Fluids

Fluids

Swirling flow in annular channel

    • cyw400
      Subscriber

      Hi,

      I am working on simulating a swirling flow through a annular channel then enter to a combustion chamber using k-w SST with curvature correction, steady state. The computational domain is 60 degrees sector with peridoic boundary. The swirl is represented by axial velocity component and tangential velocity component. 

      The results of axial velocity profile near the exit of the annular channel does not show any recirculation but the magnitude value match with experimental data. 

      The swirler number is 1.2 and from the user guide, RSM is recommended. The epsilon-based RSM and w-based RSM do not provide the profile to match the experimental data as well.

      Could someone please adivse on how can I resolve this and get the recirculation near the outlet of the annular duct? Many thanks

    • Prashanth
      Ansys Employee

      You can use k-omega to model turb. as directed in the literature. And btw does the periodic domain allow the use of hexa mesh near the wake region/recirculation zone/nozzle expansion region ? Also did you make sure the expansion chamber/domain is big enough allowing the gas to fully expand.

    • cyw400
      Subscriber

      Hi Prashanth,

      Thank you for your reply. I have used watertight workflow to genetate polyhedral mesh with 5 boundary layers. The mesh size is 1mm in the wake region and recirculation zone. The computational domain is 10D long and the gas is fully expanded. 

      I have then re-generated poly-hex core mesh and the results were similar to the polyhedral mesh. 

      Could you please advise that is there a meshing problem or something else? Many thanks

    • Prashanth
      Ansys Employee

      Hey

      You can read up on numerical diffusion here. I think adding boundary layers won't help much. As you are trying to capture recirculation zones, you'd be better off using structured hexa mesh for your model. You can decompose the domain into multiple blocks and obtain it. You can also try an axis-symmetric case (if it's applicable for your geometry) before moving on to 3d periodic. Also, you'll find more similar numerical modeling work in journals (use keywords like "gas atomizer, atomization nozzle") to get an idea of how people approach this kind of problem.

      Can also check out this flow in a nozzle on Ansys Innovation Courses (open source) and transient flow (this one uses poly mesh). Not sure if it'll help for your geometry, but check it out anyway.

    • cyw400
      Subscriber

      Hi Prashanth

      Thank you for your advice. 

      I have tried 2D axis-symmetric swirl with hexa mesh before. For numerics, I have tried using PRESTO! for pressure and second order for the others. I am following a literature which shows unstructured mesh approach can solve the recirculation in pair with SAS. However, I do not have such computational power to solve in transient case and I believe the unstrcutured mesh approach work for steady-state RANS. 

      Could you please advise futher on this topic? Many thanks

    • Rob
      Ansys Employee

      For the outlet use "from neighbouring cell" for backflow direction. This can help as it'll stop the reverse flow damping any swirl: ideally you should extend the domain to remove the backflow. 

      Otherwise, pictures showing the flowfield, mesh and what you're expecting to see would be helpful. Add into a post as staff are not permitted to open attachments. 

    • cyw400
      Subscriber

      Hi Rob,

      Thank you for your advice. 

      I have tried different outlet setup (from neighbouring cell, prevent reserve flow) and the results still do not capture the recirculation. 

      My domain is 2m long and the burner is 500mm. 15 degrees sector with perodic boundary was used and 0.5mm poly mesh for the recirculation.

      I am excepting the recirculation shown in the red line in the axial velocity xy plot below.

      I have tried to investigate the problem by different mesh size, sector degrees and numerics and none of them would give any changes in result.

      Could you please advise if there something I did wrong? Many thanks

    • Rob
      Ansys Employee

      The flow looks reasonable. What is inlet condition? You also have a relatively long inlet channel, most swirl burners have the vanes close to the flame region and may have 2-3 inlet parts with varying swirl. The latter will be in a text book somewhere, look up gas turbine injectors (liquid fuel). 

    • cyw400
      Subscriber

      Hi Rob,

      Thank you for your suggestion.

      I have used axial velocity and tangential velocity components for velocity inlet boundary condition. The annular channel is 50mm which is suggested from literature to let the circumferrential variation of the axial velocity become axisymmetry. 

      I have found a literature which used the same approach for inlet boundary condition in LES but I cannot find any RANS literatures to report this problem. 

      I have tried inlet boundary condition senstitivity study to reduce the annular channel to 3mm and remove the annular channel. 

      I could not think of any possible approaches to solve this problem. Could you please advise more on this? Many thanks

    • Rob
      Ansys Employee

      Just checking, you did fix the swirl axis to align to the x-axis? I think default is z. 

    • cyw400
      Subscriber

      Hi Rob, Yes, I did define the rotational axis align to x-axis. Many thanks

    • Rob
      Ansys Employee

      Not sure then. Zoom in on the inlet region and investigate the flow, vectors might be a good option. 

Viewing 11 reply threads
  • You must be logged in to reply to this topic.