A concurrent and parallel processing methodology for the analysis, design and optimization of coupled thermal/structural interaction based on domain decomposition and sub-structuring techniques

Abstract This paper highlights the development of an efficient method concerning the solution of problems of a coupled nature. In particular problems of thermal/structural interaction are considered. The treatment of the problem starts by decoupling (partitioning) of the coupled domains, assigning coupling links between the elements of the fields involved and proceeding with a decomposition of each domain. Each domain is then solved on a sub-structure level using a domain decomposition algorithm based on the Schwarz alternating method of N-D (Neumann-Dirichlet) iterations of the thermal domain and a suitable sub-structuring technique for the structural domain. Implementing a staggered solution procedure and in particular an algebraic partitioning formulation (intregrate-partition-predict), the solution proceeds in time. The coupling links or linked groups of elements in each domain—called interacting macros—play here a central role in as much as they define intermediate interpolator-extrapolator macro functions across the predictor domain during and before the determination of an updated predictor field. This serves primarily an extended synchronization and parallelization of the staggered solution. Additionally, macros provide a powerful tool in all design or optimization stages of the post-processing phase.