Dear Yves,

Thank you. This is precisely the formulation I used but it raises the
following problem (in python)

0.5*sqr(Normalized(element_K).(Grad_u*Normalized(element_K)))
------------------------------------------^
The second argument of the dot product has to be a vector.
logic_error exception caught
...
RuntimeError: (Getfem::InterfaceError) -- Error in
getfem_generic_assembly.cc, line 8949 void getfem::ga_node_analysis(const
string&, getfem::ga_tree&, const getfem::ga_workspace&,
getfem::pga_tree_node, bgeot::size_type, bgeot::size_type, bool, bool,
int):
Error in assembly string

following a call such as
md.add_linear_generic_assembly_brick(mim,"0.5*sqr(Normalized(element_K).(Grad_u*Normalized(element_K)))")

Did I miss something?

I am sorry to bother you again.

Thanks

Best regards
Jean-François



2017-11-20 14:10 GMT+01:00 Yves Renard <[email protected]>:

>
> Dear Jean-François,
>
> For a vector variable 'u', each line of 'Grad_u' is the gradient of the
> ith component of 'u', each of them is tangent to the curve and length being
> the derivative with respect to the curvilinear abscissa. The linearized
> deformation is a priori  Normalized(element_K).(Grad_u *
> Normalized(element_K))
>
>
> The formulas used to compute the gradient and the Hessian can be found
> here:
>
> http://getfem.org/project/femdesc.html#geometric-transformations
>
> http://getfem.org/project/appendixA.html#derivative-computation
>
> The hessian of a vector valued variable is also the hessian of each
> component.
>
> Best regards,
>
> Yves.
>
>
>
>
>
> Le 20/11/2017 à 02:10, Jean-François Barthélémy a écrit :
>
> Dear Yves,
>
> Thank you very much for your answer.
>
> It's OK for scalar variables but I do not really understand how Grad_u is
> built when u is a displacement vector field of 3 components. I thought
> Grad_u would represent the vector du/ds ([dux/ds,duy/ds,duz/ds]) with s the
> local curvilinear abscissa (so that Normalized(element_K).Grad_u would give
> the linearized longitudinal deformation) but it seems that Grad_u is
> actually a 3x3 matrix field. Then I do not see how to build the
> longitudinal deformation. What would be the best way please? And by the
> way, what would be the right syntax to get the second derivative of the
> transverse displacement by means of Hermite elements and the Hessian?
>
> Thank you again for your help.
>
> Best regards
> Jean-François
>
>
>
> 2017-11-17 20:52 GMT+01:00 Yves Renard <[email protected]>:
>
>>
>> Dear Jean-François,
>>
>> There is no specific tool yet for that.
>> You can have access to the tangent with 'element_K' in the generic
>> assembly language (the unit tangent is then 'Normalized(element_K)')
>> If you define a scalar quantity "u" on your 1D structure, then "Grad_u"
>> will be the gradient of the quantity in the sense that it is a tangent
>> vector whose norm is the derivative of the qunatity along the curve. So
>> that "Grad_u.Grad_Test_u" is still the stiffness term for a curvilinear
>> second derivative. For a vector quantity "u", "Grad_u" is the componentwise
>> gradient.
>>
>> Best regard,
>>
>> Yves.
>>
>>
>>
>> ----- Original Message -----
>> From: "Jean-François Barthélémy" <[email protected]>
>> To: [email protected]
>> Sent: Friday, November 17, 2017 6:17:13 PM
>> Subject: [Getfem-users] Curvilinear structures in Getfem
>>
>> Dear Getfem users,
>>
>> I wonder whether it is possible to model simple linear elastic curvilinear
>> structures submitted to traction, bending, torsion etc... in 2D or 3D in
>> Getfem. I haven't found a way to have access to the tangential or normal
>> parts of vectors in the local basis of a beam and their derivatives with
>> respect to the curvilinear abscissa needed to build the formulation. Does
>> someone have an answer please?
>>
>> Thanks in advance
>>
>> Best regards
>> Jean-François
>>
>
>
> --
>
>   Yves Renard ([email protected])       tel : (33) 04.72.43.87.08
>   Pole de Mathematiques, INSA-Lyon             fax : (33) 04.72.43.85.29
>   20, rue Albert Einstein
>   69621 Villeurbanne Cedex, FRANCE
>   http://math.univ-lyon1.fr/~renard
>
> ---------
>
>

Reply via email to