[Hooman]

Hi,

Can anyone please tell me way and reasons why results of a simulation could have a zig-zag behaviour and possible ways to remove them?


Thanks!

[Lurapa]

Are you using incompressible or compressible Navier Stokes?

[FMDenaro]

Originally Posted by Hooman

Hi,

Can anyone please tell me way and reasons why results of a simulation could have a zig-zag behaviour and possible ways to remove them?


Thanks!

depends on the kind of model you are using, compressible or incompressible flows, viscous or not, then check for the stability contraints of your numerical method, chech for the grid regularity...
search for "odd-even decoupling", "numerical instability", "wiggles"

[Hooman]

Thanks,
Can you please suggest a book for checking the stability of a numerical scheme.

[Hooman]

My flow is compressible at low speeds of sound.

[Azita]

Originally Posted by Hooman

Hi,

Can anyone please tell me way and reasons why results of a simulation could have a zig-zag behaviour and possible ways to remove them?


Thanks!

Increasing the resolution of your grids, may also help to get rid of un-physical zig-zag behaviour.

[Hooman]

tanx, what do you mean by increasing the resolution? do u mean refining the mesh?
I refined the mesh and the zig-zag become very small, I don't zoom in, the results almost seem smooth, but they still exist.
reducing the time step size does not have any effect.

[Azita]

Originally Posted by Hooman

tanx, what do you mean by increasing the resolution? do u mean refining the mesh?
I refined the mesh and the zig-zag become very small, I don't zoom in, the results almost seem smooth, but they still exist.
reducing the time step size does not have any effect.

I mainly meant mesh-resolution, but depending on the problem, there might be other ways to increase the resolution. For example if you are solving the Radiation equation, resolution can be increased by increasing the "Angular Discretization".

First off, try to increase your mesh resolution and if you are having high computational expense issue, there are various techniques to overcome like

i. Decrease the dimension from 3D to 2D if possible
ii. Start with high mesh sizes and then refine the mesh based on the initial solution as required

Later after you generated the results using the best mesh possible, simply take the average of the variable of interest over the area where physically it should remain constant. This way you can come up with more realistic values for variables of interest.

[Hooman]

Originally Posted by Azita

I mainly meant mesh-resolution, but depending on the problem, there might be other ways to increase the resolution. For example if you are solving the Radiation equation, resolution can be increased by increasing the "Angular Discretization".

First off, try to increase your mesh resolution and if you are having high computational expense issue, there are various techniques to overcome like

i. Decrease the dimension from 3D to 2D if possible
ii. Start with high mesh sizes and then refine the mesh based on the initial solution as required

Later after you generated the results using the best mesh possible, simply take the average of the variable of interest over the area where physically it should remain constant. This way you can come up with more realistic values for variables of interest.

Hi,
Thanks again. my problem is actually just 1D.
The results are realistic when the mesh is very fine, but it still has the zig-zag behaviour, not sure if that would be acceptable.
About taking the average of the area, is this something that is acceptable to do?

[Aeronautics El. K.]

Which equation are you solving and what methods are you using?