Parallelized structural topology optimization for eigenvalue problems

摘要

A parallel topology optimization method is proposed to deal with large-scale structural eigenvalue-related design problems. While existing parallel methods have been employed mainly for finite element analysis, the present work deals with the entire parallel process of analysis and topology optimization using a parallel processing technique. In the framework of the physical domain decomposition method, the sensitivity analysis and design variable update are performed independently in each subdomain with minimum data communication among subdomains. A successive estimation strategy over design iterations is proposed to expedite the design optimization process. The preconditioned conjugate gradient method and the subspace iteration method are used as parallel solvers. Several numerical examples are considered to show the scalability and effectiveness of the present parallel approach in handling large-scale design problems. (C) 2003 Published by Elsevier Ltd.