Dear Phuoc,
I think the correct expression with Contract is simply
div_sigma = "(Contract(Grad(first_Piola_Kirchhoff_stress), 2, 3))"
Contract makes a summation with respect to the given indices, so that
Contract(A, 2, 3) = A_{i,j,j} with summation on j.
Can your try again with the fix I commited to the master branch ?
Best regards,
Yves
----- Original Message -----
From: "Huu Phuoc BUI" <address@hidden>
To: "yves renard" <address@hidden>
Cc: "franz chouly" <address@hidden>
Sent: Friday, 13 July, 2018 12:00:55
Subject: Re: [Getfem-users] Divergence of the first Piola-Kirchhoff stress
tensor
Dear Yves,
Thank you very much for your email.
Please find enclosed a minimal code.
I am not sure that I use correctly the Contract operator, you will find
therefore both the original code and the new one for div_sigma:
/div_sigma = "[Contract(Grad(first_Piola_Kirchhoff_stress)(1,:,:),1,2),
Contract(Grad(first_Piola_Kirchhoff_stress)(2,:,:),1,2),
Contract(Grad(first_Piola_Kirchhoff_stress)(3,:,:),1,2)].[1,1,1]"/
Please let me know if you find the problem. Thank you!
Best regards,
Phuoc
On 13/07/2018 08:59, Yves Renard wrote:
Dear Phuoc,
No, a priori this should work. To simplify a bit your expression, you
can use Contract(Grad(...), 2, 3).[1,1,1] instead of the three traces
in div_sigma.
If you give me a minimal code, I could have a look to the problem.
Best regards,
Yves
Le 12/07/2018 à 15:36, Huu Phuoc BUI a écrit :
Dear GetFEM-users,
As Grad(expression) is available in the new version of GetFEM++, I
compute the divergence of the first Piola-Kirchhoff stress tensor as
(python code):
/div_sigma = "( Trace(Grad(first_Piola_Kirchhoff_stress)(1,:,:)) +
Trace(Grad(first_Piola_Kirchhoff_stress)(2,:,:)) +
Trace(Grad(first_Piola_Kirchhoff_stress)(3,:,:)) )"/
in which first_Piola_Kirchhoff_stress is a macro of a model /md/,
which is defined as:
/md.add_macro("second_Piola_Kirchhoff_stress",
"Compressible_Mooney_Rivlin_sigma(Grad_u, [c1;c2;d1])")//
//md.add_macro("first_Piola_Kirchhoff_stress",
"(Id(meshdim)+Grad_u)*second_Piola_Kirchhoff_stress")/
Then I use generic assembly to assembly this quantity as:
/eta_L_1_tmp = gf.asm_generic(mim,1,"(("+div_sigma+").delta)*Test_t",-1//
// ,md//
// ,"delta", False, mfz, delta//
// ,"t", True,mfrj,np.zeros(mfrj.nbdof()) )/
with /delta/ a given vector on a meshfem /mfz/.
When executing the code, GetFEM++ only gives the error as
/segmentation fault (core dumped)/
If you see any problem in my code, please let me know. Thank you in
advance.
Best regards,
Phuoc
--
Yves Renard (address@hidden) 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
---------