Mesh refinement vs P-Elements ?

[MaikTheBike]

Hi,

I've heard that there are two method for refining the FEA-mesh:

1. Smaller Elements (Autogem Control)

and 

2. Increasing P-Level from 1 to 9, which is done in multipass-adaptive

Now I wonder when it makes sense to use smaller elements  and when to use elements with high P-levels.

Is it ok to use the standard-mesh with high P-Level to get a accurate result ?

Or do I need to use a mesh with high P-Level AND small Element-Size to get accurate results ?

When using high P-Level and small element-size the solution takes a lot of time.

So whats the best way to refine the mesh and to know that the solution is correct ? 

Does it depent on the part I want to analyze (e.g. stress in a weld; or stress in a hook (crane) or stress in a pressure-tank ?)

Thanks a lot,

Maik

[DavidBigelow]

H-Elements are historically and traditionally used and replied on.

P-Elements are the "new-old kid on the block (ahem .. as of 20 years ago)" 

H-P Elements are a hybrid approach.... (not all FEA systems support these)

Voxels seems to be the newest trend - Real-Time GPU Powered Solutions.

This really depends on your needs and goals for the type of analysis you are interested in performing (statis, dynamic, contact, collision, meta physics, etc.)

While FEA is fundamentally a great thing, each technology has its benefits and gaps.  Edge cases tend to showcase the limitations better (e.g. column deformation is usually a good reference test case).

Higher level mesh resolution vs compute time vs accuracy is a balance... experience with your problems will determine the balance.   Typically higher boundary resolution conditions are more desirable (especially in acute geometry - because that is where the stress tends to be more pronounced).

My recommendation - start out coarse, and refine as you learn more from faster analysis to add resolution where necessary to obtain more relevant analysis results.

Hope this helps.

Dave