# What is exact reason behind choosing basic primitive shapes for elements in FEA?

1. Apr 29, 2010

### mvpunekar

what is exact reason behind choosing basic primitive shapes for elements in FEA?

Why it cant be hexagonal, pentagonal, octagonal, etc. in both cases 2D and 3D.....

2. Apr 29, 2010

### Mech_Engineer

All of the shapes you've described can be decomposed into pyramids (triangles).

3. Apr 29, 2010

### mvpunekar

Its true...but why those small elements can take any other shape....

why only those basic shapes are standardised?

4. Apr 29, 2010

### mvpunekar

how interpolation functions are affected by choosing higher primitive shapes??

5. Apr 29, 2010

### Mech_Engineer

They're the only ones neceaary to describe any shape under the sun, especially when you take into account midside nodes which allow arbitrary shapes and quadratic interpolation. take for example this ANSYS element:

An element that was shaped like a dodecahedron would still be subject to the limitations of triangular elements, since it has faceted sides with discrete nodes; there is no advantage to using a huge element with a million faces when you can use a million elements with single faces.

#### Attached Files:

• ###### gELEM152-1.gif
File size:
3.3 KB
Views:
80
6. Apr 29, 2010

### Mech_Engineer

For some more examples:

Pyramid with midside nodes:

Brick with midside nodes:
[URL]http://research.me.udel.edu/~lwang/teaching/MEx81/ansyshelp/graphics/gELEM95-1.gif[/URL]

Last edited by a moderator: May 4, 2017
7. Apr 29, 2010

### mvpunekar

Last edited by a moderator: May 4, 2017
8. Apr 29, 2010

### Claws

A follow-up question from me: Why does it seem like pretty much every mesher aims to create quads/hexahedrons, and from my own experience quads/hexahedrons blows away triangles/tetrahedrons in terms of density/precision of result; fewer quads/hexa are required to reach a comparable analytical solution.

9. Apr 30, 2010

### mvpunekar

Thanks for help...

but can u relate ur answer to interpolation functions??

why elements like octagon, pentagon,etc are not used for 2D meshing(also in solids)?

10. Apr 30, 2010

### minger

Quads are typically easier to "control". The mesh lines are smooth, so I can typically avoid erroneous results from rapidly changing element shapes and sizes. In addition to that, you typically find that you can get a better mesh with much fewer elements using "brick" elements.

11. Apr 30, 2010

### Mech_Engineer

As I explained, both an octagon and pentagon can be broken down into triangles; there is no advantage to use a larger element when it will have a much larger stiffness matrix to solve. There are fundamental advantages between brick and pyramidal elements in terms or solution accuracy and mesh quality, but I can't think of any reasons a more complex element would be better.

12. Apr 30, 2010

### mvpunekar

now i got it perfectly ...thanks Mech engg...

13. Apr 30, 2010

### mvpunekar

can u provide some links to clear basics of FEm from practicality point of view.......

14. Apr 30, 2010

### Mech_Engineer

You can just do a search on Google. Some results that will come up are like this:

Last edited by a moderator: May 4, 2017
15. Apr 30, 2010

### mvpunekar

Thanks again

by the way ,,,whats ur specialisation ??

n where r u from??

Last edited by a moderator: May 4, 2017
16. May 1, 2010

### mvpunekar

This books has more mathematical approach ..

can u provide me links which teaches basic from pratical point of view ...like there is one book called "Practical Finite Element Analysis" by Nitin Gokhale, Finite to Infinite Publications....

Last edited by a moderator: May 4, 2017
17. May 1, 2010

### Mech_Engineer

In case you didn't realize it let me clear something up for you- finite element analysis is ALL math. You can't learn about FEA without math, period. Even if you want a book with "pratical" in the name, it doesn't mean it won't cover the math aspects of FEA.

I'm a mechanical engineer, my job involves a lot of FEA analysis. Structural, thermal, electromagnetic, CFD, and coupled-physics.

18. May 2, 2010

### mvpunekar

Thanks

But i have seen in some big companies that the Sr FEA engineers r just good at application part ...and not good in basics of maths of FEA....... i guess only practical application of FEA is necessary & not its all core maths.....its necessary for ppl who r developing those softwares?? isnt it?

anyways myself i have started learning FEA, ansys n Hypermesh ......will ask u whenever i get some doubts .....