Computationally efficient dynamic modeling of MEMS

Traditional modeling work in MEMS includes simplified PDE/ODE formulation, based on physical principles, and Finite Element Analysis. More recently, reduced order modeling techniques using Krylov subspace decomposition have been proposed in the context of nodal analysis. This modeling technique makes it possible to predict the dynamic behavior of more complex MEMS, but the computational engine is still a traditional cubic order solver. In this paper we apply a new modeling approach for complex MEMS based on a linear O(n+m) (n- number of bodies, in - number of constraints) solver for rigid multibody dynamics. As direct applications, we present simulation and experimental results of models for thermally driven MEMS actuators, compared against established simulation tools, namely FEA (Intellisuite), AUTOLEV, and SUGAR 3.0.