Parallel hierarchical algorithm for identification of large-scale industrial systems

Abstract

Parallel execution is the most powerful method of speeding up numerical analysis. However, numerical complexity increases much faster than the problem size. For this reason the computational effort required for the analysis of systems of big dimension,· or generally speaking, large-scale systems, will be substantial even with application of parallel computer architectures. This leads to necessity for proper model reduction, as well as decomposition of the physical problem into a set of smaller-size problems. In this paper we propose a method which comprises both of these demands and results in an algorithm for parallel hierarchical identification of reduced-size large-scale models.