Optimization of Fem Models by Stochastic Methods *

Abstract

Stochastic shape optimization methods in combination with general numeric field computation method like FEM offer the most universal approach in automated design optimization (AOD). The paper discusses methodology, characteristic features and behavior of two compromising methods, the Evolution Strategy and Simulated Annealing. The application to a non-linear magneto-static problem of technical significance is shown by minimizing the weight of a lifting magnet.