Dynamic load balancing for large‐scale mortar contact formulations

Abstract

Parallel computing plays a vital role when solving large problems in computational mechanics, which might not be tractable by serial computers. Yet, in order to utilize parallel computing hardware effectively, appropriate balancing of the computational work load among all parallel processes is crucial to speed up computations and to avoid waste of computational resources. The present contribution addresses the question of optimal load balancing in the context of contact problems with a particular focus on dynamic contact scenarios, i.e. contact problems with large relative motion and significant changes in the contact configuration over time. To this end, we analyze the issue of load imbalance in contact problems, propose a load balancing strategy for dynamic contact problems, and demonstrate optimal balance and scalability for some application-driven test cases.