Robust topology optimization for structures under bounded random loads and material uncertainties. (August 2021)

• Record Type: Journal Article
• Authors: Bai, Song; Kang, Zhan
• Is Part Of: Computers & structures. Volume 252(2021)
• Publication Date: 2021-08
• Subjects: Bounded randomness -- Random field -- Ellipsoid convex model -- Robust optimization -- Topology optimization; Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2021.106569
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: Robust topological design under bounded loads and material uncertainty is studied. A new ellipsoid convex model for bounded random loads is proposed. Loads are assumed uniformly distributed according to the maximum entropy theorem. Spatially correlated material uncertainty is modeled as a random field by EOLE. Polynomial chaos expansion with sparse grid rule is used in uncertainty analysis. Abstract: Design optimization considering uncertainties arising from different sources has been one of the hotspots in the area of structural optimization. This paper presents a robust topology optimization method for structures with bounded loads and spatially correlated material uncertainties. We first develop a new model for random structural loads with bounded nature, in which the uncertain load components are assumed to be uniformly distributed within a convex set represented by an ellipsoid model. This model is then combined with the random field discretized by means of the expansion optimized linear estimation to provide uncertainty modelling of structures exhibiting both load and material uncertainties. The robust topological design problem is formulated as a bi-criteria optimization problem for minimizing the mean value and standard deviation of the structural compliance under these uncertainties. The polynomial chaos expansion, in conjunction with the sparse grid quadrature rule, is employed for calculating the statistical moments of the structural responses. The adjoint sensitivity analysis scheme is derived, and a gradient based optimizer is employed for updating the design variables. Numerical examples show that the structural designs obtained by the proposed robust topology optimization method exhibit higher robustness against uncertainties as compared with their deterministic counterpart.