Integration by parts-I can't reproduce a given answer

  • Thread starter Thread starter bzz77
  • Start date Start date
  • Tags Tags
    Integration
AI Thread Summary
The discussion revolves around a user's difficulty in reproducing a given answer for an integration by parts problem in finite element analysis. The user initially presents their integral and the provided solution, noting a discrepancy that includes an additional term in their calculation. After receiving feedback, it is clarified that the extra term arises from boundary conditions, which are often neglected in finite element analysis. The user confirms that their integration is between a specific pair of nodes and acknowledges the importance of boundary terms in their calculations. Ultimately, the conversation highlights the nuances of applying integration by parts in the context of finite element methods.
bzz77
Messages
33
Reaction score
0
Integration by parts--I can't reproduce a given answer

I am working through an example problem that involves integration by parts. The answer is given, but the answer I get is different.

Can anyone help me identify the problem? Is it me or the given answer?

Question:
∫ [ (Ni) (d2 Nj / dx2) ] dx

Given answer:
- ∫ [ (dNi / dx) (dNj / dx) ] dx + boundary terms

Answer I get (additional term):
Ni * dNi/dx - ∫ [ (dNi / dx) (dNj / dx) ] dx + boundary terms

Thanks a lot for any help!
 
Mathematics news on Phys.org
hi bzz77! :smile:

you have an extra [Ni dNj/dx]0, which is correct

what is N ?

I assume it's defined so that [Ni dNj/dx]0 = 0
 


Hi tiny-tim:

Thanks a lot for your help!

Ni = (1 - x/L); Nj = x/L

Am I making a silly mistake? I still can't see where I'm going wrong! Thanks for any assistance!
 
hi bzz77! :smile:

(i don't know where i got those limits from :redface:)

what problem does this come from?

are the limits from 0 to L ?

i suspect that at the boundaries either N or N' has to be zero
 


Hi tiny-tim:

Thanks again! It comes from an example problem I'm working through in finite element analysis. Sorry, I should have mentioned that the limits are 0 to L! I'm an old fart and I have totally forgotten my integral calculus! Sorry for the silly question!

P is a continuous variable that we are approximating within an element in terms of the P at two nodes P1 and P2.

So: N1 = 1 - x/L and N2 = x/L
L is the length of an element and x is the spatial variable that varies from 0 at node 1 to L at node 2.

N1=1 at node 1 and N1=0 at node 2.
N2=0 at node 1 while N2=1 at node 2.
N1 + N2 = 1 (over the entire element).
 
hmm …

it looks like N1 = 0 at one limit, and 1 at the other, but dN2/dx is constant :confused:

are you integrating just between one pair of nodes, or over a large number of nodes?
 


Well, I'm integrating it between one pair of nodes... Then doing it over another pair of nodes until I have covered my whole mesh.

I think I have an inkling about what the issue is now... It seems like that extra term has to do with the boundary conditions. And that they somehow cancel out.
 


Hey tiny-tim:

Just letting you know that the boundary terms (my extra terms) are often ignored in finite element analysis. So maybe that's why that term is not included in my example. Anyway, thank you so much for all your help!
 
Back
Top