Genetic Algorithms for Multicriteria Shape Optimization of Induction Furnace

0438640 - ÚT 2015 RIV US eng C - Konferenční příspěvek (zahraniční konf.)
Kůs, Pavel - Mach, F. - Karban, P. - Doležel, Ivo
Genetic Algorithms for Multicriteria Shape Optimization of Induction Furnace.
NUMERICAL ANALYSIS AND APPLIED MATHEMATICS (ICNAAM 2012), VOLS A AND B. Melville: AMER INST PHYSICS, 2012, s. 2344-2347. 1479. ISBN 978-0-7354-1091-6. ISSN 0094-243X.
[International Conference of Numerical Analysis and Applied Mathematics (ICNAAM). Kos (GR), 19.09.2012-25.09.2012]
Institucionální podpora: RVO:61388998
Klíčová slova: optimization * coupled problems * hp-FEM
Kód oboru RIV: BA - Obecná matematika

In this contribution we deal with a multi-criteria shape optimization of an induction furnace. We want to find shape parameters of the furnace in such a way, that two different criteria are optimized. Since they cannot be optimized simultaneously, instead of one optimum we find set of partially optimal designs, so called Pareto front. We compare two different approaches to the optimization, one using nonlinear conjugate gradient method and second using variation of genetic algorithm. As can be seen from the numerical results, genetic algorithm seems to be the right choice for this problem. Solution of direct problem (coupled problem consisting of magnetic and heat field) is done using our own code Agros2D. It uses finite elements of higher order leading to fast and accurate solution of relatively complicated coupled problem. It also provides advanced scripting support, allowing us to prepare parametric model of the furnace and simply incorporate various types of optimization algorithms.
Trvalý link: http://hdl.handle.net/11104/0242068