Campelo, Felipe, Guimarães, Frederico G., Ramírez, Jaime A., and Igarashi, Hajime
Subjects
ELECTROMAGNETIC devices, ALGORITHMS, APPROXIMATION theory, MATHEMATICAL optimization, PROBABILITY theory
Abstract
In this paper, we introduce an approach for the design of electromagnetic devices based on the use of Estimation of Distribution Algorithms (EDAs), coupled with approximation-based local search around the most promising solutions. The main idea is to combine the power of EDAs in the solution of hard optimization problems with the faster convergence provided by the local search using local approximations. The resulting hybrid algorithm is tested on a numerical benchmark problem. [ABSTRACT FROM AUTHOR]
MATHEMATICAL optimization, MAGNETIC resonance imaging, DIMENSIONAL analysis, MAGNETIC fields, ALGORITHMS, COMPARATIVE studies, APPROXIMATION theory
Abstract
The pole pieces of a permanent-magnet system for magnetic resonance imaging (MRI) are optimized for maximum magnetic field homogeneity in the imaging volume. Axisymmetric and nonaxisymmetric optimized pole piece designs based on the same geometrical parameterization are presented. The optimization algorithm demonstrated is based on field calculations by the three-dimensional finite-element method and takes into account the nonlinearity of the magnetic materials involved. The necessity of the nonaxisymmetric design in obtaining suitable field homogeneity over a large imaging volume is demonstrated through comparison of various pole piece designs for a novel biplanar permanent-magnet assembly with reduced pole dimensions as required in the development of an integrated linear accelerator and MRI system for real-time image-guided adaptive radiotherapy. The sensitivity of the field inhomogeneities to geometrical variations in the nonaxisymmetric design surface is explored and statistical parameters quantifying this sensitivity are approximated. [ABSTRACT FROM AUTHOR]
MATHEMATICAL optimization, ALGORITHMS, PARETO optimum, MAGNETICS, APPROXIMATION theory
Abstract
Many real world optimization problems turn out to be multi-objective optimization problems revealing a remarkable number of locally optimal solutions corresponding to the chosen objective function. Therefore, it seems desirable to detect as many of those solutions with as few objective function calls as possible. A Niching Higher Order Evolution Strategy (NES) can successfully be applied to locate a large number of these local solutions during a single optimization run. Additionally, it turns out that all of these solutions can be found next to the front of non-dominated solutions. Therefore, evaluating more than one objective function (in parallel or in series) yields a good approximation of the Pareto-optimal front. The proposed method will be tested against several test functions and then applied to the solution of a magnetic shunting problem. [ABSTRACT FROM AUTHOR]