Minimizing the variation in performance by optimizing the design parameters

We present a design algorithm that minimizes the variation in performance by optimizing the design parameters of micro electro mechanical systems (MEMS). Due to variations from causes such as fabrication processing, packaging, actuation signals, and external disturbances, the true performance can be greater than 100% different than what may be predicted. Given the range and uncertainties of the design parameters, our algorithm finds a set of parameters that minimizes the variation in performance. That is, the resulting design is least sensitive to process variations in geometry and material properties, yet still meets the desired performance specifications. Prior work in this area by others includes a statistical framework for quantifying uncertainties through probability densities. The difference in our method is that it only reduces the variation bounds of desired performance. Due to its lack of distribution detail, our algorithm is computationally efficient. For test cases, we demonstrate the reduction of performance variation of an AFM cantilever and a comb-drive resonator. Our algorithm will be made available online through SugarCube.