Non-intrusive reduced-order modelling for studying the dynamics of structures with uncertain geometries and material properties: a review
摘要
Structural systems, particularly composite structures, often exhibit uncertainties in material properties and geometry due to manufacturing variability. These uncertainties can significantly influence the dynamic response of structures, especially in scenarios involving resonance. While finite element (FE) models are widely used for structural analysis, their high computational cost limits their practicality in uncertainty quantification studies that require numerous simulations. Reduced-order models (ROMs) offer a computationally efficient alternative, enabling accurate approximations of structural behaviour with significantly reduced computational effort. This paper presents a comprehensive review of non-intrusive ROM techniques used to study the variability of structural dynamic responses under uncertainties in geometry and material properties. Unlike intrusive methods, non-intrusive ROMs do not require access to FE code, making them well-suited for industrial applications using commercial FE software. Both linear and nonlinear non-intrusive ROMs are reviewed and assessed in terms of accuracy, computational cost, and applicability.