MEMS optimization incorporating genetic algorithms

Abstract

Micromechanical sensors are routinely simulated using finite element software. Once a structure has ben proposed, various parameters are optimized using experience, intuition, and trial-and-error. However, using proven finite element modeling coupled with a genetic algorithm (GA), optimal designs can be 'evolved' using a hands-free approach on a workstation. Once a problem is defined, the sole task required of the designer is the specification of a mathematical objective function expressing the desired properties of the sensor; the sensor geometry that maximizes the given function is then synthesized by the algorithm. We have developed an optimization tool and have applied it to the design of tuning fork gyroscopes (TFG). In this paper, we demonstrate how a TFG was optimized using GA's. TFG suspension beam lengths were adjusted through the robust search technique, which is resistant to trapping in local maxima. Desired vibration mode order and mode frequency separations were governed by the objective function as specified by the designer. This multidimensional nonlinear optimization problem had a solution space of over eight million possible designs. Industry-standard mechanical computer-aided engineering tools were integrate along with a GA toolbox and a web-based control interface. Designs offering reduced vibration sensitivity and increased sensor dynamic range have been produced. A tenfold decrease in total sensor optimization time has been documented, resulting in reduced development time.