Mesher refinement

[Ben Fernando]

Hey all, 

I'm using pymesher to build a grid for the Sun, but am having some issues with the meshing. 

1) In the tripling refinement case, the elements at the centre of the AxiSEM mesh are not quadrilaterals. Is there a way that I can make them be quadrilaterals? I notice that if I mesh the VPRMoon.bm model that's on GitHub this is done automatically, but is there a way to force it to happen? 

2) In the doubling case, there's something very odd going on at the refinement layers which I don't really understand - the layers appear to double, and then go back to the way that they were? 

If it helps this is a very big mesh (600,000km radius, and very long period (up to 2000s). I've tried different values of hmax but that doesn't seem to change the odd behaviour. Photos are attached. 

Thanks 

Ben 

[Martin van Driel]

Hi Ben,

can you provide us with the model file and the exact inputs you used to reproduce the behaviour?

Thanks,
Martin

--
You received this message because you are subscribed to the Google Groups "Salvus" group.
To unsubscribe from this group and stop receiving emails from it, send an email to salvus+un...@googlegroups.com.
To post to this group, send email to sal...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/salvus/daec3b6d-a18a-488d-9245-f0b8c0b78d2b%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Ben Fernando]

Hi Martin, 

In the tripling case: $ python -m pymesher.interface AxiSEM --basic.period=1000 --chunk2D.max_colatitude=180 --basic.model solar_test4.bm --overwrite  --refinement.hmax_refinement 100 --refinement.refinement_style tripling 

In the doubling case: $ python -m pymesher.interface AxiSEM --basic.period=1000 --chunk2D.max_colatitude=180 --basic.model solar_test4.bm --overwrite  --refinement.hmax_refinement 100 --refinement.refinement_style doubling

.bm file is attached as well. 

Thanks 

Ben 

On Friday, October 6, 2017 at 10:24:44 AM UTC+1, van Driel Martin wrote:

Hi Ben,

can you provide us with the model file and the exact inputs you used to reproduce the behaviour?

Thanks,
Martin

On 06.10.2017 11:21, 'Ben Fernando' via Salvus wrote:

Hey all, 

I'm using pymesher to build a grid for the Sun, but am having some issues with the meshing. 

1) In the tripling refinement case, the elements at the centre of the AxiSEM mesh are not quadrilaterals. Is there a way that I can make them be quadrilaterals? I notice that if I mesh the VPRMoon.bm model that's on GitHub this is done automatically, but is there a way to force it to happen? 

2) In the doubling case, there's something very odd going on at the refinement layers which I don't really understand - the layers appear to double, and then go back to the way that they were? 

If it helps this is a very big mesh (600,000km radius, and very long period (up to 2000s). I've tried different values of hmax but that doesn't seem to change the odd behaviour. Photos are attached. 

Thanks 

Ben 

--
You received this message because you are subscribed to the Google Groups "Salvus" group.

To unsubscribe from this group and stop receiving emails from it, send an email to salvus+unsubscribe@googlegroups.com.

[Martin van Driel]

Hi Ben,

thanks for the file, I am able to reproduce the problem:

the first issue is due to the discontinuity that you enforce by repeating the 0.1 km radius in the model file. Removing it should fix it.


Actually, if you zoom in enough you can see the quads also with the original model, they are just tiny compared to the domain size:


this is not a bug, but a feature: there can't be a discontinuity in the central most part of the mesh, because that would not be correctly meshed.

The second issue with doubling is also avoided by using the fixed model, although it still seems to hint at a bug that need to be fixed. Even with the extremely small 'inner core' model it should produce a correct mesh. I'll need to take a closer look, but that should not stop you from continuing your work:



The model file I used is attached, command lines are the same as yours.

Cheers,
Martin

To unsubscribe from this group and stop receiving emails from it, send an email to salvus+un...@googlegroups.com.

To post to this group, send email to sal...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/salvus/8542c062-f40f-4e9d-b524-f678b6a0d725%40googlegroups.com.