Decomposition and parallelization of strongly coupled fluid–structure interaction linear subsystems based on the Q1/P0 discretization. (September 2016)

• Record Type: Journal Article
• Authors: Toma, Milan; Oshima, Marie; Takagi, Shu
• Is Part Of: Computers & structures. Volume 173(2016)
• Page Start: 84
• Page End: 94
• Publication Date: 2016-09
• Subjects: Decoupling -- FSI applications -- Condition number -- Parallel computing; Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2016.06.001
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: A domain decomposition based method for solving a coupled FSI problems is proposed. P0 related DOFs are moved to the interface subdomain which provides a new advantage in parallelization of the algorithm. Step by step parallelization of the FSI decomposition and solution is presented. Abstract: Traditional fluid–structure interaction techniques are based on arbitrary Lagrangian–Eulerian method, where strong coupling method is used to solve fundamental discretized equations. After discretization, resulting nonlinear systems are solved using an iterative method leading to set of linear subsystems. The resulting linear subsystem matrices are sparse non-symmetric indefinite with small or zero diagonal entries. The small pivots of these matrices may render the incomplete LU factorization unstable or inaccurate. To improve that, this paper proposes an approach that can be used together with a stabilization and offers an additional feature to be used to gain an advantage in its parallelization.