used to generate the super-grid. The intent is to yield a more realistically constrained assessment of the current capabilities of the method, and insight into possible ways for improving it. Four simulations were conducted using three SG cluster sizes. The finest resolution was tested with a global 2-step mechanism, showing good agreement with experimental data for temperature and velocity, particularly near the bluff body. The two larger cluster sizes used a 66-step skeletal mechanism for more detailed chemical closure but led to unphysical quenching due to splicing inaccuracies. To mitigate these issues, two novel additions were introduced: an intra-cluster-stirring routine and a method to control SG cluster shapes to reduce numerical dissipation. These methods improved flame stability with coarser SG clusters and more detailed mechanisms. Comparison with experiments showed good agreement for temperature and velocity, though elevated CO levels were observed in the recirculation region. Potential methods for further improving SG-LEM's capabilities are discussed.
Read full abstract