Re: [Libmesh-users] A mistake in the miscellaneous example 12?

[Yuxiang W. <yw...@vi...>]

Hi David,

Thank you so much for talking to Sylvain! It's really nice that I can hear
from the thoughts of the original implementer. That makes sense to me.

For future reference for anyone who might be interested, the original
author in the book recommended a value of 1e-4 to 1e-7 (maybe for more
general cases). For this specific problem, I tried to compare the results
of alpha = 1 and alpha = 1e-6, the results are *marginally* improved.

With alpha = 1:
z-displacement of the point C: 156.824
Analytic solution: 164.24
y-displacement of the point D: 4.05575
Analytic solution: 4.114

With alpha = 1e-6:
z-displacement of the point C: 157.149
Analytic solution: 164.24
y-displacement of the point D: 4.06329
Analytic solution: 4.114

I suggest add a comment of author-recommended alpha values in the example.
I could do a pull-request if anyone thinks that it is worth it?

Best,
Shawn




On Mon, Oct 22, 2018 at 5:23 AM David Knezevic <dav...@ak...>
wrote:

> On Sun, Oct 21, 2018 at 3:48 AM Yuxiang Wang <yw...@vi...> wrote:
>
>> Dear all,
>>
>> I seem to find a mistake in the implementation of the MITC4 shell
>> element, in the miscellaneous example 12.
>>
>> In the reference used, "Modelisation des structures par elements finis,
>> Vol. 3: Coques.", on page 372 equation 6.3.90 it only added a small
>> drilling stiffness. This was done by a coefficient alpha that is
>> sufficiently small (1e-4 to 1e-7). However, in line 779
>> <
>> https://github.com/libMesh/libmesh/blob/master/examples/miscellaneous/miscellaneous_ex12/miscellaneous_ex12.C#L779
>> >of
>> the code (corresponding to this equation), this coefficient is dropped.
>>
>> Could anyone please help discuss this issue?
>>
>
> Hello,
>
> My colleague Sylvain implemented that example, and I passed your question
> onto him. His reply was "The drilling dof coefficient value is very
> empirical, and I think in that specific example I just got good results
> with a value of 1." So there is no bug here, it's just that he had to set
> the value in a way that worked well for this example.
>
> Best,
> David
>


-- 
Yuxiang "Shawn" Wang, PhD
yw...@vi...
+1 (434) 284-0836