Abstract
Currently, vibroacoustic problems can be solved using a wide range of numerical techniques. In the low-frequency range, element-based deterministic methods, such as the Finite Element Method (FEM) and Boundary Element Method (BEM) are regularly employed to define the structural and acoustic domains, respectively. The fully coupled FEM-BEM is a classic, vastly popular method. In the high-frequency range probabilistic methods, such as Statistical Energy Analysis, tend to be more efficient and produce more reliable results. Although new techniques are becoming available (e.g. Hybrid FE-SEA Method), the characterisation of the mid-frequency behaviour still poses some challenges, as the computational cost of element-based techniques is often prohibitive, and the modal density is not sufficiently high for statistical approaches to be applicable. This paper discusses an approach aimed at improving the efficiency of the classic FEM-BEM method and potentially extending its usability to the mid-frequency band, specifically in the context of space-craft structural design. The iterative coupling between Craig-Bampton reduced finite element models and BEM is considered as an alternative to directly solving the FEM-BEM coupled equation, allowing the use of efficient procedures for either domain separately. A pre-process enabling the method’s computational implementation is presented, which is based on a manipulation of the reduced mass and stiffness matrices. It is used to allow the application of a distributed load to a Craig-Bampton condensed structure, while mitigating the need to retain a large number of physical degrees of free-dom. The efficiency of the aforementioned matrix modification procedure is compared to that of performing a full Craig-Bampton reduction, and its cost is expressed in terms of floating point operations. An iterative coupling scheme is used on a test-case structure for both a full physical model and a reduced one to verify the concept, and check whether convergence is susceptible to initial conditions, such as the shape of the acoustic field. Finally, the perturbation of the condensed matrices is shown to produce results consistent with those for the full physical model, while substantially reducing the computational effort required for the simulation.