An Hp-Adaptive Finite Element Procedure For Fluid-Structure Interaction In Fully Eulerian Framework

This thesis attempts to implement a fully automatic hp-adaptive finite element procedure for fluid-structure interaction (FSI) problems in two dimensions. This work hypotesizes the efficacy of Fully Eulerian framework of FSI in hp-adaptivity on an a posteriori error estimator and adaptation for mini...

Full description

Saved in:
Bibliographic Details
Main Author: Abas, Mohamad Aizat
Format: Thesis
Language:English
Published: 2013
Subjects:
Online Access:http://eprints.usm.my/46298/1/Mohamad%20Aizat%20Bin%20Abas24.pdf
http://eprints.usm.my/46298/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This thesis attempts to implement a fully automatic hp-adaptive finite element procedure for fluid-structure interaction (FSI) problems in two dimensions. This work hypotesizes the efficacy of Fully Eulerian framework of FSI in hp-adaptivity on an a posteriori error estimator and adaptation for minimization of error in energy norm. Automatic mesh adaptation over triangular elements is handled by red-green-blue (RGB) refinement method. An effective mesh adaptivity to avoid excessive growth of unknowns is also addressed. Since the hp-method uses high order polynomials as approximation functions, the resulting system matrices are less sparse leading to the notion of FSI computation with parallelism. The parallel hp-adaptive computation is assessed with the conventional uniform and h refinement on a number of benchmark test cases. Subsequently, the efficacy of the fully Eulerian framework is compared to the well known Arbitrary Lagrangian Framework( ALE) for two different material models, namely, the St. Venant Kirchoff and the Neo-Hookean models. It was found that the fully Eulerian framework provides accurate FSI predictions for large deformation without need of frequent remeshing. The hp-adaptive method was also found to be a viable approach in obtaining accurate solutions without much compromise in computer memory and time. Furthermore, the integration of parallelism is successful in reducing the computation time by up to two orders of magnitude relative to the serial solver. For the comparisons between the ALE and the fully Eulerian frameworks, the computed solutions in all test cases are observed to be in agreement with each other.