Fluids

Fluids

Topics related to Fluent, CFX, Turbogrid and more.

RPI Model Diverges at high superheats

    • zain.qazi
      Subscriber

      Hello.

      My simulation is a channel flow for subcooled boiling. I am using the standard RPI model and have tried it with Tolubinski-konstanchuk dor Departure Diameter and Site Desnity and also with Kocamostafa. Kocamostafa predicts a lower volume fraction in the domain and converges well to a certain wall superheat. But instantly diverges when moving from applied heatflux of 4.5  to 5 MW/m^2.

      Tolubinski predicts a much higher volume fraction but has trouble converging at these heat fluxes and does not aloow further increase. I see that there is dependancy on the force models chosen. Ishii-Zuber predicts that increase in heat transfer and drop in wall temperature due to boiling phenomenon very well in conjuction with Antal and Burns Wall Lubrication and Turb Dispersion. This effect is not seen while using Kocamustafa.

      Also switching to Schiller Naumann, massively stabilizes the simulation but also loses this cooling effect that I am looking for trying to predict.

      So my preferred choice is Ishii-Zuber but the simulations diverge as I said earlier. Any insights and tips on how to work with this?

    • Rob
      Forum Moderator

      Check time step and mesh quality: they're typically good starting points. The RPI models try and dump all of the vapour into the near wall cell so source terms can be large, and that causes stability problems. 

    • zain.qazi
      Subscriber

       

      The mesh is quite coarse near the wall. The y+ for the vapor phase starts at 30 and goes upto 120, increasing along the channel. The domain is a straight channel with solid and fluid zones with a hexahedral mesh, uniform, no inflation layers, aspect ratios very close to one in the cross section. 

      I’m using Psuedo Time Step Method – Global Time Step and the Time Step Method has been set to Automatic with a Time Scale Factr of 1. This is a steady state simulation. This method works okay till high heat is introduced as I stated above. Why do you think this happens and how do you advise me to change it?

       

    • zain.qazi
      Subscriber

      And also, switching to Schiller Naumann would relieve us of this problem but will also show lower wall lubrication leading to wall temperature rising sharply at the later stages of the channel. My interpretation is that Schiller Naumann applies a smaller drag force as compared to Ishii-Zuber or Tomiyama and that prevents it from actively removing the bubbles from the near wall region, something that Tomiyama and Ishii-Zuber do very well, showing a decrease in wall temperature. I'm not sure what strategy to employ to get Ishii-Zuber or Tomiyama to function at high heat fluxes

    • Rob
      Forum Moderator

      Mesh - good. I see far too many high aspect ratio cells as everyone takes the y+ bit of their CFD training far too seriously and never considers the resulting cell shape. 

      Drop the Time Scale Factor. I suspect "something" ie boiling is occurring too quickly and that's causing problems for the solver. Whilst the RPI models are designed for steady state, the resulting multiphase flow is often very transient. 

Viewing 4 reply threads
  • You must be logged in to reply to this topic.