Model order reduction is known to work reliably for MEMS simulations and can beused to speed up transient and harmonic computations tremendously. However, using standard model reduction techniques the generated reduced models are only able to compute one special case. If any changes to the model are made, e.g. boundary conditions are changed, a new reduction step is necessary. To generate compact models for system simulation it is indispensable, that the reduced model leaves some parameter for the engineer to set. In this paper we describe an approachfor a parameter preserving model reduction, e.g. a reduction technique that preserves parameters defined in the original model during the reduction step. To demonstrate this technique we chose the example of an anemometer where we compute conductive and convective heat flow, modeled using a forced convection approach and dicretized with finite elements. We show that we are able to generate reduced models of order typically about 50 that leave the fluid speed as an input parameter and that show very good accuracy compared to the original model. While the resources needed to generate this reduced model are comparable to theresources needed for a steady state solution of the original model, we achive tremendous speed increase when comparing the transient and harmonic simulations of the reduced and the full model.
Journal: TechConnect Briefs
Volume: 3, Technical Proceedings of the 2005 NSTI Nanotechnology Conference and Trade Show, Volume 3
Published: May 8, 2005
Pages: 684 - 687
Industry sectors: Advanced Materials & Manufacturing | Sensors, MEMS, Electronics
Topics: Informatics, Modeling & Simulation