Plateforme Sophemis

For more than ten years now, the developments of multi-physics models and optimisation methodologies have been gathered together on the ‘SOPHEMIS‘ platform (Optimisation supervisor of electrical machines in their environment).

Version 1 of Sophemis used to propose strategies with experimental design methodologies [Thesis by S. Vivier – 2002].

Currently, version 2 offers many multi-objective optimisation methodologies based, for example, on genetic algorithms [Thesis by F. Moussouni – 2009], multi-granularity aspects (space mapping) and discrete variables [Thesis of Vu Tran – 2009], multi-level methodologies [Thesis by S. Kreawan – 2008], multi-physics [Thesis by A. Berbecea – 2012] and multi-level multi-granularity [Thesis by R. Ben-Ayed – 2012]: development of multi-objective and multi-level space mapping. This development was achieved with the help of more than ten theses as well as the recruitment of a computer engineer. Furthermore, this platform uses distributed computing on 32 cores.

The models associated with this platform use:

• Finite element models: coupling with commercial tools as well as with ‘code_Carmel’ developed by the ‘modelling’ team;
• Static (Matlab) and temporal (Simulink) analytical models.

Different inter-team work can be cited: the platform is interfaced with the code Carmel of the ‘modelling’ team and is used by the ‘network’ team (thesis by P. Pankovits, SNCF). This platform is also used by all the doctoral students.