Cluster identification algorithms for the decomposition of large systems

Many large aggregated systems found in process control, environmental systems and communication networks consist of groups of strongly bounded clusters or sub-systems. These special structural properties can be explored and a large system can be decomposed into many smaller sub-systems which can be handled easier and hence improve integrity. The decomposition of a large system into clusters of sub-systems belong to the class of block assignment problem which is NP-hard. In this paper, a grammian based inter-system coupling measure is used for measuring the input-output interaction. A recursive tree search algorithm is developed to find the best decompositions. GA heuristics are used for the search and the efficiency is enhanced by a new evolution process based on experiment design. Copyright © 2006 SoftMotor Ltd.