[akshaykumar]

I am running a direct numerical simulation of a supersonic transitional wake flow to match experimental results. However, my initial coarse (quasi)DNS results show a significant overprediction of the wake size compared to the experiments. By overpredicting, I mean that the shear layers which merge to form the wake reattach later. Before proceeding with refining the mesh, I wanted to get insights from the community on whether this discrepancy could be due to the insufficient mesh resolution or spanwise domain size.

Here are some specifics of my setup:

Numerical schemes: 4th-order central spatial scheme with TVD RK3 time integration, CFL ~ 0.7
Shock treatment: Near shocks, the scheme reverts to 2nd-order upwind via a Ducros sensor

I’m concerned that:

1) Insufficient spanwise resolution or domain size might be causing the wake overprediction, especially given the transitional nature of the wake flow.

2) The Ducros sensor may be activated in smoother regions of the flow, leading to unintended 2nd-order upwinding there, which could impact wake structure predictions.

Could insufficient spanwise resolution or domain size alone cause an overestimated wake size, or might the sensor's sensitivity be degrading the accuracy in transitional regions? I'd appreciate any insights or suggestions in the simulation setup from those with experience in DNS for transitional wake flows. Thank you!