Distributed multi-objective optimization methods for shape design using evolutionary algorithms and game strategies
Year: 2012
Type: dissertation
Abstract: Leskinen, Jyri Distributed Multi-Objective Optimization Methods for Shape Design using Evolutionary Algorithms and Game Strategies Jyvaskyla: University of Jyvaskyla, 2012, 86 p.(+included articles) (Jyvaskyla Studies in Computing ISSN 1456-5390; 15 ) ISBN 978-951-39-4843-6 (nid.) ISBN 978-951-39-4844-3 (PDF) Finnish summary Diss. This research investigates innovative newmethods for increasing the efficiency of shape design optimization are studied. Considerable improvements on algorithmic convergence can be achieved by splitting the geometry of a single-objective problem into an equivalent multi-objective problem and solving it by using competitive Nash games from the field of game theory. Further efficiency improvements can be achieved using a “distributed one-shot” method introduced in this work by decomposing the computational domain and solving both the geometry and domain simultaneously in a single “global” Nash game. Themethod is inherently parallel and is suitable for distributed platforms. Optimization is done using evolutionary algorithms allowing global and non-smooth optimization. Implementation of modern graphics processing units for the introduced methods is also studied. The methods are evaluated on academic model problems from the field of computational fluid dynamics, available in the Finnish Design Test Case Database. The numerical results presented validate the new approach and open the door to more complex optimization problems of industrial interest.
Author: Jyri Leskinen
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: Jyväskylän Yliopisto
OpenAlex ID: W2252227343