Re: [Libmesh-users] Gradient of C1 error

[Ataollah M. <am...@ti...>]

Roy, 

1st as always thanks for the quick response. 

I see what you mean. And no I didn't I did the gradient calculation in VisIt. I don't use/calculate this gradient in my code. 

Here is the reason I suspected gradients:

the problem I'm solving is complicated but in short what I solve for is a fracture mechanics problem and the solution I showed you is one step of an iterative solution for the displacement field; long story short when I have a discontinuity of gradient of displacement (i.e. a kink in the displacement field) that will initiate cracks. In this case when the gradient is not continuous  along mesh partitions I get fracture along mesh partitions. 

I'll try to run some more checks and then get back to you before spending time on a test case.

Ata
On Dec 4, 2012, at 2:04 PM, Roy Stogner <roy...@ic...> wrote:

> 
> On Tue, 4 Dec 2012, Roy Stogner wrote:
> 
>> On Tue, 4 Dec 2012, Ataollah Mesgarnejad wrote:
>> 
>>> I found a potential bug with Clough elements. I'm solving a 4th order
>>> problem using C1 Clough elements and when I plot magnitude of gradient of
>>> the solution I found out that it is discontinuous on mesh partition
>>> boundaries. I would appreciate if any comments.
>> 
>> Can you put together a test case reproducing the problem?
> 
> Actually, before you do that: are you sure you localized the solution
> before using it to calculate the gradients?  That's the only way I
> could picture encountering a partitioning-specific gradient
> discontinuity in C1 elements - the gradient components aren't
> calculated (and thus aren't potentially miscalculated) from some other
> degrees of freedom, they *are* degrees of freedom.
> ---
> Roy