Find answers, ask questions, and connect with our <br>community around the world.

Home Forums CFD foundation course: Forum Several Doubts in General

  • Several Doubts in General

    Posted by María Manuela on August 24, 2023 at 9:58 am


    I have listed some doubts, hope anyone can help me with them:

    1) I’ve experienced that continuity equations takes longer to achieve convergence is we compare it with velocity components, why happens that?

    2) How a Fluent user sets in the solver the use of WallFunctions? or is it something intelligent that Fluent chooses by its own if checks that y+ is between 30 a 300?

    3) Which mesh program would you recommend to fix a 3D model in case we can not use ICEM or Fluent?

    4) For academic use, which is the % or error between CFD results and experiments?

    5) For industrial use, which is the % or error between CFD results and experiments?

    6) To monitor residuals, for industrial porpuses which are the tolerance acceptable?

    7) Is there a free database online with wind-tunnel experiments?

    8) If we take a wind tunnel experiment, reproduce it with CFD, can we take the same numerical-methods and govern equations if we want to scale the simulation into a prototype?

    9) Would FlowthermoLab have an example model of a report of a CFD simulation? Or may a list of recommendations… in order to use as guide to know how to structure a report in our jobs

    Thank you!

    Sandeep replied 7 months, 2 weeks ago 2 Members · 1 Reply
  • 1 Reply
  • Sandeep

    September 4, 2023 at 10:36 pm

    Really a long list!!!

    let me try to answer it one by one.

    1. Continuity usually takes a long time if there is any kind of recirculation in the domain. Either from outlets or inlets. So in such cases in CFD we try to extend the inlets and outlets to avoid such recirculation flows.

    2. Yes there is inbuilt wall function. You can also modify it if required.

    3. Any meshing software can be used in 3D. Structured mesh is preferred as the convergence is very fast usually and you can get mesh convergence with less number of cells compared to unstructured mesh Also the mesh generation is under your control. So I suggest using ICEM in general applications, Turbogrid for turbomachines, and also GridPro as another option for structured meshing. But these days for Fluent even unstructured meshes are giving good results, especially using polyhexcore or polyhedral mesh (prepared using Fluent Mesher).

    4. Usually there will be at least 5% error between CFD and Experiments. If it matches perfectly then we should doubt CFD because there are many factors like turbulence modeling etc. which will cause the difference. There are complicated problems where even 50% deviations or order of magnitude results are also acceptable. So it is purely dependent on the class of problems.

    5. In industry accuracy is not as important as in academia where publishing data is given more importance. So In industry people will look for trends with minimum effort. Time and money is key in the industry. If you can get between 5-20% is good enough.

    6. In industrial problems we really don’t care about residuals much. We will monitor all the variables that are important over timesteps or iterations along with residuals. And if the monitor points come to a steady state of quasi-steady state conditions then we can consider stopping the simulation.

    7. Not just one place. There are many University repositories, NASA repositories, research communities like Ercoftac

    8. Yes that should work unless there is some massive difference coming due to scaling.

    9. We can discuss a structured report in one of our sessions/live classes.

    Hope this is useful.


Log in to reply.

error: Content is protected !!