Robust Design Optimization in Parallel Computing Environment

Abstract eng:
In engineering problems, randomness and uncertainties are inherent. Robust design procedures, formulated in the framework of multi-objective optimization, have been proposed in order to take into account the various sources of randomness and uncertainty. These design procedures require orders of magnitude more computational effort than the conventional analysis or the optimum design processes since a very large number of finite element analyses is required to be dealt. It is therefore an imperative need to exploit the capabilities of the computing resources in order to deal with this kind of problems. In particular, parallel computing is implemented at the level of the metaheuristic optimization, by exploiting the physical parallelization feature of the nondominated sorting evolution strategies method, as well as at the level of the repeated structural analyses required for assessing the behavioural constraints and for calculating the objective functions. In this study an efficient dynamic load balancing algorithm (DLBA) for the optimum exploitation of the available computing resources is proposed and, without loss of generality, DLBA is applied to the task of computing the desired feasible Pareto front. In such problems the computation of the complete Pareto front with feasible designs only, constitutes a very challenging task. The proposed algorithm achieves linear speedup factors and almost 100% speedup factor values with reference to the sequential procedure.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket