Large scale thermal-solid coupling analysis using inexact balancing domain decomposition

In this research, a system of thermal-solid coupling analysis is developed with the implementation of Inexact Balancing Domain Decomposition with a diagonal scaling (IBDD-DIAG) in both thermal and solid analysis. The IBDD-DIAG is an improved version of Balancing Domain Decomposition (BDD), where an...

Cijeli opis

Bibliografski detalji
Glavni autori: Mukaddes, Abul Mukid Mohammad, Shioya, Ryuji
Format: Članak
Izdano: BRAC University 2011
Teme:
Online pristup:http://hdl.handle.net/10361/891
Opis
Sažetak:In this research, a system of thermal-solid coupling analysis is developed with the implementation of Inexact Balancing Domain Decomposition with a diagonal scaling (IBDD-DIAG) in both thermal and solid analysis. The IBDD-DIAG is an improved version of Balancing Domain Decomposition (BDD), where an incomplete factorization based parallel direct method is employed to solve a coarse space problem, and the diagonal-scaling is employed to precondition local fine space problems instead of the Neumann-Neumann preconditioner. The developed system performed heat conductive analysis to have temperature distributions in solid models and then performed the structural analysis to see deformation or expansion due to temperature differences. Both of the analyses employed the Hierarchical Domain Decomposition Method (HDDM) with parallel IBDD-DIAG. It is shown that the iterative procedure converges rapidly and the convergence is independent of the number of subdomains, namely, numerical scalability is satisfied. The present system is implemented on massively parallel processors and succeeds in solving a thermal-solid coupling problem of 12 millions of nodes.