Keywords: structural optimization, parallel computing, evolutionary algorithm.

In this paper the use of parallel computing in the structural optimization which requires a large amount of computational time is presented. The parallel computing techniques are combined with evolutionary algorithm methods which are used for solving optimization models. The software package called Evolutionary Optimization System (EOS) [1], which was enriched with the parallel computing module, was applied to solving a fairly complicated structural optimization problem.

The Evolutionary Optimization System is designed to solve single and multicriteria optimization problems for nonlinear programming models. The system is equipped with various evolutionary optimization methods. Any optimization problem can be easily introduced to the EOS by means of an ANSI C function. The Evolutionary Optimization System (EOS) and ANSYS, which is the finite element method (FEM) analysis program, were combined to create new possibilities for optimizing structures [2].

The idea of using parallel computing results from one of the characteristic features of evolutionary algorithms, namely, that natural distribution of computational tasks among various individuals within one generation is possible. This feature is a prerequisite for effective parallel computing [3]. For parallel computing some modification to the EOS was introduced [4]. This consisted in adding an external module which was built in a client-server architecture with the support of HTTP protocol. This parallel computing module includes the Parallel Computing Server (PCS) and the Parallel Computing Client (PCC). The module enables parallel computation on all accessible computers in a local network. No expensive multiprocessor computers are necessary.

The optimization of the mass of a railway vehicle structure was chosen as a numerical example. The optimization model was based on the parametrical FEM analysis of the structure studied. The FEM model of the structure consisted of 25,303 elements of shell63 type and 1,711 beam44 type. The FEM analysis was written in the APDL programming language, which makes batch processing possible. The optimization problem was to find the minimum weight of the structure with the von Mises equivalent stress as the constraint. As a result of optimization process the structure mass was considerably reduced. Due to parallel computing the computational time was three times shorter when compared with the calculations performed on the fastest single computer of the set.

Evolutionary Algorithms combined with Parallel Computing techniques are a new effective tool for solving complicated optimization problems in which the FEM analysis is used. Thus, the approach presented is likely to find wide application in industry. The EOS can be combined not only with the ANSYS software package but also with other CAE simulation tools (see [2,4]).

1

Osyczka A. Evolutionary Algorithms for Single and Multicriteria Design Optimization, Springer Verlag Phisica, Heidelberg, Berlin, 2001.

2

Osyczka A., Mrzyglod M., Evolutionary Optimization of Mechanical Structures in Computer Simulated Environment, 6th World Congress on Structural and Multidisciplinary Optimization, Rio de Janeiro, 2005.

3

Alba E.; Nebro A.J.; Troya J.M., Heterogeneous Computing and Parallel Genetic Algorithms, Journal of Parallel and Distributed Computing, Vol. 62 (9), pp1362-1385, 2002. doi:10.1006/jpdc.2002.1851

4

Mrzyglod M., Parametrical optimization of structures working in high-cycle load conditions, (in Polish), Cracow University of Technology, PhD Thesis, 2005.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £80 +P&P)