Digital filtering inflow turbulence generation and artificial shocks
 

Greetings,

I am using the digital filtering technique proposed by Touber & Sandham (2009) for inflow turbulence generation for a supersonic turbulent boundary layer over an isothermal wall under the WRLES framework.

However, I've encountered a shock-like structure right at the leading edge. The figure below is a time-averaged pressure field.

Attachment 85945

In the figure, one can see a shock-like structure standing there right near the inflow plane. This artificial shock creates huge pressure fluctuations which survive all the way to the downstream, causing the prediction of pressure fluctuations very poor compared to the corresponding DNS data. I couldn't find a way to mitigate this structure and to reduce the pressure fluctuations.


I am wondering has anyone ever used this inflow technique before, and encountered a similar issue?

Or any suggestions/thoughts?

Thanks ahead.

Yes, I have seen the same issues with a turbulent inflow using a filtering technique for supersonic flow.


The issues are mainly caused by three factors:

• The averaged velocity profile you are using does most likely not match the "correct" one very well. Correct here means for the Re_tau and Re_theta you are setting at your inlet
• The basic method of Touber & Sandham does not produce a realizable turbulent flow field. This is why some induction distance is needed to produce realistic turbulence and is the main cause for the increase pressure fluctuations (not the shock).
• You are probably using a too simple Reynolds analogy to calculate the thermodynamic fluctuations from the velocity field. A higher-quality method might improve your results somewhat, especially if you are using strongly heated or cooled walls.

I agree with the first one.

But from what I observed, the pressure fluctuations is actually strongly coupled with velocity fluctuations (not surprising actually), which is directly related to the Reynolds stresses provided. Somehow I found the prescribed integral length scales do not have that significant impact on the fluctuations than the Reynolds stresses. Anyway, I tried to "artificially" reduce the magnitude of the Reynolds stresses, and the pressure fluctuations did reduce as well. However, I've observed a shifting in overall pressure magnitude as the flow develops downstream. Don't know why.

I use profiles from DNS for the thermodynamic quantities. Not sure using some other relations (e.g., Waltz relation) would give improvement.