Discussion Forum

Hi Yao,

I believe the reason you are seeing more data points than mesh nodes is because you have set a "Refinement" in the Quality tab; the default is set to normal. I think the refinement evaluates more points between mesh nodes. You can set this to No Refinement and should get the right amount of data points.

Let me know if this works,

Alex

Hi Yao,

In principle, Alexandre is right.

There is however one subtle detail that maybe should be mentioned: With No Refinement, only the corner nodes of the elements are used. This matches the mesh nodes only if the discretization is linear.

If you want to use values also at the midside nodes when quadratic shape functions are used, you need to set refinement to Custom, and the corresponding value to 2.

Hi Yao,

I believe the reason you are seeing more data points than mesh nodes is because you have set a "Refinement" in the Quality tab; the default is set to normal. I think the refinement evaluates more points between mesh nodes. You can set this to No Refinement and should get the right amount of data points.

Let me know if this works,

Alex

Thanks Alex! That was the right setting.

Hi Yao,

In principle, Alexandre is right.

There is however one subtle detail that maybe should be mentioned: With No Refinement, only the corner nodes of the elements are used. This matches the mesh nodes only if the discretization is linear.

If you want to use values also at the midside nodes when quadratic shape functions are used, you need to set refinement to Custom, and the corresponding value to 2.

Thank you for your answer, Henrik!

Are the data points always chosen to be the corner nodes?

For instance if with your example of quadratic elements and Resolution: Custom with Element refinement = 2, would the data point inbetween reflect the middle node of the element? Or would the position be an interpolation based on the chosen refinment, with the value being an evalution of the field at that point?

Edit: I looked into the reference and it says that the predefined values would do a refinement based on 'heuristics' and a custom value makes the refinment by subdividing the edges. So I think that the positions are kind of an interpolation and the data is then evaluated? Is this the right thinking?

Cheers, Yao

Hi Yao,

The value is taken at interpolated points, yes. But in my example, that would coincide with the midside node. As long as Element Refinement matches the shape function order, you will get values at the nodes (except possibly for some round-off errors), since they are placed on the line using the same type of interpolation.

Note also that that 'nodal values' is a meaningful statement only when you evaluate the DOF itself, but not when evaluating other expressions like gradients and fluxes.