Parallel Evolutionary Algorithm for Computationally Expensive Single Criteria Design Optimization

Abstract:

The paper presents an approach to design optimization using parallel evolutionary algorithms. The only use of a simple evolutionary algorithm in order to generate the optimal solution for complex problems can be ineffective due to long calculation time. Thus a tournament evolutionary algorithm (EA) and a parallel computation method are proposed and used. The proposed EA does not require an analysis of the optimization model for each potential solution from evolutionary populations. The second element of the method consists in parallel running of evolutionary algorithms using multi-threads approach. The experiments were carried out for many different single design optimization problems and two of them are presented in the paper. The first problem considers a task of robot gripper mechanism optimization and the second one deals with the optimization of a shaft based on Finite Element Method analysis. From the generated results it is clear that proposed approach is a very effective tool for solving fairly complicated tasks considering both the accuracy and the time of calculation.

• Periodical: Applied Mechanics and Materials (Volume 555)
• Pages: 586-592
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.555.586
• Online since: June 2014
• Authors: Stanisław Krenich*
• Keywords: Computationally Expensive Functions, Computer Aided Design, Evolutionary Algorithm (EA), FEM-Based Optimization, Parallel Computation

References

[1] Ansys - Theory Manual , 12th Edition, Ansys Inc., (2001).

[2] Burczynski T., Dlugosz A., Kus W., Parallel Evolutionary Algorithms in Shape Optimization of Heat Radiators. Journal of Theoretical and Applied Mechanics 44, 2, pp.351-366, Warsaw, (2006).

[3] Krenich S., Genetic Algorithms in Parametrical Optimization of Robot Gripper Mechanisms, Department of Mechanical Engineering. Cracow University of Technology (2002).

[4] Lis J., Lis M., Self-adapting Parallel Genetic Algorithm with Dynamic Mutation Probability, Crossover Rate and Population Size. [In: ] Proceedings of the First Polish National Conference on Evolutionary Computing, Arabas J., Ed., Warsaw Polytechnic, pp.324-329, (1996).

[5] Miki M., Hiroyasu T., Hatanaka K., Parallel Genetic Algorithms with Distributed-Environment Multiple Population Scheme. [In: ] The 3rd World Congress on Structural and Multidisciplinary Optimization, Buffalo 17-22 May, USA, (1999).

[6] Mrzyglod M., Zielinski AP, Parametric structural optimization with respect to the multiaxial high-cycle fatigue criterion, Structural and Multidisciplinary Optimization 33 (2), 161-171, (2007).

DOI: 10.1007/s00158-006-0045-7

[7] Osmera P., Lacko B., Peter M., Parallel Evolutionary Algorithms. Proceedings IEEE International Symposium on Computational Intelligence in Robotics and Automation, (2003).

DOI: 10.1109/cira.2003.1222193

[8] Osyczka A., Krenich S., Evolutionary Algorithms for Global Optimization, Chapter in: J. Pinter (Ed. ) Global Optimization - Selected Case Studies, Kluwer Academic Publishers, (2007).

[9] Osyczka A.: Evolutionary Algorithms for Single and Multicriteria Design Optimization. Springer -Verlag Physica, Berlin Heilderberg, (2002).

[10] Syrek M., Małopolski W.: Analiza wpływu zastosowania maszyn wirtualnych na wydajność obliczeń współbieżnych, Logistyka 6/2011, str. 3601-3610, (2011).