A parallel algorithm for unilateral contact problems. (15th October 2022)

• Record Type: Journal Article
• Authors: Guillamet, G.; Rivero, M.; Zavala-Aké, M.; Vázquez, M.; Houzeaux, G.; Oller, S.
• Is Part Of: Computers & structures. Volume 271(2022)
• Publication Date: 2022-10-15
• Subjects: Contact mechanics -- Dirichlet–Neumann boundary conditions -- Finite element method -- High performance computing -- Domain decomposition approach; Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2022.106862
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: This contact algorithm is designed for High-Performance Computing. The multicode approach allows using different time integration schemes for each body. The size of the linear system of equations remains fixed. The mesh partitioning is performed once at the beginning of the simulation. The contact algorithm is simple and flexible for large-scale problems. Abstract: In this paper, we introduce a novel parallel contact algorithm designed to run efficiently in High Performance Computing based supercomputers. Particular emphasis is put on its computational implementation in a multiphysics finite element code. The algorithm is based on the method of partial Dirichlet–Neumann boundary conditions and is capable to solve numerically a nonlinear contact problem between rigid and deformable bodies in a whole parallel framework. Its distinctive characteristic is that the contact is tackled as a coupled problem, in which the contacting bodies are treated separately, in a staggered way. Then, the coupling is performed through the exchange of boundary conditions at the contact interface following a Gauss–Seidel strategy. To validate this algorithm we conducted several benchmark tests by comparing the proposed solution against theoretical and other numerical solutions. Finally, we evaluated the parallel performance of the proposed algorithm in a real impact test to show its capabilities for large-scale applications.