Scicos + Modelica for multi-body-simulations
p....@repower.de
I would like to ask a few questions concerning the use of Scicos and
the Modelica language for modeling + simulating dynamic
multi-body-systems.
I am working as an engineer in the loads&simulation group at a German
manufacturer of multi-megawatt wind turbines and actually searching for
a new, more sophisticated software for the dynamic simulation of our
turbines and found the Scicos-solution as one favourable way offering
the possibilty to build a comprehensive multiphysical model of a
complete tubine including structural dynamics, electrical and
controller systems - that's my hope.
On the Scicos internet page (www.scicos.org) under 'Related Projects'
it is mentioned that Scicos supports Modelica-models as 'implicit
blocks' (?) and comes with its own integrated Modelica-Compiler. -->
1.) Is it possible to use the free mechanical libraries from the
Modelica home page under Scicos to build up the structural model as
multi-body-system ?
--> is there any direct graphical support for these libraries or is it
just indirectly possible by using a sort of black box in the model
graph ?
2.) Is it possible to use the Open-Modelica-Compiler alternatively as
'second source' in addition to Scicos build in ?
The list of Scicos' features states a 'model-conversion to C-code'. -->
3.) Does this mean, that it is possible to build a Modelica model of
the complete tubine within Scicos (mechanical + electrical + controller
system) then export / convert it completely into C-code and hereafter
link it with a solver such that a self-standing, executable program can
be created / compiled using some standard C-compiler ?
Creating such a 'model-executable' would be a very strong feature as we
need a program which can run on a Cluster simultaneously on serval
machines (each time started with a different set of input parameters).
I hope that some Scicos expert can answer my questions. Any information
regarding the use of Scicos in the field of multi-body-simulation will
be appreciated.
Kind regards, Patrick Rix
Ramine Nikoukhah
p....@repower.de wrote:
> Hello Scicos developers & users / !
>
> I would like to ask a few questions concerning the use of Scicos and
> the Modelica language for modeling + simulating dynamic
> multi-body-systems.
> I am working as an engineer in the loads&simulation group at a German
> manufacturer of multi-megawatt wind turbines and actually searching for
> a new, more sophisticated software for the dynamic simulation of our
> turbines and found the Scicos-solution as one favourable way offering
> the possibilty to build a comprehensive multiphysical model of a
> complete tubine including structural dynamics, electrical and
> controller systems - that's my hope.
>
> On the Scicos internet page (www.scicos.org) under 'Related Projects'
> it is mentioned that Scicos supports Modelica-models as 'implicit
> blocks' (?) and comes with its own integrated Modelica-Compiler. -->
>
> 1.) Is it possible to use the free mechanical libraries from the
> Modelica home page under Scicos to build up the structural model as
> multi-body-system ?
Modelica compiler incorporated in the current version of Scilab/Scicos
is rudimentary and does not cover all the language features. In particular
it does not handle inheritance. Because of this, most blocks from Modelica
librairies don't work unless they are "flattened". For someone familiar
with the Modelica language, this is pretty straightforward in general.
We are currently working on a much more sophisticated compiler. The
first version will be released with Scilab 5 next year. It will hopefully
cover most, if not all, language features. Note that the "flattening" of the
models can also be realized using freemodelica.
> --> is there any direct graphical support for these libraries or is it
> just indirectly possible by using a sort of black box in the model
> graph ?
Scicos provides graphical display of blocks including implicit blocks.
It is also possible to translate Modelica icons into Scilab graphics
automatically. The code exists and will be made available in Scilab 5.
>
> 2.) Is it possible to use the Open-Modelica-Compiler alternatively as
> 'second source' in addition to Scicos build in ?
The compiler is not part of Scilab/Scicos executable. It is called through
a host command and the result is a C file. Because of that, it would be
easy to use any compiler. However, the resulting C file must have the
appropriate calling sequence (scicos block function). This would require
a wrapper in case Scicos is not a target of the compiler.
>
> The list of Scicos' features states a 'model-conversion to C-code'. -->
>
> 3.) Does this mean, that it is possible to build a Modelica model of
> the complete tubine within Scicos (mechanical + electrical + controller
> system) then export / convert it completely into C-code and hereafter
> link it with a solver such that a self-standing, executable program can
> be created / compiled using some standard C-compiler ?
It is almost possible except for one thing. When using Modelica blocks,
the resulting scicos block has an internally implicit dynamics. This
means that it contains a system of DAE (differential algebaric
equations). The
code generator only works for ODEs for which it uses fixed step solvers.
What needs to be done is to force the generation of an explicit block
(ODE based) whenever possible, in which case the C generator will work.
And find a reliable fix step solver for DAEs for cases where this cannot
be done. I have not been able to find any....
>
> Creating such a 'model-executable' would be a very strong feature as we
> need a program which can run on a Cluster simultaneously on serval
> machines (each time started with a different set of input parameters).
>
I hope this answers some of your questions. We are aware of this limitations
that is why we are working hard on the new version.
Cheers
Ramine
--
Ramine Nikoukhah
INRIA, Rocquencourt BP 105, 78153 Le Chesnay Cedex, France
Email: ramine.n...@inria.fr
Tel: (33) 1-39-63-53-08 Fax: (33) 1-39-63-57-86
Matthias Zenker
Ramine Nikoukhah wrote:
> The compiler is not part of Scilab/Scicos executable. It is called through
> a host command and the result is a C file. Because of that, it would be
> easy to use any compiler. However, the resulting C file must have the
> appropriate calling sequence (scicos block function). This would require
> a wrapper in case Scicos is not a target of the compiler.
Will there be any improvement concerning the C-compiler in the new
Scicos version?
AFAIR, at the moment only VC++ ist supported. Are there plans to
support other (free) compilers?
Matthias
Ramine Nikoukhah
I am going to rely on Scilab. I have been asking for a free built-in
compiler
and I hope the Scilab team will include one for version 5 of Scilab.