Parallel BDD-based monolithic approach for acoustic fluid-structure interaction

Parallel BDD-based monolithic algorithms for acoustic fluid-structure interaction problems are developed. In a previous study, two schemes, NN-I + CGC-FULL and NN-I + CGC-DIAG, have been proven to be efficient among several BDD-type schemes for one processor. Thus, the parallelization of these schemes is discussed in the present study. These BDD-type schemes consist of the operations of the Schur complement matrix-vector (Sv) product, Neumann-Neumann (NN) preconditioning, and the coarse problem. In the present study, the Sv product and NN preconditioning are parallelized for both schemes, and the parallel implementation of the solid and fluid parts of the coarse problem is considered for NN-I + CGC-DIAG. The results of numerical experiments indicate that both schemes exhibit performances that are almost as good as those of single solid and fluid analyses in the Sv product and NN preconditioning. Moreover, NN-I + CGC-DIAG appears to become more efficient as the problem size becomes large due to the parallel calculation of the coarse problem.