A multi-scale LES technique for coupling near-field and far-field domains for a jet flow

Abstract

When using Large Eddy Simulations (LES) for a jet flow in a domain which is more than a hundred jet diameters long, the computational requirements are dictated by the near-field region immediately downstream of the jet exit. This is because high spatial and temporal resolution is required to accurately transition the flow to a fully-developed state. In particular, the high flow velocities combined with small grid-scales required for the thin shear layers in this region necessitate a time-step which can be an order of magnitude smaller than that required for the downstream region. To eliminate this near-field time-step restriction, a Recorded Interface Boundary Condition (RIBC) technique is developed herein to represent the unsteady three-dimensional flow emanating from the near-field region. The RIBC technique prescribes an upstream boundary condition for the far-field domain to allow a much larger time-step, reducing CPU resources by an order of magnitude. Two variants of the multi-scale RIBC are considered: one is a recording that plays only forwards and repeats over several cycles, while the other is a similar recording but plays both forward and then backwards (so as to eliminate temporal discontinuities). To evaluate the performance of this method, these variants were compared to a full-field simulation (which encompasses all time and length scales of the near-field and the far-field regions). Despite its simplicity, the RIBC technique reasonably reproduces the mean and turbulent velocities in the far-field jet without requiring the small time steps and high resolution domain associated with the near-field. This allows the far-field domain to be simulated with larger time steps than the full-field simulation reducing computational time for a converged solution. Of the two variants considered, the forward-and-backward technique tends to be more stable but more work is needed to determine the robustness for other flows (e.g. wakes and shear layers), and to investigate the potential non-physical influence of the recording cycle period on the flowfield turbulent spectrum.