| Summary: | Driven by economic, environmental and ergonomic concerns, porous composites are increasingly being adopted by the aeronautical and structural engineering communities for their improved physical and mechanical properties. Such materials often possess highly heterogeneous material descriptions and tessellated/complex geometries. Deploying commercially viable porous composite structures necessitates numerical methods that are capable of accurately and efficiently handling these complexities within the prescribed design iterations. Classical numerical methods, such as the Finite Element Method (FEM), while extremely versatile, incur large computational costs when accounting for heterogeneous inclusions and high frequency waves. This often renders the problem prohibitively expensive, even with the advent of modern high performance computing facilities.
Multiscale Finite Element Methods (MsFEM) is an order reduction strategy specifically developed to address such issues. This is done by introducing meshes at different scales. All underlying physics and material descriptions are explicitly resolved at the fine scale. This information is then mapped onto the coarse scale through a set of numerically evaluated multiscale basis functions. The problems are then solved at the coarse scale at a significantly reduced cost and mapped back to the fine scale using the same multiscale shape functions. To this point, the MsFEM has been developed exclusively with quadrilateral/hexahedral coarse and fine elements. This proves highly inefficient when encountering complex coarse scale geometries and fine scale inclusions. A more flexible meshing scheme at all scales is essential for ensuring optimal simulation runtimes.
The Virtual Element Method (VEM) is a relatively recent development within the computational mechanics community aimed at handling arbitrary polygonal (potentially non-convex) elements. In this thesis, novel VEM formulations for poromechanical problems (consolidation and vibroacoustics) are developed. This is then integrated at the fine scale into the multiscale procedure to enable versatile meshing possibilities. Further, this enhanced capability is also extended to the coarse scale to allow for efficient macroscale discretizations of complex structures.
The resulting Multiscale Virtual Element Method (MsVEM) is originally applied to problems in elastostatics, consolidation and vibroacoustics in porous media to successfully drive down computational run times without significantly affecting accuracy. Following this, a parametric Model Order Reduction scheme for coupled problems is introduced for the first time at the fine scale to obtain a Reduced Basis Multiscale Virtual Element Method. This is used to augment the rate of multiscale basis function evaluation in spectral acoustics problems. The accuracy of all the above novel contributions are investigated in relation to standard numerical methods, i.e., the FEM and MsFEM, analytical solutions and experimental data. The associated efficiency is quantified in terms of computational run-times, complexity analyses and speed-up metrics.
Several extended applications of the VEM and the MsVEM are briefly visited, e.g., VEM phase field Methods for brittle fracture, structural and acoustical topology optimization, random vibrations and stochastic dynamics, and structural vibroacoustics.
|
|---|