Learning Forum

[Marius Carrier]

Hello Ansys Team,

I am simulating airflow through a labyrinth seal using Ansys Fluent. I am utilizing the k-omega-sst model and defining the outlets as pressure outlets. I have observed that my residuals start off well but then spike and remain constant (see below). This is just one example, as I have encountered several similar cases. The spikes occur between 100 and 120. Based on my readings of various discussions in this forum, I understand that residuals do not necessarily need to reach 10^-5, and instead, convergence definition can be used to monitor convergence. However, for the purpose of my bachelor thesis, it would be preferable to have an official recommendation from Ansys, such as those found in the Ansys User and Theory Guide. Unfortunately, these sources only suggest running the simulation until convergence. Could you provide any tips on where I might find an official recommendation from Ansys? Additionally, could the convergence issues be related to the following factors:
- I have only included boundary layers in the labyrinth seal to save time.
- There is reversed flow at the pressure outlet.

 

Thank you for your help.

Regards,
Marius

[Rob]

I know there's something about that in the training materials as I've been involved in writing/revising/teaching/reviewing the content at various times. However, I don't think there's anything in the formal documentation. You may want to check the Fluent course on here (look in Learning) as I'm told it's very similar to the course I am used to. 

Back flow can have an adverse effect on residuals, check it's not reaching any of the areas of interest in your model. 

[Marius Carrier]

Thank you,

do you have any recomendations, on how to lower the omega and k residuals?

 

 

[Rob]

As a guess, alter the outlet backflow to "from neighbouring cell" and check the backflow values are sensible. Pictures will help. 

[Marius Carrier]

Hey,

Thank you for your help.

Pictures will be available soon.

Could the high k and omega residuals be due to the fact that I only have boundarie layers at a small part of the mesh, which I did to save time later on?

[Rob]

Possibly, but if you plot the contours you can overlay mesh and/or use node values (contour panel) to see. Hence wanting pictures!