Fluids

Fluids

Choosing the right time step size and judging convergence in transient simulations

    • MarkusP
      Subscriber

      Dear community members,


      I would really appreciate your advice and experience concerning the following questions:



      • What are best practice guidelines to choose an appropriate time step size in transient simulations (apart from always performing a sensitivity study)?

      • How can I judge convergence within (at the end of) each time step (apart from residuals dropping below a certain value)?


      Thanks a lot in advance!


      Best regards,
      Markus

    • DrAmine
      Ansys Employee

      1/ time step resolving the underlying physics like for natural convection the time step has to be in the order of squre root of the domain extent or cell size divided by thermal gravity. Or a portion of the inverse of angular frequency for rotating machineries..


      2/ judge convergence of each time step by looking into to the evolution of monitor points versus iteration. If the monitor does not change anymore then it is likely that time step has converged. 

    • MarkusP
      Subscriber

      Thanks, Amine, for your answers!


      However, these are very general rules, but applying them to a specific simulation is not always straight forward:


      1.)  How can I find a suitable time step to resolve the underlying physics (e.g. when simulating residence time distribution for a packed bed)?
      And what does "a portion" mean?


      2.) Where should the monitor points be located (e.g. when simulating residence time distribution of a packed bed)?
      And how can I automate this convergence criterion (i.e. how to tell Fluent to stop iterating if "monitor does not change anymore")?


      Your comments are highly appreciated!


      Thanks,
      Markus


       

    • DrAmine
      Ansys Employee

      1/Residence time: I will check first of all the terminal velocity of the particle and or the free fall velocity. Based on that and using a characteristic length scale in the domain one can provide a max.value for the time step size.


      2/You create a report definition for static pressure at certain interior surface in the bed /above the bed or volume average on the cell zone confining the packed bed (as examples). If there is heat transfer then I would rather stick to monitoring thermal quantities, if mass transfer then to the mass transfer rate.. 


       


      In Fluent there is an option to use the created report definitions as convergence criteria (Convergence Condition).

Viewing 3 reply threads
  • You must be logged in to reply to this topic.