Fluids

Fluids

Wild Oscillations in Cp and Cf?

    • vedama
      Subscriber

      Hello,

      I'm using adaption to refine an existing solution but my final result plots along the wall oscillate wildly. They're oscillating around the right answer (from other papers etc.) but they seem highly unsteady. See images below for Cp and Cf along the wall. It's a 2D transonic flow case, and I'm using second-order Upwind scheme with density-based solver. I've tried adjusting Courant number from 0.1 to 5 and that didn't seem to help. I've run over 100,000 iterations per adaption iteration (10 iterations total) so it should be converged. The oscillations actually get progressively worse over each adaption iteration. I've tried the pressure-based solver with velocity coupling, didn't help. I tried third order MUSCL, also didn't fix it.

      I've really exhausted a lot of options and I'm still unsure if it's a problem with the numerics or the grid or just something fundamentally wrong. I looked on other forums and some people have faced a similar issue but without a solution. I'm hoping someone here can help fix this, please? Or at least point to maybe where I'm going wrong.

    • vedama
      Subscriber
      It seems part of this issue lies in the hanging-node adaption for 2D grids, which Fluent is phasing out (according to the latest release). Extruding a 2D grid by one cell and using the PUMA (3D) adaption seems to have fixed this issue (see images below)
      Curious as to why this is happening. I am interested in the math behind it. PUMA has 4:1 face interactions just like hanging-node has 2:1 which I suspected was the main issue. But I seem to be wrong in my hypothesis. Any explanations? Or papers I can refer to?


    • Karthik R
      Administrator

      Thank you for posting a solution to this. This is indeed a creative solution. PUMA is always going to be more robust than a simulation on a hanging node adaption. This is intimately connected to the algorithm. The biggest advantage with PUMA is that it does not create any hanging nodes and this is very critical for problems like the one you are solving.
      Karthik
    • vedama
      Subscriber
      Thanks, Karthik. Do you have any resources on PUMA other than the Fluent manual, by any chance? I'm looking at my grid using PUMA and it still has "hanging nodes" in the sense that there are nodes not attached to two separate cells. How does this work any different in 3D than in 2D? I am curious about the math behind it, especially given the 4:1 cell faces interactions.
Viewing 3 reply threads
  • You must be logged in to reply to this topic.