A sequential metamodel-based method for structural optimization under uncertainty. (August 2020)

• Record Type: Journal Article
• Authors: Dutta, Subhrajit
• Is Part Of: Structures. Volume 26(2020)
• Page Start: 54
• Page End: 65
• Publication Date: 2020-08
• Subjects: Optimization under uncertainty -- Metamodel -- Polynomial chaos expansion -- Kriging -- Particle swarm optimization; Structural engineering -- Periodicals; 624.1
• DOI: 10.1016/j.istruc.2020.04.009
• Languages: English
• ISSNs: 2352-0124

Abstract

Highlights: Two optimization under uncertainty problems: robust design optimization, and reliability based design optimization problems are successfully implemented, verified and validated. Implementation of large-scale structural optimization test cases to show the robustness, efficiency, and versatility of the proposed optimization under uncertainty algorithm. Design of experiments are implemented in an adaptive way such that the region(s) of interest for objective and constraint functions are adequately sampled. Large-scale structural optimization under uncertainty problems are efficiently solved using preconditioning of geometric nonlinear numerical (finite element) solver. Abstract: Optimization under uncertainty (OUU) provides robust optimal design solutions for real engineering problems considering uncertainties. These OUU problems involves a costly inner loop uncertainty quantification, involving a computation-intensive numerical solver for large-scale real systems with significantly higher degrees of freedom. The current work is aimed at reducing this cost of computation in OUU. To this end, a sequential polynomial chaos expansion (PCE) and kriging based metamodel is used. This metamodel is later adopted to substitute the actual expensive true numerical model solver in the uncertainty analysis computation phase. Particle swarm optimization (PSO) is used for optimization, leveraging on the properties of stochastic search. The effectiveness of PCE-kriging metamodel combined with PSO is demonstrated for optimization of two transmission towers. It has been observed that the proposed metamodel-based approach for OUU of a 244 member large-scale tower provides significantly faster and accurate solutions.