127 / 2021-03-31 10:55:18

Multiphysics mode synthesis for vibro-acoustic coupled model reduction

Vibro-acoustic simulation; Fluid-structure interaction; Reduction-order modeling; Component mode synthesis; Multiphysics mode synthesis

A vibro-acoustic problem is a typical fluid-structure interaction (FSI) in noise and vibration societies [1, 2]. The coupled physics model reduction is then essential to effectively handle the problem. In the linear vibro-acoustic problem, its reduced-order modelling (ROM) could be conventionally conducted by the modal projection. It is categorized as two main schemes depending on the types of mode. One is based on the coupled modes computed by the coupled eigenvalue problem. This is the simplest approach in the mode superposition manner, but solving the coupled eigenvalue problem with large DOFs and non-symmetric matrices is quite demand. For the reason, synthesing mono physics modes computed by eigenvalue problems in separate physics domains has been widely used. Using the interface constratint modes is a conventional way to synthesize the mono physics modes [1, 2]. This approach provides accurate reduced matrices in the low frequency range, but it may cause relatively large error in the mid and high frequency ranges. It is because using the interface constraint mode that reflects the interface behaviors only may not be enough to describe the coupled modes of fluid and structure inside the domains. A robust multiphysics mode synthesis, known as strongly coupled model reduction, has been proposed to handle this issue [3, 4]. A two-step sequential projection from structure to fluid is a main idea of the multiphysic mode synthesis. It allows the structural modal energy transform to fluid domain, and then the fluid modes could be updated. The newly derived transformation matrix also satisfies the matrix orthogonal condition both structure and fluid domains, and thus one can expect a good sparsity of the final reduced matrices. The multiphysic mode synthesis has been developed to the well known displacement-pressuere (u, p) and the disaplcement-pressure-potential (u, p, phi) formulations [1, 2]. In this presentation, those are introduced, and the performance (both accuracy and efficiency) is investigated with some numerical examples.