Convergence observed

  • 64 Views
  • Last Post 10 October 2018
hugo CFD posted this 06 October 2018

Hello

When we are conducting a mesh convergence study, it is necessary to assign a theoretical convergence index. The observed convergence index is then calculated, based on the results obtained from the variables of interest. Due to different types of error, the order of convergence observed is often less than theoretical.

I got an order of convergence observed superior to theoretical. Is this a plausible result? What is the origin of this result?

 

With respect to the extrapolation of richarsdon, this is to calculate a value of the variable under study when the refinement of the mesh tends to zero. Does this mean that this value is the real value of the solution? or just the actual value of the numerical simulation?

Thank you

Order By: Standard | Newest | Votes
Kremella posted this 09 October 2018

Hello,

Could you please elaborate on the problem your are trying to do mesh independence studies? What model are you using? Which modeling tool are you using?

Please provide us with more details to help us understand the context.

Thank you.

Best Regards,

Karthik

hugo CFD posted this 09 October 2018

Hello.

I'm simulating a flat plate heat exchanger, almost counter flow.

In the study of knit independence, I used the richardson method. The ratio between the dimension of the mesh element is 2. Three mesh sizes 0.5mm, 1mm, 2mm were studied. The shape of the elements used is hexagonal.

According to the variable in study, heat flux, the three meshes studied converge to the value extrapolated by the richardson method. However, what is the reason for the p-value, order of convergence is higher than the stimulated ratio? Is it supposed to be smaller due to the associated errors, or am I wrong?

 

Thanks

mcoderon posted this 10 October 2018

Hello, 

If you are simulating a turbulent flow you need to be careful on how the mesh refinement modifies the y+ value at the wall. Changing the size of your mesh may drastically change the y+ value and your mesh may or may not verify the requirements of your turbulence model.

Let's say that hypothetically you want y+=1 and the coarser meshes do not verify that condition. The solutions are affected by a bad physics modeling and the extrapolation may give you unexpected results.

Regards,

Marco

hugo CFD posted this 10 October 2018

No. the regime is laminar. 

What I want to clarify is not the extrapolated value. is the value of p, of the order of convergence being greater than the chosen ratio between meshes.

Close