Optimization of an electromagnetic linear actuator using a network and a finite element model
Abstract
Model based design optimization leads to robust solutions only if the statistical deviations of design, load and ambient parameters from nominal values are considered. We describe an optimization methodology that involves these deviations as stochastic variables for an exemplary electromagnetic actuator used to drive a Braille printer. A combined model simulates the dynamic behavior of the actuator and its nonlinear load. It consists of a dynamic network model and a stationary magnetic finite element (FE) model. The network model utilizes lookup tables of the magnetic force and the flux linkage computed by the FE model. After a sensitivity analysis using design of experiment (DoE) methods and a nominal optimization based on gradient methods, a robust design optimization is performed. Selected design variables are involved in form of their density functions. In order to reduce the computational effort we use response surfaces instead of the combined system model obtained in all stochastic analysis steps. Thus, MonteCarlo simulations can be applied. As a result we found an optimum system design meeting our requirements with regard to function and reliability.
 Publication:

Active and Passive Smart Structures and Integrated Systems 2011
 Pub Date:
 March 2011
 DOI:
 10.1117/12.885637
 Bibcode:
 2011SPIE.7977E..21N