[Kratos] new constitutive law interface, second draft

Janosch Stascheit janosch.stascheit en rub.de
Jue Jun 10 08:44:21 CEST 2010


Hi all,

find my replies below:

Am Mittwoch, den 09.06.2010, 23:24 +0200 schrieb Riccardo Rossi:
> just a few remarks:
> 
> 1 - if we like we can make the implementation shorter and more general
> by making for example
> 
> template< TDataType >
> virtual bool Has( const Variable< TDataType >& rThisVariable )
>             {
>                 return false;
>             }
> 
> the use of the template would allow avoiding specifying all of the
> cases, furthermore it would be automatically extended for new datatypes
> (for example bool)
> 
good idea, yet not possible: virtual functions cannot be templatized
> 
> 2 - concerning the volumetric stuff we could also consider the option
> 
> virtual void CalculateVolumetricResponse( const double VolumetricStrain,
>                                                       double&
> StressVector,  //PROPOSED CHANGE HERE
>                                                       double&
> algorithmic_bulk ,  //PROPOSED CHANGE HERE
>                                                       const ProcessInfo&
> CurrentProcessInfo,
>                                                       const Properties&
> props, 
>                                                       const
> GeometryType& geom,
>                                                       const Vector&
> ShapeFunctionsValues,
>                                                       bool
> calculate_stresses = true,
>                                                       int
> calculate_tangent = true,
>                                                       bool
> save_internal_variables = true )
> 
> 
> i see the advantages of yours ... but this also looks quite natural
> 
the volumetric response stuff does not affect my constitutive models.
Hence, I am willing to add any kind of parameters that suit you best.
For the moment I will implement it as proposed by Riccardo
> 
> 3 - ResetMaterial ... i would also pass geom, N, and processinfo ... you
> never know...
> 
good idea. I will do it.

> 4 - why don't you like having Cauchy function public?
> 
Actually, I thought that this would only be called internally. The
problem is, that this would somehow foil the concept of having only one
function for the material response. Are the cauchy stresses definitely
needed? 
> 
> for the rest i think your proposal looks quite clean!
> 
> well .. that's all from my part.
> 
> Nelson any suggestion?
> 
> Pooyan?
> 
> on my side i will ask Michele Chiumenti who is an expert on the subject,
> and i may also speak with Sergio Oller to have an opinion
> 
> ciao all
> Riccardo
> 
> 
> 
> 
> On Tue, 2010-06-08 at 10:43 +0200, Janosch Stascheit wrote:
> > Dear all,
> > 
> > attached, you will find the second draft of the new constitutive laws
> > interface. Please note that the definition of pure virtual functions
> > failed due to the fact that a base class wrapper is needed in the python
> > interface. This, in turn, requires the instantiation of the constitutive
> > law base class, not being possible if pure virtual functions are
> > present. I created KRATOS_ERROR statements instead.
> > 
> > Best regards
> > Janosch
> > _______________________________________________
> > Kratos mailing list
> > Kratos en listas.cimne.upc.edu
> > http://listas.cimne.upc.edu/cgi-bin/mailman/listinfo/k
> 
> 
> > ratos
> 

I'd like to implement the final version as soon as possible.

Best regards
Janosch



More information about the Kratos mailing list