feMEMS: parametric optimization using semi-lumped parameter models

The design of MEMS devices usually involves several Finite Element (FE) analyses, each with different parameters (e.g. beam thicknesses) to tune the characteristics of the sensor. This trial-and-error procedure is tedious and time-consuming, even more so considering that typical FE models may count 100k-1M degrees of freedom. For this reason, our group developed a Matlab code, feMEMS (see the old GUI on the left), to model gyroscopes with beam elements and lumped masses, which has been successfully used for parametric optimization (average iteration time: 0.5s).

In this thesis, the candidate will have to reorganize feMEMS using an object-oriented framework to develop new features, possibly integrating and/or making it compatible with other tools used inside our group (e.g. yaFEc). Possible extensions include the implementation of von Karman beams, (special) shell elements, and sub-structuring techniques (depending on the candidate’s commitment). Code and features will be tested on a real MEMS layout to be optimized.