Fluids

Fluids

How to edit the data in .ip (interpolation) file?

    • liangerscu
      Subscriber

      Hey, all. 


       


      I am a student working on a half-domain geometry simulation, using ANSYS FLUENT (the center plane of my geometry is with symmetric boundary condition, i.e. the variable gradient on this plane is 0). I have got the results on this half-domain simulation, and NOW, what I want to do is to use the results from this half-domain simulation as the initial condition for a full-domain simulation, by mirroring the half-domain results and import it into the full-domain case. 


      I am able to export the .ip file for a half domain and tried to mirror the data in it, but cannot find a way to edit the .ip file (Notepad does not work). 


      Any idea on editing .ip file from FLUENT. OR is there a way to create a new .ip file from the scratch? (I actually tried to create my own .ip file, BUT Notepad++ cannot handle a file bigger than 350 mb, and I am stuck). 


       


      Any thoughts would be greatly appreciated! 

    • Rob
      Ansys Employee

      Try with a small file (100 cells maximum) to understand the format. You can write out non-binary so it'll be human readable. No reason why you can't then write your own file, but stitching it together could be difficult: I'll leave you to discuss that with IT. 


      One thought, on your test model, what happens if you write out two interpolation files containing different data and then interpolate them both back onto the model? You may want to initialise everything to zero to make sure you're actually reading in new data. 


      Finally, interpolation only accelerates convergence (in most cases) so you may just be quicker running the model. 

    • liangerscu
      Subscriber

      Thanks for your detailed reply. Thanks for your innovative ideas! I appreciate your insightful comments. 


      1. I think I understand the file format already. The problem is that the file I am trying to write up is too big to be handled by Notepad++, which is upset. 


      2. Thanks, I am looking to talking to IT, how can I discuss with them? Will they reach out to me?


      3. Thanks for your creative idea. I actually have tried to write out two interpolation files and interpolate then both back onto the model. Problem with this method is that FLUENT can only interpolate one file at a time, and the new data will overwrite the old data. So, unfortunately, this method is not feasible. But thank you all the same! 


      4.Yes, I totally agree with you that interpolation only accelerate convergence. In my case, we are having convergence issue for full-domain simulation, but having very good convergence (1e-08) for half-domain simulation, which is why we are striving to interpolate the data from a half-domain case to a full-domain case. 


      Please don't hesitate to reach out when any new though jumping up in your brain. 


      Thanks,


      Mingyi

    • Rob
      Ansys Employee

      3) Not if you have different fields?  You must have the x, y, z location data but try adding pressure to one file & u, v, w velocity to a second & see what happens. 


      2) IT are your end: nothing to do with us! Some of the UNIX text editors can handle larger files but it'll depend on your system. 


      4) If half of the model works and full model doesn't carefully look at the flow field. What does the symmetry plane do that stabilises the flow? 

    • liangerscu
      Subscriber

      Thanks, rwoolhou!


      I solved the problem! The file size is too big because I am using 16 significant figures. Reducing it to 6 does the trick in my case! I now can edit the .ip file from a half-domain simulation and interpolate it onto a full-domain mesh!


      Thanks for your reply!


      2) Thanks for letting me know. I'll try to figure that out.


      4) On the symmetry plane, gradient of all variables are zero. While on a full domain, asymmetrical flow pattern may occur, which means that the flow would wabble around the center plane due to turbulence and causes convergence issue. Technically, if mesh topology and quality are good enough, convergence should not be a concern even on a full-domain mesh!

    • Rob
      Ansys Employee

      Correct, re the symmetry definition. What you may find is the symmetry plane prevents swirl or significant transient eddy shedding. So, 2d and symmetry based models may converge well but full domain may fail. 

Viewing 5 reply threads
  • You must be logged in to reply to this topic.