Substructuring based parametric reduced order modelling for large-scale dynamical systems containing viscoelasticity with application to bonded assemblies

Abstract

Adhesive bonding plays an increasingly important role in mechanical systems to connect different components and multi-material assemblies, and usually exhibits viscoelasticity. To accurately predict their structural dynamic behavior, finite element models are often used. When large systems are considered, these models result in high computational costs. To reduce the computational costs, substructuring combined with model order reduction can be applied. In this context, the well-known Craig-Bampton method is often used. However, due to the frequency-dependency of viscoelastic material properties, it cannot be directly applied to assemblies with adhesive joints. To enable its use for viscoelastic substructures, this work proposes several improvements to the Craig-Bampton method. Moreover, since the dynamic performance of adhesive bonded structures can strongly depend on the viscoelastic properties, the efficient assessment of different material parameters is highly desirable. Therefore, this work also extends the improved Craig-Bampton method towards parametric model order reduction in which a global reduction basis is constructed by singular value decomposition and interpolation of sampling-based local bases. Two numerical examples of adhesive bonded systems are employed to demonstrate the accuracy and computation time reduction which can be achieved with the proposed approach.