Fluids

Fluids

Topics related to Fluent, CFX, Turbogrid and more.

CFD Simulation: Reverse Flow Issue at Outlet2

TAGGED: , ,

    • jingyuliu
      Subscriber

      I’m currently running a CFD simulation in ANSYS Fluent with the k-epsilon turbulence model for a steady-state solution. My model geometry consists of one inlet surface and two outlet surfaces. I’ve set up boundary layers with 5 and 7 layers, respectively. However, during the simulation, I keep encountering a "reverse flow" warning at Outlet2.

      Even though the model eventually reaches stability, this reverse flow issue significantly delays my computational time. I run the LES model once the steady-state solution is achieved. This warning "reverse flow" still exists

      My questions are:

      1. Does the reverse flow at Outlet2 matter, especially in terms of solution accuracy or physical relevance?
      2. How can I address or minimize the reverse flow issue to reduce computational delays?

      Any suggestions or advice would be greatly appreciated!

      Thanks!




    • Rob
      Forum Moderator

      It depends. If the outlet is too close to the area of interest then, yes, it may effect the solution accuracy. Avoiding reverse flow is usually done by extending the domain, but may also require some thought about the solved domain. 

    • jingyuliu
      Subscriber

      Thank you for your response. Typically, I use the same geometry but with a coarser mesh, which makes it difficult to encounter such an issue. However, after increasing the mesh density by refining the boundary layers, I encountered the reverse flow problem. I'm considering further refining the mesh in the outlet2 area. Does that seem like a good approach?

      Best regards, 
      Leo

    • jingyuliu
      Subscriber

      a followed question:
      the convergence criteria i set      continuity 1e-5; x-velocity 1e-6, y-velocity 1e-6, z-velocity 1e-6.
      but in the k-epsilon stable simulation, the countinuty cannot decrease to 1e-5. It is stuck in the 1e-3. But the x, y, z velocity is already convergent. Should I decrease the convergence criteria of continuity from 1e-5 to 1e-3 to make the flow stable.

    • Rob
      Forum Moderator

      No. The value you're setting triggers the convergence check, but doesn't alter the solution accuracy (well, assuming the check value is 1e-3 or lower). I could set the check as 10, the result would be useless but the convergence check would trigger as complete. 

      In your case continuity is stuck slightly above the default convergence check value. So, we then look to flux balances (is what enters the domain leaving) and point monitors (is the solution changing) to figure out why the residual is stuck. You may find some very slight changes in the flow with iteration and you then need to decide if the change is within an acceptable tolerance. This is why we don't rely on just residuals and also look at monitors and look at the flow results in detail. 

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