Preprocessing

Preprocessing

Difference between domain independence study and Mesh independence study

    • gishnutr
      Subscriber

      Hi, All


        I am doing fluid flow simulation 2D in an open space (not all boundary open) with two boundary walls. 


      So to find the effect of the two boundary walls , i think i need to conduct a domain independence study. 


      1. Is there any standard procedure to do that? 


      2. Which one do i have to do first Mesh independence or domain independence?

    • Vishal Ganore
      Ansys Employee

      Hi, 


      I will get my mesh right first before playing around a domain. To avoid the effects of open walls, you can place the walls far away from the body to get started. For eg., If you are solving an external flow over a car in a wind tunnel then you could follow the standard practice. Let's say L is the length of the car.


      1. Place inlet at a distance of 1.3L from the body


      2. Place outlet at a distance of 5L to capture the physics & wakes accurately. 


      3. The top boundary at a distance of 3L to make sure that boundary reflects will not come in.


      Carry out the iterations, get the results and then you could try to even increase this distance further to understand whether results are changing. 


       

    • peteroznewman
      Subscriber

      I have been studying mesh independence and came across a useful NASA site that give some guidance for CFD models. I'm applying similar method for structural models as described in this discussion.


      It would be helpful if you could include a sketch of the model geometry.

    • raul.raghav
      Subscriber

      Domain independence: When your domain doesn't affect the major flow features (not a standard definition and sometimes cannot be applied due to certain geometrical restrictions). In your case it could be (i) the length of the domain or the distance between the inlet and the outlet, and (ii) the distance between the two boundary walls. Again a true domain independence should also include the effects of the 3D domain on the flow.


      A neat and easy way of doing your case would be parametrization of the above mentioned dimensions (length and distance between the walls) and setup a standard "Mapped face meshing" with "Face Sizing" being a parameter. Then you can setup the simulation in Fluent and monitor your results. If you want to measure velocity or pressure drop or any other variable of interest could be the output parameters. So all you do it setup your simulation once and then everything is streamlined by just changing the parameter values. The following youtube tutorials might be useful if you are interested in this approach.


      Parameterization of geometry in designmodeler (WB)


      Parametric meshing in ANSYS Meshing


       


      Grid/mesh independence: When your results stop changing with successive mesh refinements. The idea is to use an optimal mesh size that would best represent the flow scenario in the shortest time possible without significant loss in accuracy. Say you use a mesh size of 10k, 20k and 30k. If your 10k mesh provides you results that are within a certain error % of the 20k and 30k mesh, it makes sense to use the 10k mesh. Below is good blog post on grid independence:


      Convergence and Mesh Independence Study

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