Abstract.
The unconstrained global programming problem is addressed using a multi-start, multi-algorithm infrastructure, in which different algorithms compete in parallel for a contribution towards a single global stopping criterion, denoted the unified Bayesian global stopping criterion. The different algorithms compete in parallel on a cluster of up to 128 machines.
The competing algorithms are motivated by the observation that no single (global) optimization algorithm can consistently outperform all other algorithms when a large set of problems in different classes is considered. The Bayesian stopping criterion is based on the single, mild assumption that the probability of each algorithm in the infrastructure to converge to the global optimum is at least as large as the probability of convergence to any local minimum.
Numerical results are presented for both analytical test functions, and composite structures modelled using the finite element method. In the finite element model, we use a new, highly accurate, flat shell finite element with drilling d.o.f. This new element exploits coarse, yet highly accurate finite element meshes.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Author information
Authors and Affiliations
Additional information
Received August 31, 2001 Revised manuscript received November 16, 2001
RID="*"
ID="*"Presented as paper 148 at the Fourth World Congress of Structural and Multidisciplinary Optimization, held in Dalian, China, May 2001
Rights and permissions
About this article
Cite this article
Groenwold, A., Hindley, M. Competing parallel algorithms in structural optimization. Struct Multidisc Optim 24, 343–350 (2002). https://doi.org/10.1007/s00158-002-0246-7
Issue Date:
DOI: https://doi.org/10.1007/s00158-002-0246-7