Résumé
MEMS are now entering an industrial era. Fabrication processes are usually based on integrated circuits ones, but they can also include non-electronical parts such as mechanical or optical ones. As multiphysics applications, MEMS appear to be very complex and hard to be accurately modelized, which lead their industrialization to be delayed until today. Now, they have to face the aggressivity of semiconductor market, and yield management is a critical success factor for this new technology. In this dissertation, we detail the development of a software for MEMS parametric yield statistical optimization which can be easily included in integrated circuits or MEMS conception flow. By optimizing layout parameters, it allows to reduce MEMS performances variations and dispersion which are due to unavoidable technological parameters fluctuation during the different process steps, and lead to yield losses.